Abstract
Numerous studies have shown that the choice of measurement strategy (number and position of measurement points) when measuring form error on a coordinate-measuring machine (CMM) depends on the characteristics of the machining process which was used to machine the examined surface. The accuracy of form error assessment is the primary goal of verification procedures and accuracy is considered perfect only in the case of the ideal verification operator. Since the ideal verification operator in the “point-by-point” measuring mode is almost never used in practice, the aim of this study was to examine a relationship which had not been examined in earlier studies, namely how the machining process, surface roughness and a reduced number of points in the measurement strategy affect the accuracy of flatness error assessment. The research included four most common cutting processes applied to flat surfaces divided into nine different classes of roughness. In order to determine functional dependency between the observed input variables and the output, statistical regression models and neuro-fuzzy logic (artificial intelligence tool) were used. The analyses confirmed the significance of all three input parameters, with surface roughness being the most significant one. Both the statistical regression models and neuro-fuzzy models proved to be adequate, matching the experimental results. The use of these models makes it possible to determine flatness error measured on a CMM if input variables considered in the paper are known.