Method for Evaluating Surface Defects on Cast Parts Using Principal Component Analysis for 3D Point Cloud

Abstract

This paper proposes a method for evaluating surface defects on cast parts. A laser profilometer measures the shape of parts as point cloud data and principal component analysis estimates the surface variation. The idea to detect defects is that they usually have a high surface variation value. Experiments were conducted, applying our method to simulated defects and cast parts. The simulated defects at the estimated surface variation at the defect is characteristically larger than the flat surface. For the cast parts, a defect on its cast surface was evaluated with SN ratio of 5.9. This method is robust to the noise caused by the surface roughness or shape non-uniformity of cast parts.