基于RBMO优化SVMD-CNN-BILSTM的滚动轴承故障诊断  

Fault Diagnosis of Rolling Bearings Based on RBMO Optimization SVMD-CNN-BILSTM

作　　者：丁伟 赵飞龙 Ding Wei;Zhao Feilong(School of Automotive and Intelligent Transportation,Jiangsu Vocational College of Information Technology,Wuxi,Jiangsu 214153,China;XCMG Excavation Machinery Co.,Ltd.,Xuzhou,Jiangsu 221000,China)

机构地区：[1]江苏信息职业技术学院汽车与智能交通学院,江苏无锡214153 [2]徐州徐工挖掘机械有限公司,江苏徐州221000

出　　处：《黑龙江工业学院学报(综合版)》2025年第1期99-106,共8页Journal of Heilongjiang University of Technology(Comprehensive Edition)

基　　金：江苏省高等学校基础科学(自然科学)研究面上项目(项目编号:23KJD430006);江苏高校“青蓝工程”资助项目(项目编号:苏教师函〔2022〕51)。

摘　　要：为解决滚动轴承故障类型难以提取和识别的问题,提出一种基于红嘴蓝鹊优化算法(RBMO)优化连续变模态分解(SVMD)联合卷积神经网络(CNN)提取故障特征,并与双向长短期记忆(BiLSTM)相融合的滚动轴承故障诊断模型。首先,采用RBMO算法优化SVMD中平衡参数;其次,根据得到的优化参数对故障信号进行SVMD分解,通过包络熵值最小准则提取出最佳的信号分量,然后利用卷积神经网络(CNN)提取信号的时域和频域特征,并将特征进行融合;最后,利用双向长短期记忆网络(BiLSTM)学习了故障的序列模式,完成故障训练和故障识别。实验结果表明:基于RBMO优化SVMD-CNN-BILSTM模型在滚动轴承故障识别方面表现出明显的优势,平均识别准确率可达99.33%,与CNN-LSTM、CNN-BiLSTM、SVMD-BiLSTM模型进行了对比,该方法对滚动轴承故障诊断的识别准确率高、计算速度快,具有良好的应用价值。To solve the problems of difficulty in extracting and identifying fault types in rolling bearings,a rolling bearing fault diagnosis model is proposed based on the Red Billed Blue Magpie Optimization Algorithm(RBMO)optimized Continuous Variable Mode Decomposition(SVMD)combined with Convolutional Neural Network(CNN)to extract fault features,and integrated with Bidirectional Long Short Term Memory(BiLSTM).Firstly,the RBMO algorithm is used to optimize the balance parameters in SVMD;secondly,based on the optimized parameters obtained,the fault signal is subjected to SVMD decomposition,and the optimal signal component is extracted using the minimum envelope entropy criterion.Then,a convolutional neural network(CNN)is used to extract the time-domain and frequency-domain features of the signal,and the features are fused;finally,a bidirectional long short-term memory network(BiLSTM)is used to learn the sequence patterns of faults,completing fault training and recognition.The experimental results show that the RBMO optimized SVMD-CNN-BILSTM model exhibits significant advantages in rolling bearing fault recognition,with an average recognition accuracy of 99.33%.Compared with CNN-LSTM,CNN-BiLSTM,and SVMD-BiLSTM models,this method has high recognition accuracy and fast calculation speed for rolling bearing fault diagnosis,and has good application value.

关 键 词：轴承故障诊断 红嘴蓝鹊优化算法 连续变模态分解 卷积神经网络 双向长短期记忆 

分 类 号：TH133.3[机械工程—机械制造及自动化] TP306.3[自动化与计算机技术—计算机系统结构]