AC 2007-2770: A COLLECTIVE UNDERGRADUATE CLASS PROJECT RECONSTRUCTING THE SEPTEMBER 11, 2001 WORLD TRADE CENTER FIRE Andre Marshall, University of Maryland James Quintiere, University of Maryland American Society for Engineering Education, 2007 Page 12.16.1
A Collective Undergraduate Class Project Reconstructing the September 11, 2001 World Trade Center Fire Abstract Fire Protection Engineering undergraduate students enrolled in a fire assessment laboratory course conducted their own investigation of the September 11, 2001 World Trade Center disaster by simulating the fire that followed the aircraft impact. The project focused on characterizing the fire on the 96 th floor of WTC1 (North Tower) and evaluating the contribution of the fire to the structural collapse. Students contacted vendors and suppliers for the World Trade Center to get information regarding construction details and fire properties of building materials and furnishings. Students also obtained information reported from the National Institute of Standards and Technology Building and Fire Research Laboratory investigation of the World Trade Center collapse. A 1/20 th scale model of the original structure (including damage effects from the aircraft and liquid fuel dispersed from the aircraft impact) was designed, constructed, and instrumented over ten weeks corresponding to the last half of the semester. Students held briefing for invited guests from the university, government agencies and industry prior to the actual scale model test. Results from the test were recorded continuously with video and with an automated data acquisition system for detailed analysis. Analysis of the results in the scaled spatial and temporal coordinates provided insight into peak temperatures, smoke production rates, and fire growth behavior that may have occurred in the actual WTC1 fire. This classroom study provided an excellent opportunity for students to apply classroom principles to a problem of significant social and engineering relevance. 1. Introduction 1.1. Fire Protection Engineering The Department of Fire Protection Engineering at the University of Maryland offers a unique program of study in Fire Protection Engineering at the undergraduate and graduate levels. We offer the only fully accredited undergraduate program and one of the two graduate degree programs in the U.S. in this area. Fire Protection Engineering includes the thermal and fluid sciences, combustion, materials, human behavior, egress modeling, toxicity, and reliability and risk analysis. In this field, we focus on reducing the burden of fire losses through engineering design, development, and research. Fire Protection Engineers may be involved with the design of fire protection systems; the analysis of fire protection performance in buildings, nuclear power plants, or even aerospace vehicles; or research in areas such as fire propagation, suppression, or detection. Fire Protection Engineering is a multi-disciplinary field that is not fulfilled by any other branch of engineering. Our pedagogical aim is to provide the technical knowledge and skills to either practice or pursue advanced studies in the field. Our curriculum shares a high degree of commonality with the other UM engineering departments in the first two years of study. In the last two years of study, our students take their major courses which range from fluid mechanics and fire dynamics to hazard analysis. This paper showcases a special class project that was conducted in our ENFP320 Fire Assessment Methods and Laboratory Course offered to Fire Protection Engineering students in their third year of study. 1 Page 12.16.2
1.2. Course Objectives and Organization The ENFP320 Fire Assessment Methods and Laboratory Course is designed to give students knowledge and awareness of test standards and measurement techniques in fire. Standard fire tests developed by a number of institutions (e.g. ASTM, NFPA, UL) are often called by prescriptive fire safety codes such as the NFPA101 Life Safety Code [1]. These codes often specify fire protection design requirements based on material classifications determined using standard fire tests. As the field advances beyond prescriptive codes to performance based fire protection design, fire protection engineers must learn to view the fire problem outside of codes and their associated standard tests. Engineering judgment will be required to determine when prescriptive codes or a performance based design analysis is more appropriate. This judgment will require an understanding of the strengths and limitations of the standard tests and how they relate to real fire behavior and the associated Fire Protection Engineering problem. The special class project focused on the World Trade Center fire was intended to provide students with a clearer perspective on the challenges of fire assessment and the various methods available to support Fire Protection Engineering design, analysis, and research. Although the project was central to the course, other laboratory and research assignments were required to offer balance and to ensure that the specific course objectives were met. In this course students are prepared to determine appropriate test methods for assessing a particular hazard, to apply basic calculation techniques to assess fire performance, and to document technical reports in a clear and concise manner. Students were required to