基于1-DCNN的滚动轴承退化预测研究

Prediction for Rolling Bearing Performance Degradation Based on 1-DCNN

传统的滚动轴承退化特征提取方法高度依赖于预先研究和专业知识,对于学习退化特征与大量测量数据之间的复杂关系的能力有限,很难构建一个单一的指标预测轴承的退化状态。针对这一问题,提出了基于一维卷积神经网络(1-DCNN)的轴承退化预测模型,以原始振动信号作为输入,构建健康指标。以PHM 2012轴承全寿命数据对原始振动信号、频谱信号、3种模态分解预处理后的信号等5种处理方法进行测试。实验结果表明,相对于其他几种处理方法,以原始振动信号直接作为所提模型的输入,提取出的健康指标能更好地反映轴承的退化状态。

Prediction for rolling bearing performance degradation is complex.Traditional methods are highly dependent on prior research and expertise,and have limited ability to learn the complex relationships between degraded features and large amounts of measured data.Therefore,it is difficult to construct a single indicator to predict the degradation state of the bearing.In order to solve this problem,a bearing prediction model based on convolutional neural network was proposed,and the original vibration signal was used as input to establish health indicators.Five kinds of processing methods,such as original vibration signal,spectrum signal and three modal decomposition pre-processed signals,were tested with PHM 2012 bearing life data.The experimental results show that the original vibration signal is directly used as the input of the 1-DCNN model,and the extracted health indicators can better reflect the degradation state of the bearing compared with other methods.