基于Kriging的燃油离心泵智能优化设计方法研究

Intelligent Optimization Design Method of an Aero-Fuel Centrifugal Pump Based on Kriging-Model

关键词:  燃油离心泵  参数化仿真  代理模型  智能优化  流体仿真

Key words:  Aero-fuel centrifugal pump  Parametric simulation  Surrogate model  Intelligent optimization  Fluid simulation

摘要:

为了快速化、智能化地实现燃油离心泵设计，提出一种改善期望准则下的Kriging智能优化方法。对离心泵的性能参数、轴面投影控制参数等进行分析，确定优化变量及优化目标。利用Python脚本调用实现对离心泵一维设计、三维建模、内流场仿真及智能优化设计等联合参数化仿真，完成叶轮型线实时修改及效率全局优化。其中，重点给出Kriging建模及遗传算法优化加点相结合的智能优化方法。算例验证表明：相比多项式响应面及径向基函数，所提出的优化方法对常规数学算例具有更好的效果；同时，仿真预测的优化泵效率为82.52%，与优化方法理论计算的效率（82.56%）高度一致，表明该方法能够实现离心泵的优化设计。此外，对比了优化前后离心泵的性能，在相同流量工况下优化泵的扬程和效率均高于原型泵，小流量工况的效率提高幅度较小，设计流量工况最大，提高了2.65%。且优化泵内部流动更为均匀，不利流动得到了一定改善。

Abstract:

In order to realize the design of fuel centrifugal pump quickly and intelligently， a Kriging intelligent optimization method under expected improvement criterion is proposed. The performance parameters and axial projection control parameters of centrifugal pump are analyzed to determine the optimization variables and objectives. The Python script is used to realize the joint parametric simulation combined one-dimensional design， three-dimensional modeling， internal flow field simulation and intelligent optimization design of centrifugal pump， to complete the real-time modification of impeller profile and global optimization of efficiency. Among them， the intelligent optimization method combining Kriging modeling and genetic algorithm optimization is given. Cases show that compared with polynomial response surface and radial basis function， the proposed optimization method has better effect on conventional mathematical examples. Meanwhile， the optimized pump efficiency predicted by simulation is 82.52%， which is highly consistent with that calculated by the optimization method （82.56%）， indicating that this method can realize the optimal design of centrifugal pump. In addition， the performance of the centrifugal pump before and after optimization is compared. Under the same flow-rate condition， the head and efficiency of the optimized pump are higher than those of the prototype pump. The efficiency improvement under small flow-rate condition is small， and one under the design flow condition is the largest， which increased by 2.65%. The flow inside the optimized pump is more uniform， and the adverse flow had been improved to a certain extent.