Conformity assessment using Monte Carlo methods

Abstract

Conformity assessment is the activity to determine whether specified requirements relating to a product, process, system, person or body are fulfilled. Often measurements are used to show that the measurand is within (legal) tolerances. Currently analytical methods are available to test whether tolerances are met with a preset level of confidence, e.g. 95%. The test requires the availability of the overall measurement uncertainty and the statistical distribution of the measurand. In absence of better information this distribution is assumed to be Gaussian. The new point in this paper is that Monte Carlo methods can be applied directly to perform the conformity assessment. The reason is that the Monte Carlo process generates the cumulative distribution, whereby the (legal) tolerances can be compared directly. The advantage of this process is that the type of distribution does not need to be known and the (worst case) assumption of the distribution being Gaussian can be avoided. Consequently, for a Monte Carlo method the difference between tolerances and acceptance criteria is slightly smaller than for analytical methods. A test of the Monte Carlo method applied to a calibration of a high-pressure gasmeter meeting MID tolerances demonstrates the applicability of the method in practice.