3 —Research and Technologies for Evacuation Systems

Chapter 3 RESEARCH AND TECHNOLOGIES FOR EVACUATION SYSTEMS CHAPTER 3 Research and Technologies for Evacuation Systems

The aircraft evacuation system has three key In 1980, FAA’s Special Aviation Fire and elements: exits and slides, efficient means of Explosion Reduction (SAFER) Advisory reaching the exits, and the crew and passengers Committee published several recommendations who use them. To be able to leave one’s seat, to improve fire safety and survivability.2 FAA move toward an exit door or hatch, and escape used the committee’s recommendations to di- from the aircraft depends on the passenger’s rect its research and development (R&D) physical and mental condition, and tolerance to efforts, and produced new and modified regu- crash and fire hazards. These hazards, in turn, lations in a number of areas.3 The success of depend on the strength of seat attachments and FAA’s programs rests primarily on the devel- restraints, airframe energy absorption, and the opment of representative fire scenarios and test fire resistance of the cabin lining and seating methods. Currently, research is concentrated in materials. two categories: in-flight fires, where safety is measured by the ability to prevent, detect, and Evacuation performance thus requires en- contain a fire in the immediate vicinity of igni- hanced cabin safety to preclude incapacitation tion as well as discriminate from false alarms; from impact, smoke, heat, and toxic gases be- and postcrash fires, which in turn involve either fore egress can be achieved. Evacuation per- making the environment inhabitable for a formance also depends on the design and longer time or evacuating passengers more operation of emergency equipment and flight 4 quickly. The key programs relating to cabin attendant training. Cabin safety research and materials, emergency equipment, and training evacuation testing are essential elements of any are discussed below. effort to assess and improve evacuation safety.

CABIN SAFETY RESEARCH AND Cabin Materials TECHNOLOGIES According to FAA, the most important recent The Federal Aviation Administration (FAA) improvement in cabin safety was the addition of researches and regulates several facets of cabin fire-blocking layers to seat cushions.5 FAA, safety for transport airplanes, rotorcraft, and with NIST participation, established in the mid- general aviation aircraft. The majority of the 1980s the methodology for determining the rate research and testing is accomplished at the at which hot gases are emitted from burning Technical Center and the Civil Aeromedical seat cushions. The fire blocking has been Institute (CAMI). FAA also relies on the shown to extend evacuation and survival time National Aeronautics and Space Administration by at least 40 seconds in one representative .“ and the National Institute of Standards and Technology (NIST) for contract or cooperative 2 Federal Aviation Administration, Find Repoti of the work in crashworthiness and fire safety, re- Special Aviation Fire and Explosion Reduction (SAFER) spectively. In passenger transport, after the Advisory Committee, vol. 1, Report FAA-ASF-80-4 ashington, DC: June 26, 1980). United States, the United Kingdom is the sec- !?’ ond largest contributor to cabin safety research R.G. Hill et al., “Aircraft Interior Panel Test Criteria Derived From Full-Scale Fire Tests, ” DOT/FAA/CT- and technology (R&T). Other foreign investiga- 85/23 (Atlantic City, NJ: Federal Aviation Administra- tors in fire safety research include Canada, tion Technical Center, Septemkw 1985), p. 1. Germany, the Nordic countries, Japan, and 4 Constintlne p. Sarkos, manager, Fire Safety Branch! Australia. 1 FAA Technical Center, personal communication, Apr. 22, 1992. 5 U.S. Depafiment of Transportation, Federal Aviation 1 Richard B~kowski, senior researcher, Building and Administration, Aircraft Safety Research Plan (Atlantic Fire Research Laboratory, Natioml Institute of Standards City, NJ: Federal Aviation Administration Technical and Technology, persoml communication, Apr. 8, 1992. Center, November 1991), p. 123.

