A PROBABILISTIC VIEW OF PROBLEMS IN FORM ERROR ESTIMATION

Form error estimation techniques based on discrete point measurements can lend to significant errors inform tolerance evaluation. By modelin g surface profiles as random variables, we show how sample size and fi tting techniques affect form error estimation. Depending on the surfac e characteristics, typical sampling techniques can result in estimatio n errors of as much as 50 percent. Another issue raised in the fitting approach is the metric p selection for the fitting objective. We show that for p = 2 and p = infinity, the selection does not appear to sig nificantly affect the estimation of form errors.