MIS POCUMfNT

nf~ii2iü Volume 14

THE NUCLEAR SOCIETIES OF ISRAEL

TRANSACTIONS

TECHNION - ISRAEL INSTITUTE OF TECHNOLOGY HAIFA

December 21-22, 1987 Volume 14

THE NUCLEAR SOCIETIES OF ISRAEL

TRANSACTIONS

TECHNION - ISRAEL INSTITUTE OF TECHNOLOGY HAIFA December 21-22, 1987 Organizing Committee

E. Elias, (Chairperson), Department of Nuclear Engineering, Technion

A. Ketter, Israel Atomic Energy Commission L. Tepper, Israel Electric Corporation, Ltd. A. Galperin, Department of Nuclear Engineering, Ben-Gurion University of the Negev S. Kaizerman, Department of Nuclear Engineering, Technion M. Blau, Nuclear Research Centre - Negev

This Meeting was made possible by the assistance and generosity of: Technion - Israel Institute of Technology Israel Electric Corporation, Ltd. Israel Atomic Energy Commission Ministry of Energy and Infrastructure The Israel Academy of Science and Humanities National Council for Research and Development Israel Military Industry Ministry of Tourism Table of Contents

Page

Keynote Addresses (Invited)

Chairmen - M. Katz and G. Amir

Foreword to the 14th Meeting of the Israel Nuclear Society Louis Tepper 1-1

Utility Requirements and System Aspects of Next Generation Power Plants Walter B. Loewenstein 1-4 The BOT (Build-Operate-Transfer) Model for Building Nuclear Power Sta- tions in Developing Countries Shimon Yiftah 1-6

Non - Intuitive Thermal - Hydraulics in Reactor Safety and Containment Tech- nology F. J. Moody 1-9

High-Temperature Materials Chemistry in Severe Fuel Damage in Nuclear Reactors Don Olander 1-12

The Challenge of Developing Radioprotective and Photoprotective Drugs Emanuel ftiklis, Rina Kol, Michal Green 1-14

Reactor Technology and Safety I (Invited) Chairmen - A. Einav and E. Greenspan

Artificial Intelligence Expert Systems in the Nuclear Industry Shimon Yiftah II-l Nuclear Reactor Safety Research in Austria Gerald Sonneck II-4

Unsteady Thermal-Hydraulics of Boiling Water Reactor Safety and Contain- ment F.J. Moody II-7

i Activities of IAEA in Probabilistic Safety Assessment L. Lederman II-8

Small Modular High Temperature Reactors (HTGR): A Far-Reaching Inher- ently Safe Concept W. Kroger 11-22

Irradiation Behaviour of Advanced Fuel Elements for the High Temperature Reactor Under Normal and Accident Conditions G. Pott, H. Nabielek, H. Nickel and W. Sdienk 11-23

Nuclear Technology and Safety II

Chairmen - D. Maron end L. Reznik

Utilization of the IRR-1 Reactor H. Hirshfeld and A. Nagler III-l

Evaluation of LOCA in the IRR-1 Using the 3D-AIRL0CA Code A. Nagler and H. Hirshfeld III-5

Homogeneous Non Equilibrium Critical Flow Model S. Dickman, S. Kaizerman and E. Elias III-9

Critical Two-Phase Flow R. Dagan, E. Wacholder and E. Elias Ill-14

Experimental Apparatus for Quantitative Measurements in the Precursory Cooling Regime During Bottom Reflooding Y. Barnea, E. Elias and I. Shai Ill-19

The Effect of Precursory Cooling on Rewetting of a Solid Cylinder Shmuel Olek 111-23

Isolation of Small Malfunctions in a Nuclear Power Plant Using a Boolean Signature Algorithm Zvi Covaliu and Yakov Ben-Haim 111-28

The Israel Electric Corporation Meteorological Tower for the Nuclear Power Plant in the Negev Y. Balmor, A. Gutman and S. Kovacs 111-32

ii Stochastic Parameter Evaluation in Modelling Underground Transport of Ra- dioactivity Leib Reznik and Louis Tepper 111-35

Sensitivity Analysis of CRAC-2 Results Regarding the Characteristics of a Hypothetical Nuclear Plant Accident M. Rambam, L. Reznik and L. Tepper 111-37

Severe Accident Scenarios Following Externally Initiated Events in a Commer- cial Nuclear Power Plant D. Marouani and S. Weiman 111-40

PWR Loss of Feedwater ATWS Analysis Using the RELAP4/MOD6 Code D. Hasan, S. Kaizerman, E. Elias and E. Wacholder 111-42

Advancing DSNP to Simulate ATWS in PWR. D. Saphier and D. Gal 111-45

A Loss of Offsite Power Event Simulation D. Gal and D. Saphier 111-49

High Temperature Gas Cooled Reactors

Chairmen - D. Saphier and I. Kis

Israel's Interest in HTGRs E. Greenspan (Invited Paper) IV-1

A Review of the Economics of HTGRs Amitzur Z. Barak IV-5

Application of High Temperature Reactors to Israeli Shale Processing Irving Spiewak IV-9

Nuclear Safety Implications of Water Ingress Accidents in HTGRs J. Szabo, E. Greenspan, A. Ketter and S. Ron IV-13 Loss of Coolant Accident in a Modular HTGR D. Saphier and D. Gal IV-17

On the Graphite Oxidation Rate in HTGRs Following an Air Ingress Accident Shlomo Ron IV-21

m On the Consequences of Fission Products Plateout and Liftoff in HTGRs Shlomo Ron and Ehud Greenspan IV-24 Analysis of HTGR Vulnerability to External Missiles A. Ketter and J. Szabo IV-28

Reactor Physics, Concepts and Calculational Methods Chairmen - W. Rothenstein and A. Galperin

Accurate Multigroup Cross Sections for Resonance Reactions W. Rothenstein V-l Introducing NOXER-A 3D Nodal Diffusion Code M. Segev V-4

Equivalence Method for Resonance Shielding Calculations in Duplex Pellet Rod Lattices S. Carmona and M. Segev V-8 The Inverse Uncertainty Analysis Yigal Ronen V-12

Impact of Pn Approximation in Calculations of transfer Cross Sections and Scalar Fluxes for Anisotropic Scattering of D-T Neutrons in Graphite R. Ofek, A. Tsechanski and G. Shani V-16 Minimum Thickness Blankets for Fusion Reactors Y. Kami and E. Greenspan V-21 Liquid-Metal MHD Conversion of Nuclear Energy to Electricity Revisited E. Greenspan, A. Barak, L. Blumenau, H. Branover, A. El-Boher, E. Spero and S. Sukoriansky V-25 MHD Heat-Transfer Enhancement Possibilities in Fusion Reactors

S. Sukoriar Ay, H. Branover, D. Klaiman and E. Greenspan V-29 Direct Thermal Neutron Capture in Mg Isotopes S. Kahane, S. Raman and J. E. Lynn V-33 Pressure Vessel Fluence Reduction Possibilities Using Effective Shielding Re- flectors Z. Shayer and E. Greenspan V-38

IV An Example for the Potential Applications of Americium-242m as a Nuclear Fuel Yigal Ronen and Melvin J. Leibson V-42

Shielding Benchmarks J. Celnik V-45

The Collapsed Cross Section in Different Eigenvalue-Type Formulation of Boltzman Equation Z. Shayer V-65

Reliability and Non Destructive Testing

Chairmen - A. Notsa and Z. Alfassi

Dynamic Character of Failure State in Damage Accumulation Processes Dov Ingman and Leib Reznik VI-1

Software for the TNA Explosive Detection System T. Gozani, P. Shea, J. Adir VI-3

Reconstruction of Radiographic Images by Means of an Adaptive Polynom Fitting D. Ingman and Y. Merlis VI-9

Assay-System Design for Borehole Logging of Uranium Uzi Vulkan and Yakov Ben-Haim VI-14

Epithermal Neutron Activation Analysis for Determination of Trace Amounts of Gold in Rock Samples from Makhtesh Ramon E. Ne'eman, N. Lavi, A. Itamar and G. Baer VI-18

Reduction of Blur in Coplaner Rotational Laminagraphic Images B. Cohen and Y. Segal VI-20

Ultrasound Inspection of the Adhesive Properties of Metal to Metal Bondings P. Dickstein, E. Segal and Y. Segal VI-25

Single Projection Tomography of Objects with Cylindrically Symmetric Den- sity Distribution A. Notea, D. Pal and M. Deutsch VI-31

v Solution of the Inverse Problem in Radiographic Interpretation by Theoretical Model and Convolutable Function Y. Bushlin and A. Notea VI-39

Dual Energy Approach for Mass and Elemental Composition Distributions in Ceramics F. Tricher and A. Notea VI-43

Nuclear Medicine

Chairmen - E. Lubin and A. Kushelevsky

Preparation of r3Se by Proton Irradiation of Br Target Zeev B. Alfassi, Peter Jones-Smith and Regin Weinreich VII-1

On the Use of 5SCo in Nuclear Medicine - Its Disadvantage and Why t7Ni is Preferable Zeev B. Alfassi and Regin Weinreich VII-5

Disposal of Radioactive Hospital Wastes to the Negev and Fatal Road Acci- dents Y. Kalish VII-9

The Alpha-Beta-Gamma Scintillation Spectrometer and the Measurement of H 3/1 125 and C-14/I-125 Samples Y. Kalish VII-10

Adaptive Assay of Radioactive Pulmonary Aerosol with an External Detector A. Talmor, Y. Leichter, Y. Ben-Haim and A. Kushelevsky VII-14

Multi-Field Total Skin Irradiation Y. Mandelzweig, M. Yudelev, D. Sapir and M. Tacher VII-18

Recent Developments in the Radiation Sterilization of Pharmaceuticals Geoffrey P. Jacobs (Ya'akovi) VII-21

Portal Film Charts for a 6 MeV Linear Accelerator Sergio Faermann, Yehiel Leser and Eli Regev VII-25

VI Keynote Addresses (Invited)

Chairmen - M. Katz and G. Amir 1-1

FOREWORD TO THE 14 TH MEETING OF T H E ISRAEL N U C L EAR S 0 C I E T Y

At this Meeting of the Israel Nuclear Society, the first after Chernobyl, it is impossible to Ignore an event that, though it happened in the USSR, shook the western world in spite of the fact that its consequences were much less disastrous than many other, far less reported accidents. The reasons for the commotion are quite obvious: on the one hand, for the first time people were known to have died of radiation resulting from an accident at a commercial nuclear power plant; on the other hand, a sizable segment of the media blew up the accident out of all proportions, giving the impression at times that apocalypsis was here, now. To top it all, the ideologically anti-nuclear groups in Western Europe and the U.S. used the event as a heaven's sent omen, meant to confirm their long standing message of nearing doomsday, and whipped up emotions on a largely uninformed public.

Few people know.that to this day not a single person has died of radiation resulting from an accident at a commercial nuclear power plant in any western country, in an industry that after more than thirty years of production sold over five hundred billion dollars worth of electricity. Few people know that the Soviet RBMK reactor is unique and unstable, and unlicensable in the Vest; that its design was rejected in Britain long ago due to intolerable safety shortcomings; that no western reactor has the physical properties that made the Chernobyl accident possible.!

Few people know that Nuclear Safety is inherent in the very essence of the' industry in the Vest; that it is a profession comprising tens of thousands of engineers, scientists and technicians, working in all phases of the design, construction and operation of reactors, as well as in research; that not a stone is left unturned to assure the public's health. F«w people know that the human toll of Chernobyl, tragic as it was, comprised only workers at the plant or the fire brigade and amounted in all to 32 dead and 203 hospitalized, most of whom released at some later stage; that the latent health effects projected for the next 50-70 years are based on overconservative calculations yielding results well within the statistical errors of the numbers of cases due to "natural" causes, and therefore totally unidentifiable if existing at all. 1-2

Why do so few people know these and hundreds of other facts, such as the nature of risk, the need of energy for feeding a growinglv hungry world population, the irrelevance of linking the issues of nuclear weapons and nuclear-electric power, the necessity of nuclear power as part of a balanced energy basket on a national and global rcale, the fact that there is no such thing as a radioactive-free environment, that a week of "taking the waters" at one of the renowned health spas such as Baden-Baden or the Dead Sea produces more radon-induced radioactivity than that received from Chernobyl at the same places, that on the whole the nuclear industry has an unrivalled safety record and potential?

I believe that the nuclear community is largely to blame for this lack of information. True, the public is no better informed about many other industrial activities, not to speak about other areas such as military or political, even if it appears otherwise. But the fact is that for some reason nuclear matters are shrouded in an aura of mystery, perhaps because radioactivity cannot be seen, touched or smelled. In addition, the secrecy surrounding the development of nuclear weapons somehow found its way, perhaps unnecesarily, to the nuclear-electric power program in its early stages. Those among us old enough to remember, will recall the awe and respect that nuclear scientists undeservedly earned in those days. Today the scenario is all turned upside down and we appear as the villains of the play, even more undeservedly.

I believe that 3 decades after Calder Hall and Shippingport, 27 years after Nahal Soreq and 20 months after Chernobyl, we should take stock, look back at the road behind and more importantly, at the road ahead. Nuclear-electric power is here to stay, as part of a balanced energy basket. It is still the safest and cleanest among the commercially available sources of energy. I believe the breast-beating in part of Western Europe to be an ephimeral phenomenon of a society of plenty. Energy-starved countries such as Japan forge ahead with their nuclear program in a no-nonsense approach. The same is true with the Soviet Union and the eastern group of countries. France's massive nuclear program (over 70* of its electricity production), launched after the 1973 oil crisis as a result of its heavy dependence on imported energy sources, has not been interrupted. Nuclear power will undoubtedly come back to the West once the energy pinch is felt again, maybe even in a rush. Pity that people - and many governments - have such short memories.

Absolute safety or perfection in any human endeavour, just as a mathematical asymptotic limit, is unattainable. "How safe is safe enough?" has for a long time been a question discussed and worked upon by Nuclear Safety experts. In actual fact, however, nuclear power plants around the western world are being built in an ever improving pattern of safety, which by all standards is far better than in any other industry. Is this safe enough? 1-3

In this quest for getting ever closer to absolute safety, the nuclear industry as a whole has constantly been upgrading the design of existing types.of reactors, on the basis of their vast accumulated operating experience. The impressive record speaks for itself. Simultaneously with this effort a different approach has been taken by part of the industry, a search for new reactor designs which could assure safety inherent in their very concept.

The High Temperature Gas Cooled Reactor, to which a special session is being devoted at this meeting, falls somewhere between these two approaches. Not entirely a new concept, having been around in Germany and the US for quite a many years, it is however very different from the commercially available water reactors, its design being based on ideas intended to make it inherently safe, even without resorting to engineered devices. Its THTR-300 version has been connect 3d to the German grid for two years now, therefore cannot be compared to concepts still on the drawing board, such as PIUS or others.

It is not easy to foresee now which of these two approaches will eventually prevail in the commercial power reactor market, in that interim period before the next generation of reactors takes over, probably the fast breeders and later maybe even fusion reactors. On the one hand proven technology that withstood the test of time and varied operating conditions with a wealth of experience, on the other hand newer concepts with built-in intrinsic safety. One thing however must be clear: commercial reactors today are safe, as safe as can reasonably be expected, safer on the whole.than other industrial installations."I do not believe that the fears which reactors instill in part of the uninformed public will be allayed by a safer type of reactor. For those who oppose them, nuclear reactors are seen as demons which must be exorcised. There are no safe demons.

I believe the whole nuclear subject can be demystified by a massive information effort, starting by the nature of risk. I think that all of us in the nuclear community should devote some time to this, in order to explain to the public - and to a good part of its leaders - the necessity and safety of nuclear power. If anything, Israel is* more needy of nuclear energy than most other industrialized countries. As in Japan, there are no large scale indigenous sources of energy. Strategically, the advantages of small once-a-year shipments of fuel are too obvious to need further explanation. Advanced technology is yet another reason. As for the ether advantages - cheap, clean, safe, abundant - they are the same as all over the world. Technological progress cannot be stopped. Sooner or later we are bound to have nuclear-electric power. It would be better sooner than later.

Louis Tepper President The Israel Nuclear Society 1-4

Utility Requirements and System Aspects of Next Generation Power Plants

Walter 13. Loewenstein

Electric Power Research Institute Palo Alto, California

It is with considerable humility that one must approach the cloudy crystal ball to predict the future of electric utility requirements, system aspects and the next generation of nuclear power plants. To first approximation, the utility must provide electricity that is reliable, safe and economic. The utility must also be able to cope efficiently with the varying daily and seasonal demands for electricity. To the extent that nuclear electric power plants play a role in responding to the demand for elec- tricity, they must conform to the overall utility requirements for providing a reliable, safe and economic product. At the present time, nuclear power plants provide almost 20% of the electricity in the USA. There are regions 'vithin the USA where the nuclear electric fraction exceeds 50%. The number of operating nuclear electric power plants approaches 110 capable of producing almost 100,000,000 KWe. More than 50 U.S. electric generating utilities have nuclear commitments including both publicly and investor-owned organizations. Within the USA, the installed NSSS face the continuing challenge of excellence in operations. On a day-to-day basis, this means enhancing capacity factor, reducing operation and maintenance costs and, where applicable, reducing capital costs while maintaining, assuring and enhancing overall safety. These challenges manifest them- selves in different ways as they pertain to individual utilities and as the framework of the utility industry evolves. Current appraisals show that the economic advantage of new nuclear plants is likely to be small, if it exists at all, relative to existing and projected coal plants in the USA. Of course, it is difficult to predict the costs of compliance with emission regulations which are expected to be substantial and growing. This evolution will impact the choices facing the investor in future electric generating plant hardware. The future favored generating options will be dominated h}' (.hose with prospects of good capital investments (e.g., short construction time) and good operating and maintenance costs. Clearly, both of these factors (capital and O&M) are facilitated by 1-5

simplicity of systems. Recent trends have been otherwise, leading to more complexity requiring more O&M activities. Another vital factor is plant size for new construction. It is very desirable that any new addition within a utility system be consistent with the size of the system. As a result, near-term projected capacity additions tend to focus in the several hundred megawatt range. This is a significant departure from the nuclear plant commitments in the USA during the early 1970s. These tended toward the 1000 MWe or greater plants. In part, these commitments were based on the perception and evidence on economy of scale. However, there is new evidence suggesting that the "economy of scale" can also be achieved through innovations in construction (e.g. factory assembly maximized leading to a shorter construction time) if scale can be made up by number of units. As one looks at the utility needs for early new capacity addition during the 1990s and beyond, generation, transmission and distribution hardware options must be quite flexible. Competition for extra-territorial markets will impact generation choice. More load-follow, frequency control and system stability requirements must be met than the routine base-loaded operations that dominated the recent years. (Fortunately, experience is being generated on these questions in Europe and the Far East.) The prospect of independent nuclear plant generating companies cannot be dis- missed in these deliberations. A recent proposal by U.S. utilities was not implemented; it is likely that similar proposals will gradually appear. (Some precedent exists in the Federal Republic of Germany.) In the same vein, strategic decisions on generating capacity in Europe and the Far East are not and need not be mirrored in North America. For example, the French decisions based on "No coal, no gas, no oil and no choice" have a different basis from those in the USA. The economic advantage of nuclear generation in Europe and the Far East is tens of percent (OECD studies are eloquent on this). 1-6

The BOT (Build-Qperate-Transfer) Model for Building Nuclear Power Stations in Developing Countries

Shimon Yiftah Department of Nuclear Engineering Technion, Haifa 32000, Israel

In 1383 Turkey issued letters of intent for the construction of a 600 Mwe power reactor on a turnkey basis to AECL of Canada and

Kraftwerk Union of Germany. The site, selected in 197b( was the Akkuyu site on the Mediterranean coast. Negotiations with AECL and KWU continued until August 1984, for a turnkey project. Then the Turkish Prime Minister, Turgut Ozal, announced the policy of BOT, build-operate-transfer.

What is BQT?

BOT is a scheme, or a model, by which a country which does not have hundreds of millions or billions of dollars can initiate large projects, to be constructed and operated at little or no initial capital cost to the government. Under the scheme a foreign vendor or consortium forms a joint venture with a national company to first build and then operate a project for an agreed time. The foreign consortium arranges financing of the project, then sells the output at an agreed price. Once cost has been recovered and a reasonable profit realized, the project is transferred to the national company.

Thus, considering the BOT model of building the first Turkish nuclear power station at Akkuyu since August 1985, AECL formed an international consortium with the turbine supplier and with an Istanbul-based engineering contractor, which in turn formed with the Turkish Electrical Authority a joint venture arrangement to finance build and operate the nuclear power plant at Akkuyu. The idea was that this joint venture would operate the plant for 15 years during which time the project financing would be recovered, including reasonable profits. The nuclear plant would then be transferred to the Turkish Electrical Utility (TEK) to be operated as one of its own national plants. This deal has not yet been concluded because the parties have been unable to agree on an appropriate financial arrangement. The Turkish government is trying to get other projects to be built under the BOT model, including natural gas pipelines from 1-7

the Soviet Union and Qatar, an oil pipeline from Iraq, a network of highways, several hydro electric projects, coal power plants and a telephone exchange complex with 555,000 lines. Altogether there are projects with a total cost of S billion dollars.

The Indonesian BOT Nuclear Project

A second country that is actively considering initiating a nuclear power project under the BOT model is Indonesia. Three consortia, Framatome/Kraftwerk Union, AECL, and Westinghouse/ Mitsubishi/Ansaldo submitted in the summer of 1387 to BATAN, Indonesia's Nuclear Energy Agency, their "pre-feasibility" studies of the BOT 500 to 1000 Mw reactor supply concept, 3 months after the issue of the bid order. The three bids were then submitted to Motor Columbus Consulting Engineers of Switzerland for evaluation. After two months of study the Swiss company concluded that the nuclear power station to be built on the island of Java under the BOT model is feasible. All three bids assume plant construction time of five to five and a half years and average plant capacity factor during operation of about BOH. Ths three bids propose construction of at least one conventional thermal station to be finished before the nuclear plant. This conventional station would start generating cash flow to keep construction of the nuclear project on schedule.

Motor-Columbus believes that the BOT scheme will shorten construction time and provide a smoother technology transfer. Also, if the BOT model works in Indonesia, it could work in other Third World countries.

Indonesia's Nuclear Agency is scheduled to submit its report of the bids and consulting engineer evaluation to the Indonesian president at the end of October 1987. By the end of 1987, it is hoped that a decision will be taken whether or not to pursue studies of the nuclear project, including negotiations with the vendors.

Based on previous experiences in Iran and Turkey, the vendors will insist on guarantees that the Indonesian government will back up any financial commitment from BATAN. 1-8

Concluding Remarks

The BOT Model, if successful in one country, could bt, applied in other developing countries. It may become almost "the only game in town" for large nuclear vendors who face the prospect of reducing their manpower by thousands of people because of the lack or scarcity of nuclear orders. On the other hand, for developing countries who need nuclear power and can't finance it - this may be the only possible scheme.

Because of several reasons, Israel should follow very closely developments of the nuclear BOT model. It may well be a scheme capable of solving several problems and removing or easing constraints that have delayed the initiation of a nuclear power program in the country. 1-9

NON-INTUITIVE THERMAL-HYDRAULICS IN REACTOR SAFETY AND CONTAINMENT TECHNOLOGY

F. J. Moody GE Nuclear Energy San Jose,California USA

ABSTRACT

Intuition is used to project physical behavior into regions where the parameters have not yet been tested. However, thermal-hydraulic analyses sometimes yield results which are not intuitive. This paper describes several interesting, non-intuitive aspects of thermal-hydraulic phenomena which occur in reactor safety and containment.

DISCUSSION

Steam/Water Critical Flow The discharge of steam/water mixtures from reactor vessels through ruptured pipes is an important phenomenon in reactor safety analysis. Maximum pressure in the containment is determined by the discharge rate. It is well-known that two-phase steam water flows from a pressure vessel are limited by a critical flow condition. But the possibility of two critical flow conditions in a single ruptured pipe has been demonstrated. When saturated water flows from a vessel into a pipe entrance, its pressmre decreases, which causes homogeneous bubble formation. The mixture is bubbly in the pipe entrance. But as flow precedes through the pipe, the vapor and liquid divide into two separate streams. The bubbly mixture reaches critical flow in the pipe entrance, and the separated flow can reach a second critical flow condition at the pipe exit. This discovery made it possible to interpret two-phase critical flow data which agreed with the homogeneous model for discharge from short tubes, and with the separated flow model for discharge from long tubes.

Vessel Decompression Rate

The two-phase critical flow models show that for a given discharge area, the mass discharge rate of saturated liquid is three or four times the discharge rate of saturated steam. Consequently, the mass loss from a pressure vessel discharging water occurs four times faster than a vessel which discharges steam. Intuitively, one might think that the decompression rate would be faster for water discharge than for steam discharge. However, the opposite is true! Consider a vessel with equal volumes of saturated steam and water at 7 MPa pressure. The decompression rate for steam discharge is 7.5 times that for water discharge! When water 1-10

discharge occurs, steam in the vessel expands to occupy that volume formerly occupied by water, and pressure reduction is slight.. When steam discharge occurs, the steam space pressure is reduced more significantly by the loss of steam mass, which causes a n.ore rapid decompression rate. Snow in the Drywell? A steam lines may contain flow limiting venturi nozzles for measuring steam flow rates, and also for limiting the discharge flow rate if a steam line rupture should occur. The pipe/flow limiter area ratio is about 4.0. It can be shown from gas dynamics that for hot steam at 7 MPa pressure and 285 °C in the vessel, discharge from the flow limiter will be supersonic at a temperature of - 70 °C l That is cold enough to form snow! Then why worry about cooling the drywell after a steam line rupture? When the supersonic, cold steam discharge comes to rest in the containment, its high kinetic energy is converted to internal energy, which results in a temperature approaching that in the vessel, before it mixes with existing contents. Containment Pressurization Rate When steam/water discharge occurs through a postulated pipe rupture, the drywell pressure rises in a pressure suppression containment until the vent water is expelled and venting to the pool begins. Critical flow models show that saturated steam with its higher energy discharge rate causes a higher drywell pressurization rate than saturated water discharge. Cold water discharge, which does not flash in the drywell, causes the lowest pressurization rate. However, between highly subcooled and saturated water, there is a vessel water temperature for which the energy discharge rate is maximum, which causes the drywell pressurization rate to achieve a value of more than twice the pressurization rate for steam discharge 1 Vessel Thrust during Water Slowdown The thrust on a ruptured pipe caused by fluid discharge must be accommodated by the structural design in order to prevent pipe whip and further damage to nearby equipment. If a vessel at pressure P and pipe of area A contained water, rupture discharge could be stopped by holding a plug in the end of the pipe with a force PA . Thus, the required axial force exerted on the broken pipe to hold the plug in place would be PA . However, if the plug slipped out, the momentum principle shows that liquid discharge would cause the pipe axial force to increase to 2PA, or twice the Nuclear Technology and Safety I (Invited)

Chairmen - A. Einav and D. Saphier II-l

ARTIFICIAL INTELIiIQENCE EXPERT SYSTEMS IN THE NUCLEAR INDUSTRY

Shimon Yiftah Department of Nuclear Engineering Technion, Haifa 32000, Israel

Nuclear technology has always been characterized by being at the forefront of advanced technology. The recent rise in Artificial Intelligence (AI) interest, leading edge of computer science, and widening development and use of expert systems in. several scientific, aerospace, defense, engineering, medical, computing, telecommunica- tions, finance, management and manufacturing areas, are having an impact in the nuclear industry. It is a growing edge.

It is estimated that more than one thousand expert systems are currently prototyped all over the world in various areas. ("Expert Systems 19B7" by Graeme Publishing, dated April 1987, lists over 1025 expert systems and over 168 expert system shells.) Much important work on a wide variety of AI and expert systems applications in different areas of the nuclear industry is being done worldwide, in the U.S.A., Japan, France, the U.K., the Nordic countries, Italy, Belgium, Finland, Canada, Taiwan, and others.

France through Framatome's subsidiary for AI applications, Framentec, is developing about 30 expert systems of which, to—date, 3 systems are operational, 15 have completed field-testing or at the advanced prototype stage and 12 are research and demonstration grade systems. The main fields covered are: diagnosis, advising, optimization, monitoring and regulations.

In the United States, there is much activity in the expert system field by vendors, architect engineer firms, universities, national laboratories, Federal agencies, and the electric utility industry. The Electric Power Research Institute

In Finland, the Imatran Voima QY

In Taiwan, the Taiwan Power Company, and EPRI, co-sponsored a project to develop and prove a decision support system for emergency procedure tracking for application to the Kuosheng BWR nuclear power plant <2x948 Mwe). The software was developed by Nuclear Software Services of Los Gatos, California, and the system is being integrated into the overall Safety Parameter Display System (SPDS) developed by the General Electric Company. The production rule expert system (a rule-based knowledge base and inference engine) was written in the "C" language for use on a DEC VAX computer containing the SPDS data base. The system is to be evaluated in tests at the Kuosheng simulator and, when accepted, will be installed at the Kuosheng nuclear power plant.

The development of so-called expert system shells, or expert system development tools, can be used to prepare, relatively easily, expert systems for different areas. Also, the development of several expert system shells that can be run on microcomputers, coupled with the growing memory and computing power of these micros, seem to indicate that the field of expert systems in the nuclear industry will be growing rapidly in the next few years.

An example of one of the most widely used expert system shells for microcomputers is EXSYS. We have started experimenting with the Demo of EXSYS. A "Shell Review Monthly" in AI Today reviewing EXSYS in February 1967 is attached as an example. II-3

The Nuclear Regulatory Commission (USNRG) is sponsoring a few •xpert systems applications in the areas of accident management for containment assessment and Reactor Safety Assessment. The Reactor Safety Assessment System (RSAS) is resident on a Xerox 1186 AI work- station and is installed at the NRC Operations Center for testing and review by NRC personnel.

Among the numerous axpert systems being developed in the U.S., we mention as examples "ADRA" an ALARA design Review Assistant, by Sargent & Lundy; an expert system for planning steam generator inspections, by Combustion Engineering; an expert system for diesel generator diagnostics, by Pickard, Lowe and Garrick, Inc. and the Pilgrim Nuclear Power Station of the Boston Edison Company; and an expert system for fault tree analysis being developed by Expert - EASE Systems, Inc. Another interesting example is XTIP, a prototype version of the Expert Thermal Information Program, an expert advisor for the operator of a coal-fired steam power plant. XTIP, being developed by the General Physics Corporation, includes the rules elicited from expert plant operators to diagnose plant conditions and to recommend specific actions based on the present state of the plant to improve efficiency. If desired by the operator, XTIP provides interactively an explanation of its reasoning.

Canada with its high level of plant computerization in CANDU stations, is aiming the development of expert systems at assisting reactor operators in areas such as: fault diagnosis, identifying limiting conditions of operation, heat sink availability, optimizing on-power fueling schemes, computerized procedures and communication. The "Operator Companion" - an expert system designed to diagnose plant faults and to advise the operator on appropriate restoring and corrective actions is a major undertaking which is receiving support from AECL and McMaster University in Hamilton, Ontario.

In Japan a national project of "Advanced Man-Machine System Development for Nuclear Power Plants" (MMS-NPP) has been carried out since 1984. This eight year project consists of Phase I (Conceptual Design, 1984-1986) and Phase II (Detailed Design and Implementation, 1987-1991). The participants of the project are six private companies engaged in NSSS supply in Japan. They contribute to the development of MMS—NPP on the division of work basis according to the reactor type and the proposed functions. II-4

NUCLEAR REACTOR SAFETY RESEARCH IN AUSTRIA

Gerald Sonneck OFZS (Austrian Research Center Seibersdorf)

1959, in the wake of the Atoms for Peace Programme, a research center for atomic energy was built in Austria so that there was already some experience also in safety research when it was decided in the early seventies to build the nuclear power plant GKT, a Kraftwerk Union boiling water reactor of S00 MWe. Our main interest then was of course the licensing of this reactor which was to be supported by safety research. 197B a referendum led to a law which bans the production of electricity from nuclear fission, so the nuclear power plant was mothballed. The research center for atomic energy became the Austrian Research Center Seibersdorf (OFZS) and only a small group still works in reactor safety.

The reasons for maintaining such a group are:

- Austria is a net energy importer also importing electric energy. This electric energy is produced mainly in countries which operate nuclear power plants such as Germany, Switzerland, Czechoslovakia, Hungary, the Soviet Union and Yugoslavia. There is, of course, an ethic position that whoever uses nuclear energy should also contribute to make it as safe as possible.

- A number of NPPs is situated next to the Austrian borders and the Chernobyl accident has shown how near the Soviet Union is. This calls for domestic expertise for the public and the government.

- Nuclear technology is high technology. Its spin-off products in the non-nuclear field are not to be neglected.

The means for this work comes from the government, from sponsors such as the Austrian National Bank and from income from spin-off products:

The work is done mainly in the OFZS. Some contributions come from the universities. II-5

A small country like Austria cannot be involved in everything. He choose to actively collaborate in a few topics rather than try to be informed on everything. These topics arei

- fuel behavior in accident conditions and - thermal hydraulics

International collaboration is, of course, indispensable for a small country. We are a partner in a number of multi- and bilateral agreements and research contracts. These include the following:

OECD: Austria is a member of the OECD-LOFT Programme, which aims at investigating the effect of simulated loss of coolant incidents and transients on thermal-hydraulics and fuel. One of us is currently assigned to the LOPT-reactor at Idaho Falls, Id., U.S.A. In Austria, we used our own code EALO-2A to calculate ballooning and rupture at the test LP-FP-1.

An important part in the nuclear safety programme of the OECD is played by the International Standard Problems (ISPs), where pretest-calculations are compared with test results. We participated in a number of these ISPs at the following facilities;

REBEKA (Karlsruhe), a German electrically heated loss of collant test for fuel behaviour (BALO—2A).

PHEBUS (Cadarache), a French nuclear in-pile test for fuel behaviour (BALO-2A).

FIX II (Studsvik), a Swedish experiment simulating the thermal-hydraulic behaviour of a BWR at intermediate breaks (RELAP 4/Mod S).

LOBI (Ispra), an experiment done by the European Community to investigate the behaviour of a PWR at small breaks (RELAP 4/Mod 6).

PIPER I (Pisa), a thermal hydraulic test for a BWR, where we use the Japanese code THYDE-B.

IAEA: The IAEA also conducts standard problems, up to now using two tests on the Hungarian PMK-NVH-loop simulating the thermal- hydraulics of the soviet WWER-440. We participated in both using RELAP 3/Mad 6. II-6

Israel: Two research projects were funded by the Austrian National Bank: The first, a nodalization study using RELAP 4/Mod 6, was done in collaboration with the Tel Aviv University. The second, on the thermal-hydraulic behaviour of BWRs was done together with the Technion.

Hungary: Together with KFKI, the Central Institute for Physics of the Hungarian Academy of Sciences, we are developing and testing two-phase flow instrumentation.

USA: We are just beginning a collaboration within ICAAP, the International Code Assessment and Application Program.

As mentioned above, spin-off effects in the non-nuclear field are important for us. This includes stress analysis, plant analysis, risk analysis, handling of complex computer codes, and measurement of liquid and/or gas flows. In all these fields our nuclear experience has prompted and supported non-nuclear application in Austria. II-7

UNSTEADY THERMAL-HYDRAULICS OF BOILING WATER SAFETY AND CONTAINMENT

F.J. Moody G.E. Nuclear Energy San Jose, California, U.S.A.

Abstract

Boiling water reactor safety and containment has been the subject of analyses, research and development in the United States and other countries for three decades. These notes provide a survey and summary discussions of various thermal-hydraulic phenomena with safety and containment applications. Topics discussed include valve and piping transients, loss of coolant accidents, and the response of pressure suppression containment systems. II-8

ACTIVITIES OF IAEA IN PROBABILISTIC SAFETY ASSESSMENT L. Lederman International Atomic Energy Agency Vienna.

ABSTRACT

The prograEines of the International Atomic Energy Agency in the field of Probabilistic Safety Assessment (PSA) are described in terms of the Agency's basic functions. The paper reviews those activities related to a level 1 PSA analysis. Among then are the establishment of quantitative probabilistic safety criteria, the publication of a series of case studies documenting in a tutorial manner actual PSA applications in safety decisions and the preparation of guidelines for conducting PSAs. Other activities include the assistance to Member States under Technical Cooperation Programmes, trainning courses, and the development of a software package for fault tree and event tree analysis in personal computers. Programmes to promote international coordination of research in areas such as "Data Collection and Analysis" are also addressed. Finally future trends including the use of PSA as a tool for operational safety management are mentioned. II-9

1. INTRODUCTION Over the past 12 years probabilistic safety assessment (PSA) matured and became an important tool for evaluating reactor safety. It is however far from being a new idea. In the early 194O's quantitative safety requirements were already proposed for the aeronautics industry.

In 1942, when the high failure record of the German V-2 rocket programme was studied, the concept of dependency between parts of a system was introduced as a step ahead of prevailing belief that a system was only as good as its weakest part. That started a logical and integrated system analysis approach. In I9 60 further development was achieved with the use of logical analysis in connection with the Apollo programme of the US National Aeronautics and Space Agency (NASA). An important tool called "failure mode effect and criticality analysis" was then introduced. In 1962 the Bell Telephone Laboratories in the USA used the fault-tree technique in a study for the US Air Force on the reliability of launching and controlling Minuteman missiles. In the nuclear power field, a limit line for accidental iodine releases was first proposed in terms of the probabilities of occurrence by F.R. Farmer of the United Kingdom in 1967.

The pioneering report in the application of probabilistic methodology to nuclear reactor safety was WASH-14 00, generally known as the Reactor Safety Study, published in 1975 in the U.S.A. Unfortunately, the study's complexity and difficulties in reporting results combined to downplay its potential for evaluating reactor safety. However, when it was found that WASH-14 00 had foreseen sequences similar to those leading to the TMI accident, the use of probabilistic methodology in nuclear safety gained additional momentum. To date, over 30 PSAs have been completed in various countries and the results have brought invaluable insights for plant design and operation.

Some studies have been sponsored by governmental organizations - as in the cases of WASH-1400 in the USA and the German Risk Study in the Federal Republic of Germany. Many others were conducted entirely by the industry or as joint ventures. Most recently a comprehensive assessment of severe accident risks for a set of five commercial light water reactor nuclear power plants (NPPs) (NUREG 1150) was commissioned by the U.S. Nuclear Regulatory Commission. The reports, currently under review, are intended as benchmarks for the evaluation of NPP's vulnerabilities to severe accidents. 11-10

2. IAEA PROGRAMME

In line with current developments, the IAEA, within the framework of its nuclear safety programme, pays due attention to the field of probabilistic safety assessment. Activities in this area are structured in accordance with the Agency's basic functions, namely: - Information exchange

- Development of Standards

- Training - Technical cooperation and assistance - Promotion of research and development - Operations and service

Next, those activities related to a level 1 PSA are described.

2.1 Information exchange

In response to current interest of Member States, the Agency convenes technical meetings and publishes state of the art reports on various aspects of PSA. Table 1 lists the meetings planned for 1988.

2.1.1. Probabilistic Safety Criteria (PSC).

The status of PSC in countries with nuclear power programmes varies considerably. One group is already applying them to support safety decisions. A second group of countries is at present debating the basic concepts and implications. A third group is monitoring the developments for possible future use. To assess the status, experience and future aspects for the development of PSC, two technical committee meetings and consultants' meetings have been organized by the Agency in 1986 and 1987. The committees addressed PSC at the level of public health and at plant level. Table 2 (Ref.l) indicates the relation between the level of PSC imposed and the level of PSA required to demonstrate compliance. Probabilistic safety criteria at the safety function/system level are treated in reference 2. 11-11

2.1.2. Case Studies

This programme intends to provide Member States with a comprehensive set of case studies reports, documenting in a tutorial manner actual applications of PSA in safety decisions. There are at present 12 case studies under preparation. Table 3 indicates the status of these studies. Drafts are peer reviewed by an oversight committee before publication by the IAEA. A number of additional topics have been identified as necessary for completing the programme. Among them are reports on " Insights gained from PSAs on fires, floods and seismic events" ; " Allocation of inspections resources by a utility to improve reliability"; " Use of PSA results for training" and " PSA results for plant availability optimization".

2.1.3. Human Interactions

This area of activity deals with aspects for human error modelling and quantification in PSA. It also includes the use of simulators for training operators for beyond design basis situations and the use of advanced operator aids based on PSA insights (Refs. 3,4,5). Related meetings are also shown in Table 1, including the "International Conference on Man-Machine Interface in the Nuclear Industry" (Japan,88).

A review of the state of the art for considering human errors in PSA is also being undertaken by the Agency in cooperation with the Ispra Establishment (Community of European Countries) and a joint meeting is scheduled for 1988.

2.2 Development of Standards

Aiming at promoting the standardization of PSA reports and to provide a consistent framework to compare results with safety criteria an effort was initiated to prepare a series of documents on PSA guidelines. A reference report is under preparation addressing the various elements of a level 1 analysis. The review of the initial draft was recently completed. The final report is planned for 1988. The report describes seven major procedural steps for performing a level 1 11-12

PSA, namely: 1. Management and organization; 2. Identification of Radioactivity sources and Accident Initiators; 3. Accident Sequence Modelling; 4. Data Acquisition, Assessment, and Parameter Estimation; 5. Accident Sequence Quantification 6. Documentation of Analysis, Display and Interpretation of results; 7. Technical Quality Assurance. The programme is complemented by a series of guidelines on specific PSA elements. Specific guides to be initiated in 1988 address: computer codes for level 1 PSA; human error modelling and quantification; treatment of external hazards and common cause failure analysis.

2.3. Training

International PSA training courses are an important element of IAEA's activities. A six week course is offered to PSA practitioners on a yearly basis. The course covers the basic elements of a PSA including a two week long workshop where participants exercise practical applications of systems modelling and accident sequence quantification. The use of computer codes is introduced at this stage. In 1987 the course was offered in Spain for Spanish speaking candidates. In 1988 it is scheduled to take place in the U.S. A two and a half week long course is also offered every other two years for managers whose responsibilities include PSA activities. This course will next be offered in 1988 in the United Kingdom.

2.4. Technical Cooperation and Assistance Technical assistance is provided to developing Member States in the various aspects of PSA programme planning and implementation.

2.4.1. Inter-regional PSA Programme

Until very recently PSAs were being carried out mostly by a small group of countries with advanced nuclear power programmes. The complexity of such studies requires substantial resources and expertise. Aware of the fact that even if a probabilistic analysis is limited to the assessment of the most important accident sequences it can be of great value in the evaluation of the safety of a 11-13

nuclear power plant, the IAEA started in 1985 an inter-regional technical cooperation programme in PSA. The objective of the programme is to establish in developing Member States teams capable of performing PSAs, and of using the results to support safety decisions. The programme, which was originally limited to nuclear power plants, was extended to research reactors in 1986.

The implementation of PSA programmes in the various countries participating in the programme is described in reference 6.

The assistance provided by the IAEA for the countries participating in the programme includes expert missions to support PSA programme planning and the organization of workshops tailored to specific needs of Member States. In this framework a workshop is planned for 1988 on the use of Fault Trees and Markov methods to evaluate plant's technical specifications. Workshops held in previous years covered subjects such as "Methods for reliability data collection and retrieval" and "Use of the SETS computer code".

