Residual Evaluation without Derivative Computation

Autor: Patrick Taillibert
Rok vydání: 2000
Předmět:
Zdroj: IFAC Proceedings Volumes. 33:605-610
ISSN: 1474-6670
DOI: 10.1016/s1474-6670(17)37426-8
Popis: Fault detection and isolation in dynamic systems, when based on parity space approaches, requires the evaluation of the derivatives of the input and output (observations) of the system to be diagnosed. Because of inaccuracies and noise in measurements, this evaluation is always very imprecise which leads to false alarms or missed detection of faults. We show that if the residuals are of the form “f+g′+h″+.=0” where f, g, h. are functions of observable quantities and parameters of the system to be diagnosed and if observations are represented by temporal band sequences (piecewise polynomial envelopes encompassing the observations), it is then possible to evaluate safely the residuals, that is without derivative computation. This work stems from research in the field of analog electronic circuits diagnosis for which derivating measurements is very hasardous because of the huge variance in the temporal characteristics of electronic signals.
Databáze: OpenAIRE