Behavior of Two Big Rockfill Dams, and Design Aims

Missouri University of Science and Technology Scholars' Mine

International Conference on Case Histories in (1984) - First International Conference on Case Geotechnical Engineering Histories in Geotechnical Engineering

11 May 1984, 8:00 am - 10:30 am

V. F. B. de Mello ISSMFE

Follow this and additional works at: https://scholarsmine.mst.edu/icchge

Part of the Geotechnical Engineering Commons

Recommended Citation de Mello, V. F. B., "Behavior of Two Big Rockfill Dams, and Design Aims" (1984). International Conference on Case Histories in Geotechnical Engineering. 4. https://scholarsmine.mst.edu/icchge/1icchge/1icchge-theme9/4

This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

This Article - Conference proceedings is brought to you for free and open access by Scholars' Mine. It has been accepted for inclusion in International Conference on Case Histories in Geotechnical Engineering by an authorized administrator of Scholars' Mine. This work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the permission of the copyright holder. For more information, please contact [email protected]. Behavior of Two Big Rockfill Dams, and Design Aims· V.F.B. de Mello Professor, D.Sc., Consulting Engineer, President ISSMFE

SYNOPSIS The case histories of two major compacted embankment dams are analysed with regard to the problem most amenable to reasonings and computations of statistics of averages, the problem of deformations, discussed regarding specific laboratory testing compared with use of generalized cor relations. Improved mental models for predictions are proposed. In the earthcore-compacted rock fill section the problem of crest cracking suggests the interest in a significant change of zoned section in the topmost stretch. For the concrete-face rockfill dam it is suggested that one needs a significant revision of instrumentation and monitoring orientation. Both dams behaved extraordi narily well on questions of consequence and served to show that unfavourable observations, if too indireqt, lead to no benefit, but may sometimes prove the inexistence of the consequential misbeha vior to be guarded against.

1. INTRODUCTION that in all probability the threshold of unde sirable behavior will not be crossed. Most Over the past few years I have deepened my often the self-same material must satisfy two convictions, and repeatedly emphasized, that opposite limits: for instance, a compacted clay the basic concepts of design of important civil core must ensure that it will not suffer differ engineering structures comprise five steps ential settlements higher than a certain value, sununarizable as .follows:- whereupon the design decision is based on the upper confidence limit of compressibilities not 1) firstly seek and adopt for the aesired being exceeded; the self-same compacted clay functional need a type of structure that phy core may, in certain zones, be sought to ensure sically precludes behaviors of significance that brittleness and rigidity are minimized, subject to the statistics of extreme values, and, therefore, for such zones the compres truly unquantifiable, especially if and when sibility (often considered to accompany plastic capable of progressive accelerative degener deformability) may be specified not to fall ation towards failure (de Mello, 1977). Phe below the lower confidence limit. Thus, the nomena such as the weakest link tension failure engineering decisions may be said to be based case should, in a good design, be precluded; on prediction of what will most probably not not by increasing nominal computed factors of happen, rather than on the direct aim at safety, which always generate extremely un computationally reaching more accurate pre economical conditions and often are truly il diction of what will presumably happen in lusory, but rather by choice of change of reality (deterministic) . physical universe, so as to work with other statistical histograms of definable limiting (2.3) the basic needs for design computations quantification. are, therefore, to improve the predictability on average parameters and adjusted methods of 2) having chosen the functional model, whose computation, and to narrow the width of our significant behaviors are dictated by the confidence bands, as rapidly and efficiently as statistics of averages, engineering acquires possible. the right to meaningful calculations, and a four-pronged attack occupies the gross of For any forthcoming job this is achieved on the apparent energies of engineering workers:- basis of well digested data from monitored projects; and for any given project -under con (2.1) the prediction of what is likely to be struction, by rapidly adjusted iterative de the behavior is associated with the statistical cisions by Bayesian prior-to-posterior proba average behavior, equated, adjusted, and extra bility adjustments from greater intensity test polated: the ultimate aim would be to be able ing and monitoring right at the bottom, for ju t.o predict "exactly"; adjustments and extra dicious extrapolation to the nevralgic con polations reqt~ire "laws", that is, repetitive ditions when the job is crowned and subjected responses to meaningful index parameters; to first-filling. (2.2) however, for many inexorable realistic reasons, the engineering design decision is (2.4) finally, it recognizedly often happens based on estimates of the unfavourable "proba that cumulative experience ("common sense" being bility limits" (let us say, the 90% confidence a. cumulative intuition developed by Bayesian bands) around the average, and on the guarantee "natural selection") on complex lumped para-

923

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu meters of prototype to prototype comparison decisions, and as leading to suggested revision will yield better design decisions and behavior for future cases. estimates (Hynes and Vanmarke, 1977), than a specific new set of test data and/or design Note that both cases also supplied very computations. interesting data regarding grouting treatments and efficiencies in sealing horizons of open 3) thus it is that the analysis of case-histo -fractured rock: and it is recognized that the ries acquires significance not only with regard foundation treatment problem is another one of to the sensational attention to patent failures the dominant ones in such dams; however, for (indicating physical models to be guarded convenience in concentrating on a single mes against) but also with regard to the large sage, the paper will not touch on these facets, silent majority of cases of varying degrees of which were most successfully handled. misbehavior which will permit quantifying the decision thresholds. The intent herein is to consider the case his 2. BASIC DATA ON SOME COMPACTED EARTH DAMS tories of two recent dams of about 160m in height (Figs. 1, 2) and impressive data. It may Brazilian experience on test compressibilities be accepted that both the dam sections are of and monitored compressibility settlements of chosen physical models well conceived to avoid dams ranging between 50 and 160m has often sensational failure, and therefore it is enough been published case by case, generally understandable that attention has concentrated showing that observed settlements have been on problems of deformations which have in significantly smaller than those predicted. various projects generated varying degrees of Table I summarizes the principal data on the misbehavior. Deformations result from integrat dams under consideration and settlement ed effects and are therefore well included in predictions based on oedometer tests. It may be the group dominated by the statistics of remarked that the principal concern arises from averages. The paper concentrates on such the fact that the ratios of observed: predicted problems, both for predictions and design (cf. Fig. 19) or differences of predicted minus

TABLE I

HEIGHT SOIL CHARACTERISTICS COMPACT ION ENO•CONSTR KEY IN (m) · FIGS.(TESTS STANDARD PROCTOII ( Fl ELD) SE;r,T,;~M DAM SOIL GEOLOGIC

TYPE ORIGIN "') •/em" 'Sd mGa (0 ') ( 0 EST WL (o/o) lp (,. ~(. • ,-, ( g/eJI Wopt 1o PC Yo ~~~~* T

~-101 SAL TO RED SILTY BASALT :56/:52 70 - 8:5 2:5 - 3:5 2,9- 3,0 1,3 - 1,~ 34• 38 -I I +I HILF 41 54 0 OSORIO CLAY DECOMP. 100 • SAL TO RED SILTY 80/80 BASALT 62 • n 17 • 22 2,t ·3p 1,35 -I,~S ~ • 34 97 • 9B 0 1+1 HILF 81 244 fj, SANTIAGO CLAY DECOMP.

