中文摘要：

针对复杂工程多目标优化中求解精度受近似模型精度影响大的问题，提出了一种基于信赖域近似模型管理的多目标优化方法。该方法在每一个优化迭代步中首先采用拉丁超立方实验设计方法获得样本点，并基于这些样本点建立各目标和约束的二阶响应面模型，接着用所建立的响应面模型代替真实模型进行多目标优化，优化算法采用微型多目标遗传算法，然后通过信赖域模型管理方法来管理近似模型。该方法大大降低了近似模型对求解精度的影响。该方法在车身薄壁构件的耐撞性优化中的应用验证了其解决复杂工程多目标优化问题的能力。

英文摘要：

A multi-objective optimization method based on the trust region model management framework was suggested to reduce the dependence on the approximations. In each optimization iteration, the approximation models were constructed by the quadratic response surface approximation technique with the samples obtained from the Latin hypercube design. And a Pareto optimal frontier predicted by the approximations was identified through the micro multi-objective genetic algorithm, and the trust region model management framework was employed to manage the approximations. Eventually, the present method is successfully applied in the thin-wailed sections for structural crashworthiness.