Uncertainty in Smoke Transport Models

Data from full scale experiments are collected and organized in a database. A statistical method is developed to evaluate the uncertainty in predictions of smoke transport models. The method is based on a regression analysis of measured and predicted data. The computer program CFAST is evaluated to exemplify the statistical method. The uncertainty is quantified with a regression coefficient and the residual variance When the model uncertainty is quantified it is possible to adjust the model predictions for the model error. The uncertainty in CFAST’s predictions of smoke gas temperature and position of the interface is investigated for a number of different scenarios