conduct two laboratory reports focused on standard test methods, which were prepared individually during the first five weeks of class. These assignments introduced students to safety practices, measurement techniques, and reporting methods used in the laboratory in addition to introducing students to standard test methods. A research report and class presentation was also required documenting the history, measurement goal, and application of a selected standard test method. The test method projects and presentations were conducted in small groups during weeks six thru eleven. This activity exposed students to a wide range of standard test methods well beyond the two that were conducted in the first weeks of class. The same small groups were also given assignments for the WTC project conducted from weeks six through fifteen, which will be discussed in the following. 2. Project Objectives The focus of the World Trade Center fire reconstruction project was to investigate the September 11, 2001 World Trade Center disaster by simulating the fire that followed the aircraft impact. Fire assessment techniques were selected to gain insight into peak temperatures, smoke production rates, and fire growth behavior that may have occurred in the actual World Trade Center fire. This project provided a rich context for teaching fire assessment methods. The power of having students work on a project of such monumental social and engineering relevance cannot be underestimated. This project filled students with curiosity and motivated them to explore many Fire Protection Engineering design, analysis, and assessment issues. These issues included topics ranging from the suitability of structural fire proofing insulation requirements and associated testing to the ability of computer models to accurately simulate fire behavior in this scenario. The specific pedagogical objectives of this project were to 2 Page 12.16.3
provide students with an opportunity to observe and assess fire behavior in its full complexity; teach students to appreciate the limitations of standard tests in predicting fire performance; expose students to fire assessment techniques used in research and development; introduce students to instrumentation capable of measuring the most critical quantities that affect fire behavior. 3. Approach The project specifically focused on characterizing the fire on the 96 th floor of the WTC1 and evaluating the contribution of the fire to the structural collapse. Students studied WTC1 construction plans and contacted office designers and suppliers for the World Trade Center to get information regarding fire properties of building materials and furnishings in the actual model. Next, a 1/20 th scale model of the original structure (including damage effects from the aircraft and liquid fuel dispersed from the aircraft impact) was designed, constructed, and instrumented over the last ten weeks of the semester. The model was developed using physical modeling scaling concepts to determine thermodynamically appropriate construction geometry, materials, and quantities to achieve model similarity. Students also obtained information reported from the National Institute of Standards and Technology Building and Fire Research Laboratory investigation of the WTC collapse to help with the design of the model and evaluation of the results [2]. 3.1. Technical Approach The World Trade Center fire reconstruction project is founded on the basis that full-scale fires can be studied with physical models at smaller scales. This concept is exploited extensively in aerodynamics where lift and drag measurements are performed on small-scale models to determine dimensionless quantities that are used to describe the aerodynamic performance at full scale. Small scale physical models can also be designed to produce fires that behave similarly to full scale fires when comparing dimensionless quantities [3]. Because it was prohibitive to model an entire World Trade Center tower and the fire was a localized event in both of the two towers, a single floor of WTC1 (North Tower) representing the center of impact was chosen as a focus for our investigation. The overall dimensions of the 96 th floor of WTC1 are geometrically scaled by a factor of 1/20 th resulting in overall model dimensions of 3 m x 3 m x 0.15 m as shown in Figure 1. The openings in the tower corresponding to damage from the aircraft and fire were also geometrically scaled. The walls were constructed using materials and thicknesses that were thermodynamically appropriate for achieving model similarity. The fuel load on the 96 th floor was estimated from an analysis of office plans and discussions with office suppliers in addition to estimates of the fuel dispersion after aircraft impact. After the fuel load was determined it was scaled to achieve similarity. The office load was represented using a total of 43 wood cribs made of 19 mm square pine sticks and 5 322 ml pans of fuel mixture (85% kerosene, 15% heptane) representing the dispersed jet fuel. The scale model was fitted with instrumentation throughout its interior. An automatic data acquisition system was used in this study providing a wealth of information regarding the fire spread and growth. Thermocouples were used to measure gas temperatures and surface 3 Page 12.16.4
