Automatic Wear Detection on Normalized Gears Made by Additive Manufacturing from Dense 3D Point Clouds

Abstract

A low-cost method based on macro-photogrammetric reconstruction is presented to automatically detect wear and other defects in small gears created with additive manufacturing. This novel approach is oriented to preventive and predictive maintenance of gears in order to avoid faults in machines and devices. The experimentation has been conducted using three defective gears produced in Nylon PA-12. First, a robotic platform and a systematic macro-photogrammetric data acquisition procedure were used to accomplish the 3D reconstruction and generate the dense point clouds. Subsequently, a comparison between the dense point cloud and the ideal solid CAD model of the normalized gear has been carried out. For this aim, the models have been alignment in the same spatial system. The computation of the distances between solid models and point clouds allows the automatic visualization of different types of defects even for defects that are not visible to the naked eye. This conclusion has been checked from a statistical point of view considering the discrepancies obtained in the comparison and their distribution.