Measurement uncertainty: a soft computing approach

Measurements of any kind are characterized on one hand by their uncertainty due to modeling and measurement errors. Unfortunately for several reasons this characterization is not easy and requires further (human and/or machine based) considerations and intensive computing. As an alternative measurements are characterized also by their accuracy which can be improved also at the price of further data acquisition and computation. All these computations require time and therefore additional requirements like speed, costs, etc. may strongly limit the system designer in achieving the specified precision. Moreover the complexity of the measurement problems of current interest has considerably increased. Recent advances in time-critical computing and new modeling techniques provide promising tools to meet these requirements. Due to some of their features hereafter these techniques will be referred as "soft" computational methods. For a system engineer the most important knowledge is when and how to apply such new tools, what are the decisions not covered by the theory and how to characterize the final results. Based on the analysis of some measurement problems the authors discuss these questions and point out that the importance of these "soft" calculations is much higher than anticipated. The investigations are followed by an example demonstrating the application of fuzzy logic to a particular measurement problem.