基于PIRD-CNN的航空发动机轴承故障诊断方法研究  

Fault diagnosis method for aeroengine bearings based on PIRD-CNN

作　　者：张搏文 庞新宇 程宝安 李峰[3] 宿绅正 ZHANG Bowen;PANG Xinyu;CHENG Baoan;LI Feng;SU Shenzheng(State Owned Sida Machinery Manufacturing Company,Xianyang 610400,China;School of Mechanical and Transportation Engineering,Taiyuan University of Technology,Taiyuan 030024,China;School of Aeronautics and Astronautics,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]国营四达机械制造公司,陕西咸阳610400 [2]太原理工大学机械与运载工程学院,太原030024 [3]太原理工大学航空航天学院,太原030024

出　　处：《振动与冲击》2024年第18期201-207,231,共8页Journal of Vibration and Shock

基　　金：山西省重点研发项目(202102010101009,202102010101006);山西省基础研究计划项目(20210302124204)。

摘　　要：航空发动机结构与系统的复杂性导致轴承的故障诊断方法通常面临特征提取与模式识别的困难。针对以上不足,考虑实际工程诊断的实时性与准确性,提出了一种新的基于转子位移概率密度信息(probability density information of rotor displacement,PIRD)的航空发动机轴承智能故障诊断方法。其主要对一维卷积神经网络(1-dimensional convolutional neural network,1DCNN)模型进行改进,在传统的卷积层前面增加了PIRD的提取层,可以提取转子振动位移信号的概率密度信息,有效地降低了数据的冗余度,同时保留了故障监测的重要指标。提出的PIRD-CNN诊断模型保留了1DCNN端到端的故障诊断优势,将该模型在航空发动机试验台产生的轴承故障数据进行测试,其对轴承故障诊断精度可达96.58%,与基准研究相对比表明,PIRD-CNN能够快速且更加精准地诊断航空发动机轴承的故障。The complexity of aircraft engine structures and systems often leads to difficulties in feature extraction and pattern recognition in bearing fault diagnosis methods.In response to the above shortcomings and considering the real-time performance and accuracy of actual engineering diagnosis,a new intelligent fault diagnosis method for aviation engine bearings baseds on probability density information of rotor displacemen(PIRD)was proposed.It mainly improved the 1-dimensional convolutional neural network(1DCNN)model by adding an PIRD extraction layer in front of the traditional convolutional layer,which can extract the probability density information of the rotor vibration displacement signal,effectively reducing data redundancy,while retaining the important indicators in fault monitoring.The proposed PIRD-CNN diagnostic model retains the end-to-end fault diagnosis advantages of 1DCNN.The model was tested by using the bearing fault data generated on an aviation engine test bench,and its accuracy in bearing fault diagnosis reaches 96.58%.Compared with the benchmark research,PIRD-CNN can quickly and more accurately diagnose aviation engine bearing faults.

关 键 词：航空发动机 轴承 转子位移概率密度信息(PIRD) 卷积神经网络 故障诊断 

分 类 号：TH212[机械工程—机械制造及自动化]