SANOY CLAY IIOTITE INEISI 65 - 73 30 • 3S z,n-2,83 I,S • 1,6 23 • 26 PARAIBUNA ~4/61 99 -1/+o. 30 72 - Q SANOY SILT RESIDUAL SOIL ~4 ·:51 NP • 23 2,69·2,79 1,6 • 1,7. 16 • 19 ~~

SANOY CLAY IIOTITI tNEI .. 53 • 87 23- ~2 2,78·2,8E 1,42· 1,6~ 26 - 29 ~ STD 104/64 PARAITINGA 99 2 47 7t 0 SANDY SILT UIIOUAL lOlL 34 ·48 NP-19 2,69·2,791,64•1,80 14 • 20 ' ·II+C1,5fAOclai • STD SANOY SILTY BASALT CAPIVARA 60/60 ~ • SO 15 • 20 2,83•2,9 1,:55- I~ 20 • 25 ~7 • 98 +1 I+2P~ 57 ~8 liil CLAY DECOMP. IAIALT DECOIIP. SANOY SILTY E~BORCA<;AO ISO/ISO ...D IUtA.NITI- 42 • 52 15 • 25 2,6 - 2,8 1,55 ~1,75 20 - 25 101- 104 -1,510 HILF 180 285 CLAY v 8NIISS OECOIIII'. ' SANDY SILTY BASALT 911· KlO STO AGUA 63/47 38 • 45 13 • 19 2,8$- 2,9! - 21 ·0,61 +I 10 45 CLAY COLLUVIAL 1,6 • ~· 11 99,7 ji'IIXTCJ 0 VERMELHA • 91- 100 ITAUBA 93/93 SILTY CLAY BASALT 50 • 80 30 • 40 2,7· 2,115 1,3 - 1,45 211 - 33 +1 I +2 HILF 92 240 X DECOMP. 98,5 - SILTY CLAY PASSO R"EAL 54/54 ANO BASALT so • 70 10 - 22 2,65-~ 1,3 - ~4 32•34 97- 98 -1 I +2 HILF 40 9ll - .. CLAYEY SILT DECOMP.

MARIMBONOO 70 SILTY CLAY BASALT 29 - 35 12 • 20 2,80 1,83- 1,96 IZ • 15 -- - - - DECOMP. - - e MAJIIITI~-1:11. HILF+ PEORA 00 SANDY SILTY 130/95 IIICOMII. AND 1M 32 - 40 1,8 • 1,9 • 16 12 - 22 211!1'2;~ 12 98 ·101 -21-10,1 60 86 ~ CAVALO CLAY lt!IIIIAI~ ~ " SAO SIMAo 127/64 CLAYEY SAND BASALT 25. 40 10. 22 Z,T • 2,115 1,8·2,0 II - IS 100· IOl -~2/·o,2 HILF 28 31 - DECOMP. + SILTY CLAY 62. 72 S. SANTIAGO 17 - 22 2,9. 3,0 1,35 -1,4!5 33- 38 97. 98 If>/+~ HILF 60/60 BASALT 77 11'11 IAUliiLI.U DIQ) - -- CLAYEY SILT DECOMP. 55- 64 17 - 20 2,1111- 2,9!! 1,31!1-1,45 33-38 97- 98 2PI•Z.6 HILF

* OBSERV. • SPECS CONVENTIONALLY WRITIEN AS < llW< PRESENTLY PREFERRED AS 0,95< WCQMP < 1,05 (exl WOPT

924

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu ROCKI"ILL LEV£L (M) : CONSTRUCTION ~ WATER LEVEL ( "' ) 710 1-- rr- no v-- 700 _.../ / 110 ~ ~ I / v j I-- 140 .-/ v triM E ls--0 N D J F M A M J J A S 0 N D .I I" M A lot J J A s 0 N D ITr ME 1100 A- M- J- ,j- A- s- 0 ~1 f':,.- ---f-.-:-t-r" 10 ...... _--U !' _\ "~ "-...... :: ,.. ,.12 "--'-. 1-- ' ~ 10 I\: ~·---- ["\ ...... '\ 1:--- (2-8) tO ' '10-15 f"'"' I IL 1'41- ~\ - ~\ 40 ~) 100 '--=' I~ ._1 '"\, "\ " 1'- "! 1'.. t--- ~ 18 '- eo ...... , -- 140 ~ f- 38 I ------~- f:::::: ~ - [~ ~ - , I:S 115 10 (2·8),(F-2): COMPR SSIO S 8Y \.il.ft_ COMPMS. :53 TYPICIU. rD IS PLACE MEN TS c• SECTION DIFFERENCES OF (CIII) 7 •8 tO SETTLEMENTS (NORMAL TO FAICE,COMPUTED FROM VERTICAL 27 ·2 SETTLEMENTS MEASURED). ,~ 1 I ' F 1 -

Fig. 1 - Summary Data, Foz do Areia Dam

~bserved settlements accentuate in cases of laboratory and field data on compressibilities ~reater settlements,thus indicating errors of for the purpose of improving the basis for trends and extrapolation, in contrast to er design estimates. ~atic errors (statistical dispersions) ; since 1ifferential settlements have been associated In previous papers (de Mello, 1977; 1980) it has ~ith generating cracking misbehavior at the been emphasized that the most fruitful mental vulnerable top, the noted trend is of definite model for interpreting behaviors of compacted Unportance to design and performance. clayey soils is to consider compaction as equivalent to producing a certain preconsolidat rhe first part of the present study is thus ion pressure cr'p. Such a preconsolidation pres lirected towards the statistical digestion of sure should be directly related to the percent

EMBANKMENT LEVEL(m) WATER LE'vf:L(m)

SWEDtsH- 80X SETTLEMENT MARKERS ll 7·8 , 10- B TRANSITION SECTION • 12-B, 13·8, 16·8 ROCKFILL

SETTLEMENT (CIII) 0 8·8 CORE • DAMAGED AS INDICATED Fig. 2 - summary Data, Emborca~ao Dam

925

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu compaction PC%, and, presumably, within the nar deformations varies as one moves from sandier row ranges of PC% in which a given project to more clayey behaviors: and, on the other works, beyond the cr'p value the nominal virgin hand, the capacity to retain such precompres compression index Cc would be essentially sion benefits also varies as a function of the constant. Within such idealized preliminary expansivities on release of applied stress, and reasonings the following analyses and im retentions of pore suctions that avoid heave provements are submitted. Some second-order (themselves varying from sandier to clayey adjustments are discussed as appear justified. soils}.

3. STATISTICAL CORRELATIONS INVESTIGATED 3.1. Compaction Preconsolidation Pressures The available data have been analysed separately In Fig. 4 we begin by analysing the most clayey in soil groups as distinguished by the re soils (ydmax < 1,5 t/m3) in which the above spective maximum dry densities. In previous -r.lentioned hypothesis would seem most applica papers (de Mello, 1977; 1980) it has been shown ble. All the available data were used for a that the attempt te relate soil types to best fit semilog linear statistical regression. Atterberg limits as appropriate index tests It is indiscriminate and consequently spurious: would be much less fruitful. Fig. 3 reproduces firstly because there was no attempt to employ generalized data on compression indices Cc a theoretically suggested reasoning for the against a background of hypothetical correla inclination of the semilog linear statistical tion with liquid limits; one should note that regression; secondly, because there was pur the more "plastic soils" (higher liquid limits} posely no attempt to distinguish between speci have systematically given much smaller set mens molded in the laboratory and those trim tlements than would be inferred from routine med from undisturbed block samples from the classifications evolved from saturated sedi compacted fill, thus disdaining the fundamental ments. The reasoning concerning compaction is tenet that everything is different unless and that on the one hand the absorption of pre until proven reasonably similar. compression effects by volumetric vs. shear

Cc 0

I I I [J I

9, +I ETC ... SEE TABLE I FOR .JOB IDENTIFICATION ANO SOIL DATA.