27 28 . Aircraft Evacuation Testing: Research and Technology Issues postcrash fire scenario by delaying the onset of fire performance, and toxicology research to material ignition and reducing the spread of NIST. NIST conducts in-house research at the flames and toxic products of combustion.6 Building and Fire Research Laboratory and funds additional research through its University The FAA Technical Center developed the 11 Grants Program. According to NIST staff, standard test protocol for assessing cabin mate- recent gains in scientific knowledge and the rial flammability through comparison of labora- advent of measurement technology will shift tory studies and fill-scale fire testing (using a fire safety regulation toward performance stan- reconfigured C-133 fuselage). In simulated dards rather than design criteria. postcrash fires, evaluation of combustion gas and temperature profiles indicated that the oc- The measurement technology required for currence of cabin flashover7 dictated surviv- quantitatively assessing evacuation system per- ability, and that flashover can be best formance, including human factors, has not characterized by heat release levels.8 This been developed to the same degree. The Avia- prompted the development of the current heat tion Rulemaking Advisory Committee efforts to release standard instead of limits on specific replace evacuation design criteria with per- combustion products.9 Today FAA continues to formance standards suffer from the lack of investigate fire behavior, smoke toxicity, the sophisticated analytic tools and human per- behavior of composite materials, and the effec- formance data. tiveness of potential safety improvements using the FAA Technical Center’s DC-10 and B-707 10 Emergency Equipment test craft. Analysis of the 1985 Manchester aborted CAMI has extensively studied the effects of takeoff and subsequent fuel-fed fire prompted fire on aircraft interiors, supporting rulemaking several recommended design changes, includ- for crew member protective breathing equip- ing improved access to overwing exits and ment (PBE). Continuing fire safety research cabin interior hardening, most of which have topics include smoke release and relative toxic- been implemented.12 The accident also renewed ity of materials used in cabin finishings, and interest in cabin water sprays and passenger methods to improve evacuation under toxic protective breathing equipment. The relative smoke conditions. merits and disadvantages of these proposals are discussed below, along with the topic of Over the years, FAA’s Technical Center risk/risk assessment. contracted out portions of its materials safety, Protective Breathing Equipment 6 Job J, Petrakis, “FAA Occupant Protection and Time and the thermo-toxic environment are Cabin Safety Overview, ” in Proceedings of the Flight two critical aspects of survival in aircraft acci- Safety FoundationlFederal Aviation Administration 13 International Aircrt@ Occupant Sa$ety Conference and dents involving fire. Based on R&D done at Workshop, DOTIFAAIOV-89-2 (Washington, DC: U.S. CAMI, criteria for PBE for air transport crew Department of Transportation, August 1989), p. 43. 7 F]ashov~r is the sudden, rapid, and uncontrolled growth of fire throughout the cabin, generating high temperatures and toxic gases and robbing the cabin atmos- 11 Bukowski, op. cit., fOOtnOte 1. here of oxygen. 12 Arthur Reed, “Technology Safety . . . For Cabin-Fire ! Constantine P. Sarkos, manager, Fire Safety Branch, Survival, ” Air Transport Worki, October 1991, pp. 101- FAA Technical Center, personal communication, Jan. 15, 106. 1993. 13 Garnet A. Ivfch et al., Civil Aeromedical Author- 9 Also known as the 65/65 rule, which refers to the ity, l%e Eflects of Wearing Passenger Protective Breath- maximum allowable rate of heat release, in kW/m2, and ing Equipment on Evacuation Times Through Type III and the total heat reiease, in kW-min/m2, under specified test Type IV Emergency Aircraft fiits in Clear Air and Smoke, criteria. Final Repoti, DOT/FAA/AM-89/12 (Washington, DC: 10 Alan S. Brown, “Fire Rule Changes Aircraft Mate- U.S. Department of Transportation, November 1989), p. rials Mix, ” Aerospace America, March 1991, pp. 20-24. 1 Chapter 3—Research and Technologies for Evacuation Systems . 29 members were issued in June 1983.14 Consist- spent donning smokehoods during the period ing of a full-face oxygen mask or combination when conditions permit the fastest egress re- smoke goggles and oxygen mask, crew member duces their potential to save lives and may even PBE is required equipment for all aircraft op- result in more deaths. 18 erating under 14 CFR 121. Water Spray Although the investigation of the Manchester FAA commissioned an early cost/benefit accident resulted in a recommendation for pro- study of fire management systems and safety vision of passenger PBE, or smokehoods, rules improvements, completed in 1983. CAA re- mandating their installation on transport aircraft 15 viewed worldwide accidents involving fire-re- have not been issued. Two general types of lated deaths over the 1966 to 1985 period, and smokehoods, filter and oxygen-generating, concluded that the benefit attributable to having have been proposed. The lightweight filter type an onboard cabin fire suppression capability is susceptible to carbon monoxide contamina- (e.g., a water spray system) is likely to be sub- tion and becomes ineffective when cabin oxy- stantial and exceeds the benefit attributable to gen is depleted. Either type can delay evacu- systems that do nothing to delay the onset or ation because passengers stop moving toward progress of fire.19 In June 1989, FAA began the exits to don the masks. Smokehoods can working with CAA and Transport Canada to also impede egress through smaller doors, pre- develop and evaluate a cabin water spray sys- vent passengers from hearing crew instructions, tem (CWSS).20 and reduce vision. 16 The present heat release standard has driven In addition, while the Civil Aviation Author- technology to the point where it is unlikely that ity (CAA) of the United Kingdom issued a draft further cabin materials research and improve- specification for passenger smokehoods in ments over the near term will lead to apprecia- 1986, it rejected requiring smokehood equip- ble delays of flashover.21 Water spray works ment after a joint review of regulatory policy independently of fire origin and has more by U. K., U. S., French, and Canadian authori- potential to delay flashover under a variety of ties showed that the implementation of other fire scenarios; its benefits include cooler cabin safety measures (e.g., seat fire blocking and temperatures, suppressed ignition of cabin ma- cabin material improvements) has improved terials and delay of flashover, absorption of survivability to the extent that smokehoods combustion gases, and the washout of smoke have become less useful .17 Because the time particles. 