2.4.2. Regional WER Programme

Considering the need to foster cooperation between countries operating or planning to operate similar reactor type or sharing common interest because of their geographical proximity,the IAEA is initiating in 1988 a regional programme. The programme is intended to assist owners of WER reactors in the development of a level 1 study under a common framework. To facilitate the review of results and exchange of insights, aspects related to reliability, data base, nomenclature, selection, grouping of accident initiating events, treatment of human errors, dependencies and use of computer codes are discussed at the early stages of the project. Additionally, the programme relies on the use of the PSA guidelines being prepared by the Agency.

2.4.3. Regional Human Factors Programme

The objective of the proposed programme is to assist developing Member States to identify human factors that can lead to accident situations and to recommend preventive measures.

The programme scope includes the evaluation of operational experience to identify human errors; development of a classification for human errors and of dedicated reporting schemes; modelling human actions in PSA; utilization of simulators to analyze human actions; use of computer aids including expert systems to improve diagnosis of abnormal 11-14

situations? operator training for severe accident conditions; human factors in NPP operational management and the impact of performance shaping factors.

2.4.4. Computer Codes

As a part of the inter-regional PSA programme a small library of PSA related computer programmes has been established at the IAEA. In addition, support for the implementation and use of the programmes is provided. The selection of specific computer codes depends to a large extent on the computer hardware available in a country or accessible to a specific organization. Therefore adaptation of some of the codes to different computer equipment was done by the IAEA or by the users. Table 6 shows a list of the PSA related computer programmes contained in the library and the computer types on which they are operable. Because Personal Computers (PCs) are increasingly utilized as convenient work stations for PSA analysis and considering that such packages are generally not of public domain, the IAEA has initiated a project to develop a PC software package for fault tree and event tree analysis. The package is developed in cooperation with Member States and is primarily intended for instructional purposes. The software known as PSAPACK (Ref.7) is being developed in a modular form. The package includes: - reliability data base; - full screen graphic fault tree editor; - fault tree codes (PREP, MOCUS, ALLCUTS, FTAP) ; - event tree development code; 1 - uncertainty calculations; - minimum cut set processing files; - link modules to mainframe computers

2.5. Promotion of Research and Development

As a means of promoting and coordinating international research, a number of coordinated research programmes have been initiated by the IAEA in 1987. One of such programmes addresses the area of "Data Collection and Analysis for PSA". This programme recognizes that available 11-15

data is often unusable in PSA studies because of the way in which the data have been collected. Frequently events are recorded without providing any information concerning the time over which the data have been collected or the number of demands experienced during this time. Other shortcomings in available data sources are the lack of sufficient detail concerning the event or operating conditions and the lack of consistent and well specified definitions. These include component boundaries, operating environment, design specifications and available information on failure mode or root cause.

The main goal of the proposed Coordinated Research Programme is the development of systems for data collection, data retrieval and data analysis for PSA which will facilitate the exchange of information among IAEA Member States.

The use of expert systems and fuzzy sets to develop intelligent interrogation techniques; data collection for common cause failure analysis; the development of procedures for data collection and development of statistical analysis techniques to process data are some of the proposed subjects of research.

A second programme on "Reference Studies on Probabilistic Modelling of Accident Sequences" aims at fostering the development of standard methods for modelling accident sequences in PSAs. The coordination includes selection of a common initiating event; identification of a reference data set; assessment of fault tree codes and the analysis of reference accident sequences using realistic assumptions about operator behavior and best estimate calculations of system response.

A third on-going programme deals with PSA methods as applied in Nuclear Research Reactors. It is scheduled for completion in 1988 with the publication of a technical document by the IAEA.

2.6. Operations and Services

A newly introduced programme provides Member States with international expertise to review PSAs. Review teams are organized by the IAEA and missions are sent to Member States to conduct a peer review of PSAs. The programme known as IPERT (Independent Peer Review Teams) will be available starting 1988. 11-16

3.- FUTURE TRENDS

Despite its potential, the actual use of PSA insights in decision making is still modest. One reason is that PSA reports mix useful results with a great deal of technical information that is irrelevant to decision makers. Another reason is that PSA reports are understood only by those who are well versed in PSA methodology.

Some years ago work has started to allow for a more immediate and interactive use of the information contained in a PSA. The objective of these efforts was to create a "living PSA model", readily available for operational safety management.

Integrated systems which have been developed to structure PSA information make use of recent technical developments in the area of small computers. Due to its highly interactive and user's friendly characteristics these systems are particularly suitable for updating PSA information and for responding to "what if" questions.

Two such systems are the PRISIM (Probabilistic Safety Information Management System) developed for the U.S. Nuclear Regulatory Commission to assist plant inspectors and the ESSM (Essential System Status Monitor) developed in the United Kingdom.

Future Agency activities shall consider this trend by promoting needed research and facilitating the exchange of information.

A recent IAEA meeting reviewed the international experience in using small computers for PSA (Ref.8). Another area of growing interest deals with the use of PSA insights to evaluate the safety significance of operational occurrences. The combined NPP operational experience of Agency Member States represents a much richer source of operational experience than that which could be provided by any single country. Development of a consistent framework for evaluating this experience and for preventing accidents is the goal of activities to be initiated in this area.

PSA is today an invaluable tool to achieve a balanced plant design for current and for future plant concepts and to improve operational safety management. IAEA's programmes are reflecting the growing demands of Member States in this field. 11-17

References

1) "Status, experience and future prospects for the development of probabilistic safety criteria", Technical Committee Meeting IAEA, Vienna Jan. 1986 (TECDOC - draft).

2) "Probabilistic safety criteria at the safety function/system level", Technical Committee Meeting IAEA, Vienna Jan. 1987 (TECDOC - draft).

3) "Identification of accident sequences sensitive to human errors", Technical Committee Meeting IAEA, Vienna May 1986 (TECDOC 424, 1987).

4) "Experience with simulator training for emergency conditions" Technical Committee Meeting IAEA, Vienna Nov. 1986 (TECDOC draft).

5) "Improving Nuclear Power Plant Safety Through Operator Aids: Guidelines for selecting Operator Aids", IAEA, Vienna, 27-31 October 1986 (TECDOC draft).

6) "International experience in PSA", Lederman L. and Gubler R. International Topical Conference on PSA and Risk Management, Zurich, Aug.1987.

7) "Computer software for PSA level 1 analysis developed within the framework of the IAEA's inter-regional PSA programme". Gubler R. and Lederman L,, International Topical Conference on PSA and Risk Management, Zurich, Aug,1987.

8) "The use of Probabilistic Safety Assessment on Personal Computers for Operational Safety Management". Technical Committee Meeting Report, IAEA, Vienna, March, 1987 (TECDOC Draft). 11-18

TABLE 1 List of Major Meetings for 1988 Technical Committee on Evaluation Vienna, Austria of Reliability Data Sources February 1-5 (Lederman) International Conference on Man Machine Tokyo, Japan interface in the Nuclear Industry (Control February 15-19 and Instrumentation, Robotics, and Artificial Intelligence) in co-operation With OECD/NEA and CEC (Niehaus) Research Co-ordination Meeting on Comparison New Delhi, India of Cost-effectiveness of Risk Reduction February 22-26 among Different Energy Systems (Novegno) Research Co-ordination Meeting on PSA Vienna, Austria data collection from operating experience March 7-11 (Lederman) Technical Committee on Use of PSA for Vienna, Austria Probabilistic Safety Criteria Mar 28 - Apr 1 (Niehaus) Research Co-ordination Meeting on Vienna, Austria Standard Problems in Modelling May 2-6 Accident Sequences (Cullingford) Technical Committee on Use of PSA for Vienna, Austria Guidance in Re-licensing of Extended June 6-10 Lifetimes of Power Plants (Cullingford) Technical Committee on use of Expert Vienna, Austria Systems in Nuclear Safety September 5-9 (Cullingford/Swaton) Technical Committee on Hazard Control Location to be (Novegno/Asculai) decided October 17-21 11-19

Level of Probabilistic Level of Probabilistic Safety Criteria (PCS) Safety Assessment (PSA) Safety components Level 0 PLANT Safety Systems (Reliability Studies) Safety Functions Initiating Events Accident Sequences Core Melt Level 1 Containment Performance Large Release of Level 2 Radioactive Materials

PUBLIC Individual Risk HEALTH Societal Risk Level 3 Cost-benefit (effectiveness)

Table 2: Levels of PSC and required level of PSA to show compliance. n-20

Table 3: STATUS OF PSA STUDIES

Case Study Title Status/Comments & Author 1. Identification of First and second drafts reviewed by systems and oversight committee (OSC). components important Awaiting final version. to safety. H. Lambert

2. Backfitting Second draft reviewed by OSC. J. Young 3. New safety issues, Final draft ready. identification and ranking. R. Budnitz 4. Formulation and First draft reviewed by OSC. use of reliability Second draft received. data for a PSA. J. Fragola

5. A PSA peer review Final draft ready. with focus on specific issues. H. Lambert

6. Investigation of First draft reviewed by OSC. station blackout Revisions now being made. A number risk at the of other case studies have been Milestone unit 3 written based upon the PSA work of PWR. Northeast Utilities. These will be J. Bickel reviewed and published.

7. Making Nuclear First draft reviewed by OSC. Power Plant Opera- Revisions now being made. tional decisions using PSA informa- tion on personal computers. J.B. Fussel 11-21

Table 3 (Cont.) : STATUS OF PSA STUDIES

Case Study Title Status/Comments & Author 8. Use of Human Second draft reviewed by OSC. Reliability Analysis Few Changes needed. for PSAs and Plant High priority for publication. applications. G.M Hanaman 9. Use of PSA methods Second draft reviewed, to analyze common Ready. cause and common mode failure. P.J Amico

10. Identification of First draft reviewed, dominant accident Changes being made, sequences and components important for safety from PSA results. P. Kafka G. Reichart

11. Evaluation of First draft reviewed by OSC. operating experience: The German percursor study. P. Kafka G. Reichart

12. Probabilistic safety First draft reviewed, evaluation for a Changes being made, one-time variance in technical specifications. J. Bickel 11-22

Small Modular High Temperature Gas-Cooled E,eactors (HTGR): A Far-Reaching Inherently Safe Concept

W. Kroger

Institute for Nuclear Safety Research Nuclear Research Center, Jiilich Federal Republic of Germany

The behaviour of a reactor under loss-of-forced-cooling conditions is the central feature of its safety characteristics and illustrates its inherent safety features. It is known from comprehensive analyses that the efficiency of inherent heat transport mechanisms depends on design details, pressure level and boundary conditions of the primary circuit. Design characteristics of recent German modular HTGR's (HTR- Model, HTR-100) are such that even without any forced convection for after-heat removal maximum fuel temperatures stabilize in the range of 1600° C. These temper- atures do not cause any radiological danger to the environment. The TRISO-coated fuel particles remain intact up to 1700-1800° C, and a total failure is expected at about 2500° C. Furthermore, colder parts of the graphite core and reflector have a higL re- taining capacity for metallic fission products. Due to a strong negative temperature coefficient and high temperature resistance of the reactor, failures of shutdown sys- tems do not influence this statement significantly. Nevertheless, simple active systems are needed to control water ingress accidents following steam generator leaks. But even after additional failure of these systems (frequency of 10~6/aor less) only small amounts of fission products (e.g., 50 - 150 GBq of Cs-137), detached from metallic surfaces, would be released into the environment. Due to the chemical behaviour of hot graphite, corrosion effects in the case of air ingress have been studied carefully; the necessary conditions for "graphite burning" and subsequent massive release of fission products can be credibly ruled out for the HTGR-concepts. Small effort PRA's result in very low risk figures. No early fatalities and practi- cally no lethal cancer cases are computed for the very low accidental dose estimates. No emergency measures to mitigate damage in the vicinity of the plant, have been taken into account, and do not need to be considered. 11-23

Irradiation Behavior of Advanced Fuel Elements for the High Temperature Reactor under Normal and Accident Conditions

G. Pott, H. Nabielek, H. Nickel and W. Schenk

Julich Nuclear Research Centre (KFA) Federal Republic Germany

The Julich Nuclear Research Centre (KFA) has be>;n involved in development and irradiation testing of fuel elements for High Temperature Reactors (HTR) for 25 years. During this time several reactor concepts have been designed and developed employing a variety of fuel element concepts with various fuel operating conditions. Following the development, testing and utilization in AVR and THTR of mixed oxide

thorium/high enriched UOa (LEU) fuels. The UO2 fuel kernels are coated with an additional silicone carbide interlayer sandwiched between pyrocarbon layers. Coated particles and fuel element matrix components, as well as the integral fuel elements have undergone extensive irradiation testing in various reactors. The tests are carried out in instrumented capsules purged with He/Ne gas sim- ulating HTR nominal and extreme operating conditions. The fission gases produced during irradiation are swept out and analysed for their activity. For testing larger amounts of fuel elements the AVR Prototype Reactor is used. These experiments have shown in-pile failure fraction below 2xlO~5 and fission product-release orders of magnitude lower than in previous HTR fuels. The measurements indicated that not a single coated particle failed during irradiation (fig. 1). The comparison of calcula- tion and measured fractional release rate of Kr 85 confirms that the fission gases are generated from the Uranium contamination. The change of material parameters during neutron irradiation is determined by post irradiation examinations in hot cell facilities. The fuel element performance under temperature excursion accidents is investigated by heating experiments with irradiated fuel elements where the activity release is measured on-line. The tests are carried out in specially designed furnaces at temperatures between 1600 °C and 2500 °C. Examples of the resulting fission product release profiles are shown in fig. 2. The fraction of the end-of-irradiation inventory accumulated on a cold finger is displayed versus heating time. As expected releise is much faster at 1800 °C than at 1600 °C. 11-24

The sequence of Ag 110m, Cs 137 and Kr 85 release is quite typical for all heating tests.

PUilon 8M Aeltm Equivalent itltjte of R/B KrIB Jpn£e«£ticla

900° C

HFR"Is3 ""^* ~~* "?p™"'—™" —~

irradiation Tlm«

100 200 300 d

Pig. 1: FISSION GAS RELEASE KR 88, COMPARISON EXPERIMENT AND CALCULATION UO» LEU TRISO PARTICLES 11-25

1800°C HFR - K3/3 1600°C 11Om Ag HFR - K3/1 10"2-

10-4- 90 A Sr 137,C3 10'6*

65Kr

10 100 1000 Hasting Time (h)

Fig. 2: Flsalonproduct Raieasa from two Fuel Elements UOj Triso Parties Nuclear Technology and Safety II

Chairmen - D. Maron and L. Reznik III-l

Utilization o-F -tr.t-i

H. Hirshfeld and A. Nagler

Soreq Nuclear Research Center Yavne 7600, Israel

General Description

The Israel Research Reactor #1 is a 5 MW swimming-pool type reactor. It is -fueled with 93% enriched MTR elememts. The reactor is located at the Soreq Nuclear Research Center;it is operated by the Israel Atomic Energy Commission since June I960. The core consists o-f 24 to 30 -fuel elements, manufactured by CERCA (France) containing highly enriched uranium purchased in the U.S.A. The reactor is operated 4 days a week, 6 to 7 hours a day. The reactor is mainly used for: a) neutron radiography. b) activation analysis. c) radioisotope production. d) various industrial purposes.

The following irradiation facilities are available at the IRR-1: 6 radial beam tubes. 2 tangential beam tubes. 1 slow pneumatic transfer systen (rabbit). 1 fast pneumatic transter system (flexo-rabbit). 1 gamma cell. 6 irradiation positions in the reflector. in-2

The maximum thermal neutron flux in the reflecfor is 5xlO*3 n cm~a sec"1. Reactor utilization

1. Neutron Radiography Two neutron radiography facilities are installed outside the reactor pool using two of the radial beam tubes. These facilities are used for NDT purposes and are operated by the Non-Destructive Testing group at Soreq. Both facilities have thermal fluxes of about 10* n cm-= sec-1 and cadmium ratios of about 10. The L/D collimation ratios range from 250 to 800. The geometric image unsharpness of the facilities is below 20pm. The size of the analyzed objects are limited to 100 x 100 cm in one faciliity and to 40 x 40 cm in the other.

2. Activation Analysis Neutron activation analysis is available by means of a fast pneumatic system

3. Radioisotope Production Radioisotopes for industrial and medical purposes are produced in the reflector irradiation positions. Various tracing techniques using radioactive tracers are commonly II1-3

used by the R&D group at Soreq. For example, Br—92 in KBr used for water flow measurements at aqueducts and mapping of large water reservoirs. IRR-1 facilities were used to produce radioactive Qs*** for the development of an Os-Ir generator— a project that has b»en carried out in the last 5 years by the Radiopharmaceutic Department at Soreq. The Os-Ir generator has an advantage over Tc"***" as a medical tracer (mainly in cardiology) because of ir**lm short half life <4.88 sec).

4. Patino of oeoloaical samples A facility for the irradiation of geological samples for Ar—Ar dating has recently been established at the IRR-1 reactor. The Ars*-Ar*° dating method is based on the K3"-Ar*° radiometric clack and surpasses the latter in both precision and range of rocks and minerals that can be analyzed. The geological samples are irradiated in a rotating cadmium -shielded capsule. The thermal to fast neutron flux in the capsule is less than 1/50.

5. Archeoloav Irradiation of ceramics from archeological excavations is used to determine the origin of the ceramics. This method is very helpful in understanding the commercial relations between peoples several thousand years ago.

6. Enhancement of precious stones A technique was developed for changing the color of diamonds and topaz gemstones. The procedure is based on irradiation in the reactor together with appropriate supplementary laboratory treatment. The basic colors which can be achieved in diamonds are: green, gold and cognac. Topaz stones turn blue. The color centers formed by this process arm stable,. III-4

7. Neutron-Transmutation-Dooina of Silicon NTD of Si is a well established technique to manufacture well defined n-type silicon crystals. The nuclear reaction Si30

Future possibilities

The irradiation facilities are presently not fully exploited. The following list presents some topics which we are capable of handling, should there be enough economic motivation. — Study of the behaviour of nuclear fuel and construction materials for nulear power plants under irradiation conditions. This kind of work has been carried out intensively in research reactors in Europe over the last years. - Basic research using the beam tubes. — Color enhancement of other gemstones. - NTD of silicon on a commercial scale. III-5

i-fc *• «=>«"» o-P IRFt—1

A. Nagler and H. Hirsh-feld

Soreq Nuclear Center, Yavne, Israel

E. Elias and S. Kaizerman

Department o-f Nuclear Engineering, Technion, Haifa, Israel

Introduction Until recently it was believed that melting o-f fuel elements -fallowing a Loss o-f Coolant Accident cannot occur in a 5 MW research reactor core. To study the problem of core heatup -following LOCA, the 3D-AIRL0CA code has been developed by a group in the RERTR program (Reduced Enrichment for Research and Test Reactors). The code was used to model the IAEA generic reactor core <3S>, after being validated by the RERTR group through comparison with experiments *3>. The data from thirteen tests at powers from 0.39 MW to B.87 MW were compared with calculated results, in which the code predicts the experimental maximun temperature rise within error of -10.17. to +14.4%. In the present work an input data model was prepared for the Israel Research Reactor (IRR—1) and the sensitivity of II1-6

the calculated peak core temperature, to some input parameters, was investigated. The final results reveal that in the current operation schedule at a power level higher than 2.5 MW, the maximum fuel temperature in the core may exceed the melting point during a LOCA; therefore, an ECCS is essential in the IRR—1.

Core model A LOCA was postulated to occur after eight weeks (4 days a week, 6 hours a day) of operation, in a reactor which is not equipped with any kind of an ECCS. Time zero in the LOCA transient is taken as being that instant at which all coolant water is completely lost from the core fallowing a 100X break of the 10 inch inlet/outlet pipe. The difference in time between reactor shutdown and loss of coolant (lag time) is about 20 minutes. The assumption is that as long as the fuel plates are partly covered with coolant the fuel temperature cannot esxceed 100 QC. It is also assumed that no blockage is present: after the water loss, so that air cooling by free convection can occur. The residual decay-heat (as calculated by the ANB 5.1 standard) was applied to the IRR-1 operation schedule. Decay-heat distribution during the accident is assumed to be the same as at normal operating conditions. In this study, no account was taken of any decay energy loss from the core by gamma radiation. The heat transfer models in the code include axial (vertical direction) heat conduction along each of the reflector and fuel elements, and internal free convection heat loss. Heat is transfered from one element to the neighboring elements by thermal conduction and radiation across the air gap separating the elements. Heat is also lost to the environment by free convection and thermal radiation rii-7 on exposed surfaces, such as tops and sides of core edge elements. The grid plate serves as a horizontal conduction path between elements and heat is transfered from the bottom of each element to the grid plate by means of conduction as well as radiation across the air gap between the element end fitting and the grid plate hole. This code calculates the transient temperatures for the fuel, air, grid plate and reflector element.

Results and discussion Fig. 1 shows the peak temperatures vs. time curve for the IRR-1 core under standard operating conditions (8 weeks with 4 days a week? 6 hours a day) at various power levels.

Hfit TEMP-MELTINS POINT TEMPERATURE tow

too

aoo

700 jte 600

Ju 400

MO

& 200

too

3 * TIME

Fig. 1 Calculated peak core temperatures vs. time during LOCA III-8

The core temperature behaviour as a function of reactor power can be divided into three distinct regions: a) below 2 MW the reactor is safe under LQCA conditions, b) in the 2.5 MW range the temperature reaches the Al melting, and a critical time can be defined in terms of an engineering margin; c) above 3 MN core power, a time of onset of Al melting can be defined; beyond that time the code calculation continues assuming that no melting takes place. Therefore, the curves ar& hypothetical (shown with dashed lines in Fig. 1). For 4.5 MW, which is the normal operating p^wer of IRR-1, the temperature reaches the melting poing point after about 20 minutes. Sensitivity studies reveal that the temperature rise is relativity insensitive to variation in the following parameters: lag time, fuel channel Nusselt number, ambient temperature and initial temperature distribution. The temperatures were found to be more dependent on the operating schedules.

Refrences C13 Y.K CHEUNG and J. GLOVER, "3D-AIRL0CA, LQCA Analysis Code for Research and Test Reactors", ANL,1980.

C23 "Research Reactor Core Conversions from the Use of Highly Enriched Uranium to the use of Low Enriched uaniwn Fuel Guidebook", IAEA-TECPOC-233, International Atomic Energy Technical Document, 1980.

C33 D.K. Warinner, "Comparison of the Aerospace Systems Test Reactor Loss-of-Coolant Test Data with Predictions of the 3D-AIRL0CA Code" JAERI-M 84-073, 1984. III-9 Homogeneous Non Equilibrium Critical Flow Model

S. Dickroan, S. Kaizerman and E. Elias

Department of Nuclear Engineering Technion, Haifa

In the hypothetical event of a pipe break in the primary heat transport system of a light water nuclear power plant, the discharge flow rate largely determines the decompression rate of the system, the pressurization rate of the containment and the forces on structures and various equipments. Therefore, the evaluation of the blowdown flow rate is essential for the analysis of system response during a hypothetical Loss of Coolant Accident (LOCA). As a result of the actual conditions (pressure and temperature) in the primary system, the flow is choked (critical) during most of the blowdown transient.

The purpose of the present report is to present a new theoretical two-phase critical flow model. In fact, the proposed model is an extension of the Homogeneous Equilibrium Model (HEM) [1], in which the thermal equilibrium constraint between the phases is relaxed.

In the present model, the two-phase mixture is assumed to behave like a homogeneous fluid in which the two phases could have different temperatures. The vapor is assumed to be in saturated state while the liquid, due to the finite rate of vapour gent-ation, could also be in metastable superheated states.

The thermal non-equilibrium phenomena are considered to be significant and are taken into consideration in the present work for the cases of subcooled liquid stagnation conditions. The thermal non- equilibrium effect is taken into account by a new non-dimensional parameter, DD, defined as:

T - Tg(p) DD = f~=-f where:

= liquid temperature

g a saturation temperature corresponding to the system pressure T = stagnation temperature 111-10 The non-equilibrium coefficient ranges between 0 .<. DD < 1. The HEM model is obtained in the present model as a particular case for DD-0. DD=1 corresponds to the case of isothermal expansion,

In the proposed one-dimensional model, the flow is assumed to be isentropic. A schematic description of the model is given in Fig. 1. First an arbitrary value is chosen for DD and an exit pressure is searched for to maximize the mass flow rate. The value of DD is then varied in order to obtain a Mach number of unity at the exit. The model predictions were compared with experimental data from the large scale Mar*/iken tests [2].

The influence of the thermal non-equilibrium coefficient DD on different critical flow parameters was analysed. It was found that, for a specified value of the thermal non-equilibrium coefficient, DD, the critical flow velocity becomes equal to the isentropic speed of sound, even for subcooled stagnation conditions, as depicted in Fig. 2. This is a very important feature of the present model {in contrast with HEM) which is also consistent with the homogeneity as&umption, i.e., the maximum mass flow rate in the critical section is always sonic as in a single phase flow. Thus, the model avoids the paradox of the HEM model which predicts critical Mach numbers greater than one for subcooled stagnation conditions due to the jump in the isentropic speed of sound in the region of very low flow qualities.

Maps of the non-equilibrium coefficient corresponding to critical sonic flow as functions of the stagnation conditions were derived. Moreover, compared to the Marviken experimental data, it was found that the present model provides better predictions for the specific values of DD, than the HEM model.

Due to the fact that the critical flow is always sonic, the present two-phase critical flow model was successfully applied to find the boundary conditions for a two-phase free jet model in which methods similar to those from single phase free jets were used. III-ll

References

[1] Lahey,R.T., and Moody,F.J., "The Thermal Hydraulics of a Boiling Water Nuclear Reactor", American Nuclear Society, 1977

£2] The Marviken Full Scale Critical Flow Tests, "Summary Report", EPRI NP-2370, vol.1, RP956-1, Marviken, MXC-301, 1982. in-12

\ 1?, h0 / stagnation conditions

DD assumed assumed

TS = TS(P) nonequilibrium T^VDDa-TJ liquid temperature

state eqs. x =(so-sl)/ h = h, + x (k-h.)

= v=vL +x (v,-vL)

energy eq.

continuity eq.

maximum flow rate conditions

Yes speed of sound

Mach number

sonic condition

critical flow

Fig. 4.. - Flow Chart for the Computation of Critical Flow with the Present Model 111-13

R - 35.01 ( BAR )

4 •

"•\._

"\ 3 "\ -O.O82. "\. XU = 6 '•\.

a \ 5 • \ =5° \ \

\

\

\

~~~"^——.-^.____ -0.O2.5 \

\

—-_^ \

~~~~~-— ^-^__ _/ =: ::=: 1 - . ^ —. .—-— 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 DD - Non Equil COG/. i. XU--O.OI7D >'."-D.0?<6 tU'-O.OII7

Fig. £. - Variation of the Mach number in the critical cross section as a function of the nonequillbrium coefficient for various degrees of subcooling in-14

Critical Two-Phase Flow

R. Dagan, E. Wacholder and E. Elias

Department of Nuclear Engineering Technion, Haifa Israel

The critical flow of a single component two-phase mixture has already been studied 50 years ago, but only in the last two decades the subject was accelareted due to its importance in reactor safety analyses. A sudden break in the reactor (especially in PWR) cooling fluid circuit may cause a sudden drop in pressure followed by flashing of the fluid and inducing a two phase flow. The knowledge of the maximum flow coming out from the broken circuit is vital in this case as it determines the time for core uncovery. This maximum two-phase flow is known as a critical or choked flow. In single-phase flow the phenomena of critical flow is well understood and can be easly computed once the initial stagnation conditions and the geometry of the flow channel are given. However, critical two-phase flow is more complicated due to the existence of different flow patterns (flow regimes) such as bubbly, slug, churn or annular flow and the possibility of thermal non-equilibrium between the phases. The different aspects of the problem were disscused in several articles and some analytical models had been suggested [1-5].

Analytical models for two-phase critical flow calculations are usually divided into three categories:

1. Homogeneous equilibrium models (HEM) which assume an isentropic flow with both phases at thermal and mechanical equilibrium.

2. Equilibrium non homogeneous (slip) models which consider thermal equilibrium with the phases traveling at different velocities.

3. Thermal non equilibrium models in which the phases are at mechanical equilibrium but thermal equilibrium is not achieved.

The thermal non equilibrium effects sstsem to have major influence in calculating the two phase flow mainly as subcooled fluid expands through short pipes. The reason is that there is not enough time for heat and mass transfer to achieve equilibrium due to heat resistances (e.g., at the interface). Moreover, the relaxation time at such rapid expansion may be of the same order of magnitude as the traveling time 111-15

of the fluid in the circuit. An extreme model which allow no vapor generation during the expansion (frozen model) generally overestimates the critical mass flux. The HEM, which is the other extreme case, usually undevpredicts the experimental results. Therefore, partial evaporation is expected.

Henry and Fauske [6] allowed, in their work, the exit quality to differ from the equilibrium quality by using an apropriate empirical coefficient. The model assumes that the liquid temperature is constant and no heat transfer occurs during the expansion. This, in turn, results in a large temperature difference between the phases at the throat where violent heat transfer takes place followed by a rapid change of the vapor quality and density. The liquid density is assumed to remain constant and the no-slip assumption was considered valid although the pressure drops sharply. The amount of phase exchange (at the throat) is, as mentioned before, determined by an empirical coefficient. Henry and Fauske's model was coded in this work and tested against experimental data covering a wide range of conditions. About 1500 data points were used in the validation.

The results shown in Fig. 1 demonstrate the quality of the model's prediction. Similar comparison with the HEM (not shown in Fig. 1) emphasizes the importance of the thermal non equilibrium effect. There is a significant improvement in the predicted critical flux compared with HEM. Nevertheless, for some cases there are large deviations between the experimental and predicted results as shown in Fig. 1. The thermal non equilibrium effect which is governed mainly by the pressure cannot explain these deviations. The large scattering of the data by Sozzi and Sutherland [7] shown in Fig. 1 points out that other effects such as frictional and irreversible entrance losses may also be important. For a limited range of flow channel configurations and stagnation conditions the model of Henry and Fauske provides an adequate results for the two-phase critical flow as shown in Fig. 2 comparing the large scale Marviken [8] experimental data.

References

1. Ardron, K. H. and Furness, R. A., 1976, "A Study of critical flow models used in reactor blowdown analysis" Nucl. Eng. Design 3S, 257-266. 111-16

2. Burnell, J. A., 1947, "Flow of boiling water through nozzles, orifices, and pipes" Engineering, 12, 572-576.

3. Saha, P., 1978, "A Review of Two Phase steam-water critical flow models with emphasis on Thermal Non Equilibrium" NUREG/CR-0417.

4. Wallis, G. B., 1980, "Critical Two-Phase FLow" Int. j. Multiphase flow, 6, 97-112.

5. Isbin, H. S., 1980, "Some observations on the status of two-phase critical flow models" int. J . Multiphase Flow 2, 131-138.

S. Henry, R. E. and Fauske, H. K., 1971 "The two-phase critical flow of one component mixtures in nozzles, orifices, and short tubes" J. of Heat Transfer, 179-187.

7. Sozzi, G. L. and Sutherland, W. A., 1975, "Critical Flow a* Saturated and Subcooled Mater at High Pressure" NEDO 13418.

8. The Marviken Project, 1982, The Marviken Full Scale Critical Flow Tests, EPRI Report NP-2370. Henry Fauske

90

80-

70

30 40 50 60 70 80 90 G predicted (Mg/m2-s) + + + Ardron * * Bo i v I n X X X Bryen O D D Fink, O O O Jeandey ^ & Reocreux « Jt »t Sc»nK»e»t Y Y Y So 2 2 ; Henry t'austce (Marviken)

10

60

^40

0)

t/) o £ o2o

10

0 10 20 30 40 50 60 70 G predicted (Mg/m2-s) I + + + Tea tO4 * * * Tea tO6 X X X Testl3 • D • T«» t 18 o o o Its ( 111-19

Experimental Apparatus for Quantitative Measurments in the Precursory Cooling Regime During Bottom Reflooding

Y. Barnea and E. Elias

Dept. of Nuclear Engineering Technion, Haifa

I. Shai

Dept. of Mechanical Engineering Ben-Gurion University, Beer-Sheva

During the early phases of a LOCA, the fuel rods wall temperature increases quickly because of the stored and decay heat and the reduced cooling effectiveness. Bottom reflooding by cold water must provide an acceptable coolabilifcy as soon as pos- sible to avoid overheating of the fuel above metallurgically prohibited values. There- fore, it is necessary to understand the heat transfer mechanisms in the various flow patterns taking place in the core during LOCA. Two Row regimes are particularly important, namely the Inverted Annular Flow (IAF) and Dispersed Flow Film Boil- ing (DFFB). These regimes, which may exist ahead of the "quench front" provide a precursory cooling effect and enhance the overall heat transfer from the fuel rod. Numerous experimental and theoretical studies were carried out worldwide, in the recent years, to understand the regimes mentioned above [1-3]. It is the purpose of this study to concentrate on the following aspects which have not been yet completely understood [3]: A. The thickness of the Vapor layer and vapor-liquid interface waviness in IAF. B. Temperature and velocity of the liquid in IAF. C. Vapor layer temperature in IAF and DFFB. D. Droplets diameters and velocities in DFFB. E. A criteria of transition between IAF and DFFB. To accomplish these tasks, an experimental facility was designed and built in the heat transfer laboratory of Nuclear Research Center Negev. A series of tests were planned using an annular test section of different initial wall temperatures, inlet velocities, inlet liquid temperatures and ambient pressures. A schematic description of the experimental apparatus is given in Fig. 1. It consists in a main open flow 111-20

path through a centrifugal pump, a filter for the demineralized water, a preheater of inlet water, a flowmeter and the lower plenum of the test section. Upstream, the test section is equiped with a phase separator where the vapor and the liquid are measured and collected simultaneously. An auxiliary by-pass closed loop includes the presurizer and filling vessel, an additional flowmeter and the ion-exchanger tank. The required flow conditions are established in the by-pass loop and then applied to the test section. The test section is a 45 mm OD, 2.5 mm thick and 2 m long AISI 304 steel or Inconel 600 tube, surrounded by a 56 mm ID, 3 mm thick Plexiglass tube. A detailed description of the test section is presented in Fig. 2. The present study concentrates on novel local measurement. Microscopic measure- ments include sixteen 0.5 mm wall temperature ungrounded thermocouples spot- welded into the inner metal surface at twelve elevations by a very complicated tech- nique. There are also 10 bare thermocouples (no. 30) inserted through the external tube, at four elevations, to measure vapor temperatures during various stages of re- flooding. Macroscopic local measurements include the measuring of: void fraction by a Gamma-Ray Densitometer (home maid), drop diameters and velocities by filming and waviness of liquid-steam interface by the densitomeier and a device based on the "pulse-echo" method. Additional common measurements are : a) Inlet water temperature, flowrate and static pressure; b) Outlet flow conditions, namely vapor temperature and separate phase flowrate; c) Heat input, by measuring direct current and net potential difference on the test section. Sampling, recording and data processing through an IBM-PC programed as an advanced data-logger, is also one of the achievements imprecedented in this study.

1. Yadigaroglu, G., Nuclear Safety 19 (1), 20, 1976. 2. Barnea, Y. et al, Trans. Israel Nucl. Soc, 12, 1986. 3. Barnea, Y., Proposal for D.Sc. Thesis, Technion, Haifa, July 1986 (in Hebrew). PRESSURE

I to

F'%- I

r r ;כ· ;כ·

G.I:I ' .2 ן.

~._~_._[ Qכ i ~ jך~~ ;;=- I -. r -=..ר·- o

,-~,= -.:-fPJ ,

L, _l_YJ- ז-ר. ·1 ·1

· 1· I I

! !!

@) @)

I • • I I I @

o o 1

o [

j

I@ @ I@

@ @ ~.: ~.: -- • :'1 - I .•

,,-,Oi' jי t-·(V t-·(V • . ~~-- o .- '. _".::.': .., -ן -- ~ .-.@ (!)

---.. ,

@ o . .

-'~"0 0-~,_-'~"0 .

u u . . • רכח למ מ'ם כנ'םח 12 12

• z" C!.OU U.O c,unLCU ם

. . • חחחוו מלו' מ'כל ii ' 11 cttou a

'' uyUC,I., uumUUUIIl ז.. u"'' ii.ccU.C u u . . • בוורם טרמ' לצר,ר בר מ 10 10

' ' cLU., CI1L!:'ILDQוt'LL OLU .ז~UHCun!:LL CKU '"010

. .

הלחץ מפל למד'רח ח I'!J I'!J

• r 9כ

' ' L C,UtlMנE.IG.,UU CUL.Lr.UL.LUU UGCUGc,. UCU ''6 ! ! . . 8 8 U I ,.(. הרופו ל יים רט ן צ 'lJ 'lJ I) I) [11 [11

''·(' IJ..DD Hnc,C L'LtGL'LIGIl i iSI") 1 .וCH.U.a ט.KU1..~ !JL D ''8 ~~ן~ג~ ~~ן~ג~

ji ji . . . הח'מום נוף

' ' O IO.גt UU.UI!U rנl<זttt ו'" כ' '1E':'~ ~:::~:,~,..~:

6 6 U PYRE א טפח .מ

.. ..

ד

ddג 3H א*c GUU X3bו UtICIGun ''9 ::~'!' ::~'!' . . . הדנם ף א'ם מ'כל אל S S I

' ם L u.0נ.U L'Uא.ם,

אר'tב הפךרח

" ם.LQ א.GLLUU NL ו.בUGLL מ'כל axU.Cc,t '" !.>ר

. . . וחושח אבקטרורח ג

Q'::: Q'::: .'

& & ". ". "to

~ ~ 1

<.!) <.!)

' UlmU ז Cdt1LILUU אr ' NCClClLtL ~ ~

. . . .

צג'רר, פס'

. ונ L ~:ו. • GQ ן'Z . וi>

. . . הךסופק ארון אל

~:~~'"~' I

_- -- ......

' I.'OGdונ LII א ,cא ' L:.:-.:f ~~ ::~ ._ ._-:= ~ '.. : ':: ~~.... :i l: : : I-22 I-22

(:i: - 111 11 111-23

The Effect of Precursory Cooling on Rewetting of a Solid Cylinder

Shmuel Olek Swiss Federal Institute for Reactor Research (EIR) CH-5303 Wurenlingen, Switzerland (present address)

1 Introduction

The problem of rewetting of a hot surface is of interest mainly in the nuclear indus- try, regarding the emergency core cooling of a nuclear reactor after a postulated loss of coolant accident (LOCA). Thi3 problem finds its applications also in a number of industrial processes, such as metalurgical treatment, start-up of liquefied natural gas (LNG) pipe lines, filling vessels at room temperature with cryogenic liquids, or cooling of large superconducting magnets by such liquids. When a very hot surface is suddenly cooled by a liquid, the latter boils and a blanketing vapor film is generated, which separates the liquid from the wall. Convective and ra- diative heat transfer remove heat from the solid, decreasing its temperature, and after a while the vapor film becomes unstable. Local intermittent wetting of the surface in the transition boiling regime that follows increases the cooling rate and a wet patch is formed. This wet patch eventually spreads and results in the propagation of a stable quench front. The quench front region is one of vigorous boiling. It is the boundary between the wet region, the temperature of which is lower than a threshold tempera- ture (the rewetting temperature) and the dry region whose temperature is above this threshold temperature. In the wet region heat is removed from the wall by convection to the liquid, nucleate and transition boiling, whereas in the dry region heat is transferred by dispersed-flow and/or inverted annular film boiling in the case of bottom flooding, or by a mixture of vapor and droplets in the case of top-spray cooling. The cooling of the dry portion of the rod by direct convection and radiation to the coolant is usually termed the precursory cooling. A number of two-dimensional models were proposed for slab geometry, where pre- cursory cooling was taken into account [1,4]. The present model considers rewetting of an infinitely long solid cylinder, including precursory cooling. It is shown that pre- cursory cooling may greatly increase the rewetting velocity. The results of the present solution by separation of variables are in excellent agreement with a Wiener-Hopf tech- nique solution to this model by Olek [5] and to the solution by Evans [6] for the case of no precursory cooling. 111-24

1.1 The physical model

It is assumed that the wet region is quenched to a temperature Tt, while the far pre- quenched zone is still at the initial wall temperature Tw. The heat transfer coefficient h in the wet. region is assumed to be constant, whereas the heat flux from the surface of the dry side is supposed to have the following functional form:

a" — —eaz q N

where Qo is the heat flux at the quench front associated with the difference between the rewetting temperature Jo and the saturation temperature T,. N is a model parameter assumed to depend on the flow rate, a is a model parameter controlling the solid length influenced by the precursory cooling, and I is the axial coordinate of the (f,z) coordinate system, which moves along the rod at the constant rewetting velocity M.

2 Mathematical formulation and solution

The quasi steady state heat conduction equation in the aforementioned moving frame of reference can be written in the form:

d29 1 d9 d20 d9

with z==l K } Tw-Ts RR R k * where R is the radius of the rod and B is the dimensionless temperature, p, c, k are the rod density, its specific heat at constant pressure, and thermal conductivity, respectively. Eq. (1) is subjected to the following boundary conditions:

z —> —oo 9 ->0 (3)

z —* oo fl- • 1 (4)

86 T = 0, ail .z = 0 dr (5) d$ r = 0, --> 0 = B(9-1) dr (6) 59 r=0, z < 0 dr (7)

z = 0. r = 1 9 — 9n (8) where 111-25

0oR (B\TO-T, hR

Separting variables and solving for the temperature distribution in the wet region (region A) gives:

ll2 9A = 1 + f] AnJQ{\nv) exp {[f - (4" + K) \z} (10) where j4n are constants and the eigenvalues \n are the positive roots of the following equation:

An/i(An) - BJo(Xn) = 0 (11) The solution for the temperature distribution in the dry region (region B) consists of two steps: 1. homogenizing condition (7), while keeping homogeneous condition (5) by introducing a new dependent variable (for similar treatment see, e.g., [7]). 2. solving the resulting nonhomogeneous differential equation (see, for example, [8]). Performing these steps, one finally obtains the following expression for the temperature distribution in the dry region: 2 b 9B = \r Ae * + ^{BJ^WI)1'3)* + ^g"_ ^UOM (12) with the eigenvalues /?„ being the non-negative roots of:

MPn) = 0 (13) and qBn are constants given by:

n=1

O n = 2,3,... (15)

Proceed by equating the temperature distributions in the wet and dry legions and their axial derivatives at z — 0. Acting with /„' J0(/3nr)rdr on both sides of the resulting equa- tions and eliminating the constants Bn,n = 1,2,... between them after using the noted operator, results finally in the following linear system of equations for the constants An:

CO

71=1 where

7T5 7^>~\ m = *>*)•.• in-26

For numerical computation only a finite number ,say N, of simultaneous equations in eq.(16) are used for determining the constants A.n. The temperature distribution must satisfy either of the following equations:

N A 0A(l,O)^t>o = l + Y, »HK) (17) n=l

( ffl +^_ y) (18) to a desired accuracy. The guess of a Peclet number (or rewetting velocity) is changed accordingly in an iterative procedure.

