基于GADF-MDSC的特大型轴承深度迁移故障诊断方法

PDF(3411 KB)

振动与冲击 ›› 2024, Vol. 43 ›› Issue (19) : 10-18.

论文

基于GADF-MDSC的特大型轴承深度迁移故障诊断方法

姜烨飞1，王华1，潘裕斌1，王天祥1，傅航2

作者信息 +

Deep transfer fault diagnosis method for extra-large bearings based on GADF-MDSC

JIANG Yefei1, WANG Hua1, PAN Yubin1, WANG Tianxiang1, FU Hang2

Author information +

文章历史 +

摘要

针对工程应用中特大型轴承运行工况复杂以及故障数据匮乏，导致其故障特征提取不全面的问题，提出了一种基于格拉姆角差场与多尺度深度可分离卷积(Gramian angular difference field and multi-scale depthwise separable convolutions，简称GADF-MDSC)的特大型轴承深度迁移智能诊断方法。首先，构建GADF-MDSC故障诊断网络，该网络分为三大模块：图像转换、特征提取、输出部分。图像转换模块采用格拉姆角差场编码方式将振动信号转换为二维图像；特征提取模块通过多尺度深度可分离卷积提取综合故障特征信息，并利用双向门控循环单元筛选融合特征；输出部分由Softmax函数预测轴承故障类型的概率分布。然后，利用源域数据预训练模型，将预训练模型权重参数作为目标域训练模型初始化参数，冻结除底层外的所有参数，使用目标域数据微调模型，实现深度迁移故障诊断任务。最后，通过两种特大型轴承试验对深度迁移模型进行验证。试验结果表明，所提方法在目标域样本仅有5%的条件下，仍能保证较高的跨工况精度，达到86.04%，且迁移效果优于其他方法。

Abstract

Aiming at the problem of incomplete fault feature extraction due to complex operating conditions and lack of fault data of oversized bearing in engineering applications, an oversized bearing intelligent diagnosis method of deep migration based on Gramian angular difference field and multi-scale depthwise separable convolutions (GADF-MDSC) was proposed. Firstly, the GADF-MDSC fault diagnosis network is constructed, which is divided into three modules: image conversion, feature extraction and output. In the image conversion module, the vibration signal is converted into two dimensional image by using the GADF. The feature extraction module extracts comprehensive fault feature information by MDSC, and filters fusion features by bidirectional gated cycle unit. The output part predicts the probability distribution of bearing failure types by Softmax function. Then, the source domain data is used to pre-train the model, the weight parameters of the pre-trained model are used as the initial parameters of the target domain training model, all parameters except the bottom layer are frozen, and the target domain data is used to fine-tune the model to achieve the deep migration fault diagnosis task. Finally, the deep migration model is verified by two kinds of oversized bearing tests. The experimental results show that the proposed method can still guarantee high cross-working accuracy of 86.04% when the target domain sample is only 5%, and the migration effect is better than other methods.

导出引用

姜烨飞1, 王华1, 潘裕斌1, 王天祥1, 傅航2. 基于GADF-MDSC的特大型轴承深度迁移故障诊断方法[J]. 振动与冲击, 2024, 43(19): 10-18

JIANG Yefei1, WANG Hua1, PAN Yubin1, WANG Tianxiang1, FU Hang2. Deep transfer fault diagnosis method for extra-large bearings based on GADF-MDSC[J]. Journal of Vibration and Shock, 2024, 43(19): 10-18

