Metamodeling techniques are commonly used to replace expensive computer simulations in complex engineering optimization problems. Due to the discrepancy between the simulation model and metamodel, the prediction error in predicted responses may lead to a wrong solution. To balance the predicted mean and prediction error, the efficient global optimization (EGO) algorithm using Kriging predictor can be used to explore the design space and find next sample to adaptively improve the fitting accuracy of the predicted responses. However in conventional EGO algorithm, adding one point per iteration may be not efficient for the complex engineering problems. In this paper, a new multi-point sequential sampling method is proposed to include multiple points per iteration. To validate the benefits of the proposed multi-point sequential sampling method, a mathematical example and a highly-nonlinear automotive crashworthiness design example are illustrated. Results show that the proposed method can efficiently mitigate the prediction error and find the global optimum using fewer iterations.
High-Performance Computing Strategies for Complex Engineering Optimization Problems
