基于POD和多层感知器的跨声速压气机叶型精细优化设计方法  

Precise Optimization Method of Transonic Compressor Airfoil Based on POD and Multilayer Perceptron

作　　者：茅晓晨[1] 焦英辰 陈璇[1] 李民 高丽敏[1] MAO Xiaochen;JIAO Yingchen;CHEN Xuan;LI Min;GAO Limin(School of Power and Energy,Northwestern Polytechnical University,Xi’an shaanxi 710072,China;The 91729th Unit of PLA,lingshui Hainan 571800,China)

机构地区：[1]西北工业大学动力与能源学院,陕西西安710072 [2]91729部队,海南陵水571800

出　　处：《海军航空大学学报》2024年第6期715-725,共11页Journal of Naval Aviation University

基　　金：国家自然科学基金(52106057,92152301);航空科学基金(2022Z060053001);气动院流动显示与测量重点实验室基金(D5110220177)。

摘　　要：压气机叶型的精细化设计发展趋势对优化设计方法提出了更高要求,以性能为导向的传统优化设计方法难以满足未来高性能压气机的精细化设计需求。文章提出一种结合本征正交分解(POD)和多层感知器(MLP)的代理模型,它能够快速准确获得叶型流场结构。该方法首先利用拉丁超立方采样法获得样本叶型,并通过CFD计算得到叶型表面等熵马赫数分布,经POD处理后作为样本数据对MLP进行训练。结合POD-MLP和改进人工蜂群算法构建了完整的优化系统,对符合表面等熵马赫数分布约束的叶型进行寻优。基于该方法,对某跨声速叶型在设计点进行了优化。结果表明,该方法可以兼顾非设计点条件,且优化叶型的表面等熵马赫数分布符合所给约束。优化叶型的非负攻角性能明显提升,设计点下总压损失系数降低14.1%,可用攻角范围拓宽4.5%。The development trend of precise design for compressor airfoil requires higher demand for optimization methods,and the performance-oriented optimization method is difficult to achieve the precise design requirements for future high-performance compressors.Therefore,a surrogate model combines the proper orthogonal decomposition(POD)and multilayer perceptron(MLP)is proposed,which can obtain the flow characteristics of airfoils rapidly.The method first uses the Latin hypercube sampling to obtain the sample airfoils,and calculates the surface isentropic Mach number distribution through CFD,which is used as sample data for training MLP after POD processed.After that,a complete optimization system is constructed by combining POD-MLP and revised artificial bee colony algorithm.The final step is to use the system to select the optimal airfoil that comply with the surface isentropic Mach number distribution constraints.Based on the method,a certain transonic compressor airfoil at the design operating condition is optimized.The optimization result shows the method can take into account the off-design operating conditions,and the surface isentropic Mach number distribution of optimized airfoil reaches the given constraints.The optimal airfoil performances at zero and positive incidence angles effectively improve,and the total pressure loss coefficient drops by 14.1%while the available incidence angle range widens by 4.5%.

关 键 词：跨声速压气机叶型 代理模型 本征正交分解 多层感知器 精细优化设计 

分 类 号：V231.1[航空宇航科学与技术—航空宇航推进理论与工程]