3 Results and discussion The variation of the Peclet number with the dimensionless temperature for different magnitudes of the precursory cooling is presented in Fig. 1. The results of a one- dimensional solution are also shown. The latter solution takes the form: P = Y-B/Y (19) where

From Fig.l it can be realized that precursory cooling may greatly increase the rewetting velocity. The present solution was compared with a Wiener-Hopf technique solution by Olek [5] to the same model. For the special case of no precursory cooling it was compared to the solution of Evans [6]. In both cases an excellent agreement was obtained.

Acknowledgements The author wishes to thank Prof. George Yadigaroglu for motivating this study. 111-27

100.00

10.00-

E 3 1.00- C 1-D solution 2-D solution 7£5 o.. 0.10-:

0.01- 0.0001 0.0010 0.0100 0.1000 1.0000

Nondimensional temperature , i90

Fig. 1. The dimensionless rewetting velocity P vs nondimensional temperature 0O> for various magnitudes N of the precursory cooling.

References [1] Edwards, A.R. and Mather, D.J., 1973, "Some U.K. Studies Related to the Loss of Coolant Accident", Water Reactor Safety Meeting, Sale Lake City, March 26-28, CONF 730304, pp. 720-737. [2] Sawan, M.E. and Teraraz, H.M., 1981, "A Three Regions Semi-Analytical Rewet- ting Model", Nud. Eng. Des., Vol. 64, pp. 319-327.

[3] Hsu, C.-H.,Chieng, C.-H. and Hua, T., 1983, "Two-Dimensional Analysis of Conduction-Controoled Rcwetting with Internal Heat Generation", International Conference on Numerical Methods in Nuclear Engineering, Montreal, Canada, Sept. 6-9, CNS/ANS(spons. G.J. Phillips (Ed)), ISBN-0-919784-02X, Vol. 1. [4] Dua, S.S. and Tien, C.L., 1976, "Two-Dimensional Analysis of Conduction- Controlled Rewetting with Precursory Cooling", Trans. ASME, Ser. C, J. Heat Transfer, Vol. 98, pp. 407-413.

[5] Olek, S., 1987, "Rewetting of a Solid Cylinder with Precursory Cooling". Submitted for publication in Appl. Sci. Res.

[6] Evans, D.V., 1984, "A Note on the Cooling of a Cylinder Entering a Fluid", IMA J. Appl. Math., Vol. 33, pp. 49-54.

[7] Yeh, H.-C, 1977, " Solving Potential Field Problems in Composite Media with Complicated Geometries", J. Appl. Phys., Vol. 48, p. 4428.

[8] Yeh, H.-C., 1976, "Extension of the Method of Solving Potential Field Problems for Complicated Geometries", J. Appl. Phys., Vol. 47, p. 2927. in-28

ISOLATION OF SMALL MALFUNCTIONS IN A NUCLEAR POWER PLANT USING A BOOLEAN SIGNATURE ALGORITHM

Zvi Covaliu Licensing Division, Israel Atomic Energy Commission Tel-aviv, Israel

Yakov Ben-Halm Department of Nuclear Engineering Technion - Israel Institute of Technology Haifa, Israel

Introduction The management of malfunctions in complex systems gains increasing importance as a tool for improving reliability and performance stability. This is particularly true of nuclear power plants, where adding the safe- ty and economic aspects makes the detection and isolation of incipient failures a vital need. Among the tasks involved in malfuction management, the isolation of malfunctions is concerned with identifying the malfunctioning system's subunit or sensor. This is usually preceded by an alarm and followed by an estimation of the failure's extent. Several approaches to malfunction isolation exist: from frequency domain techniques [6J through expert sys- tem methods [2] to algorithms based on a state-space formulation [3,7,4, 5,1]. In this work an existing state-space approach algorithm was applied to an extensive model of a PWR power plant.

Methods Basically, the malfunction-isolation algorithm [1] generates at each time step, on the basis of measurements, a binary residual, which is then compared with a precalculated catalog of binary malfunction signatures. By this comparison a very broad range of malfunctions may be isolated. 111-29

The catalog of binary failure words is prepared on the basis of the nor- mal dynamics and measurement models alone, and, unlike with many state- space based methods, no specification of the form or magnitude of the an- ticipated malfunctions is required. The only limitation of the algorithm Is that the linearity of the system must be preserved, so only small pro- cess malfunctions are dealt with succesfully. Binary failure words (vec- tors) have the same dimensions as the state vector; process malfunctions are characterized by a single specifically-located "1" bit, while sensor malfunctions - by a typical constellation of "l"s. When some of the state variables are non-measurable, an estimation technique is normally em- ployed to extract their values from the measurement vector. • The algorithm was applied to the H B Robinson PtfR power plant, des- cribed and modelled by Kerlin et al. [3]. The dynamic model, consisting of a set of linear first-order coupled differential equations, describes the reactor's core, pressure vessel, pressurizer, steam generators and various pipings by 21 state-variables. As the first stage of this work, the model was discretized and its adequate response to various perturba- tions, including process noise, was numerically verified. Measurement noise was also included In the dynamic time-invariant model. The algorithm's isolation capability was tested by simulating a wide variety of malfunction?. This included diverse locations (subunits and sensors corresponding to various state-variables), forms (bias, step, ramp, wave, etc.), intensities and combinations (two or more simultaneous malfunctions). To begin with, the typical time stability of the (analog) residual mean was explicitly confirmed for many failures. Then, complete measurement of the state-variables was assumed and the isolation of (a) process and (b) sensor malfunctions was tested. As the first results of sensor malfunction isolation were disappointing, several alterations and extensions of the original algorithm have been devised and examined. Among others these included: - using a partial compliance criterion (instead of the complete one) when comparing the binary word with catalog signatures; employing a batch of 17 threshold-catalogs (instead of and including the basic one). Finally, the problem of partial measurement was addressed. As state estimation by standard optimal methods, such as the Kalman filter, is at 111-30

odds with malfunction isolation [1,4], a different approach was used: the residual elements unbiased by the non-measured state-variables were iden- tified and only these were used by the algorithm. In practice, only 15 (at most) of the 21 state-variables of the nuclear plant model are meas- urable, resulting in a partial binary word of only 13 bits. This, of course, somewhat degrades the performance of the algorithm.

Results and Conclusions

Some representative results of the application of the algorithm in the case of complete measurement are summarized in the table below with respect to both process and sensor malfunctions. The performance degrada- tion in the incomplete measurement case manifests itself mainly when at- tempting to isolate process failures in the core region and by an in- creased degeneracy problem with some sensor malfunctions. It has been shown that the extended algorithm detects and isolates accurately a very wide range of process malfunctions with high sensitivi- ty, extremely low false-alarm rate and almost promt response. The only limitation concerns some core-region failures in the partial-measurement case. As to sensor malfunctions, the performance of the method is satis- factory, except in some particular, predefinable cases. However, some consequent work is yet to be done in order to overcome these and other minor drawbacks. Several possible approaches to it are suggested accor- dingly.

References

[1] Y Ben-Haim, Malfunction isolation in linear stochastic systems: ap- plication to nuclear power plants, Nuc. Sci. Eng., 85., pp. 156-166, 1983. [2] J Christenson, T Graae and H Roggenbauer, Implementation of an auto- mated status analysis system in an operating nuclear power plant, Nuc. Eng. Des., 67, pp. 297-304, 1981.

[3] T V Kerlin, E M Katz, J Z Thakker and J E Strange, Theoretical and experimental dynamic analysis of the H B Robinson nuclear plant, Nuc. Tech., 30, pp. 299-316, 1976. 111-31

[4] T H Kerr, Statistical analysis of a two-ellipsoid overlap test for real-time failure detection, IEEE Trans., AC-25, pp. 762-773, 1980.

[5] F Hurata, K Kato, F Tomizava and I Sumido, Development of a diagnosis system for a boiling water reactor, Nuc. Tech., 44, pp. 104-117, 1979. [6] J A Thie, Power reactor noise, American Nuclear Society, 1981. J7] J L Tylee, A generalized likelihood ratio approach to detecting and identifying failures in pressurizer instrumentation, Nuc. Tech., 5_6_, pp. 484-492, 1982. [8] A S Willsky, A survey of design methods for failure detection in dy- namic systems, Automatica, 12., pp. 601-611, 1976.

Table of representative results of algorithm's performance

Malfunction characteristics Location i capability type source form relative (*> intensity

process Tf step .001 100 process T* step .0004 88 process T£ step .0003 10 process Tf wave (3 sec) .005 97 process Tf ramp (4 sec) .005 97 process Tf random .005 87 process Tm step .001 100 process PP step .002 100 sensor Tc step .01 10 sensor T« step .01 100 sensor TLp step .01 50 sensor P» step .02 0 sensor PP ramp (4 sec) .32 99 sensor TC2 step .1 100

State-variable notation; T - temperature subscripts; P — pressure LP — Lover Plenum f - fuel s - steam P - Primary loop Ci - coolant node 1 m - metal C2 - coolant node 2 ni-32 The Israel Electric Corporation Meteorological Tower in the Negev - a First Examination of the Data

Y. Balmor, A. Gutman and S. Kovacs

Israel Electric Corporation Ltd., Environmental and Efficiency Control Department, P.O.Box 10, Haifa, ISRAEL

Since March 1987 the Israel Electric Corporation operates a meteorological tower in Shivta, in order to collect data for the environmental impact evaluation of a nuclear or a coal-fired power station in the Negev area. The meteorological tower has been designed and constructed on the basis of US NRC, IAEA guides and IAEC requirements. (See list of abbreviations below.) Wind, temperature and relative humidity are measured at three levels: 10, 60 and 90m above the ground. Rain, net and global radiation are measured at the base level. Comprehensive QA and QC programs are applied in order to achieve at least a 90i level of data capture. All data sensors are scanned each second and logged data are transmitted to a programable data logger. The latter averages the data for 10 minutes periods and checks for data consistency. The 10 minutes averages for each of the 34 channels are transferred via cassette tape recorder to an IBM/PC computer for further processing. From these data, the 1 hour averages are computed with specific attention given to the maximization of data completeness. Similar data bases are commonly used all over the world in both licensing and operational phases of nuclear power plants, as shown in figure 1. It is obvious that a high quality meteorological data base may be used for purposes in addition to those primarily intended. In this paper an attempt is made to use the measured data to learn about the local conditions of atmospheric stability in different air layers at the Shivta site. A preliminary study of atmospheric stability in the region has been performed within the framework of the PSAR [1] prepared for the IEC by the Desert Research Institute. The frequencies of occurrence of different stability categories for the whole 5 years of data set are shown in figure 2. The classification scheme adopted for this study is the Pasquill-Turner technique, which is based on the routine observations of wind speed, solar radiation and cloud cover in the region. The data have been picked from different meteorological stations scattered within tens of kilometers around the Shivta site. The two curves seen in figure 2 are based on wind data from two stations. II1-33

The results show that the most frequent atmospheric stability is the neutral (D) category over the 5 year period as a whole. The least frequent are the extremely unstable (A) and the extremely stable (G) stability categories. We have used the on-site meteorological data, gathered over the summer period for a similar study. The results are presented in figure 3. The three curves show stability category frequencies calculated by different methods. Two curves are obtained using the NRC recommended method based on a vertical temperature gradient between 10m and 60m, and between 10m and 90m. The third curve is obtained with the modified Pasquill technique, recommended by the European guide, which is based on the wind speed at 10m elevation and global/net solar radiation data for day/night. The difference between the curves, which is best seen for the extreme A and G categories, can be explained by the known fact, that the 60m deep atmospheric layer near the surface is more unstable, than the higher ones. In order to make a closer examination of the extremely unstable (A) conditions, which are typically developed under high solar radiation and low wind speed, the data for only the noon hours (11-16) over the summer period have been processed. The results, seen in figure 4, can be explained as follows: in a shallow air layer below 60m unstable atmospheric conditions prevail. If the layer is somewhat deeper (90m), neutral stability appears, by definition. The remaining curve can be related to a much deeper layer, hence it resembles the curves seen in figure 2, where neutral stability prevails. Advanced computer models for evaluation of pollutant transport and dispersion should account for particular meteorological properties of air layers at different heights. For instance, formulas are generally used to account for the change in wind speed with elevation. In this paper we have described the sensitivity of stability categories to the depth of the air layer. This might be taken into consideration in different evaluations.

Abbreviations:

US NRC - United States Nuclear Regulatory Commission IAEA - International Atomic Energy Agency IAEC - Israel Atomic Energy Commission QA - quality assurence QC - quality control ER - environmental report PSAR - preliminary safety analisys report PRA - probabilistic risk assessment Reference: [1] PSAR draft for Shivta site, Meteorological chapter Prepared by O.Miron, Ben-Gurion University, Desert Research Inst,, Sde Boker. June 1987. 111-34

LICENSING OPERATION

E R PSAR P R A Operational Accidental releases releases

Multi Year Joint Specific Wind & Stab, Wind & Stability distributions(day/night distribution seasonal etc.)

Full year hourly data On-line data

Figure 1. Scheme of different nuclear power plant evaluations using a meteorological data base

Atmospheric Stability in Shivta ROUtlNE OBSERVATIONS FOR 19/6-1930

i r STABILITY CATEGORIES HAR-KEREN KETZK3T

Figure 2. Frequency distribution of atmospheric stability categories in the Negev area, based on the Pasquill-Turner classification Atmospheric Stability in Shivta ON-StTE MASJMCIKNrS tOH SUMMER I»f7

J &

STABILITY CATEOOWES TSO-TIO TSO-TIO

Figure 3. Frequency distribution of atmospheric stability categories in Shivta, obtained by different methods from on-site meteorological data for the summer 1987

Atmospheric Stability in Shivta ON-SITE MEASUREMENTS FOR SIMMER 1987

STABWTV CATEOORCS Teo-no + TMV-TIO WIND* RAD

Figure 4. Frequency distribution of atmospheric stability categories in Shivta, obtained by different methods from on-site meteorological data for noon hours during the summer of 1987 111-35

STOCHASTIC PARAMETER EVALUATION IN MODELLING UNDERGROUND TRANSPORT OF RADIOACTIVITY

Leib Reznik and Louis Tapper Nuclear Engineering Department, Israel Electric Corporation, P.O.B. 10, Haifa SUMMARY

A realistic assessment of potential groundwater contamination, in the case of a major hypothetical accident in a nuclear power plant, requires approplate modelling of the transport processes involved. Some of the difficulties frequently encountered are related to the lack of field data necessary for evaluating accurately certain key parameters of the transport models, such as soil hydraulic conductivity, dispersivity tensor components and retardation factors governing the anion exchange rate of various radioactive constituents during their chemical transport through the geo-medla.

Even with an ample amount of experimental laboratory and field data, a serious problem remains regarding the parameter averaging or "lumping" procedures: this should be done prior to solving the transport equations for such variables as mean flow and isotop:j concentrations. The difficulty is primarily due to to the fact that the relevant soil system is usually a three-dimensional, highly heterogeneous layered system, both in unsaturated and saturated zones, while most of the present numerical simulation codes available are based on simplistic two-dimensional models for spatially homogeneous hydro-geological units. in-36

Attemting to determine detailed spatial variations of hydraulic properties, in a soil having ^ complex spatial structure which

includes irregularities such as water Iense3 and layers, is

obviously impractical. An appiopiate statistical method is therefore sought, which could provide an adequate large-scale characterisation of the key transport parameters mentioned above.

Several stochastic approaches have recently been developed and applied to the problem of low-level-waste site assessment. Some of the most important among them are the "spectral" and the "semiovariagram" approaches, and to a lesser extent the "maximum likelihood" and "maximum entropy" approaches.

In the present work and attempt is made to apply some of the stochastic modelling mentioned above to the problem of realistic •estimation of key parameters necessary for the assessment of potential groundwater contamination following a severe reactor accident.

The different stochastic approaches and their relative merits will be discussed at the meeting in the context of their relevance to the postulated problem. 111-37

SENSITIVITY ANALYSIS OF CRAC-2 RESULTS REGARDING THE CHARACTERISTICS OF A HYPOTHETICAL NUCLEAR PLANT ACCIDENT

M. Rambam, L. Reznik and L. Tepper Nuclear Engineering Department, Israel Electric Corporation, P.O.Box 10, Haifa.

SUMMARY

CRAC-2 was developed for the U.S. NRC for the purpose of assessing population risks in the event of a major reactor accident resulting in significant atmospheric releases of radioactivity. Most applications of CRAC-2 (or some of its versions such as CRACIT) have been performed for nuclear plants located in densely populated areas, such as Zion or Indian Point. In contrast, the relatively large distance of the Shlvta site (*) from population centers in Israel, leads to changes in the relative importance ranking of the various CRAC-2 models depicting processes of radioactive release, atmospheric transport and health effects.

For example, the impact of initial plume buoyancy rise height is totally different when considering the consequences for plant staff and nearby population on the one hand, as opposite to those for inhabitants of cities located at distances of 20 or 30 km away.

Calculation of the initial plume rise height is performed in CRAC-2 according to the Briggs method:

(*} The proposed site for che future first NPP in Israel. 111-38 a) For stable atmospheric conditions

tnltlil _ i/3 »*«bi. = 2.6 [3.7x10-* Q.u.s] + h (la) b) For unstable and neutral conditions

initial 1/3 _ s 2 x hunBt.bi. = 1.6 [3.7xl0- Q.X ] .u~ + h (lb)

2/5 where X = 2.08 (3.7xl0-sQ) .h Q = sensible heat rate of release

u e wind speed s 3 stability index

h B building release height

In addition, the time duration of release T is important in the CRAC-2 computational models. The lateral plume dispersion (in terms of the relative concentration C) is calculated according to the Gaussian model

-1/2 r C(y) = [2n.oy(x)] .exp - (2) where x and y are the geographical coordinates related to the center of release. The dispersion coefficient ay{K), dependent on the atmospheric stability conditions, is determined from the Gifford curves . An important correction in av, taking into account a prolonged release rather than a "puff" type release (3 minutes), is performed in CRAC-2 as follows: III-39

Oy(T) - oy(3 mln).[ T ] (3) L 3 min J

where a • 0.2 for 3 mln < T < 60 min a » 0.25 for 60 min < T < 600 inin

Equations (1-3) show the need for sensitivity analysis of CRAC-2 results regarding parameters such as plume heat rate Q and duration of release T. In addition, due to nonlinearlties in the dosimetric models and methods of health effects assessment, sensitivity checks on other input parameters are required as well. Such parameters include release fractions of radioactive isotopes out of the total core inventory and the effective inversion height. This last parameter depends on the meteorological conditions prevailing at the time of release and subsequent atmospheric transport. For population centers located at remote distances from the site it is important to assess directly the results' sensitivity to any mitigating measures to be taken. These measures are represented in CRAC-2 by effective radiation shielding factors.

In the v/ork described, the sensitivity analysis was performed on a reference PWR reactor of 3000 MWt nominal power. The CRAC-2 output parameters chosen for sensitivity examination included early and latent health effects and equivalent whole body dose levels. Results will be reported at the meeting. 111-40

SEVERE ACCIDENT SCENARIOS FOLLOWING EXTERNALLY INITIATED EVENTS IN A COMMERCIAL NUCLEAR POWER PLANT

D. MAROUANI , S.I. WEIMAN Israel Electric Corporation ltd.,Nuclear Project Department Haifa .Israel.

To define possible core meltdown scenarios in the aftermath of externally initiated events in a commercial Nuclear Power Plant,a variety of accident scenarios were analysed.The final aim of the analysis is to develop specific safety features to be incorporated in the plant design in order to reduce the likelihood of core meltdown following such events.

The reference plant used for the analysis is a Framatome three-loops 900 MWe Pressurized Water Reactor[l].

External events considered were those natural and man made events which could affect limited areas at the plant site. Particular attention has been paid to those individual areas where all redundant trains of a safety system could be compromised. Accident scenarios were analyzed taking into account the existing means for residual heat removal in the reference plant.The analysis was done by using a deterministic method. Significant progress has been achieved with relatively modest efforts in evaluation of a wide range of external events. Various external events were screened and accident scenarios which could lead to core meltdown were selected.Events both initiated at the containment building and outside of it were considered.The analysis was based on plant/system responses to design-basis accidents taking place in the containment (small LOCA,steam generator tube rupture,steam line break etc.) Outside of containment ,beyond-design-basis accidents featuring total loss of specific safety systems, were of particular interest.Total loss of ultimate heat sink,total loss of steam generator feedwater and total loss of electrical power supply [2] were analyzed. 111-41

Concomittant occurrence of selected piping and component failures inside the containment ,and specific safety piping and components failures outside the containment was assumed.All other components and systems ,not selected as mentioned fwere assumed to continue to operate rexcept for those which have lost their vital electrical and water supplies in the aftermath of the external events.The number of safety components which could loose their function concomittantly was limited.

It was concluded that in an externally initiated event .core meltdown occurrence is more susceptile to failures taking place outside of containment rather than to failures inside containment. The following accident scenarios were identified to be the main Contibutors to core meltdown for the analyzed externally initiated events: A-small LOCA concomittant with loss of the Refueling Water Storage Tank (i.e.,loss of safety injection system)

B-loss of the Refueling Water Storage Tank concomittant with loss of the Auxiliary Feedwater Tank (i.e.,total loss of steam generator feedwater)

C-loss of the Essential Service Water pumping station ( i.e., total loss of ultimate heat sink) concomittant with loss of the Auxiliary Feedwater Tank.

REFERENCES

[1] FRAMATOME STANDARD THREE-LOOPS PRELIMINARY ANALYSIS REPORT. [2] P. TANGUY , The French approach to nuclear safety , NS Vol.24,No. 5 sept-oct 83 IIT-42

PWR Loss of Feed-water ATWS Analysis Using the RELAP4/MOD6 Code

D. Hasan, S. Kaizerman, E. Elias and E. Wacholder Department of Nuclear Engineering Technion - Israel Institute of Technology, Haifa

The RELAP4/MOD6 code has been used to simulate the transient thermal hy- draulic behavior of a typical large Westinghouse 4-loop PWR during a loss of feed- water ATWS (Anticipated Transient Without Scram). Previous analyses have shown that this transient produces some of the more limiting overpressure conditions in the primary system [1]. The basic input nodalization is based on a previous model developed for use in small break loss of coolant analysis [2]. However, unlike the small LOCA analysis in which an immediate reactor scram occured, no scram was activated in the present transient. The point-kinetics nuclear feedback model in RELAP was, therefore, im- plemented. In some of the cases analyzed, the reactor power actually increased during the later stages of the incident. A number of additional systems, important in overpressurization transients, were represented and added to the basic input model, e.g., the pressurieer spray line and safety/relief valves capable of intermittent operation. For the pressurizer, at least two control volumes were necessary to represent the pressure response to the fluid expansion [3]. The transient is initiated by a complete loss of feedwater to all 4 steam generators during the first second of the transient. As a result, the primary system pressure increases and the pressurizer sp-ay is activated. When the pressure increases further the PORV (Power Operated Relief Valve) and the SRV (Safety Relief Valve) open according to their set points. As the transient proceeds, other systems are activated when their set points are reached. In Fig. 1 are depicted the short term results of the basic ATWS case. The loss of main feedwater causes a large imbalance in the heat transfer rate, as the secondary side can no longer remove all the heat gcnerateu in the core. Therefore, a rapid pressurization of the primary system is observed in the first 50 sec. Activation of the pressurizer spray system (a*, about 5 sec) and the PORV (at about 20 sec) ni-43 does not reverse this trend although reducing its rate. At about 50 sec the reactcr power drops sharply, due to the negative void coefficient in the core, causing a rapid depressurization of the primary system to a level of about 2750 psia. The reactor coolant pumps keep running throughout the transient. Some variations on the basic case were analyzed. It was found that stuck open PORVs and SRVs have only a minor effect on the initial stage of the transient. This is not surprising since in the base case analyzed, those valves remain anyhow open during most of the time. In another case, in which the reactor coolant pumps were tripped (at 2370 psia), the critical pressure peak was "flattened11', i.e., lowered and delayed to about 120 sec.

Acknowledgements This work was partly supported by the Israei Electric Corp.

References 1. R. Salvatori et al., Anticipated Transient Without Trip Analysis for a Four-Loop (3817 MWt) Westinghouse PWR, WCAP-8440, 1975. 2. S. Kaizerman et al., Parametric Analysis of a Loss of Feedwater Incident Followed by a Small Break LOCA using the Code RELAP4/MOD6, Trans. INS Vol. 13, Feb. 1986.

3. M. Andreani et al., The Analysis of LOFT Test L9-3 using RELAP4/MOD6 and ALMOD-JRC Computer Codes, KFK-3880/3, Dec. 1984. o o Figure i Variation of Pressurizer Pressure'

It

~izo 3FO 400 111-45

Advancing DSNP to Simulate ATWS in PWR.

D. Saphier, D. Gal

Soreq Nuclear Research Center. Yavne 70600, ISRAEL

A major problem with large system codes such as RELAP, RETRAN or TRAC in simulating ATWS (Anticipated Transients Without Scram) is the long computer times necessary for the simulation of these events. In addition, a large amount of time is necessary to achieve steady state conditions. The user is required to modify many of the unknown state variables and input parameters in a iterative manner until an approximate steady state is achieved. In addition the process of setting up an ATWS simulation requiring detailed nodaMzation is very tedious.

Recently the DSNP<1> simulation language wa" \\ < raded to permit the simulation of a variety of AT'JS events. Tin upgrading was achieved by taking the following actions:

1) Development of advanced PWR modules.

?) Inclusion of advanced integration techniques.

3) Verifying the DSNP PWR options.

Modelling a power plant with DSNP can be achieved either by combining "control volumes" and "junctions" or "pipe" elements, as done with the RE1.AP code or by using integral components. The second mode is much faster and efficient, both in computer CPU time and in the time required to develop a full scale power plant simulation. Three major PVR component models were recently added to DSNP: An advanced pressurizer model**', an advanced moving boundaries U-tube steam generator Model<*>, and a tnultinode PUR core model that can be combined with any number of detailed furl pins. m-46

The DSNP system was verified in the past for LMFBR<*> and HTGR<»> simulations by comparing calculated transients to experimental results. Recently the simulation of the NRC standard problem #1, the "Edward's Pipe" was performed, and the result compare well with experimental measurements as shown in fig. 1. This test verifies the DSNP Hydraulic Network Solver. A detailed DSNP simulation of the Trojan PUR vas also developed using the DSNP. The ATVS simulation performed with this model compare well vith similar calculations performed by Collier<7> at Battelle Columbus as can be seen from figs. 2 and 3.

o

< 4

u as to two os a.

10 20 30 TIME-(SEC -10"*-2,) Fig. 1: Experimental and calculated pressure at the first gauge station of the Edwards pipe.

Fig. 2, shows the pressurizer pressure during a loss of feedwater with stuck open safety valve ATUS. Fig. 3, shows the pressure inside the secondary side of the U-tube steam generator during the same transients. In borh cases the accident consequences were mitigated by the various systeir. safety valves.

Hit ahove ATUS simulations verity the advanced models developed for DSNP. Based on the extensive model development and verification, it can be concluded that DSNP can now be used to simulate a large variety of ATVS event.'- and other transients in PVRs. 111-47

Pump trip

U-tubes are uncovered RELAP3B

10 TIME(SEC'K)1)

Fig. 2:Pressure inside the piessurizer during a loss of feedvater with stuck open safety valve ATVS type event.

U-tubes are uncovered

DSNP

RELAP3B

10 20 30 46 60 1 TIME(SEC-10 )

3:Pressure inside the U-tube steam generator secondary side during a loss of feedvater with stuck open safety valve ATWS type event. 111-48

REFERENCES

1. D. Saphier, "The DSNP User Manual", RASG-112-85, Georgia Institute of Tecnolgy, Atlanta (1985)

2. D. Saphier, J. Kalfelz, L. Belblidia, "Response of a DSNP Model Under Accident Conditions", Trans. Am. Nucl. Soc, 52, 475 (1986).

3. D. Gal, D. Saphier, E. Elias, "A UTSG Model for DSNP", ANS Topical Meeting on Anticipated and Abnormal Transients in Nuclear Power Plants, Atlanta (1986).

4. W. K. Lento, E. M. Dean, H. A. Larson, J. F. Koenig, "Experimental Breeding Reactor II, Dynamic modeling and Code Verification", Trans. Am. Nucl. Soc. 44. 310 (1983).

5. W. Croenbroeck, "Nachrichtung von AVR-Betriebstransienten Unter Vervendung des DSNP-Programsystem", KFA Julich, KFA-ISF-LB-5/86 (1986)

6. D. Gal, D. Saphier, "A Loss of Offsite Power Event Simulation", This meeting.(1987).

7. R. P. Collier, et al, "Final Report on Selected ATWS Audit Calculations for three PWR Designs", Battelle Columbus report to BNL (1982). 111-49

A LOSS OF OFFSITE POWER EVENT SIMULATION

D, Gal and D. Saphier

Soreq Nuclear Research Center Yavne 70600, Israel

A loss of offsite power (LOOP) event in a PWR was simulated and analyzed using the DSNP simulation language. The plant analyzed was a four loop

3^11 MWt Westinghouse PWR^1'. The immediate result of this event is the loss of reactor coolant system (RCS) pumps and the main feedwater supply to the steam generators.

The DSNP model includes 2M fixed and moving boundary control volumes,

21 junctions and 10 valves as shown in fig. 1. The flowchart in fig. 1 shows the DSNP modules used in the simulation representing: core

(neutronics and thermal hydraulics), U-tube steam generator, pressurizer, primary pump, primary pipes and cavities, steam line and the principal valves.

The simulation began with 1 sec of steady state operation followed by the loss of all the pumps. According to ATWS (Anticipated Transient Without

Scram) philosophy, the reactor trip (Scram) signal did not activate the control rods. However, an isolation of the main steam line was assumed to take place, resulting in the secondary and primary pressure increase.

DSNP - Dynamic Simulator for Nuclear Power Plants. 111-50

At 30.3 sec the pressurizer became liquid filled and a second pressure

increase occurred as shown in fig. 2. The auxiliary feedwater system

started delivering water to the steam generator at 61.0 sec. Between 44.!?

sec 235.0 sec the pressurizer pressure decreased due to flow through

relief and safety valves and due to decreasing neutronic power caused by

the negative reactivity feedback. At t - 235.0 sec steam generator dryout

began (U tubes were exposed). The loss of heat sink induced another

pressure increase. At 500 sec the power was reduced to 7.2%, the outlet

core coolant temperature was 349 C and the pressurizer pressure was 16.2

MPa. No ECCS* actuation was assumed during the first 500 sec.

Figure 2 presents the pressurizer pressure during the first 500 sec of the

transient. As can be seen, the results as calculated by DSNP are in good agreement with the RELAP3B calculation of the transient performed by

Collier et al.^2^ The major difference between the RELAP and DSNP simulations is in the steam generator nodalization approach. The steam generator model in the RELAP code includes 32 control volumes and 45 junctions while the DSNP uses an integral steam generator model^) using

7 control volumes with moving boundaries. The DSNP savings in computational effort both in setup and CPU time are significant, in particular, since the DSNP searches automatically for the system initial steady state while with RELAP the input data have to be modified by the user in a long iterative process until an approximate steady state is achieved.

As can be seen from the results the consequences of the LOOP accident are mitigated by the system as a result of negative feedback reactivity and the proper operation of the relief and safety valves.

ECCS - Emergency core cooling system. 111-51

References

1) FSAR, Trojan Nuclear Plant, Portland General Electric Co., Oregon

(1975).

2) R.P. Collier et al., Final Report on Selected ATWS Audit Calculations for Three Power Designs, Battelle Colorabus Lab., Ohio (1982).

3) D. Gal, D. Saphier, E. Elias, "A UTSG MODEL FOR DSNP", Proceedings of the ANS Topical Meeting on Anticipated and Abnormal Transients in Nuclear Power Plants, Atlanta GA, April 12-15 (1987). 111-52

Fi%. It DSMP modules diagram of a 4 loop PWK plant.

Pressurizer fills up.

< w an

as a. S

158 UTSG tubes are uncovered. W PS a. 154- 10 20 30 50 T1ME(SEC-1O1)

Fig. 2; Pressurizer pressure during LOOP event. High Temperature Gas Cooled Reactors

Chairmen - E. Greenspan and I. Kis IV-1

ISRAEL'S INTEREST IN HTGRs

E. Greenspan

Israel Atomic Energy Commission P.O.Box 7061 Tel-Aviv, 61070

Up until recently, the only nuclear power reactor technologies that have been seriousely considered for adoption in Israel vere LW. and HWR technologies. So far Israel made no committment to any of these technologies. It now appears that the first opportunity for the introduction of nuclear power plants into the Israeli electricity grid will come at the end of the century. It is likely that by that time new and improved nuclear reactor technologies will become commercial. In view of this set of conditions, compounded by the crisis the world nuclear industry is presently undergoing, the Israeli Atomic Energy Commission (IAEC) set forth to assess whether it should base its nuclear power program from the beginning on one of the emerging reactor technologies.

A preliminary evaluation carried-out indicated that of the emerging technologies, the HTGRs appear to be one of the most promising for the needs and conditions of Israel. Consequently, a feasibility study aimed at clarifying the interest Israel should have in HTGRs was recently undertaken. The present paper describes the rational for the adoption of the HTGR technology for this feasibility study. SAFETY

HTGRs offer a very high degree of safety due to a unique combination of the following features: * A very good retention of the fission products in the ceramic coated fuel particles up to a fuel temperature exceeding 1600°C. The nominal average fuel temperature is * 700°C. * Very high core heat capacity due, primarily, to the heat capacity of the graphite (and its close contact with the fuel). * Large negative temperature coefficient of reactivity. Realizing that a AT>900°C is available for core heatup IV-2

before the fission product release rate becomes of concern, it is estimated that the total negative reactivity worth associated with the fuel heatup is close to 2% and with the core (i.e., fuel + graphite) heatup is about 10%!

* Very slow fission products release from the fuel even if a major core heat-up (to >1600°C) accident would occur. The radiological hazards of a source term released (in a more or less constant rate) over a period of a number of days is orders of magnitude smaller*x> than that from a source term of a comparable total magnitude released in a short duration (a "puff" release). Nevertheless, the consequences of low probability beyond design-basis air and water ingress accidents need still be thoroughly examined<2-3>. * Power excursion leading to unacceptable fuel temperatures appears to be practically impossible due to the large negative temperature reactivity worth. Nevertheless, the consequences of low probability beyond design-basis water ingress accidents in cold cores need still be thoroughly investigated*2'.

* If designed to have large enough surface area-to-volume ratio cores (such as in "modular" designs), the HTGRs can have their decay heat removal by passive means only.

* It appears that the design proposed for certain HTGRs can withstand external events (including acts of war) without a need for an extra investment***. * Moreover, these designs are likely to offer a relatively high investment protection. This is due to two factors in addition to the high safety level of HTGRs: (a) The compactness of the nuclear island and its enclosure within a relatively thick concrete structure (which, in certain cases, is located underground), and (b) The relatively small power level per unit. UNIT SIZE AND ECONOMICS

Of the two HTGR sizes being developed commercially - «100MWe (representing a typical modular HTGR) and 550MWe (the HTR-500), the latter turns out to have the same power level as the coal fired power plants under construction in Israel. Moreover, as the steam conditions from the HTGRs is similar to these from coal (or oil) fired power plants, the BOP for the HTR-500 can be very similar to that for the coal fired power plants with the design, construction and operation of which the Israel Electric Coorporation (IEC) has much experience. IV-3

On the other hand, modular HTGRs can enable expanding the Israeli grid with minimum over-capacity. This is due to the fact that the average increase in the peak demand of electricity by the end of this century is expected to be about 150MWe/year. Thus, by constructing one to two modules per year one could follow very closely the growth in demand. Such a rate of construction might give an economic justification for the local industry to get into the manufacturing of a large fraction of the reactor components (A similar situation also exists with regards to a number of major components of the HTR-500, such as the steam generators and circulators; they are designed to take about lOOMWe per unit, just as :ln a modular HTGR).

Economic evaluation recently carried-out in the USA concluded*3' that the American modular HTGR is competitive with coal fired power plants. Independent evaluation carried-out in the Federal Republic of Germany concluded<=> that HTR-500 is more economical, under the FRG conditions, than a 500MWe coal-fired power plant and than a 125OMWe PWR. Whether or not HTR could be economically competitive with coal fired power plants in Israel is a matter for further investigation.

SITING It is presently perceived t^at, due to the relatively high population density in the coastal region, nuclear power plants will have to be located inland. However, the combination of high safety, relatively small unit size and resistance to acts of war may enable siting modular HTGRs near the coast. Coastal location offers reduced cooling and power transmission costs.

On the other hand, due to their relatively high temperature for energy delivery, HTGRs appear more suitable for inland siting than LWRs. It is estimated56', for example, that the economic penalty to HTGRs associated with the use of dry cooling is only about half that expected for LWRs. INDUSTRIAL APPLICATIONS The high temperature operation-ability of HTGRs (helium outlet temperature of 950°C has been demonstrated in the AVR) makes these reactors uniquely suitable nuclear heat sources for a variety of industrial applications. One of these applications of particular interest to Israel is the utilization of oil-shales - the only significant indigeuous fossil fuel resource found so far in Israel. A preliminary assessment indicates49* that by using HTGR to provide the energy (and extra hydrogen) for converting the organic materials and carbon of the shales into liquid and gaseous fuel, it might be possible to increase by more than 50% the energy value of the chemical fuels obtained from a given quantity of shales (relative to the conventional approach, which involves burning a part of the IV-4

shales for providing the energy for the process) while reducing the environmental pollution and improving the economics.

HTGRs may be also an attractive heat source for water desalination. A preliminary assessment indicates<6> that the HTGR makes a significantly more attractive match to Multi-Effect-Distillation (MED) process than LWR; it offers a simpler, cheaper and more efficient coupling to the MED plant while losing less on its electricity production efficiency.

SUMMARY

The HTGR technology offers a collection of features which might justify its adoption for the nuclear power program of Israel. Final conclusions on the interest Israel should have in this technology should, however, await the completion of a thorough feasibility study.

REFERENCES

1. S. Ron, Trans. Israeli Nucl. Soc. H. (1983).

2. J. Szabo et al, "Nuclear Safety Implications of Water Ingress Accidents in HTGRs", These Transactions.

3. S. Ron, "On the Graphite Oxidation Rate in HTGRs Following an Air Ingress Accidents", These Transactions.

4. A. Ketter, and J. Szabo, "Analysis of HTGR Vulnerability to External Missiles", These Transactions.

5. A. Barak, "A Review of The Economics of HTGRs", These Transactions.

6. A. Barak et al., "Israeli Perspective on HTGRs", Proceedings IAEA Technical Committee Meeting on Gas Cooled Reactors, KFK Julich, Oct. 1986.

7. E. Ziermann, Nucl. Eng. and Design. 78 (2), 99 (""84).

8. R. Schulten, K. Kugeler and W. Frohling, Progress in Nuclear Energy. 14, 227 (1984).

9. I. Spiewak, "Potential Industrial Applications of High Temperature Reactors in Israel". These Transactions. IV-5

A REVIEW OF THE ECONOMICS OF HTGRs

An.itzur Z. Barak

Israel Atomic Energy Commission P.O.Box 6071, Tel-Aviv, Israel

This review presents comparisons of the anticipated economics of HTGR with those of coal power stations and PWRs. Two figures of merit are examined: (1) the average (estimated) cost of the net. KWhe at the station exit (to transformer), and (2) the system capitalization of all expendictures.

COST OF KWh APPROACH tHE Cost of KWhe is described as the sum of the following terms:

1. Capital cost (per KWhe), Cc;

c ~ N "8760 CF ' / v ' ' Where I is the Total Capital Investment Cost of the unit, ($) N is the net capacity of the unit (KWe) ^ is the capital recovery factor which is a function of d - the (annual real) discount rate - and of T - the life time of the unit. CF- is the capacity factor.

2. Fuel specific cost C«: cost of unit fuel ($) f ~ Heating value (KWh^) per unit fuel xn Where W is the net overall efficiency.

3. Operation and maintenance, Com: r £ om = N x 8760 CF Where £ is the sum of all current expenses (except fuel) in average, per year, convering expenses such as salaries, insurance, materials and chemicals, transportation, communication etc. IV-6

In addition to these three major components, nuclef. units have two more minor components: 4. Decommissioning cost, Co:

CD = N X 8760 CF - 1] Where D is tha total cost of decommissioning. 5. Indirect costs - additional expenses that are caused to the system by the nuclear unit and other expenses caused to the unit by the rest of the system due to its special characteristics. It is quite difficult to evaluate this cost component. CAPITALIZATION APPROACH

This criterion (which is favored by Israel Electric Co.) is obtained by capitalizing all costs of investments, fuel, operation, maintenance, etc. during the period of interest, say between 1987 and 2018. By comparing various scenarios (e.g. one with expansion by coal units only to one with 50% expansion by coal and 50% expansion by nuclear units) the relative economics can be estimated. Each of the two criteria is based on a definite method and each method has is limitations. A combination of br>*' expected to deliver a reasonable clear economic picture. COMPARISONS U.S. AND GERMANY

Many economical comparisons have been published in the FRG<«-«- !.*> and in the U.S.A.<-•«• 3.*.=># The published information has been worked out into the comparisons shown in Table 1 and 2,

Table 1: Economics of different electricity generating technologies in the FRG (d=8%/year; T=30 years; 1.92 DM/$).

HTGR PWR COAL UNITS Net capacity N 500 1230 625 MWe Total Investment I 2560 5750 1280 MDM Capacity factor CF 0.8 0.8 (C•7) 0.8 (0,63) 0per.+ Maint. £ 63.7 138.6 78.7 MDM/yr Fuel Spec. Cost 2.32 4.43 11.5 Pf/KWh c£ H 0 + M loin 1.89 1.61 (1.83) 1.80 (2.28) Capital Cost Co 6.48 5.93 (6.78) 2.60 (3.30) If Decommissining Cn> 0.13 0.06 (0.07) II Cost of KWh 10.82 12.03 (13.17) 15.90 (17.08) II ii ii n 5.64 6.27 (6.83) 8.27 (8.90) 4/KWh IV-7

Table 2: Economics of different electricity generating technologies in the USA (d=8%/year; T=30 years).

MHTGR PUR COAL UNITS

Net Capacity N 4x135 1140 406 MWe Efficiency n .385 .334 .34 Total Investmeni11 1050 2050 670 10* $ Capacity Factor CF .8 .6 .8 (0.63) Oper. + Maint. Z 31.5 47 20.0 106$/year Fuel Spec. Cost c* 1.1 .70 3.0 4/KWh 0 + M ( .84 .78 ( .67) .70 (0.89) it Capital Cost Co 2.47 3.04 (2 .60) 2.09 (2.65) ii H Decommissioning CD 0.02 0.02 -• -- Cost of KWh 4.43 4.54 (3 .99) 5.79 (6.54) II

Note: The costs of Ktfh in the above Tables for the nuclear alternatives may be higher by about 0.1 - 0.2 $/KWh due to indirect expenses vhich covers most of the differences between the two methods of evaluation.

CONCLUSIONS:

1. According to the presented comparisons the cots of KWh from HTGR and PWR are roughly the same. However considerable differences of up to about ±15£ may be found due to variations in the various parameter.

2. According to the cited comparisons the costs of KWh from coal were considerably higher, but it should be kept in mind that they are very sensitive to possible fluctuation in coal prices and depend on the possible requirement for scrubbers.

3. Adaptation of the data for the local conditions in Israel is due in the very near future, as the properties and size of HTGR units suit the local grid better than PWR. Coal prices and construction costs in Israel seem to be considerably lower so that the comparison may lead to different conclusions.

REFERENCES:

1. G. Wittchow, K. Bode and V. Schrumpf "HTR-500 - an Alternative to the PWR", Energiewirtschaftliche Tagesfragen No. 5 pp. 344-349, May 1985.

2. J. Schoening, K. Knizia and D. Schwarz "Die Zukunft der HTR-Bauline" VGB Kraftwerkstechnik 65, pp. 11-17, January 1985. IV-8

3. J.C. Scarborough and L.D. Mears "Economic Comparison of Modular HTGRs in the U.S. with LWR" GCRA Report, 1985. 4. J. Euss "An Economic Evaluation of the HTGR", GCRA Report, 6th Annual International Conference on HTGR, August (14) 1984. 5. "Modular HTGR Demonstration Project Definition Study" GCRA Report 86-005, p. 11-25, October 1986. IV-9

APPLICATION OF HIGH TEMPERATURE REACTORS TO ISRAELI SHALE PROCESSING

Irving Spiewak

Weizmann Institute of Science, Rehovot, Israel

Currently, 40% of the fuel used in Israel goes into the generation of electricity and 60% into direct fuel consumption. Whereas nuclear power has thus far been used primarily for electric power production, the graphite-core high temperature reactor (HTR) is a promising source of high-temperature process heat. The steam cycle HTR used for electric power generation has a helium outlet temperature of 700°C and can superheat steam to 550°C. Helium temperatures of 950°C have been reached routinely at the AVR at Julich. West Germany for many years; consequently it is anticipated that HTRs can be used to heat processes up to the range

A survey of process applications in Israel indicated that the most promising was for the production of synthetic fuels from oil shale. There are large deposits of oil shale in the Negev that are expected to become commercially useful in the twenty-first century, after oil prices reach a new plateau above $35 per barrel. Processes have been developed for producing liquid fuels from the shale. Gases and solid wastes from the retort are used as the sources of heat in conventional processes; surplus gas and solid energy is used to generate by-product electric power<2>. In the nuclear shale processing plant suggested (Figure 1), the HTR supplies process steam for heating the retort, high temperature energy for reforming hydrocarbon gases, and miscellaneous steam and power needs. Surplus HTR energy, when available, is used for by-product power production. The chemical energy of the retort gases and solids is converted to the form of synthesis gas (carbon monoxide plus hydrogen) or as in the case of Figure 1, to hydrogen. The gas product may be distributed by pipeline and/or converted to methanol, ammonia or methane. IV-10

In the process example of Figure 1, 84,755 tons per day of shale are crushed and sized, about 24,000 tons of fines going to a gasifier and the larger fraction of shale to the retort plant. The retort is heated by steam from the HTR and oxygen produced in an on-site oxygen plant not shown. The retort gases are reformed in the KTR plant and converted to hydrogen. The raw oil from the retort is hydrotreated on-site. The products from the plant consist of 18,000 barrels per day of oil suitable for feed to a refinery and 1.26X1011 Btu per day of hydrogen (equivalent to 21,000 barrels per day). The HTR plant, in this case, is assumed to consist of four modular reactors, two rated at 350 MUt of steam and two rated at 297.5 MWt of 850°C process heat. The nuclear plant is estimated to cost $ 1.02 billion 1986 dollars while the process plant is estimated to cost $ 1.2 billion. Table 1 summarizes the costs and product flows of the nuclear plant compared to reference conventional plants.

It is concluded that the use of HTGRs might increase the amount of fuel recoverable from a given amount of Israeli oil shale by over 50£ at a lower unit cost of product. However, a market must be created for large blocks of gas, if the proposed concept is to be commercial.

REFERENCES (1) IAEA, "Status of and Prospects for Gas-Cooled Reactors", Technical Reports Series No. 235, 1984. (2) R. Schulten, K. Kugeler and W. Frohling, Progress in Nuclear Energy, U., 227 (1984). IV-11

Hj,»CO h 950" PRODUCT STEAM 1) I I 9 1 •> X NH3,H( 1— 8 NH i HjS sCo! 1 4 V \ 5CO2 STEAM } H2 & Z2 \ 8 \r HELIUM 9 3 STEAM > PRODUCT ' 250° ) +0 \ ii \ 2 / 2 t OFF-GAS STEAM FINE GAS \ /

OIL PRODUCT OIL SHALL 2 10 1 OiL'

STEAM +0 i- ^-— STEAM —»j 13 f2

.ASH

Figure 1. Flowsheet of an HTR Assisted Shale Processing System

(1) Shale feed preparation, (2) Retort, (3) Gasifier, (4) High temperature reactor, (5) Hydrocarbon reformer, (6) Steam generator, (7) Helium circulator, (8) Gas purification system, (9) Hydrogen shift reactor, (10) Oil hydrotreater, (11) Gas compressor, (12) Heat recovery, (13) Steam turbine-generator. IV-12

Table 1: Summary of Parameters and Costs of Nuclear and Reference Conventional Shale Plants

Paraho Paraho+ Gasifier Nuclear plant Gasifier Plant Shale plant Plant

Shale mined, tpd 84,755 84,755 84,755 84,755

Product oil, bpd 18,000 18,000 18,000 Gas, equivalent, bpd 7,000 25,000 21,000 Total, bpd 18,000 25,000 25,000 39,000

Efficiency, % Gas+oil/raw shale energy 36 50 50 77.4

Gas+oil/shale+nuclear energy 36 50 50 63

Product cost, $/106 Btu Fixed charge rate=9£ 7. 17 7.06 8.49 6.54

Fixed charge rate=12X 8.07 7.95 9.36 7.47 IV-13

NUCLEAR SAFETY IMPLICATIONS OF WATER INGRESS ACCIDENTS IN HTGRs

J. Szabo, E. Greenspan, A. Ketter and S. Ron

Israel Atomic Energy Commission P.O.Box 7061, Tel-Aviv 61070

Economic considerations force High Temperature Gas Cooled Reactor (HTGR) designs to have undermoderated cores. Consequently, the addition of water to the HTGR core can have a positive reactivity effect. Potentially, water ingress accidents can be one of the most severe safety hazards of HTGRs. The purpose of this work is to review the published information on water ingress and to attempt an upper bound assessment of the assertion commonly found in the literature that water ingress does not pose a realistic threat to the HTGR safety. This assessment uses a deterministic approach and assumes that no active systems compensate for the reactivity effect of water ingress. REACTIVITY EFFECT OF WATER INGRESS The general reactivity effect of water ingress is shown in Fig. I*1* for three different fuel types having a heavy metal content of 6.6 gram heavy metal per fuel element. The 0.1 water volume fraction in the abscissa of Fig. 1 corresponds to 2600 kg of water homogeneously distributed in the core (the overall coolant space, 26000 liter, is 39% of the core volume). The positive reactivity effect of water ingress is due to the reduction in the neutron leakage and to the increase in the resonance escape probability caused by the addition of water. The high sensitivity of the temperature coefficient of reactivity (TCR) to the type of fuel (see Fig. 1) reflects the fact that the resonance absorption is higher the higher is the 238U concentration. Beyond a certain water concentration (to be referred to as the critical concentration) the increase in thermal neutron absorption due to water addition outweighs the reduction in the resonance escape probability, thus causing a decline in the TCR.

Other studies arrived at a similar dependence of the core reactivity on the insertion of water. This includes the study by Druke<2> pertaining to LEU/Th fuel; a study conducted by the Russians<3> with HEU (21%); and a study performed at IV-14

1 1 1 i 1 1 "^V. L€U LEU oos \

- V LEU/Th - \ 0.08 V \- "^V- HEU/Th V HEU/Th 0.01 \ \ 1 \ '/ \ 1 \ 1 i 1 1 1 f 01(00 O0» 0050 0.0T9 0.100 oo» oflse oon WATER FRACTION IN COOLANT SPACE WATER FRACTION IN COOLANT SPACE

Fig. 1 The reactivity addition Fig. 2 Fuel kernel temperature from the insertion of required to compensate water for reactivity increase from water ingress INTERATOM for LEU. The latter study also demonstrated the strong dependence of the TCR on the HM loading in the fuel; the higher the loading (i.e., the more undermoderated the core), the more pronounced is the reactivity effect of water ingress. The critical water concentration depends on the core size; the smaller the core the higher is the neutron leakage probability and the higher is the critical water concentration. FUEL TEMPERATURE REACTIVITY COMPENSATION ABILITY Fig. 2 shows the asymptotic fuel temperatures necessary to compensate for the positive reactivity effect of the water ingress events referred to in Fig. 1. Fission products retention requirements limit the fuel upper operating temperature to 1600°C. Thus it is observed that whereas a HEU fuelled reactor can accomodate, asymptotically, any quantity of water, the LEU fuelled reactor can only accomodate the equivalent of less than 0.02 cf water fraction in the core. If larger quantities of water are to penetrate the core, the design may be forced to rely upon active means to compensate for the water reactivity effect. IV-15

MAXIMUM AMOUNT OF WATER THAT CAN GET INTO THE CORE Consider a modular HTGR of the pebble bed side-by-side type. Suppose the water ingress occurs as a result of a breach in a hot section of the steam generator (SG) while the reactor is at nominal operating temperatures. Then steam enters into the helium coolant system and is being carried into the core. As the pressure of the secondary coolant is significantly higher than that of the primary, the breach in the SG will cause the primary circuit pressure to increase and, possibly, to the openning of the pressure relief valve resulting in loss of part of the helium. An upper limit to the quantity of steam which might enter the core under these conditions is obtained by assuming that steam, at the primary system relief-valve pressure, completely replaces helium in the core. If the core graphite is near its nominal operating temperature distribution, the average temperature of the steam in the core will be in the vicinity of 450°C. At this temperature the steam density is 18 kg/m3 — 10% lower than the density requiring a kernel temperature of 1600°K to compensate for its reactivity effect.

However, a larger amount of water can hypothetically get into the core in case the water ingress accident occurs when the reactor operates at a reduced power level and lower operating temperatures or in a cold shutdown (but near critical) conditions. Even at or near the nominal operating temperatures the upper limit to the water inventory in the core might exceed the value estimated above due to (1) the contribution of the pores in the graphite to the effective free volume - increasing it from 39% to 47%, and (2) partial condensation of steam in the relatively cold "corners" of the core and on the graphite surface. IMPLICATION TO HTGR SAFETY

The reactivity effect of the maximum amount of water that can be contained in an HTGR pebble bed core the graphite temperature of which is nominal, can be compensated by a fuel temperature increase to less than 1600°C. Moreover, the assymptotic temperature the HTGR core will get to theoretically, following any water ingress accident will be below 1600°C. However, if the water ingress accident is to start with relatively large amount of liquid water getting into a relatively cold core, the fuel temperature can, theoretically, increase rapidly to significantly overshoot the 1600°C limit before the water content is reduced (by evaporation and expansion) to the tolerable level. Hence, the deterministic IV-16

upper limit approach used in this preliminary analysis can not prove (neither can it disprove) that KTGRs are passively safe against every vater ingress accident. It is concluded that time dependent analysis based on sophistlcatd simulation codes need be undertaken in order to arrive at firm conclusions on the safety of HTGRs against vater ingress accidents. It ought be mentioned, though, that all the experimental evidence and numerical analysis published so far indicate that HTGRs are safe against vater ingress accidents.

REFERENCES

[1] D.R. Vondy, and R.D. Timmerman, "Estimated Reactivity Effects from Vater Ingress and Temperature Changes for a Modular Pebble Bed HTGR Core", ORNL/TM-9716; DOE/HTGR-85-170, Nov. 1985. [2] Y. Druke, et al. Nucl. Sci. & Ene.. 57, 328 (197). (3] V.N. Grebennik et al., "Analysis of Some Accident Conditions in Confirmation of the HTGR Safety" IAEA Specialists Meeting on Gas-Cooled Reactor Safety and Licensing Aspects - Lausanne, pp. 214-221, Sep. 1980. [4] G. Lohnert, "The Corrosional and Nuclear Effects of Water Ingress into the Primary Circuit of an HTR-Module", IAEA Specialists Meeting on Safety and Accident Analysis for Gas-Cooled Reactors, IAEA-TECDOC-358; Oak Ridge, May 1985.

[5] H. Muler et al., "Investigations on the Vater Ingress in a Pebble Bed High Temperature Gas-Cooled Reactor", ibid, p. 227. IV-17

Loss of Coolant Accident in a Modular HTGR.

D. Saphier, D. Gal

Soreq Nuclear Research Center. Yavne 70600, ISRAEL

The available pebble bed HTGR module C0RPB2 of the DSNP<2> simulation language was expanded to permit the analysis of complete loss of coolant accident in a modular HTGRCPi;>. In addition a cylindrical shells module was developed to surround the reactor core to permit the heat transport by conduction and radiation to the environment through the side reflector, the reactor vessel and the concrete structure.

The major concern in this study was to investigate whether during the postulated accident of loss of coolant, and loss of all active heat removal devices, the maximum fuel temperature will stay below the temperature leading to the release of fission products. In other terms, the question to be answered is, whether the concept of inherently safe reactor can be applied to the proposed system.

In the past the DSNP system was used to analyze accidents in several HTGR plants and appropriate modules of HTGR components were developed. The analysis of the PNP-500 pebble bed reactor is summarized in reference 4, while the analysis of the AVR reactor is summarized in reference 5. In the first study**' the HTGR basic modules were developed and a preliminary comparison to other calculations was performed. In the second studyc3> the DSNP HTGR modules were verified by simulating experimental transients measured in the AVR reactor. In both cases good agreement was obtained, thus giving a high level of confidence in using DSNP to a selected range of HTGR transients.

In the present study a total loss of flow is assumed, and therefore only the core thermohydraulics, neutronics, feedback, decay heat and radial structures were involved in the simulation as shown in the flowchart IV-18

) segments the cote can have also M radial cylindrical segments to permit heat transport into the radial direction.

HIM IMH PLANT CONTROL CORE SAFETY SYSTEM FEEOBACK SYSTEM

t f r CORE NEUTRONICS

CORE CORE RADIAL POWER TKERMOHYORAULICS REFLECTOR DISTRIBUTION

DECAY HEAT PREDETERMINED FLOW FUNCTION

Fig. 1: Schematic flowchart of a DSNP program to simulate loss of flow transient in modular HTGR.

The core is represented by a matrix of N*M radial rings. In each ring the heat transfer equation for an average fuel sphere is solved. The heat transfer for the fuel sphere is given by

ST 1 & (kr2—) + q(r,t) v/heve d, C,,, and k are the graphite sphere density, specific heat, and conductivity, T is the temperature and q is the power distribution inside the core. The power distribution resulting from the neutronic power and decay heat are constant in space but vary with time according to the solution of the kinetic and decay heat equations.

While flow prevails, heat is transported from the spheres to the coolant using appropriate pebble bed heat transfer correlations. During loss of flov.1 a special semi-experimental correlation accounting for heat transported by radiation conduction and local convection is used.

In the radial shells the heat conduction equation in cylindrical geometry is used. IV-19

U 6T (kr—) + q(r,t) r6r hi

Shells of different materials with different conductivities as shown in fig 2 can be used.

rv-n

RADIAL REFLECTOR CORE AIR

2: Cross section through the cylinder of structures surrunding the pebble bed core.

In the DSNP simulation with the new HTGR modules, the core parameters as given in reference 3 were used. The inlet flow was assumed to coast down from full flow to zero in 10 minutes. As can be seen in fig. 3, due to the strong negative feedback the neutronic power is reduced to zero.

The maximum fuel temperature increases slowly during the first forty hours but does not reach the 1600C, at which level fission products are being released. No credit was given in this simulation for natural circulation loops that will be probably established, reducing further the maximum temperature. In this study only a single event was investigated and the results do not preclude other scenarios achieving higher temperatures. IV-20

i I < TIME-UEC) «19>

Fig. 3: Response of a pebble bed core power to a loss of flow accident.

REFERENCES

1. D. Saphier, J. Rodnizky, Dynamic modeling and Simulation of a High Temperature Gas Cooled Pebble-Bed Reactor" ANS/CNS Topical Meeting On Numerical Methods in Nuclear Engineering, Montreal (1983).

2. D. Saphier, "The DSNP User Manual", RASG-112-85, Georgia Institute of Technolgy, Atlanta (1985)

3. INTERATOM "An Assesment of the INTERATOM/KWU Modular HTGR Concept", Interatom-KVU, (1985).

A. D. Saphier, "Transient Analysis of the Pebble-Bed HTGR with the DSNP Simulation Language", Final Report, Vol-I, Soreq NRC (1984).

b. W. Croenbroeck, "Nachrichtung von AVR-Betriebstransienten Unter Vervendung des DSNP-Programsystem", KFA Julich, KFA-ISF-LB-5/86 (1986) IV-21

ON THE GRAPHITE OXIDATION RATE IN HTGRs FOLLOWING AN AIR INGRESS ACCIDENT

Shlomo Ron

Atomic Energy Commission, P.O.Box 7061 Tel-Aviv, Israel

If the HTGR graphite is exposed to air at its nominal operating temperatures it can become oxidized at a significant* rate. Consequently, air ingress into the primary cooling system following a depressurization accident is recognized as one of the potentially hazardous (even though a low probability) accidents which the HTGR safety analysis has to address. Indeed, the impact of massive air ingress events was investigated in several studies<1>"<'4). All of these studies came to the conclusion that even in the highly hypothetical case in which a double break at both the top and bottom of the reactor vessel occurs (leading to the so-called "chimney effect"), public safety is assured. The purpose of the present work is to make an independent estimate of the graphite oxidation rate in HTGRs, assuming an event which is even beyond the upper limit of severity considered in the earlier studies, and to assess its consequences.

Consider a modular HTGR. The very low probability event postulated is the creation of a chimney effect. The effective area of the lower rupture is assumed to be 550 cm2, corresponding to 2% of the effective cross-section area to air flow through the pebble-bed core (the radius of which is 1.5 m2; fuel occupies only 7\% of the core volume). In other words, an upper limit to the rapture size, as far as the air ingress flow rate is concerned, is 2.75 m2. The maximum rapture cross-section area considered in previous studies*1* of air-ingress accidents was about 500 cm2.

An upper limit of the oxygen flux penetrating through the lower hole is obtained by assuming that the graphite is continuously exposed to air at STP conditions (i.e. that the oxidation products do not interrupt the oxygen flow to the graphite). According to the kinetic theory of gases the flux, F, equals to 1/4 v, where v is the average velocity of the air molecules. Thus, the flux of 02 molecules in air at 25°C will be: IV-22

-16 1.38x10 x 298 1/2 =1.1x10 cm/sec -24 L2nx32xl.672x10 As air under standard conditions contains 0.20946 volumetric fraction of oxygen, the partial oxidation of the graphite releasing CO will occur at a rate, Qc, (upper limit) of: 3 1 Qc=Fx550x0.20946xl2x2x(22.4xl0 )- = 1.36 kg/sec=4.9 tons/h. . The rate of graphite consumption is similar to the values obtained by previous studies*1*-*** using sophisticated simulation models, for the first few hours, after which the graphite burn rate is drastically decreased. Volters et al.*1* attribute the reduction in the graphite burn rate to a combination of the following factors: (1) Pressure buildup of the C+02—>2C0 reaction products, interfering with the inflow of air (2) Limited amount of oxygen in the structure containing the reactor - in case this structure is maintained intact. The endothermic reaction C+CO2—>2C0, as they claim, contributes to graphite temperature decrease. This, indeed, is the case as long as the graphite temperature is below 700°C. However, above 700°C the free energy of this reaction is reduced. As the graphite temperatures in the oxidizing zone is expected to be above this 1 value* *, the C+C02 reaction in expected to increase rather than reduce the graphite temperature.

To evaluate the safety consequences of the air ingress event let us assume that the graphite oxidation proceeds at a constant rate without inhibition of any kind until all the graphite is burned up. The HTR-Module contains about 280 tons of graphite . With an upper limit on the burn up rate of 4.9 tons/h, the consumption of this graphite inventory will take about 2.5 days. This long time duration is a key to the safety of HTGRs against air ingress accidents. First, it provides ample time for the actuation of various countermeasures. Secondly, the fission products release will be extended over a number of days so that the risk to individuals will be drastically lower than in case of a puff release*s> (which characterize light water reactor accidents). Moreover, it may be possible to incorporate simple yet effective special design measures to counteract graphite fire, if it happens to start. Finally, the slow burn rate of graphite provides ample time for evacuation, in the very unlikly event that such an action will be necessary. A more detailed analysis of air ingress accidents and their consequences is initiated in order to arrive at a more realistic estimate. IV-23

REFERENCES

1. J. Volters, G. Breitbach and R. Moormann, "Air and Water Ingress Accidents in a HTR-Module of Side-by-Side Concept", IAEA-TECDOC-358, Oak-Ridge (1985).

2. R. Moormann, "Effect of Delays in Afterheat Removal on Consequences of Massive Air Ingress Accidents in High-Temperature Gas Cooled Reactors", J. Nucl. Sci. Technol.. 21, 11, 824 (1984).

3. R. Moormann, "Graphite Oxidation Phenomena During Massive Air Ingress Accident in Nuclear High Temperature Gas Cooled Reactors with Pebble Bed Core", Ber. Bunsenges Phys. Chem., 87, 1086 (1983). 4. J. Wolters, "Aspects of Water and Air Accidents in HTRs", IAEA, IWGGCR/1, Lausanne (1980).

5. S. Ron "The Influence of Averaging Atmospheric Conditions on Dose Levels Calculations", Trans. INS, Vol. 11 (1983). IV-24

ON THE CONSEQUENCES OF FISSION PRODUCTS PLATEOUT AND LIFTOFF IN HTSRs

Shlomo Ron & Ehud Greenspan

Atomic Energy Commission, P.O.Box 7061 Tel-Aviv, Israel

The ceramic coating of the High Temperature Gas Cooled Reactor (HTGR) fuel elements is one of the major contributors to the high safety of these reactors. This coating, consisting of pyrocarbon and SIC layers, is capable of good fission products retention up to at least 1600°C. However, the ceramic coating is not perfectly leaktight even under normal operating conditions; a small fraction of certain fission products, notably Cs, Sr, Ag, Ru, Eu and Sb, can diffuse through and get into the primary circuit. Fission products can get into the HTGR primary circuit also via fissioning of the very small quantities of uranium which contaminates the coated fuel particles during the manufacturing process.

The fission products diffusing out of the fuel graphite matrix are transported by the coolant gas and deposited on the primary circuit surface area (the so-called "plateout" phenomenon). The deposition occurs especially on the cooler surfaces of the steam generator. Subsequently, the deposited fission products are absorbed into the metallic lattice. This penetration is governed by slow diffusion processes*1>. If a depressurization accident occurs, a certain fraction of the plateout inventory can be lifted off the surface of the primary circuit (the so-called "liftoff" phenomenon) thus contributing to the total release of radionuclides to the environment (i.e., to the source term).

The plateout (and therefore, the liftoff) phenomenon may effect the attractiveness of HTGRs via contribution to (a) the source term, (b) maintenance difficulties and (c) embrittlement of the steam generators. The purpose of the present study is to assess the significance of these phenomena on the safety of HTGRs.

CONTRIBUTION TO THE SOURCE TERM

Let us consider first a modular HTGR. The equilibrium inventory of Cs-137 (the dominant contributor to the plateout IV-2S

inventory with respect to radiological consequences*2>) circulating in the helium of the HTR-Module is estimated<2> to be 3.8xlO~6 Ci, whereas the design basis inventory is taken<2>, conservatively, to be 1.9xlO"3 Ci. The expected*2* plateout equilibrium inventory of Cs-137 is 150 Ci, whereas the design basis is taken*2' to be 1500 Ci. During a depressurization accident the circulating inventory and the liftoff fraction from the plateout inventory will be released from the primary circuit with the escaping helium flow. According to Hanson*3' Cs-137 appears to be the hardest nuclide to lift-off; only 3% of the plateout activity is expectd to be blown off. This fraction of the design basis Cs-137 plateout inventory, i.e. 45 Ci, will be released in a short time duration, i.e. in a "puff" release. To appreciate the significance of the liftoff contribution to the source term, let us compare it with the direct release from the fuel elements during a depressurization accident. The equilibrium inventory of Cs-137 in the fuel of the HTR-Module is 6.9xlOs Ci*2'. During a severe depressurization accident the fuel average maximum temperature increases to about 1050/1510°C*A>. It is kept above the nominal operating temperature 700/850°C for several days. The increase in the fuel temperature leads to an increase in the fission products diffusion rate; it is estimated that the Cs-137 fraction directly released from the "hot" fuel elements will be 0.03% of the total inventory*'1'. Of the 210 Ci released (during more than a week) from the fuel, no more than 70 Ci of Cs-137 is expected*5' to escape to the environment. Thus, the contribution of the direct release and of the liftoff to the source term might be comparable. However, as the individual radiological risk drastically decreases when the release occurs during a long period (over one day)*6>, the liftoff activity of 45 Ci poses a much higher individual risk than the 70 Ci directly released from the fuel elements. It should be emphasized that in both cases the risk to the public is negligible.

Consider, next, the integrated reactor FJTR-500. The liftoff risk in this medium power reactor (550 MVe vs. 80 MWe of the HTR-Module) is relatively insignificant as the estimated direct release from the fuel elements during a depressurization accident includes 21% of the total Cs inventory*A'. This is due to the higher temperatures (up to 2350°C<*>) the HTR-500 fuel reaches during such an accident.

The plateout Cs inventory might get out of the primary circuit also as a result of a "washout" accident. If a massive penetration of water into the primary circuit is to occur, it is expected that, due to the high solubility in water, the plateout inventory fraction to be washed-out is larger than the lift-off fraction. However, as only a small fraction of the Cs dissolved in the water is likely to get out of the IV-26

containment, the contribution of a washout accident to the source term is expected to be negligible.

It is concluded that the plateout phenomenon typical to HTGRs can make a significant relative contribution to the source term of modular HTGRs. Its absolute contribution to the radiological hazards of all types of HTGRs is, however, negligible. EFFECT ON PRIMARY CIRCUIT MAINTENANCE

Contamination of the primary circuit may interfere with the control instrumentation and with the access to components for maintenance and repair. However, the experience accumulated so far with the operating HTGRs is very encouraging: the radiation exposure of operating and maintenance personnel is one order of magnitude lower than in light water reactors<7>. The experience to be accumulated in the operation of the recently commissioned 300 MWe THTR-300 HTGR<7> is likely to shed more light on the plateout effect on maintenance. EFFECT ON STEAM-GENERATOR EMBRITTELMENT Steel embrittelment may be caused by higher level interaction of the plated out Cs with the metallic additives of the steel (such as Mg, Mn and Mo) which leads to desegragation of some metals, removing them from the surface and grain boundaries, and leaving there the Cs. For Cs in an iron matrix, the bond strengths ratio of the solute versus solvent matrix is 18.7/99.5 Kcal mole-1 = 0.19 < 1 indicating therefore a strong Cs segragation to the surface and grain boundaries. According to the experimental results of Hanson*31, 362 of the total plateout in the primary circuit deposits on the steam generator surface. Based on the data of Ref. 8 for the 350 MWth MHTGR, it is estimated that the surface area of HTR-Module steam generator tubes is approximately 2000 m2. Therefore the average Cs-137 plateout density on the steam generator surface area will be (using the design basis value for the plateout inventory), 1.4xl015 atoms <=«»-2. The outcome is of the order of magnitude of the saturated monolayer adsorption value. Although all the plateout inventory could have been in the adsorption state, the very low liftoff fraction shows that most of the Cs penetrates into the metallic lattice.

It should be realized that Cs-137 constitues only about half of the total Cs equilibrium inventory. Moreover, the Cs-137 decays into the stable Ba-137 which causes embrittelment in iron matrix of a similar magnitude to Cs. It appears, nevertheless, that the total plateout density is too low to bring about a worrisome embrittelment. A closer examination of the embrittlement issue is, nevertheless, desirable. IV-27

CONCLUDING REMARKS

Based on the information available so far and on the simple analysis performed, it appears that the plateout and liftoff phenomena will not have a significant effect on the safety, maintenance and integrity of HTGRs. A thorough understanding of these phenomena is, however, highly desirable for a reliable assessment of the HTGRs and for the establishment of good operation and maintenance practices. Thus, for example, whenever possible it might be preferable to control the HTGR power level using absorption control rather than temperature control (as the fission products release rate depends exponentially on the fuel temperature).

REFERENCES

1. S. Ron, Z.B. Alfassi and M. Baer, Nucl. Technol.. 2£> 326 (1986), and Chem. Phvs. 117, 39 (1987).

2. Bechtel National, Inc., "Radiological Dose Evaluation, Modular HTGR," (1984).

3. D.L. Hanson, "Results of the General Atomic Deposition Loop Program", General Atomic, GA-A13140, UC-77 (1976).

4. W. Rehm, J. Altes, G. Breitbach, R. Nabbi and K. Verfondern, "Safety Analysis of Small and Medium HTRs under Core-Cooling Accident Conditions", IAEA-TECDOC-358, Oak-Ridge (1985).

5. J. Wolters, G. Breitbach and R. Moormann, "Air and Water Ingress Accident in a HTR-Module of Side-by-Side Concept", IAEA-TECDOC-358, Oak-Ridge (1985).

6. S. Ron, "The Influence of Averaging Atmospheric Conditions on Dose Levels Calculations", Trans. INS, 11. (1983).

7. IAEA, "Status of and Prospects for Gas-Cooled Reactors", Technical Reports Series No. 235, Vienna 1984.

8. "HTGR-Concept Description Report, Reference Module HTGR Plant", DOE-HTGR-86-118 (1986). IV-28 ANALYSIS OF HTGR VULNERABILITY TO EXTERNAL MISSILES

A. Ketter, J. Szabo

Israel Atomic Energy Commission

HTGRs (High Temperature Gas Reactors) have some outstanding safety features in comparison to light and heavy water reactors. This paper discusses its enhanced protectability against external missiles. Although the Israeli licensing does not predefine a "reference threat" (but rather considers it an applicant responsibility), a postulated threat can be determined for the present purpose, in order to enable a first-level analysis on a practial and resonable basis. Once such a reference threat was determined, an engineering analysis was carried out in order to examine the station vulnerability. Analysis of HTR-Module This concept consists of two standard 80 MUe units with steel pressure vessels enclosed in an internal concrete compartment and an external containment building. A penetration analysis is based on nine categories of approach directions as shown in Figure 1. They cover all horizontal, vertical and diagonal possibilities that were found relevant to the analysis. Table 1 shows the concrete thickness for each penetration possibility.

protection width (m) direction compartment external total A cont./pr. vessel 2.05 5.60 B cont./pr. vessel 2.73 6.44 C cont./pr. vessel 2.93 4.77 D cont./steam gen. 2.93 4.10 E cont./pr. vessel 3-72 5.70 F cont./steam gen. 3.22 4.70 G cont./pr. vessel 1.81 4.10 H cont./steam gen. 2.05 3.80 I Cont./pr. vessel 1.95 4.35

Table 1; Penetration analysis for HTR-Module IV-29 The numbers in the column "external" represent the width of the containment walls, while the column "total" represents the total shielding width including the inner compartment, e.g., pressure vessel or steam generator compartments. The penetration analysis reveals that there is no possibility of penetrating the pressure vessel or the steam generator compartments. Furthermore, there is no reasonable approach direction that results with a penetration into the containment itself, nor a possibility of a consequent internal explosion.

Analysis of HTR-500

The HTR-500 is a large integrative design of 550 MWe that mainly consists of a large prestressed concrete Reactor Vessel (PCRV), having its heat transmission components around it. This integral unit is enclosed, as a whole, within a very thick concrete envelope (averaging 5.5 meters thick).

Here, eight approach categories were defined, representing all typical penetration possibilities, as shown in Figure 2.

Table 2 shows the concrete thickness for each penetration possibility:

direction component protection width (m)

A pressure vessel 4.19 B pressure vessel 7.18 C pressure vessel 7.64 D steam generator 1.68 E pressure vessel 6.70 F pressure vessel 6.82 G steam generator 6.23 H steam generator 4.19

Table 2; penetration analysis for HTR-500 IV-30 The penetration analysis for this design shows that the primary system, i.e., pressure vessel and steam generator, is well protected against aerial approach direction, with a very large safety margins (in the order of magnitude of hundreds of percents). Approach direction "D", however, should be carefully considered in order to enhance its safety margins against penetration to an equivalent level.

Analysis of HTGR - SILO This concept consists of a 130 MVe modular unit that its primary system, i.e., pressure vessel and steam generator, is located completely underground in a silo, having a massive concrete shield, approximately 1 m thick, atop of it. The only approach direction for this design is from the top, since the ground provides sufficient protection against aerial missiles. The top itself might be a weak point, but since it is only about 10 m in diameter, it could easily be redesigned for purposes of better protection. Summary As a general conclusion, it could be stated that the above HTGR concepts have satisfactory protection potential against aerial missiles. Only minor modifications (if any) will be needed when external threats are considered. IV-31 HTR-IOO

1 Reactor pressure vessel 2 Steam generator pressure vessel 3 Connecting pressure vessel 4 Primary circuit blower 5 Primary cell 6 Protective shell 7 Surface cooler

Fig. 1: HTR-Module cross-section with penetration directions TV-32

Incore rod

H generator (vf) Cooling water -Thermal shield

Auxiliary - "Ceramic circulator internals

Fuel element discharge pipe HTRSOO BBC Reactor Pressure Vessel with Internals 85.57-2

Figure 2: HTR-500 cross-section with penetration directions Reactor Physics, Concepts and Calculational Methods

Chairmen - W. Rothenstein and A. Galperin V-l

ACCURATE MU.'YTIGROUP CROSS SECTIONS FOR RESONANCE REACTIONS

W. Rothensteia Department of Nuclear Engineering Technion, Haifa 32000, Israel

The preparation of multigroup cross sections for reactor analysis from basic nuclear data by preprocessing codes cannot include the relevant group values for resonance reactions, because they are not problem independent. Even for homogeneous assemblies, the flux depressions within the resonance groups depend or a number of factors which include the slowing down properties of the nuclides contained in the assembly, the nature of the resonances of the fuel isotopes and the interference effects between them, and the extent to which the flux recovers between the individual resonances. For heterogeneous assemblies, the spatial flux changes, and the energy dependence of these changes, may also be significant. It is clearly impossible to tabulate group resonance cross sections in such a way that their dependence on all the characteristics of a reactor region, such as a fuel assembly, including the fuel temperature and burn-up are fully taken into account, as the number of independent variables would be far too large.

A Bondarenko type prescription is usually adopted to describe the group resonance date which may be shielded cross section or effective resonance integrals. The data are tabulated as a function of two variables only, the temperature and a background cross section. The preparation of group data, which depend on such a limited number of variables, must refer to a very simple system; it is generally taken to consist of a single resonance absorber and scatterer mixed homogeneously with a hydrogenous moderator in different proportions. The application of the data to different problems, homogeneous and heterogeneous, requires care and experience in the specification of the background cross section which is appropriate for a given assembly.

The question arises how such data can be used, without complicating them unduly, to provide accurate group cross sections for resonance reactions in an actual assembly, which is in general heterogeneous and contains a mixture of resonance absorbers. It is shown that three ingredients are needed: V-2

a) An accurate transport theory code which can calculate resonance reaction rates in assemblies of the type encountered in lattice analysis, possibly including two dimensional effects, and which can also be used to prepare the group resonance data, of the Bondarenko type, needed for multigroup reactor analysis.

b) Specification of the multigroup cod«* in which the group resonance data are used.

c) Introduction of an optional routine into the multigroup code, which can evaluate cross section corrections for the values resulting from the tabulated group resonance data. This should be done for suitably selected assemblies, by appropriate comparisons of the information resulting from the transport code on the one hand, and the resonance algorithm from which the group cross sections are derived in the multigroup code on the other.

It is clearly desirable to select the resonance shielding algorithms in such a manner that the correction factors, referred to in (c) above, are as close as possible to unity. If this is the case, they have to be calculated only for a few benchmark problems, and then be introduced into the multigroup code in the form of tables which lend themselves readily to simple interpolation.

The comparisons between results obtained by the transport and multigroup codes, which are needed to determine the cross section corrections to be applied to the group resonance cross sections of the latter code for a particular assembly, involve a well known problem in reactor analysis, in which an accurate calculation for a relatively small reactor region and a limited range of energies precedes a far less detailed calculation, which is to be made many times on a routine basis for all neutron energies. The resonance absorption problem is similar, from this point of view, to the problems encountered in fuel assembly homogenisation, prior to the determination of flux and power distributions throughout a reactor core. The quantities which must be preserved between the codes are reaction rates, not average cross sections. The latter are merely convenient quantities which are listed in the output of one code, and used as input in the other. The cross sections which preserve the reaction rates are however the reaction rates, calculated by the first (accurate) code, divided by the flux values determined by the subsequent {simpler) code. These V-3

flux values are clearly not known a priori, since the desired cross sections are needed for their determination.

In the light of these observations, the optional routine referred to in

Corrections for two Bondarenko type resonance shielding algorithms applied in a lattice physics code - the LEOPARD program - will be presented and discussed. They allow for inevitable shortcomings in the equivalence relations used to apply the algorithms to heterogeneous assemblies, as well as for interference phenomena between resonances belonging to different fuel uuclides. The correction factors are temperature and burn-up dependent. As they are close to unity for the important resonance groups, specially for one of the shielding algorithms, and vary only slightly with burn-up in the groups in which interference; affects are appreciable, they can be readily applied to a range of problems of a similar type, in this case PWR lattices.

Acknowledgement

This project was partially sponsored by the Israel Electric Corporation. V-4

INTRODUCING NOXER-A 3D NODAL DIFFUSION CODE

M. SEGEV . C.E.N./SACLAY (On leave from the Ben Gurion University)

If the interior of LWR assembly be homogenized in two neutron groups for the XY plane, and if slices of 20 cm or more along the Z (fuel) axis can be considered homogenous, then the two-group flux is separable in most of the volume of such an XYZ slice, or node. This fact can be used to advantage by assuming strict flux separability in all of the node interior. A simple, first order, diffusion nodal theory then follows , serving as a basis for an efficient numerical nodal code. This, the NOXER code, executed with one numerical point per node, generates the XY assembly-wise powers of a LWK in 15 to 18 iterations, consuming about 5 seconds on the IBM 4361, or 0.25 seconds on the CRAY/XMP. These distributions of assembly powers come out typically with an average error of ^ 1.5 % and a maximal error of ^ 3.5 %. A 3D case, corresponding to a 2D case with added eight to ten Z slices, converges in about 20 to 25 iterations, a tenfold in computing time, and its Z-integrated XY powers-map comes out again with ^ 1,5 % and "v 3.5 %, average and maximal, errors.

The efficiency of the programm derives from the basic separa- bility stipulation and is manifest in a twofold manner. One - the 3D differential equations are reduced to 1D algebraic equations for the nodal fluxes. Two - each such 1D equation can be treated as almost homogenous, namely as meakly inhomogenous, resulting in a low dominance iteration ratio, in the realm of 0.6 to 0.8.

The NOXER code has been extensively tested with familiar benchmarks, sach as the 2D and 3D IAEA, as well as with numerous examples of core loadings. In these examples the "exact" solutions were V-5

derived from a finit-element programs l2' l' run with 4 cubic polynomials in XY in each node, and with ^ 40 Z slices, each treated by a quadratic polynomial.

Work on flux reconstruction inside a node is in progress. The very functions (trigonometric or hyperbolic), used to express the sepa- rable flux solution in a node, are used for the reconstruction with excellent results. In particular absent are the oscillations typical to polynomial reconstruction.

In what follows we present the NOXER performance vs. the 2D and 3D IAEA benchmarks. To obtain the results shown, NOXER was run with 1 point in XY per assembly, and with 6 Z-intervals for the 3D case.

REFERENCES

[!] M. SEGEV Two group diffusion theory based on separation of variables. To appear in Annl. Nucl. Energy

[2] J.J. LAUTARD, SERMA/SACLAY, private communication. V-6

I THE IAEA BENCHMARK 2D AND 3D CORES

I Table 1 : An XY core layout for the IAEA benchmarks.

t w w w w / 1 1 1 w w w Fuel Type

1 1.11 2 2 1 1 1 w W 2 1.05 2 2 2 2 3 3 w w 3 0 .69

5 3 2 2 2 2 1 1 w

2 2 2 2 2 2 1 w w * 2 2 2 2 2 2 1 1 w

2 2 2 2 2 2 2 1 w

3 2 2 2 3 *> 2 1 w

6 7 8

The 2D Problem : Z infinite ; the XY layout is as in Table 1 (w s "water") ; broken lines designate symmetry lines. Assembly pitch is 20 cm.

* The 3D Problem : Z height is 380, of which 20 cm water at bottom and 20 cm water on top. The XY layout for the 340 cm high active-fuel core is as in Table 1, except that at position (3.3) the top 80 cm are replaced by fuel type 3. V-7

Figure 1 : Assembly Powers (<> »1000) for the IAEA benchmarks.

xxxx exact power XX Noxer error

585 + 10

2D 471 686 597 -16 + 10 - 6

1193 967 906 846 + 13 +17 + 1 -10

1469 1345 1179 1071 975 692 + 13 + 4 + 14 -10 -14 -32

1435 1480 1315 1090 1036 950 736 • 23 + 9 + 16 + 16 + 5 -29 -32

746 1310 1454 1211 610 935 934 755 -23 + 29 + 28 + 23 -25 - 1 -16 -27

597 +24 3D

476 700 610 -12 -23 - 5

1177 971 923 866 + 10 +20 +13 - 2

1367 1309 11R0 1088 999 709 + 4 + 1 + 13 - 5 - 9 -29

1396 1430 1290 1071 1054 975 756 + 10 + 3 + 10 + 15 + 3 -18 -37

729 1281 1421 1193 610 954 959 777 -24 + 20 • 22 + 19 -22 0 -12 -37 V-8

EQUIVALENCE METHOD FOR RESONANCE SHIELDING CALCULATIONS IN DUPLEX PELLET

ROD LATTICES

S. Carmona

Soreq Nuclear Research Center, Yavne 70600, Israel

M. Segev

Ben-Gurion University of the Negev, Bcer-Sheva, Israel

(1 2) Duplex pellet rod lattices ' have recently been studied for improving fuel utilization and material endurance in LWR cores. The treatment of resonance shielded cross sections in these non-uniform rods is cruicial in analyzing their lattices.

An equivalence principle for resonance shielding calculations was developed for the annular fuel zone of the duplex-pellet. We generalised this problem to a three-region heterogeneous nuclear fuel system con- sisting of a fuel lumps in an infinite moderator medium. The fuel lump consists of two fuel regions: the central one and surrounding it an annular (or ring) region. The fuel compositions include the same absorber.

The treatment of the integral transport equations for this heterogeneous system proceeds in the usual way, leading to reduced equivalent forms as for two region systems.

