Quantifying Error and Uncertainty in CFAST 2.0 Temperature Predictions

Abstract

In this paper the predictive capability of the smoke transport model CFAST 2.0 is evaluated for five different scenario configurations. The evaluation is made by statistical analysis according to a methodology presented in an earlier paper. Model predictions and experimental data, previously published, are compared and quantitative measures of the predictive capability are thus derived. With the quantitative knowledge of the model error, future predictions from the two-zone model can be adjusted so that the error is taken into account. The suitable method of adjustment depends on how the uncertainty is treated in a specific application. This can be done either by using conservative adjustments or by treating the uncertainty in the predictions as a stochastic variable. The evaluation shows that the statistical method is applicable for this purpose and that reduction of the model error can be achieved with the adjustment model for the types of scenarios subject to analysis.

On Quantification of Error and Uncertainty in Two-zone Models used in Fire Safety Design

“…For a given scenario configuration, a prediction can be adjusted according to Equation (17) and the uncertainty in the prediction will be expressed stochastically or a conservative adjustment can be made according to Equation (20) if it is known whether high or low temperature is considered hazardous. In the uncertainty analysis framework that has been briefly presented previously, this information can be used to include model uncertainty in the analysis.…”

“…It is important to point out that these scenarios have not been used in the development of the two-zone model subject for evaluation. Since the purpose of this paper is to focus on the development of the statistical model, the results from the evaluation are presented in a subsequent paper in Journal of Fire Sciences [20].…”

“…The 95% percentile for a normally distributed variable is given by 97.5% % 1.96 % 2. For simplistic reasons, the input needed for the conservative adjustment model can be reduced to two parameters, U adj and à , once the statistical analysis for the scenario configuration has been performed, see Equation (20).…”

A Monte Carlo analysis of the effect of heat release rate uncertainty on available safe egress time

“…Notarianni [3] analyzed the effect of uncertainty related to the critical criteria determining untenable time in the model CFAST using a Monte Carlo simulation. Lundin [4] compared the predicted results using CFAST with the experimental results under the same conditions and discussed the uncertainty of smoke layer temperature predicted by CFAST. Uncertainty analysis is time-consuming as it usually requires many runs of fire models.…”