基于Kriging代理模型的约束差分进化算法

广泛应用的高精度分析模型使得飞行器设计优化的计算成本不断增加，为了缩短优化耗时，基于代理模型的进化算法（SAEAs）近年来得到了广泛关注。针对现有SAEAs处理约束优化问题优化效率低下的缺陷，提出了一种基于Kriging代理模型的约束差分进化算法（KRG-CDE），结合约束改善概率与最优适应度定制了一种改进的可行准则，从而提高新增样本点的潜在可行性与最优性，并根据种群改善情况，平衡算法全局探索与局部搜索性能。标准测试算例对比研究结果表明，相比于基于全局与局部代理模型的差分进化算法、（μ+λ）-约束差分进化算法，KRG-CDE算法在优化效率、全局收敛性及鲁棒性等方面具有显著优势。最后，运用KRG-CDE算法求解全电推卫星多学科设计优化问题，验证了该算法的工程实用性。

Kriging-assisted constrained differential evolution algorithm

High-fidelity analysis models have been widely used in modern design, significantly increasing the computational budget of engineering design optimization. To reduce the computational cost, researchers have paid extensive attention to Surrogate Assisted Evolutionary Algorithms (SAEAs) recently. A Kriging-assisted Constrained Differential Evolution algorithm (KRG-CDE) is developed in this study to improve the efficiency of SAEAs in solving constrained optimization problems. Based on constraint improvement probability and optimality fitness, an improved feasibility rule is tailored to enhance the potential optimality and feasibility of the infill sample points. Moreover, the global exploration and local exploitation capacity of the KRG-CDE are balanced according to the population improvement. The proposed method is tested on several standard benchmark problems and compared with global and local surrogate-assisted differential evolution and (μ+λ)-constrained differential evolution to verify its optimization performance. The comparison results illustrate that the KRG-CDE outperforms the competitors in terms of efficiency, convergence, and robustness. Finally, the KRG-CDE is successfully applied to an all-electric propulsion satellite multidisciplinary design optimization problem, demonstrating the practicality and effectiveness of the proposed KRG-CDE in engineering practices.