GUIDE AND BAYESIAN APPROACH TO POLYNOMIAL REGRESSION MODELS IN FLOW METER CALIBRATION DATA

Olga S. Yoshida, Nilson M. Taira

Abstract

The polynomial regression model has been largely used in flow meter calibration. The document ISO 7006 - Part II describes how to adjust a polynomial expression to non linear calibration data. The Guide establishes general rules for evaluating and expressing uncertainty in measurement, but does not demonstrate methods of evaluating uncertainty in flow meter calibration. In this paper we concentrate on the uncertainty analysis for calibration of flow gas meters by using polynomial regression models. A detailed Guide analysis of this model will be given with selected examples of non linear calibration data. The Guide analysis of this model is dictated by assumptions made on the distribution of errors. Typically, it is assumed that the errors have normal distributions but it is not true in many situations. Alternatives to the normal distribution for regression errors are discussed such as Student-t with small number of degree of freedom using Bayesian statistics. The rules recommended in the Guide are viewed in the light of Bayesian concepts, and we conclude that the interpretation of the results are very natural. The Bayes calculations was done with the Bugs (Bayesian Using Gibbs Sampling) software.

Keywords

flowmeter calibration, uncertainty, guide, polynomial regression, Bayesian statistics

Download

IMEKO-TC9-2000-075.pdf

DOI

-

IMEKO TC

TC9 - Flow Measurement

Back to the proceedings