Application of Multimodal Deep Learning Method in Rolling Bearing Fault Diagnosis
-
摘要: 滚动轴承在实际运行中负载多变且噪声干扰较大, 导致故障特征提取及诊断困难, 针对此问题本研究提出一种用于机械设备故障诊断的深度学习方法(MF-CNN), 该方法将多模态融合技术(MFT)与卷积神经网络(CNN)结合, 用卷积神经网络对一种工况下的滚动轴承故障数据分别提取时域、频域两个模态特征并融合, 将融合后的特征作为故障分类的依据来构建整个网络, 对变工况下的未知故障类型的数据进行测试, 实现时域、频域双模态对轴承故障类型的联合诊断。大量实验结果表明, 在变载荷和噪声下, MF-CNN模型用于故障诊断的准确率相对传统单模态的时域CNN和频域CNN均有提高, 对由重载荷向轻载荷变化的工况下准确率提升更为明显。Abstract: In the actual operation of rolling bearing, the load is changeable and the noise interference is large, which often lead to the difficulty of fault feature extraction and fault diagnosis. To solve this problem, this study proposed a deep learning method (MF-CNN) for mechanical equipment fault diagnosis, which combines multi-modal fusion technology (MFT) with convolutional neural network (CNN). The convolutional neural network is used to extract and fuse the time domain and frequency domain modal features from the fault data of rolling bearing under one working condition. The fused features are used as the basis of fault classification to construct the whole network. The data of unknown fault types under other working conditions are tested to realize time domain and frequency domain Dual domain fault diagnosis of the bearing type. A large number of experimental results show that under variable load and noise, the accuracy of MF-CNN model for fault diagnosis is better than that of traditional single-mode CNN in time domain and frequency domain, and the accuracy rate of MF-CNN model is more obvious especially when the load condition changes from heavy load to light load.
-
Key words:
- multimodal deep learning /
- fault diagnosis /
- rolling bearing /
- variable load /
- noise
-
表 1 实验数据集组成
数据集 训练负载/ HP 测试负载/ HP 训练样本个数 测试样本个数 D0/123 0 1、2、3 4 000 12 000 D1/023 1 0、2、3 4 000 12 000 D2/013 2 0、1、3 4 000 12 000 D3/012 3 0、1、2 4 000 12 000 表 2 D3/012实验准确率
信噪比/ dB 准确率/% 时域CNN 频域CNN MF-CNN 无噪声 91.4 88.9 93.4 25 74.8 76.4 90.1 24 67.0 72.9 87.3 23 65.5 73.9 90.5 22 68.0 72.0 90.0 21 59.8 67.7 86.6 20 55.8 63.4 85.0 19 51.0 62.6 81.0 18 55.6 60.7 83.7 -
[1] GAN M, WANG C, ZHU C A. Construction of hierarchical diagnosis network based on deep learning and its application in the fault pattern recognition of rolling element bearings[J]. Mechanical Systems and Signal Processing, 2016, 72-73: 92-104 doi: 10.1016/j.ymssp.2015.11.014 [2] 徐乐, 邢邦圣, 郎超男, 等. LMD能量熵和SVM相结合的滚动轴承故障诊断[J]. 机械科学与技术, 2017, 36(6): 915-918 doi: 10.13433/j.cnki.1003-8728.2017.0615XU L, XING B S, LANG C N, et al. Fault diagnosis of rolling bearing combined LMD energy entropy and SVM[J]. Mechanical Science and Technology for Aerospace Engineering, 2017, 36(6): 915-918 (in Chinese) doi: 10.13433/j.cnki.1003-8728.2017.0615 [3] 汪瑾. 滚动轴承变工况状态评估的特征融合技术研究[D]. 南京: 南京航空航天大学, 2016: 1-7WANG J. Feature fusion research on variable operating condition assessment of rolling bearing[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2016: 1-7 (in Chinese) [4] BHATTACHARYYA S, SNASEL V, HASSANIEN A E, et al. Deep learning: research and applications[M]. de Gruyter, 2020 [5] 雷亚国, 贾峰, 孔德同, 等. 