Application of a Versatile "Real-Space" Validation Methodology to a Fire Model

Abstract

This paper applies a pragmatic approach to validation of a fire-dynamics model involving computational fluid dynamics, combustion, participating-media radiation, and heat transfer. The validation problem involves experimental and predicted steady-state temperatures of a calorimeter in a wind-driven hydrocarbon pool fire. Significant aleatory and epistemic sources of uncertainty in the experiments and simulations exist and are transformed to a common basis of interval uncertainty for aggregation and comparison purposes. The validation comparison of experimental and simulation results, and corresponding criteria and procedures for model substantiation or refutation, take place in real space as opposed to space where various transform measures of discrepancy between experiment and simulation results are calculated and assessed. The versatile model validation approach handles difficulties associated with representing and aggregating aleatory and epistemic uncertainties (discrete and continuous) from multiple correlated and uncorrelated source types, including 1) experimental variability from multiple repeat experiments, 2) uncertainty of experimental inputs, 3) experimental output measurement uncertainties, 4) uncertainties that arise in data processing and inference from raw simulation and experiment outputs, 5) parameter and model-form uncertainties intrinsic to the model, and 6) numerical solution uncertainty from model discretization effects.