Subset selection in predictive modeling of the CMM digitization uncertainty

Abstract

Abstract A number of factors affect digitizing accuracy, such as the travel speed of the probe, pitch values, probe angles (part orientations), probe sizes, and feature sizes. A proper selection of these parameters in a digitization or automatic inspection process will improve the digitizing accuracy. Factorial design of experiments is used to plan the experiments. Regression analysis is applied to developing the empirical models for prediction of digitizing uncertainty. Four criteria, namely the PRESS statistic, the adjusted R 2 , the C p statistic, and the residual mean square s 2 , are employed to select the best regression model. Hypothesis testing is conducted to check the goodness of each model in construction and to qualify the validation of the model. It is shown that the prediction model has a satisfactory goodness of fit. The method for model selection and validation as well as the best model selected can be used in both computer-aided reverse engineering and automatic inspection for error prediction and accuracy improvement.