Real-time detection of multi-wedge belt tooth surface defects based on 3D laser technology

Abstract Defects in multi-wedge belts are a key factor in triggering transmission failure and efficiency reduction of mechanical equipment, and defect detection plays a key role in the quality monitoring process of multi-wedge belts. It has been proved that the three-dimensional defect detection method is better than the traditional two-dimensional method and manual detection method in terms of accuracy and comprehensiveness. Therefore, this paper proposes a real-time detection method of multi-wedge belt tooth surface defects based on 3D laser technology. Firstly, the acquired point cloud is background filtered to extract the tooth face point cloud. In order to solve the problem of uneven distribution of the point cloud at the tooth inflection points, a 3D edge detection algorithm based on asymmetry evaluation is adopted to extract the features of the tooth apex surface and uniformly remove the outliers; taking into account the geometric differences between different tooth apex surfaces of multi-wedge belts, a DBSCAN clustering algorithm based on the grouping approach to accelerate the neighbourhood search is introduced to speed up the clustering analysis process and effectively distinguish between different tooth apex surfaces of multi-wedge belts. Finally, for the meandering tendency of the tooth top surface point cloud, the adaptive moving window technique is combined with multi-dimensional statistical analysis to detect the defects in the clustered point cloud of the tooth top surface. The experimental results show that the overall detection rate of the proposed method for various defect types is more than 96%, and the average detection time is 1.68 s. This performance meets the requirement of efficient detection in actual production, and provides a new technical way for defect detection of multi-wedge belt.