0 20 80 ao

Fig. 3 - Routine Attempt to Infer Cc Compressibility from Liquid Limits

926

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu 3 ~quation extracted from fewer data as published % COM~IONI '((dmox ~ 1,5 t/m J · J (P.fl . J by de Mello (1980), which is herein conside.;;ect revokable. 104 PC• 57,06 +58,82 loo~'p PC=~ 38+6 43 Often we are led to insist on tests for- a spe 102 c~fic soil and project and to adopt conserva t1ve parameters based on specific test results. Such_a directive would suggest the attaching of spec1al importance to the confidence band on single values. In cases such as that of compressibility of porous granulates, especially as reflected by specimens of truly small di mensions, it is evident that individual results may be subject to a wide scatter that would never be reflected in bigger masses of the prototype. Thus, experience from many projects may lead to much more valid predictions: in statistical terms one should place much more faith in confidence bands on the averages. Theoretical idealizations are required for ar riving at more meaningful best fit regressions, and moreover, a judicious decision must be made regarding applicability or not of dispersions 9 of averages, or of individual values. Therefore 2 " 6 8 one should conclude that although each ad ALL DATA : UNDISTURBED.+ MOLDED SAMPLES ditional case-history tested for design, and monitored for interpretation, cannot but be Fig. 4 - Statistics for cr'p, Most Clayey Soils required, it must be more wisely taken as of The figure is submitted on purpose, showing fering its contibution to the appropriate firstly the widely different regressions that statistical evaluation of desired behaviors, can be obtained. Secondly, it is intended to than as necessarily indicating specifically ap emphasize the big distinction between confidence plicable design parameters irrespective of the limits on averages and those on "single values". global experience digested until that moment. I submit the admonition that in order for mathematical statistics to lead to fruitful re However, I shall discuss below the relevance of sults, it is indispensable to use all possible single layers in affecting differential settle indications that separate deterministically ments and cracking at the top of the dam section. distinct universes, and also to use available Therefore we should distinguish between indi "laws of behavior" to preestablish the type of vidual test points and the averages extracted appropriate equation for desired best fit. On from individual tests (based on dispersions of the same figure,I also show a regression test values within a specific lift). The aver-

0 P.C.(%) v ,.. 102 le V/ ,.. c .-c) . c •01 c". kiv/// . o,z~;e..::,.. ------1;-~ --·-·!--

100 1-' / ----- ·I-- ~Jf ·-·71\: c1o;, c // ( UHDIS:U~• FROM FIG.&) • ! ,, .. " • v: ' x ll.e • .... t'-• rx-:- ...!.._ - ...!.._ =.lJl.mn . •• PC, PCz fs I //"1 "--- ' e do,zec 7 ' ' !r' I PC1 --- I ' .i .i' ' ' ~z:s:-- Iff I ' I PCz --·- 94 ------_,_- _..,..__. :' I''\_ ~i"7-'' P.C : logU G'PI UP& • I j j j l (i 'p(kQA:Jl 0 4 • • 10 Fig. 5 - Theoretically Oriented Statistical Regressions for Compaction cr'p, Molded Speci~ns

927

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu ~ I'W.l Cr • O.ICc P.C. ( 0 Yol '( d max 40 1,5 t lrl Cr• 0 THEORETICALLY ~ OIIIENTEO: 102 I{'''(II l Cr... •_C!JCe , V ~ ·-~:.t :/ o c• ~. ~ a .... -~ .... a':~ , ~- a / 0 , Vr 0 aV / Cc (A) 100 O,t ..., 0 0 ·~/ \ _,_ Cel IIC, • __ I• 0, II , 4j (.. Ccl -- "a·, v \ I' A.i ~ , I , o••• Ill y , ' a ' ,, C.l P.C. . ITA• II IC; I. ' 1•0,11 ~ / Cc2 -----'t ' ' ..... 98 • o·, •• I • 4 ,, 1 4,00 ~" I ..... } ...... o' i ' 0_/ 1 64 .. 1 .. ~ I: al;t ~- ·: J/' ~diiiCil. I 96 V) Col P.C:. I I·C 44 O,t /! c f.c.l £;/I • I" ' I y '; .., -: ~viN • t(F ~I .lj / l(~ ~~ ·~~ 141 p.C: •• M ~~ ••• * p 14 / J f.c.2 -~~~~ Cc•f ( C) ,:Jl I a// (A) c P.C. 104 rr /;'' u~ kQ/C ~) •• p(k911~ 92 0 2 4 I I 10 I 2. 4 6 a 10 Fig. 6 - Theoretically oriented regressions on undisturbed block specimens, comparison with Fig. 5

age behavior of a given lift definable from a second order. limited number of tests from that specific lift would present less dispersion than the indi To cover the range of materials of more common V.idual tests themselves, while maintaining a use the same types of analyses were extended to broader dispersion band than obtains from many cover the moderately sandier materials (Fig. 7, tests on many similar lifts. The use of dif 1.5

928

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu !MOLDED I UNDISTURBED P.C.. (Oj,)0 P.C.(%1 " ,/ _...~ 04 ,/ v~" v'lv . 02 02 )EJOI j,./ I/ ~ / v v 00 o/ _). 00 v 16 v" - , v· 1/ ; l:t /" I ~ Po / / )a ~ / •• • ·y , 1/ ~ •• / // ee ~ je ~~ c ~ 6 • 6 / • j;;" i / • / I tot I ~ t~ / 1/ - - I 0 2 • ~J'd IIIGll>ll t/•'1- •2 ·.o II I O"'p( g/an2 a'pJ Kg/cm1l •" I ' 4 6 • 10 20 4 I 8 10 20

Fig. 8- Comparative Oriented Trends for Sandiest Materials- ydmax > 1.7 t/m3 3 • 2 • Comp ress~on· I n d"~ces Cc a ·more appropriate index test for classification Having adopted the mental model of precompres- of the behaviors traditionally associated with sed behavior for compacted materials and more sandy vs. clayey materials. The same satis- having established the statistical e~timates of factory type of equation published previously a'p = f(PC), it now behoves us to establish the was reused directly, Fig. 9. Subsequently an statistical regression for the Cc compressibili- exponential form was tried, but did not result ty. Our experience has favoured the use of the in improvement. Finally, because of the fact St~ndard Proctor ydmax of the various soils as that most of the very clayey materials (ydmax <1.5 t/m3) were red residual soils of

Cc o,sor------~------_,------~------+------~------~------+------~------+-- •

STA DARD t t d-. 01-~----~----~~----~~-----*~----~------*-----~------~------~~1,1 1,5 1,4 I,S 1,6 1,7 1,1 1,1 ·a.p J.l Fig. 9 - Regressions for Cc = f(ydmax) for all Oedometric Qata Collected

929

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu basalts, with higher than average o values (about 2.9 t/m3) a correction was introduced Ill)· MODULUS NUMIIIER reducing ~11 ydmax values to a nominal value of 2.67 because of the direct association of compressibilities with porosities.

Although the final equation Cc ~ 0.23 (2.55 - ydmax) changed little, both its form and its coefficients suggest themselves as most ap propriate.

