Evaluation of the Computer Fire Model DETACT-QS

摘要

The proper use of engineering design methods requires an understanding of their applicability and limitations, since all design methods are, at least to a certain extent, empirically based. Equations or constants used within design methods are frequently based on curve fits to data from experiments. Typically, the experiments used to develop the correlations were conducted under a limited set of conditions, e.g., compartment sizes, heat release rates or fire growth rates. If the design method is used for an application that falls outside of the bounds of the experiments used to develop the correlations used in the design method, uncertainty may be introduced. The potential for uncertainty in computer models is greater than within basic closed form equations. Errors can be introduced in the numerical methods used to solve integral or differential equations, or more simply in math errors that were created during coding of the program. To facilitate the use of engineering design methods and the review of designs developed using engineering methods, the Society of Fire Protection Engineers has summarized and evaluated a number of engineering methods, including several of those which predict radiation from pool fires, predict the effects of thermal radiation to people, and predict ignition of objects when exposed to thermal radiation. These methods are typically simple algebraic or differential equations. Given the added potential for the introduction of uncertainty, it is also necessary to evaluate computer models. In many cases, existing computer models have been released to the engineering community without sufficient technical guidance for the user to understand the capabilities or limitations of the model. As noted by Howard Emmons in 1991, "there should be a fire model validation group ..." to review computer models that are used to demonstrate public safety.