为了拓宽余热排出泵设计高效区的范围，提出了一种基于Kriging近似模型和遗传算法的优化方法。采用拉丁超立方试验设计方法对叶轮叶片的进口冲角∆β、包角φ及出口安放角β2进行16组方案设计，并采用ANSYS CFX14.5对16组叶轮方案进行定常数值模拟，选取离心泵设计工况1.0Qd和大流量工况1.62Qd下的效率为水力优化设计目标，建立了效率与叶片三个参数之间的Kriging近似模型，并应用多目标遗传算法对近似模型进行寻优，得到了最优的叶片参数。对原始方案进行外特性试验，数值模拟结果与试验结果基本吻合。优化后，叶轮在两工况下的效率均高于原始泵，效率分别提高了5.53%和2.29%。同时对比优化前后的泵内部速度分布，表明在设计工况和大流量工况下，优化后的叶轮内部相对速度分布更均匀，水力损失较小。提出的叶轮优化方法对泵性能提高提供了有效参考。

To enlarge the high efficiency area of the residual heat removal pump (RHRP), an optimization process is proposed based on Kriging model and neighborhood cultivation genetic algorithm (NCGA). 16 design cases containing three main parameters, which are incidence angle ∆β, are angle φ and outlet blade angle β2, are designed by using the Latin hypercube sampling (LHS) methods. The efficiencies are calculated by steady numerical simulation. The efficiencies under the design point 1.0Qd and 1.62Qd are selected as the objective functions. Approximation models between the efficiencies and the parameters are built by Kriging model. The neighborhood cultivation genetic algorithm is applied to optimize the approximation models to obtain the best combination of parameters. The results show that the predicted performance has a good agreement with that of the experiment of the original pump. The efficiencies of the optimal pump are 5.53% and 2.29% higher than those of the original one under 1.0Qd and 1.62Qd, respectively. Meanwhile, the relative velocity distributions are more uniform in the optimal impeller. The optimization method can provide an effective reference to the improvement of pump’s performance.