面向高维昂贵优化问题的Kriging辅助进化算法研究

昂贵优化问题广泛存在于汽车、船舶、航空、机械制造等领域，提高昂贵优化问题求解能力对提升智能设计水平和推动智能制造发展有着重要作用。Kriging辅助进化算法是求解昂贵优化问题的重要方法，但随着问题维数增高，Kriging模型训练时间急剧增加而Kriging加点准则优化效率逐渐降低，严重影响了其在高维昂贵优化问题上的计算效率和求解精度。针对上述问题，本项目研究基于增量学习的Kriging构建方法，加快算法迭代过程中高维Kriging模型训练过程；构建基于组合优化的Kriging加点准则，提高Kriging辅助进化算法优化效率；分析算法在实际高维昂贵优化问题上的优化性能，验证算法在实际问题上的适应性。项目研究有望为工程中日益增多的高维昂贵优化问题提供新型高效的求解方法。

Expensive optimization problems commonly exist in vehicle engineering, ship engineering, aeronautical engineering, mechanical engineering and so on. Improving the optimization efficiency of these expensive optimization problems has a great effect on improving our intelligent design ability and intelligent manufacturing ability. Kriging-assisted evolutionary algorithms are important methods for solving the expensive optimization problems. However, as the dimension of the problems increases, the training time of the Kriging model increases rapidly and the efficiency of the Kriging-based infill criteria decreases gradually, which affects the computation efficiency and optimization accuracy of the Kriging-assisted evolutionary algorithms on high-dimensional problems. In order to tackle this problem, the project first introduces the incremental learning method to reduce the computation time of the high-dimensional Kriging model. Then, this project builds a new infill criterion based on combinatorial optimization to improve the optimization efficiency of the Kriging-assisted evolutionary algorithm. Finally, this project applies the proposed Kriging-assisted evolutionary algorithm to a high-dimensional expensive engineering problem to test the adaptability of the algorithm on real-world problems. This project will provide new and efficient methods for solving high-dimensional expensive optimization problems rising up increasingly in engineering.