A self-adaptive method for the assessment of dynamic measurement uncertainty

Abstract

Measurement uncertainty is as important as measurement in metrology and industry. The GUM and its supplements provide a widely accepted framework for evaluating measurement uncertainty; but don’t provide a reasonable assessment method for some special circumstances, especially for dynamic measurement. Several emerging methodologies with different mathematical approaches are used for evaluating the dynamic uncertainty in a specific application, such as knowing the characteristics of data. To expand the applicability, a self-adaptive method is proposed. This method evaluates measurement uncertainty by analyzing the compositions of dynamic data, regardless of linearity, stationarity, or stochasticity. Information entropy on spectra combined with EDM algorithms is presented to divide dynamic data into deterministic and stochastic components; and then a Bayesian model and a time-varying auto-regression model are used to analyze decomposed components, respectively. Synthetic noisy signals and experimental data from a double-rotor table are utilized to demonstrate the effectiveness of the proposed method.