基于PVC-CAE的轴承剩余寿命预测方法  

Remaining life prediction method of bearing based on PVC-CAE

作　　者：张远亮[1] 李海浪 Zhang Yuanliang;Li Hailang(College of Intelligent Manufacturing&Transportation,Urban Vocational College of Sichuan,Chengdu 610110,China;School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]四川城市职业学院智能制造与交通学院,成都610110 [2]西南交通大学机械工程学院,成都610031

出　　处：《计算机应用研究》2023年第10期3070-3075,共6页Application Research of Computers

摘　　要：轴承由于在制造、安装以及工况上的不同,退化过程有很大差异,在轴承的剩余使用寿命预测中,特征的轴承个体差异会直接影响到后续模型的训练难度和预测精度。为了在提取特征时减小轴承的个体差异性,提出了一种并行方差约束卷积自编码(parallel variance constrained convolutional auto-encode,PVC-CAE)的轴承退化特征提取模型。具体方法是在卷积自编码的框架内定义并行方差约束,通过改进卷积自编码的损失函数,搭建出PVC-CAE模型。PVC-CAE模型可以有效地降低同标签特征的轴承个体差异性,提升预测精度。预测流程为:先用PVC-CAE模型在频域信号内提取特征,再用LSTM网络预测。通过PRONOSTIA实验平台所获取的实验数据集以及西安交大轴承数据集对所提方法进行了验证,同时与另外三种方法进行对比,实验结果表明,所提方法在轴承剩余使用寿命预测中取得了较好的结果,并且在不同的工况下具有一定的泛化性。Due to the different manufacturing,installation and working conditions of bearings,the degradation process is very different.In the prediction of the remaining service life of bearings,individual bearing differences of characteristics will directly affect the training difficulty and prediction accuracy of subsequent models.In order to reduce the individual differences of bea-rings during feature extraction,this paper proposed a PVC-CAE model for bearing degradation feature extraction.The specific method defined the parallel variance constraint within the framework of convolution self-coding,and built the PVC-CAE model by improving the loss function of convolution self-coding.The PVC-CAE model could effectively reduce the individual bearing differences of the same tag characteristics and improved the prediction accuracy.The prediction process is as follows:firstly,the PVC-CAE model was used to extract the features in the frequency domain signal,and then the LSTM network was used to predict.The proposed method was verified by the experimental data set obtained from PRONOSTIA test platform and the bearing data set of Xi’an Jiaotong University,and compared with the other three methods.The experimental results show that the me-thod achieves good results in the prediction of the remaining service life of bearings,and has certain generalization under diffe-rent working conditions.

关 键 词：寿命预测 轴承 特征提取 个体差异性 方差约束 

分 类 号：TH17[机械工程—机械制造及自动化]