A data-driven model for milling tool remaining useful life prediction with convolutional and stacked LSTM network

Abstract

This paper introduces a hybrid model that incorporates a convolutional neural network (CNN) with a stacked bi-directional and uni-directional LSTM (SBULSTM) network, named CNN-SBULSTM, to address sequence data in the task of tool remaining useful life (RUL) prediction. In the CNN-SBULSTM network, CNN is firstly utilized for local feature extraction and dimension reduction. Then SBULSTM network is designed to denoise and encode the temporal information. Finally, multiple fully connected layers are built on the top of the CNN-SBULSTM network to add non-linearity to the output, and one regression layer is utilized to generate the target RUL. The cyber-physical system (CPS) is used to collect the internal controller signals and the external sensor signals during milling process. The proposed hybrid model and several other published methods are applied to the datasets acquired from milling experiments. The comparison and analysis results indicate that the integrated framework is applicable to track the tool wear evolution and predict its RUL with the average prediction accuracy reaching up to 90%.