A. Zolghadri

Structured H∞/H_ LPV filters for fault diagnosis: Some new results

In this paper, a method is presented to design robust fault detection and isolation filters for Linear Parameter Varying (LPV) systems modeled in a Linear Fractional Representation (LFR) fashion. It consists in designing an optimal FDI filter that minimizes the influence of unknown inputs on the residuals, in the H∞-norm sense and simultaneously maximizes fault sensitivity performance, in the H_-index sense. The design problem is formulated so that all free parameters are optimized via Linear Matrix Inequality techniques. Computational aspects are discussed and it is shown that the proposed solution is structurally well-defined. An illustrative example demonstrates the potential of the proposed method.

FAULT DIAGNOSIS IN LINEAR PARAMETER VARYING SYSTEMS

Abstract This paper presents a new approach to solve the FDI (Fault Detection and Isolation) problems in Linear Parameter-Varying (LPV) systems. A sufficient condition is established for guaranteeing fault sensitivity performance. The design problem is formulated within the LPV polytopic modeling framework and solved using LMI optimization techniques. The overall method is applied to a secondary circuit of a Nuclear Power Plant (NPP). Experimental results show the efficiency of the proposed method.

Specification of a Real-time Knowledge Based Supervision System

Abstract We propose a hierarchical structure for the supervision of an industrial process. The general frame for the specification of the knowledge based is presented through the specific example of a supervision system for the robust control of a robot. The role devoted to the supervision system lies in the surveillance of a few indexes of good functioning of the path-planner, the robust controller and the electro-mechanical plant. After detection, analysis and diagnosis of a failure situation, the supervision layer decides to set off appropriate actions as readjustment of the path-planner design parameters, either self-tuning of the robust controller through an identification phase, or switch to a different operating mode

Fault Detection in Induction Motors via Parameter Estimation Techniques

Abstract This paper deals with the application of model-based fault detection in an induction motor. Two different techniques (based on the extended Kalman filter) for estimating the physical parameters of the induction motor are presented A high performing simulator is used to perform identification experiments. In order to decide whether the discrepancy between the identified parameter values and the reference ones is significant in the presence of random disturbances and modelling errors, a robust decision test is implemented

A METHOD FOR DESIGNING FDI FILTERS FOR POLYTOPIC LPV MODELS

Abstract The aim of this communication is to introduce a methodology to solve the FDI (Fault Detection and Isolation) problem in Linear Parameter Varying (LPV) systems. The design problem is formulated within the LPV polytopic modelling framework. A sufficient condition is established for guaranteeing robust fault sensitivity performance level. It is then shown that the problem can be formulated and solved using a reasonable number of LMIs. The overall method is applied to an experimental data set, measured from the secondary circuit of a French nuclear power plant.

A robust FDI design procedure for multivariables feedback systems

Abstract The paper presents a new and general scheme to design robust Fault Detection and Isolation (FDI) filters for multivariable uncertain systems under feedback control. The design procedure ensures simultaneously robustness of the FDI output against disturbances and model1ing errors, and nominal sensitivity to faults. Robust sensitivity of the residual signals is analysed by means of a test based on the generalised structured singular value. The feedback controller is directly included in the design procedure, making the proposed approach very appealing in most practically relevant situations. Simulation results based on an engine failure scenario of the RCAM benchmark illustrate the potential of the proposed approach.

Monitoring Ground-Level Ozone Concentrations in Bordeaux (France).

Abstract The paper describes the status of an on-going research program to develop a highly reliable operational public warning system for air pollution monitoring in Bordeaux, France. Experimental results are presented for ground-level ozone concentrations. Meteorological variables are used as input in order to obtain an estimate of the next days maximum ozone concentration as well as a measure of its temporal extent. The proposed approach combines new adaptive nonlinear state-space modelling techniques, a gain scheduling strategy and multi-layer perceptron neural networks.

Model Based Fault Detection in a Multivariable Hydraulic Process

Abstract This paper deals with the application of model-based fault detection in a hydraulic process under digital control The approach is based on modelling and estimation methods and uses some known statistical tools as well as geometric arguments. The test decision is based on the overlap between the confidence regions associated with two estimates, one obtained using on-line measurements and the other based on a priori information. Practical results illustrate the potential of the method for detection of slowly acting faults.