Generalised inverse uncertainty quantification in numerical models

Since structures are nowadays designed up to their performance limits, uncertainty quantification (UQ) is being increasingly applied in computer aided engineering to account for the fact that not all parameters of the computer models can be exactly quantified. Especially when these parameters vary within a single component, and the considered quantity is not directly measurable, such quantification is challenging. In this case, inverse uncertainty quantification methods, where the uncertain parameters are quantified based on measurements of the system’s responses, are gaining scientific popularity. In this context, two complementary approaches already exist. The first approach, based on a Bayes’ rule, is matured and provides detailed information on the uncertainty, but its accuracy suffers greatly when insufficient system response data are available. On the other end of the spectrum, a method based on interval theory was introduced by the applicant during his PhD. This method excels under limited data but offers less information as compared to the Bayesian approach. This project starts from the idea that both approaches are complementary and aims at unifying them in a generalised framework for UQ. In this framework, the available data decides which of the two yields the best results, or even if a hybrid method that combines both approaches should be applied. Moreover, information is provided on which data should additionally be collected, given the selected iUQ method.