Aiming at the problem of inaccurate failure judgment of the rolling contact fatigue detection method currently adopted

an on-line detection system based on machine vision is researched and fabricated. Obtain the surface fatigue point of the roller sample and its area quantitative statistics to realize the accurate judgment of fatigue failure. The detection accuracy of the system is determined by comparing the area of the fatigue point obtained by manually marking the edge of the fatigue point with the metallographic microscope and the area of the fatigue point detected by the system. The comparison results show that the maximum relative detection error for a single fatigue point is 5. 56 % and the average error is 3. 21 %. The system complete a test within 1. 97 s to meet the actual detection requirements

and achieve quantitative detection of the fatigue point of the roller sample surface and accurate judgment of the failure state.