基于机器学习XGBoost算法的跨音速离心压气机扩压器气动优化设计  

Optimization Design for the Diffuser Vanes of a Transonic CentrifugalCompressor Based on the XGBoost Algorithm

作　　者：李庆阔 张子卿 张英杰 韩戈[1] 张燕峰[1,2] 卢新根 LI Qing-Kuo;ZHANG Zi-Qing;ZHANG Ying-Jie;HAN Ge;ZHANG Yan-Feng;LU Xin-Gen(Key Laboratory of Light-duty Gas-turbine,Institute of Engineering Thermophysics,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院工程热物理研究所/轻型动力重点实验室,北京100190 [2]中国科学院大学,北京100049

出　　处：《工程热物理学报》2021年第8期1970-1978,共9页Journal of Engineering Thermophysics

基　　金：国家科技重大项目(No.2017-Ⅱ-0002-0014);国家自然科学基金(No.51706223)

摘　　要：本文以某跨音速离心压气机扩压器为研究对象,提出了一种叶型参数化方法和优化流程,采用机器学习XGBoost算法以及全局寻优模型对其进行了气动优化。研究结果表明,该回归优化模型可以在较少的迭代步数内获得全局最优解,优化速度比传统代理模型具有显著的提升,优化精度与CFD仿真结果相当。优化后的扩压器能够在保证离心压气机压比基本不变的条件下,峰值效率提高1.03%,稳定裕度提高4.79%,对优化后扩压器内部流动进行了分析,讨论了不同设计参数对离心压气机的影响,为离心压气机扩压器设计提供了指导。A transonic centrifugal compressor with vaned diffusers is numerically studied in this paper,and the parametric method and the optimized procedure for the vaned diffusers are proposed,which were optimized by XGboost and Global optimization algorithms.The result of this study show that the comparing with the other surrogate models,this method can obtain the global minima with less iterations and its accuracy is comparable with that of the CFD results.Optimizations with the method showed that for the same diffuser inlet and outlet radius,the pressure ratio of the compressor almost kept unchanged,but the peak stage efficiency was increased by 1.03%,and the surge margin was increased by 4.79%.Furthermore,the channel flow of the optimized diffuser is analyzed,and the influence of different design parameters on the centrifugal compressor is discussed,which provides guidance and reference for the design and optimization of the similar transonic centrifugal compressor with vaned diffusers.

关 键 词：离心压气机 扩压器 叶片参数化设计 XGBoost 机器学习 

分 类 号：TK05[动力工程及工程热物理]