3.3. Overall Generalized Compressibility If on the one hand it be contended that often (as in the present case of compressibility set~ tlements) statistical average behaviors deduced from a broader spectrum of data might be more representative than individual specific test results, on the other hand the engineer must guard against generalizations too broad, except for the purpose of "educated guesses" for feasi bility estimates. An appropriate example of this extreme is configurated by the generalized expression on volumetric compressibility propos ed by Janbu (1963): Fig. 10 reproduces Janbu's suggested broad general band for all mineral granulates, and on it we have superposed our data as per Fig. 9. Although the nature of the problem is obviously similar, one can well see that when the generalizations are too sweeping, the practical interest to the engineer facing a given project is totally lost: the dispersion would permit estimates varying more than ten -fold for a given porosity. We well know from test results .that although the selfsame porosiy might be reached either by preparing (mixing, depositing) a sample at a given porosity or by precompressing a looser sample to that con dition, the two "structures" will yield dif ferentiated compressibilities, and in the lat ter case there will be a more perceptible dif ference between the precompressed and the virgin compression ranges. There is, thus, an optimum point of judicious decision between too much credence to specific test results, and too much reliance on "generalized experience". Another direct observation is that compacted materials tend to behave (in oedometric test results), in the case of clayey materials somewhat more incompressible than anticipated, and in the case of sandier materials, much more compressible than anticipated: the latter re sult is known to be due to sampling, trimming 20 40 eo and testing errors in more brittle specimens. Fig. 10 - Generalized Compressibility Trends as Two aspects yet to be discussed concern, on the function of porosity, all soils compa~ one hand, the extent to which increased numbers ed, virgin compressions, Cc conditions of tests might improve definitions of design parameters, and further, the improved appli bands of some sort, it is of considerable rele cability of more sophisticated tests hitherto vance to develop for any given problem the ap occasionally considered. plicable rate of improvement of design bases with increased routine efforts. The first of these items will be discussed forthwith, and firstly on the basis of no more It is my conviction, however, that in our than statistical reasonings. Figs. 11 A, B cultural advances in general we do not fail to present for a sample condition the curves of interact with the very statistical universe of percent confidence bands and how they tighten the data we are progressively collecting: in as one increases the number of tests, assuming other words, we should and generally do progress a constant statistical universe. Under the in the accumulation and digestion of progressive premisse that engineering parameters and deci data and concomitant consequent decisions in a sions are based on upper and lower confidence Bayesian manner. Fig. llA reproduces some of

930

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu BAYESIAN CONFIDENCE LIMITS (BCL)OFTHE POSTERIORI NUMBER OF ADDITIONAL TESTS ( n ) VS. MEAN VS. NUMBER OF ADDITIONAL TESTS ( n ) ® r DISPERSION (cf l OF THE POSTERIORI n ! - -~;~~ENCr-----r·--r- MEAN 120t----:----.-·-L--···-----~------L-L ! f : I : ' ~-·--+-----1----1--- 1 1 i I ' I 90%, ! CONFIDENCE INTERVAL +. --~-- - -

Uo = MEAN OF THE PRIOR POPULATION I I i U MEAN OF THE GLOBAL POPULATION = d'C%1=~ ! u* : MEAN OF THE POSTERIOR POPULATION u * ------r--- (FROM n ADDITIONAL TESTS I i = n TESTS MEAN . +···------1 SIMPLIFYING HYPOTHESIS : 40'r---- I I / INCREASING NR. OF

T. U02 - I

6 8 10 Cf'<%1

HYPOTHESIS (j 2 __ r. 2 V 0 (CONSTANT VARIANCE I . i = u0

ASSUMED VALUES IN EXAMPLE

uo • !I

2 !I 10 20 !10 100 n

Fig. 11 - Bases for Evaluation of Gains in Decision Bands by ;I;nc;~;eased, 'I'esting the data in a manner purporting to elucidate therefore engineering) decisions are assym- the advantages of such Bayesian progress from metrical (greater fear of a certain result,than prior to posterior probabilities through improv- desire for an equivalently opposite result). ed "educated guesses" (what I have called quantified Observational Method) . It can be seen that whereas in a universe of constant dispersion (T = 1.0) the gain in confidence 4. FIELD COMPRESSION DATA FROH SETTLEMENTS, interval ICB from 1.04 to 0.5 would require COMPARED WITH TEST RESULTS increasing the number of tests from 4 to 20-4 16, if after the first step we incorporate a Figs. 12, 13, 14 and 15 summarize the compara more educated guess for the second step (smaller tive information on compressive vertical strains dispersion, T changed from 1.0 to 8) the cor as derived from laboratory tests, both routine responding number of additional tests could be and sophisticated, and from settlement obser limited to 14 - 4 = 10 instead of the 16. vation of the construction settlements.

Note that the Bayesian calculations carried out Firstly it must be mentioned that for the are based on the routine mathematical hypothe purpose of approximation with current routines, sis of symmetry, whereas we are presently the· field data on a have not been corrected searching for the desirable and necessary in with regard to I. Fig. 16 summarizes, for com corporation of computed bias, assymetrical, be parison, typical data on some settlement gages, cause it is inescapable that human (and presented both with respect to yz and with

931

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu pesuvED SETTLEMENts ~WEDISH -801C \ SALTO osORIC "' \ @

0,1 (J k9A:nf 10

OEDOMETER MOLDED

COMPARATIVE STRAINS FOR o - 1 Kg/cmz· F_ig. 12 Field Compressive Strains Compared with various Lab. Tests Tried. Salta Santiago Clay Data.

U(kg/c~) 0,1 10

____J.A!!QRATORY C.OMPA.G.nQ_ SPECIMENS

··-1-···-·-l . ·····- - . i ! i 140 . -··-·-·-·--· ·--····· -·-· .. ···--1·····- ---··· ·-·-·-··--- C.!xl031 6 (xi03 ) Fig. 13 - Compressions, Laboratory arid Field, Fig. 14 - Comparative Data, Salta Santiago Dam, Salta Osorio. Indications of cr'p. Lab. and Field.

932 First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu 'r-----~O=E~D~OM~·~U~N=D~IS~T~U~R~B~E~D~B~LOC~K~S~----~ B-17 S-2

~·~------4------~----~ ~I 1,0 10 (j (Kg/cwli) e r------~O~EDO~M!!!E;_!TC!:E~R~..::_~M!.!,O~L:_!:!O~E!:!..D__,.. ____

Fig. 16 - Rockfill Compressibilities. Need to include Influence I Factors for Stresses. Convenience of plotting s vs. log. a almost exactly the same indications: the scatter in laboratory data is apparently great, but the consistency of field observed behavior was generally so very close as to be surprising (and to render impossible packing more curves into the same drawing).

The principal conclusions from these obser~ vations stand in surprisingly emphatic confir ~1+------~------~----~0,1 1,0 10 mation of some of the points discussed re garding the need for very significant revision of the routine test procedures and corresponding parameters. .... Laboratory molded specimens would seem useless, 1,17 OEDOM. both if tested in oedometer compression and in MOLDED triaxial compression. One suggestion for the oedometer test might be to mold the specimens by a.•• compacting directly into and against the I, II oedometer ring, incorporating into the specimen an initial lateral confining stress: it would 1,14 require research and adjustment. The use of specimens cut from undisturbed block samples .. Ia from the compacted fill offers some improvement: FIELD that is why in our design and construction ex 1,12 perience with compacted earth dams we prefer to use the first few thousand cubic meters of IJI COMPARATIVE STRAINS 0,1 actual placement and compaction as a field 1,0 r. IO FOR 0- 7 KQ!cm2 u 1Kt/e•2l compaction test for field adjustments ~nd for Fig. extraction of undisturbed block samples. One 15 - Compressibilities, Lab, and Field, suggestion would be to resort more to field Salto Santiago Clay. tests (plate.load tests, duly conducted and interpreted, or pressuremeter tests, etc ••• ) respect to the nominal elastic Iyz: the ~rinci rather than so-called undisturbed block samples: pal consequence is that the log (Iyz) graph ac thereby one approaches somewhat more the in situ centuates more definitely the apparent nominal condition that should retain residual stresses preconsolidation pressure. from compaction. Secondly, it must be mentioned that overburden sophisticated laboratory tests on 4" diameter stresses were measured in the core (because Of undisturbed specimens from block samples do fears of redistributions due to incompatible offer some improvement (cf. the Ko tests and the deformabil.ities and possible "hang-up" of the anisotropic triaxial compression with crl/cr3 cores) and the pressure data attributed to the ratio of 1.5). The predicted strains (and set field behaviors are well confirmed. Finally , tlements) would still be of the order of 2 to 3 it must be repeated that only one or few points times higher than the field behavior observed. of fiel.d observation are represented in each It hardly seems worth the trouble: moreover case because al.l of the oth~r points gave