22 Full-scale tests of one cabin sprin- available to evacuate an aircraft is the most critical element of survival, the additional time 18 ~ulse Speltel and Richard G. Hill, St@ of Be~fits of Passenger Protective Breathing Equipment From 14 Federal Aviation Administration Technical Standard Anaiysis of Past Accidents, Final Report, DOTIFAAICT- Order, TSO-C99, June 27, 1983. 88/03 (Atlantic City, NJ: FAA Technical Center, March 15 The aviation industry first concentrated On the role of 1988), p. 4. smoke and toxic gases in hindering evacuations after the 19 Lionel Virr, “The Feasibility of Improved Fire Pro- November 1965 aircraft accident at Salt Lake City, Utah. tection Systems for Aircraft Occupants, ” in Proceedings of FAA published a summary of CAMI studies related to the Flight Sa$ety Foundation/Federal Aviation Admini- passenger smokehoods performed during the period of stration International Aircraft Occupant Safety Conference November 1965 to February 1987. See E.A. Higgins, and Workshop, DOTIFAAIOV-89-2 (Washington, DC: Summary Report of the History and Events Pertinent to the U.S. Department of Transportation, August 1989), p. 200. Civil Aeromedical Imrtitute’s Evaluation of Providing 20 Water spray is Urgeted because the use of foams and Smoke/Fume Protective Breathing Equipmenl for Airline Halon-based suppression systems present health and/or Passenger Use, DOT/FAA/AM-8715 (Washington, DC: environmental obstacles. U.S. Department of Transportation, Federal Aviation 21 Sarkos, op. cit., footnote 8; and Kent pofler~ con- Administration, June 1987). tributing engineer, New Large Airplane Payload Systems, 16 Helen Gavaghan, “Aircraft Fires: Living Through the Boeing Commercial Airplane Group, personal communi- Smoke, ” New Scientist, Aug. 6, 1987, pp. 54-57. cation, Aug. 11, 1993. 17 Reed, op. cit., fOO~Ote 12. 22 Sarkos, op. cit., fOOtnOte 4“ 30 . Aircraft Evacuation Testing: Research and Technology Issues kler concept (the SAVE system) have suggested gas concentration levels were lower, and the that survival times can be extended by 2 to 3 survival times greater than those in a fully minutes.23 By comparison, fire blocking and sprayed cabin.28 The optimal amount of water improved cabin interior materials provided ex- and its distribution requirements have yet to be tensions on the order of 40 seconds and 17 determined. The drawbacks associated with seconds, respectively .24 using a system with a small fraction of the water required by the original concept should The possibility of inadvertent system dis- be reassessed. charge during flight, the weight/cost of system implementation, and reduced visibility during FAA is also evaluating the effectiveness of evacuation are key drawbacks. At FAA’s re- another CWSS concept, one which employs quest, manufacturers participated in a disbenefit sheets of water to act as curtains between sec- study (i.e., estimating the consequences of both tions of the aircraft and contain the fire within a commanded and accidental use). Estimated small region of the cabin. Using nozzles fash- weight penalties for narrow- and wide-body ioned by British Petroleum and sensor/activa- aircraft were on the order of 600 and 2,100 tion systems developed by GEC Avionics, the pounds, respectively.25 Boeing estimated the BP/GEC system would function similarly to the costs of installing SAVE CWSS to be approxi- first design (see figure 3-l). Relative system mately $800,000 for a 757 airplane, and nearly effectiveness for equivalent water supplies has $1.7 million for the newest model 747.26 Es- not yet been determined. Other options for timated costs of retrofitting the world’s fleet of minimizing the weight penalty of CWSS in- current production aircraft exceeded $6 bil- clude the use of potable water and, in the long lion.27 term, water reclamation systems. Recognizing that these penalties and risks A CAA study of turbine-engine aircraft acci- must be reduced before system implementation dents involving fire deaths compared the po- is feasible, FAA has explored zoned use of the tential benefits of five improvements to cabin sprinklers, or spraying water only in the im- safety .29 Assuming each improvement was mediate vicinity of the fire, to decrease the applied uniquely, CAA found that the expected amount of water required. Full-scale effective- saving of life was much higher for water spray ness tests with the zoned CWSS showed that, and smokehoods than the other options. Indus- along with improved visibility, temperature and try has argued that the study was biased toward water spray because the majority of the aircraft included in the assessment were first- and sec- 23 Richard G. Hill et al., “Evaluation and Optimization ond-generation models that lacked many of of an On-Board Water Spray Fire Suppression System in today’s fire safety improvements and had Aircraft, ” paper presented at the Twelfth Meeting of 30 United States-Japan Panel on Fire Research and Safety, higher accident rates. Changing demograph- Tsukuba and Tokyo, Japan, Oct. 27-Nov. 2, 1992, p. 1. ics indicate that the average age of airplane 24 Petrakis, op. cit., footnote 6, p. 43; and U.S. General passengers will be increasing, suggesting that Accounting OffIce, Aviation Safety: Slow Progress in the ability of passengers to move about and Making Aircrafl Cabin Interiors Fireproof, GAOIRCED- 93-37 (Washington, DC: U.S. Government Printing Of- fice, January 1993), p. 26. 28 Hill et al., op. cit., footnote 23, P. 1. 25 Tho~s L. Reynolds, “Study of the Disbenefits Cre- 29 The five proPsed improvements were: smokehoods, ated by the Installation of Water Spray Systems for Fire CWSS, improved access to overwing exits, more fire-re- Protection of Aircraft Cabins, ” Final Report and Industry tardant cabin lining materials, and minimum spacing Concerns Preface, December 1992, p. iv. Figures cited requirements for seats. Ron Ashford, “Air Safety Regula- for systems installed aboard Boeing 757-200 and 747-400 tion and Its Commercial Impact, ” l%e Aeronautical airplanes, respectively. For the 747, the corresponding Journal, vol. 95, No. 943, March 1991, p. 85. revenue loss, based on $1 million and 200 lbs. per passen- 30 Thomas L. Reynolds, “Study of the Disbenefits er seat, would be roughly $10 million per year. Created by the Installation of Water Spray Systems for 5 6 Ibid. Fire Protection of Aircraft Cabins, ” draft report, July 27 Ibid., p. vii. 1992, p. 63. Chapter 3 —Research and Technologies for Evacuation Systems . 31

