Repeatability of pre-flashover fire patterns on gypsum wallboard.
Thesis DisciplineFire Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Unwanted fires result in loss of life and property. These fires can also create an adverse economic impact on a community. The investigation of fires provides a means to identify the cause of the fire in order to develop a knowledge base that could enable the elimination of that cause and thus reduce the losses from unwanted fires. Questions about the lack of science in the practice of fire investigation have been raised during the review of several arson homicide cases and a forensics science review by the U.S. National Academy of Sciences. Specifically, the National Academy of Sciences indicated that “...research is needed on the natural variability of burn patterns...” This study addresses that need in two ways: 1. Examining the repeatability of several small fire sources and the fire patterns that were generated by those fires. 2. Examining the capability of numerical models to simulate the fires and the resulting fire patterns based on input data collected from engineering reference sources, bench-scale experiments, and full-scale fire experiments. This manuscript highlights the many uncertainties involved in what appear to be simple fire experiments. Uncertainties related to fuel type, measurement methods, and analysis techniques were examined. The objective of the research was to examine the repeatability of pre-flashover fire patterns generated from exposure to short duration (300 seconds maximum), well characterized fires. Three different fuels were used: natural gas, gasoline, and polyurethane foam. Each fuel had a similar top surface area. The heat release rate data showed that the variability between replicates was greater for more complex fuels. The variation in peak heat release rate with the natural gas was similar to the expanded uncertainty of the measurement system, 11%. However, the variation in the peak heat release rates of the gasoline and the polyurethane foam increased due to increased uncertainties in the burning behavior of the fuels. The variability in the heat release rate of the fires resulted in higher levels of variability of the replicate fire patterns. The maximum fire pattern heights generated from the natural gas and gasoline fires were shown to have uncertainties of 18% or less based on a Type A statistical analysis with 95% confidence limits. The comparison of the fire pattern heights and the mean flame height demonstrated that the steady state natural gas fires exhibited the highest level of agreement and polyurethane foam fueled fire exhibited the least agreement, with the gasoline fueled fires in between. A range of methods used to simulate characteristics of fires were applied using data from the source fire experiments as input. The empirical-based predictions for heat flux to a target under-predicted the measured values. The gasoline fueled fires exhibited best agreement of 22% or less for the heat flux predictions. The polyurethane foam results exhibited the worst agreement with a difference of at least 40% between the measured and calculated values. Improved understanding of the radiative fraction from different fuels would improve the capability of the predictions. The best agreement between measured flame height, the measured height of the fire pattern, and flame height predictions occurred with the gasoline fueled fires. Given the overlap of the expanded uncertainty and predicted range of values, the measured and predicted heights would be considered similar. The gasoline fueled fires released the highest amount of energy and had a higher radiative fraction than the natural gas and polyurethane foam fueled fires. The agreement between the computational fluid dynamics predicted values and the measurements from these experiments demonstrated examples of agreement that were within the measurement uncertainty estimates as well as examples with poor agreement. The differences were driven in some cases by limitations within the combustion sub-model close to the burner surface; as the distance between the burner surface increased, the predicted values tended to converge with the measured values. The wall located adjacent to the burn had a significant impact on the flow field in and around the flame and plume region. The horizontal flame movements observed during the experiments was not simulated to the same degree by the model. Given the areas around the flame have steep thermal gradients; small differences in position can result in large differences in temperature and heat flux. The models are useful tools for gaining insight into fire behavior but in many cases many require specific experimental data for input or for validation. The availability of well characterized input data is limited, and data for newer materials is unavailable. Additional research is needed for modeling data development protocols. Improved data repositories are also needed. Further research is needed to understand the appropriate resolution for implementing the results based on fuel type, fuel geometry, and ventilation. Most importantly, there is a critical need to extend this research to better understand the capabilities and limitations of predictive methods using fuels and scenarios that fire investigators are likely to encounter during a fire investigation.