933

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu there is no plausible the0retical reasoning in used to digest statistically the records of favour of such sophistication. Therefore, in existing dams within a moderately plausible other cases the comparison could be either less physical and computational model, thereupon favourable or more so, purely by coincidence. furnishing for each new project the means not merely to reach more probable design estimates, Note that the recompression strain values from but also, principally, to adjust such estimates oedometer tests on the undisturbed block samples through a "quantifiable Observational Method" come adequately close to representing the field right from the start of construction through behavior. Therefore, a technique may well be appropriate inspection testing and behavior adjusted for using such simple tests, but re monitoring. sorting to two, three, or more cycles of com pression-decompression, before extracting the The important point to emphasize is that one appropriate recompression E value, much as was cannot backanalyse published data from past done for building settlements on Stuttgart, cf. projects without inquiring what data were ob Schultze and Sievering (1977). tained and how, and without readjusting such data to the new mental model. And that is why, One very remarkable observation is, of course, once again, the use of the most modern sophisti the fact that the clay compressibility is not, cated solution, although theoretically much bet as was early feared, significantly greater than ter, may find itself unable to establish the that of the clean sound dense basalt, compacted link with.the "experience" to be extracted from in 0.8 to 1.2m lifts, which tallies reasonably the past; whereupon at each stage of progress with the observations that stress distributions preference may have to fall upon a moderately between core and rockfill at the points of set theorizable improved mental model, applicable tlement observations were not significant· to greater numbers of case-histories, reasonably adjustable. In conclusion, it would appear that since the principal problem would derive from statistical At the extreme of the empirical use of as many dispersions in the tests (possibly because of cases as possible one might cite such proposals small specimens etc ..• ) and since irrespective as that of Soydernir and Bjaernsli, 1979: all of test-type sophistication the factors of ad displacements were related merely to H, and on justment laboratory-to-prototype are of major ly post-construction data were analysed. Thus reductions, most often it will prove preferable there is no mental model tying during-construc to employ the more routine (oedometer) test and tion decisions and observations with future .greater numbers of tests, for more rapid and consequences, which is frustrating to the very economical determinations of the applicable ad essence of engineering. Crest movements (set justments in individual cases. It is irrefuta tlements and horizontal displacements) and ble that we gain much by pursuing investigations displacements normal to the upstream deck were as to which laboratory test (oedometric by-di considered, for reservoir filling, and subse verse applicable procedures, or triaxials of quent: these are indeed the points of conse various stress-strain-time trajectories) might quence, but the engineer faces important deci best represent a clayey core behavior. In such sions that -cannot be tied passively merely to a case, however, the subject falls into the H and the categories of compacted vs. dumped areas of 1) the systematic error between labo rockfills. ratory soil element behavior and corresponding field soil element 2) the above-mentioned At the other extreme there have been individual calculations of the numbers of tests (each with ized back-analyses by finite element programs, its dispersions, which unfortunately can often adjustable with evident success. There has be greater, the greater the sophistication) been satisfactory correction of test results necessary to achieve average parameters within from oedometric and triaxial conditions to the desired tolerances. plane strain conditions for the dam through in clusion of the v corrections; but there has been insufficient recognition of the influence of v on field observations, partly because of 5. NECESSARY IMPROVEMENTS EMPLOYING DATA FROM inertia in adjusting assumptions comprehensibly CASE HISTORIES applied in earlier settlement analyses of earth dams. Moreover, one might question if often the It is recognized by the practising profession data to which the finite element analyses were als that the principal source of information adapted did not incorporate an incompatible should derive from field data from case histo simplification such as was the routine assump ries. Especially in the case of rockfill it tion of field vertical stress crv = yz (Boughton, is emphasized that "characteristics cannot be 1970, on Lower River Bear Darn n9 1). determined beforehand by laboratory tests" (Soydemir and Bjaernsli, 1979). However, it The routine assumption used to be that incre is very important to recognize that field com ments of overburden stress yz are applied and pressibility data have most often been com transmitted as if pertaining to infinite loaded puted, plotted, interpreted, and published un area conditions, I = 1.00, and yet are limited der oversimplified assumptions. At the other strictly to the vertical comprising the point end finite element analyses have been employed, (and "compressible column") under consideration yielding valuable conceptual indications but for each settlement gauge. If slopes are flat often leading to undesirable conclusions when and the dam construction does not incorporate tied do back-analyses of questionably derived varied phasing within the section (longitudinal case-history data. and transverse) , such an assumption may not be come patently unacceptable. In fact however,as It is the intent herein to suggest to the prac the fill rises, a given midpoint between verti tising professional a middle course that may be cally positioned settlement gages within the

934

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu dam body only receives Iyz pressure increments; on: but in our dams herein discussed such situ and the least that can be done is to use the ations could be set aside as negligible; and it elastic model charts (Poulos and Davis, 1974) is better to analyse the simpler case of instan for estimating approximate I values. taneous deformations. - One of the illusions generated by the wrong ear In short, the proposal herein is to include the lier practice of interpreting and plotting con interference of: a) the Influence factor I (pre struction settlement data is that at some moment sently for homogeneous elastic bodies}; b) the when the fill height has reached the outer slope explicit effect of the Poisson's ratio v so as above the point, the presumed ~cr is considered to permit adjusting for its importance and im to have ceased, and the continuing settlement portant variations; c) recognition of the vari has been presumed indicative of consolidation ation of compressibility with a, and this (in clayey materials, wet, high dams) or as preferably through the mental model of com "secondary compressions" and "creep" in rock paction precompression cr'p and semilog straight fills. There must be conditions in which creep line in the "virgin compression line". occurs in more crushable rocks, in cases of m~ intense wet-dry and frost-defrost cycles, and so Fig. 16 presents some illustrative data along such lines. Fig. 17 reproduces data from recent research testing on modelled rockfill (Veiga Ko Pinto, 1983), which patently reinforce the ind! cation of higher cr'p for higher relative den sities, even without perceptible crushing; 2,6tt--+---t---11---+-----I obviously the effect should be greater through 10 crushing of contacts during compaction. Finally (%) 2,2H---il-- ...... 60 m (MUL.TIPLYJNG FACTOR s-., =m X S~_ 1 -03 1 il --- 80 ON SETTLEMENTS) • ·~~~~-~'===~·~0~0~+--~--\\! ~·0,2 1,4t--T'::-i--__,..--f--..j._-~-- ..... / \;! 1,2 - • \} •0,3(REF.- 1,0 I,Ot--,-...._"'1' ~~~;:-·,-,,--t-~-. --t----~-- + ~~PTEO) Opr---~~~··.. ~-~~~~-,~~·-..~~ .. ~~~-~~~--- 0,1 ·.:.·,,-~;;;-.. J:.- + ... 'y. 0:4 -- ·--- 0,1 ~r---r-~---4---+--~--- + L----- 0,4 ~ + ~-----'"" ~ 0,2 + ·------·~ '\.~•0 41- -- REF. (POULOS A~ 0 BOOKER, t72) l " I 0,4 0,6 0,1 1,0 "M • ·- "'F"' . HYPOTHESIS 0,511) -~ SETTLEMENTS ® I U• Ih OBSERVED E • l • f( h) PREDICTED.;) •0,48--• o,.wo fiG' E • 347,8•2,!1611 +0,014h2 !Kt/caft~ (NARDELLI,..;) •0,35 ...... 821. 'V •(),30- . 74B 'V •0,211·---• o.- ' ~ II ! \ 'i) •O,ZO-·-· •!i•j· \ o,- -· ·+- I' l 'I ! i , I 0,310 I:! li l ~-- -· .. flfU. .. .l. . 0,110 '! li --- -- '1~ i.k ---- O,HO · i' I ~ ....., ...... i'='-"'-'- ! ,. "...... , 0,140 'o (%) (KII/,.2) I ---eo HO FOZ DO AREIA '',',,, o- -·- eoo 100 --TI 1lllliil0 DAM l Lll I tilt 10 I 4 liM I 4 ••'Ill I 4 ••iit Ur I.KN/.-2)