=—--. .——

/,// “’z~, -: 32 ● Aircraft Evacuation Testing: Research and Technology Issues

rapidly exit the aircraft if necessary will be Flight attendants’ spokespersons cite fatigue diminished. In general, then, technologies that from lengthy duty times as providing potential further mitigate the thermo-toxic effects and for diminished capability during emergencies. extend cabin survivability periods would have However, the quality of their initial and recur- greater benefits than attempts to further speed rent training is perhaps more crucial. Flight the evacuation rate. attendants rely heavily on this training in emergency situations because real emergencies are rarely encountered in commercial aviation, Risk/Risk Assessment providing little opportunity to practice the nec- When the interactive effects of introducing a 32 essary skills. Technologies assuming a larger new technology are considered, the overall re- role in training flight attendants include motion- sult may be less rather than more safety. Water based cabin simulators, fill-scale cabin/cockpit spray systems may reduce the risk of fire-re- evacuation trainers, cabin evacuation simula- lated fatalities but could contribute to an overall 33 tors, and actual aircraft. Some operators also increase in risk to passenger safety-for exam- 34 use computer-assisted instruction. However, ple, inadvertent discharge during takeoff or the training provided in mockups does not test landing phases of flight may distract pilots or the flight attendants’ ability to manage passen- cause critical avionics to fail. Similarly, while ger flow, which has become increasingly im- smokehoods could extend survivable conditions 35 portant as seat density has increased. for a fraction of passengers, other passengers who might also have survived may, by delaying NTSB recommends that FAA require their escape in order to don smokehoods, be evacuation drills and group exercises during overcome by fire and smoke despite the recurrent training, and that flight attendants breathing assistance. demonstrate proficiency in managing passenger flow with verbal commands when competitive In addition to technical feasibility and 36 behavior is displayed. cost/benefit analyses, risk/risk assessments must be an essential part of the decisionmaking No matter how well-designed an aircraft or process when the likely safety improvement well-trained the flight attendants, passengers afforded by new technology is marginal. This is can undermine the safety capability by bringing especially true of commercial aviation, where on board excessive or inappropriate carry-on the overall fatality risk to passengers is already baggage, damaging safety equipment, or less than 1 in 10 million per flight. drinking to the point of becoming unable to respond to emergency instructions. In the 1992 evacuation from an L-1011 (see box 3-A), one Training and Operations passenger insisted on keeping a set of large The ability of flight attendants to quickly 37 animal horns while he exited the plane. More .“ assess and respond to an in-flight or ground emergency affects passenger safety as much as the design of the aircraft and the performance 32 Ibid., p. 1. of emergency equipment. The National Trans- 33 Ibid., p. 18. portation Safety Board (NTSB) believes that as 34 Ibid., p. 19. 35 Nora Marshall, senior accident investigator, National the crashworthiness of aircraft and survivability Transportation Safety Board, persoml communication, continues to improve, flight attendants “. . . are NOV. 16, 1992. assuming a more critical role for ensuring pas- 36 Natio~] Transportation Safety Board, Flight Afleti~ senger safety. “31 Training and Pe@ormance During Emergency Situations, Special Investigation Report, NTSB/SIR-92/02 Washington, DC: June 9, 1992), p. 31. 31 NatjOm] Trmspwlalim Safety Board, ~lighf~fleti $ 7 Independent Federation of Flight Attendants, Training and Performance During Emergency Situations, “Recommendations to the National Transportation Safety Special Investigation Report NTSB/SIR-92/02 Board Concerning Trans World Airlines Flight 843, ” July (Washington, DC: June 9, 1992), p. 37. 30, 1992, p. 3. Chapter 3-Research and Technologies for Evacuation Systems . 33

Box 3-A–TWA Flight 843 Evacuation

On July 30, 1992, shortly before 6 pm, TWA Flight 843 from New York to San Francisco aborted a takeoff from JFK airport. The plane quickly came to rest to the left of the runway and caught fire. Despite having but three of eight operable exit doors, there were no fatalities, in part due to the presence of off-duty flight attendants. According to preliminary National Transportation Safety Board (NTSB) investigations, the Lockheed L-101 1 took off as normal and rose 50 to 100 feet before returning to the ground. Some passengers and flight attendants commented that something felt amiss with the plane prior to and during liftoff, but they could not be any more specific. Crew members and witnesses indicated that the aircraft landed very hard, causing the wings to flex excessively. A crew member in a plane awaiting takeoff reported that he saw and smelled jet fuel emanating from the plane immediately after it came down. A fire quickly ensued and engulfed the aft portion of the plane, preventing the evacuation of passengers from all but three forward exits. By all accounts, the flight attendants responded swiftly, and evacuation was complete in approximately 2 minutes. Of the 273 passengers, 10 were injured, only 1 seriously. Flight attendants aboard the L-1011 stated that some passengers panicked and left their seats before they were told to do so and before the plane completely stopped. Investigators noted that a significant number of passengers climbed over the seat backs in order to exit the plane. Nine flight attendants were assigned to flight 843, three more than the six required by Federal Aviation Regulations, and five off-duty flight attendants were on board as passengers. According to an Independent Federation of Flight Attendants report, the eight additional flight attendants played a significant role in the safe evacuation of the passengers. For example, the on-duty flight attendant assigned to the L-2 exit could not see if there were flames outside through the door’s prismatic window. When she moved to a passenger seat window to get a better view, an off-duty flight attendant took over her post and prevented passengers from crowding the exit. The off-duty attendant then opened the hatch when the on-duty flight attendant verified that it was safe to do so. Subsequently, passengers became jammed at L-2, and the on-duty attendant instructed them to proceed to the L-1 exit.

SOURCES: National Transportation Safety Board, Office of Aviation Safety, “Factual Report of Investigation, ” unpublished report, 1992; and Independent Federation of Flight Attendants, “Recommendations to the National Transportation Safety Board Concerning Trans World Airlines Flight 843, ” July 30, 1992.

thorough enforcement of carry-on luggage rules evacuations may in fact indicate that additional has also been sought by flight attendant unions. passenger training could have a negative effect on overall passenger safety .38 Rather than In 1985, the Training and Operations Work- withhold information that may assist passengers ing Group, established for the FAA’s technical in surviving a real emergency, crew coordina- conference on emergency evacuation, recommended that FAA conduct research in communication techniques, behavioral sciences, and 38 Federa] Aviation Administration, Emergency Evacu- ation Task Force, “Report of the Training and Operation optimum learning situations to further improve Working Group Meeting, December 3-4, 1985, comprehension and retention of safety instruc- Washington, DC, ” Td.rk Force Report on Emergency tions by passengers. FAA responded that the Evacuation of Transpoti Airplanes, DOTIFAAIVS-8611,11 number of passenger-initiated unwarranted (Washington, DC: U.S. Department of Transportation, July 1986), p. 18. 34 . Aircraft Evacuation Testing: Research and Technology Issues

