Designing small samples for form error estimation with coordinate measuring machines

We propose a method for designing inspection plans of small size for use in form error evaluation of parts from batch or mass production using coordinate measuring machines. The method exploits a priori information coming from large sample inspection of one or a few parts of the lot. It is based on a regression model fitted to the large sample. The model is used twice. First, it provides the dominant pattern of the surface. Then, the deviations from the model that are relevant to form error are captured by weighting the points of the convex-hull of the large sample with the regression residuals. Based on two case studies, we show that the method provides a good accuracy in the estimation of straightness and flatness with very few measurement points. Comparative results also indicate that the method outperforms both the typical sampling schemes used in industry (random sampling, latin hypercube sampling), which do not exploit a priori information, and a recently proposed method using the same kind of a priori information.