Fig. 17 - Lab. Confirmation, on Model Rockfill, Fig. 18 - Importance of v Values in Interpret Compa~tion Influence on cr'p (Veiga ing Self-weight Construction Settle Pinto, 1983) ments

935

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu ~ig. 18 reproduces some recent indications loading. For the reevaluations the average (Nardelli, 1982) on the great importance of v field inspection data on ydmax and PC% were used on self-weight settlements. The upper graph in accordance with the procedures herein recom furnishes indications for preliminary estimates, mended. Fig. 19a summarizes the results showing in adjusting observed settlements to different that estimates always tend to be pessimistic, v values. As seen in the lower graph most of often up to 3 times higher than observed, and the surprising data on very low E values obtain this principally in the higher dams and more ed in the Foz do Areia Dam might be associated clayey materials, wherein the fears are greater with v values, which, incidentally would affect and decisions of much greater consequence and construction settlements and upstream deck responsibility. Fig. 19b summarizes comparative movements in a totally differentiated manner. data on three dams of almost geometrically simi lar sections, employing the very clayey material of red porous basalt clays (SO, SS and IT Dams). Finally, Fig. 19c summarizes the corresponding 6. OBSERVATIONS FROM EMBANKMENT DAMS COMPARED data for the EMB Dam,in which the intuitive pre WITH PREDICTABLE SETTLE~~NTS BASED ON caution was taken of requiring higher PC% in a REGRESSIONS lower stretch, presuming to improve conditions in the uppermost stretch. The projects listed in Table I were used for reevaluation of the predictable construction These data, based on the hypothesized v=0.3, and period settlements compared with the prototype especially the comparisons in Fig, 19b, lead to observations. In all cases settlements were the impression that the estimated a'p precom essentially instantaneous with incremental pressions based on oedometer tests, undisturbed

s-• EST. (cal CJ:' ~E

1:1

'0 d max ':! 1,4 t/m3 { ® '\) = 0,3 ADOPTED

100 IT"ITAUU. EM· EMBORCACAO SS ·SALTO SANTIAGO Pt • PARAITINGA AV- AGUA VEIIIIIELHA P8 • MRAI8UNA CP·CAPIVARA PO· PEORA 00 CAVAU) PR • PASSO REAL ( PIIRTIAL HEIGHT) SS (dq)· S.SANTIAGO LA"LA ANGOSTURA ( AUXILIAR DIKE) NE·NETZAHUALCOYOTL SO· SAL.TO OsORIO EI"EL INFERNILLO Sl ·sAo Sllil~O CH·CHICOASEN

200 S 111011. OBS. (QII) H ( m )

CONSTRUCTION SETTLEMENTS

------PREDICTED OBSERVED

PREDICTED· PC1• 102%, PC2 = 104% COIIIPACTACION FIELD DATA .

RANGE OF OBSERVED SETTLEMENTS

PREDICTED • PC1= 100% , PC2• 102% FROM COMPACTION SPECIFICATION DATA

Fig. 19 - sample cases: Predictable vs Observed Max Settlements; Predictions Based on Present Suggestions

936

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu lls APPROXIMATE TENDENCY SIIIOXUt. - S lftOXOba. (CM) ill APPROXIMATE TENDENCY OJSERVED ( DECREASING I lllll •t. - I 1110¥ 0111. ( c111) OBSERVED ( DECRE.UI Nt 200------t------~O~d~m~~V~A~W~E~S~)~.----- PC VALUES) 200------+------~------

+ ss (1,4) ~~\/ I 'Sd -~~ VALU S +.IT(I,I71 / ( ) PC(%1 VALU S (t/IIIS) I / I / EM (1,1$) + 100 ------~- + SS DIKE ( 1,40 I 100------+- / / / / .~ +PR (1 155) tt.~ .I::+N(I,Sil / AVU.TOI .,:tCP(I,eo) _.- +PT(I,OBI -+frO ( 101) CI,IS)- +EI (1,70) I 115

100 IOD Fig. 20 - Bases for Estimation of Corrections, to arrive at smax observed from smax estimated. 11/H ® 11/H

HYPOTHESIS

H•IODIII

SOIL CHAR{As.TERISTICS: 2 PC ot 100% (j p• 7,0 KtJa- Cc•0,3 son LAYER: ( 2011) PC" til% {oP •1,11 Ktlc•l Cc. 0,5

2 SETTLEMENT S (II) LARGER SETTLEMENTS ON CREST 11/H 11/H @0 / .A ® 1,0 , " 1,0 / .... ~ t. ~ ,t. \ ~ , \, ...... i'..... , \, t o,a " 0, 1'- ' 't, "' o.• v 0,• ICALCI , CO INC DENT ~~/ 0,4 If' ""/ 0,2 / 0,2 ~ v / ../ .p I I 4 S(lll) I S 4 S(lll)

Fig. 21 _ Analysis of Position of Greatest Consequence of a "Softer Layer"·

937

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu ed as most significant in the present mental HYPOTHESIS model. Obviously the errors (and nec~ssary H•IOOm; '6 •2,0 tf/1113 corrections) increase with increasing H (and V: 0,3 1 0(. • 300 increasing predictable settlements) • C) INITIAL A'- FINAL ----r- .,.. ---.- Intuitions have sometimes led to requiring (j z = 33,5 + 8,5 = 42,0 heavier compaction at the base of high cores, Ciz =70,8 + 7,4 = 78,2 and have tended to relax such requirements in Gz = 104,4 +6,8 = 111,2 one or two steps up to the top. However, if we /-j------+-=------~G'z =140,7 + 6, 7 =147,4 look at the typical internal settlement diagram ( tf/rro2J (reaching a maximum at about mid-height) and if we emphasize that the only settlements (and differential settlements) that matter (to the top of the core) are those suffered by the top, we conclude that the first intuition needs immedi ate revision: a more compressible layer matters e most where settlements are greatest, close to mid-height (Fig. 21).

