A FPGA Deployment System Based on Convolutional Neural Network for Rolling Bearing Diagnosis

Real-time fault diagnosis of rolling bearing is a challenging issue for industry. Although artificial intelligence-based technologies could be well used for fault diagnosis of rolling bearing, the factories may not take into account the deployment of diagnosis algorithms. To tackle the issue, this work proposes a flexible deployment system of diagnosis algorithm for rolling bearing, where the required Convolutional Neural Network (CNN) model is deployed on the Field Programmable Gate Array (FPGA) to identify the working conditions of rolling bearing using vibration signals. The experimental results show that the deployed system performs accurately and efficiently on the test set, while a real-time prediction of FPGA could be guaranteed, indicating its potential as a powerful auxiliary system of rotating machinery.