tion and communication could be improved to during evacuation testing by offering financial reduce the potential for unwarranted evacu- incentives to limited numbers of test partici- ations. Other technologies and training aids, pants. Additional work using smoke and cabin . including computer simulation, that may foster water spray has been recently completed. Ac- better communication between the flight crew cording to Lionel Virr of Europe’s Joint Avia- and attendants should be explored. In addition, tion Authority: “. . . the issue of competitive some operators use videos (on newer model behavior must be resolved to allow harmoniza- aircraft) to heighten passengers’ attention to the tion of future evacuation standards. “41 CAMI airline’s safety briefing.39 investigators are considering initiating cooperative research with the United Kingdom’s Passenger education is only briefly mentioned Cranfield Institute of Technology (CIT) to in the National Plan for Aviation Human Fac- 42 compare motivational techniques. tors; flight attendant training is not. According to FAA, a forthcoming revision to the National Building evacuation research has included the Plan is expected to address the cabin environ- design and development of several computer ment. models to predict egress under various fire scenarios. These have some application to the de- EVACUATION RESEARCH AND velopment of models for aircraft. FAA has TECHNOLOGIES supported evacuation model development in Little information is available on the behavior previous years. In 1991, the Air Transport of evacuation systems and passengers in real Association (ATA) began funding research by accidents except for data recounted by wit- the Southwest Research Institute (SwRI) into nesses and survivors after the fact. It is impos- simulation of passenger behavior in aircraft sible to realistically simulate an emergency en- accidents. vironment without exposing participants to This section discusses evacuation R&T pro- considerable danger. To study the effects of grams, including testing to support rule changes various human behaviors on overall evacuation (e.g., CAMI test of seat row separation stan- performance, researchers have used controlled dard for overwing Type III exit), from which and carefully staged emergency scenarios. In an improved understanding of evacuation issues addition, researchers have developed and im- may be derived. It also discusses developments plemented complex evacuation models whose in computer modeling and simulation of pas- results depend on pre-set values or random senger response during an emergency evacu- variables representing human behavior. A ation. number of persons interviewed for this paper felt that more “realistic” evacuation tests that attempt to introduce panic by exposing test Evacuation Testing participants to significant hazards would be In 1986, CAMI studied flow rates through unethical.40 the overwing exits and exit preparation times under the following four different seating con- In the United Kingdom, CAA has attempted figurations: to introduce competition among passengers

39 The sou~em California Safety Institute, a spinoff of the University of Southern California’s Safety Science Department, produces aviation safety videos and conducts training programs, audits airlines’ safety programs, and so 41 Advanc~ Ru]emaking Advisory committee, Emer- forth. gency Evacuation Subcommittee, minutes of the Jan. 24, 40 AI1 Federal biomedi~l and behavioral research utiliZ- 1992, meeting, Washington, DC. ing human test subjects is governed by a common rule for 42 Jeffrey H. Marcus, mawgcr, protection and su~ival the protection of human subjects. See 14 CFR Part 11, Laboratory, Civil Aeromedical Institute, persoml com- June 18, 1991. munication, Jan. 13, 1992. Chapter 3—Research and Technologies for Evacuation Systems . 35

Figure 3-2-- Schematic Representation of the Four Evaluated Seating Configurations

A c FAA standard CAA proposal . Exit hatch

B CAA standard

r\ 37” a ~~ard Seat removed.

KEY: FAA= Federal Aviation Administration CAA=United Kingdom, Civil Aviation Authority

SOURCE: Federal Aviation Administration, 1992.

the existing CAA minimum require- figure 3-2).43 FAA observed no statistically ments; significant difference in exit preparation times the minimum requirements of the CAA for the various configurations. airworthiness notice (see section on After releasing a notice of proposed rulemak- “Exits” in chapter 1); ing for improved access to overwing exits in the existing Federal Aviation Regulations April 1991, the Regulations Branch of the (FAR) minimum requirements; and, Transport Airplane Directorate requested that an alternative proposed by FAA, in CAMI conduct a second study of egress effi- which the seat adjacent to the exit is re- ciency for different seating arrangements.44 moved. Observed egress rates were faster for the 43 pau] G. Msmussen and Charles B. Chittum, me proposed CAA configuration and FAA’s alter- Influence of A~”acent Seating Configurations on Egress llrough a Type III Emergency Exit, Final Report, native arrangement than for the configuration DOT/FAA/AM-89/14 (Washington, DC: December specified in the existing FAR (see diagrams in 1989). 44 Garnet A. Mc~n et a]., Civil Aeromedical Institute, Eflects of Seating Configuration and Number of Type III Exits on Emergency Aircrajl Evacuation, Final Report, 36 . Aircraft Evacuation Testing: Research and Technology Issues

