为了揭示炉水循环泵内部非定常流动特性，并提高其水力性能及运行稳定性，针对运行背景压力为36.6MPa，输送360℃的高温水的炉水循环泵，提出了基于反向传播（Back Propagation，BP）神经网络代理模型及非支配排序遗传算法II（Non-dominated Sorting Genetic Algorithm II，NSGA-II）的炉水循环泵多目标优化方法。采用FLUENT软件对0.6Qd、Qd、1.4Qd流量工况下泵内部流动进行了模拟，并根据试验结果验证了数值模拟的准确性，揭示了炉水循环泵内部非定常流动规律，在此基础上，采用拉丁超立方抽样方法对叶轮结构参数取样，根据抽取样本构建BP神经网络代理模型，以压力脉动系数主频振幅最小、扬程和效率最大为优化目标，结合NSGA-II算法进行结构优化，对比分析了优化前后的扬程、效率及压力脉动系数。研究结果表明：设计流量下内部流场流动最稳定，随着流量增大压力脉动强度呈逐渐增大的趋势，导叶进口附近压力脉动最大；叶轮与导叶内部的压力脉动频谱中，主频由彼此叶片数量决定；优化后模型泵与原模型泵相比，扬程增加了4.15%，效率提升了0.4%，监测点gv1处压力脉动系数主频振幅减小了8.19%，优化效果较好。

In order to reveal the unsteady flow characteristics in the boiler water circulating pump and improve its hydraulic performance and operational stability, a multi-objective optimization method for the boiler water circulating pump was proposed. This method is based on a Back Propagation (BP) neural network surrogate model and the Non-dominated Sorting Genetic Algorithm II (NSGA-II), targeting the boiler water circulating pump operating under the background pressure of 36.6MPa and conveying high-temperature water at 360℃. FLUENT software was employed to simulate internal flow within the pump under the flow rates 0.6Qd, Qd, and 1.4Qd. The accuracy of the numerical simulations was validated against experimental results. The unsteady flow patterns of the boiler water circulating pump were revealed. On this basis, the Latin hypercube sampling method was employed to sample the structural parameters of the impeller. A BP neural network surrogate model was constructed based on the extracted samples. Structural optimization was performed using the NSGA-II algorithm, with the optimization objective of minimizing the main frequency amplitude of pressure pulsation coefficient, maximizing the head and efficiency. A comparative analysis was conducted on the head, efficiency, and pressure pulsation coefficient before and after optimization. The research results show that under the design flow rate the internal flow field has the most stable flow. As the flow rate increases, the intensity of pressure pulsation gradually increases. The pressure pulsation is maximum near the inlet of the guide vane. In the frequency spectrum of pressure pulsation inside the impeller and guide vanes, the main frequency is determined by the number of blades in each other. Compared with the original model pump, the optimized model pump has an increase of 4.15% in head, an increase of 0.4% in efficiency, and a decrease of 8.19% in the main frequency amplitude of the pressure pulsation coefficient at the monitoring point gv1. The optimization effect is quite good.

TH311

江苏省市场监督管理局科技计划项目（KJ2024026）；国家自然科学基金项目（51306087）；江苏省高等学校自然科学研究重大项目（A类）（17KJA480003）