Analysis of Comparison Data from Acceleration Measurements

Abstract

The Evaluation Guidelines [1] specify general analysis methods for key comparisons of measurement data where their reasonable application is subject to some preconditions. We discuss a real example of comparison data that can be analyzed in formal accordance with the Guidelines. But we prefer a model-based analysis benefiting from the intrinsic relation of the data. Indeed, this method is founded on the principles of the Guidelines. In our example, the laboratories participating in the key comparison have carried out a series of calibration measurements of a particular accelerometer. The values from this series correspond to the measurements at different excitation frequencies. A physical model describing the frequency dependence is known. The proposed model-based analysis estimates the unknown model parameters taking the values from the series of the laboratories simultaneously into account. If the derived model and the data are consistent, the model function gained represents the reference values. The advantage of the model-based analysis is that we utilize the a priori knowledge we have at our disposal. We obtain better reference values, i.e., the uncertainties associated with the reference values are considerably lower.