Fiber Optic Sensor-Based Bearing Defect Detection and Its Usages in Computer Vision

Abstract

This work focuses on the investigation of surface defects in small bearings. Based on the theory of rough surface scattering and the dual-beam ratio measurement method, fiber optic sensing technology is applied in identifying surface defects in bearings. To facilitate the extraction of features for surface defects in bearings, a sensor probe fiber array with three concentric circles around the central emission is determined. A reflective intensity-modulated fiber optic sensor (FOS) is employed to detect surface defects on bearings. The structural parameters of the FOS are simulated through Matlab, considering the inner/outer diameter, numerical aperture, and axial spacing of the sensor. This work involves designing modulation light source excitation circuits, photoelectric conversion module circuits, pre-amplification differential amplifier circuits, infinite gain bandpass filtering circuits, and window function comparison circuits. This effectively amplifies the defect feature signals and eliminates noise interference. In experiments, the sensor probe is fixed on the support of a micro-displacement measurement platform. By adjusting the distance between the probe and the side surface through rotation, initial tests are conducted using standard roughness samples. The results indicate that installing the sensor probe at a distance of 0.92 mm from the side surface provides better measurement of surface roughness. The oscilloscope waveform reveals that the FOS can identify defects on different bearing surfaces. Furthermore, the bearing surface is divided into sections with engraved text (seal cover part) and without engraved text (inner and outer rings of the bearing). Using computer vision (CV) technology, a FOS detection system is designed, achieving a defect recognition rate of 99% for bearings, in line with the intended design goals.