By applying the Wigner rational approximation, a formula was derived for £e_, the escape cross section in the ring fuel zone of the duplex-pellet. For a single fuel lump: R«

where R. and R? are, respectively, the inner and outer radii of the fuel ring sone.fl flyi s an adjustable factor having the same function as the V-9

Bell factor, a, i.e. to relate a "grey" - real - medium to a black medium. At the black medium limit 3,-1 and Eq. (1) is similar to that for a black fuel annulus of a hollow pellet. This similarity implies that the corresponding effective absorber surfaces for neutron escape are the same for both annul!i in this limit.

Figure 1 shows £e~ curves derived from exact effective absorption cross (3) sections computed by the RABBLE code for the annular fuel zone of two rods (R« = 0.5 cm, 0.6 cm) as a function of the inner ring zone radius. These curves are compared with those obtained for E „, the black hollow cylinder escape cross section. The proximity of the two curves is evident.

The 6_ factor, needed for the escape cross section expression, was found to be a smooth function of the different geometrical and physical parameters on which it depends. Thus, adjusted averaged values can be fixed to fit large ranges of these parameters.

Table 1 summarizes a comparison of self shielding factors obtained by accurate calculation (with RABBLE) and by the equivalence method. The dis- crepancies are very slight. The values of the factor (averaged) <3-> are seen to be very close, although there is a wide range of N-/N- values for the different cases.

Another validation of this equivalence method was carried out by computing the spatial distribution of resonance integrals in uniform fuel rods. We obtained good agreement between computed values and measured data from Hellstrand's experience with uranium metal rods, as shown in Fig. 2.

The equivalence method for resonance shielding calculations for duplex - pellet rod lattices was assessed by further treatment. The results ob- tained are accurate to a few percent; the level of accuracy is typical of equivalence treatments of regular fuel rods. This practical equivalence relation can be implemented in cell lattice calculations to make the analyses of duplex pellet rod lattices possible with reasonable use of computer resources. V-10

in In

1.0 -

1.0

K, IKNER RADIUS

1. - Escape cross sections In annulus of uniform fuel rods (computed from RABBLE) and In annular region of black hollow rods.

c Hunt fv.ilua! l(»n u JO o oea•lured (llellnt r;in •J.O •».» •s.o m i I t 1 1 1 i i 20 •, 1 £ i <1 8 il s 4 ^1 /•' 1 s 10 J 1 A u i 1 J! I 1 i 0 5 0 Radial Distance Rj (ca) Figure 2. - The spatial distribution of the effective absorption resonance integrals for rods of different diameters V-ll

REFERENCES:

(1) R.J, Allen and B.J. Wrona, "Duplex Fuel Pellet Manufactoring Feasibility Study", Final Report, EPRI NP - 2653, RP 1581-6 (1982). (2) J.B. Aincough, et al., "Preliminary Study of Cost Benefits Associated with Duplex Fuel Pellets of the LOWI Type", Nucl. Technol., J51, 3, 521 (1983). (3) P.H. Kier and A.A. Robba, "RABBLE, A Program for Computation of Resonance Absorption in Multiregion Reactor Cell", ANL - 7326, Argonne National Laboratory (1967). (4) E,, Hellstrand, "Measurements of the Effective Resonance Integrals in Uranium Metal and Oxide in Different Geometries", J. App. Phys., 28, 1493 (1957).

238 Table 1. - U effective resonance shielding factors evaluated: exactly- f(R) (by RABBLE) and by equivalence relation method - f(E), in rings of single duplex pellets, for different relative concentrations of the absorber in the two regions of the pellet, .24 At 10 N, - Constant m 0.0174 N, cm

< 27.7 eV 9.9 eV 366.9 eV Nl V (barn) 16.0 eV 4.0 eV 4.0 eV

f(R) 0.0833 0.0852 0.1004 0.1 1.10 84.3 1) X -0,5 -0.6 -0.1

f(K) 0.0822 0.0841 0.0990 0.5 1.05 80.5 I) X 0.5 0.4 0.6

f(K) O.O8.-5 0.0835 0.0981 1.0 1.05 80.5 I) X -0.1 -0.4 -0.3

f(K) 0.0812 0.0831 0.0976 1.5 1.00 76.6 I) X 1.8 1.6 1.4

f(K) 0.0809 0.0829 0.0973 2.0 1.00 76.6 ) X 1.4 1.3 1.1 V-12

THE INVERSE UNCERTAINTY ANALYSIS

Yigal Ronen

Department of Nuclear Engineering Ben-Gurion University of the Negev Beer Sheva, Israel

Uncertainty analysis of linear systems is usually referred to as a problem of calculating CR

(the covariance-variance matrix of the responses)w/ie/t Ca (the covariance-variance matrix of the input parameters) is known. Namely the calculation of the uncertainties of the responses (integral parameters) when the uncertainties of the cross sections and the other nuclear data are known. The inverse uncertainty analysis for linear systems is an analysis used to calculate (or estimate) Ca when CR is known. Namely we know the uncertainties of some integral parameters (responses) usually from the differences between the measured and the calculated values, and we would like to calculae (estimae) the uncertainties of the system cross sections. The purpose of this paper is to present a method for estimating the covariance-variance matrix Ca of the input parameters for a given matrix CR when the number of input parameters M is larger than the number of responses N.

Let us assume that a vector Aoce can be determined by a linear model

Act e = AAR + b (1) where A is a matrix of NxM. The vector b has M components. The vector Act* is denoted as the estimation vector of the input parameters, where AR is the change in the responses.

The estimation vector Aoce estimates the vector Aa of the input parameters when two conditions are satisfied (1) E(Aoc e-Acc ) = 0 (2a) (2) Q = E(Acc e-Aa ) (Aa e-Aa )T = minimal (2b) V-13

Requirement (1) [Eq. (2a)] asserts that the expected value of the difference between the estimated vector Aae and the real vector Aa should be zero.

Requirement (2) [Eq. (2b)] asserts that the variance of the difference between Aae and Aa should be minimal.

Requirements (1) and (2) enable us to obtain the estimaed vector Aoce. Once this vector is known the estimated covariance-variance matrix of the input parameters can be obtained

e eT (fa = E(Aa Aa ) (3)

Using (Eq. 2a) we get for Q of Eq. (2b): Q=E(AAR-Aa )(AAR-Aa )T=

E(AARARTAT) + E(Aa Aa T)-E(A ARAa T)-E(Aa ARTAT) (4) Assuming no correlations between the uncertainties of the input parameters and the responses we have E(ARAa T) = E(Aa ART) = 0 (5) and Eq. (4) will have the form

T Q=ACRA + Ca (6)

We would like to find the matrix A so that the matrix Q will be minimal subject to the constraints of requirement (1). Thus we require the functional QQ to be minimal

T +X T(STAT-1) (7) where X is the Lagrangian multiplier matrix.

The variation of QQ is given by

+ 5A) - (^(A) = (A + 8A) (^(A + SA)T + C + [I - (A + 8A)S] X

T T T T T T T X [S (A+5A) -I]-ACRA -Ca-(I-AS)X -X (S A -1)

T T 7 T T T 8ACRA + ACR 8A + SAC^A " - X S 8A - 6AS X (8) V-14

Rearranging Eq. (8) we have

T T T T X S )8A +8ACR8A (9)

in order that the variation SQQ will be equal to zero we must have « T T AUp - A, a = U (1U)

Thus A-xVc? (11)

Multiplying Eq. (11) by S on the right hand side we can obtain AS=Jl TSTq1S = >. TB (12)

From Eq. (12) we have X T = B+ (13) where B+ is the pseudoinverse for B. Combining Eqs. (11) and (13) we have for the matrix A the following relation A = B+STC^ (14)

We also have

+ T I Aa • = B "S CI AR (15)

Thus the estimation of the covariance-variance matrix of the input parameters is given by

+ T 1 T +T

T =_, Bn oc /C^U c n(AKA/ AD A i>K \ )r ^L ^c oio l —=J. n) co rC^^ r ^ p^ C^D" =- *o> **DD" = *X*» ***I>O* (16/1 ^)\

Thus

C*=B+ (17)

For the cases that it is impossible to obtain the matrix B+ which is the pseudoinverse cf S) we can use a different approach. V-15

The general solution of the vector Eq. (2) is Aa e = S+AR + (I-S+S)y (18) where y is arbitrary, provided that the equation has a solution. The equation Aa C = S+AR (19) is the usual least square solution for Eq. (2). The least square solution is the best approximate solution. Thus the estimation of the covariance-variance matrix of the input parameters is obtained from Eq. (19) (f = E(S+ARARTS+T) = S+E(ARART) STT (20) a Thus, e + +T C =S CBS (21)

since the rows of the matrix S are linearly independent the marices S+ and S+T has the form

S+ = ST(SST)"1 (22) and = (SST)*1TS (23)

e Thus the estimated covariance matrix Ca has the form

(24)

Consider a special case of one response and M input parameters. Thus,

where cR is the standard deviation of the response. Furthermore

/CO \ /CO \ /C i O i i O \ O£L\ (aJ» ) = (aa ) = (b. + o, + ... + o..) (26) and Eq. (24) has the form

C =

Impact of Pn Approximation in Calculations of Transfer Cross; Sections and Scalar Fluxes for Anisotropic Scattering of D-T Neutrons in Graphite

R.Ofek, A. Tsechanski and G. Shani Department of Nuclear Engineering,Ben Gurion University of the Negev, Beer Sheva 84105, Israel

The desired "order of scattering", namely, the order of the Pn expansion of the transport equation, is of major importance for the accuracy of neutron transport calculations with Pn,multigroup, discrete-ordinates transport codes. In the present work, the accuracies of the Legendre components of the multigroup transfer cross sections, and their effects on the scalar fluxes, are investigated. In our study we have calculated the Legendre components of multigroup transfer cross sections with the code NJOY^. That, for scattering of neutrons by carbon, either elasdcally or inelastically via the first three discrete levels of carbon. The neutrons are scattered from a source energy group of 14.5 - 14.8 MeV into equally spaced sink groups with widths of 0.2 MeV at the energy range of 2.5 -14.8 MeV. The N-th "order of scattering" multigroup transfer cross section, for a scattering from the source group into a sink group via the i-th level (where the elastic scattering is considered as a zero level scattering), is given as:

N (i) n ON (g) = S (2n+l) a ' (g) Pn(ftg) (1) V-17

1 where o' ^ is the n-th Legendre component of the transfer cross section, Pn is the

n-th Legendre polynomial with an argument p.g, the cosine of scattering angle in the laboratory coordinates system which is averaged on the source group and on the group g.

Now we define the parameter PCSDN(g) as the percentage deviation of the sum of the N-th transfer cross sections (N~ 4,5,6) via all the scattering level, relative to that with N= 7, for a scattering into the group g. Namely:

i=0 PCSDN(g) = (2)

i=0

The values of PCSDN with N=4,5,6 are shown in Fig.l. First, it is seen that the spread of the parameter with 1,ht energy groups is quite arbitrary. Second, the average values are: PCSD4 =22%, PCSDg ~ 10%.

B D El B BB a HOB B|3HQ El D EI 40- EP h ° 13 B QP 9L 1 •£I1 30- % •a f -• 13 Q* EL B 8 11 * N = 5 • " • u 20- "V*"*.V Q. fir 10- •I • «• • n • _ • QB • •• B • B 0- 1—0 Cl 5 10 15 ENERGY (MeV) y-18

Moreover, a comparison has been made between the 7th order of scattering multigroup transfer cross sections and the rigorous values of the transfer cross sections , for each of the scattering levels, that is shown in Fig.2.

ENERGY (M«V)

Figure 2

It might be seen that even the 7th order Pn approximation, which is a fairly high "order of scattering", underestimates considerably the values of the transfer cross sections.

Nevertheless, the convergence of the scalar flux with the Pn appriximation is much faster. It might be seen that the accuracy of the scalar flux calculation is less affected by the accuracy of the scattering cross section. The transport calculations of the fluxes have been carried out by the discrete- V-19

ordinates code DOT 4.2 ^\ with a geometry comprising a graphite cylinder bombarded by a collimated D-T neutron beam. The details of the calculations are given elsewhere @\ The fluxes calculated with P4, P5 and P^ "orders of scattering" were compared to those of P7, taken as a reference. The percentage flux deviation (PFD) of the neutron spectra, for some detector locations are shown in Fig.3. Two conclusions may be drawn from Figure 3: 1) The calculation error in the flux is much smaller than that of the transfer cross-sections (< 1% for P5 and <8% for P4). 2) The PFD are well correlated with the forward and backwards anisotropies of the elastic and the discrete-level inelastic scattering, particularly at the detector location close to the axis of the system.

That may be understood by examining the Pn approximation of the Bolzman transport equation: the Legendre components of the flux appear in all the tenns of the equation (except the "source" term), while the Legendre components of the scattering cross section appear only in the " in-scattering" term, multiplied by the Legendre components of the flux. Thus, besides being affected by "in-scattering", which is determined by the scattering cross-section, the flux is also affected by the "leakage" and the "out-scattering" terms. Hence, in an energy group where the anisotropy is less pronounced, die Legendre components of the flux vanish faster than the Legendre components of scattering cross-sections. Thus the contributions of high-order products of Legendre components of the flux with those of the scattering cross-section are negligible at energy groups in which the angular distribution var™ «i«'"ly, and where the Legendre components of the flux are very small. Therefore, the "in-scattering" term, and the scalar flux as well, at those energy groups are determined almost entirely by the low order distributions of the V-20

Legendre components. That explains the difference between the convergence of the flux to that of the transfer cross-section, and the correlation of the calculational error in the fluxt o the anisotropy of the scattering.as well.

43x15 MESH Z=47.5cm R =1.5 cm 1°

if.

a

I I i i i i I i I i i I

(b) R = 23cm

J-SI COLLISION H-SILEVEtlNE-^ LASTK SCIVimr, B<»5E

X ,1-STCCUISION, (•2-KD IEVEM INELASTIC SCATTEBHG ,1-STCCUISION *m£ H-SOLEVEH INELASTIC SCATTERING UJ RANGE a o

i i i i I i i i i I -2 LLJ (c) o 3 on R=46cm 2

I -

0

- I -

-2 - . P6RUN to P7RUN

-3 x P5RUN to P7RUN

-4 - a P4RUN to P7RUN 1 1 I -5 - 5 10 ENERGY (MeV)

Figure 3

References 1. R.E. Mac Farlane, D.W.Muir and R.M. Boicourt, "The NJOY Nuclear Data Processing System . Volume I: Users Manual", LA-9303-M Vol.1 (ENDF- 324)(1982) 2. W.A. Rhoades, "DOT - IV Version 4.2 - Two Dimensional Discrete Ordinates Radiation Transport Code System", CCC-320 , RSIC (1982) 3. R. Ofek, A. Goldfeld, A.Tsechanski, and G. Shani, Ann. Nucl. Energy, 13,23 (1986). V-21

MINIMUM THICKNESS BLANKETS FOR FUSION REACTORS

Y. Kami S. Greenspan

Nuclear Research Center-Negev Atomic Energy Commission P.O.Box 9001 P.O.Box 7061 Beer-Sheva 84190, Israel Tel-Aviv 61070, Israel

There is mounting evidence*1"*' that the thinner the blanket (and shield), the more economically viable the fusion reactor can be. The present work reports upon the results of a recent systematic nucleonic optimization study aimed at identifying the minimum possible thickness of a number of blanket concepts. This study is an extension, to new blanket concepts, of the blanket optimization studies reported upon in Refs. 5 and 6; it uses the same methodology. The first blanket concept considered is of the Be/-/TBM/He/SS (standing for neutron multiplier/neutron moderator/tritium breeding material/coolant/structural material) type. Recently this blanket type having Lii7PbB3 for the TBM was found the most promising for the MINIMARS fusion reactor<7>; It was significantly thinner than blankets using Li and Flibe for the TBM. Seeing no physical justification for the superiority of the LiPb TBM (when used in addition to Be), we set-out to identify, systematically, the minimum thickness of Be/-/TBM/He/SS blankets using either LiPb, Li or Flibe.

Fig. 1 compares the tritium breeding ratio ). All the blankets use a PCA first-wall (FW), 0.4 cm in effective thickness, and are followed by a reflector-shield made of W(90v/o) + H20(10v/o). Liz0 is used as one of the TBM as it offers the highest lithium density attainable. Throughout this work the lithium is taken to be fully enriched. Realizing that the lithium density in Li20, Li, Flibe and LiPb is, respectively, 8.09, 4.60, 1.65 and 0.55xl022 atoms/cm3, it is concluded that the higher the lithium-6 density of the TBM, the thinner the Be/-/TBM/He/S blanket can be for a given TBR. Hence, LiPb is the inferior TBM for the blanket type under consideration. The fact that a LiPb blanket was found the most promising for MINIMARS<7> is likely to be attributed to the specific design approach rather than to the TBM used. V-22

I 1.6-

10 20 30 10 20 30 40 EFFECTIVE BLANKET THICKNESS(cm) EFFECTIVE BLANKET THICKNESS (cm)

Fig. 1: Maximum Tritium Fig. 3s Max. TBR for Zr/H20/ 6 Breeding Ratio (TBR) Li20 blankets attainable from He-cooled (>10v/c Zr) characterized by Be blankets using different (a) No restrictions (b) Li2O/ TBHs. Blankets contain H20 volume fraction (VF) is 6 lOv/o PCA, preceded by a fixed at 1/9; VF (H20+ Li20)<10% PCA first-wall (0.4 cm in (c) Uniform composition? ^LizO/ effective thickness) and H20 VF is variable (d) Same as followed by a "C" but 6L12O/H2O VF = 1/9. 6 W(90v/o) + H20(10v/o) wYH20/ I.i20 blankets are of type reflector-shield. a. Reflector-shield as in Fig.l. The second blanket concept considered is of the M/TiH2/Li20/He/S type (where M stands for either Be or tungsten (W) and S for either SS or W FW and structural material). The purpose of this study was to find out whether the use of an hydrogenous moderator in addition to a neutron multiplier may enable improving the blanket design. Some of the studies 9 reported upon found that Be was superior to TiH2< > while other studies recommended using an hydrogenous material (specifically ZrHi.7 at the outer part of the blanket) along with a neutron multiplier*1O>.

The investigation is done by applying SWAN<3), for the first time, to search for the optimal distribution, across the blanket, of 3 constituents - Be/M/TBH (previously SWAN was applied only to 2-component optimization). Fig. 2 compares the optimal 3-component and 2-component blankets identified. It is concluded that the inclusion of TiH2 as a blanket constituent increases the TBR attainable from a given thickness. Moreover, it can significantly reduce the required inventory of both M VOLUME FRACTIOH VOLUKS FRACTION

rt 01 o 3" c 3 12 a. (B _w CO a. T3 •? 3" > fl) H P> 3 0) H o to < O a. rr 3" fl> to IB IB • *• .^ «: i—» WO P> n LO (B o o w rt rt 01 O •a 3 3 3 3" 3* (3B Ui-O < rHt - p) O ota H. (B IB 3 i-i fl) It I-i 3 IB O O Q. OQ •a 3" a. Hi 3 it M» I-" H- it fD td Ml i^* J3 3 01 IB fl) OQ o P) Mi o • o IB o H' •o Q o 3 (-{ 3 01 H r> m xs t-j 3 H* 3 & CB o 1 o 03*0 o IB o r- 01 PO o o =TT3 Hi P) P) M. IB i-3 (X 3 « hi O a. CO P- M- fl) ho o 3 n 33 03 rt (B CO r^ o 10 > CO o If i-3 3 ON ^ o •a If to (A 0) Hi e Hi |-|* S£ M* a. IB ft H p> IB Q. a, 3 Hi 3" VOUffC FRACTION •a H hJ. f> f IB I-I H o. a. O < (0 Sal H- 3 M- M o 0> <^ O to Hi o 01 M- P) IB a> Hi o 01 H P3 rt 3 IB 3 3" p> rt i-t a. It) r* IB 3- 3 3" 3" a- a> O a o •a IB i-l r> Q- l-h c3 (B IB !-•• «: §" CO Hi P> ^J IB >/ •o a> i at O it n> i-I CO 3 If IB o o CO a rt 3 if O c 3 5" f CO O W O p> M rt o O CO 3 i-i C o o fl) m Q. 3 l_iT3 I—" (B o- *^ •o P) O Hi IX N3 o 3 M 1 CO P> ^ 01 M n M« 3 P> M- • H* 3 Ln 01 r-r cr o (B 1 1 o t- * i— (—1 it O ts rt rt o H-* fl) 3" 3 to CO O 3 )—i a. (B • V-24

find out to what extent one can increase the TBR by designing the ASCB to have a non-uniform composition. Another purpose of the work was to clarify the relative effectiveness of Hz0 and 41Z Da0 for ASCB applications; earlier studies > found D20 to offer a higher TBR. A third purpose of the work was to check the suitability of W as the structural constituent of ASCB; earlier studies considered only Zr or SS<11'12>.

Fig. 3 compares the maximum TBR attainable from ASCB subjected to different design constraints. Of particular interest is the comparison between the uniform composition blanket (Curve "d") and the variable composition blanket (Curve 6 "b"), both using a 90v/o H20+10v/o Li20 aqueous TBM. It is concluded that by designing ASCB to have a non-uniform composition, it is possible to improve their tritium breeding ability. Ha0 was found to offer a somewhat higher TBR than Da0. Tungsten was found to be as good a neutron multiplier as Zr, up to a TBR of « 1.25. For higher TBR Zr shows a clear superiority.

REFERENCES 1. R.A. Krakowski et al., Nucl. Technol./Fusion, .4, 342 (1983). 2. D. Steiner, Nucl. Technol./Fusion, 4, 332 (1983). 3. L.M. Waganer, Fusion Technol.. 8_, 55 (1985). 4. L.J. Perkins, Trans. Am. Nucl. Soc, 52., 125 (1986).

5. E. Greenspan, A. Kinrot and P. Levin, Fusion Technology. 8_, 619 (1985). 6. A. Kinrot and E. Greenspan, Trans. Amer. Nucl. Soc, 4£, 104 (1985). 7. L.J. Perkins, Trans. Am. Nucl. Soc., 52., 125 (1986).

8. E. Greenspan, W.G. Price, Jr. and H. Fishman, "SWAN - A Code for the Analysis and Optimization of Fusion Reactor Nucleonic Characteristics", MATT-1008, Princeton Plasma Physics Laboratory (1973).

9. L.A. El-Guebaly, Trans. Am. Nuc.l. Soc, 52, 300 (1986). 10. S. Taczanowski, Ann. Nucl. Energy 9, 331 (1982).

11. L. Deutsch et al., Trans. Am. Nucl. Soc. 52, 129 (1986). 12. T.A. Parish et al. Nucl. Technol./Fusion, 4, 811 (1983). V-25

LIQUID-METAL MHD CONVERSION OF NUCLEAR ENERGY TO ELECTRICITY REVISITED

E. Greenspan1, A. Barak1, L. Blumenau2, H. Branover2, A. El-Boher2, E. Spero2 and S. Sukoriansky2

1 Israel Atomic Energy Commission, P.O. Box 7061, Tel-Aviv, Israel 2 Center for MHD Studies, Ben-Gurion University of the Negev, P.O. Box 653, Beer-Sheva 84105, Israel

The development of liquid metal (LM) magneto-hydro-dynamic (MHD) energy conversion (EC) technology is associated with nuclear energy since its inception/1) A number of LMMHD power conversion system (PCS) concepts, most of them aimed for space applications, were proposed in the sixties, whereas in the seventies the LMMHD EC related activities (carried out primarily at the ANL) concentrated on terrestrial (and marine) applications (See Refs. 2 and 3 for review). Recently, a re-assessment of possibilities for improving the design and performance of nuclear reactors was undertaken as part of a more comprehensive evaluation of the promise of the LMMHD EC technology. This work reviews our present perception of possibilities for improving the performance of nuclear reactors with the aid of the LMMHD EC technology.

LWRs Having an indirect cycle and higher primary coolant temperature, it appears that the PWRs could benefit more than BWRs from coupling to the LMMHD EC technology. By incorporating the LMMHD EC technology in the PWR PCS, it is possible to improve the performance of PWRs in a number of ways: (a) Reducing the wetness level the steam turbines will have to operate at; (b) Replacing the steam generators by smaller and simpler

H2O-to-LM (lead alloy) heat exchangers and; (c) Increasing the EC efficiency, by virtue of the near isothermal expansion of steam in the MHD generators (see Fig. la) and, possibly, by some increase in the steam high temperature. In order to realize a cycle efficiency gain, the MHD generator efficiency should exceed -80%. V-26

HTGR One of the unique features of HTGRs is that their helium coolant can be provided at very high temperatures - possibly exceeding 950"C. The upper steam temperature of the Rankine cycle (550°C) is well below the temperature delivery-ability of HTGR energy. A novel LMMHD cycle using Mercury for both the electrodynamic (i.e., LM) and thermodynamic (i.e., vapor) working fluids appears to make a perfect match to the HTGR at above the steam cycle temperatures. This so called "wet-vapor" cycle W is proposed as a topping cycle (See Fig. lb). In addition to increasing the HTGR EC efficiency by ~7%, the use of the LMMHD PCS eliminates the possibility of water ingress into the reactor pressure vessel*5) heat exchanger. However, the compatibility of Hg with structural materials and graphite at the temperatures of interest (~600°C + 950°C) is not clear yet.

-p-S

Ranldne steam b. Dual cycle for HTGR: c. LMMHD Ericsson cycle for PWR T: Hg LMMHD)wet-vapor cycle cycle for LMR using conventional B: Rankine steam cycle conventional LMMHD (also applicable to LMR) compressor - - - conventional LMMHD LMMHD m VJ1V11M» kUT compressoVrUAljpi WJ9Ur1

Fig. 1. T-S diagrams for LMHD EC cycles found promising for different types of nuclear power reactors.

LMR Another novel LMMHD EC cycle, referred to as an "all LMMHD Ericsson cycle"*6), is proposed for liquid metal reactors, provided they could be designed to deliver their energy at £650°C (See Fig. lc). Fig. 2 illustrates the coupling of this EC cycle to a LMR in comparison with a conventional 3-loop LMR. The primary merit of the all LMMHD Ericsson cycle is that it enables eliminating the (relatively complicated) tertiary water cycle thus significantly simplifying the PCS and eliminating the hazard of H2O-Na interaction altogether. The efficiency of the all LMMHD Ericsson cycle is estimated at ~40%, when the cycle high/low temperature is 650°C/25°C.(6) V-27

516-0 5,6-c STEAM DRUM 27MN/m8

HP. I.P. SECONDARY SODIUM PUMP• y

370"C

PRIMARY SODIUM PUMP CONDENSER 3.4 KN/m*

REACTOHVESSa 400"C REACTOR COHE 3 H.P. FEED HEATERS 6 L.P. FEED HEATERS

LIQUID METAL

SEPARATOR

2-PHASE LMMHD GENERATOR CHANNEL PRIMARY SODIUM PUMP ? 1 REGENERATIVE i I HEAT EXCHANGER

REACTOR VESSEL REACTOR CORE

Fig. 2. A schematic layout of a conventional liquid metal reactor power conversion system (top) vs. an all LMMHD Ericsson cycle system (bottom)

Space Reactors A Cs wet-vapor cycle, similar to that proposed for an HTGR topping cycle (Fig. 1b) was found most attractive for space nuclear power system. A preliminary study indicates*7) that the specific mass (kg/KWe) attainable with this LMMHD PCS can be significantly lower than that attainable with other types of PCSs proposed. V-28

CONCLUDING REMARKS The LMMHD EC technology possesses a collection of attractive features including: (1) Nearly isothermal expansion (offering infinite reheat without the need for reheaters), (2) Nearly isothermal compression (offering infinite intercooling without intercoolers), (3) No moving machinery and relatively simple components, and (4) Good thermodynamic matching with different energy sources. These features might improve the performance of nuclear power plants by simplifying their design, improving their safety and reliability, and increasing their efficiency. However, before a firm conclusion could be drawn on the attainable magnitude of these improvements, a thorough techno-economic feasibility study is necessary and the performance of different LMMHD PCS and components need be demonstrated.

REFERENCES 1. D. Elliot, "Two Fluid Magnetohydrodynamic Cycle for Nuclear Electric Power Conversion", ARS Journal, June 1962. 2. M. Petrick and H. Branover, "Liquid Metal MHD Power Generation - Its Evolution and Status", in Single and Multi-Phase Flows in an Electromagnetic Field, H. Branover, P.S. Lykoudis and M. Mond, Eds., Progress in Astronautics and Aeronautics, Vol. 100, p. 371 (1985). 3. H. Branover, "Liquid Metal MHD", Proc. 9th Int. Conf. on MHD Electrical Power Generation, Tsukuba, Ibaraki, Japan, Nov. 1986. 4. L. Blumenau et al., "Liquid Metal MHD Power Conversion Systems with Conventional and Nuclear Heat Sources", p. 33 in Proc. 24th Symp. on Eng. Aspects of MHD, Butte, Montana, June 1986. 5. L. Blumenau et al., "Liquid Metal MHD Energy Conversion in High-Temperature Gas Cooled Reactors", Proc. 4th Int. Conf. on Emerging Nuclear Energy Systems, Madrid, Spain, June 30 - July 4, 1986. 6. A. Barak et al., "Improvement Possibilities in Liquid-Metal Reactors Using Liquid- Metal MHD Energy Conversion", Proc. Int. Conf. on Fast Breeder Systems - Experience Gained and Path to Economical Power Generation, Richland, Washington, Sept. 13-17, 1987. 7. L. Blumenau et al., "Liquid-Metal MHD Power Conversion Systems for Space Electric Systems", Proc. 4th Symp. on Space Nuclear Power Systems, Albuquerque, N.M., 12-16th January, 1987. V-29

MHD HEAT-TRANSFER ENHANCEMENT POSSIBILITIES IN FUSION REACTORS

S. Sukoriansky, H. Branover and D. Klaiman Center for MHD Studies Ben-Gurion University of the Negev P.O.B. 653, Beer-Sheva 84106, Israel

and

E. Greenspan Atomic Energy Commission P.O. Box 7061, Tel-Aviv 61070, Israel

An extensive experimental study of turbulence in Liquid Metal (LM) flow recently established W that a magnetic field can enhance, as a result of inverse energy cascade, both the velocity fluctuations perpendicular to the field and the integral scale of turbulence. It was asserted^2* that the resulting strongly anisotropic turbulent field might significantly increase the heat-transfer coefficient and thus improve the design of self-cooled LM blankets for fusion reactors/2) The present work reports upon the first experiments designed for directly measuring the enhanced anisotropic turbulence effect on the heat- transfer coefficient (or Nusselt number^ and perfcmed at the Ben-Gurion University (BGU) Center for MHD Studies (CMHDS). It also reports upon preliminary results from the first measurements of enhanced turbulence and heat-transfer performed at the BGU CMHDS in channels made of conducting walls. The experiments are performed in a rectangular flow channel the test section of which is shown in the insert to Fig. 1. The top face of the test section can be electrically heated (uniformly) and the entire surface of the test section is thermally insulated. A turbulence inducing grid can be inserted at the entrance to the test section; it consists of 7mm diameter plastic bars placed 10 mm apart (between centers). Mercury is used for the LM. The Hg flow velocity and magnetic field strength domain covered in the measurements are 3.8 £ V < 17.7 cm/sec and 0 < B < 0.9 T. These correspond to the dimensionless parameters domain 4-103 < Re < 2-104 and 0 < Ha < 300 (based on the V-30

channel half-width in the field direction). The heat flux to the test section is controlled by varying the current through the heating element; it is limited to q £ 0.1 MW/m2. The

Nusselt number is deduced from the measurement of the temperature difference ATwb between the heated channel face (at the interface with the Hg, near the downstream end of

the heating plate) and the bulk LM, downstream from the heating elements; Nu <* q/ATwb. Thermocouple Heating Element / 70 cm

200

I 100-1

Ha/Re*10

Fig. 1. A schematic diagram of the flow channel used for the measurements and effect of magnetic field strength (proportional to Ha/Re) on the Nusselt number in a channel made of non-conducting walls for flow velocities of 7.6 cm/sec [A] and 15.1 cm/sec [•].

Figure 1 shows representative results obtained using a flow channel made of nonconducting walls and with the bars of the grid oriented parallel to the magnetic field direction. The decrease in Nu as B is increased from zero is due to the suppression of turbulence - an expected MHD effect on regular locally isotropic turbulence. However, the increase in Nu as B is increased beyond a certain value is due to the anisotropic turbulence enhancement caused by inverse energy cascade. 0«2) An indirect but convincing evidence V-31

that this, indeed, is the correct interpretation is given by tine, results from measurements carried out with the grid rotated 90° to its previous orientation; No Nu enhancement was observed at all. Figure 2 shows the Nu number deduced from the enhanced anisotropic turbulence experiments carried-out with B = 0.9T as a function of Reo - the Reynolds number based on the channel hydraulic diameter. Also shown in the figure are the corresponding Nu numbers obtained from experiments with normal shear turbulence (i.e., for a grid and magnetic field free channel) as well as from a couple of theoretical predictions corresponding to (a) laminar flow that is not fully developed (representing the conditions in the experiments, if the magnetic field was to completely suppress all turbulence), and (b) enhanced turbulence, using the simple model suggested in Ref. 2. Briefly, this model first calculates the turbulent thermal diffusivity coefficient for a "normal" locally isotropic turbulence. This coefficient is then scaled by an amplification factor which accounts for the enhancement by the magnetic field (via the inverse energy cascade), of both the integral scale of turbulence and the turbulence intensity. The amplification factor used for the present analysis is 25; it was selected to give good agreement with the Nu values measured in the high velocity range.

10*

Computation for enhanced turbulence

z 102-

Computation for laminar flow 10 10'

Rer

Fig. 2. Comparison of the Nusselt number measured with maximum magnetic field intensity (B=0.9T) [A] with the Nu number measured with no field and no grid [•] and with the Nu number calculated for a laminar and for enhanced turbulence flows. V-32

Comparing the results of Fig. 2 it is observed that: 1. The experimentally measured Nu number is significantly higher, for all flow velocities considered, than the Nu number predicted for laminar (not fully developed) flow (commonly taken to represent the flow conditions under the influence of a magnetic field). 2. The inverse energy cascade increases the Nu number also beyond the value attainable in conventional locally isotropic turbulent flows. 3. The turbulent thermal diffusiviry amplification factor found to match the measured Nu number at high velocities (25) is in surprisingly good agreement with the amplification factor estimated <2) from detailed measurements of the velocity field (20) A second series of measurements is being carried out using a flow channel made of thin tin plated copper walls (the plating is to assure good wetting of the walls, i.e., good conductance). Even though contemporary MHD theory predicts a strong turbulence and heat-transfer suppression in such a flow channel (when subjected to a magnetic field), the measurements reveal as strong an enhancement of the Nu number as with the non- conducting walls. Moreoever, this heat transfer enhancement was obtained even without the use of a flow perturbation inducing grid. A close investigation of the velocity field in the copper channel revealed that a strong anisotropic turbulence is being developed by inverse energy cascade and that the turbulence prevails in about half of the channel width in the vicinity of the channel walls. It is concluded that, if properly used, a magnetic field can significantly improve the heat transfer coefficient in flow channels made of both non-conducting as well as conducting walls. As illustrated in Ref. 2, this heat transfer enhancement can enable simplifying the design and improving the efficiency of self-cooled LM blankets.

REFERENCES 1. S. Sukoriansky and H. Branover, "Enhancement of Turbulence in a Magnetic Field," Proc. 5th Beer-Sheva Int. Seminar on MHD Flows and Turbulence, Jerusalem, March 2-6,1987. 2 H. Branover, E. Greenspan, S. Sukoriansky and G. Talmage, Fusion Technology, 1Q, 822, 1986. See also H. Branover et al., Trans. Israeli Nucl. Soc., H, 36 (1986). V-.J3

Direct Thermal Neutron Capture in Mg Isotopes

S. Kahane Oak Ridge National Laboratory, Oak Ridge, TN 37830, U.S.A., and

Nuclear Research Center, Negev, Beer-Sheva, ISRAEL

S. Raman

Oak Ridge National Laboratory, Oak Ridge, TN 37830, U.S.A.

J.E. Lynn

Los Alamos National Laboratory, Los Alamos, NM 87545, U.S.A.

In a series of papers on slow neutron capture by light nuelides, we have quantita- tively analyzed the importance of the direct capture mechanism in an optical model framework. In simple terms, this mechanism involves the transition of a single neu- tron, orbiting in an s-st&te in the overall potential field of the target nucleus, to a bound p-wave orbit. The basic theory of this mechanism was developed by Lane and Lynn [1] and by Cugnon and Mahaux [2]. This theory was employed in the analysis of experimental data in a fully quantitative way in a series of papers beginning with Raman et al. [3] on the S isotopes (this paper also contains a full resume of the theory) and following with Lynn et al. [4] on "Be and C isotopes and Kahane et al. [5] on the even Ca isotopes. In these papers, it was demonstrated that direct capture is indeed the predominant mechanism in these nuclides and that the remaining (usu- ally small) discrepancies between these quantitative estimates and the experimental data can be attribuied plausibly to contributions from the much more complicated and statistically oriented "compound nucleus" contributions from local compound nucleus levels. In the original Lane and Lynn work [1], a very simple formula was established for "pure" direct capture, namely "hard sphere" capture, in which the scattering wave function of the neutron was assumed to have a node at the radius of the nucleus; the nucleus, in turn, was assumed to be a hard sphere with negligible internal penetration of the neutron wave function permitted. This simple concept could be easily general- ized to account for actual cases of neutron scattering, for which the thermal neutron scattering length is generally different from the potential radius; this generalization V~34

is known as "channel" capture. This channel capture formula was shown in the Lane and Lynn paper to be successful in explaining semiquantitatively a considerable vol- ume of capture data extant at that time. Since then this formula has been shown by Mughabghab [6] and many others [see ref. 5 for a detailed list of references] to be approximately valid for many new and more precise data that have been published in the ensuing two decades. This success has even led to the use of this formula as a tool for determining nuclear quantities of interest (nuclear potential radii, for example). We have discussed elsewhere, both in general terms [7] and in specific terms [8], the somewhat uncritical use of the channel capture formula. We calculate the cross sections of the main primary El transitions from the data oa final state excitation energies, (d,p) spectroscopic strengths, and scattering lengths. We employ two different approaches: (a) a combination of a global optical model plus a valence contribution from local levels, which we call the [G+V] approach, and (b) a specialized optical model [S] approach in which the optical model parameters are chosen to reproduce the scattering length of the particular nudide in question. In both approaches, we vary the real well depth of the optical potential in order to reproduce the binding energy of the final state. The cross sections calculated by these two approaches differ at most by 6%, thus reinforcing our confidence in the methods of calculation. For the nuclides that we have studied (8Be, 12C, 13C, 32S, 33S, 34S, 40Ca, and 44Ca), the calculated cross sections for most of the primary transitions are in good agreement with the data. We attribute any differences to a compound nucleus component in the capture amplitude from the tails of nearby resonance states. The term "compound nucleus radiative amplitude" is used in our work as a generic term for mechanisms involving more general features of the wave functions than the simple projections of neutron motion in the field of the unexcited core of the target. We can attempt to assess the plausibility of this compound nucleus hypothesis in the following way. From the theoretical value of direct capture cross section calculated by either the [G+V] or the [S] method and from the experimental value, we deduce the compound nucleus capture cross section using

(1)

and remove from this quantity the factor E*. From the average value of o-7|Cjv/^ for each case, we deduce a quantity proportional to the compound nucleus radiation V-35

width from the relation [see eq. (12) of ref. 5]

(2) 2itRRl?e where k is the neutron wave number and RR'fc is the deviation of the actual scattering length from the potential scattering length. This quantity can then be compared with the Cameron semi-empirical relation- ship [8] deduced from a wide range of neutron resonance radiation widths:

(3)

The resonsmce energy Ey (which can be positive or negative) can be expected to be of the ordeir of or rather smaller than the level spacing D. Therefore, we expect the quantity on the left-hand side of eq. (2) to be about equal to or somewhat larger than the quantity on the right-hand side of eq. (3).

Table 1: Direct capture cross sections for primary El transitions, calculated with specialized optical model parameters, are compared with experimenal values. The compound nucleus contributions are deduced using eq. (1).

Primary (fy(exp) EAMeV) mb mb mb (A) J4Mg(n,7) reaction 3.918 17 36 3.5 3.054 5 9 0.6 2.610 0.2 0.1 0.02 (B) 35Mg(n,7) reaction 4.217 14 12 0.08 3.832 16 35 3.7 3.744 12 12 0.0 3.551 3 9 1.6 (C) a8Mg(n,7) reaction 2.884 40 24 2.0 1.617 10 5 0.9 0.535 0.3 0.01 0.2 V-36

Before testing this conjecture, we present in table 1 the results of some recently completed optical model calculations of the direct capture cross sections for the Mg isotopes. When we combine this with our previously published results, we are in a position to assess the overall situation concerning the compound nucleus contribu- tions for the nuclides that we have studied. The results are summarized in table 2. It is apparent that the qualitative kind of agreement anticipated in the preceding para- graph is indeed found for nearly all nuclides in that table. As an explanation for the abnormally low value in the case of l3C, we note that only one transition is included in the average and its radiation width is subject to Porter-Thomas fluctuations.

Table 2: Summary of compound nucleus contributions

Our Result Cameron

Number of Ex D Nucleus transitions (MeV~3) (MeV~a) "Be 3 -0.5 x 10-9 1.4 x 10-9 »c 2 -1.3 x 10-e 1.7x10-° 13C 1 -0.01 x 1G-" 1.8 x 10-9 "Mg 3 -13 x 10-B 2.7 x 10-" "Mg 4 19 x 10-B 2.8 x 10-9 a8Mg 3 540 x 10"° 2.9 x lO"9 s»s 7 40 x 10~9 3.3 x 10-9 9 33g 12 -23 x 10-° 3.4 xlO" 9 9 34S 5 53 x 10- 3.5 x 10- 40Ca 6 -6.4 x 10-9 3.8 x 10-B 44Ca 10 -1.8 x 10-9 4.1 x 10-9

In only one case - that of s8Mg - does it appear that the quantity listed in col- umn 3 of table 2 is excessively greater than the Cameron estimate listed in column 4. Furthermore, the compound nucleus effects deduced for all three transitions are sep- arately much greater than the Cameron value, suggesting that there is a mechanism other than compound nucleus capture operating here and that this mechanism could well be anticorrelated with the direct capture mechanism. We note that, unlike the other nuclides listed in table 2, the Mg isotopes are deformed or deformable targets. V-37

This fact gives rise to the possibility of coupling in a correlated manner certain in- elastic channels to the elastic scattering of the incident neutron. A radiative capture component will be associated with these coupled channel wave functions and this component will affect the calculated capture cross section. If it reduces the overall direct capture cross section, it might be unnecessary to hypothesize a large compound nucleus contribution. We are currently carrying out detailed calculations to test this idea.

References 1. Lane A. M. and Lynn J. E., I960 Nucl. Phys. 17 563-585; 1960 Nucl. Phys. 17, 586-608.

2. Cugnon J. and Mahaux C, 1975 Ann. Phys. (New York) 94 128-183.

3. Raman S., Carlton R. F., Wells J. C, Jurney £. T. and Lynn J. E., 1985 Phys. Rev. C 32 18-69.

4. Lynn J. E., Kahane S. and Raman S., 1987 Phys. Rev. C 35 26-36.

5. Kahane S., Lynn J. E. and Raman S., 1987 Phys. Rev. C 36 533-542.

6. Mughabghab S. F., 1979 Phys. Lett. 81B 93-97.

7. Raman S. and Lynn J. E., 1986 Proceedings Fourth International Symposium on Neutron-Induced Reactions, Smolenice ed. J. Kristiak and E. Betak (Dodrecht: Reidel) pp 253-274.

