基于多代理模型的离心叶轮高效优化设计方法  

Efficient Optimization Design Method of Centrifugal Impellers Based on Multi-surrogate Model

作　　者：刘基盛 计良 李威[1] 贾志新[1] 成金鑫 方鹏程 LIU Jisheng;JI Liang;LI Wei;JIA Zhixin;CHENG Jinxin;FANG Pengcheng(School of Mechanical Engineering,University of Science and Technology Beijing,Beijing,100083;Beijing Shiny Tech.Co.,Ltd.,Beijing,100074;Department of Industrial Engineering,Tsinghua University,Beijing,100084)

机构地区：[1]北京科技大学机械工程学院,北京100083 [2]北京航天新立科技有限公司,北京100074 [3]清华大学工业工程系,北京100084

出　　处：《中国机械工程》2023年第8期899-907,共9页China Mechanical Engineering

基　　金：国家自然科学基金(51775036);广东省基础与应用基础研究基金(2022A1515110055)。

摘　　要：针对自动优化设计时迭代过多、耗时过长等问题,建立了基于多代理模型优化算法的离心压气机叶片高效寻优模型,研究了全局/局部模型管理策略和样本填充方法,对比分析了多代理模型与常用的元启发式优化算法(粒子群算法、多岛遗传算法等)的综合性能,验证了多代理寻优模型的高效性。离心叶轮优化结果表明:等熵效率提高了0.73%,总压比增大了0.18%,喘振裕度提高了1.1%;与经典粒子群算法相比,优化时间缩短54.9%。Aiming at the problems of excessive number of iterations and time consumption in automatic optimization design,an efficient optimization search model for centrifugal compressor blades was established based on multi-surrogate model optimization algorithm.A global/local model management strategy and a sample filling method were studied.The comprehensive performance of multi-surrogate model and common meta-heuristic optimization algorithms(particle swarm algorithm,multi-island genetic algorithm,etc.)were compared and analyzed,and the effectiveness of the efficient optimization search model was verified.The results of the centrifugal impeller optimization show that the isentropic efficiency is improved by 0.73%,the total pressure ratio is increased by 0.18%,and the surge margin is improved by 1.1%.Compared with the classical particle swarm algorithm,the optimization time may be reduced by 54.9%.

关 键 词：离心叶轮 叶片几何外形 高效寻优 优化算法 多代理模型 

分 类 号：TH122[机械工程—机械设计及理论]