Abstract: Surrogate-Assisted Evolutionary Algorithms (SAEAs) are the most popular methods to solve the design optimization problems of expensive and complex engineering systems,which can accelerate the search for a set of Pareto solutions.However,the performance of the existing individual surrogate model is problem-dependent,and the predictive uncertainty will be increased with the increasing number of objectives.Therefore,an ensemble surrogate assisted evolutionary algorithm for complex system many-objective optimization was proposed.The ensemble surrogate model combined with the reference vector replace mechanism was adopted to select the Pareto solutions further.The improving lower confidence bounder utility criterion and the adaptation of sampling selection strategy were used to choose new samples for the actual function evaluation.The newly added samples were used to update the ensemble surrogate model to find the best Pareto solutions.Compared with the existing algorithms,the algorithm’s performance was verified by several used benchmark problems and practical engineering optimization problems.The results showed that the proposed algorithm had well performance and potential.

Key words: many-objective optimization, ensemble surrogate model, lower confidence bounder utility criterion, sampling selection strategy

摘要： 代理模型辅助进化算法广泛用于昂贵的复杂工程系统优化设计,能够加速找到问题的最优解集。然而,单个模型预测性能依赖于具体问题,并且随着目标个数的增加,预测性能的不确定性将随之增加。因此,提出一种基于组合模型的复杂系统超多目标优化算法。首先,建立组合代理模型并结合随机参考向量替代机制,以更好地搜索超多目标问题的非支配解集。其次,基于改进的统计下限最小值（LCB）准则及自适应个体选择策略选择优秀个体进行真实评估,以更新组合代理模型,使其能更好地辅助算法找到最优解集。最后,通过所提算法与已有代理模型进化算法在一系列测试函数和工程优化实例上的对比结果表明,所提算法具有良好的性能和潜力。

关键词: 超多目标优化, 组合代理模型, 统计下限最小值准则, 个体选择策略