Test results indicated that, of the total time re- using an actual aircraft, a Trident Three; quired to evacuate through a single Type III training and dressing researchers as cabin exit, the amount of time a passenger needs to staff; and move from the center aisle through the seats providing pre-flight briefings and playing and out the exit depends greatly on the ergo- back a sound recording of an aircraft nomic restrictions encountered at the exit starting up and taxiing to a runway, ex- opening (i. e., increasing the pathway width or periencing an aborted takeoff, and being decreasing the restricted distance to be shut down.49 traversed results in shorter egress times) .45 comparing evacuation rates between the Based on the results from evacuation trials series, CIT researchers concluded that increas- with a dual Type 111 exit configuration, FAA ing the width of the bulkhead aperture leads to hypothesized that arranging the seat rows such an increase in passenger flow rates through the that only one pathway leads to each exit would adjacent Type I exit. CIT researchers also maximize the flow rates to and through the concluded that changes to the distances between hatches. Aircraft with exit centerlines 29 inches seat rows on either side of an overwing exit apart (e.g., the Fokker 100) would have diffi- influence flow rates; however, complete re- culty achieving this configuration.46 In May moval of the seat row adjacent to the Type 111 1992, FAA issued a final rule revising seat exit allowed passengers to pool together and spacing standards for rows that lead to over- resulted in slower evacuation rates than those wing exits; the implementation deadline was measured for vertical projections between seat December 1992.47 rows in the range of 13 inches to 25 inches (see figure 3-2).5° In 1987, CIT commenced a CAA-sponsored program of research into passenger behavior A preliminary investigation into effects from during emergency evacuations. Analyses of the presence of nontoxic smoke was initiated in aircraft accidents indicated significant conges- 1989, during which CIT again conducted a tion occurred during some emergency evacu- series of evacuations using varying bulkhead ations at galley entrances and overwing (Type apertures and distances between seat rows next III) exits.48 CIT research sought to determine to overwing exits. CIT found that the presence whether an optimum aisle width through the of smoke significantly reduced the rate at which cabin divider (bulkhead) near the Type I exit or test volunteers were able to orderly evacuate an optimum seating configuration adjacent to the aircraft. At CAA’s request, CIT also inves- Type III exits existed. Two independent series tigated the effects of nontoxic smoke and cabin of evacuation trials using different bulkhead configuration using competitive behavior. CIT apertures and seating configurations were per- found significant differences in egress rates for formed, with one series employing financial four alternative seat spacings adjacent to Type incentives to foster competitive behavior among III overwing exits, but observed no statistically test participants. significant differences for evacuations through various bulkhead apertures.51 CIT efforts to introduce as much realism as possible during the test included: After comparing the results of these tests with data from the earlier noncompetitive evacuation DOT/FAA/AM-92/27 (Washington, DC: U.S. Depart- trials involving nontoxic smoke, CIT research- ment of Transportation, August 1992), p. 1. 45 Ibid., p. 5. 49 Ibid., p. 8. 46 Ibid., p. 6. 50 Ibid., p. 12. 47 57 Federa/ Register 19220 (May 4, 1992). 51 Hoc. Muir et al., Cranfield Institute of Technology, 48 Helen C. Muir and Trevor J. Gilpin, “Egress UP- Applied Psychology Unit, Aircrafl Evacuations: Competit- date, ” paper presented at the Eighth AMuid International ive Evacuations in Conditions of Non-Toxic Smoke, CAA Aircraft Cabin Safety Symposium, Costa Mesa, CA, Feb. paper 92005 (London, England: Civil Aviation Authority, 4-7, 1991, p. 2. March 1992), pp. 9, 11. Chapter 3-Research and Technologies for Evacuation Systems . 37 ers determined that the presence of a competi- More complex network and queuing models tive element had a significant impact on egress have been used to represent the characteristics w rates for evacuations through the bulkheads, of evacuation systems. Network models, but did not affect the rate of evacuation through graphic representations of paths by which ob- the Type III exit.52 In the latter case, the dif- jects may move from one point to another, are ference in seat spacing (vertical projection) was useful for minimizing the time or distance of the dominant factor in egress rates. point-to-point travel but can quickly grow too complex for efficient use on computers .55 CAA also commissioned a study of human Queuing models describe the dynamics of . factors aspects of water spray system use dur- waiting lines, time-dependent processes that ing cabin evacuations. Using a 707 aircraft 56 obey the laws of probability. The initial frame, CIT conducted eight full-scale evacu- population distribution and the probability of a ations, half in dry conditions. Mean evacuation person moving from one station in an evacu- times for the two conditions were virtually ation system to another determine the waiting identical, suggesting the operation of the CWSS times and exit throughput. did not affect evacuation rates.53 CIT researchers identified no significant visibility Simulation relies on computer-generated ran- problems or hazards from wet cabin furnishings dom numbers to represent processes whose and floor surfaces. values cannot be approximated analytically. Parameter variability can be modeled with Human behavior in actual emergency probability distributions; step-by-step and item- evacuations or even demonstrations for FAA by-item, the simulation predicts what is likely certification cannot be extrapolated from the to happen by running the model through several results of these series of CIT/CAA tests (e.g., 57 conditions. For example, the influence of because of the differences in participant demo- various hesitation times in the face of a grow- graphics and small sample sizes). However, the ing fire threat could be observed using com- data have provided insight into the effects of bined simulation models of aircraft evacuation changes in human motivation and the cabin and fire performance. environment on evacuation capability. “A model is only as good as the parameters which describe the system . . . any evacuation Computer Modeling and Simulation models developed and used will need an exten- The mathematical models used by aircraft manufacturers to predict evacuation times are sive program of parameter determination and sensitivity analysis, and an equally extensive simple calculations of total escape times based 8 validation effort. “5 For example, if the pres- on empirical relations for equipment prepara- ence of passengers with disabilities is assumed, tion and deployment times and the average simulation results are of little use unless good throughput of exits. (These relations are de- approximations (distributions) of seat exit and rived from the results of research experiments aisle flow rates are incorporated. Both general and demonstrations for evacuation certification, evacuation models and simulation efforts not from actual emergency evacuations. ) specific to aircraft are discussed below.

52 Ibid., pp. 18-19. 53 Researchers noted that the sample SIZeS were small 54 John M, watts, Jr., and that the test results are not as statistically reliable as “Computer Models for Evacu- those derived from a larger sample. D.M. Bottomly and ation Analysis, ” Fire Safety Journul, vol. 12, 1987, p. H.C. Muir, Cranfield Institute of Technology, Applied 241. Psychology Unit, “Aircraft Evacuations: The Effect of a 55 Ibid., pp. 237-238. Cabin Water Spray System Upon Evacuation Rates and 56 Ibid., p. 240. Behaviour, ” report prepared for the Civil Aviation 57 Ibido, pp. 242-243. Authority, February 1993, p. 5. 58 Marcus, op. cit., fOOtnOte 42” 38 . Aircraft Evacuation Testing: Research and Technology Issues

