冲击冷却是涡轮冷却中常见的方式，其优化设计涉及多种几何参数，是典型的高维问题。在冲击冷却结构的设计过程中，需要根据涡轮的热负荷情况适应性地设计冷却结构，以提高综合冷却效率和表面温度的均匀性。实验或数值模拟耗时长且成本高，而代理模型可以快速预测结果，配合计算机自动寻优算法可显著提高设计效率和效果。为了降低优化设计的成本、提高优化过程的效率，以平板冲击冷却为研究对象，同时考虑非均匀热负荷的影响，通过数值模拟构建数据集，建立了基于迭代算子神经网络的代理模型，并使用遗传算法对斑状非均〖JP2〗匀热载荷条件下孔位置排布进行了优化。优化结果显示：对于优化潜力较低的结构，优化策略保持了靶板平均温度水平不变；对于优化潜力较高的结构，可以降低靶板平均温度约2.6 K；所研究各结构的表面温度标准差普遍降低70%以上。

Impingement cooling is a prevalent method employed for turbine cooling, which necessitates the optimization of multiple geometric parameters, leading to a highdimensional problem. Efficiently optimizing impingement cooling involves designing adaptive cooling structures to handle turbine thermal loads, enhancing overall cooling effectiveness and ensuring surface temperature uniformity. Traditional experimental or numerical simulation methods are timeconsuming and costly, making surrogate models a valuable tool for fast predictions, facilitating automatic optimization algorithms and significantly improving design efficiency. In order to reduce the cost of optimization design and improve the efficiency of the optimization process, this paper focuses on the study of impingement cooling on plate while taking into account the influence of nonuniform thermal load. This paper utilizes numerical simulation results as the dataset and builds a surrogate model based on an iterative operator neural network. The optimization of hole positions under porphyritic nonuniform thermal loads is performed using a genetic algorithm (GA). The results show that in cases with low optimization space, the optimization strategy maintains the average temperature of the target plate unchanged; while in cases with larger optimization space, the average temperature can be reduced by approximately 2.6 K; the standard deviation of surface temperature is generally reduced by 70% in all cases.

V231.1