CNC Machine-Bearing Fault Detection Based on Convolutional Neural Network Using Vibration and Acoustic Signal

Purpose To detect bearing faults in CNC machine tools, this study proposes an intelligent vibration-based fault diagnosis approach. Flexible manufacturing systems (FMS) make extensive use of computer numerical control (CNC) machine tools. Bearings are one of the essential components of a CNC machine tool, and their failure is one of the most prevalent reasons for machine failure. Bearing problems must be detected and diagnosed for rotating machinery to work properly. Methods Experimental vibration data for different bearings and operational needs were studied to develop a structure for monitoring and classifying bearing problems to assess the machine's health. This paper presents a bearing fault diagnosis method based on a convolutional neural network that can diagnose CNC machine faults early. The STFT technique converts raw signals such as vibration and acoustic signals into time-frequency analysis. Results Extensive experiments suggest that the proposed method provides 100% classification accuracy on vibration and acoustics signals for CNC machine-bearing fault detection. The proposed model outperformed the other classical diagnostic algorithms on acquired datasets for different bearing faults. Conclusion The presented CNN technique has been validated on different datasets. Findings show that the CNN-based approach on vibration and acoustics has a classification accuracy of 100%, exceeding ANN and classic machine learning methods.