大数据下机械智能故障诊断的机遇与挑战[J]. 机械工程学报, 2018, 54(5): 94-104 https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201805011.htmLEI Y G, JIA F, KONG D T, et al. Opportunities and challenges of machinery intelligent fault diagnosis in big data era[J]. Journal of Mechanical Engineering, 2018, 54(5): 94-104 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JXXB201805011.htm [6] 赵小强, 张青青. 改进Alexnet的滚动轴承变工况故障诊断方法[J]. 振动测试与诊断, 2020, 40(3): 472-480, 623 https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCS202003008.htmZHAO X Q, ZHANG Q Q. Improved Alexnet based fault diagnosis method for rolling bearing under variable conditions[J]. Journal of Vibration, Measurement & Diagnosis, 2020, 40(3): 472-480, 623 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCS202003008.htm [7] 姜战伟. 变分模态分解方法及其在机械故障诊断中的应用研究[D]. 马鞍山: 安徽工业大学, 2017JIANG Z W. Variational modal decomposition method and its application in mechanical fault diagnosis[D]. Maanshan: Anhui University of Technology, 2017 (in Chinese) [8] 李录平, 邹新元, 陈荐, 等. 汽轮发电机组碰磨故障的典型特征研究[J]. 振动测试与诊断, 2001, 21(4): 281-285, 296 https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCS200104009.htmLI L P, ZOU X Y, CHEN J, et al. Typical features of rubbing fault of turbo-generator unit[J]. Journal of Vibration, Measurement & Diagnosis, 2001, 21(4): 281-285, 296 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCS200104009.htm [9] CHEN Z Q, LI C, SANCHEZ R V. Gearbox fault identification and classification with convolutional neural networks[J]. Shock and Vibration, 2015, 2015: 390134 [10] GUO L, GAO H L, HUANG H F, et al. Multifeatures fusion and nonlinear dimension reduction for intelligent bearing condition monitoring[J]. Shock and Vibration, 2016: 2016: 4632562 [11] 赵柄锡, 冀大伟, 袁奇, 等. 于采用时域与时频域联合特征空间的转子系统碰磨故障诊断时与分分[J]. 西安交通大学学报, 2020, 54(1): 75-84ZHAO B X, JI D W, YUAN Q, et al. Rubbing fault diagnosis of rotor system based on combined feature space in time and time-frequency domains[J]. Journal of Xi'an Jiaotong University, 2020, 54(1): 75-84 (in Chinese) [12] THIRUKOVALLURU R, DIXIT S, SEVAKULA R K, et al. Generating feature sets for fault diagnosis using denoising stacked auto-encoder[C]//Proceedings of the IEEE International Conference on Prognostics and Health Management (ICPHM). Ottawa: IEEE, 2016: 1-7 [13] 何俊, 张彩庆, 李小珍, 等. 面向深度学习的多模态融合技术研究综述[J]. 计算机工程, 2020, 46(5): 1-11 https://www.cnki.com.cn/Article/CJFDTOTAL-JSJC202005001.htmHE J, ZHANG C Q, LI X Z, et al. Survey of research on multimodal fusion technology for deep learning[J]. Computer Engineering, 2020, 46(5): 1-11 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJC202005001.htm [14] Case Western Reserve University. Bearing data center[EB/OL]. [2018-02-25]. https://engineering.case.edu/bearingdatacenter [15] 那晓栋. 基于深度学习的变负载下滚动轴承故障诊断方法研究[D]. 哈尔滨: 哈尔滨理工大学, 2018NA X D. Research on fault diagnosis method of rolling bearings under variable load based on deep learning[D]. Harbin: Harbin University of Science and Technology, 2018 (in Chinese)