A hierarchical kriging approach for multi-fidelity optimization of automotive crashworthiness problems

Abstract Multi-fidelity optimization schemes enriching expensive high-fidelity functions with cheap-to-evaluate low-fidelity have gained popularity in recent years. In the present work, an scheme based on a hierarchical kriging is proposed for large-scale and highly non-linear crashworthiness problems. After comparison to other multi-fidelity techniques infill criterion called variable-fidelity expected improvement applied evaluated. This complemented by two innovative techniques, new approach regarding initial sampling novel way generate model crash problems are suggested. For former, modified Latin hypercube sampling, pushing samples more towards design space boundaries, increases quality of selection. latter, projection-based non-intrusive order reduction technique accelerates simplifies evaluation. The investigated application from field automotive crashworthiness—a size problem lateral impact shape frontal impact. use compared baseline single-fidelity saves computational effort while keeping acceptable level accuracy. Both suggested modifications, independently especially combined, increase performance result presented examples.