| The increasing cost of full-scale testing makes model-based computational methods very important for the reliability assessment of large complex systems. However, physical, statistical, and model uncertainties make it difficult to have high confidence in model-based reliability prediction. Hence, there is an important need to validate model predictions using test data. However, for large-scale systems, availability of test results is rare. This kind of problem can be validated from obtaining test data to validate smaller modules (subsystem and component-level models) of the overall reliability computational model.;A framework for Validation and Uncertainty Quantification (VUQ) of a model for an overarching problem with no prior experimental data is implemented on one such problem, measurement of combustion efficiency for industrial flares. The hierarchical VUQ process begins with defining lower subsystems. For the current overarching problem, a nonreacting buoyancy-driven turbulent mixing experiment was selected as a component-scale case and wind-tunnel flare experiments as a pilot-scale case.;A 6-step systematic validation framework is adopted from the literature and applied to provide upper and lower bounds of the prediction. Each brick/level in the hierarchy is validated individually as well as together as one big system to propagate the uncertainties and to build confidence in the model. Monte-Carlo method and consistency constraints are used to analyze surrogate models, constructed for complex and expensive multiphysics simulators. The analysis refines the parameter space where the model makes valid predictions and with certain confidence. |
