Uncertainty estimation by the concept of virtual instruments H. Haitjema, B. van Dorp, M. Morel and P.H.J. Schellekens Precision Engineering section, Eindhoven University of Technology, PO Box 513, 5600 MB Eindhoven, The Netherlands Published in: proceedings SPIE 4401 (2001) 'Recent Developments in Traceable Dimensional Measurements', Decker, Brown (ed) ABSTRACT For the calibration of length standards and instruments, various methods are available for which usually an uncertainty according to the GUM [1] can be set up. However, from calibration data of a measuring instrument it is not always evident what the uncertainty will be in an actual measurement (or calibration) using that calibrated instrument. Especially where many measured data are involved, such as in CMM measurements, but also in typical dimensional geometry measurements such as roughness, roundness and flatness measurements, setting up an uncertainty budget according to the GUM for each measurement can be tedious and even impossible. On the other hand, international standards require that for a proof of the conformance to specifications, the measurement uncertainty must be taken into account (ISO 14253-1 [2]). Apart from this it is not so consistent that a lot is invested in the calibration of instruments where it is still unclear what the uncertainty is of measurements carried out with these ‘calibrated’ instruments. In this paper it is shown that the 'standard' GUM-uncertainty budget can be modified in several ways to allow for more complicated measurements. Also, it is shown how this budget can be generated automatically by the measuring instrument, by the simulation of measurements by instruments with alternative metrological characteristics, so called virtual instruments. This can lead to a measuring instrument where, next to the measured value, also the uncertainty is displayed. It is shown how these principles are already used for roughness instruments, and how they can be used as well for e.g. roundness, cylindricity, flatness and CMM measurements. 1. THE STANDARD UNCERTAINTY BUDGET In general, a measurement result M is a function of N input quantities m i (i=1,2..N). This leads to the general functional relationship, known as the 'model function' [1]: ) m ,... m , m ( f M n 2 1 = (1) The model function incorporates the measurement and the calculation procedure. It can an analytical function, but also a complicated, iterative, computer algorithm. The measurement data m i can be grouped into two categories, dependent on the way they, and their uncertainty, are obtained: A Quantities where the value and its uncertainty are directly obtained from the measurements B Quantities where the uncertainties are obtained from other sources, such as calibration data, used material constants, previous measurements etc. However this grouping does not influence the uncertainty evaluation; it is just essential that a standard uncertainty u mi is attributed to any influencing quantity m i . The quantity M is best approximated by using the best approximations for m i , which are usually the measured data, in equation (1). Now, the uncertainty u M can be written as: j i j i j i i m M i M m m M u m M u i ∆ ∆ ∂ ∂ ∂ ∂ ∂ = ∑ ∑ < 2 2 2 2 2 ) ( (2)