Numerical analysis of large-scale structures can be computationally expensive, thus hindering the design optimization. Surrogate modeling techniques are studied to alleviate computational costs. First, engineering knowledge is used to reduce the modeling dimension. Then genetic algorithm is performed to search for the best regression functions. The quality of surrogate models is improved adaptively and parallelized computing is carried out. It is shown in an aircraft wingbox and an automobile front crash system that the techniques can greatly reduce computational effort.
«
Numerical analysis of large-scale structures can be computationally expensive, thus hindering the design optimization. Surrogate modeling techniques are studied to alleviate computational costs. First, engineering knowledge is used to reduce the modeling dimension. Then genetic algorithm is performed to search for the best regression functions. The quality of surrogate models is improved adaptively and parallelized computing is carried out. It is shown in an aircraft wingbox and an automobile fr...
»