A new sorting feature-based temporal convolutional network for remaining useful life prediction of rotating machinery

Abstract A new sorting feature is designed to improve the remaining useful life (RUL) prediction accuracy of rotating machinery. First, the life-cycle signals of rotating machinery are decomposed into several subsignals through local mean decomposition. Second, the amplitude values of each subsignal are sorted from minimum to maximum, and the sorting feature is constructed. The sorting features of all subsignals are then arranged to construct a feature map. Lastly, the feature maps are inputted into a temporal convolutional network to train a prediction model. The well-trained model is used to predict the RULs of other rotating machines. The effectiveness and efficiency of the proposed method are validated using bearing and gearbox datasets in comparison with other conventional techniques. The new sorting feature reflects the intrinsic connections among signal amplitudes and therefore improves prediction accuracy. The proposed method exhibits great potential applications in the RUL prediction of rotating machines.