General Models and Assessment changing fire conditions. Neither panic, push- Assessment and modeling of flow problems ing, nor falling was assumed. involving people began in the early 1980s. Certain methodological problems limit the Several models were developed to estimate the study of human behavior in fires: experimental time required for groups of people to evacuate subjects cannot be placed in real fire (or crash) a given space or building. Building evacuation situations; testimony obtained after the fact was modeled for situations in which the number from participants in fires may contain errors; of people inside a lobby affected the rate of exit 59 and conclusions must be drawn cautiously from the lobby, and where inhabitants may or 60 where sample data are limited or not represen- may not be alerted before beginning egress. 64 tative. In each case, the network flow solution method assumed egress occurred through well-defined In general, egress research (to fill models’ passageways.6l Other critical assumptions data gaps) has fallen into three main categories: typical of the general approaches to solving field studies of circulation facilities in non- related flow problems included: emergency conditions; laboratory studies (e.g., sign visibility in smoke); and post-incident sur- • Any congestion will occur at doorways, 65 veys of human behavior in emergencies. The and flow through vertical and horizontal nature of case studies has progressed from passageways will be relatively free flow- mainly descriptive to more complex, analytical ing; ventures that attempt to identify typical behav- • Doors serve to meter flow to about one 62 ior patterns or correlate behavior and fire de- person per second per door. 66 velopment . The building models do not consider damage 63 Despite the frequent use of the term “panic” to exits as flow obstructions. Implicit as- to describe human response to emergency sumptions about nonvarying door and passage- situations, particularly fire, researchers have way dimensions and stairwell and hallway flow concluded that “. . . people generally respond rates do not apply to cabin evacuation, and the to emergencies in a ‘rational, ’ often altruistic models are inappropriate for conditions involv- manner, in so far as is possible within the con- ing aisle congestion. None of the models at- straints imposed on their knowledge, percep- tempted to incorporate human decisionmaking 67 tions, and actions by the effects of the fire. “ into the process, particularly in response to Continued research into the reasoning and motivation behind individuals’ actions, altruistic or not, is necessary. Existing models typically do 59 R.L. ~~a~ci~, A Negative fipone~la[ SOhUIOn to ~ not represent the perception of cues, investiga- Evacuation Problem, Report NBS-GCR-84482 (Gaithers- tive behavior (e.g., looking for the fire), and burg, MD: Natioml Bureau of Standards, December general coping behaviors.68 1984). 60 D*M. Alvord, ?7w Fire Emergency Evacuation Si~- These data have limited application to aircraft lation for h4ultijhmily Buildings, Report NBS-GCR-84483 evacuation. For example, some of the indeci- (Gaithersburg, MD: Natioml Bureau of Standards, December 1984). 61 T.M. Kisko and R.L. Francis, Network Models of M R.L. Paulsen, “Human Behavior and Fire Emergen- Building Evacuation: Development of So@are System, cies: An Annotated Bibliography, ” Report NBSIR 81-2438 Report NBS-GCR-82-417 (Gaithersburg, MD: Natioml (Gaithersburg, MD: National Bureau of Standards, Bureau of Standards, December 1982). December 1981), p. 3. 62 Haro]d E. Nelson, ‘Fireform’-A Computerized Col- 65 Hamish A. MacLennan, “Towards an htegrakd lection of Convenient Fire S@ety Computations, Report Egress/Evacuation Model Using an Open Systems Ap- NBSIR 86-3308 (Gaithersburg, MD: National Bureau of preach, ” in Proceedings of the First International Sympo- Standards, April 1986), p. 27. sium of Fire Safety Science, n.d., pp. 581-590. 63 Matthew McCormick, chief, Survival Factors Divi- 66 Paulsen, op. cit., footnote 647 P. 5“ sion, National Transportation Safety Board, persoml 67 Ibid., p. 4. communication, Nov. 16, 1992. 68 Mac~~n, op. cit., footnote 65. Chapter 3—Research and Technologies for Evacuation Systems . 39 sion attributable to not knowing for certain if a language, developed by IBM.70 To estimate and fire has broken out is often absent, time scales analyze the escape process, the model used are different, and escape modes differ from the statistical functions to control passenger move- well-defined hallway model typically used. ments and to advance time related to each However, data on how smoke and the sight of event.71 First applied to evacuation tests in two fire affect decisionmaking skills, sex- or age- single-aisle, narrow-body aircraft configu- related effects, and so forth likely would be rations,72 FAA further developed the model to transferable. represent evacuations from wide-body aircraft.73 The model provided for passenger Computer-based models that incorporate both reassignment to equalize the length of queues human performance parameters and system before exits. An average of 20 runs was used to characteristics are being developed. Simulation evaluate each scenario. requirements include: assessment of risk from, for example, crash-related injury, smoke/gas FAA executed simulations of evacuations toxicity, and fire; typical human behavior un- from DC-10, L-101 1, and B-747 aircraft during der stress, darkness, and smoke; basic response the same period the wide-body aircraft were times under best conditions; and decision- undergoing evacuation certification tests. The making parameters. The generality and validity simulation results correlated well with full-scale of these models must be determined. Models demonstration times.74 However, the simulation must address performance in dynamic large- model could not assess a priori the effects of scale, multiperson systems, as well as the ef- human behavior. Although FAA’s model fects of stress and emergencies. predicted the total evacuation time for 527 passengers aboard a 747 would be 84 seconds, in a In addition, the NIST Building and Fire demonstration for certification participants ex- Research Laboratory has developed the 75 ited the aircraft in under 67 seconds. FAA HAZARD1 model of evacuation from burning attributed the difference to the motivation of buildings using fire survivor psychological data 69 passengers and crew. to construct human behavior parameters. HAZARD1 analyzes the fire environment for Recent and Continuing Efforts. Under a allowable egress time and demonstrates FAA/CAMI-sponsored contract initiated in evacuation of building occupants based on 1987, Gourary and Associates developed a behavioral rules obtained from interviews with clock-driven simulation model of the aircraft fire survivors. NIST staff acknowledge the data evacuation process for use on a computer. Each are skewed in favor of successful behavior; cycle, the model recalculates the position of those who did not survive cannot be inter- each passenger subject to initialized variables: viewed. The software can be modified to do exit preference; endurance, or probability of probabilistic branching for non-universal be- surviving heat or smoke; agility; and “wake-up havioral patterns; the current version uses only a deterministic approach, Other work at NIST TO J*DO Gamer et al., GPSS Computer Simulation of Aircrafi Passenger Emergency Evacuations, FAA-AM-78- relates to congestion in large buildings. Uni- 23 (Washington, DC: U.S. Department of Transporta- versity of Florida/Gainesville researchers have tion, Federal Aviation Administration, June 1978), p. 1. developed a version of HAZARD1 that allows 71 ~rl D. Fo]k et al., GPSS/360 Computer Models TO optimization of building and fire safety designs. Simulate Aircraft Passenger Emergency Evacuation, FAA- AM-72-30 (Washington, DC: U.S. Department of Trans- Aviation-Specific Models portation, Federal Aviation Administration, September 1972), p. 1. In the early 1970s, FAA developed a com- 72 Ibjdo, p. 8. Cornpmkon of test rfNIkS fOr 134- and puter model of aircraft evacuation using 234-passenger loads showed that larger exits used in the General Purpose Simulation System (GPSS) latter case allowed higher flow rates through the doors, 73 Garner et al., op. cit., footnote 70, p. 1. 74 Ibid., p. 5. 69 Bukowski, op. cit., footnote 1. 75 Ibid., p. 6. 40 . Aircraft Evacuation Testing: Research and Technology Issues