I In fact, by use of the mental model proposed, if I I we assume that attention concentrates on the top I 10m (rarely could tension cracks descend below) Ae f J it is possible to "theorize" on what would be I for each elevation the PC values that should be Ae t------respected as important (Fig. 22). At great +- ,------depths both the initial crv (before adding the GREATEST DEPENDENCE OF final lOrn of dam) and the Acrv (Icrz corresponding SETTLE~ENTS ON ZONE AROUND l.)'p to the additional final height) lie in the Cc straight line and the compression should be ap proximately independent of PC. Also at the up permost elevations, both initial crv and crv + AOV would lie in the precompressed range for a given PC, and the importance of the incremental com pression would be secondary. It is, therefore, in the intermediate range of elevations that Z(m I specifications and inspection have to be of greater consequence to the top. ZONE OF GREATEST SETTLE There is at present sufficiently wide recognit MENTS(STRESSES AROUND Ci'p ion that field moduli E (computed with I values ....._ DEPENDING ON FIELD PC 08- but assuming constant v) present a steady vari ation with cr. The data extracted from the two --~~;--- @:RVED.) T ~ major dams are plotted in Fig. 23. The somewhat surprising indication has been that compacted clayey fill did not prove disparagingly compres ---- L{J~N~T~A-~ ....._ '-r sible compared to the soundest clean rockfills. Of course, the I values need to be corrected I ' ~ I :-._ because of the not-constant E: computer solu I ::-._ tions changing E by steps would be the typical I ~ I ::-._ modern approach. However, one can easily rea 2.0 <10 60 80 . 100 120 140 160 son that the stiffer outer "arch" should de Gz crease stresses in the center and increase them 97 PC C%1 (tf/m2) outward. Without too much added effort one can use the homogeneous elastic body Influence ASSUMED 1,5 <'6'd 111ax< 1, 7t/m3 values, by superpositions, to revise the erst -FROM (PC, ~dm xl ~ !J'p (SEE FIG. 7) while estimated stresses, Fig. 24 shows how this was done assuming but two zones and a ratio Fig. 22 - Idealized Deduction of stretch Wherein of moduli of 2. The fact is that inevitably the Compaction PC Reflects Most on Top 10m true curve of E vs cr drops more rapidly. blocks, and PC% data, do not reflect adequately the presumably higher internal cr'p prestress conditions that prevail in the dams, partly re 7. TRANSITION FILTER-DRAINAGE MATERIALS. tained by the K'o lateral compressions, and partly retained by suctions. Instrumentation It has been repeatedly published, and is by now and monitoring efforts will have to dedicate well recognized, that the most incompressible much more attention to problems of in situ re materials in an earth-rock dam section are those sidual stresses and transverse strains. of intermediate grainsize, the typical "transi tion strips" used for filter-drainage. Moreover, At present there seems to be no way to reach a the appreciably different behavior of rounded better mental model, generally applicable via gravels vs. angular crushed rock aggregates has estimation of in situ residual stresses as func been emphasized, as shown in Fig. 25. tions of PC% and ydmax. Fig. 20 has been pre pared to indicate the possible trends of the One observes, there~ore 1 that problems of hang interference of the two basic parameters adopt- up due to differentiated settlements tend to

938

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu '3

E .M. DAM • EMBOfiCACAo C.R. DAM • FOZ DO AREIA

o+------~------~~----~~------~------1------1-~u~c~t~21 . 5 10 15 20 ~5 30 Fig. 23 -Field Moduli. Varying E with cr. Modest Di~~erentiation between Core and Rockfill

( • ) HOMOGENEOUS (+ l NON - HOMOGENEOUS

IO VERTICAL STRAIN Dv (%0 ) L.. :::=-l- L.. 1--- •8 •10 8 [:...- -[::--1- I •2 • 4 ,;. ~~ __....-:_ v F 6 !l¥1J ORA WELS ® vP OED ~ ( ROU 4 IOOO·r-----~~-----f------}------~------~ v / ® 2 V RTIC LSn ESS ( v (kalicm2J 0 v II 10 II II 7 0 2 4

STRAIN &v (0/ool

( E1 z E2) (CF. F G. 23) II 2 7

0~----~------~----~------~----~-- 0 10 15 20 25 (}. 1:1"1 (from llomot-• :I I Fig, 25 - Comparative stress-strain Behaviors of Transition Gravels vs. Crushed Fig. 24 - Simple: Correction for Varying E. Rock (Kjellman & Jakobsen 1955) • Example Assuming 2 Zones of Constant E

939

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu depend very much on the transition sections. filter; the upstream rip-rap and transition Needless to say conditions are worsened if the zones could well be narrowed. Ideally any dif design requires wider transitions, broader ferential settlements of the differently sup well-graded grainsizes, and compaction: such ported zones should be absorbed along modestly requirements,that would appear to add to the compacted chimneys of (alluvial) sand. safety of the dam,can frequently be concluded to be exactly the opposite. The remarkable difference in behaviors evidenc ed with regard to non-consequential longitudin al cracking and the transverse cracking that would be critical, points to the need for clos 8. OBSERVED BEHAVIOR OF THE MAJOR EARTH-ROCK er attention to lateral stresses cr'2 or cr'y in DAM comparison with cr'3 or cr'x that has dominated attention because of routine slope stability Presumably the crucial attentions are directed analyses. Fig. 27 schematically compares, on at predicting and avoiding deleterious US-DS the basis of elastic behaviors (E, v parameters) cracking across the core. The dam in case gave what lateral stresses would be required for the best of indications of predictable satis closing a given crack, either longitudinal or factory design and monitored behavior to the transverse. Because of settlement, and assum end. However, shortly aft~er the crest was ing that longitudinally the dam's crest has to (rapidly) paved, while re'servoir filling was compress as it descends into the wedge between occurring (at an unexpected rate corresponding the two abutments, the transverse crack can be to 1:100 yr recurrence flows) two longitudinal closed by a very small incremental cr'y (resi cracks developed along the US and DS transttions dual stresses due to compaction were not con in a remarkably homogeneous manner along nearly sidered, conservatively assumed equivalent in 2/3 of the length of the crest. A posteriori the two cases). The necessary incremental cr'x analyses would suggest that the upper 20m of stresses to close the longitudinal crack would, the dam rose very rapidly, with compacted clay for the case in question, be of the order of 33 fill stiffer than desired; and the hurried times higher. For both cases another rough com finish of crest grading and paving served to putation that seems to bear interest is what make the cracking more salient. In short, the height of added fill (and consequent incremental magnitudes of deferred movements that will af cr'x or cr'yl would cause the strain to counter fect the upper few meters are important. More the crack. The additional heights of fill re over, reservoir filling did cause incremental quired for the two cases are even more blatantly rockfill settlements even in points on the different because of different I coefficients. downstream slope (although in modest amounts), By inverting the reasoning such indication of and on hindsight one wonders how much such additional fill reflects how deep a tensjle movements would have been attenuated if the crack can propagate. rock-fill had been watered during compaction. An important inference from the case is that In a practical sense the case merely repeated instrumentation needs on such dams have to be the oft-quoted platitudes: the tension cracks significantly reconsidered. Across the shell (up to 10 em wide) went down only up to about -to-core transitions it wouJdbe of interest to 3-4m and transformed into shear cracks (some of employ profilers for simultaneous vertical and them into the core); there was, of course, no horizontal displacements, but one must guard consequence to the dam from such longitudinal cracks, which, moreover, partly moved back, and mostly were treated by appropriate filling. It was remarkable that not a hairline fissure de veloped in the US-DS direction across the crest pavement, which indirectly served as an excel lent visual index for comparative monitoring: the specific differential settlements longi tudinally ranged in the order of 1:600 to 1:800, which are strains too minute for design estimates but have been discussed with regard to choices of appropriate materials for avoid ing crest transverse cracking. What would have been the important transverse behavior if total settlements had been significantly bigger as predictable, and especially so in the final up per stretch ? I submit that it is mUch too PREFERABLY TO BE NARROWED AND SIMPLIFIED specious to discuss different tensile strains to fissuring as achievable by materials of dif ferent plasticity indices and compaction speci fications. For weakest link tensile conditions the reliance on meticulous homogeneity is ~ar -fetched and too dangerous. One practical design decision (for a subsequent project) arose both from the desire to minimize differentiated narrow strips of placement and compaction near the top, and from the theore tical desire to employ as homogeneous a section 0,10 as possible in the upper 20 meters, Fig, 26 DET. A illustrates one such case wherein the only do~ minant feature to be retained is the chimney Fig. 26 - Suggested Design Revision Near Crest