temperatures throughout the enclosure. Heat flux gauges were used to determine floor heat fluxes throughout the enclosure. External and internal video was also used for qualitative assessment of the fire behavior. A load cell was also used to determine the fuel mass loss rate over time, which can be used to quantify the fire size. Finally smoke samples were collected during the test to evaluate smoke concentrations in the fire. The previously described instrumentation provides qualitative and quantitative information that can be used to evaluate the fire dynamics. The growth of the fire is a key parameter that is often quantified in terms of event times. For comparison of small scale and full scale fire growth, a dimensionless time, t * = t(g /l) 1/ 2, must be used. Temperature is another key factor in evaluating the severity of a fire. However, scale analysis reveals that temperatures between the scale model and the full scale configuration can be compared directly, a direct result of the manner in which similarity was enforced in the model design. 3.2. Project Organization The 45 students in the class were divided into nine teams, with each team assigned a specific role in the fire reconstruction. Five of the teams focused on designing the model to produce similar fire behavior to that which occurred in WTC1. The remaining four teams focused on the design and placement of instrumentation for assessing the fire dynamics within the model. Planning began in the sixth week of the semester. Construction was set for the twelfth week of the semester, culminating in the fire simulation in the fourteenth week of the semester. On the day of the test, students held briefing for invited guests from the university, government agencies and industry prior to the actual scale model test giving student an opportunity to share what they had learned and accomplished during the semester in a professional environment. A brief description of each group s responsibilities is provided in the following discussion. 3.2.1. Floor Construction The students were tasked with researching the floor plan and materials used in the actual construction of WTC1. Information was obtained from the NIST study [2]. Students developed a floor plan for the scale model and selected materials using scale analysis and guidance from the instructors. Although a 1/10 th scale model was initially planned, funding and test logistics necessitated a reduction to 1/20 th scale. The model was constructed in the laboratory in four parts as illustrated in Figure 2. The day before testing the model was moved to the test site for setup and instrumentation. 3.2.2. Damage by Aircraft, Windows Assessing the extent of the damage to WTC1 was a crucial aspect in designing the scale model because openings in the structure control the fire ventilation. The ventilation of the fire is of paramount performance in determining the fire dynamics. The students obtained information from the NIST report [2] and FEMA report [4] focusing on the damage to the ceiling, floor, façade, windows and building core by the aircraft in addition to progressive window failure as the fire progressed. Initial damage was included in the model before testing as shown in Figure 1 and vents were opened remotely during the fire test based on times scaled from observations made during the actual WTC1 fire. 4 Page 12.16.5
3.2.3. Office Load Design Fuel load is the amount of combustible material in a space that will support and spread a fire. It was important to estimate the fuel load characteristics on the 96 th floor of WTC1 in order to design the scale model to produce similar fire behavior. The students determined that the 96 th floor of WTC1 was leased by Marsh and McLennan Companies, a professional services firm specializing in risk and insurance services. Based on reviews of the fire literature,and understanding of the use of the space, and a survey of office plans, students selected an office load of 52 kg/m 2 as a representative fuel load. Realizing that flames were still observed at the collapse time of 102 min, students estimated a burn time for the 96 th floor of WTC1 of 120 min. Finally, students used the estimated office load, total floor area of 2873 m 2, and estimated burn time to approximate an overall burning rate of 21 kg/s. Scale analysis was then used in order to design a small scale office load producing similar dimensionless local and overall burning rates and burn times. The students designed 150 mm x 150 mm x 75 mm wood cribs out of 12.5 mm pine sticks shown in Figure 3 to achieve local burning behavior similar to the actual office fuel load. A total of 43 wood cribs were placed throughout the model providing the appropriate overall mass required to maintain similar overall dimensionless burn rates and times. 3.2.4. Jet Fuel Dispersal Based on estimates from the NIST report, the students determined that there was about 28,400 kg of fuel on the aircraft at the time of impact. Also following NIST analysis, it was determined that 7023 kg of fuel was consumed in three 60 m diameter fire balls generated outside of the building upon impact. Students assumed that about 24,000 kg of fuel spread onto the floor in WTC1 and 9,900 kg or 3,400 gal was allocated to the 96 th floor. Based on likely aircraft maneuvers and the location of impact, the fuel was assumed to be dispersed mostly on the North side of the tower. Assuming a maximum burning rate for the jet fuel and assuming that it was dispersed over 10% of the floor area, an overall burning rate of 10.2 kg/s was determined resulting in a jet fuel fire of 16 min. The students performed scale analysis to determine that a fuel burning rate of 5.7 g/s with a burn time of 218 s would be required to achieve similar behavior. This burning behavior was achieved by distributing 5 pans containing 322 ml of fuel (85% kerosene, 15% heptane) throughout the North side of the model. To simulate the actual event, the jet fuel was used to ignite the fire within the model. Photographs of ignition and pool fire tests are provided as Figure 4. 