Remaining Useful Life(RUL)prediction of machinery is an important part of system maintenance strate-gy.In the process of establishing the target function of the deep learning prediction approach,the RUL degradation model is usually established as a piece wise linear function.The influence of outliers on the predicted results is eas-ily amplified.This paper proposes a temporal convolutional network regression model with piece wise nonlinear deg-radation.The nonlinear function can better depict the sensor degradation trend and reduce the systematic deviation caused by linear model prediction.The framework is validated in NASA's Company-Modular Aero-Propulsion Sys-tem Simulation data sets(C-MAPSS)dataset,experiments show that this model has lower error than the model whose target function is piecewise linear function,and is better than some existing prediction methods.
关键词
剩余使用寿命预测/深度学习/非线性目标函数/时间卷积网络
Key words
remaining useful life prediction/deep learning/nonlinear target function/temporal convolutional network
维普
万方数据