Statistical diagnostics for industrial processes in a context of dynamic linear models
This work deals with systems which can be suitably modelled within a linear stochastic pattern. It refers to the management of industrial installations and production lines, where it is practicable to have a continuous, or quasi continuous, monitoring of the working operations. The methodology can be applied either to technological or environmental processes when they can be classified as dynamic systems and suitably represented by input and output matrix differential equations. The technique of statistical diagnostics can be considered as complementary to other methods of engineering because it may cover areas of intervention which are not yet reached by the other approaches. In particular, some important differences from classical control engineering can be noted. This work deals with two main topics concerning the methodology and the application. In the first one, the methods for the identification and analysis of systems as multivariate auto regressive moving average (ARMA) processes within the "black box" philosophy are introduced. In the second part, the two application aspects are studied: simulation models and laboratory facilities.
Bibliographic Reference: Paper presented: TOOLDIAG 93 - International Conference on Fault Diagnosis, Toulouse (FR), April 5-7, 1993
Availability: Available from (1) as Paper EN 36961 ORA
Record Number: 199210998 / Last updated on: 1994-12-02
Original language: en
Available languages: en