航空学报 > 2024, Vol. 45 Issue (13): 629037-629037   doi: 10.7527/S1000-6893.2023.29037

机理⁃数据混合驱动的叶片加工变形预测方法

董立卓1,2, 张思琪1,2, 张钊1,2, 吴宝海1,2()   

  1. 1.西北工业大学 航空发动机高性能制造工信部重点实验室,西安 710072
    2.西北工业大学 航空发动机先进制造技术教育部工程研究中心,西安 710072
  • 收稿日期:2023-05-23 修回日期:2023-06-12 接受日期:2023-08-20 出版日期:2024-07-15 发布日期:2023-09-27
  • 通讯作者: 吴宝海 E-mail:wubaohai@nwpu.edu.cn
  • 基金资助:
    国家重点研发计划(2020YFB1710400);民机专项(MJZ4-2N21);中央高校基本科研业务费专项资金(D5000220135)

Prediction method of blade machining deformation driven by mechanism⁃data hybrid

Lizhuo DONG1,2, Siqi ZHANG1,2, Zhao ZHANG1,2, Baohai WU1,2()   

  1. 1.Key Laboratory of High Performance Manufacturing for Aero Engine,Ministry of Industry and Information Technology,Northwestern Polytechnical University,Xi’an 710072,China
    2.Engineering Research Center of Advanced Manufacturing Technology for Aero Engine,Ministry of Education,Northwestern Polytechnical University,Xi’an 710072,China
  • Received:2023-05-23 Revised:2023-06-12 Accepted:2023-08-20 Online:2024-07-15 Published:2023-09-27
  • Contact: Baohai WU E-mail:wubaohai@nwpu.edu.cn
  • Supported by:
    National Key Research and Development Program(2020YFB1710400);Civil Aircraft Special Project(MJZ4-2N21);the Fundamental Research Funds for the Central Universities(D5000220135)

摘要:

航空发动机叶片作为一种典型的薄壁类零件,在铣削加工过程中易发生弹性变形与加工残余应力变形,影响其加工精度与质量。针对叶片复杂加工过程的变形机理建模问题和依赖于有限采集样本的数据驱动模型预测问题,建立了机理引导的麻雀优化极限学习机(SSA-ELM)薄壁叶片加工变形预测模型。首先,建立了综合考虑弹性变形、加工残余应力变形的薄壁叶片加工变形机理模型。其次,设计极限学习机神经网络(ELMNN),并采用麻雀搜索算法(SSA)优化ELMNN隐含层网络参数。然后,构建了数据驱动模型的数据集,并利用蒙特卡洛模拟对加工变形数据样本进行增强,建立了数据驱动模型。最后,采用机理引导的SSA-ELM预测薄壁叶片加工变形值,以加工变形均方根误差与相关系数作为模型精度评价指标,结果表明机理引导和SSA优化后的加工变形预测模型的均方根误差分别减小了77.25%和30.5%,证明模型有良好的预测性能。

关键词: 薄壁叶片, 弹性变形, 加工残余应力变形, 加工变形预测, 机理模型, 数据驱动模型

Abstract:

As a typical thin-walled part of aviation, blade is prone to elastic deformation and machining residual stress deformation in the process of milling, which affects the machining accuracy and quality of blade. To address the deformation mechanism modeling problem of complex machining process of blade and the data-driven model prediction problem relying on limited collected samples, a mechanism-guided Sparrow optimized Extreme Learning Machine (SSA-ELM) thin-walled blade machining deformation prediction model is established. Firstly, the machining deformation mechanism model of thin-walled blade considering elastic deformation and machining residual stress deformation is established. Secondly, the Extreme Learning Machine Neural Network(ELMNN)is designed, and the Sparrow Search Algorithm (SSA) is used to optimize the parameters of the ELMNN hidden layer network. Then, the dataset of the data-driven model is constructed, and the Monte Carlo simulation is used to enhance the processed deformation data samples to establish the data-driven model. Finally, the mechanism-guided SSA-ELM is used to predict the thin-walled blade machining deformation, and the root mean square error and correlation coefficient of the machining deformation are used as the model accuracy evaluation index. The results show that the root mean square error of the mechanism guided and SSA optimized machining deformation prediction model is reduced by 77.25% and 30.5% respectively, which proves that the model has good prediction performance.

Key words: thin-walled blade, elastic deformation, machining residual stress deformation, machining deformation prediction, mechanism model, data-driven model

中图分类号: