Intelligent fault diagnosis of hoist bearing based on feature transfer learning
-
摘要: 针对提升机复杂实际工况导致的现有故障诊断方法准确率低和适应性弱的问题，提出了一种基于深度迁移特征选取（DTF）与平衡分布自适应（BDA）的提升机轴承智能故障诊断方法。对不同工况下的轴承故障信号进行时频分析，提取时域、频域统计特征，采用深度置信网络进行高维深度特征提取。为从高维深度特征集中选取出既有利于故障模式识别，也有利于跨域故障诊断的特征，采用基于ReliefF与域间差异的迁移特征选取(TFRD)方法对各特征的可迁移性进行量化评估，利用TFRD方法对各特征进行类别区分度和域不变性量化评估，采用ReliefF算法处理各类特征数据，获得表征类别区分度的权重值；计算同一特征在不同域间的最大均值差异，构建一种新的特征可迁移性量化指标。基于TFRD 方法，选取特征可迁移性大的深度特征构建特征子集，利用BDA对源域和目标域的特征子集进行分布适应，降低两者间的分布差异。采用源域特征集训练故障模式识别分类器，对目标域样本进行故障识别与分类。采用经典机器学习方法、深度学习方法和迁移学习方法构建了8种故障诊断模型，用于与提出的DTF−BDA故障诊断模型进行故障诊断准确率对比。结果表明：① DTF−BDA故障诊断模型能够取得明显优于其他对比模型的性能，最高故障诊断准确率可达100%。② TFRD方法能有效提高基于迁移学习方法构建的故障诊断模型的性能，与迁移成分分析和联合分布自适应相结合情况下的最高故障诊断准确率分别可达96.46%和97.67%。Abstract: The complex actual working conditions of the hoist causes the problems of low accuracy and weak adaptability of existing fault diagnosis methods. In order to solve these problems, an intelligent fault diagnosis method of hoist bearing based on deep transferable feature selection(DTF) and balance distribution adaptation(BDA) is proposed. The bearing fault signals under different working conditions are subjected to time-frequency analysis. The time and frequency domain's statistical characteristics are extracted. The high-dimensional depth characteristics are extracted by adopting a deep belief network. In order to select features that are beneficial to fault mode identification and cross-domain fault diagnosis from a high-dimensional depth feature set, the transferable feature selection based on ReliefF and differences between domains(TFRD) method is adopted. The method carries out the quantitative evaluation of the transitivity of each feature. The TFRD method carries out the quantitative evaluation on the class discrimination and domain invariance of each feature. The ReliefF algorithm processes various feature data to obtain weight values representing class discrimination. This method calculates the maximum mean discrepancy of the same feature between different domains, and constructs a new quantitative index of feature transferability. Based on the TFRD method, depth features with high feature transferability are selected to construct feature subsets. The balance distribution adaptation is applied to carry out distribution adaptation on the feature subsets of the source domain and the target domain, so as to reduce the distribution difference between the two domains. The source domain feature set is used to train the fault pattern identification classifier, and the target domain samples are used for fault identification and classification. Eight fault diagnosis models are constructed by using the classical machine learning method, deep learning method and transfer learning method. The models are used for comparing the fault diagnosis accuracy with the proposed DTF-BDA fault diagnosis model. The results show the following points. ① The DTF-BDA fault diagnosis model can achieve better performance than other models, and the highest fault diagnosis accuracy can reach 100%. ② The TFRD method can effectively improve the performance of the fault diagnosis model based on the transfer learning method. The highest fault diagnosis accuracy can reach 96.46% and 97.67% respectively when combined with the transfer component analysis and joint distribution adaptation.
