Kriging模型并行加点策略的多目标优化方法

A Multi-objective Optimization Method using Kriging Model and Parallel Point Adding Strategy

针对复杂机械装备多学科多目标优化设计成本高、周期长等问题,提出一种近似模型与并行加点策略相结合的多目标优化方法。基于Kriging模型,将添加更新样本点定义为同时考虑Pareto最优解和预测误差的动态多目标优化问题,应用改进NSGA-II优化算法和极大极小距离准则,确定最优的并行更新样本点,在提高Kriging模型精度的同时实现多目标优化。测试函数验证和实例结果表明,该方法可有效提高复杂系统多目标优化效率,同时获得收敛性和分散性俱佳的Pareto最优解。

Aiming at reducing computational cost and time for expensive multi-objective optimization problem of complex mechanical systems, a multi-objective optimization method based on approximate model with a parallel update point adaptively adding strategy is put forward. In this approach, based on a Kriging model, the point adding strategy is formulated as a dynamic multi-objective optimization problem which considers both Pareto optimal solutions and model uncertainty. By applying an improved NSGA-II and a maximum distance criterion, multiple update points can be identified for further parallel evaluation. The proposed approach is tested on six numerical functions and one engineering example. The results show that, this method not only is able to obtain Pareto optimal solutions of good convergence and diversity, but also possess the advantage in optimization efficiency.