ENBIS: European Network for Business and Industrial Statistics

Statistical methods for analyzing dynamic measurements

21 September 2009, 16:05 – 16:25

Abstract

Submitted by

Peter Hessling

Authors

Peter Hessling

Affiliation

SP Technical Research Institute of Sweden, Measurement Technology, Box 857, SE-501 15 Borås, Sweden

Abstract

Until recently, evaluation of measurement uncertainty has been limited to stationary measurands. ‘Stationary’ refers to quantities that are either constant, or can be described by constant parameters (as for instance harmonic signals). The widely applied guide to the expression of uncertainty in measurement (GUM) appears to only directly address this case.

Measurements of non-stationary or transient signals are nevertheless made in abundance. The measurement system is usually calibrated for stationary signals and the result is strongly dependent on the signal. As a consequence, any transient measurement will completely lack any trace of traceability – no matter how accurate the stationary calibration is, it is not generally applicable to transient measurements. A claimed uncertainty of around 0.5% may in practice be 20%. Perhaps 15% originate from systematic errors but at most 10% of it should be corrected for to avoid noise amplification without control. This non-existence of good practice jeopardizes the whole concept of calibration and for many measurements confines traceability to only exist in calibration laboratories.

Methods of conventional dynamic analysis are here merged with the statistical methods of GUM. The calibration procedure is extended as required and practical solutions are proposed. Methods for evaluation of uncertainty are generalized to accommodate non-stationary measurements and modern DSP tools such as customized digital filters will be utilized. As the ambition never should be to completely eliminate the complex dynamic error, a simple solution to the subtle question of optimal correction will be suggested. Examples will be used to illustrate how the methods are applied but also transferred to potential customers.

Return to programme