-
Keywords:
- mine hoist /
- bearing fault /
- fault diagnosis /
- transfer learning /
- deep features /
- balance distribution adaptation
-
【编者按】机械设备是矿山生产运行的基础，其结构复杂，工况环境多变。随着智慧矿山的提出，煤矿对于矿山机械设备的安全性、可靠性及经济性等方面提出了更高的要求，加强矿山机械设备的状态监测与故障诊断成为矿山机械设备安全、高效和稳定运行的基础条件。为进一步总结、交流我国矿山机械设备健康状态监测与故障诊断技术最新进展，《工矿自动化》特邀安徽理工大学郭永存教授担任专题客座主编，胡坤、姜阔胜和马天兵教授担任客座副主编，于2022年第9期策划出版“矿山机械设备健康状态监测与故障诊断技术及应用”专题。在专题刊出之际，衷心感谢各位专家学者的大力支持！
-
![]()
图 1 基于 DTF−BDA 的提升机轴承智能故障诊断流程
Figure 1. Flow of hoist bearing intelligent fault diagnosis based on deep transferable feature selection and balance distribution adaptation
![]()
图 2 基于3层RBM构建的DBN
Figure 2. Deep belief network based on three layer restricted Boltzmann machine
表 1 凯斯西储大学轴承故障数据集
Table 1 Bearing fault dataset of Case Western Reserve University
轴承状态 缺陷尺
寸/cm不同轴承工况下的样本数 类别标签 域1 域2 域3 域4 正常状态 0 60 60 60 60 1 滚动体缺陷 0.017 78 60 60 60 60 2 0.035 56 60 60 60 60 3 0.053 34 60 60 60 60 4 0.071 12 60 60 60 60 5 内圈缺陷 0.017 78 60 60 60 60 6 0.035 56 60 60 60 60 7 0.053 34 60 60 60 60 8 0.071 12 60 60 60 60 9 外圈缺陷 0.017 78 60 60 60 60 10 0.035 56 60 60 60 60 11 0.053 34 60 60 60 60 12 
下载: 导出CSV
表 2 不同故障诊断模型在4个任务下的故障诊断准确率对比
Table 2 Comparison of fault diagnosis accuracy of different fault diagnosis models under 4 fault diagnosis tasks
% 故障模型 故障诊断准确率 任务1 任务2 任务3 任务4 FS−SVM 95.00 73.13 87.50 78.96 FS−KNN 96.88 82.50 90.13 85.00 FS−DBN−Softmax 85.21 85.63 82.50 80.13 FS−DAE−Softmax 59.17 53.96 53.33 51.67 FS−TCA−SVM 77.50 78.75 72.67 76.13 FS−JDA−SVM 83.33 81.67 79.17 77.50 FS−TFRD−TCA 96.88 96.67 95.42 95.00 FS−TFRD−JDA 98.13 98.96 97.71 97.08 DTF−BDA 100.00 100.00 100.00 100.00
下载: 导出CSV
表 3 不同故障诊断模型实验结果
Table 3 Experimental results of different fault diagnosis models
可迁移特征
选取数故障诊断准确率/% FS−TFRD−TCA FS−TFRD−JDA DTF−BDA 任务1 任务2 任务3 任务4 任务1 任务2 任务3 任务4 任务1 任务2 任务3 任务4 20 66.04 67.08 60.67 63.17 68.67 65.83 61.67 60.50 71.33 69.83 70.33 68.17 40 71.67 73.13 66.83 68.50 74.50 73.17 72.50 71.33 77.50 76.67 75.00 74.83 60 79.79 81.88 76.00 78.67 80.00 81.67 80.33 79.29 83.33 82.67 82.67 82.50 80 83.54 86.04 82.67 84.50 89.83 87.00 87.50 88.13 91.67 92.00 92.50 93.13 100 88.96 88.33 84.83 86.50 95.17 96.83 94.83 94.50 99.50 99.17 99.67 99.50 120 95.00 95.63 91.33 89.17 97.67 96.00 97.50 96.33 98.33 98.17 99.13 98.67 140 96.46 96.00 96.33 95.50 95.67 95.17 94.83 94.33 96.67 96.50 97.67 96.46 160 89.38 92.67 86.50 84.83 88.00 86.83 89.33 87.00 94.83 93.33 95.50 94.83 180 80.83 82.50 76.33 82.67 84.33 82.50 86.17 84.67 92.17 91.50 93.13 92.67 200 77.50 78.75 72.67 76.13 83.33 81.67 79.17 77.50 88.75 86.46 83.75 82.08
下载: 导出CSV
-
[1] 李娟莉,闫方元,梁思羽,等. 基于卷积神经网络的矿井提升机制动系统故障诊断方法[J]. 太原理工大学学报,2022,53(3):524-530. LI Juanli,YAN Fangyuan,LIANG Siyu,et al. Fault diagnosis method of mine hoist braking system based on convolutional neural network[J]. Journal of Taiyuan University of Technology,2022,53(3):524-530.
[2] 康守强,刘旺辉,王玉静,等. 基于深度在线迁移的变负载下滚动轴承故障诊断方法[J]. 控制与决策,2022,37(6):1521-1530. KANG Shouqiang,LIU Wanghui,WANG Yujing,et al. Fault diagnosis method of rolling bearing under varying loads based on deep online transfer[J]. Control and Decision,2022,37(6):1521-1530.
