Predictive Modelling

The goal of this book is to present procedures that experimentalists and analysts can adopt in order to verify and hierarchically validate their computational models. Doing so requires establishing the domain over which the model can be used with confidence. This work addresses model validation from start to end, i.e. From the design of validation experiments to the assessment of predictive accuracy.

Verification activities aim to demonstrate that the overall numerical errors and uncertainties are low relative to both the experiment variability and the fidelity of the model outputs to the experiments. There are two categories of verification: code verification and solution verification. Validation activities aim to demonstrate the model's predictive accuracy with respect to a reference to reality, generally provided by a suite of physical experiments. Confidence in this predictive accuracy is gained by quantifying uncertainty and determining robustness against lack of knowledge. The general process of validating engineering models includes test-analysis correlation, sensitivity analysis, model updating, uncertainty quantification and analysis of the robustness of the predictions to lack of knowledge.

The author teams experiences with real applications of practical importance, "hierarchical validation" approaches are presented that can prioritize the validation effort according to the budget, deadlines and deliverables that an engineering project must satisfy. The applications will be drawn from several disciplines, such as engineering mechanics, structural dynamics (both modal/stationary and transient/fast dynamics), material science and computational physics.