time, ” or the time it takes for a passenger to SwRI’s literature search yielded no system- begin to move purposefully (reflects shock and atic overview of the evacuation process; rather, the capability of opening one’s lap-belt).76 the effort produced references to work on spe- Increases in heat or smoke, passenger cific issues and concerns.79 The SwRI model “fatalities,” and disabled exits affect the flow variables address situational characteristics; rates through aisles and doorways (i.e., transi- passengers’ physical and psychosocial charac- tional probabilities). teristics, including motivational variables; and behavioral outcome variables. The latter in- The Gourary model is not comprehensive in cludes initial response, helping another passen- terms of the human behavior assumptions, but 80 ger, panic, and competitive behavior. it does portray passenger evacuation under a variety of crash/fire scenarios described by 40 Data Issues. Although FAA’s simulation crash characteristics and narrow-body aircraft model provided for variations in passenger cabin layouts. mix, seating and exit configuration, door- opening delay, time on the escape slide, and None of the simulation models described slide capacity, insufficient data were available above directly addressed psychological factors. to establish appropriate variables representing Manufacturers and researchers lack the data to 8l the different influences on evacuation rate. determine how much of a role these factors Also, the lack of data on the effects of adverse have in the overall success of evacuation. With conditions (e.g., smoke and debris) prevented development and validation of adequate pa- 82 their simulation. rameters, the simulation may closely approxi- mate an emergency evacuation.77 Boeing said that its own simulation effort in the 1980s was dropped in the belief it would ATA-sponsored research by the Southwest not significantly improve the evaluation of Research Institute seeks to simulate passenger evacuation systems and procedures, given the egress under a variety of evacuation conditions lack of evacuation data to substantiate the and passenger characteristics. ATA hopes to simulation model and the reliability of its exist- develop safety requirements that are sufficient 83 ing mathematical models, for all passengers, including persons with disabilities, in all evacuation circumstances. The Today, as in the 1970s, no central clearing- SwRI four-phase effort aims to create an air- house for evacuation data exists.84 The largest craft evacuation (AIREVAC) computer model collection of data published to date, the Aero- to simulate passenger behavior during emer- space Industry Association’s report on its year- gency evacuations. Phase 1, completed in Sep- long evacuation system study, was completed tember 1991, entailed a literature search to prior to the conduct of most wide-body aircraft identify variables, mathematical relations, and certification tests.85 Phase 3 of the SwRI other information necessary to construct the model, scheduled for Phase 2 of the project. The model validation will be based on either archival evacuation data or on data from a new evacuation exercise .78 79 Ibid., p. 8. 80 Ibid., pp. 9-11. 76 GouraV Ass~iates, Inc., “Evacuation of passengers 81 Gamer, et al., op. cit., footnote 70, p. 2. From Transport Aircraft: A Microcomputer-Based 82 Ibid., p. 11. Model, ” User’s Manual for Evacuation Simubtor Dem- 83 George Vewiogiou, senior mamger, 747/767 payload onstration Diskette, Version 7.03, May 31, 1991, pp. 7-9. Systems, Boeing Commercial Airplane Group, personal 77 Marcus, op. cit., footnote 42. communication, Dec. 12, 1991. 78 James E. Schroeder and Megan Tuttle, Development M Gamer et al., op. cit., fOOtnOte 70, P. 4. of an Aircraft Evacuation (AIREVAC) Computer Model, 85 ~id., p. 4. See also Aerospace Industries Association SWRI Project No. 12-4099 (San Antonio, TX: Southwest of America, Inc., “Evacuation, ” Technical Group Report Research Institute, Sept. 30, 1991). AIA CDP-4, hdy 1968. Chapter 3—Research and Technologies for Evacuation Systems . 41 simulation task, designated for filling data environment information, parameterization of “holes,” is yet unfunded by ATA.86 flight attendant and passenger behaviors, and a test bed for evaluating wide-body aircraft sce- Studying events related to single-aisle, nar- narios are required to validate evacuation row-body aircraft is possible with test beds in simulations. the United States and elsewhere. However, there is no research facility in the world that Even augmented, validated simulations may can be used for investigating wide-body, dual- have their detractors. One passenger advocacy aisle aircraft evacuation issues .87 The fiscal group has expressed alarm at the possibility that year 1995 FAA capital budget contains funding the SwRI computer simulation models spon- for such an evacuation facility. (A 747-100 has sored by ATA will be used to rationalize limit- been offered to CAMI--the difficulty lies in ing the number of passengers with disabilities getting it to Oklahoma City.) Just as certifica- allowed on board transport aircraft.88 One can tion of the 747, with its dual-aisle configura- expect that this or any other test or analysis of tion, introduced more complexity into analyti- evacuation performance is likely to produce cal methods, the proposed super jumbo aircraft slower egress times as the percentage of older (seating 550 to 800) will also stretch the capa- passengers, children, or persons with disabili- bility of existing models and facilities. ties on board aircraft increases. This fact of life, along with equity and other issues, will The need for more data to extend the utility affect those finally making policy decisions. and reliability of the simulation technique is apparent. Improved accident data analysis, passenger demographics information, thermo-toxic

88 Fred Cowell, executive director, Paralyzed Veterans 86 James Schroeder, president, Applied Human Factors, of America, statement before the Emergency Evacuation Inc., personal communication, Oct. 28, 1992. Subcommittee of the Aviation Rulemaking Advisory 87 Marcus, op. cit., fOOtnOte 42. Committee, May 14, 1992.