Clues from Burning Furniture
An SwRI-led study of how upholstered furniture burns will help fire investigators reduce uncertainty in determining the cause of a fire
SwRI members of the Dr. Marc L. Janssens is a senior engineer in the Center for Nuclear Waste Regulatory Analyses of SwRI’s Geosciences and Engineering Division. He has approximately 30 years of experience in fire research and testing, computer fire modeling, codes and standards development, fire hazard and risk assessment and teaching. He is a Fellow of the Society of Fire Protection Engineers and chairman of the ASTM International Committee E05 on Fire Standards.
A furniture calorimeter test was performed on a CAL TB 133 three-seat sofa mockup placed directly under the hood of the oxygen consumption calorimeter.
Cone calorimeter tests were performed on 4 x 4-inch upholstered furniture fabric-padding mockups. The microscale combustion calorimeter (not shown) was used to obtain basic ignition and heat release rate data for milligram-size specimens of furniture component materials.
This figure shows a visualization of FDS flame spread and burning rate predictions for a three-seat sofa mockup ignited in a corner with a small match-like flame.
A screen capture is shown from an instructive video of a test on a used single-seat sofa ignited with an accelerant. The chart in the lower right-hand corner shows the heat release rate (blue curve) and HGL temperature (red curve) and is updated as the test progresses. The remaining quadrants show the burning sofa from three angles.
Comparison between the measured heat release rate as a function of time for a three-seat sofa mockup and corresponding heat release rate predictions from four different upholstered furniture burning rate models.
Fire investigations generally focus on two questions: Where did the fire originate and what was its cause? When the fire occurred under suspicious circumstances, the investigator’s main challenge is to prove beyond a reasonable doubt that the cause was not accidental. In recent years a number of sophisticated scientific tools such as computer fire models and advanced fire test methods have made it easier to meet this challenge.
Upholstered furniture is very often involved in residential fires, either as the first item ignited or as a significant component of the fuel load. The reconstruction of residential fires, therefore, often requires reliable estimates of the heat release rate of upholstered furniture. With an official determination of cause at stake, investigators need to reduce as much as possible the uncertainty of quantifying the burning rate of upholstered furniture.
With funding from the National Institute of Justice, which is a part of the U.S. Department of Justice, a team of engineers at Southwest Research Institute (SwRI) developed guidelines for how to best estimate the burning rate of upholstered furniture and to quantify, or at least optimize, the uncertainty of the predictions.
Under ideal circumstances, for example in a hotel guest room fire, fire investigators might test items identical to those involved in the fire. Necessary data could be obtained from experiments in a furniture heat release calorimeter and small-scale flammability tests. However, even when the test articles are identical, the test data are subject to uncertainty because of measurement errors and an unknown ignition scenario.
It is usually not possible to obtain undamaged items for furniture calorimeter testing; more likely some specimens would be available for small-scale tests. The extent of small-scale testing that can be performed depends on the quantity of available material. If there is enough for cone calorimeter tests, it may be possible to predict the burning behavior of the furniture item with reasonable accuracy. More often, however, not enough material is available for cone calorimeter tests, but only for microscale combustion calorimeter tests, which provide limited information about the ignition and heat release characteristics of the material.
In a worst-case scenario, small-scale tests cannot be performed at all due to lack of funding, time or test material. Here, the best an investigator can do is to determine the general characteristics of the furniture items involved in the fire based on a detailed survey of the fire scene or interviews with people who can identify the type of furniture in the home. From that, investigators can search the literature for heat release-rate data for similar furniture items. But if the tests in the literature used an ignition scenario inconsistent with the one postulated for the fire under investigation, the use of literature data may not be justified without some adjustments. In addition, there are virtually no heat release-rate data in the literature for upholstered furniture that has been ignited with an accelerant. The SwRI team’s study addressed guidelines for these four situations.
The uncertainty associated with quantifying the burning rate of upholstered furniture consists of two components: aleatory, related to uncontrolled (and uncontrollable) random effects; and epistemic, related to lack of knowledge. Aleatory uncertainty can be estimated using standard mathematical techniques. Quantifying epistemic uncertainty, which is often by far the larger of the two components, is much more difficult. Primary sources of epistemic uncertainty of the heat release rate of upholstered furniture include the lack of knowledge of the ignition scenario and limited understanding of enclosure effects.
Full-scale mockup furniture fire tests
The SwRI research team conducted a series of 79 full-scale fire tests on CAL TB 133 upholstered furniture mockups. In the first 19 tests, the furniture specimen was placed under the hood of an open furniture calorimeter. In the remaining tests, the furniture specimen was placed in a room approximately 15 feet long, 11 feet wide and 8 feet high. Construction was light wood-frame with two layers of gypsum board on the inside. The test object was placed on a scale located in a corner opposite an open doorway. The heat release rate for the specimen was measured based on the oxygen consumption technique. Thermocouples were distributed throughout the compartment and in the doorway to characterize the thermal environment in the room during the tests. Heat flux gauges were used to measure the heat flux to the floor and to the walls in the vicinity of the test specimen. Also, video and photographic documentation were obtained for every test.
Mockup cushions were constructed with either of two fabrics (flameretardant and nonflame- retardant treated cotton) and one of six padding materials (low-density polyurethane foam, high-density polyurethane foam, CAL TB 117 compliant fire-retardant treated foam, chloroprene foam and two types of polyester fiber fill). Three ignition sources (small flame, large gas burner and liquid pool fire) and three ignition source locations (top, front bottom and back) were used.
Based on the results of the 79 mockup tests, SwRI investigators concluded the following.
