Geometrical Inspection Point Reduction Based on Combined Cluster and Sensitivity Analysis

Abstract

Analyzing inspection data is an important activity in the geometry assurance process, which provides vital information about product and process performance. Since inspection is related to a significant cost, it is desirable with an intelligent inspection preparation where the motive is to gather as much information as possible about the product and the process with a minimum number of inspection points. In many situations, a large number of inspection points are used despite the fact that only a small subset of points is needed. The reason for this redundancy is that most systems have only a few principal causes affecting groups of variables. In this paper, we use methods of cluster analysis to find these natural groupings of inspection points and to select one representing point from each cluster. Furthermore, if the relationship between some of the process parameters and inspection points are known from experiments or from computer simulations, then the cluster analysis is combined with sensitivity-based reduction. In this way, an efficient reduced inspection plan is built up. The practical relevance of the proposed methodology for reduction is verified on an industrial case study and by computer simulations.