TABLE OF CONTENTS A wide range of analytical tools are available to the engineer. [1], [2] When selecting a particular calculation method, its predictive capability must be considered. Guidance for evaluating the predictive capability of computer fire models can be found in ASTM E1355-96. [3] The process of model evaluation is important for determining the appropriate use as well as the limitations for a particular application. Although a model might be appropriate for one fire scenario, it might be inappropriate for another.
Determining suitability for use also requires an evaluation of the sensitivity of the given model. Sensitivity analysis determines the effect of changes in individual input parameters on the results of a given model. The analysis might be done by holding all but one input variable constant and systematically studying the effects of that one variable on the predicted result.
If a high degree of uncertainty about the magnitude or variability of a given input variable for the design fire or the fire scenario exists, the model might be inappropriate if the output is particularly sensitive to that variable. On the other hand, if the given input has little effect on the output, then the tolerance for uncertainty is increased. In most cases, the most significant variable will be the heat release rate history input for the design fire. In some situations, however, heat losses to the compartment boundaries, specification of the physical properties of compartment boundaries, or the size of vent openings might significantly affect the result.