The repeatability of furniture calorimeter tests with a large flame ignition source is very good. Based on four repeat tests, the coefficient of variance of the peak heat release rate at the 95 percent confidence level was approximately 8 percent. This is comparable to the measurement uncertainty of the peak heat release rate, which for most items was between 7 and 9 percent.
The time to the onset of a self-propagating fire was found to be considerably more variable in repeat tests with a small flame ignition source. Peak heat release rate was also more variable than for large-flame ignition tests, although the effect on peak heat release rate was not as pronounced.
The heat release rate of a triple-seat sofa is very sensitive to the location on the top surface where the ignition source is applied. Peak heat release rate with a large-flame ignition source in the center was approximately 2.5 times the peak observed for ignition of one of the side cushions. A similar trend was observed for the small-flame ignition tests.
Based on a statistical analysis of a subset of the full-scale mockup data, it was determined that the ignition delay is affected primarily by the type of ignition source used, and that the peak heat release rate is affected primarily by the type of padding.
Ignition at the back of the furniture generally resulted in a shorter ignition delay but a slower fire growth rate and lower peak heat release rate.
Finally, comparison of heat release data of items tested directly under the hood of a calorimeter versus in the room indicated that enclosure effects were negligible. However, peak heat release rates in the tests under the hood were well below those required for room flashover.
Small-scale flammability tests and computer modeling
Small-scale tests were performed to obtain fire properties of the two fabrics and six padding materials and specific fabric-padding combinations used to construct the mockups. Tests were done in the cone calorimeter and the microscale combustion calorimeter.
Based on studies conducted at the National Institute of Standards and Technology (NIST) in the early 1980s, it was observed that many upholstered furniture items have heat release rate vs. time graphs that are roughly triangular in shape; that is, where peak heat release rate is the triangle’s peak and the duration of flaming combustion forms its base width. This relationship formed the basis for a simple model to predict the heat release rate vs. time of upholstered furniture. The model’s advantage is that the heat release rate can be estimated based on some generic characteristics of the furniture item, the total combustible mass and the effective heat of combustion of the soft materials. Only a few milligrams of material are needed to measure the heat of combustion in the microscale combustion calorimeter. Heat of combustion can also be estimated with reasonable accuracy from tabulated values. The model was slightly modified to improve agreement between calculated and measured peak heat release rates. Two similar but slightly more sophisticated models to estimate the heat release rate of upholstered furniture based on cone calorimeter data were explored as well.
The SwRI team also investigated the NIST Fire Dynamics Simulator (FDS) field fire model to better account for the effect of the exact location of the ignition source on flame spread over the seating surface. Although it was the most advanced model, because it is based on physics rather than correlations, the FDS model has some unique challenges that resulted in consistently under-predicting the heat release rate curve.
Additionally, the team investigated the NIST zone fire model CFAST to determine how the use of the upholstered furniture burning rate models affects the accuracy of hot gas layer (HGL) temperature and heat flux estimates in the room. CFAST predicts the HGL temperature with remarkable accuracy when the measured heat release rate is specified. This implies that the accuracy of the CFAST temperature predictions for the HGL depends on how well the burning rate model predictions agree with the actual heat release rate. CFAST consistently under-predicts the heat flux to the gauges in the test room, even when the measured heat release rate curve is specified.
Used furniture fire tests and modeling
A second series of full-scale tests involved 27 items selected from 22 sets of used upholstered furniture that were obtained from SwRI employees. A reduced number of cone calorimeter and microscale combustion calorimeter tests were performed on the soft component materials of the used furniture. Specimens of the padding materials were also tested to verify their noncompliance with CAL TB117.
The small- and full-scale test data were used to assess the predictive capability of the models for upholstered furniture burning rates. In this case, the models generally significantly under-predicted the peak heat release rate.
Dissemination of results
Results of the SwRI study will be made available to fire investigators, including a database that can serve as a central repository for other relevant data that are now at many places in different formats. In addition, videos were created as training materials to give arson investigators the opportunity to witness the full-scale fire tests. Information on enclosure temperatures and heat release rate will help arson investigators develop an understanding of fire dynamics in upholstered furniture fires.
More work is needed to refine burning rate models developed on the basis of mockup data so that they are more useful in estimating the heat release rate of actual upholstered furniture.The FDS flame spread and burning model showed the most promise because it is based on physics and not correlations. As such, it has the potential of being able to account for ignition source strength, source location and enclosure effects. However, more work is needed to address the challenges that were encountered in initial attempts at using FDS to model furniture fires. A more detailed algorithm is needed to predict opposed-flow flame spread at the sub-grid scale. Also, a better method is needed to account for thickness and heat flux effects. Until the necessary improvements have been made it is recommended that the data and findings from the study be used to guide fire investigators in conducting a sensitivity analysis to determine the most plausible ignition and fire growth scenario.
Questions about this article? Contact Janssens at (210) 522-6655 or firstname.lastname@example.org.
This project was supported by Award No. 2010-DN-BX-K221, awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect those of the Department of Justice.
The author acknowledges the contributions to this research project of Research Engineer David M. Ewan of the Applied Power Division, Research Engineer Christina Gomez and Senior Research Engineer Jason P. Huczek, both of the Fire Technology Department of the Chemistry and Chemical Engineering Division, Institute Analyst Dr. Robert L. Mason of the Fuels and Lubricants Division and Research Engineer J. Marshall Sharp of the Mechanical Engineering Division, all within Southwest Research Institute; consultant Dr. Marcelo M. Hirschler of GBH International; and graduate assistant Kristopher J. Overholt of The University of Texas at Austin. The author also wants to express his gratitude to the technical and administrative support staff at the SwRI Fire Technology Department who worked on this project, to SwRI employees who provided furniture for testing, and to the National Institute of Justice for the financial support that made this research possible.