SENSOR DIAGNOSTICS BASED ON DYNAMIC CHARACTERISTIC CURVE ESTIMATION

Kimmo Latva-Käyrä, Risto Ritala

Abstract

This paper presents a method for optimal detection of need for re-estimation of the parameters of characteristic curves based on infrequent and irregular reference measurements. We assume that the validity of parameters deteriorates through a stochastic process. In particular, we give analytic distributions of parameters when deterioration is through random-walk diffusion. The detection of need for re-estimation of parameters is similar to two-variable Statistical Process Control scheme. Method is demonstrated and its detection capabilities analyzed in cases where the actual deterioration is either a step change, a linear drift or a random walk.

Keywords

calibration, probability, diagnostics, Bayes, Statistical Process Control

Download

IMEKO-TC10-2005-03.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings