A neural network approach to determining optimal inspection sampling size for CMM

Abstract

This paper reports a neural network approach to determining the optimal inspection sampling size of 'hole' features using the Coordinate Measuring Machine (CMM). Factors which could affect sample size due to design, manufacturing, and measurement related factors, i.e. size, dimensional and geometrical tolerances, machining processes, and confidence levels, have been studied. Machining process type, size, and tolerance band have been identified as the known factors which may affect the sample size required. Experiments have been carried out to collect sampling size data versus the variation of these factors for different 'hole' features. The implicit correlation between the sample size and these factors, has been achieved by training a back-propagation neural network using the collected data. The neural network architecture is described, and the test of the trained neural network on a few new 'hole' features is presented to highlight the applicability of this approach. Copyright © 1996 Elsevier Science Ltd.