Data fusion for Uncertainty Quantification with Non-Intrusive Polynomial Chaos

This work presents a framework for updating an estimate of probability distribution, arising from uncertainty propagation using Non-intrusive Polynomial Chaos (NIPC), with scarce experimental measurements Quantity Interest (QoI). In recent years much has been directed towards developing methods combining models different accuracies in order to propagate uncertainty, but the problem improving propagations by considering evidence both computational and experiments received less attention. The described here uses Maximum Entropy Principle (MEP) find updated, least biased distribution maximising entropy between original updated estimates. A constrained optimisation is performed coefficients Expansion (PCE) that minimise Kullback–Leibler (KL) divergence estimates, while ensuring new conforms constraints imposed available QoI. this novel constraint used, based upon Dvoretzky–Kiefer–Wolfowitz inequality Massart bound (DKWM), as opposed more commonly used moment-based constraints. Such allows data be informing distribution. • fusing results physical tests simulations employed method demonstrated composite coupons