Identification of wheel hollow-worn under complex environment based on mutability and continuity

Accurate diagnosis of wheel hollow-worn is crucial before reprofiling to minimize cutting. However, complex train environments cause maintenance site images with strong noise and uneven illumination, reducing recognition accuracy. This article proposes a diagnosis method based on axial mutability and circumferential continuity. It involves two stages: First, the image is projected by grey value, and the rough location of hollow-worn is determined based on the first- and second-order differential mutation interval. Second, to mitigate noise and illumination interference, a coherence index using Spearman rank correlation coefficient enables secondary verification within the rough area. This two-stage approach detects hollow-worn effectively. In order to verify the effectiveness and robustness of the proposed method, the measured data from the freight car workshop were used for verification, the method achieved a 96.93% detection rate at 0.08s per piece. Results demonstrate its effectiveness, robustness against noise and uneven illumination in complex environments, and suitability for real-time field applications.