940

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu 9. OBSERVED BEHAVIOR OF CONCRETE-FACE DAM TRANSVERSE SECTION The only concern in the case of upstream face ( OBSERVED CRACkS ) dams lies in the displacements of the face upon reservoir filling: the need is to minimize them in order to avoid cracking and leakage. The systematic observation of construction-period deformations (hitherto almost exclusively set tlements) for calculation of E moduli has only C·!-~~·~ one purpose, which is the prediction of the de formations of consequence to the slab. The very low moduli computed as approximately valid to wards the end of construction should have led to CONFININ8 PRESSURES { 0., = AVAILABLE deep concern regarding the possible excessive (ELASTIC THEORY) O.Ec: " NECESSARY opening of joints. In fact, however, the ob served deformations during the first reservoir filling were only about l/3 of the values that PRESSURES) TO CLOSE CRACK would be computed from the final self-weight H't OF FILL settlement E values, Fig. 28. 'V. 0,3 I Ua•~:c:= ( ll- 10) Uxav The design solution for such dams has already successfully adopted the practice of employing l HT. z ZOWI a well-compacted finer-graded transition in the upstream zone. Doubtless this is a factor of significant benefit (as shown in Fig. 28 for an idealized case calculated by finite elements), and there might be insufficient interest or need to delve into better comprehension of the real behaviors. I have emphasized above the great relevance of v, and the need to monitor much more intensely the internal horizontal displacements accompa nying the construction settlements. Depending on such indications one may well conclude that LONGITUDINAL SECTION much of the difference (favorable) in behavior HYPOTHETICAL CRACKS (NOT OBSERVED) during "active" se:u!-weight movements and "par tially passive thrust" reservoir loading move ~-~---.-~--..-t ments may be explainable by the very significant l~z II! o,40• changes of v for compacted rockfill under the partial stress reversal. The basic need and interest is to reach a satisfactory comprehen sion of the geomechanical behaviors at play. Of course, if the proportional benefits due to the finer-graded zone are better assessed, there will be conditions for optimizing the use of PRESSURES ) TO CLOSE CRACK this special crushed-rock zone from technical HT. OF FILL and economic considerations.

"\) • 0,3

741

..:) =0,3 AOOP'TED

Fig. 27 - Comparison of lateral stresses re quired to close longitudinal vs. transverse cracks.

against the great danger of such instruments themselves being conducive to concentrated leaks and piping. Much more attention should USING AVAILABLE SOLUTION (REF.: POULOS a BOOJ<£R, 73) be given to deferred settlements affecting the top, and to behaviors at the top. Fig. 28 - Concrete face displacements due to US loading,

941

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu 10. CONCLUDING REMARKS Tokyo, val, I, 711,-714. Soydemir, c. & Kjaernsli 1 B, (1979) Deformation The road to knowledge ahead of us stretches of Membrane Faced Rockfill Dams, 7th ECSMFE 1 ever further and wider. There is a natural Brighton, U.K., vol, 3, 281-284, tendency to stop at the oases, and possibly 1 to Sharma, H.D. et al. (1976). Generalisation of delay too long before proceeding onwards. The Sequential Non-Linear Analysis - A Study o;f milestones of satisfactory engineering deci Rockfill Dam with Joint Elements, 2nd Int,Cont. sions have to be ahead of our trekking in on Numerical Methods in Geomechanics, ASCE, collecting quantifiable data for digestion of Virginia, USA, val. II, 662-685, the laws of the behaviors that have proven ade Tanaka, T, & Nakano, R. (1976) Finite Element quate. It is strongly suggested that the most Analysis of Miyama Rock;fill Dam, 2nd Int,Conf. profitable directions and advances require a on Numerical Methods in Geomechanics, ASCE, balance between the extremes of sophisticated Virginia, USA, val. II, 650-661, June 1976. theoretical developments applied to single special cases, and the attempt to digest nomin al behavioral information that constitutes "practical experience".

REFERENCES Boughton, N. (1970). Elastic Analysis for Be havior of Rockfill, Journal Soil Mech. and Found. Division, ASCE, vol. 96, nQ SM5, Sept. 1970. Clough, R.W. & Woodward, R.J., (1967). Analysis of Embankment Stresses and Deformations, Journal Soil Mech. and Found. Division, ASCE, vol. 93, nQ SM4, July 1967. de Mello, V.F.B. (1977). Reflections on Design Decisions of Practical Significance to Embankment Dams, Geotechnique, vol. 27, nQ 3 279-355. de Mello, V.F.B. (1980) Comparative Behaviors of Similar Compacted Earthrock Dams in Basalt Geology in Brazil, International Symposium on Problems and Practice of Dam Engineering, Bangkok, vol. I, 61-79, Dec. 1980. de Mello V.F.B. (1983) Design Trends on Large Rockfill Dams and Purposeful Monitoring Needs, International Conference on Instrumentation, Zurich, Sept. 1983. Hynes, M.E. & Vanrnarke, E.H. (1977) Reliability of Embankment Performance Predictions, Me chanics in Engineering, University of Water loo Press. Janbu, N. (1963) Soil Compressibility as De termined by Oedometer and Triaxial Tests, ECSMFE, Wiesbaden, vol. 1, 19-25. Kjellman, W. & Jakobsen, B. (1955) Some Re lations between Stress and Strain in Coarse -Grained Cohesionless Materials, Royal Swedish Geoth. Inst., nQ 9, Stockholm 1955. Pinto, A.A.V. (1983) Previsao do Comportamento Estrutural de Barragens de Enrocamento, LNEC, Lisboa, PORTUGAL. Julho 1983. Poulos H.G. & Booker, J.R. (1972) Simplified Calculation of Embankment Deformations, Univ. of Sydney, Research Report nQ 192, May 1972, Sydney, AUSTRALIA. Poulos H.G. & Booker, J.R. (1973) Embankment Deformations due to Water Loads, Research Report nQ 224, Sept. 1973, Univ. of Sydney, AUSTRALIA. Poulos H.G. & Davis, E.H. (1974) Elastic So lutions for Soil and Rock Mechanics, John Wiley & Sons, Inc., 1974, New York, USA. Nardelli Rosi, M. (1982) Foz do Areia - Retro analise pelo Metodo dos Elementos Finites, Tese de Mestrado, Dez. 1982, PUC, Rio de Janei ro, BRAZIL. Schultze, E. & Sievering, W. (1977), Statistical Evaluation of Settlement Observations, 9th Int. Conf. on Soil Mech. and Found. Eng.,

942

First International Conference on Case Histories in Geotechnical Engineering Missouri University of Science and Technology http://ICCHGE1984-2013.mst.edu