基于自适应代理优化算法的射雾器叶轮优化设计

Optimal design of the impeller of fog sprayer based on self-adapting surrogate optimal algorithm

为了提高射雾器的射程,同时提高Kriging代理模型的优化效率和精度,提出一种新的自适应代理优化算法,包括全局探索阶段和局部探索阶段。在全局探索阶段,提出改进的最大期望概率提高准则(Improved Probability Improvement,IPI)和并行加点策略;在局部探索阶段,通过最小响应面准则(Minimizing Prediction,MP)获取新样本点;各加点准则均采用差分进化算法进行寻优,并根据新样本点和已知样本点的关系来实现全局探索和局部探索的自适应切换,直至找到最优解。算例实验表明:和最大期望提高准则(Expected Improvement,EI)、多点期望提高准则(q-Expected Improvement,q-EI)、最大期望概率提高准则(Probability Improvement,PI)相比,优化算法的加点次数至少减少了6.32%,优化效率提高了11.61%以上,求解问题的有效性更好。采用优化算法对射雾器叶轮结构进行优化,射雾器射程提高了24.77%。

In order to improve the range of the fog sprayer and at the same time to improve the optimization efficiency and accuracy of Kriging surrogate model,a new self-adapting surrogate optimal algorithm is proposed,which includes global exploration stage and local exploration stage.In the global exploration stage,Improved Probability Improvement(IPI)filling criterion and parallel filling strategy are proposed.In the local exploration stage,a new sample point is obtained through Minimizing Prediction(MP)filling criteria.Each filling criterion is optimized by differential evolution algorithm.According to the relationship between the new sample points and the known sample points,the self-adapting switch between global exploration and local exploration is realized until the optimal solution is found.Numerical example experiments show that:Compared with the EI,q-EI,PI,the filling number by the optimization algorithm is reduced by 6.32%or more,and the optimization efficiency is increased by more than 11.61%,and the effectiveness of solving the problem is better.The impeller structure of fog sprayer is optimized by the optimization algorithm,and the range of the fog sprayer is increased by 24.77%.