基于CEEMDAN-SVD和CNN的轴承故障识别
1. 大连交通大学, 计算机与通信工程学院电子与通信工程系, 辽宁大连 116028
2. 大连交通大学, 机械工程学院机械电子系, 辽宁大连 116028
| 摘 要： | 轴承是电机的核心组件，轴承工作时往往要高速旋转，极易产生磨损和疲劳。滚动轴承出现故障会导致设备停机、生产延误和维修成本增加。滚动轴承的故障诊断已成为确保电机系统安全性、效率和可用性的关键。滚动轴承故障诊断通常对其振动信号进行处理、分析做出判断，本文使用完全集合经验模态分解（Complete Ensemble Empirical Mode Decomposition with Adaptive Noise, CEEMAND）、奇异值分解、Renyi熵和卷积神经网络对轴承振动信号进行处理进而进行故障诊断。滚动轴承振动信号经过CEEMDAN分解后得到的本征模态函数（Intrinsic Mode Function, IMF）分量存在无法描述特征信息的虚假分量，所以根据计算原始信号与各IMF分量的相关系数，筛选出较高相关系数的IMF分量，并对筛选出来的较高相关系数的IMF分量进行奇异值分解得到其奇异值，再计算奇异值的Renyi熵以组成故障特征向量。最后将故障特征向量输入到卷积神经网络（Convolutional Neural Network, CNN）进行故障类别的识别。当训练集比例占20%时，别准确率为99.3%。故障信号检测准确率为100%，表明本文方法有很好的故障识别效果。 |
| 关 键 词： | 故障识别; 奇异值分解; 完全集合经验模态分解; 卷积神经网络 |
| DOI： | 10.57237/j.mse.2024.01.002 |
View:
Downloads:
收稿日期: 2023-11-30
接受日期: 2023-12-28
在线出版日期: 2024-01-20
Bearing Fault Identification Based on CEEMDAN-SVD and CNN
1. Department of Electronic and Communication Engineering, School of Computer and Communication Engineering, Dalian Jiaotong University, Dalian 116028, China
2. Department of Mechanical Electronics, School of Mechanical Engineering, Dalian Jiaotong University, Dalian 116028, China
| Abstract: | Bearings are the core components of motors, and they often rotate at high speeds during operation, which can easily cause wear and fatigue. The failure of rolling bearings can lead to equipment downtime, production delays, and increased maintenance costs. The fault diagnosis of rolling bearings has become the key to ensuring the safety, efficiency, and availability of motor systems. Fault diagnosis of rolling bearings usually involves processing and analyzing their vibration signals to make judgments. In this paper, Complete Ensemble Empirical Mode Decomposition (CEEMAND), Singular Value Decomposition, Renyi Entropy, and Convolutional Neural Networks are used to process the bearing vibration signals for fault diagnosis. The intrinsic mode function (IMF) components obtained from the vibration signal of rolling bearings after CEEMDAN decomposition contain false components that cannot describe the characteristic information. Therefore, based on the calculation of the correlation coefficients between the original signal and each IMF component, IMF components with higher correlation coefficients are selected, and singular value decomposition is performed on the selected IMF components with higher correlation coefficients to obtain their singular values, Calculate the Renyi entropy of the singular value again to form the fault feature vector. Finally, the fault feature vector is input into a Convolutional Neural Network (CNN) for fault category recognition. When the training set accounts for 20%, the accuracy rate is 99.3%. The accuracy of fault signal detection is 100%, indicating that the method proposed in this paper has good fault recognition performance. |
| Keywords: | Fault Identification; Singular Value Decomposition; Complete set Empirical Mode Decomposition; Convolutional Neural Network |
基金
| 1. | 大连市重点科技研发计划项目 (2022YF11GX008) |
参考文献
| [1] | Zhiwei C, Xuejiao L, Guangbin P, et al. Transfer Deep Learning Network for Rolling Bearing Fault Diagnosis of Wind Turbines [J]. Journal of Physics: Conference Series, 2023, 2503(1). |
| [2] | 许立环, 王春, 徐翔等. 基于自适应变分模态分解的滚动轴承故障诊断研究 [J]. 工业控制计算机, 2021, 34(12): 86-88+91. |
| [3] | 吴漫, 黄国勇, 周卫兵. 基于CEEMD-SVD和ELM的滚动轴承故障诊断 [J]. 化工自动化及仪表, 2018, 45(10): 784-788. |
| [4] | 曾梦洁, 李舜酩, 陆建涛等. 基于深度神经网络的嵌入式轴承故障智能诊断系统 [J]. 工业控制计算机, 2021, 34(10): 41-43+47. |
| [5] | 李俊卿, 张承志, 胡晓东, 等. 基于CNN-ViT的滚动轴承故障类型识别方法 [J]. 电力科学与工程. 2023, (2): 64-71. |
| [6] | 肖俊青, 金江涛, 李春, 等. 基于CEEMDAN模糊熵CNN轴承故障诊断研究 [J]. 机械强度. 2023, (1): 26-33. |
| [7] | 苏文胜, 王奉涛, 张志新, 等. EMD降噪和谱峭度法在滚动轴承早期故障诊断中的应用 [J]. 振动与冲击. 2010, (3): 18-21. |
| [8] | 杨杰. 声发射技术在超低速轴承故障诊断中的应用研究[D]. 兰州理工大学, 2017. |
| [9] | 李军, 李青. 基于CEEMDAN-排列熵和泄漏积分ESN的中期电力负荷预测研究 [J]. 电机与控制学报. 2015, (8): 70-80. |
| [10] | Dou S. K., Zhang L. Y., Li C. X.. Improved EMD-PSO-LSSVM train wireless network time-delay prediction [P]. Dalian Jiaotong University (China), 2022. |
| [11] | 窦东阳, 李丽娟, 赵英凯. 基于EEMD-Renyi熵和PCA-PNN的滚动轴承故障诊断 [J]. 东南大学学报: 自然科学版. 2011, (S1): 107-111. |
| [12] | 张丽艳, 温万钦. 一种基于DWT-HD-SVD的数字图像水印算法 [J]. 大连交通大学学报, 2022, 43(06): 105-109+115. |
| [13] | 罗洁思, 张绍辉, 李叶妮. 多分辨奇异值分解在滚动轴承振动信号解调分析中的应用 [J]. 振动工程学报. 2019, (6): 1114-1120. |
| [14] | Srikrushna S B, Damodar S S, Rameshpant S G. Renyi entropy and deep learning-based approach for accent classification [J]. Multimedia Tools and Applications, 2021, 81(1). |
| [15] | 刘禛. 基于深度学习的轴承故障诊断研究 [D]. 集美大学, 2022. |
| [16] | Bengio Y, Glorot X. Understanding the difficulty of training deep feed forward neural networks [C]. proceedings of the thirteenth international conference on artificial intelligence and statistics. 2010: 249-256. |
| [17] | 宋立业, 孙琳. EEMD-GSSA-SVM滚动轴承故障诊断方法研究 [J]. 传感器与微系统, 2022, 41(04): 56-59. |
