A novel motor fault diagnosis with Chladni-based vibration visualization

Abstract To synchronously realize the requirements of effectively integrating time–frequency domain information and enhancing the richness of fault information in motor vibration signal feature extraction, this paper proposes a novel fault diagnosis method based on Chladni pattern visualization and deep feature matching. Firstly, this method by converting one-dimensional vibration signals into two-dimensional Chladni patterns, which not only preserve critical time–frequency characteristics but also intuitively display distinct resonance patterns corresponding to different fault types. Secondly, SuperPoint and SuperGlue—are employed to identify and match salient image features, establishing a robust mapping relationship between specific fault conditions and Chladni patterns. And experimental results demonstrate that the proposed method achieves a diagnostic accuracy of 98%, significantly outperforming traditional image-matching algorithms. This method offers an efficient and highly applicable solution for motor fault detection and monitoring, particularly in smart industrial settings where accuracy and real-time performance are critical.