基于一维噪声增强卷积神经网络的轴承剩余使用寿命预测  

Remaining Useful Life Prediction of Bearings Based on One-Dimensional Noise Enhanced Convolutional Neural Network

作　　者：丁伟 陈律 王骁贤 宋俊材 陆思良[1] DING Wei;CHEN Lyu;WANG Xiaoxian;SONG Juncai;LU Siliang(School of Electrical Engineering and Automation,Anhui University,Hefei 230601,China;School of Electronic and Information Engineering,Anhui University,Hefei 230601,China)

机构地区：[1]安徽大学电气工程与自动化学院,合肥230601 [2]安徽大学.电子信息工程学院,合肥230601

出　　处：《轴承》2025年第5期71-78,共8页Bearing

基　　金：国家自然科学基金资助项目(52375522)。

摘　　要：基于一维卷积神经网络,采用无条件的噪声注入方法提高网络模型的训练速度和预测精度,将得到的一维噪声增强卷积神经网络模型(1DNECNN)用于轴承剩余使用寿命预测,避免了复杂的数据预处理过程,可从原始振动信号中直接分辨出轴承的退化程度。在IEEE PHM Challenge 2012轴承数据集上的对比试验表明,与无噪声注入的一维卷积神经网络、二维卷积神经网络和卷积注意力神经网络相比,1DNECNN预测结果的均方误差降低了24%~49%,具有更高的预测精度和更优的拟合性能。Based on one-dimensional convolutional neural network,an unconditional noise injection method is employed to enhance the training speed and prediction accuracy of network model.The resulting one-dimensional noise enhanced convolutional neural network(1DNECNN)model is applied for prediction of bearing remaining useful life,circumventing the complex data preprocessing procedures and enabling direct discernment of bearing degradation state from raw vibration signals.The comparative experiments conducted on IEEE PHM Challenge 2012 bearing dataset demonstrate that,in comparison to one-dimensional convolutional neural network without noise injection,two-dimensional convolutional neural network and convolutional neural network with attention mechanism,the 1DNECNN model achieves a reduction in mean squared error by 24%to 49%,indicating superior prediction accuracy and enhanced fitting performance.

关 键 词：滚动轴承 剩余寿命 预测 神经网络 卷积 数据预处理 

分 类 号：TH133.33[机械工程—机械制造及自动化] TM307.1[电气工程—电机]