On the validation of variable fidelity multi-physics simulations

Abstract The difficulties encountered in applying current normative approaches for validation to computational models of complex multi-physics engineering systems are identified and are associated with the untestable, and sometimes unprincipled, nature of these models. The behaviour of a structural panel on the surface of a hypersonic flight vehicle when subject to complex interactions between aerothermal, aeroelastic and material responses is employed as a key exemplar. A wide range of positions in the philosophy of science, that are applicable to validation, are discussed within the context of a schematic matrix, which allows models to be categorised according to whether they are testable and principled. In the absence of test data against which to assess the accuracy of predictions, it is proposed that a model's credibility should be established based on its epistemic values, theoretical ancestry and the credentials of the modelling techniques. This shift from an objectivist to a relativist approach requires the assignment of experts who acknowledge their biases while engaging intellectually and ethically with the model, the community of knowledge and stakeholders, in a hermeneutical approach.