Data validation in large scale steady state linear systems
Résumé
Data reliability is of fundamental importance for process diagnosis, identification and control. Measurements having large, random or biased errors which go undetected lead to poor control of processes. Detection of such errors is therefore very important, but can only be carried out on the basis of a certain knowledge of the process, of its structure, of the location of the data sources (observations), and a certain degree of redundancy. Here we present a method of classifying the variables of steady state linear systems into: observable, unobservable, redundant and no-redundant variables. This classification gives information on the state of the system, the consistency of the data and leads to a way of validating the observable part of the process. A recurrent estimator is developed on the basis of an estimation of the maximum likelihood. An application of the method to material balance is presented.
Origine : Fichiers produits par l'(les) auteur(s)