A Data Clustering Approach for Blister Detection in Coordinate Metrology of Casting Products

Abstract

Blisters are among the common defects found on the casting parts due to trapped gas porosities under high pressure near the surface of the casting. Existence of blisters on some exterior surfaces of the products or on the critical surfaces that are used as a datum or interface for the downstream assembly processes can be considered as major defects that need to be detected by inspection. Digital coordinate metrology is used as a fast and automated approach in the mass-production lines to control the geometric and dimensional accuracies of the casting products. However, using coordinate metrology data to defect some local defects, such as blisters, on the surface of the products is not seen often in the common practice. This paper presents an intelligent approach to extracts blisters directly from the digital coordinate metrology data captured using laser scanner installed on the production line A methodology is developed that uses a data clustering approach considering the manufacturing characteristics of the casting surfaces to detect the possible blisters on the surfaces. The developed algorithm is fast, reliable, an intelligent that can be employed by variety of coordinate metrology systems for online-inspection of the casting products.