Looseness of belt-pulley which are employed in industrial applications such as compressor and fan is important factor for safe machine operation. In conventional method, current/vibration sensors are attached for regularly inspecting the belt looseness. In contrast, this paper proposes a novel remote sensing method for machine health monitoring, utilizing a biaxial optical system. The bi-axial optical system consists of two groups of reflectors to accommodate stereoscopic images with disparity so that it enables 3D motion measurement with only one image sensor. 3D trajectory of belt can be reconstructed using bi-axial optical system and one image sensor. To improve the 3D accuracy of trajectory, camera calibration algorithms is improved by setting virtual left and right camera parameters to be identical considering the left and right images captured by single image sensor. As a result, 3D accuracy was more improved by 15.75% compared to Zhang’s method. For the actual experiment, four looseness level (normal, slight, moderate, severe damaged) of the belt were monitored. As the looseness level increased, the 3D trajectory of the belt exhibited a curved shape with noticeable quantitative differences. The result shows that the proposed architecture can be utilized to remote machine health monitoring with mobile image sensors.
|