基于时间卷积网络的机床齿轮箱轴承剩余寿命预测

Remaining Useful Life Prediction of Machine Tool Gearbox Bearings Based on Temporal Convolutional Networks

基于深度神经网络的RUL预测模型结构比较复杂,不能很好地满足中长期预测任务的要求。为了更好地利用时间信息,设计一种基于时间卷积网络(TCN)的轴承RUL预测模型。以振动信号的频谱特征作为输入,利用因果膨胀卷积结构提取频域特征并捕获长期依赖,从而实现对轴承准确的RUL预测。为了进一步说明所提方法的优越性,将所提方法与卷积神经网络(CNN)、门控循环单元(GRU)进行了对比。结果表明:所提出的TCN模型的RUL预测精度优于其他现有方法,具有较高的精度。

The currently available deep neural network-based remaining useful life(RUL) prediction models are complex in structure and do not meet the requirements of medium and long-term prediction tasks well.In order to make better use of temporal information,a temporal convolutional network(TCN) based RUL prediction model for bearings was designed.Taking the spectral features of the vibration signal as input,a causally inflated convolutional structure was used to extract the frequency domain features and capture the long-term dependence to achieve accurate prediction of the bearing RUL.To further illustrate the superiority of the proposed method,comparative experiments were conducted using a convolutional neural network(CNN) and a gated recurrent unit(GRU).The results show that the RUL prediction accuracy of the proposed TCN model outperforms other existing methods with high accuracy.