Remaining useful life prediction of induction motor based on hybrid temporal convolution neural network

To ensure safety across various industries including manufacturing, construction, defense, science, and medical care, the hea lth of induction motors is critical. However, existing research on motor health monitoring is not comprehensive. Moreover, commonly used deep learning techniques such as convolutional neural networks (CNNs) and recurrent neural networks face challenges in making accurate long-term predictions. To overcome these limitations, this study proposes a hybrid temporal convolutional neural network (HTCN) aimed at predicting the remaining useful life (RUL) of induction motors. The HTCN architecture consists of two parallel branches: one integrates a temporal convolutional network (TCN) with a self-attention mechanism, while the other integrates a 1D CNN module enhanced with a convolutional block attention module (CBAM). The TCN branch utilizes different features from multi-sensory signals, while the raw signals are directly fed into the 1D CNN branch with CBAM. The extracted features from both branches are concatenated to predict RUL. This integrative approach enriches the model's understanding of the input data and aims to improve the accuracy of RUL predictions. The effectiveness of the proposed model is validated using datasets derived from an accelerated aging experiment on an electric motor.