3.2.5. Temperature Measurement Students designed and fabricated nine small diameter (0.25 mm wire dia) type-k thermocouple probes for measurement of fire temperatures throughout the scale model. These small thermocouples were used to follow the transient behavior of the fire. The themocouple design is included in Figure 5 and probe locations are included in Figure 6. The probes were inserted through the top or bottom of the model and were fixed either 25 mm from the ceiling, 25 mm from the floor, or 50 mm from the ceiling depending on the location. Measurements from the nine thermocouples were recorded using a data acquisition system. 3.2.6. Heat Flux Measurement Students designed and fabricated five plate thermometers to measure the heat flux to the floor of the model. These plate thermometers were constructed out of 19 mm copper tubes having a 5 Page 12.16.6
closed end consisting of a thin suspended stainless steel plate. A small type-k thermocouple is welded to the backside of the plate while the front side of the plate is exposed to the heat from the fire. A rendering of the device is included as Figure 7. With calibration, these plate thermometers can be used to determine the incident heat flux to surfaces exposed to fire. These instruments were located at positions indicated in Figure 6 and monitored continuously with the data acquisition system. 3.2.7. Weight Loss Measurement Two methods were employed to measure the mass loss, which can be directly related to the burning rate. The students designed and fabricated two load cells using strain gauges attached to two large deflector beams as shown in Figure 8. The entire structure was supported on these two beams, which were deflected considerably. The reduction in the deflection was measured with the strain gauge and monitored with the data acquisition system. Using relationships from calibration experiments performed in the laboratory, the mass loss rate could be determined from the raw data. Additionally, these two deflector beams were supported on four bathroom scales, also shown in Figure 8, which were read manually every 5 minutes providing redundancy. 3.2.8. Smoke Measurement Students developed a laser based obscuration meter to measure smoke density and a vacuum filter technique to measure smoke yield. The laser based obscuration meter consisted of an inexpensive laser level device focused onto a photo diode detector. The attenuation of light was controlled in calibration experiments using neutral density filters to calibrate the device. This device was to be placed just outside of one of the vents; however, the photodiode failed during setup on the day of the test. Alternatively, a shop vacuum was used to sample smoke coming from the South face of the model. The sample was directed through a filter every 5 min at a known flow rate of 350 lpm for 30 s. Students later measured the mass of smoke and normalized this measurement with the sample mass flow rate to determine the smoke yield at various stages of the fire. 3.2.9. Data Collection A 20 channel Fluke NetDAQ Data Acquisition system was used to collect data from the instrumented WTC1 model. Students were responsible for learning how to setup and operate the data acquisition system and coordinating with the other measurement groups to ensure that instrumentation locations were accurately documented. Four external video cameras were placed around the model and one internal video camera was used within the model. Over a dozen still cameras were also used to take photographs during the test. Students were also responsible for collecting and archiving digital video and images from the test. 3.3. Project Research Students obtained information about the World Trade Center from publicly available reports such as NIST [2] or the FEMA study [4]. Students also used faculty contacts including Eric Lipton, a Pulitzer Prize-winning New York Times correspondent, who visited University of Maryland to do an invited lecture for the A. James Clark School of Engineering during the project. Students relied heavily on the scientific fire literature and also used Prof. Quintiere s book [4] as a guide. 6 Page 12.16.7
