A Review of Research on Industrial Time Series Classification for Machinery based on Deep Learning
Mohammad Ali Nemer, Joseph Azar, Jacques Demerjian, Abdallah Makhoul, Julien Bourgeois · 2022
This research investigates detecting machine failures in a manufacturing process using multivariate time series data. From a methodological standpoint, fault detection and diagnosis in industrial machines based on supervised deep learning could be divided into convolutional neural network-based and recurrent neural network-based methods. This paper provides a systematic overview of the most recent research developments regarding the two methods. This study uses the raw data from a public Machinery Fault Database to evaluate these models. The results of this study indicate that the Residual Network (ResNet) model adapted for time series provides the best performance. The paper concludes with a summary of the significant challenges from the perspective of industrial applications, followed by a projection of future development directions.