The main goal of the project is to develop novel Bayesian optimal experimental design (ODE) methods in order to define the most informative, cost-effective, test campaigns for selecting and validating models through the whole assembling scale from coupon to prototype. OED methodologies will be developed for building and refining models of mechanisms, as well as estimating the model parameters that are activated under disparate loading conditions at the assembled substructure or system level (e.g. models of joint behaviour). For nonlinear models, tests will be designed to optimize test characteristics to excite all nonlinearities so that all associated model parameters can be estimated. The project is expected to include: – Development of a comprehensive Bayesian OED framework with computational and software tools for cost-effective experimental design at component, sub-structural assemblies and system levels.
– Exploration of information-based measures based on expected utility functions to create useful metrics for comparing the value of different experimental designs.
– Investigation of OED techniques and means of enhancing the prediction accuracy for important output quantities of interest.
– Development of OED strategies to be robust to uncertainties arising from experimental conditions, operational variations and environmental and manufacturing variabilities.
– Investigation of asymptotic and sampling algorithms, surrogate models, parallel computing strategies etc. in order to significantly reduce the excessive computational effort arising from large number of model runs for large-order high complexity structural models.
– Demonstration and validation examples using virtual and laboratory experiments.