3.4. Equipment and Facilities Primarily existing departmental equipment and laboratory facilities were used for the World Trade Center project. However, University of Maryland s Maryland Fire and Rescue Institute (MFRI) generously provided a large test-site for conducting the test. The total budget for the project with materials and equipment including fuel, construction materials, instrumentation, and video was around $3,000. 4. Outcomes 4.1. Technical Outcomes The scale model fire test revealed many interesting fire dynamics including extreme temperatures and heat fluxes, a relatively cool core region, complete burnout of office load, a delayed involvement of the southwest corner, and under-ventilation conditions during the peak burning period. Figure 9 shows the fire from all directions. You may note that the flames are relatively higher and the smoke less opaque than observations in the actual fire. Both of these effects are likely due to scale. The openings in the model are concentrated in the middle of each face to simplify adding ventilation at the appropriate time. If the flames were distributed over the entire face, shorter flames would have been observed. It should also be noted that because the opacity of the smoke is a function of the path length, the smoke from the small scale model will appear less opaque even for similar smoke density. Figure 10 shows post-burn photographs of the scale model. Figures 10 (a) and (b) show complete burn out of the wood cribs. Figure 10 (c) shows that soot produced in the fire remained on the ceiling in the core region, but was burned off in the extremely hot region surrounding the core. Finally, Figure 10 (d) shows the failure of a scaled insulated truss, which was fixed in the model at the initiation of the test. Temperatures throughout the scale model near the ceiling are provided in Figure 11. Temperatures in excess of 1000 C are measured. These temperatures are extremely high even for a compartment fire. Temperatures increase similarly throughout the model except in the Southwest corner suggesting a delayed involvement of this region. Incident heat fluxes behaved similar to temperature also indicating delayed involvement of the Southwest corner and demonstrating peak heat fluxes in excess of 150 kw/m 2. It should be noted that the scale model behaved almost exactly as predicted producing a dimensionless burn time within 10% of the design value, t * =13000, corresponding to approximately 120 min at full scale and 27 min at the model scale. 4.2. Pedagogical Outcomes Students were graded based on their group project presentation and report. Each student was responsible for a specific portion of the group report for which the student would receive an individual grade in addition to the group report grade. The reports demonstrated that students had developed some proficiency in performing research to find relevant information that could be combined with lecture experiences, and independent laboratory exercises to formulate an approach to their specific problem. The reports showed that the students appreciated being able to observe and interrogate the complex fire within the scale model. The students also demonstrated their understanding of the limitations of the scale modeling technique. However, the reports also revealed that students had not completely mastered some of the more advanced 7 Page 12.16.8
measurement or scaling concepts. Nevertheless, when all of the nine group reports are viewed collectively, the class accomplishments are truly staggering. 5. Conclusions The World Trade Center fire reconstruction project offered students a unique opportunity to work on a problem of social and engineering relevance. The students gained important perspective on fire assessment methods, fire dynamics, and group dynamics. They were exposed to cutting edge analytical techniques in their project research and were encouraged to critically evaluate these techniques and apply them to their respective project activities. Students were able to gain insight in their specific and collective topics through hands on experience working on a focused open-ended problem. Although this project aligned well with the fire assessment course objectives, because of its time requirements it would have been more suitable as a special topics course, possibly involving both graduate and undergraduate students. Bibliography [1] NFPA 101, Life Safety Code, National Fire Protection Association, Quincy, MA, 1994. [2] Final Report on the Collapse of the World Trade Center Towers., (NIST NCS tar1) Federal Building and Fire Safety Investigation of the World Trade Center Disaster, National Institute of Standards and Technology, 2005. [3] Quintiere, J. G., Fundamentals of Fire Phenomena, John Wiley & Sons, Ltd., Chichester, England, 2006. [4] World Trade Center Building Performance Study: Data Collection, Preliminary Observations, and Recommendations, In: T. McAllister, ed. FEMA 403, Federal Emergency Management Agency: Washington D.C., 2002. 8 Page 12.16.9
North Face Figure 1: Photograph of scale model of the 96 th floor of WTC w/o ceiling. Figure 2: Design of WTC1 scale model with detail. 9 Page 12.16.10
Figure 3: Photograph of wood crib representing office load. Figure 4: Photographs of trial tests showing the ignition and burning behavior of three pools ignited by a fuel soaked cloth wick. 10 Page 12.16.11
Figure 5: Thermocouple design and photographs Figure 6: Instrumentation diagram; T thermocouple, HF heat flux gauge, WL strain gauge 11 Page 12.16.12
Figure 7: Rendering of plate thermometer to measure heat flux Strain Gauge Deflected Beam Bathroom Scale Bathroom Scale Figure 8: Photograph showing weight loss measurement instruments. 12 Page 12.16.13
South North East West Figure 9: A view of the fire from all directions. (a) (b) (c) (d) Figure 10: Post fire analysis; (a) and (b) show complete burnout of wood cribs; (c) soot deposition in cooler core region; (d) insulated truss failure. 13 Page 12.16.14
Figure 11: Upper layer ceiling temperatures throughout the fire 14 Page 12.16.15