参考文献

[1] 李云飞,苏文胜. 基于时序灰度图和分组ResNet的回转支承故障诊断[J]. 机床与液压,2022,50(12):187-191.
LI Yunfei, SU Wensheng. Fault diagnosis of slewing bearing based on grayscale time series and group ResNet[J]. Machine Tool & Hydraulics,2022,50(12):187-191.
[2] 曾耀传,林云树,吴晓梅. 基于EEMD与GWO-MCKD的门座起重机回转支承故障诊断[J]. 机床与液压,2022,50(07):170-175.
ZENG Yaochuan, LIN Yunshu, WU Xiaomei. Fault diagnosis of gantry crane slewing bearing based on EEMD and GWO-MCKD[J]. Machine Tools & Hydraulics,2022,50(07):170-175.
[3] 吴晓梅. 门座起重机回转支承振动信号分析及故障诊断[J]. 起重运输机械,2021,(12):30-35.
WU Xiaomei. Vibration signal analysis and fault diagnosis of gantry crane slewing bearing[J]. Lifting and Transporting Machinery,2021,(12):30-35.
[4] Yin Wenzhe,Xia Hong,Wang Zhichao,Yang Bo,Zhang Jiyu,Jiang Yingying,Miyombo Miyombo Ernest. A fault diagnosis of nuclear power plant rotating machinery based on multi-sensor and deep residual neural network[J]. Annals of Nuclear Energy,2023,185.
[5] 段晓燕, 焦孟萱, 雷春丽, 李建华. 基于MTF-MSMCNN的小样本滚动轴承故障诊断方法[J]. 航空动力学报, 1-14.
DUAN Xiaoyan, JIAO Mengxuan, LEI Chunli, LI Jianhua. Small sample rolling bearing fault diagnosis method based on MTF-MSMCNN[J]. Journal of Aerodynamics, 1-14.
[6] Xu Tao, Lv Huan, Lin Shoujin, Tan Haihui, Zhang Qing. A fault diagnosis method based on improved parallel convolutional neural network for rolling bearing[J]. Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering, 2023, 237 (12): 2759-2771.
[7] Zhang WeiTao, Liu Lu, Cui Dan, Ma YuYing, Huang Ju. An Anti-Noise Convolutional Neural Network for Bearing Fault Diagnosis Based on Multi-Channel Data[J]. Sensors (Basel, Switzerland), 2023, 23 (15):
[8] Zhao Shubiao,Li Xuejun,Chen Jinhua,Jiang Lingli. Fault diagnosis of rolling bearing based on cross-domain divergence alignment and intra-domain distribution alienation[J]. Journal of Vibroengineering,2023,25(6).
[9] Xu Huoyao,Liu Jie,Peng Xiangyu,Wang Junlang,He Chaoming. Deep dynamic adaptation network: a deep transfer learning framework for rolling bearing fault diagnosis under variable working conditions[J]. Journal of the Brazilian Society of Mechanical Sciences and Engineering,2022,45(1).
[10] 刘飞,陈仁文,邢凯玲,丁汕汕,张迈一. 基于迁移学习与深度残差网络的滚动轴承快速故障诊断算法[J]. 振动与冲击,2022,41(03):154-164.
LIU Fei, CHEN Renwen, XING Kailing, DING Shanshan, ZHANG Maiyi. Fast fault diagnosis algorithm of rolling bearing based on transfer learning and deep residual network[J]. Journal of Vibration and Shock, 2002,41(03):154-164.
[11] Huo Chunran,Jiang Quansheng,Shen Yehu,Zhu Qixin,Zhang Qingkui. Enhanced transfer learning method for rolling bearing fault diagnosis based on linear superposition network[J]. Engineering Applications of Artificial Intelligence,2023,121.
[12] Wang Jiujian,Yang Shaopu,Liu Yongqiang,Wen Guilin. Deep Subdomain Transfer Learning with Spatial Attention ConvLSTM Network for Fault Diagnosis of Wheelset Bearing in High-Speed Trains[J]. Machines,2023,11(2).
[13] 王琦,邓林峰,赵荣珍. 基于一维CNN迁移学习的滚动轴承故障诊断[J]. 振动.测试与诊断,2023,43(01):24-30+195.
WANG Qi, DENG Linfeng, ZHAO Rongzhen. Fault diagnosis of rolling bearing based on one-dimensional CNN transfer learning[J]. Vibration. Test and Diagnosis, 2019,43(01):24-30+195.
[14] Li Gang, Ao Jiaxing, Hu Jiayao, Hu Dalong, Liu Yang, Huang Zhiqiang. Dual-source gramian angular field method and its application on fault diagnosis of drilling pump fluid end[J]. Expert Systems With Applications, 2024, 237 (PB):
[15] 郭盼盼, 张文斌, 崔奔, 郭兆伟, 赵春林, 尹治棚, 刘标. 基于增强深度卷积神经网络的滚动轴承多工况故障诊断方法[J]. 振动工程学报, 1-14.
GUO Panpan, ZHANG Wenbin, CUI Ben, GUO Zhaowei, ZHAO Chunlin, YIN Zhi-Peng, LIU Biao. Multi-condition fault diagnosis method of rolling bearing based on enhanced deep convolutional neural network[J]. Journal of Vibration Engineering, 1-14.
[16] 张辉, 戈宝军, 韩斌, 赵丽娜. 基于GAF-CapsNet的电机轴承故障诊断方法[J]. 电工技术学报, 2023, 38 (10): 2675-2685.
ZHANG Hui, GE Baojun, HAN Bin, ZHAO Lina. Fault diagnosis method of motor bearing based on GAF-CapsNet[J]. Transactions of China Electrotechnical Society, 2023, 38 (10): 2675-2685.
[17] 丁頔, 南国防. CNN-RNN融合法在旋转机械故障诊断中的应用[J]. 轻工学报, 2020, 35 (01): 102-108.
DIN Di, NAN Guofang. Application of CNN-RNN fusion method in fault diagnosis of rotating machinery[J]. Journal of Light Industry, 2020, 35 (01): 102-108.
[18] R. Dey, F.M. Salemt, Gate-variants of Gated Recurrent Unit (GRU) neural networks, in: Midwest Symp. Circuits Syst., 2017: pp. 1597–1600. https://doi.org/10.1109/ MWSCAS.2017.8053243.
[19] 焦正杉. 基于LSTM的疲劳因子隐性异常检测及故障诊断方法研究[D]. 河北工业大学, 2020.
JIAO Zhengshan. Research on hidden anomaly detection and fault diagnosis of fatigue factor based on LSTM[D]. Hebei University of Technology, 2020.
[20] Li Xueyi, Yuan Peng, Wang Xiangkai, Li Daiyou, Xie Zhijie, Kong Xiangwei. An unsupervised transfer learning bearing fault diagnosis method based on depthwise separable convolution[J]. Measurement Science and Technology, 2023, 34 (9):