8. Raman S. and Lynn J. E., 1986 Phys. Rev. Lett. 56 398. 9. Cameron A. G. W., 1959 Can. J. Phys. 37 322-333. V-38

PRESSURE VESSEL FLUENCE REDUCTION POSSIBILITIES USING EFFECTIVE SHIELDING REFLECTORS

Z. Shayer and E. Greenspan

Atomic Energy Commission P.O.Box 7061, Tel-Aviv 61070, Israel

Pressure vessel (PV) embrittlement due to radiation damage caused by fast neutron fluence may be one of the leading factors in limiting the lifetime of PWRs. Of the approaches proposed for lowering of the flux of neutrons reaching the PV<*>, the insertion of neutron attenuating materials (such as stainless steel rods) in-between the core periphery and the PV is the most effective. Investigating the potential for ex-core flux attenuation, Tran and Turner*2* examined the effectiveness of different materials inserted (displacing water) between the core baffle and core barrel and between the core barrel and the PV of PWRs. The materials considered included heavy metals, metalic oxides and hydrides. They found<2) that replacement of water by stainless steel and other heavy metals (or oxides) increases rather than attenuates the fast flux level at the PV. Metallic hydrides, on the other hand, are very effective fast flux attenuators. Canning the hydrides in stainless steel degrades the hydrides attenuation effectiveness. However, Tran and Turner did not specify what hydrides offer the best attenuation of the fast neutrons, whether a combination of a heavy meterial and an hydride can offer a better attenuation than a hydride only and if so, what is the best heavy material-hydride combination. The purpose of the present Note is to clarify thess issues.

The investigation is done by considering the one-dimensional Light Water Reactor (LWR) recommended for shielding benchmark*3>. The reactor consists of an homogenous core, 169.59cm in radius, followed by a 1.86cm thick SS core baffle, 16.51cm thick water gap, 5.72cm thick SS core barrel, 26.03cm thick water gap and a 21.91cm thick low carbon steel pressure vessel.

The investigation of the effectiveness of the different shielding materials is carried out in two stages. For the first V-39

stage the fission source distribution is assumed to be fixed - that recommended for the shielding benchmark problem<3>. The flgure-of-merit used for this study is the flux of neutrons the energy of which exceeds IMaV integrated over the PV volume. This flux is calculated by solving, with ANISN, the transport equation using the 19 high energy groups and corresponding 4 cross-sections of the BUGLE-80 cross-section library'' '. A P3-So approximation is used for these calculations.

The second stage considers only the material combinations identified promising and accounts for the reflector composition effect on the core fission density distribution (and, through it, on the fast neutron flux at the PV). This is calculated with ANISN using the Po-Sa approximation with a 4-group structure in common use for LVR diffusion eigenvalue calculations. The 4-group cross-sections are generated by the WIMS code system using the WIMS cross section library. The 4-group cross-sections for Ti and V, not available in the VIMS cross-section library, were derived by collapsing the BU6LE-80 library into 4-group cross-sections, using a 47 group ANISN calculated representative neutron flux distribution.

All the metalic hydrides considered are assumed to be at 90% of their theoretical density (at room temperature), while the tungsten density is taken to be 95% of its theoretical limit. At least 10% of the volume used for the special neutron attenuating materials is preserved for water (to provide for adequate cooling). Being part of the primary cooling system, the water are assumed to be borated. All shielding-reflector materias used are of a uniform composition.

Fig.la compares the attenuation - ability of the PV fast neutron flux of SS versus H20 and, indirectly of TiH2 versus H20 in the 16cm gap between the core baffle and core barrel. It is observed that whereas the substitution of SS for H20 has but a relatively small (but positive!) effect on the attenuation of the fast neutron fluence, TiH2 is a significantly more effective attenuator than either H20 or SS. In fact, whereas the maximum attenuation of a SS-H2O reflector is obtained with a SS:HaO volume ratio of about 1:1, any amount of SS detracts from the attenuation effectiveness of TiH2. The TiH2 was found to offer also a significantly better attenuation than ZrH2, but comparable to UH3.

Fig.lb compares the attenuation-ability of tungsten with that of stainless-steel. It is observed that W is significantly more effective (per unit volume) than fS. Moreover, a combination of some V and TiH2 offers a better attenuation than either W or TiH2 by themselves! The optimal mix consists of about 15v/o W and 75v/o TiH2 (the other lOv/o are taken by H20). V-40

0 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 1.0 SS Volume fraction T1H2 Volume fraction

Fig. 1 Reduction factor in fast neutron flux at pressure-vessel resulting from the insertion of different combinations of (a) SS-H2O and SS-TiH2 and (b) TiHa-W and TiH2-SS into 16 cm water gap between the core baffle and core barrel. The V content in this optimal mix is significantly lower than in s 6 the W-TiH2 shield optimized against fusion neutrons« ' > (in which W constitutes about BOX of the shield volume). This reflects the difference in the average energy of fusion and fission born neutrons.

Table 1 compares the fast neutron flux attenuation - ability of the different. shielding-reflector material combinations identified promising, when taking into account the effect of the shielding-reflector on the core power distribution. It is concluded that the best attenuation is offered by the TiH2(75v/o) - W(15v/o) combination; it is 5 times better than the attenuation-ability of the SS-H2O shielding reflector being offered for PWRs<7>! The effect of the shielding-reflector on the core power average-to-maximum power ratio and reactivity is negligible.

REFERENCES

1. D. Franklin and T. Marston, "Investigating the Flux Reduction Option in Reactor Vessel Integrity", Electric Power Research Institute Report NP-3110-SR (1983). V-41

Table 1. Reduction in Fast Neutron Flux at the Pressure Vessel by Different Shieldlng-Reflector Materials Located Between the Core Baffle and Barrel. Reference is Borated Water Reflector.

Shielding-Reflector Fast Neutron Material Flux Reduction Factor

Borated H20 1.00

55v/o SS + A5v/o H20 1.43

45v/o V + 45v/o ZrH2 + lOv/o H20 5.41

15v/o W + 75v/o TiH2 + lOv/o H20 7.92

90v/o UH3 + lOv/o H20 6.15

2. K.C. Tran and R.L. Turner, Trans. Am. Nucl. Soc. 47. 379 (1984).

3. J. Celnik, "Specifications for One-Dimensional Light Water Reactor (LWR) Shielding Benchmark," American Nuclear Socity Standards Committee 6.2.1 (August 1979).

4. ANS 6.1.2 Working Group on Multigroup Cross Sections, "BUGLE-80 - Coupled 47 Neutron, 20 Gamma-Ray, P3 Cross Section Library for LWR Shielding Calculations", ORNL Radiation Shielding Information Center Data Library Collection DLC-75.

5. D. Gilai, E. Greenspan, P. Levin and Y. Kami, "Optimal Shields for Fusion Reactors" Proc. ANS Topical Meeting on Reactor Physics and Shielding. Chicago, IL. Sept. 17, 1984. 6. E. Greenspan, P. Levin and A. Kinrot, "Optimal Shield Concepts for Experimental Fusion Devices," Fusion Technology. 8. 1026 (1985). 7. R.P. Vijuk and S.N. Tower, "Reactor Coolant System Design of the Advanced (W) 600 MWe PVR", Proc. 22nd IECEC, Philadelphia, PA 10-14 Aug., 1987. V-42

AN EXAMPLE FOR THE POTENTIAL APPLICATIONS OF AMERICIUM - 242m AS A NUCLEAR FUEL

Yigal Ronen and Melvin J. Leibson Department of Nuclear Engineering Ben-Gurion University Beer-Sheva, Israel

The Americium-242 isotope has three isomers, one 242Am with a half life of 16.01 hours, the second 242mAm with a half life of 141 or 152 years and the third is 242fAm with a half life of 14.02 ms. The isomer 242mAm is characterized with the highest thermal fission cross section of all known isotopes. The thermal fission cross section of 242mAm is 6600 ± 300 barn. The thermal capture cross section of 242mAm is relatively low specifically, 1400 + 860 barn. The number of neutrons per thermal fission v is also high, 3.264 ± 0.024. All these facts (high thermal fission cross section, relatively low capture cross section, relatively high v and long half life) make the isomer 242mAm very attractive as a nuclear fuel from the nuclear data view point,

The isomer 242mAm is obtained by an (n/y) capture reaction with 24^Am which has a relatively high thermal capture cross section of 83.8 + 2.6 barn. The isotope 241 Am with a half life of 433 years is obtained from the beta decay of 241 Pu (half-life = 14.4 years). The amount of 24^Pu in the discharged fuel of power reactors is relatively high. Typically, about 9.5% of the plutonium is in the discharge from a PWR is 241 Pu. As a result, a 1000 MWe PWR reactor with a fuel discharged burnup of 32000 MWD/T produces about 31 kg of 24^Pu every year. Thus the current yearly production of 241Pu and thus of 241Am in the world is on the order of tons. The V-43

irradiation of 241 Am in a thermal (production) reactor will yL.d 242mAm with a steady state concentration of about 1% based on the ratio of the absorption cross sections of the two isotopes namely, 83.8/(6600 + 1400) = 0.0105. The separation

of 242m^m from 241 Am would probably be done isotopically. All these facts lead to the conclusion that 242mAm has a potential as a nuclear fuel, in particular where lower fuel weight (reactor size) is important. Space reactors are an example in which lower weight is important. As an example for use of Am fuel, a simple space reactor with sodium coolant was considered, with a fuel radius of 0.375 cm. Three types of fissile fuel 242m 239 235 were considered AmO2, PuC>2 and UO2 with a density of 17.25% of the theoretical density. The cladding (thickness = 0.04 cm) in all the cases was stainless steel. The geometry of the unit cells is hexagonal and the equivalent radius of the cell was 0.5229 cm, thus the V^/Vp = 0.72. The density of the sodium

used in the calculation was 0.85 g/cm3. The temperature of the fuel was 1050°K of the cladding 750°K and the coolant 650°K.

For all these fuels the k^ values as well as the bucklings for keff = 1.0 were calculated. From the bucklings the critical radius and critical volumes were calculated assuming spherical geometry. The results are summarized in Table 1. The ratio between the critical mass of plutonium to Americium is about 2 and the ratio of the critical mass of uranium to Americium is about 3.4, which indicates its suitability for space reactors. V-44

Table 1: Ko AND CRITICAL VALUES FOR

242mAmo2> 239Puo2 AND 235UO2 FUELS (Na-coolant)

2 2 Fuel k^ B [cm- ] RQ^ (cmj Vcrjt (liters)

3.2427 0.0384 16.03 17.26

2.5603 0.0239 20.32 35.15

2.0328 0.0157 24.09 58.60 V-45

Shielding Benchmarks by J. Celnik Stone § Webster Engineering Corporation, New York

The American Nuclear Society Standards Committee 6.2.1 is engaged in the documentation of problems, and their solu- tion, in areas of radiation transport. The primary objective of this effort is to test computational methods used within the general shielding community. Although purely computa- tional problems are also included, the ultimate aim is to have comparisons with experimental data.

Dissemination of benchmarks will, it is hoped, accomp- lish several goals:

1) Focus attention on problems whose solutions represent the state-of-the-art for representative transport problems of generic interest to the shielding community?

2) Specification of "standard" problems will make compari- sons of different computational methods, including use of approximate vs. "exact" computer codes, more meaningful;

3) Comparison with experimental data may suggest require- ments for improvements in computer codes and/or asso- ciated data sets; V-46

4) Test reliability of "new" methods as they are intro- duced for the solution of specific problems;

5) Verify ability of-user to apply a given computational method; and

6) Verify computer programs converted to a different computer type or facility.

We note that much of the above may be considered to be part of a good Quality Assurance program. Computer calcula- tions are used to set operating limits and design parameters, both for safety and non-safety related applications. Often these calculations have no relevant experimental data base. An incorrect calculation could result in either undue con- servatism with its- corresponding imposition of unneeded, costly, and cuirbersome restrictions, or it nay adversely affect a safetv-related item with its subsequent conse- quences. Computer answers, which later turn out to be incorrect, cause a general loss of credibility and effective- ness. A quality assurance (QA) plan is required to assure that computer calculational failures are reduced to an acceptably low level. Part of the QA program is based on reliable benchmarks. V-47

Thus, in the ANS Standard on Skyshine calculations it says: "Comparisons SHALL be made between the results of reference problem solutions provided by the methods employed in a specific power plant analysis and the results of the refe- rence calculations presented in Section 5.6.2. The relative agreement with the reference calculation results should be clearly indicated in the description of the calculational methods used."

Theory-experiment comparisons are intended to test computational methods and computer codes rather than nuclear data, although it may be difficult to disassociate the two factors. This problem is mitigated by selection of data sets in general use by the shielding community and/or recom- mended by other Standards working groups.. In this regard we are, within ANS 6.2.1, correlating our efforts with ANS 6.1 Standards Committee on Cross Section Data Sets. In all cases it is required that all relevant data be fully refe- renced so that comparisons with other computational methods be possible, within the restrictions of the method. In addition, selection of calculational parameters, e.g., order of Legendre expansion, angular quadrature, mesh interval spacing, etc. should be fully documented. V-48

The Committee is presently engaged in several benchmark areas, but is open to suggestions from the shielding com- munity at large. Present interest is focused on three specific transport problem areas:

1) Neutron and gamma ray flux results for a typical PWR system;

2) Gamma ray transport applicable to architect-engineer design of a radwaste facility;

3) 14 MeV neutron transport, applicable to fusion design problems.

For the PWR neutronics problem, several one-dimensional specifications are available including those of ESIS Table 1 presents the dose rate results outside the primary concrete shield from several institutions. The relatively large differences for an integral value suggests the need for computational benchmarks to clarify the causes for such discrepancies.

In addition to the ESIS PWR problem, the ANS 6.2.1 Shielding Benchmarks Committee has drafted specifications for a similar problem. Several members are presently sub- mitting solutions, comparing several cross section data sets. V-4 9

It is our hope to avail ourselves, and correlate our work with other benchmark efforts. These efforts include the work done by CSEWG, as just reported by Dick Maerker, European and Japanese benchmarks as well as current efforts directed toward the solution of specific problem areas. Such efforts include the Computational Blind Test on the Pool Critical Assembly Pressure Vessel Facility sponsored by the Nuclear Regulatory Commission. "This test is expected to provide an indication of the reliability of reactor pressure vessel calculations in general, and provide insight into the strength and weaknesses of particular calculational procedures chosen by the test participants." A similar goal is being sought by the NRC in its contract to the MAGI company, as well as work being done by SAI for EPRI.

It is our intention to follow the one-dimensional calculations with 2-D, and ultimately 3-D representation. Such geometric modelling will be used for reactor cavity streaming and penetration analyses.

For an evaluation of reactor cavity streaming tech- niques, the experiment reported in JAERI-M 7799 is being calculated. The Japanese have reported both experimental and analytical results, based on two-dimensional discrete ordinates, Monte Carlo and coupled calculations for neutron V-50

streaming through an annular air void simulating the reactor cavity gap. The experiments were performed in the JRR-4 swimming pool reactor.

The Committee is also benchmarking problems in the area of gamma ray tranport, specifically as they relate to architect-engineers. These include bulk shielding analysis of a cylindrical source tank as well as verification of a simple labyrinth design. Computer codes range from point- kernel (QAD and KAP-V), to one~ and two-dimensional discrete- ordinates (ANISN and DOT), to the use of Monte Carlo (MORSE), Some preliminary results are shown.

It is also the intent of the Committee to extract suitable shielding benchmarks from other sources, so that they may be centralized with a common format, and within a central location. This will, it is hoped, provide a unique place for people seeking benchmarks. Such sources include other Standards publications including ANS 6.6 on Skyshine from Nuclear Power Plants.

The ideal situation would be to have a number of bench- mark problems, of progressively increasing complexity, in a variety of areas of general shielding interest, all well- documented with accurate and complete experimental d,«ta. V-51

Since the desirability and need for benchmarks is recognized by all, the question of how to increase their number while decreasing the time required to obtain meaning- ful results, should be addressed. There already exists among individuals and several institutions, problems and their solution which could form the cornerstone of important

benchmarks. Mutual personal obligation and increased co- operation is required. Adequate funding and staffing for the solution of generic benchmark problems is a prerequisite before one can hope to obtain reasonable solutions for complex, multi-dimensional transport problems.

It is the hope of the author, as Chairman of the Shielding Benchmarks Standards Committee and as a member of the general shielding community, that increased awareness of benchmarks will replace the randomness of our present efforts with a concerted, united effort.

May I add that if anyone is interested in our benchmark efforts, please feel free to see me, or Dave Trubey (Chair- man of ANS 6), for further information. Should anyone be interested in setting up solutions to specific benchmark problems, we would make every attempt to have them run at no computer cost to them. Let us hear from you, in anyway that you may be able to cr tribute.

Thank You! V-52

Table 1: Comparison of Total Dose Rate Outside Primary

Concrete Shield for 1-D Test PWR Problem

Dose Rate Institution (mRem/hour) AEE, Winfrith 14.8

CEA, Saclay 41.1

CEE, Mol • 13.0

(Sabine Result 5.5)

EURATOM, EL4 Library 10.1

EURATOM, EURLIB3 Library ^.0

IKE, Stuttgart 25.5 KWU, Erlangen 8.3 Burns and Roe 20.0

Average 20.0*

•excluding SABINE result. V-53

;RNS 6.2 LNR BENCHMflRK PROBLEM I TOTRL DOSE RRTE VS. DISTRNCC

: '.. :; • ••••"•?• LEGEND / • • . *-DLC-41B/VITRMIN C 171N-369 LIBRRRr \ •» - DLC-47/BUGLE 45N-16G LIBRRRY

.ISO 170 ... DISTAKCe TROtt SOURCCCCH. J V-54

RNS 6.2 LWR BENCHMRRK PROBLEM I OUTSIDE P'.V. NEUTRON SPECTRUM (EPHHERMFL REGION)

LEGEND DLC-41B/VITRMIN C 171N~36G LIBRRRY *- DLC-47/BUGLE 45N-16B LIBRflRY

C3 : x t3

or to o

* 1 l'I . ti Ii i', _' • ' I ' I't'l II i; ' I ' i ' I'd I •:{ 301 10 10" 10' 10s It RNS 6.2 LNR BENCHMRRK PROBLEM I ' "b_ • ...

• LEGEND o »- DLC-418/VlTflMIN C 17IN-36G LIBRflRY • • - 0LC-47/BUGLE'45N-I6G LIBRflRY

ca -

LJ* : < i— in

cc ~": LJ Q, 5 in

i ^ 2- < tj 2

r • i n J t

»

4 ib* ibr ib* : ENERGYIEV.J

Figure 1. Results of Calculations of the Fast Neutron Spectrum Leaving the Pressure Vessel of the LWR Model Using the BUGLE and VITAMIN-C Cross Section i j \ • V-56

JAER1-M 7799

'10/ 17.5 14.6 1 On unit of cm) Fig. 3 Calculational nodel for H-II-2 problem 140.0 140.0

WATER WATER s.s VOID S.S WATER WATER 4tl 5111512 513 414

30.0 30.0 WATER WATER CORE I I (0.0.0) ?EFL£CTOr WATER s!s VOID S.S WATER CORE Ml 211 J2I2 411 |4I2|4!3 511 J5I2 611 513 414 -30.0 -30.0 -21.06 21.06

WATER WATER s.b IVOID S.S WATER

-40.0 1 -40.0 -21.06 . 44.26 84.91 128.01 153.01 -76.0 -44.26 44.26 76.0 45.76 113.41 133.01

Fig, 4 N-II-2 model for MORSE calculation (in units.of on) N Reoction id* 1 1 1 1 —1 1 1 1 ' lls 58 In{n.n'J 7^—§ ^. N-II-2 (DOT-nt) Ni(n.p) 9nM (n. p ) 6''2n(n.pJ ?rfltfn.p)

10* 10= S6fo(o.p>-

5 Experiment i) 100x108 cncshss o — Col. (100x108 meshes) ? •

D

O F^^^%i ^A o r g t 1(j2 or I 1 I I V..*j t t 1 ^e

II I _L I 10"" -20 0 20 40 60 60 100 120 140 -20 0 20 40 60 80 100 120 Axial distonce from core center (cm) Axiol distance from core center (cm) Fig.4.6 Comparison of DOT-III reaction rates with Fig.U.5 Comparison of DOT-III reaction rates of CO measured values along the dotted line Ni(n,p) with measured values along the given in Fig.1.2 of N-II-2 dotted line given in Fig.1.2 of N-II-2 r?z Exp. Col. N-II-2 (TWOTRAN-II ) 5 Experiment (R*84.91cm) I A "Ni(n!p> Colculofion (R*82.47cm) _ o **Zn(n.p) Colculotion [R= 89.91 cm) • "AUn.pJ O •»*—• MQin.p) • —*— 96Fo(n.pJ V "AUn.ol §

N-D-2 (TWOTRAN-tt) in 3B CD Ni(n.p)

S

id"

Exp. (R-121.2 cm) J_ -20 0 20 40 60 80 100" 120 140 Col. (R= 117.305 cm) Axial distance from core center (cm)

10' Fig.4.11 Comparison of TWOTRAN-II reaction rates -20 0 20 40 60 80 '00 Axiol distance from core center (cm) of 58Ni(n,p) with measured values along Fig.4.12 Comparison of TWOTRAN-II reaction the dotted line given in Fig.1,2 of N-II-2 rates with measured values along the dotted line given in Fig.1.2 of N-II-2 56" » i • r f j N-E-2 (MORSE) Exp. Col. ReocHon • -o- "5In(n.n' 59 A -a- Ni(n.p):

'O N-H-2 (MORSE) 58Nt(n.p) A A VI • Experiment ?0 o o Cotculofion < c o S id* J

I I t 1 1 1 1 -40 -20 0 20 40 60 80 100 J2O Axial distance from core center height (cm) I . I -40 -20 0 20 40' 60 .80 100 Fig.4.21 Comparison of MORSE reaction rates of Axial distance from core center (cm) CO Ni(n,p) with measured values along Fig.A.22 Comparison of MORSE reaction rates with the dotted line given in Fig.1.2 of N-II-2 measured values along the dotted line in Fig.1.2 of N-II-2 1 1 1 1 1 ; : ' 1 ' I • I : 1 ' 1 1 1 I 1 1 TTT T I N-E-2 (MORSE-PALLAS) 5l3Ni(n.p) : N-n-2 IMORSE-PflLLAS)

Exp. Col. Reocf ion 22 • Experiment 10" 1O :- • I "5ln(n.n')1 5 MORSE col. : ' I T I A $ "NUn .p) ; - • I 27 2 Al(n '.a)' TTTTT T .6 V 10 T • I (U • 0> I \ - i A I i529 — in vt '; I, o I : a B \ I I i 1 c c : i I o .2 26 i t5 IO a I a a a) CC 5 i V • I I I 1 I i n J0ZT • 1 X x I :- - t ' I x 1 1 - - ! * - ze I . 1 1 i . 1 1 1 . 1 1 !0& ,1,1 , i • I i I •, I tn ! I -20 .0 20 40 60 80 100 120 -20 0 20 40 60 80 100 Axiol distance from core center height (cm) Axioi disfonpe from core center height (cm)

Fig.4.23 Reaction rates of 58Ni(r.,p) by MORSE-PALLAS Fig.4.24 Reaction rates by MORSE-PALLAS roupled coupled calculations along the dotted line calculations along ths dotted line given in Fig.1.2 of N-II-2 given in Fig.1.2 of N-II-2 28 -27

.26 Z5 .24 1 D3 Z3 JL'D2

Z2 •

ELEVy\TION PLAN •••

SCALE 1 CM = 100 CM

The computational results are given below:

DOSE RATE (mR/hr) CODE D2 D3

KftP-V 8.01 + 04 2.46

ANISN 1.06 4- 05 3.42

ANISN/KAP-V 1.32 1.39 V-63

D4

XG

X2 X8 X8 •r».|: XI X5 D5 D6 L. . X7 Y2

CODE D4 PLAN DOT 859 MORSE 959 DOT/MORSE 0.90

ELEVATION

SCALE 1 CM ' CO CM- V-64

Neutron Flux at Outer Surface of Pressure Vessel

Energy Band (eV) CASK BUGLE CASK/BUGLE

15.0 (+6) - 10.0 (+6) 3.56 + 6 7.86 + 6 0.45 10.0 (+6) - 4.06 (+6) 5.52 + 7 4.13 + 7 • 1.34 4.06 (+6) ~ 2.35 (+6) 8.07 + 7 6.70 + 7 1.20 2.35 (+6) - 5.50 (+5) 1.54 * 9 1.77 + 9 0.87 5.50 (+5) - 1.11 (+5) 4.22 + 9 4.04 + 9 1.04 1.11 (+5) - 3.35 (+3) 3.08 + 9 3.71 + 9 0.83 3.35 (+3) - 1.01 (+2) 1.30 + 9 1.22 + 9 1.06 1.01 (+2) - 4.14 (-1) 1.19 + 9 9.84 + 8 1.21 Thermal 8.91 + 7 1.02 + 8 0.87 1.156 + 10 1.194 + 10 0.97

References

"Reactor Shielding Benchmark No. 2 for a PWR" NEACRP-L-151 (1976). ' V-65

The Collapsed Cross-section in Different Eigenvalue-typs Formulation of Boltzman Equation

Z. Shayer Israel Atomic Energy Commission,Licensing Division P.O.Box 7061, Tel-Aviv,Israel

The cross-section collapsing process is one of the basic stages to carry out a global reactor core calculations.There are few possible schemes of weighting cross-sections,but the most wide-spread used is that of weighting flux shape,such that reaction rates are unaltered by the averaging procedure,namely

2

c

where, G=collapsed energy group index (for simplicity the zone indices have been omitted)

The flux shape in Eq. 1 can be obtained from several eigenvalue-type formulation of the boltzman equation*1-*>

$n- l/n(

i.e.. if in one of the 1/n values ,n equal to k,/,or S than in all other 1/n values, n equal to unity.

The physical meaning and the properties of these eigenvalues were discussed by Cacuci et al.^^and the relative hierarchy of spectral indices and other physical parameters were also analysed in this paper. V-66

The main purpose of this note is to examine the effect of different eigenvalue-type flux shape on the collapsed cross-section.Namely what eigenvalue-type equation be used to calculate the collapsed cross-section so that the results are least sensitive to the changes which might have caused the system to deviate from criticality. The collapsed cross-section procedure can be written in terms of spectral indices as,

8 S <>•„ + 2 fn In 8-2 (3)

G B 1 + 2 In B-2

where the spectral index I* for energy group g is defined as I*= Jj^/gj1 for g=2..G,index 1 indecates the highest discrete energy group.

It is observed from Eq. 3, that alteration in the neutron spectrum will induce alteration in the collapsed cross-section,such that the following relationship holds

dln 1 + 2 ] 8-2

It follows from Eq. 4 that

d

Thus in view of the inequalities (see Cacuci et al.*) it follows from Eq. 5, that for a uniform subcritical system the following proportionality relationships holds.

Ic: (6)

A*>Ar>A6=Ac (7) and

L|>

where, the symbols Ic,Ac,Le describe the critical spectral index, absorption rate and leakage rate.

The validy of this qualitative analysis about the effect of spectral Indices on the collapsed cross-sections has been confirmed by calculations (using WIMS/D4 and ANISN codes) of a simple sphere problem with critical radius of 61.01 cm (f=k=5=l) which consist of water and uranium in the homegenous mixture. From these results the advantages of using S-eigenvalue equation for a collapsing process in the uniform system is evident. However for nonuniform system which consist of different types of zones the analysis is more complicated and is strongly dependent on the spectrum indices, magntiude and the zone size.

REFENECES

1. Ronen, Y., Shvarts, D, and Vagschal, J.J., Nucl. Sci. Eng. 60,97,1976 2. Ronen, Y., et al., Nucl. Sci. Eng., 62,80,1977 3. Velarde, G., Ahnert, C. and Aragones, J.M., Nucl. Sci. Eng. 66,284,1978 4. Cacuci, D. G., et al. Nucl. Sci. Eng. 81, 432-442, 1982 Reliability and Non Destructive Testing

Chairmen - A. Notea and Z. Alfassi VI-1 DYNAMIC CHARACTER OF FAILURE STATE IN DAMAGE ACCUMULATION PROCESSES

Dov Ingman, Department of Nuclear Engineering, Technics Israel Institute of Technology, Haifa, Israel

Leib Reznik, Nuclear Engineering Department Israel Electric Corporation Ltd, Haifa, Israel SUMMARY The process of damage accumulation and consequent failure of a component during its service life is characterized by two different aspects. The first is concerned with the mechanisms of material deterioration and loss of performance function. This aspect has won considerable attention by researchers and has been thoroughly represented by various semi empirical and phenomenological models on different levels of consideration.

The second aspect refers to the limiting condition of the component failure subject to the generalized failure criterion: applied stress achieves or exceeds the value of attained component strength. There is a multitude of interpretations of the generalized failure criterion, expressed in terms of shear and normal stresses, strain work etc. The component mean time to failure (MTTF), may be adequately predicted only by taking properly into account the dynamic character of both aspects. However, in the majority of works treating the damage accumulation processes, the dynamic features of the second asoect are neglected. The present work aims at effects due to the above dynamics. VI-2

Initially, an exact solution of the reliability evolution equation

t\• (N,t) _ ' T)r(N.t) Dj2 r^t) (

is obtained for an important class of steady-state loading conditions. The solution form for the reliability distribution function r(N,t):

U)

- ex_wnp ri - ,N-No-i 2 Ncr -Nt) Ti r(Nci r N 'n 1? v. L 4 Dt J exP t- D J (

is represented by a "bell-type" curve moving toward a critical damage state * < N with the drift velocity N, while widening due to diffusion, characterized by the parameter D. The component mean-time-to-failure (MTTF) Is obtained in the fomu

(3)

The solution (2) may be used as a Green function to analyze conditions of complex loading dynamics.

This approach leads ro prediction of observable effects of abrupt decrease in component reliability for the case of a step stress increment and effects of apparent ceasing of deterioration for the case of a step stress reduction. It is also shown that deviations from the Life-Fraction-Rule result from combined dynamics of damage accumulation/failure processes rather than from requirements of non-linearity of damage mechanisms with time. VI-3

SOFTWARE FOR THE TNA EXPLOSIVE DETECTION SYSTEM

T.Gozani, P.Shea, J.Adir

Science Applications International Corporation

Science Applications International Co. of Sunnyvale California has developed a "Thermal Neutron Analysis" system to detect nitrogen bearing explosives. The main features of the system ara a very low false alarm rate and a very high detection probability. The system has been developed for the FAA and is being used experimentally at US airports for detection of hidden explosives.

Suitcases are placed on a conveyer belt, where their weight and dimensions are electronically measured. They then travel to an irradiation area where they are exposed to slow neutrons. They become mildly radioactive. The emanating gamma rays are detected by a series of detectors which are placed around the suitcases. By analyzing the gamma spectra, a spatial picture of element concentrations inside the suitcase is obtained. The activity of the suitcase reaches natural background levels very rapidly.

Raw information is corrected for background and other reasons. Special features are amplified by using statistical models. Detected weight ratios are compared with known ratios of typical explosives and a decision is reached whether explosive compounds are present. VI-4

The system is modular, portable, shielded and easily accessed for maintenance. It is controlled by a "Digital" computer and has on line stabilization, diagnostics, jam recovery features and window reporting.

The sophisticated design of analog circuits, digital electronics, mechanical structure and computer software are the main reasons for the system's success. A simplified "Operator Panel" enables unskilled personnel to run the system. Whenever an alarm occurs, the bag is diverted and airport authorities are notified.

SOFTWARE SPECIFICATIONS

The general requirements for code development were :

1. The code should enable operation, testing, debugging and servicing the FAA system.

2. Operations should be reliable, and in accordance with the functional specifications.

3. Code should be modular and flexible.

4. Code should be well structured, with commer;ts, to guarantee clarity and maintainability.

5. Code should be written in FORTRAN 77. VI-5

6. A Hierarchy of operator levels should be available.

7. System should have real time diagnostics.

8. Display windows should show real time data frames.

9. Software development and documentation according to industry

standards.

TASKS

The computer controls all the operations of the system. It is responsible for data acquisition, data analysis and communications with all the peripheral devices. About 30 tasks run concurrently to perform all those duties.

Tasks communicate via global common and system flags. Each task is assigned proper priority.

This real time multi task system is running under the VMS operating system.

The main tasks are :

1. Initialization - Startup procedures

2. Detector_Read - Continuous data acquisition from ring of

detectors.

3. Data_Analysis - Raw data processing.

4. Bag__Processing - Analyzing bag data. VI-6

5. Decision - Material composition and decision analysis.

6. Peripherals - Continuous data acquisition and control of peripheral devices.

7. Diagnostics! - System diagnostics for

immediate response.

8. Diagnostics2 - Longer term system diagnostics.

9. Stabilization - Stabilizing detector perations.

10. Reporting - On-line data displays.

SHORT TERM DIAGNOSTICS

The following is a representative list of the tests which are done very frequently :

1. Check that all the tasks are running.

Respawn tasks that got lost.

2. Check that the "Weigh Scale" is working=

Set proper switches if not.

3. Check the performance of the "Dimensions" measurement

system. Turn on "Service Light" if all fail.

4. Check "Belt Speed" and update system. VI-7 5. Check "Temperatures". Turn on "Service Light" if out of bounds.

6. Check available free "Disk Space". Turn on "Service Light" if too little is left.

LONG RANGE DIAGNOSTICS

More extensive and elaborate testing is done on line, at greater time intervals.

In general, hardware is checked for proper functioning. On failures warnings are issued and hardware units may be bypassed or disconnected. All the events are saved on a log file. The setting of software flags may change system algorithms. Recovered hardware modules are automatically returned to service.

Typical items checked are : Detector Voltage

Detector Gain Detector Offset

Gamma Rate Neutron Flux Conveyer Belt speed Peripheral Devices (weigh scale, light screens, etc.). VI-8 REPORTING and CONTROLS

The system supplies continuous informative and operationaj datai arranged in screen windows. '

Typical information describes :

Bags - weights and dimensions. Bags - nuclear data. Bag - last out. Bags - status f?ags. Detectors - voltages, offsets, gains. Detectors - status, fail mode. Peripherals - current data, status.

Decision - analyzed data, features, statistical models. Operations - logfi.la of events.

Operations - system parameters.

Operations - statistics. Operations - operator help menus.

Operator controls (SAIC personnel only) : Screen - move (up/down/sideways). Data - refresh. Parameters - changes. Flags - set/reset. Detectors - connect/disconnect. Motors - start, stop, reverse. Sources - insert/withdraw. VI-9

RECONSTRUCTION OF RADIOGRAPHIC IMAGES BY MEANS OF AN ADAPTIVE POLYNOM FITTING

D. Ingman and Y. Merlis Department of Nuclear Engineering Technion, Haifa 32000, Israel

A real radiographic image usually differs strongly from the ideal one. This difference is a result of an interaction of X-radiation with matter and the additional noise due to the secondary scattering, electronic noise, granularity of the film, etc. [5]. In practice, these phenomena cause the image to appear distorted {usually blurred) and noisy.

There is a variety of models for the mathematical analysis of real images [1,33. This study is based on the diffusion model with additive white noise [3,4]. According to this model, the distortion is represented by the exponential LSF (Line Spread Function), i.e.

LSF(x) = 0.5Texp<-T|x| ) Q) (where T is a coefficient of distortion)

After convolution of the ideal image with the LSF of distortion, a white noise n(x) is added so that

g(x) = f (2) where f(x) is the ideal image (signal) g

It can be shown [63 that when the noise is neglectable, an ideal reconstruction of f(x) can be achieved applying Laplacian filtering.

f(x) = g(x)-l/TA2g"(x> <3) where g'Mx) stands for the second derivative of g(x).

In the presence of noise, however, this method of reconstruction requires a preliminary filtering due to the noise enhancement of the Laplacian filter. The problem is that filtering, which is usually achieved by some kind of a low-pass filter, causes a loss of information of the signal (especially at high frequencies). VI-10

Application of a fixed (noaadaptive) filter ignores the fact that the signal itself can vary strongly. In the typical radiographic picture, the shape of the signal can be arbitrary, and particularly, can differ strongly from one part of the image to another. This requires a very flexible filter that will adjust itself according to the behavior of the signal in the given area, i.e. will be a more narrow-banded low-pass filter while the signal changes slowly, and will increase its band width when the signal becomes "steep". This kind of filter is an adaptive filter.

Most of the existing adaptive algorithms, such as MMSE (Minimum Mean-Square Error) Gradient or LMS (Least Mean-Square) Stochastic Gradient, assume that the signal is stationary. Moreover, it is usually assumed that the autocorrelation of the signal (or noise) exists [2]. In radiographic images, these assumptions are not valid - the signal is neither stationary nor ergodic and can be uncorrelated, with exception to a small number of consequent samples, and the noise is uncorrelated with itself.

In the present work, a new adaptive algorithm of filtering and reconstruction is introduced. This algorithm uses only local data and thus is almost independent of the autocorrelation of the signal.

The basic idea is that at each point x we fit parabola to the samples g(x-N(x)), g(X-N(x)+l), g(x ),. .. .g(x+N(x)-l), g(x+N (2N(x)+l samples altogether)

Where N(x) denotes the number of neighbours on each side of the point x (the value 2N(x)+l is called "window size"), and is calculated adaptively.

A parabola is a second degree polynom. Thus for N(x)=l, it will pass through the 3 points (the point x itself and its two neighbours) exactly (because a parabola can be fitted to any 3 given points), thus resulting in an all-pass filter. For Nl, the parabola will pass between the points and will smooth, i.e. filter. The smoothing will be stronger as N(x) is bigger, and thus the resulting low-pass filter will be more narrow-banded. Therefore, by controlling the value of N(x), the band width of the filter can be changed.

We assume that the behaviour of g(x> at the point x+1 (namely, N(x+1>) can be predicted on the basis of the data we have at the point x. VI-11

The considerations for choosing N(x) are as follows:

1) As the noise level is higher, we'll want to smooth more, and thus will choose more neighbours (large N(x)).

2) As the blurring increases (T is smaller), the distorted function g(x) is smoother, and we can achieve better noise reduction (more narrow-banded filter) without any loss of high-frequency informa- tion (large N(x)).

3) As g{x) changes more rapidly, it will consist of higher frequencies and we'll want to use more wide-banded filter. The "steepness" of g(x) is characterized by the second derivative of the fitted parabola (we call it a(x)). As a(x> increases, the fitted parabola is more steep, and assuming that g(::) is more steep as well, we'll choose the smaller N(x).

In accordance with these considerations, we introduce the following adaptive procedure: cS/T

N(X+1) = +1 (4)

Where N(x+1) denotes half of the window size at the point x+1, S is a square root of the variance of the noise, thus

SA2 = E[n(x)A2] (5) (E is an average operator)

a(x) is the second derivative of the parabola fitted to f(x) and its N(x) neighbours on each side. c is a constant.

The reconstruction of the initial signal f(x) is achieved by substitution of au) into equation (3) instead of g''(x) [5].

We assume that the noise is stationary and therefore S is independent of x. The value of S represents the noise level.

From equation (4) follows that if S -»• 0 (low noise level), or T + » (low distortion), we'll get

N(x+1) -»• 1 VI-12

On the other hand, if a

Applying equation (4) at each point x, results in a sequence of low-pass filters, each one characterized by the respective N(x).

The constraint on equation <4) is that if NM, then we'll choose N

1) As it has already been mentioned, the correlation of the signal is usually small, and thus a large window can cause the usage of the data far from the point in question, and can yield misleading information.

2) As N(x) increases, the band width of the filter decreases, result- ing in a decrease of the slope of the fitted parabola. A small a(x) causes in its turn a raise in the value of N(x+1> at the next step and so the equation (4) diverges.

3) From equation (4) follows that the minimum window size is 1. Therefore, the maximum window size M determines the dynamic range of the sequence of the windows, and thus the continuity and the smoothness of this sequence are factors that have vital influence on the quality of the resulting image.

The value of M depends on the noise level (S) and the coefficient of distortion (T). Intuitively, as S increases, or T decreases, M will increase. There are more mathematical approaches to the estimation of M that yield very good results. Likewise, there is an analytical way to estimate c

From our experience, the correct choice of the two parameters of the process, c and M, yields much more satisfactory results than using other adaptive procedures (such as LMS or RLS-Recursive Least Square).

There are some additional improvements of the algorithm that can be suggested, such as: taking into consideration the first derivative of parabola, averaging of the sizes of the windows, additional filtering (for example, median), etc. VI-13

In our opinion, the introduced algorithm is a simple and effective way of extracting information from a noisy radiographic image.

References

1. GONZALEZ, R.C. and MINTZ, P., "Digital Image Processing"

2. HONIG, M.L. and MESSERSCHMITT, D.G., "Adaptive Filters" (Kluwer Academic Publishers, 1985).

3. PRATT, S., "Digital Image Processing", (Wiley, 1978).

4. SEGAL, E., NOTEA, A. and SEGAL, Y., "Dimensional Information Through Industrial Computerized Tomography", (Material Evaluation /40/11/1982>.

