Propagation of modeling uncertainty in stochastic heat-transfer simulation using a chain of deterministic models

When using a chain of numerical models in a stochastic simulation, the distribution of the observed output depends on both the input parameter uncertainty and the errors of the individual models in the chain. In this work, the propagation of model uncertainty is studied in a simple one-dimensional heat-transfer system. The errors in temperature are found to depend on the heat flux coupling scenario and on the type of the input parameter distributions. The radiation heat flow boundary condition limits the error propagation by compensating the gas temperature errors through enhanced heat losses. Model biases were found to be detrimental to the accuracy of the predicted probabilities of exceeding safety criteria. Finally, corrections to the predicted distribution moments are proposed and tested, showing that the error contributions can be effectively eliminated from the observed distributions if the properties of the individual models are well known.