Data Acquisition and Fusion System Using 1-D CNN Network for Real-Time Bearing Fault Detection

A classification system that utilizes the convolutional neural network (CNN) is established for real-time bearing fault detection. The system consists of acquiring the vibration signals using a tri-axial CMCP1300ST-T accelerometer, designing the interface circuit for the data acquisition board, fusing three-channel data information to extract a principal signal, and implementing machine learning algorithms in the LabVIEW platform. The designed system is validated via real-time fault detection for the following four bearing conditions: normal condition, and inner race, outer race, and ball defects.