5. FRANKEL, R., "Factors Affecting Radiographic Image", (Radiological Health /B/1968).

6. BUSHLIN, Y., "Extraction of Quantitative Information from Noisy Radiograph", (Research thesis, Department of Nuclear Engineering, Technion, 1385). VI-14 ASSAY-SYSTEM DESIGN FOR BOREHOLE LOGGING OF URANIUM

Uzi Vulkan Applied Research and Exploration Dept. Soreo Nuclear Research Center Yavne> Israel

and

Yakov Ben-Haim Department of Nuclear Engineering Technion -- Israel Institute of Technology Haifa, Israel 32000

Assaying for uranium in stratified geological deposits is typically done by borehole logging with a Nal detector. Gamma rays emitted by U daughters are measured and their intensity is related to the U concentration as a function of depth in the borehole. The spatial distribution of the U within each uraniferous stratum is typically assumed to be homogeneous Cl]. In some occurrences of uranium it has been found that the spatial distribution of the U is far from homogeneous [2,3]. Rather, the U occurs as relatively large, concentrated, iso- lated nodules sparsely distributed in a non-uraniferous matrix. The standard borehole logging procedure is based on the assumption of spatial homogeneity. In the presence of large spatial uncertainty in the distribution of U, the stan- dard logging procedures may yield unreliable results. We are developing a multi-stage adaptive borehole logging procedure, whereby the position and duration of measurement along the borehole are determined on-line, in response to the measurements obtained. The adaptive approach enables practical optimization of the assay accuracy and speed, despite the lim- ited initial information on the geomorphology of the deposit-. An adaptive algorithm contains four components [4]: VI-15

1. A simplified MODEL of the geomorphology* containing mod- el parameters which approximately characterize the structure and content of the deposit. 2. A DESIGN ALGORITHM which specifies an optimal sequence of measurement positions and durations on the basis of the current estimates of the model parameters. 3. A MEASUREMENT SYSTEM, which implements the measurements. 4. An ESTIMATION ALGORITHM which determines optimal esti- mates of the model oarameters on the basis of the accu- mulated measurements. A major challenge in designing an adaptive-assay algorithm arises from the fact that even a simplified model of the geo- morphology contains several model parameters. "Conflicts of interest" often arise in formulating the design algorithm for assessing these Quantities: a measurement regime which is optimal for estimating one set of model parameters may be sub- stantially sub-optimal for estimating another set. This dif- ficulty can be ameliorated by dividing the model parameters into small subgroups, and then estimating the subgroups one after another, provided that estimation of each subgroup is independent of knowledge of parameters in subsequent sub- groups . This has been found to be possible for the assay of a sin- gle isolated uraniferous nodule. The simplified model assumes the nodule is spherical, and contains 5 model parameters! The mass of uranium CM), the soherical radius f.R], the distance from the sphere-center to the center of the borehole CD), the depth of the sphere-center (Z) and the intensity of the back- ground gamma radiation Clb). An experimental system has been built, enabling the study of uraniferous deposits characterized by a high degree of spa- tial uncertainty. The measurement system employs a standard logging head [CDI-2, Scintrex). The gamma ray intensity as a function of depth has been measured for several geomorpholo- gies which closely approximate the simplified model. It has been found that the intensity along the borehole axis can be accurately reproduced by a gaussian function. Furthermore, the VI-16

width of the gaussian deoends nearly linearly on the sphere- to-borehole distance CD), while the center of the gaussian determines the sDhere depth (2). Computer simulations confirm that the width and center of the distribution of radiation intensity along the borehole axis are in practice independent of the remaining model parameters CM and R). These results lead us to divide the model parameters into three subgroups, and to formulate a three-stage adaptive assay procedure. In the first stage the intensity of the background radiation Clb) is estimated while moving the detector downward in the borehole. During this stage a monitoring function scans the data, to detect the appearance of a uraniferous nodule. The second stage is initiated upon detecting a nodule, and while still lowering the detector in the borehole. In the second stage the model parameters Z and D are determined by fitting the measurements to a gaussian, thereby exploiting the independence of Z and D from the remaining parameters CM and R). The determination of Z and D in the second stage is done adaptively, and on the basis of the prior determination of Ib. The Rao-Cramer theorem is exploited to define the design algorithm [4], and the estimation algorithm is based on the concept of maximum likelihood [5]. The remaining parame- ters CM and R) are estimated in the third stage> while with- drawing the detector from the borehole. This stage is also an adaptive assay, and is based on the results of the previous two stages. Much work remains to be done, primarily to verify the over- all algorithm when it is confronted with a geomorphology which differs substantially from the simplified model.

References 1. IAEA, Borehole Logging For Uranium Exploration, Techni- cal Report 212, Vienna, 1982. 2. H.C.M. Whiteside, Uraniferous Precambrian Conglomerate of South Africa, in: Uranium Exploration Geology, IAEA-PL-391, (1970), pp49-74. VI-17

6. Baer, A. Xtamar, A. Zilberfarb, M. Rapaport and U. Vulkan, Subsurface Prospecting for Uranium in Makhtesh Ramon, Soreq Nuc. Res. Report ZD106/85 (1985). Y. Ben-Haim, The Assay of Spatially Random Material/ Reidel Pub. Co., Dordrecht, Holland, 1985. H.L. Van-Trees, Detection, Estimation and Modulation Theory, Part I, John Wiley, New York, 1968. VI-18 EPITHERMAL NEUTRON ACTIVATION ANALYSIS FOR DETERMINATION OF TRACE AMOUNTS OF GOLD IN ROCK SAMPLES FROM MAKHTESH RAMON.

E. Ne'eman*, N. Lavi**, A. Itamar and G. Baer***.

Institute for Environmental Health, M.O.H., Tel-Aviv University*, Soreq Nuclear Research Center, Yavne, Israel** and Survey of Israel, Jerusalem, Israel***.

ABSTRACT

Epithermal neutron activation analysis (ENAA), followed by high resolution gamma ray spectrometry, was applied to determine trace amounts of gold in the presence of rare earth elements (REE) from vein samples in the basaltic rocks of Makhtesh Ramon, located in Southern Israel (Tab. 1). Epithermal neutron irradiation has the advantage of enhancing the production of Eu-198 via the Au-197 (n,y) Au-198 reaction over that of interfering nuclides which have lower IQ/61 ratios. Since both Eu-152 and Au-198 emit gamma rays with close energies of 411.1 and 411.8 keV respectively, a correction has to be made in the case of the presence of a relatively high concentration of europium in the matrix. The contribution of Eu-152 (411.1 keV) to the 411.8 keV peak of Au-198 may be determined using multiple gamma peak ratios derived from europium standards and mixtures of gold and europium in different ratios, irradiated for different periods of time. The method is nondestructive, accurate, highly sensitive and may be routinely applied on a laboratory scale (Fig.l). The concentration of gold was found to be in the range of 10-80 ppb. Under the experimental conditions of 15 min. irradiation and 1-h counting time, the detection limit for Au was found to be 0.3 ng, A group of rare earth elements: La, Eu, Ce, Tb, Sin, Lu, Yb was identified: the concentration of europium was found to be 0.5 ppm.

Work done in Soreq. VI-19

Table 1. Au in Israel rock samples determined by ENAA* followed by high resolution gamma-ray spectrometry of the 411.8 keV peak of Au-198.

Sample weight Au found Au (mg) (ng) (ppb)

25.61 2.20 +_ 0.08 85.9 + 3.1 3S.90 0.71 +_ 0.18 17.8 + 4.5 40.05 3.11 ^ 0.20 77.6 + 5.0 41.15 3.30 +_ 0.17 80.2 + 4.3 43.17 3.40 +_ 0.14 78.7 + 3.3 50.10 0.67 +_ 0.19 13.4 + 3.8 50.20 0.56 + 0.23 11.1 + 4.5

*The samples were irradiated for 15 min. and counted for 1-h after a cooling time of 4 days,

Fiy 1.

IOOO

t %

5oo VI-20

Reduction of Blur in Coplaner Rotational Laminagraphic Images

B. Cohen and Y. Segal

Department of Nuclear Engineering Technion - Israel Institute of Technology Haifa. Israel October 1987

Coplaner Rotational Laminagraphy (CRL) is a radiographic technique that creates an image of a cross-sectional view through a solid body. All laminagraphic techniques are based on a relative motion of radiation source, detector and object, (Kiefer 1959) in such a way as to focus a predetermined plane through the object on the radiographic film. The CRL technique requires that the object and film be rotating simultaneously while the radiation souce remains static

OBJECT source

Figure 1: Schematic of Coplaner Rotational Laminagraphy. VI-21

The cross-sectional image which appears on the developed film is essentially a grey-scale map of the objects characteristic attenuation in the plane containing the film and focal spot. The main disadvantage of this technique is that the image on the film is highly blurred. One example of this method can be seen in figure 2. The object used is composed of five slabs of Aluminum, each slab being separated from its' neighbor by a set of precision fillet guages. The spacing between each slab is .45, .55, .65, .75 mm, and each slab is 5 mm thick by 25 mm wide. The graph in the lower part of the picture shows how the film density fluctuates across the image along the horizontal white line seen in the center of the frame. This example shows that while the human eye can easily recognize the 4 slits in the object, it is almost impossible to make any kind of quantitative measurement regarding dimensions in the original object.

Figure 2: CRL image of test object. Graphic overlay shows density fluctuation across center of image.

Several groups (Bates & Peters 1971, Gore & Orr 1979, Orphanoudakis & Strohbehn 1976, Israel & Amtey 1980) have mathematically analysed the image formation process that occurs in CRL and have shown in VI-22

several different ways that the resultant image can be related to the ideal image by the expression:

Ft R } = FC ! > / k <1>

where R(x,y> and Kx,y> are the resultant and ideal images repectively, FC J is the two-dimensional Fourier Transform, and k is the radial distance from the origin in the fourier domain. Bates & Peters attempted to reverse the degradation process by applying an optical filter based on eq. 1, however, their results were not very good due to the limitations imposed by the optical filtering technique.

A different and much more flexible approach would be to use a digital computer to perform the deconvolution between the ideal image and the 1/k degradation. The inverse filter also known as a ramp filter can be modified by a suitable window function so that noise in the original image, especially high frequency noise, will not be too problematic in the deconvolved image. Figure 3a shows the original density profile through center of the image (as shown in figure 2) while the density profile through the restored image can be seen in figure 3b. The ramp filter in this case was multiplied by a gaussian shaped window in the frequency domain.

This result shows how a very simple inverse filter can be of value in reducing the blur in CRL. Research is now under way to investigate more sophisticated filters that can deal more effectively with noise, non-linearity, and other degradation effects inherent to all radiographic techiques. VI-23

' ' I ' I ' ' I I I ' ' ' • I I . i 1 I I I I I I I I I I I I I I I I .01 0.52 1.03 1.54 2.05 2.E7 ^3.08 3.S9 4. JO 4.81 S. 12 CGUitiN xlO" Figure 3a: Density profile through center of test object (units on vertical axis are arbitrary)

A...,

w i / V f\I

I i , I I A. z: v.52 l.os :.;•! 2.-;s 5.03 S.5? <•.: 4.61 S.12

Figure 5b: Restored Density profile through center of test object, (units on vertical axis are arbitrary) VI-24

References s

Peters T.M., "Spatial Filtering to Improve Transverse Tomography", IEEE Trans, on Bio. Med. Eng.. Vol. BME-21. No. 3, May 1974, pages 214-219

Bates R.T.H. and Peters T.M., "Towards Improvements in Tomography", New Zealand Journal of Science, Vol. 14, 1971, pages 883-896

Gore J.C. and Orr J.S.. " Image Formation by Back-Projection: A Reappraisal". Phys. Med. Biol.. Vol. 24, No. 4, 1979, pages 793-801

Kieffer J., "Analysis of Laminagraphic Motions and their Values", Radiology, Vol. 33. 1939, pages ?60-58?

Israel J.W. and Amtey S.R., "The Degradation Function in Conventional Tomography", Med. Phys., Vol. 7, No. 6, 1980, 677-684

Orphanoudakis S.C. and Strohbehn J.W., "Mathematical Model of Conventional Tomography", Medical Physics, Vol. 3, No. 4, 1976, pages 224-232 VI-25

ULTRASOUND INSPECTION OF THE ADHESIVE PROPERTIES OF METAL TO METAL BONDINGS

P. Dickstein, E. Segal, Y. Segal

Department of Nuclear Engineering Technion, Haifa 32000, Israel

ABSTRACT

A large variety of adhesions, representing different adhesive properties were tested both destructively and non-destruatively using the Ultrasound Pulse Echo Technique.

The ultrasound signals were analyzed in both the time and frequency domain. Correlations between features obtained from the signals and the adhesive properties of the bondings were found. Based on these correlations, an algorithm for the inspection of the adh»=vive properties of bondings is developed.

INTRODUCTION

There are certain advantages in the bonding of components through adhesion rather than through nailing or welding. These advantages include: the prevention of thermal shocks and stresses, more even distribution of stresses, etc. The limitations of adhesions arc; pronounced sensitivity to surface preparation, proper polimerization process of the adhesive and the effect of environmental conditions during' the time of manufacturing and operation. These limitations impose strict NDT procedures during manufacturing and service. Since quite reliable techniques, like Fokker, are already applied for the evaluation of the cohesive strength, it is the adhesive strength of bondings which is of major concern. The development of improved deterministic NDT methods is essential, since bonding is the only means of putting together components made of modern materials, composites and ceramics.

In the present work, a large number of specimens of bonded Aluminium plates were studied. The specimens represent a variety of suxfsrse treatments applied to the adherents, and a variety of VI-26

*dh«siv*s' thicknesses. The bonded joints were scanned and the ultrasonic signal from each point was recorded and digitized. For each signal, several features were computed, based on the signal's parameters in the time and frequency domains. These features were used for comparing specimens of different adhesive properties.

A destructive testing was carried out, of which the results were compared to those of the ultrasonic inspection. It turned out that proper analysis of the ultrasonic signal can provide information regarding the surface preparation of the adherents.

EXPERIMENTAL

The specimens were constructed of two aluminium plates, 1.62 mm. thick, and a layer of FM-73 adhesive. The thickness of the adhesive layer was controlled by a foil located between the adherents. Three series of specimens were prepared, six items in each series. The thicknesses of the adhesive layers were: 0.0, 0.05, 0.1, 0.2, 0.3 and 0.6 mm. Each of the three series was characterized by different process of surface preparation: MIL-A-8625D (recommended), aceton cleaning, and MIL-C-5541C (not recommended), accordingly.

THE MEASURING SYSTEM

The specimen is immersed in an immersion tank with an X-Y scan bridge Trienco 705 and a C-scan recorder. The location of the bridge is determined with a precision of 5 microns by two linear electronic sealers, Mitutoyo At-2 N500. Several Panametric transducers were used with characteristic frequencies of 1, 2.25, 5, 10 and 20 MHz. The pulser-receiver was Panametrics 5052 PRX100. The relevant portion of the ultrasound signal was electronically gated before digitation and processing by a Panametrics electronic gate 5052GPD-2. The signals were monitored by a 2-channel Tektronics 2445 oscilliscope. The analog ' ultrasound signal was digitized by a Tektronics 7912AD Programmable Digitizer. The sampling rate was of the order of 1 GHz. According to Nyquist, this rate ensures the preservation of high frequency components of the signal (Oppenheim, 1983). The signal before-and after digitization is monitored by a Tektronics 624, XYZ Digital Monitor, and a Tektronics 634, TV Video Monitor. The digitized signal is recorded and kept in the memory of the PDP 11/34 Mini-Computer. The analysis of the signal was carried out by a proper software. Each specimen was scanned at 10 successive points. VI-27 THE ULTRASOUND SIGNAL ANALYSIS

The Fourier transform of the ultrasound signal was obtained through a Discrete Fast Fouriar Analysis - DFFT. For each signal, several features were calculated, Among these features arei

1. The moments of the signal in the time domain.

2. The moments of the signal in the frequency domain.

3. The Spectrum parameters, including: (a) the real part

A comparison was carried out between the features of th* ten scanned points of each specimen. Such a comparison tells much about the uniformity of the bonding throughout the specimen. It should be kept in mind that when disuniformity is observed, it implies that the strength and quality of the bonding varies from one point to another. Yet, a uniformity of bonding along the specimen is not sufficient for determining whether the adhesion is strong. From the analysis, the following most important conclusions could be drawn:

1. The spectra are sensitive to the thickness of the adhesive layer, but are not sensitive to surface preparations of the adherents.

2. The uniformity test proved to be more sensitive when making use of low-frequency transducers.

3. No correlation between the degree of uniformity and the type of surface preparation was observed. VT-28

4. When making use of a low-frequency transducer, a notable shift of the mean value of the spectrum is observed in specimens of the second and third series. 2 5. HOCA's Y is sensitive to the surface treatment of the adherents when using a low-frequency transducer.

6. In general, low-frequency measurements were found to be more sensitive to variations in the bond quality.

DESTRUCTIVE TESTING

Five strips were cut from each specimen. The yielding load of every strip was measured. Several points are worth mentioning concerning the results of the destructive testing:

1. The breaking load of the series 1 specimens are slightly better than those of series 2, even though they are of the same order. The failure of the members of both series was cohesive.

2. A notable decrease in the breaking load of the specimens of series 1 and 2 is revealed only when the thickness of the adhesive layer becomes large= The failure in this case only is both adhesive and cohesive.

3. The adhesive failure of the specimens of series 3, regardless of the thicknesses of the adhesive layers, indicates that the surface treatment to the adherents of these specimens is improper, and a physical connection between the adhesive and adherents is not formed.

As to point 2 mentioned above, when the thickness of the adhesive layer is large, the curing process of the adhesive does not result in a homogeneous layer but in the formation of voids which effectively reduce the contact area between the adhesive and the adherents, thus decreasing the strength of thfc bond. This phenomenon is clearly observed in Fig. 1, which shows the two points of the specimen from series 2, with a 0.5 mm. adhesive layer after breaking. VI-29

Fig. 1. Improper curing when the adhesive layer is large.,

THE DESTRUCTIVE VS. THE NONDESTRUCTIVE RESULTS

When plotting the breaking load of the specimens vs. each of the features calculated for the specimens according to their ultrasound response, no notable correlation is revealed, regardless of the frequency of the transducer. No single feature seems to be directly related to the strength of the bonding.

When applying a multi-dimension mapping procedure to the assembly of features obtained from the ultrasound experiments, clusters are obtained, in some cases, of specimens which underwent the same surface preparation. The clusters are separated, so that a distinction between bondings of different adhesive properties is achieved. Techniques from the Pattern Recognition field are applied to determine the optimal dimension of the mapping procedure and to select the most significant features, resulting in clusters which are mostly separated. An example of a 2-D mapping is given in figure 2. In this figure, the clusters are not obvious but a separation is obtained by means of a decision-function. VI-30 "07 . muz , lMlta:

u 0 s H r_f t) o 1 t z D»- s 3 X 3 .5

4. 5w 0. 41.,

3. TV ~r- TT:-

Fig. 2 2-0 Feature Mapping of Ultrasound Experimental Results. TIPUL - 1st series, NIKUY - 2nd series

CONCLUSIONS

The analyses of the experimental results imply that there seems to be no single ultrasound feature which is directly related to the strength of the bonding, and only a suitable feature mapping procedure enables the separation of specimens with different adhesive properties.

The sensitivity of the features to the adhesive characteristics of the specimens increased when using low-frequency ultrasound probes. Modeling of this phenomenon is being prepared.

The correlations between ultrasound features and the adhesive properties are promising. Based on these correlations, an algorithm for the evaluation of the adhesive properties of bondings is established, to enable an automatic microprocessor-controlled accept/ reject test.

REFERENCES

Kedem, B. (1986a). Spectral analysis for discrimination by zero crossings. Proc. IEEE, 74, No. 11, 1477-1493.

Kedem, B. (1986b). Search for periodities by axis crossings of filtered time series. .'Signal Processing, 10, 129-144. VI-31

SINGLE PROJECTION TOMOGRAPHY OF OBJECTS WITH CYLINDRICALLY SYMMETRIC DENSITY DISTRIBUTION

D. Pal*, A. Notea*, M. Deutsch**

*Dept. of Nuclear Engineering, Technion, Haifa, Israel **Dept. of Physics, Bar-Ilan University, Ramat-Gan, Israel

Introduction

A radiographic projection image superimposes effects of shape, material density distribution and elemental composition of the product under examination. To deduce the linear attenuation coefficientyU- from the measured image represents an inverse problem. Generally, multiangular projections are required for the 3D reconstruction ofu.(J?) using any of the algorithms developed to invert the Radon transform {Herman, 19B1). When the object has cylindrical symmetry, the reconstruction is possible from a single radiographic image using the Abel transform (Bracewell, 1978), provided that the radiographic image is generated with the radiation beam travelling perpendicular to the axis of symmetry of the product. This process may be considered as a single projection tomography.

Problems involving the inversion of the Abel integral equation are found in various fields of physics to /MTC) • The present contribution is confined to the problematics of g(r) determination. Many industrial products are cylindrically symmetric and are expected to have cylindrically symmetrical distributions of density and elemental composition. Even for products which deviate from strict cylindrical symmetry, the quantitative information obtained using the present method may be sufficient for an "acceptance/rejection" decision according to a given set of criteria. VI-32

Method

The radiographs were generated at a *film-to-focus' distance, large compared to the diameter of the examined item, so that the x-ray beam may be taken to be perpendicular to the object's axis of symmetry. The film, Agfa-Gevaert Structurix D4, was in contact with the object. The radiographic image was digitized with an automatic microdensitometer, Photomation 1700 Optronics Inc., controlled by PDP 11/34. The aperture of the digitizer is either 25x25, 50x50, 100x100, 200x200 or 400x400 micrometer. The smaller the aperture or the pixel, the larger is the level of the random noise. Suppression of the noise is done by averaging over neighbouring pixels or by suitable filtering. The method to be used should be chosen so that genuine high frequency features of importance in the data are preserved. The inversion model is briefly described below and the application to the radiographic image is anticipated. The attenuation of a parallel beam of x-rays of intensity I transversing a path L within the examined object is given by I(r)/I .

(1) L. — 1 where u. is the linear attenuation coefficient (cm ) in the object. For a radiographic image generated within the "linear range" of the film characteristic curve, the film density D may be approximated by a linear relation (Notea, 1983)

= C - A ?(r) -= C -J$(r*J dL (2) L. where C is the film density measured in an area exposed to the radiation beam with no (or with known) attenuating medium present. A depends on the film representative gradient and the ratio of primary and scattered radiation intensities. The D and C are expressed in film density units (fdu), while g in fdu/cm. In the interpretation process, the inversion of g(r) is sought. For objects with cylindrical symmetry in shape and /*(?) , with the axis of symmetry in the z direction and the radiation beam path in the direction of the x axis, the attenuation along the y direction is given by the Abel transform (Barrett, 1984). a.

O ' (3) VI-33

where R is the radius of the examined object.

The inverse Abel transform is traditionally given by (Whittaker and Watson, 1948)

fiyf} (4 r However, this inversion formula requires the diiferentiation of the experimental data, which enormously amplifies the inherent random errors. This leads to severe distortions in the resultant g(r> (Deutsch, 1983; Deutsch and Beniaminy, 1983). To avoid this problem, an improved inversion formula, denoted g , was developed (Deutsch and Beniaminy, 1982):

To further decrease the influence of the random experimental fluctuatios on the determined g(r), a piece-wise cubic spline function is least square fitted to CC-D(y)]. Then two new versions of the above two versions of g(r) are obtained (Deutsch and Beniaminy, 1983).

Results

The inversion procedure is demonstrated here for a step function in JUL(x) . The object under investigation is shown in Fig. la. It consists of two concentric cylinders of radii R. and R_ having different but constant u. values. The simulated profile response

[C-D(y>] is presented in Fig. lb for R, ; R2 = 10; 2mm, JJL. ; JJL~ = 0.1; 0.01 cm and f- = 0 on the outside. The profile, being a computer simulation, is of course free from the distortions caused by the radiographic system and the inherent scattering within the cylindrical item. Therefore, the borders at 2 and 10 mm. are clearly observed. The simulated profile incorporates the effects of both the shape of the item and its ix- variation. The corrected radial profile, g, which in this case is very close toyU, was determined by the inversion of the Abel's integral. This profile is free from the shape effect. The levels of 0.01 and 0.1 were reconstructed, except for the edges where over- and undershoots appear on both sides of the discontinuity in JU- . The overshoots effect is known as the Gibb's phenomenon. When the VI-34 aero-valued points of the profile response beyond 10 mm. are not included in the inversion calculations, the corrected profile does not drop to zero at 10 mm. but rises sharply. Also, a few points for r<0 should be included in the measured radiographic profile to prevent discontinuity oscillations near r = 0 in the inverted radial profile (Deutsch and Beniaminy, 1983).

At any side of the discontinuity, the oscillations decrease in amplitude as the number of the input points increases.

0.10

- 0.09

RADIAL DISTANCE a. b = Figure 1 a. Concentric cylinders with ytt^O.l and yU2 0.01. b. Simulated profile response ("profile") and the Abel inverted radial profile ("corrected profile").

Inversion of simulated radiographic profiles including 1 to ZH randomly distributed noise showed that the spline method used suppresses such errors.

The method is next demonstrated on two cylindrically symmetric objects: a 9 mm. radius graphite cylinder and a hollow lucite cylinder of 9 mm. internal radius and 21 mm. external radius. The measured radiographic profiles D(y) were digitized across the radiographs perpendicular to the axis of symmetry of each object. The radiooraphic profile D(y) measured for the graphite cylinder was averaged over 50 lines with an aperture of 400x400 micrometers and then the profile [C-D(y)3 was calculated, see Fig. 2a. The Abel inverted profile g(r) is shown in Fig. 2b, displaying a constant density distribution as expected. VI -35

0.0 < 0123456709 to 11 RADIAL DISTANCE (mm)

Figure 2a. Profile of a graphite cylinder.

0.0*

01 13*35 7 8 9 10 H RADIAL DISTANCE (mm)

Figure 2b. Reconstructed radial profile of the graphite cylinder determined by. the inverse Abel transformi VI-36

0,0 4

RADIAL DISTANCE

3a» Profile of a hollow lucite cylinder*

0.05

-0.02 -L 10 , 20 30 RADIAL.DISTANCE (mm)

Figure 3b. Corrected profile of the hollow lucite cylinder obtained by applying the inverse Abel transform* VI-37

The profile CC-D(y)] for the hollow lucite cylinder is shown in Fig. 3a. The corrected profile g(r) is shown in Fig. 3b. Across the air channel, a value of g close to zero is obtained and a constant one is obtained across the lucite. Figures 1 and 3 can be readily compared describing similar physical situations.

Discussion

It was shown that the inversion based on the Abel transform has a pronounced advantage of separating the influence of the shape of the item from the radiographic profile. Therefore, the corrected image g(r) reveals variations in material density and elemental composition which were masked by the shape effect. The Gibb's fluctuations obtained close to the discontinuity reduce the detectability of true variations in U* in these regions.

In our opinion, the inverse Abel transform discussed here is worth closer inspection in terms of possible use for radiographic interpretation.

References

Barrett, H.H. (1984). In E. Wolf (Ed.), Progress in Optics. Vol. 19. North Holland, Amsterdam.

Beniaminy, I. and M, Deutsch (1962). ABEL: Stable high accuracy program for the inversion of Abel's integral equation. Comp. Phvs. Comm., 27, 415-422.

Bracewell, R.N. (197B). The Fourier Transform and its Applications. McGraw-Hill, New York.

Deutsch, M. (1983). Abel inversion with a simple analytic representa- tion for experimental data. Appl. Phvs. Lett.. 42. 237-239.

Deutsch, M. and I. Beniaminy (1982). Derivative-free inversion of Abel's integral equation. Appl. Phvs. Lett., 41, 27-28.

Deutsch, M. and I. Beniaminy (1983). Inversion of Abel's integral equation for experimental data. J. Appl. Phvs., 54, 137-143. VI-38

Herman, G.T. (1981). In T.H. Newton and D.G. Potts (Ed.), Radiology of the, Skull and Brain, Vol. 5s Technical Aspects qf Computed Tomography. The C.V. Mosby Comp., St. Louis, Chap. 110, pp. 3888-3503.

Notea, A. and Y. Segal (1963a). Resolving power of dynamic radiation gauges. Nuclear Technology. 63, 121-128,

Notea, A. (1983b). Evaluating radiographic systems using the resolving power function. NDT International. 16, 263-270.

Notea, A. (1987). Phenomenologycal and semi-empirical linear response theory for interpreting measurement system data. To be published.

Notea, A., D. Pal, and M. Oeutsch (1987). Density Distribution in Cylindrically Symmetric Objects from a Single Radiographic Image. 4 th European Conference on Ml on-Destructive Testing. Queen Elizabeth II Conference Centre, 13-17 September, 1987.

Nhittaker, E.T. and G.N. Watson (1948). A Course of Modern Analysis. McMillan, New York. VI-39

SOLUTION OF THE INVERSE PROBLEM IN RADIOGRAPHIC INTERPRETATION BY THEORETICAL MODEL AND CONVOLUTABLE FUNCTION

Y. Bushlin and A, Notea Department of Nuclrair Engineering Technion, Haifa 32000

Extraction of quantitative information is presently the major goal in the interpretation of NDT results. The interpretation process is actually the determination of the characteristics of the investigated object by solving the inverse problem.

In solving the inverse problem involved in the interpretation of radiographic imaging, i.e. the reconstruction of the characteristics of the investigated objects out of the obtained images, we distinguish two steps [13• The first includes a physical model describing the ideal output of the radiographic system in terms of the radiation interactions with materials. The imposed constrain on this model is that it should be invertible. The second step is a linear response theory adapted to provide a phenomenological description of the radiographic system in terms of transfer functions, T. It is preferable if such functions are shift invariant, such as the Point Spread Function (PSF), which is a two dimensional function, or the Line Spread Function (LSF), which is a one dimensional function. Those functions basically describe the degradation and distortions in the measured image due to effects such as scattering, focal spot size, imaging system characteristics, digitization method, etc. Whenever the second step is describable by the linear theory [2], the following integral equation of the first kind between the ideal and measured images may be used.

where * denotes the convolution integral.

The shift invariant [3] transfer function in our approach, should be determined from experimental data and then used for the inversion of eq.(1). In this way the ideal undistorted image is obtained by inversion from eq.(l) with a semi-empirical transfer function. From the ideal image the characteristics of the object are determined by iverting the physical model mentioned as the first step. A crucial decision in employing the approach is the separation between the effects to be included within the physical model, and those to be left VI-40

to be described by the linear transfer function. From a practical point of view, it is convenient to invert eq.U) with a transfer function which relates solely to the non-local effects [3] T(r-r'). Whenever it is possible to define such a transfer function, i.e. T(r-r') and not T(r,r'), the measurement system is characterized by a single function. Buch a T may be developed if all the local effects are included in the physical model which predicts the "ideal image".

To demonstrate this separation into non-local and local effects, we shall study the radiography of a slot with a cone beam, see Fig. 1. Obviously when the beam is parallel, the slot image is independent of its position in the image plane, while for a cone or fan beam the image of the slot does change with position. When the ideal profile to be determined should be rectangular, the transfer function includes the dependence on position [4J. However, if this dependence will be entirely included in the physical model, the transfer function will become shift invariant and the integral of eq.(l) a convolution. The inversion or deconvolution will result in an ideal image that is of rectangular profile for ^< = 0, and trapeze or triangular for o(> 0, see Fig. 1. The experimental data was obtained from a sample with highly accurate slots of width varying from 0.1 to 5 mm. The sample was positioned in the plane of the image by an accurate x-y table.

sores of radiation

slots

fcjert

irnag? density profil== ^X Fig. 1 Radiography with an angle o( between the slot and the radiation direction.

The films were analyzed using an automatic microdensitometer, Optronics 1700 controlled by a PDP 11/34. The shape of the measured density profiles for two different angles are shown in Fig. 2. The variation of the full width at half maximum as a function of the angle is shown in Fig. 3. The results were completely confirmed by Monte-Carlo simulation. VI-41

e -

-

-

s d a •

1 1 , 1 1 1 1 1 1 1 1 1 1 1 , , i 1 i 1 ! 1 !' • • ' I i • i I • • • I • • i I • i • I > !.»:• I.J> M it- V

Fig. 2. Measured density profiles of a slot (100/tom width), radiographed at two different angles (positions). kFWllM

500

400

300

200

100 J_ _L _L -2.79 -1.86 -0.93 -0.46 0.0 0.46 0.93 1.86 Fig. 3 Full width at half maximum for a slot (100 Acnwidth) as a function of the angle. The solid line represents the calculated function.

The transfer function applicable for the slot is the line spread function, LSF, which according to the above-enplained definition will include the distortions which are independent of position. This LSF will be measured once at o( = 0, e.g. by a suitable step change. The LSF is used for the inversion of eq.

extracted from two radiographic images with a known lateral shift between them.

Another example of this kind that Is under study relates to the radiography of objects with finite dimensions [5].

References

1. A. Notea Phenomenological and Semi-Empirical Linear Response Theory for Interpreting Measurement Systems Data To be published.

2. Y. Bcshlin, 0. Ingman & A. Notea Moments Analysis Method for the Determination of Dimensions from Radiographs Nuclear Technology 74, August 19B6.

3. The non-local or shift-invariance property was referred to, in some papers, as "isoplanary", e.g. K. Rossman, "Point Spread Function, Line-Spread Function and Modulation Transfer Function" Radiology 9_3 (1969) pp. 252-272; A. Papoulis, "Systems and Transforms with Applications in Optics", McGraw-Hill Book Comp., N.Y. (196B).

4. Y. Segal & F. Trichter Limitation in Gapes Width Measurements Through X-Ray Radiography To be published.

5. U. Feldman, Y. Bushlin & A. Notea Effects of Finite Dimensions of the Object on Radiographic Imaging Will be presented at the Danish-Israeli NDE Symposium, Denmark, June-July 1988.

The study is part of a joint project with the Institute of Materials and the Institute for Nuclear Technology and NDT at the University of Hannover.

The research is partially sponsored by the Niedersachsen Research Program. VI-4 3

DUAL ENERGY APPROACH FOR MASS AND ELEMENTAL COMPOSITION DISTRIBUTIONS IN CERAMICS

F. Trichter and A. Nofcea

Department of Nuclear Engineering TECHNION - Israel Institute of Technology Haifa 32000, ISRAEL

The interpretation of a radiographic image poses an inverse problem which is ambiguous and does not possess a unique solution. The physical reasons are provided by the physical nature of the interactions of x-ray radiation with materials. The grey level distribution observed in the image is, interpretable in terms of the radiation chord length distribution in the examined object, the mass density distribution and the elemental composition. The ambiguity here can be removed if some grounds are provided from other measurements for selecting one of the possible solutions and rejecting the other. In ceramic products the uniformity of either elemental composition or mass density distribution may not be taken for granted, and additional information is required for the interpretation. The dual energy technique has been suggested for the separation between the density and composition effects already by Hounsfield [1] for the early computerized tomographs (CT). Later efforts [2-7] in this direction led to empirical relations between the attenuation coefficient A* , characteristics of the x-ray energy spectrum, atomic numbers and material density, in the range of interest for biological materials. Therefore, the developed formulas are relevant for the x-ray energies of the medical CT.

For monoenergetic photon beams, the following expressions for the photoelectric/Rayleigh and Compton attenuation coefficients ares

(2) VI-44

Where Z* and N* are the effective atomic numbers and the effective number of atoms per unit volume. These values are defined for a single hypothetical substance which would give the same linear attenuation coefficient as a mixture of substances at a given energy beam, E [8]; Cc (E) is the Klein-Nishina cross section.

The power n of E for carbon, oxygen, phosphorus and calcium is 3.30, 3.2B, 3.12 and 3.08 respectively, and the power m of Z for 40, 60 and 80 KeV have been found to be 4.5, 4.G2 and 4.S3. In principle by radiographing an object with two energies (one where the Compton is dominant and the other where the photoelectric is dominant), the effective atomic density and effective atomi.c number of the object can be determined using eqs.(l) and (2).

For polychromatic energy beams from x-ray units, the spectrum varies with the penetration depth in the object. This effect known as "beam hardening" results in increasing the mean energy of the beam. The attenuation is usually described by the exponential low with an effective attenuation coefficient and a corresponding effective energy. Some results obtained by us for the effective energy are presented in Table 1.

TABLE 1

Effective energy values in different materials for 60 KeV (Radiofluor 350-072-700)

| Material 1 Lucite | Al Ti 1 F Brass 1 1 i i I i 1 | Thickness 1 2.5 mm. j 1.3 mm. 0.8 mm. | 0.6 mm. | 0.5 mm. 1 1 l i I1 Ii | E* 1 16.5 KeV | 28 KeV 46.5 KeV|52.g KeV| 56 KeV 1 f 1 i

The beam hardening imposes a serious limitation on the extraction of the effective atomic number and atomic density from the dual technique, as the E values to be substituted in eqs.(l) and (2) are unknown. Another point that should be stressed is that from radiographic images after digitization, the values of the grey level are obtained, while in eqs.(l) and (2), the attenuation coefficients are required. The situation is different with the CT as the reconstruction algorithms there are expected to provide the linear VI-45

attenuation coefficient. The additional information required for the interpretation is determined from experimental data obtained with samples of known thickness, material density and elemental composition. The samples, in a shape of a step wedge, were radiographed simultaneously with the examined object.

The relation between the film density and attenuation coefficient of the object material in the "linear density" region (density of about Z) can be written by [9]

(3)

where

D - film density; JU - attenuation coefficient; O - material density; x - sample thickness; 1C - the gradient of the characteristic curve; l( - the film density at x = 0.

In the Compton energy region, the comparison between the film densities of calibration sample - Dc and examined object DDcc provides the material density of the object. c s v = (1

In the photoelectric energy region, where the attenuation coefficient of the calibration sample is known, the attenuation coefficient of the examined object is given by:

From eqs.<5) and (1), Z* is determinable by iterations. The accuracy in the Z* value depends strongly on the matching of the suitable effective energy. Therefore, the attenuation characteristics of the sample should t»e as close rs possible to those of the examined object. For example, if the sample is an aluminum step wedge and the object is a 2 mm. thick titanium, the Z* (Ti) that was determined via the radiography was 16 instead of 22. On the other hand, the examination of two series of ceramic materials- silicon nitride (Si_N ) with different Y2O3 content: 3H (CO and SK (LC), (material density D= 3.4 g/cm ) relying on Al step wedge, gave much better results. VI-46

Radiography of the ceramic: samples in the Compton energy region at 120 JcV showed identity in the slope of the film density versus thickness curve <4.1 for LC series and 4.13 for CC series) i.e. the material density values of both fire close. While in the photoelectric energy region (50 kV), the curves showed significantly different slopes i.e. difference in Z* of the samples (see figs. 1 and 2). The theoretical Z* values and those calculated from the experimental data are shown in Table 2.

TABLE 2

1 1 Theoretical Experimental 1 L 1 1 r 15.9 ;t 1.6 15.01 + 1.5 |

lz*cc ' 14.22 jt 1.4 14.21 + 1.4 |

The close agreement with the theoretical values results from the use of Al step wedge whose characteristics are very close to those of the ceramic samples.

Dual energy radiography can be a powerful tool for the detection of nonuniformity in density and composition in ceramic objects, with the imposed limitation that the characteristics of the basic matrix are known and used in the preparation of reference samples.

References

1. G.N. Hounsfield, Computerized Transverse Axial Scanning (Tomography). Part I - Description of a System. Br. 0. Radiol. 46, 1016-1022 (1973).

2. R.A. Ratherford, B.R. Pullan and I. Isherwood, Measurement of Effective Atomic Number and Electron Density Using an EMI Scanner, Neuroradiology, ii, 15-21 (1976).

3. R.A. Ratherford, B.R. Pullan and I. Isherwood, X-Ray Energies for Effective Atomic Number Determination, Neuroradiology JUL.I 23-2B (1976). VI-47

4. L. Dubai and U. Wiggli, Tomochemistry of the Brain, J. Comput. Assist. Tomogr. I, 300-307 (1977).

5. W.H. Marshall, W. Easter and L,M. Zatz, Analysis of the Dense Lesion at Computed Tomography with Dual kVp Scans, Radiology 124, 87-89 (1977).

6. R.E. Latchaw, J.T. Payne and L.H. Gold, Effective Atomic Number and Electron Density Measured with a Computed Tomographic Scanner, J. Comput. Assist. Tomogr. 2, 199-Z08 (1978).

7. R.A. Brooks, A Quantitative Theory of the Mansfield Unit and its Application to Dual Energy Scanning, J. Comput. Assist. Tomogr. 1, 487-493 (1977).

8. B.R. Pullan, R.T. Ritchings and I. Isherwood, Accuracy and Meaning of Computed Tomography Attenuation Values, Radiology of the Skull and Brain, Technical Aspects of Computed Tomography (1981).

9. A. Notea, A. Fishman, S. Wajnberg and Y. Segal, Efficiency of Radiographic Film and Film/Screen Systems, NDT International, 14, 171-17S (1981).

The study was partially funded by the Levi Eshkol Fellowship, National Council for Research and Development. VI-48

Al

6 (rnno

Fig. 1. Changes in film density versus sample thickness for LC ceramic material at GO kV with the Al step wedge. VI-4 9

6 (mm;

Fig. 2. Changes in film density versus sample thickness for

CC ceramic material at 60 kVf with the Al step wedge. Nuclear Medicine

Chairmen - E. Lubin and A. Kushelevuky VII-1

Preparation of 73Se by proton irradiation of Br target

Zeev B.Alfassi, Peter Jones-Smith and Regin Weinreich

Department of Nuclear Engineering, Ben Gurion University ,Beer Sheva Israel EIR.Wiirenlingen Switzerland

The purpose of this work was to study the possibility to use the high energy of the proton beam (72MeV) of SIN injectors (split beams of the injectors to the 590 MeV cyclotron) to produce 73Se. In principle a nuclide with atomic number Z can be produced by a Z-l element, (p, xn) reaction, Z+l element, (p.axn) reaction, or higher Z elements with emission of both a and proton or two alphas etc. However the cross sections for the last reactions is usually low. As the ' ^As(p,3n) "Se was already studied by Nozaki et al^and was found to occur mainly below 50 MeV we decide to study also the reactions of

Br with protons, 79Br (p,a3n)73Se (Q= -27.2 Mev) and 81Br (p,a5n)73Se (Q= -45.3MeV). The use of bromide target is advantageous on using an arsenic target due to the much higher current which a bromide target can tolerate. Experimental 7 73 7 The excitation functions for the formation of ^Se, Se and ->Se for proton bombardment of Br was studied from energy of 70.6 MeV and down by the standard stacked-foils technique. As 7^Se has only very low energy x-rays it was prefered to measure its formation not through these x rays as was done by Nozaki et al^) but rather by the measurement of the activity of its daughter, '^As. Since '^As is formed also directly by the protons, the measurement for 7^Se 7^As was done after ten hdf-lives of 7^As (26.0h). The gamma activities were measured with the usual Ge(Li) -multichannel set-up VII-2

and the recorded spectra on a IBM -PC where analysed with Orect's geligam program. The foils consisted of KBr pellets, produced from 99.9% -pure powdered KBr with a conventional Perkin -Elmer pellet mold with a pressure of lOkbar. The pellets were of 13 mm diameter with weights of 200-1000 mg per pellet. Al and Cu foils of 0.05-0.125 mm thickness were used as monitors^). Results Figure 1 gives the cross sections for ' ^Se formation in thin targets of KBr. Two runs are plotted; one of foils with thicknesses of 150-265 mg.crrf^ and the second of 280-370 mg.cm"2 pellets. The results of the two runs agree quite well and their results agree also with data obtained with thicker pellets. The maximum cross section is at about 55-57 Mev, i.e. about 20 MeV higher than the threshold energy for this reaction, assuming

the main reaction in this region is the ™Br(p, a3n) '^Se reaction. For other (p,a3n) reactions studied^) the peak of the excitation function is at 25-30 meV above the threshold energy. The maximum cross section for Br is in the same range as that for other nuclides

studied in the similar A range, i.e. for Ga isotopes (p,oc3n) reactions. Assuming that the

only contributor at 55 MeV is ' ^Br means that crmax is about 40 mb, compared to 80 mb for 7*Ga and 28mb for ^Ga. However for ^Ga it is not clear if the maximum was reached at all. At about 67 MeV it seems that the cross sections start again to increase with

the energy, probably due to the ^Br(p,a5n)73se reaction, again 25 MeV above the threshold energy.

The two contaminants of ^Se ^ 72ge an

the real contamination is in the middle beween the independent yield (I.Y.) and the cumulative yield.(C.Y.) However if the target is irradiated for two hours and the sample is immediately separated only 33.1% of the MBr decays and the contamination is closer to the I.Y.(l/3 of the difference). The cumulative yield is about 3.6 times the independent yield at the higher energies (70 Mev) and only 1.3 times at 48 MeV due to the higher Q values for the 79Br (p,5n)75Kr and 79Br (p,p4n)75Br compared to the 79Br(p,an)75Se. Thus at higher energies 7^Se is formed mainly by secondary reactions while at low energies the direct formation is the main one.

