摘要
由于剩余使用寿命预测对于机械正常运转和避免意外事故的发生起到至关重要的作用,如何提高剩余使用寿命预测的鲁棒性已经成为一个广受关注的问题。然而,现如今的方法很少考虑到实际工业环境中的噪声干扰及轴承不同的退化模式对于剩余使用寿命预测的影响。为解决上述问题,该文提出基于堆叠降噪自编码器和多尺度一维卷积神经网络的轴承剩余使用寿命预测模型。首先,利用堆叠降噪自编码器来去除噪声对于剩余使用寿命预测的影响;然后,利用多尺度一维卷积神经网络提取不同工况下更为全面的退化特征;最后,使用多个尺度得到的深度特征进行回归预测。该方法在PHM2012挑战赛的数据集上进行实验验证,实验结果表明该方法能够有效地预测轴承的剩余使用寿命,其均方根误差和绝对平均误差分别达到0.063 1和0.027 8,优于其他方法。
Because the prediction of remaining useful life plays a vital role in the normal operation of machinery and avoiding accidents, how to improve the accuracy of remaining service life prediction has become a more and more popular problem. However, today’s methods rarely consider the noise interference in the actual industrial environment and the impact of different degradation modes on the prediction of remaining useful life. In order to solve the above problems, a bearing remaining service life prediction model based on stacked denoising autoencoder and multi-scale one-dimensional convolutional neural network(SDAE-MS1DCNN) is proposed in this paper. First of all, the stack noise reduction self-encoder is used to remove the influence of noise on the remaining service life prediction;then, the multi-scale one-dimensional convolution neural network is used to extract more comprehensive degradation features under different operating conditions;finally,the depth features obtained from multiple scales are used for regression prediction. The method is verified by experiments on the data set of the PHM2012 Challenge. The experimental results show that the method can effectively predict the remaining service life of the bearing, and its RMSE and MAE are 0.063 1 and 0.027 8 respectively, which is better than other methods.
出处
《科技创新与应用》
2022年第28期21-26,共6页
Technology Innovation and Application
基金
合肥工业大学国家级大学生创新项目(202110359052)
合肥工业大学省级大学生创新项目(S202110359172)。
关键词
剩余使用寿命预测
堆叠降噪自编码器
一维卷积神经网络
多尺度学习
轴承
remaining service life prediction
stacked denoising autoencoder
one-dimensional convolutional neural network
multi-scale learning
bearing