[3] 丁恩杰,俞啸,廖玉波,等. 基于物联网的矿山机械设备状态智能感知与诊断[J]. 煤炭学报,2020,45(6):2308-2319. DOI: 10.13225/j.cnki.jccs.zn20.0340 DING Enjie,YU Xiao,LIAO Yubo,et al. Key technology of mine equipment state perception and online diagnosis under Internet of things[J]. Journal of China Coal Society,2020,45(6):2308-2319. DOI: 10.13225/j.cnki.jccs.zn20.0340
[4] 张梅,许桃,孙辉煌,等. 基于模糊故障树和贝叶斯网络的矿井提升机故障诊断[J]. 工矿自动化,2020,46(11):1-5,45. DOI: 10.13272/j.issn.1671-251x.17562 ZHANG Mei,XU Tao,SUN Huihuang,et al. Fault diagnosis of mine hoist based on fuzzy fault tree and Bayesian network[J]. Industry and Mine Automation,2020,46(11):1-5,45. DOI: 10.13272/j.issn.1671-251x.17562
[5] 王保勤. 基于一维卷积神经网络的提升机轴承故障诊断方法研究[J]. 矿山机械,2021,49(9):29-34. DOI: 10.3969/j.issn.1001-3954.2021.09.007 WANG Baoqin. Research on fault diagnosis method for hoist bearing based on one-dimensional convolutional neural network[J]. Mining & Processing Equipment,2021,49(9):29-34. DOI: 10.3969/j.issn.1001-3954.2021.09.007
[6] 刘旭,朱宗玖,杨明亮. 基于小波包与隐马尔可夫的矿井提升机主轴故障诊断[J]. 煤炭技术,2022,41(1):214-216. LIU Xu,ZHU Zongjiu,YANG Mingliang. Fault diagnosis of mine hoist spindle based on wavelet packet and HMM[J]. Coal Technology,2022,41(1):214-216.
[7] 马辉,车迪,牛强,等. 基于深度神经网络的提升机轴承故障诊断研究[J]. 计算机工程与应用,2019,55(16):123-129,184. DOI: 10.3778/j.issn.1002-8331.1903-0010 MA Hui,CHE Di,NIU Qiang,et al. Research on fault diagnosis of hoisting bearing based on deep neural network[J]. Computer Engineering and Applications,2019,55(16):123-129,184. DOI: 10.3778/j.issn.1002-8331.1903-0010
[8] 俞啸,范春旸,董飞,等. 基于EMD与深度信念网络的滚动轴承故障特征分析与诊断方法[J]. 机械传动,2018,42(6):157-163. DOI: 10.16578/j.issn.1004.2539.2018.06.033 YU Xiao,FAN Chunyang,DONG Fei,et al. Fault feature analysis and diagnosis method of rolling bearing based on empirical mode decomposition and deep belief network[J]. Journal of Mechanical Transmission,2018,42(6):157-163. DOI: 10.16578/j.issn.1004.2539.2018.06.033
[9] 廖玉波,俞啸,李伟生,等. 基于深度置信网络的旋转机械迁移故障诊断[J]. 机电工程,2022,39(2):193-201. DOI: 10.3969/j.issn.1001-4551.2022.02.008 LIAO Yubo,YU Xiao,LI Weisheng,et al. Transfer fault diagnosis for rotating machinery based on deep belief network[J]. Journal of Mechanical & Electrical Engineering,2022,39(2):193-201. DOI: 10.3969/j.issn.1001-4551.2022.02.008
[10] 黄定洪. 基于深度置信网络的旋转机械设备故障诊断方法研究[D]. 武汉: 武汉理工大学, 2020. HUANG Dinghong. Research on fault diagnosis method of rotating machinery equipment based on deep belief networks[D]. Wuhan: Wuhan University of Technology, 2020.
[11] 丁思凡,王锋,魏巍. 一种基于标签相关度的Relief 特征选择算法[J]. 计算机科学,2021,48(4):91-96. DOI: 10.11896/jsjkx.200800025 DING Sifan,WANG Feng,WEI Wei. Relief feature selection algorithm based on label correlation[J]. Computer Science,2021,48(4):91-96. DOI: 10.11896/jsjkx.200800025
[12] 董飞. 基于数据驱动的滚动轴承故障诊断研究[D]. 徐州: 中国矿业大学, 2018. DONG Fei. Research on rolling bearing fault diagnosis based on data-driven[D]. Xuzhou: China University of Mining and Technology, 2018.
[13] WANG Jindong, CHEN Yiqiang, HAO Shuji, et al. Balanced distribution adaptation for transfer learning[EB/OL]. [2022-05-16]. https://ieeexplore.ieee.org/document/82156132017.
[14] 范春旸. 基于多源信息融合的井下皮带机驱动电机状态识别方法研究[D]. 徐州: 中国矿业大学, 2019. FAN Chunyang. Research on state recognition method of driving motor of mine belt conveyor based on multi-source information fusion[D]. Xuzhou: China University of Mining and Technology, 2019.
[15] YU Xiao,DONG Fei,DING Enjie,et al. Rolling bearing fault diagnosis using modified LFDA and EMD with sensitive feature selection[J]. IEEE Access,2018,6:3715-3730. DOI: 10.1109/ACCESS.2017.2773460
计量
- 文章访问数: 368
- HTML全文浏览量: 160
- PDF下载量: 95
