An investigation of the effect of sample size on geometrical inspection point reduction using cluster analysis

Since the model program in automotive industry gets more and more extensive, the costs related to inspection increase. Therefore, there are needs for more effective inspection preparation. In this paper a method for reducing the number of inspection points using cluster analysis is tested on production data. This leads to reductions evaluated up to 90 percent in the case studies considered. Furthermore, the relation between movements in locators and the resulting movements in inspection points is used to find inspection points particularly suited to monitor the fixtures and its locating points. Those same points are used as input to the cluster analysis and chosen as representatives for the clusters. Using cluster analysis, the sample size is an important matter. The sample size, of course, affects the statistical confidence, but it is also important to choose a sample large enough to contain as many effects as possible related to the different process phenomena that exist over time. Those issues are investigated and a sample size lower than 50 items cannot be recommended.