Table 1 gives the actual production rates (nCi/jxAh) for various energies ranges. As can be seen from this table lmCi/MAh can be obtained (54-46Mev) with reasonable purity (0.14% 7^Se and about 0.46% 7^Se). If the amount of impurity is not so important higher yields can be obtained by using higher bombarding energies, however this is not the common case. Table 1: Production rates of 7^Se and the percentages of contamination for various proton energies impinging on KBr pellets

Energy Pellet 7^Se -production contamination (%)

Range thickness (nCi/fiAh)

(MeV) (mg.cm^) (cumulative) (independent)

46-48 229 217 0.08 0.46 0.36 46-50 465 482 0.09 0.53 0.38 46-52 708 786 0.12 0.55 0.40 46-54 959 1120 0.14 0.51 0.42 46-56 1216 1463 0.18 0.58 0.43 46-58 1480 1812 0.23 0.67 0.45 46-60 1752 2127 0.31 0.76 0.47 VI1-4

These results indicate that this method natBr(p,axn)73Se is preferable to a or 3He irradiation of germanium, but inferior to ^^As(p,3n)^3Se from point of view of radionuclide purity. Although the arsenic target leads to higher yield per a given current, as the bromide target can tolerate larger currents the yield with the bromide can be higher than with As target. The yield of the bromide target can be increased by about 15% by substituting KBr with NaBr due to the lower stopping power of the Na atom compared to the potassium atom. 25% increased yield will be obtained by using LiBr compounds. Another way to increase the yield and mainly to decrease the contamination is by using enriched bromine. Both K™Br and K°*Br has to be studied to be sure which of the isotopes is preferable, but from the systematics of (p.axn) reaction it seems that in our energy range '3Se is due to ™Br while °*Br contribute mainly to '^Se and consequently the use of Li^Br will lead to higher yield of more pure '3Se. References 1. T. Nozaki, Y.Itoh and K. Ogawa, Int. J. Appl. Radiat. Isot. 3_Q, 595 (1979).

2. A.Grutter, Nucl. Phys. A383 . 98 (1982).

3. K.A.Keller, J.Lange, H.Munzel and G.Pfening " Excitation Functions for Charged - Particle Induced Nuclear Reactions", Landolt - Bornstein New Series Vol. 5b, Group I, Springer Verlag Berlin 1973. VII-5

On the use of ^Co in nuclear medicine - its disadvantage and why "Ni is preferable

Zeev B. Alfassi and Regin Weinreich

Ben Gurion University, Beer Sheva, Israel and EIR, Wiirenligen, Switzerland

Several works were done in the last years on the preparation of "Co and its use as 55Co - bleomycin^1'2) (or55Co - diethylenetriamine pentaacetic acid DTPA)(3). The reason for these works are that "Co - bleomycin was found to be a good scintigraphic agent for tumor detection, especially for the detection of lung cancer and malignant lymphoma^). The use of ^Co - bleomycin suffers from the disadvantage of the long half life of •* 'Co (271 d) and consequently it was suggested to use the shorter lived ^-*Co

(17.54h)(5>6), However ^^Co y lines are of too high energy for the scintigraphic camera

(931 keV (75%), 477 keV (20.3%)) and the use of the p+ emission in PET measurement was recommended and later done by several groups. It was found by Nieweg et al^ that the detection of lung cancer by ^Co - bleomycin using a positron camera was superior to that obtained with ^Co - bleomycin single photon imaging but inferior to those obtained with ^' Co - bleomycin scintigraphy. "Co - bleomycin and "Co - bleomycin were almost equally sensitive in detection of lung cancer however the tumor - non tumor ratio was significantly higher for the ^'Co case. This difference should be due to the physical properties of ^Co and not to the biological distribution as both are the same compounds. It means that in the case of ^Co, atoms which are located in the tumor are seen in the picture as not in the tumor. The recommended change from ^'Co to ^Co due to its shorter half life did not consider thouroughly its nuclear properties. ^Co is not only a bad radionuclide for single photon imaging, due to the multitude of the y lines and their high energy, it is also an VII-6 undesirable nuclide for positron camera studies. In order for a positron-emitter radionuclide to be an ideal one for PET studies it should have low energy positrons and no gamma emmision. Gamma emissions below 300-350 keV are not recorded by the PET detectors due to the gain and discriminator settings which are designed to Sll keV measurement.Hence photons with energy below 300-350 keV do not interfere with positron tomography. Photons with energy above 600 keV can interfere with the detection of the annihilation photons as due to Compton scattering, some of the photons can fall into the energy window around 511 keV and be registered as due to annihilation (The Compton scattering can be in either the patient or in the detector or in both). The worst gamma emission for a positron emitter to have is in the range of the detector window i.e. 511+50 keV. In this case the photopeak of the gamma line, which have considerably higher efficiency than compton scattering1^) is registered erroneously as one of the annihilation photons. This photon is not necessarly emitted with 180° direction to one of the annihiation photons and since in the case it is detected together with one of the annihilation photons the computer treat it as a true 180° coincidence, a strong distortion of the tomograph picture is obtained. This distortion cannot be reduced by timing of the coincidence unit or by using lower activity of the positrons. The y quantum is emitted from the nuclide produced by the P+ emission and hence their coincidence cannot be reduced by lower concentration . The life time of the levels emitting this y photon is usually very short in the order of several or at most few tens of psec. The emNsion of the annihilation quantas occur after the emission of the decay quanta but as positrons in wat : decay mostly as free positrons or p-positronium, their half life is about 300-400 psec which is in the same order of magnitude as the difference in the time between the two annihilation quanta reaching the two detectors. The time resolution of most PET systems is not less than 5 nsec and hence even the relatively small fraction of positrons forming o - Positronium which decays with a half life of 1.86 nsec cannot be separated by timing,(8>9) If both the undesired photon and the two annihilation photons reach the detectors the event is discarded. However one of the three can be emitted in a direction where there is no detector and even if the whole space is filled with detectors the efficiency of detection is ususally below 50% which renders that not all the cases will be registered as three photons events and be discarded. If the intensity of the unwanted y line is I|and the intensity of the

positron emission is I2 and the respective detection efficiencies is £j and £2, then the VII-7

statistical ratio of accidental to true coincidence is IjE./Lje,,. In getting this ratio it was taken

into consideration that only one of the 511 keV annihilation photons can form an accidental

coincidence with the undesired y, only the one that moves in the direction of the another

bank of detectors than this y. This factor is an upper limit value and the real value will be lower due to geometrical reasons as not all the detectors are connected in the coincidence net.

The 55Co has a y line of 477.2 keV with intensity of 20.3%. The detection efficiencies for the 477 keV and the 511 keV in the PET detectors are about the same and thus the statistical factor for accidental to true coincidences is 20.3/76. The geometric factors will decrease this factor but usually not more than by a factor of 2,i.e about 12% of the registered coincidences are accidental, composed of a couple of 477 keV and 511 keV. It seems that the use of the same element with another isotope, is wrong in this case due to its bad nuclear properties for PET. It seems that a better choice for PET studies with another bleomycin - metal complex will be the use of -^Ni. bleomycin. Ni was found to produce in vitro a stronger complex with bleotnycin than the

cobalt do/*™ As it was found that -^Cr - bleomycin (H»12) and in . bleomycin are also tumor seeking agents, the same might be true for Ni. No work was done on Ni - bleomycin distribution in vivo but this is due to the fact that none of the nickel radioisotopes have a good scintigraphic properties. However -^Ni have reasonably good properties for PET

studies ^Ni nas i^c 55QJ aiso y lines but its yline interfere less with the PET measurements. 55Co have the 477.2 keV (20.3%) and 931.5 kev (75%) as the main lines, while -^Ni main line is of 1378 kev (78%) and hence contributes less to the noise.

References 1. O.E.Nieweg, H.Beekhuis, A.M.J.Paans, D.A.Piers, W.Vaalburg, J.Wellewerd, T.Wiegman and M.G.Woldring, Eur. J.Nucl. Med.,_Z, 104 (1982).

2. A.M.J. Paans, T.Wiegman, EJ.de Graaf, T.Kuilman, O.E.Nieweg, W.Vaalburg and M.G.Woldring, Proc. 9tn Int. Conf. on Cyclotrons and application, Caen (France) 1981. VII-8

3. B.Maziere, O.Stulnaft, J.M.Verret, D.Comar and A.Syrota, Int. J. Appl. Radiat. Isot. 24,595 (1983).

4. R.C.Reba, W.C.Eckelman, K.P.Poulose, R.B.Grove, J.S.Stevenson, W.J.Rzeszotarski and A.Primack, in The Radiopharmaceuticals, R.Subramanian et al (ed.) p. 464.

5. O.Monod and M.Rymer, Cancer Chemother. Rep. part 3,4, 245 (1973).

6. W.C.Eckelman, RKubota, B.A.Siegel, T.Komai, WJ.Rzeszotarski and R.C.Reba, J.Nucl.Med. J2, 385 (1976).

7. J.H.Neiler and P.R.Bell, in " Alpha beta and gamma ray spectroscopy" Siegbahn (ed.) vol. 1, North Holland Publishing Co. Amsterdam 1968, p. 245.

8. HJ.Ache in " Positronium and muonium chemistry", H.J.Acle (ed.) American Chemical Society Series.

9. G.Friedlander, J.W.Kennedy and J.H.Miller, "Nuclear and Radiochemistry", John Wiley and Sons, New York 1964, p. 110.

10. H.Renault, R.Henry, J.Rapin and M.Hegesippe, "Radiopharmaceuticals and Labelled Compounds", p. 1951.IAEA. Vienna (1973).

11. J.Liniecki, J.Pertynski, Z.Kranczykowa, J.Stephien and K.Durski, Nucl. Med. 22, 306 (1983).

12. D.Brykalski, J. Liniecki, J.Dobek, T.Pertynski, St. Faindt, M.Studniarek, K.Durski and M.Mussur, Nuklearmedizin, 24, 66 (1985) VII-9

DISPOSAL OF RADIOACTIVE HOSPITAL WASTES TO THE NEGEV AND FATAL ROAD ACCIDENTS

Y. Kalish

Department of Medical Physics, Beilinson Medical Center, Petah Tiqva, Israel

* Edward W. Webster calculates that a typical barrel of carbon-14 liquid scintillation waste contains about 2 microcuries of the radioactive substance and that if a person were to consume all of the carbon-14 in one barrel, the radiation dose averaged over the whole body would be about 1 millirem. He further calculates: "How much are we spending to prevent one cancer case in this situation, the answer is...about $1 billion per cancer case averted; $1 billion to save a death from cancer. And since people are not actually drinking the radioactive li- quid, the disposal cost to save a possible death will be very much greater...with $1 billion, you could build about 50 modern cancer treat- ment centers". Our calculations reveal that if the probability of traf- fic accidents in transferring the radioactive barrels to the disposal site in the Negev is taken into consideration, at the expense of 1 billion dollars, one human being might be saved from cancer but at least 8 persons can be expected to be killed by road accidents. It is quite evident that it is non-sensical to send such wastes to the Negev and that incineration or burial with ordinary wastes is to be preferred. The author estimates that about 99% of the volume of radioactive wastes from hospitals, most of which consist of short-lived radioisotopes, and which is now sent from all over Israel to the disposal site in the Negev, should be treated as ordinary waste from similar considerations. * Edward E. Webster (Director of Radiation Safety, Massachusetts General Hospital, Prof, of Radiology, Harvard Medical School.)In: "Disposal of Low-Level Radioactive Biomedical Wastes", Academy Forum 24/11/80. National Academy of Sciences, U.S.A. VII-10

THE ALPHA-BETA-GAMMA SCINTILLATION SPECTROMETER AND THE MEASUREMENT OF H-3/I-125 AND C-14/1-125 SAMPLES

Y. Kalish

Department wf Medical Physics, Beilinson Medical Center, Petah Tiqva, Israel

The alpha-beta-gamma scintillation spectrometer was invented by the author in 1972 (1,2) and described publicly for the first time in an international conference on radiation protection in Jerusalem in March 1973 (3) and later in some other conferences (4-9) . The first manual alpha-beta-gamma spectrometer was developed and constructed during 1973 1974 and a prototype of an automatic version was displayed at the 5th International Conference on Medical Physics in August 1979 in Jerusalem A similar spectrometer was also developed by J.E. Noakes in the U.S. (10). The results we obtained with this system in the measurement of double-labelled radioactive samples of H-3/I-125 and C-14/I-12i> will be presented and the advantages of the alpha-beta-gamma scintillation spectrometer in the measurement of such samples will be discussed.

Fig. 1: The alpha-beta-gamma spectrometer (1980) VII-11

-i ;' • d_ • - . 1 VJ .. '• - i .4.1 « X ,».!. 'i .U [,t !.C I ';f <,:!.!.•. . ' "I, I '. J. t.' 1.1 V, .- t 1 (i-v-;-i-.v-) VII-12

Fig. 4 : Spectrum of H-3/I-125 double-labeled sample

Fig. 5 : Spectrum of c-14/1-125 double-labeled sample VII-13

REFERENCES

1. Kalish, Y. Scintillation spectrometer. Israel Patent No. 39434 (Filed: May 12, 1972). 2. Kalish, Y. Scintillation spectrometer. U.S. Patent No. 3,944,832. 3. Kalish Y, Proposal for a new type of alpha-beta-gamma scintillation spectrometer. Proceedings of the Regional Conference on Radiation Protection, Jerusalem, March 5-8, 1973. Vol. 1, p. 312-318. 4. Kalish Y. Proposal for a new type alpha-beta-gamma scintillation spectrometer. Transactions of the Israel Nuclear Society, 1973 annual meeting, Soreq Nuclear Research Center, June 26, 1973. Vol. 1, p. 65. 5. Kalish Y. A new type of scintillation spectrometer: physical princi- ples, construction and experimental results. Transactions of the 1974 annual meeting of the Israel Nuclear Society, December 10, 1974 Vol. 2, p. 8. 6. Kalish Y. A new type alpha-beta-gamma scintillation spectrometer. Awards in Nuclear Medicine and Radiopharmacology, Vol. 3:83-98, 1975. 7. Kalish Y. New type alpha-beta-gamma scintillation spectrometer. Digest of the Fourth International Conference on Medical Physics. Special Issue of Physics in Canada, Vol. 32:25-30, 1976, Ottawa, Canada. 8. Kalish, Y., Sadeh, Y., et. al. An automatic, computerized alpha-beta- gamma scintillation spectrometer. Proceedings of the 5th Internation- al Conference on Medical Physics (combined with the 12th Internation- al Conference on Medical § Biomedical Engineering), Jerusalem, Israel August 19-24, 1979. 9. Kalish Y., Sadeh Y, et al. The alpha-beta-gamma spectrometer - the state of the art. Transactions of the Nuclear Societies of Israel, Joint annual meeting, Ben Gurion Univ. of the Negev, Dec. 4-5, 1980. 10. Noakes, J.E., Spaulding, J.D. Pulse shape scintillation counting for beta gamma or beta-gamma counting. In: Liquid Scintillation Counting, Recent Applications and Development, Vol. 1, Academic Press Inc., 1980. VII-14

ADAPTIVE ASSAY OF RADIOACTIVE PULMONARY AEROSOL WITH AN EXTERNAL DETECTOR

A. Taltnor and Y. Leichter Safety Department Nuclear Research Centre - Negev Israel

Y. Ben-Haim Department of Nuclear Engineering Technion -- Israel Institute of Technology Haifa, Israel

A. Kushelevsky Department of Nuclear Engineering Ben-Gurion University Beer-Sheva, Israel The accuracy of the assay of radioactive pulmonary aerosols is influenced by the spatial distribution of the aerosol in the lungs. Measurements of aerosol dispersion in lungs gener- ally show a relatively uniform distribution shortly after inhalation. However, the spatial distribution changes with time, resulting in decreased aerosol concentration in the par- enchyma and increased concentration in the subpleural region [see refs. in 1]. In a previous communication [1] it was dem- onstrated theoretically that plausible spatial distributions of pulmonary aerosols could lead to large assay errors when using a standard measurement system whose calibration is based on assuming uniform spatial distribution. The purpose of this communication is to outline an "adaptive assay" approach [2] for optimizing the assay of non-uniform spatial distributions of pulmonary aerosols. Results are presented which compare the performance of an adaptive assay with that of the standard measurement procedure. The basic idea of adaptive assay is that the design of the assay system is adjusted during the measurement so as to ooti- VII-15

mize the performance. The measurements are interpreted on-line and the assay system configuration is modified in accordance with the updated understanding of the sample. An adaptive algorithm contains four components [2]: 1. A simplified MODEL of the structure and content of the thoracic region. containing model parameters which approximately characterize the spatial distribution of the aerosol. 2. A DESIGN ALGORITHM which specifies an optimal seauence of measurement positions and durations on the basis of the current estimates of the model Darameters. 3. A MEASUREMENT SYSTEM, which implements and stores the measurements. 4. An ESTIMATION ALGORITHM which determines optimal esti- mates of the model parameters on the basis of the accu- mulated measurements. In this work we study the assay of Am-241 distributed as point sources in the lungs of a Lawrence Livermore Realistic Phantom [3]. The 59.5 KeV gamma radiation is measured. The detector is a Nal-Csl phoswich detector. [Diameter = 20cm; thickness of Nal and Csl layers are 0.15 and 5.0 cm, respec- tively [4]). The simplified model of the spatial distribution of the pulmonary aerosol, upon which the adaptive assay is based, assumes that the concentration of aerosol decreases exponen- tially along lateral axes from the subpleura to the center of the lung. The design algorithm is based on the idea of mini- mizing the variance of the estimates of the model parameters, while the estimation algorithm is based on the concent of max- imum likelihood [2]. The adaptive assay is performed as a seauence of 3 pairs of 10-minute measurements. The model parameters are evaluated (by the estimation algorithm) after each pair of measurements, and the optimum positions for the next pair of measurements are then determined (by the design algorithm) on the basis of all accumulated data. The design algorithm is constrained so that the detector positions are selected at either 2.5 or 10 cm VII-16

from the body surface, and at one of four angular orienta- tions: centered on ventral or dorsal sidesi or under the left or right arm. The initial pair of measurements are always per- formed at 2.5cm from the body, at the ventral surface and under the right arm. We will compare the performance of the adaptive assay with a static measurement of 60 minutes duration. The interpreta- tion of this static assay is based on a calibration of the detector which assumes the aerosol to be uniformly distributed in the lungs. Table 1 shows the results of adaptive and static assays of Am-241, for various total activities. In these measurements the Americium is distributed exponentially as explained above, and is present in both lungs. In the static assay the detec- tor is located 2.5cm above the ventral surface. The uncertain- ties in the estimated Quantities are approximately +10%. Table 1 True Activity CBq) 553.3 398.8 347.7 281.5

Final Estimate by Adaptive Assay 554.4 401.8 352.7 286.2 CBQ)

Estimate by Static 571.0 413.9 364.2 295.9 Assay CBQ) Table 2 shows the results of a different set of adaptive and static assays of Am-241. In these measurements the Ameri- cium is distributed exponentially as explained above, but is present exclusively in the left lung. The detector in the static assay is positioned under the right arm. Table 2 True Activity CBQ] 120.3 70.6 54.1

Final Estimate by Adaptive Assay 110.4 56.9 43.3 CBQ)

Estimate by Static 24.9 13.6 10.7 Assay CBQ) VII-17

In conclusion, we can state the following? 1. The adaptive assay methodology seems to improve the accuracy of the assay of radioactive pulmonary aerosols. 2. When the aerosol is present in both lung? in comparable amounts, the static ventrally-located assay is nearly as good as the adaptive assay in the cases studied. This is in part due to the constraints on the spatial distributions consid- ered, and in part due to the fact that the 59.5 KeV radiation of Am-241 has been measured. Mote random spatial distributions, or radiations of lower energy, are expected to display a great- er distinction between the static and the adaptive measure- ments . 3. In the extreme hypothetical situation in which the aer- osol is concentrated entirely in the left lung and the static measurement is performed under the right arm, the error obtained by calibrating on the assumption of uniform spatial distribution is as largo as 80% of the true value. In the same situation the adaptive assay errs by less than 20%.

References 1. A. Talmor, Y. Leichter, Y. Ben-Haim and A. Kushelevsky, The influence of nonuniform spatial distribution on lung counting, The Nuclear Societies of Israel Trans., Tel- Aviv, Feb. 1986. PP182-5. 2. Y. Ben-Haim, IbS-A.SSay.-fif-Scslislly.-BaQd.SED-MstSCisl, D. Reidel Pub., Dordrecht, Holland, 1985. 3. D. Newton, A.C. Wells, S. Mizushita, R.E. Toohey, J.Y. Sha, R. Jones, S.J. Jefferies, H.E. Palmer, G.A. Rieksts, A.L. Anderson and G.W. Campbell, The Livermore phantom as a calibration standard in the assessment ov Plutonium in lungs, in*. Assessment of Radioactive Con- tamination in Man, Paris, Nov. 1984, pp 183-99, IAEA-SM-276/01. 4. Y. Leichter, U. German, E. Romm and G. Weiser, Nucl. Res. Centre - Negev Report (TN)-075> 1984 Cin Hebrew). VII-18 MULTIFIELD TOTAL SKIN IRRADIATION

Yuri Mandelzweig, Dov Sapir, Mark Yudelev, Morris Tateher

Rambam Medical Center, Haifa, Israel

Total body or half body skin irradiation is used for treatment of cancer patients who suffer from diseases involving the skin. In cases of local disease, we can generally use half body irradiation with electrons, (eg. Kaposi's sarcoma, Cutis lymphoma). In cases where there is widely spread skin involvement, total skin irradiation (TSI) is employed (eg. mycosis fungoides).

The major technological problems encountered in TSI are as follows:

1. Delivery of a radiation dose to a layer 10-15 ran deep which is sufficiently uniform over the patient's entire body.

2. Minimizing contamination of the beam by Bremsstrahliing X-rays generated in electron interactions with matter located between the electron source and the patient.

3. Achievement of a practical treatment session time, taking into account limitations on the maximum dose rate available and on the maximum dimensions of the treatment

In the most cannon method of TSI, patients are irradiated by beams of 3-4 MeV electrons while standing at a distance of 3-7 meters from the electron source in order to produce large treatment fields (1,2). An electron scattering layer is needed to achieve satisfactory dose uniformity. Since the beam intensity decreases with distance while at the same time theX- ray/electron dose ratio is enhanced by preferential attenuation of electrons in the scat- VII-19 tering layer and in air, this method requires an accelerator with a high electron output together with a lew intrinsic level of X-ray contamination.

At a distance of 3 meters, the maximum intensity of the standard 4 MeV electron beam from our Philips SL 75/10 linac is only about 10 rad/min when a 3 mm Perspex scattering plate is present. At the same time, the X-ray contamination increases from a tolerable value below 2% at 1 meter to about 8%. These conditions are not acceptable, hence a longdistance method is not suitable for our machine.

An alternative procedure is to treat the patient at a shorter distance with an electron beam that scans the surface of his body. For example, at Manchester a scanning system has been constructed for the SL 75/10 linac in which a moving platform replaces the usual treatment couch (3). This approach was rejected because it was felt that the adopted nethod should introduce minimum changes in the routine of an already busy treatment machine.

Treatment was performed with an electron beam having nominal energy 4 KeV from the Philips SL 75/10 linac. No electron applicator was attached and no scattering layers were present except for the beam exit window, the monitor ionization chamber and the air space between the linac and the patient. During irradiation the patient reclined on the regular treatment couch at a source-to-skin distance of 150 an. The collimator was fully open.

To cover the entire surface of the patient's body it was necessary to apply multiple fields. Because of the Gaussian nature of the beam, the matching of adjacent fields was not critical. Fig. 1 shows how a uniform dose is produced over a large cross section when 2 beams are joined at the 50% value of the profiles. An error of 1 cm in the separation distance causes an overdosage or underdosage of only 5%.

Repeated phantom studies were made using TLD's (Harshaw TLD-100), film (Kodak XV-2), and ionization chambers (Farmer and Ionex dosimeter systems using micro and thimble chambers). Measurements were made in polystyrene, water and Rando phantoms to check for acceptable uniformity (generally + or - 15%). Following these initial studies, this multifield tech- VII-20 nique was introduced into clinical practice at Rambam Medical Center and has been in use for the past several years.

In general, the clinical results have been mostly positive, although there have been a te'i cases of skin burning in the areas of tangential electron incidence.

In this report, we compare two possible methods for reducing the effects of tangential incidence: 1. Shift method for incident beams with respect to body surface (not effected by pat- ient geometrical parameters) (4).

2. Full Width Half Maximum correction method in which field sizes are chosen according to patient contours.

Results of this comparison are now in the process of being prepared for clinical introduction.

REHREHCES 1. Karanark, C.J., Loevinger, R., Steele, R.E., and Weissbluth, M., (1960): Radiol- ogy, 74, p.633. 2. Tetenes, P.J. and Goodwin, P.M. (1977): Radiology, 122, p.219. 3. Williams, P.S., Hunter, R.D. and Jackson, S.M. (1979): Brit. J. Radiol., 52, p.302. 4. Mandelzweig, Y., Tatcher, M., and Yudelav, M., (1980): Proc. of Vorld Congress of Radiation Protection, Jerusalem, p.291.

Otstince 0 _._«

DISTANCE Icml >. i Resultant profile when 2 beams are joined. VII-21

RECENT DEVELOPMENTS IN THE RADIATION STERILIZATION OF PHARMACEUTICALS

Geoffrey P. Jacobs (Ya'akovi)

School of Pharmacy, Hebrew University of Jerusalem, P.O. Box 12065, Jerusalem, Israel 91 120

The application of gamma rays to the sterilization of disposable plastic medical devices is now commonplace. The spectacular progress of radiation sterilization can be attributed to the availability of large cobalt-60 radioisotope sources, a vast spectrum of radiation-stable plastic polymer-based disposable medical supplies and packaging materials, as well as the unique features of a cold steri1izatio" method using deeply penetrating radiation. To date there are more than 130 commercial scale gamma irradiation facilities commissioned in some 40 countries, and radiation sterilization of medical products is now a well established industrial process. Proceeding and concomitant with these rapid developments, the research efforts expended in the study of the underlying phenomena of the processes, their effects on microorganisms and on the chemical and physical characteristics of the constituent materials, monitoring of facilities, and public health assessment from use of the finished products, appear to have exceeded by far the collective efforts directed to all other sterilization r/.ethods.

The use of gamma irradiation for the sterilization of Pharmaceuticals is gaining momentum. The increased acceptance of radiation sterilized drugs is no doubt connected to two recent developments. First, there has been an increased awareness of the ecological problems associated with ethylena oxide sterilization. Recently the U.S. Occupational Safety and Health Administration (OSHA) has reduced the maximum permitted daily personnel exposure dose of ethylane oxide to 1 ppm. This has prompted many large scale users of this mode of sterilizaion to revise their production methods and in some cases even abandon the system.

Secondly, there has been a legislative breakthrough in the acceptance of irradiated food. Following the recommendation of the WHO of the acceptance of food irradiated up to an average dose of 10 kGy, the US Department of Health and Human VII-22

Services approved in December 1985 the FDA rule to broaden the approved uses of food irradiation. The above legislative considerations, coupled with the recognition of irradiation as a highly effective and reliable sterilization process for, inter alia, thermolabi1e materials, has certainly added impetus to its acceptance. There are o the market today quite a number of radiation sterilized ,.%ugs. These include ophthalmics, topical ointments, ii ectables and veterinary products.

LEGISLATIVE CONSIDERATIONS Whilst sterilization by exposure to ionizing radiation has been a recognized method in the United Kingdom since 1960, medicines controlled under the Medicines Act 1968 are subjected to individual assessment by the Committee of Safety of Medicines. This Committee requires proof of sterility assurance, that the potency of the drug is unaffected by the process, and that any degradation products are not harmful. Similarly, whilst the British Pharmacopoeia recognizes gamma irradiation as a suitable sterilization process, it is the responsibility of the manufacturer to prove that no degradation of the product has taken place. Similarly in the United States, the FDA looks upon a radiation sterilized drug as a "new product", with the onus upon the manufacturer to prove ibs safety.

CHOICE OF STERILIZATION DOSE Regarding choice of sterilization dose, the U.S.P. states: "Although 2.5 Mrad of absorbed radiation was historically selected, it is desirable and acceptable in some cases to employ lower doses for devices, drug substances, and finished dosage forms. In other cases, higher doses are essential. In order to validate the efficacy particularly of the lower exposure levels, it is necessary to determine the magnitude of the natural radiation resistance of the microbial population of the product." According to the British Pharmacopoeia (1980): "... A dose requirement of 25 kGy (2.5 Mrads) is generally accepted as adequate for this purpose although other dosage levels may be employed provided that they have been authorised by the appropriate authority. If lower doses are used additional controls to those normally applied may be necessary in order to assess the adequacy of the procedure; such controls will VII-23

include additional microbiological monitoring of the product before and after exposure." An approach similar to that of the British Pharmacopoeia has recently been approved by the Ministry of Health in Israel. At a recent I.A.E.A. Advisory Group Meeting on the Code of Practice for Radiation Sterilization of Medical Supplies (Colombo, November 1986), a compromise approach to the selection of the minimum sterilizing dose was adopted.

IRRADIATION OF WATER

Studies on the feasibility of radiosteri1izing water in a variety of container materials have been carried out in various laboratories (Du Plessis, 1977; Hilmy & Sadjirun, 1975; Jacobs and others, 1977). The main drawback has been that irradiated water fails the limit test for oxitHzable matter stipulated by the European Pharmacopoeia because of hydrogen peroxide formation. However, this official specification is probably intended to limit the content of organic impurities.

DECONTAMINATION BY IRRADIATION

Many powders used in the pharmaceutical and cosmetic industries are heavily contaminated with microbes because of their natural source, and thus present a health hazard to the patient or user. Such powders will often not withstand heating processes to reduce the initial microbial load, and ethylene oxide treatment is becoming increasingly stringent and problematic. In such cases low radiation doses (less than 10 kGy) may be quite sufficient to reduce the bioburden by several orders of magnitude.

RADIOSTERILIZATION OF DRUG POWDERS

Our own investigations on the radiation sterilization of drugs has focused on powders of the beta-lactam antibiotics (for example, Jacobs, 1983). The rationale for these studies is based on the susceptibility of the beta-lactam antibiotics to hydrolysis, particularly at elevated temperatures, which eliminates sterilization of injectables by conventional methods such as autoclaving. The necessary practice of sterilizing powders for injection by techniques that involve costly and highly demanding aseptic processes, makes sterilization by irradiation most desirable. These powders may be sterilized by gamma irradiation, having been packed VII-24

under clean (as opposed to sterile) conditions in the final contai ner. Even when radiolysis products are within the acceptable compendial limits, it has to be conclusively established that they are not toxic. However, it has been shown that such radiolysis products are generally not unique to irradiation (for example, Kane and Tsuji, 1983).

MINIMIZATION OF RADIATION DAMAGE Radiolysis can be minimized by choice of appropriate treatment conditions such as irradiation in anoxia or at low temperatures, or alternatively (or in addition) by incorporation of suitable additives, providing that degradation pathways are known. Such additives must not be toxic or interfere with the efficacy of the drug. This may be achieved by the use of energy transfer systems, -SH containing molecules, scavengers of radiolysis products of water, or reagents that convert radiolysis products to the parent compound. Our observations in this field are being applied to aqueous Pharmaceuticals whose sterilization or decontamination is problematic.

REFERENCES Du Plessis, T.A. (1977). The Radiation Sterilization of Pyrogen-Free Water in Polyethylene Sachets, Report PER—17— 1. Atomic Energy Board, Pretoria, Republic of South Africa.

Hilmy, N.t and S. Sadjirun (1975). In Radiation Sterilization of Medical Products 1974, IAEA, Vienna, p.145. Jacobs, G.P. (1983). Int. J. Pharmaceut., 17, 29. Jacobs, G.P., M. Oonbrow, E. Eisenbera, and M. Lapidot (1977). Acta Pharm. Suec., 14, 287. Kane, M.P., and K. Tsuji (1983). J. Pharm. Sci., 72, 30. VII-25

PORTAL FILM CHARTS FOR A 6MeV LINEAR ACCELERATOR

Sergio FAERMANN#, Yehiel LESER*. and Eli REGEV,*#

#Dept. of ONCOLOGY, SOROKA MEDICAL CENTER, BEER-SHEVA *Dept. of ONCOLOGY, BEILINSON MEDICAL CENTER, PETAH-TIKVAH

1. Introduction

The portal film technique plays an important role in the patient set-up for radiotherapy treatment, due to the frequent localization errors arising as a result of incorrect positioning or incorrect machine parameters . Altough the portal films obtained with high energy radiotherapy machines suffer from an inherent poor image ,as compared to that obtained in conventional radiography, they provide valuable information for the radiotherapist and the technician.In order to obtain high quality portal films, it is necessary to choose the optimum film exposure. This depends on various parameters such as patient thickness, patient to film gap field size, SSD, etc. For each set of parameters the optimum dose in cGys at the film location is measured and the generated tables are called portal film charts.The present work describes an analytical method to construct these charts for a 6MeV linear accelerator .The selection of the best film-screen combination was greatly simplified by adopting the combinations used in industrial MV radiography for years.Also,a method to avoid the influence of the processor parameters on the film optical density is sug- gested. Both localization and medical films were used. VII-26

Theory The analytical method is based on the formalism developed by Khan(1). The calculated total dose is obtained by

2 = Dpr(1+[TMR(F,d)-TMR(0,D)]) (SSD + d ) (1) (SSD + d+g)2

Dpr = the primary dose at the film position TMR(F,d) = tissue-maximum ratio for a field size PxP and the depth "d" P = side of equivalent square field, defined at a distance SSD + d from the focus. TMR (0,d) = tissue-maximum ratio for a 0x0 field and depth "dM in water SSD = source- to- skin -distance g = air-gap between the patient's exit and the film position.

3. Experimental Two types of films were used: CURIX RP1 medical film (AGFA) and X-OMAT TL localization film product (KODAK). The selection of the metal intensifying screens were based on the criteria of Domanus(2) that requires front and rear screens of copper and tungsten with 0.5 mm minimum thickness. In our case we used front and rear screens of 0.5 mm thickness copper. The range of optical densities of 1.3 to 1.8 is considered to give the best accepted portal film images (3,4) and an optical density of 1.6 was selected for comparison and calculation purposes. The dose at the film position was measured with a Farmer Dosimeter model 2570 (Nuclear Enterprises) coupled to a 0.6 Farmer ionization chamber and a suitable build-up cap, in air. The film readings were performed with a Victoreen Digital Densito- meter Model 07-424. Extensive measurements were performed with a water phantom and different imaging geometries, with the ion chamber and build-up cap attached to the position of the film cassette. V1I-27

4.1 Sensitometric curves The characteristic curves for both films, with front and rear screens, are shown in Fig. 1. The solid lines represent the best fitted curves obtained using a least square fitting procedure. The equations found were: O.D. = 2.612(1 -exp[0.556 D] for the Curix RPI film (2) O.D. » 3.502(1 -exp [0.0941 DJ for the X-OMAT TL fi!m (3) where : O.D. optical density (including base and fog); D - dose in the film (cGy) leading to an optimal dose of D = 1.7 cGy for the Curix RP! film and d = 6.5 cGy for X-OMAT TL film, at the O.D. = 1.6. 4.2 Measurements with water phantom The calculated dose with to = 1.5 cm and SSD = 100 cm (according to eq.(1)] and the measured ones (in cGy/MU) are presented in Table 1. The quoted errors are one standard deviation of the mean measured values. Excellent agreement between the theoretical and experimental results was achieved with a maximum discrepancy of 15%. 4.3. Portal Film Charts. A derived portal film chart is presented in Table 2 for the X-OMAT TL film. The doso is expressed in monitor units (MU). Also presented (in parenthesis) are the partition doses for the double exposure technique. It was found that allowing 1/3 of the total dose for the portal radiography and 2/3 to the bigger field provides the best result, confirming the findings of Droege and StefanakosW. 4.4 Processing considerations Because processing conditions have a strong effect on the developed density, it is necessary to check the sensitometric curve on a daily basis^5). The sensitometric curves are normalized to a reference dose, for example, the dose that produces an O.D. = 1.6 on the film, its parameters remain insensitive to developing changes. Then, by adopting a curve like: A (1-e-BD/D1.6) will yield optical density

O.D. = 2.612 (1-exp [-0.948D/D1 6]) and O.D. = 3.499 (1-exp [0.61D/D1 6]), for the CURIX RP1 and X-OMAT TL film respectively. VII-28

Bibliography 1. Khan, F.M. "The Physics of Radiation Therapy" - p. 182, Williams & Wilkins, 1984. 2. Domanus, J. "Industrial Radiography and Related NDT problems". Report No. B. 307 (1973) - Danish Atomic Energy Commission. 3. AAPM Report No. 13. "Physical Aspects of Quality Assurance in Radiation Therapy"-May 1984. 4. Droege, R.J. and Stefanakos, T.K. Int. J. Radiation Oncology Biol. Phys., Vol. 11, pp. 2027-2031. 5. Reinstein, L.C. Persona! Communication. Nov. 1985. VII-29

Characteristic Curves-Mevatron 6 MEV LIN. AC

25- 2-«-KODAK X-OMAT"

i i 3 4 5 6 h ib DOSE (rads) VII-30

TABLE 1. MEASUREMENTS WITH WATER PHANTOM AND COMPARISON WITH CALCULATED DOSES AT THE FILM LOCATION, SSD = 100 cm

Field Size Phantom-to-Film Water Measured dose Calculated FxF(cm^) Distance,g(cm) Thickness (cGy/MU) Dose (cGy/MU) g (cm) d(cm) D™ D'FC

10 X 10 2.75 10 0.567+0.0014 0.572 15 X 15 2.75 10 0.600+0.0027 0.587 20 X 20 2.75 10 0.619+0.0022 0.604 30 X 30 2.75 10 0.636+0.0022 0.624

5 X 5 3 15 0.358+0.0013 0.375 10 X 10 3 15 0.415+0.0015 0.410 20 X 20 3 15 0.462+0.0016 O.44O 30 X 30 3 15 0.479+0.0017 0.455

5 X 5 3 20 0.258+0.001 0.270 10 X 10 3 20 0.305+0.0011 0.296

20 X 20 3 20 0.351+0.0012 0.320 30 X 30 3 20 O.367+O.OOI3 0.330

5 X 5 20 20 0.177+0.001 0.204 10 X 10 20 20 0.197+p.OOl 0.223 20 X 20 20 20 0.232^0.001 0.238

30 X 30 20 20 0.254ip.001 0.245

10 X 10 20 15 0.267+0.001 0.305 30 X 30 20 15 0.328+0.0011 0.335

10 X 10 10 15 0.333+9.0012 0.361 30 X 30 10 15 O.4O6+O.OOI4 0.399 TABLE 2. PORTAL FILM CHART FOR THE X-OMAT TL KODAK FILM,*,**

FIELD SIZE

5x5 i 10 x 10 15 x 15 20 x 20 30 x30

3 10 20 3 10 20 3 10 20 10 20 10 20

10 13 14 17 12 13 16 11 13 15 11 12 15 10 12 14

(4+9) (5+9) (6+11) (4+8) (4+9) (5+11) (4+7) (4+9) (5+10) (4+7) (4+8) (5+iO)|(3+7) (4+8) (5+9>

18 20 23 16 20 24 21 14 17 20 14 16 20

(6+12) (7+13) (8+15) (6+10) (7+13) (8+16) (5+10) (6+11) (7+14) (5+9) (6+11) (7+13) (5+9) (5+H) (7+1!

20 25 32 38 21 25 33 21 24 28 19 23 28 18 23 28

(8+17( (11+21) (13+25) (7+14) (8+17) (11+22) (7+14) (8+16) (9+19) (6+13) (8+15) (9+19) (6+12) (8+15) (9+1

* THE DOSES ARE EXPRESSED IN MONITOR UNITS (HO) ** DOUBLE EXPOSURE DOSES BETWEEN PARENTHESIS Authors Index

Adir, J., VI-3 Kahane, S., V-33 Alfassi, Z, B., VII-l,VH-5 Kaizerman, S., III-9, 111-42 Balmor, Y., 111-32 Kalish, Y., VII-9, VII-10 Barnea, Y., Ill-19 Kami, Y., V-21 Baer, G., VI-18 Ketter, A., IV-13, IV-28 Barak, A. Z., IV-5, V-25 Klaiman, D., V-29 Ben-Haim, Y., 111-28, VI-14, VII-14 Koi, R., 1-14 Blumenau, L., V-25 Kovacs, S., 111-32 Branover, H., V-25, V-29 Kroger, W., 11-22 Bushlin, Y.. VI-39 Kushelevsky, A., VII-14 Carmona, S., V-8 Lavi, A., VI-18 Celnik, J., V-45 Ledertnan, L., II-8 Cohen, B., VI-20 Loewenstein, W. B., 1-4 Covaliu, Z., 111-28 Leibson, M. J., V-42 Dagan, R., 111-14 Leichtet, Y., VII-14 Deutsch, M., VI-31 Leser, Y., VII-25 Dickman, S., III-9 Lynn, J. E., V-33 Dickstein, P., VI-25 Mandelzweig, Y., VII-18 El-Boher, A., V-25 Marouani, D., 111-40 Elias, E., III-9, 111-14, 111-19, 111-42 Merlis, Y., VI-9 Faermann, S., VII-25 Moody, F. J., 1-9, II-7 Gal, D., 111-45, 111-49, IV-17 Nabielek, H., 11-23 Gozani, T., VI-3 Nagler, A., III-l, KI-5 Green, M., 1-14 Ne'eraan, E., VI-18 Greenspan, E., IV-1, IV-13, IV-24, V-21, Nickel, H., 11-23 V-25, V-29, V-38 Notea, A., VI-31. VI-39, VI-43 Gutman, A., 111-32 Ofek, R., V-16- Hasan, D., 111-42 Olander, D., 1-12 Hirshfeld, H., III-l, III-5 Olek, S., 111-23 Ingman, D., VI-1, VI-9 Pal, D., VI-31 Itamar, A., VI-18 Pott, G., 11-23 Jacobs, G. P., VII-21 Raman, S., V-33 Rambam, M., 111-37 Sonneck, C, II-4 Regev, E., VII-25 Spero, E., V-25 Reznik, L., 111-35,111-37, VM Spiewak, I., IV-9 Riklis, E., 1-14 Sukoriansky, S., V-25, V-29 Ron, S., IV-13, IV-21JV-24 Szabo, J., IV-13, IV-28 Ronen, Y., V-12, V-42 Tacher, M., VII-18 Rothenstein, W., V-l Talmor, A., VII-14 Saphier, D., 111-45, 111-49, IV-17 Tepper, L., 1-1, 111-35, IH-37 Sapir, D., VII-18 Tricher, F., VI-43 Schenk, W., 11-23 Tsechanski, Y., V-16 Segal, E., VI-25 Vulkan, U., VI-14 Segal, Y., VI-20, VI-25 Wacholder, E., 111-14, 111-42 Segev, M., V-4, V-8 Weiman, S., 111-40 Shai, I., 111-19 Weinreich, R., VIM, VII-5 ShanJ, G., V-16 Yiftah, S., 1-6, III Shayer, Z., V-38, V-65 Yudelev, M., VII-18 Smith, J. P., VIM