A global–local meta-modelling technique for model updating

The finite element model updating procedure of large or complex structures is challenging for engineering practitioners and researchers. Iterative methods, such as genetic algorithms and response surface models, have a high computational burden for these problems. This work introduces an enhanced version of the well-known Efficient Global Optimisation technique to address this issue. The enhanced method, refined Efficient Global Optimisation or rEGO, exploits a two-step refinement and selection technique to expand the global search capability of the original method to a global–local, or hybrid, search capability. rEGO is tested and validated on four optimisation test functions against the original methods and genetic algorithms with different settings. Good results in terms of precision and computational performance are achieved, so an application for model updating is sought. A penalty function for the finite element model updating is identified in residuals of the modified total modal assurance criterion. Finally, rEGO for finite element model updating is implemented on a hybrid, numerical and experimental, case study based on a well-known experimental dataset and on a higher dimension finite element model of a wing spar. Satisfactory results in terms of precision and computational performance are achieved when compared to the original methods and genetic algorithms, needing two orders of magnitude fewer evaluations and achieving comparable results in terms of precision.