Frédéric Hamelin

Brief Robust fault detection in uncertain dynamic systems

The work reported in this paper focuses on the design of optimal filters for fault detection in uncertain dynamic systems. A design procedure dealing with interval type parameter uncertainties is proposed for residual generation via a frequency domain approach. Robustness enhancement is based on the design of optimal detection filters maximizing the effect of faults while minimizing the influence of disturbances. In this framework, in order to cope with parameter uncertainties, Kharitonov polynomials and Dasgupta geometry are introduced and an extension of previous results to worst case situation is proposed. An example is also included to illustrate the method efficiency.

Frequency-domain optimization for robust fault detection and isolation in dynamic systems

The paper deals with the design of optimal filters for fault detection and isolation (FDI) in dynamic systems. The proposed method, involving the generation of directional residuals derived from systems input-output description model, facilitates the enhancement of robustness even in cases where it is impossible to decouple fault effect from system perturbations. The proposed solution is based on the optimization of a performance index in the frequency domain. The resulting detection filter is capable of maximizing the effect of faults while minimizing the influence of disturbances.

Robust Residual Generation Via Lmi

Abstract This paper presents a new approach that attempts to optimise two contradictory objectives: disturbance attenuation and fault sensitivity. The proposed method is applied to linear time-invariant systems subjected to unstructured perturbations and critical faults that involve a frequency change. This approach is based on H 2 estimation which allows optimising the attenuation of the worst-case effects of disturbances on the residual. To enhance the sensitivity of this residual to faults, a parametrisation of the initial observer is proposed. This parametrisation intends to improve the amplification of critical faults on residual with regard to disturbances attenuation.

Brief paper: State and input observability for structured linear systems: A graph-theoretic approach

This paper deals with the state and input observability analysis for structured linear systems with unknown inputs. The proposed method is based on a graph-theoretic approach and assumes only the knowledge of the systems structure. Using a particular decomposition of the systems into two subsystems, we express, in simple graphic terms, necessary and sufficient conditions for the generic state and input observability. These conditions are easy to check because they are based on comparison of integers and on finding particular subgraphs in a digraph. Therefore, our approach is suited to study large-scale systems.

A fault-tolerant control design against major actuator failures: application to a three-tank system

This paper deals with a new fault-tolerant control method for nonlinear systems without hardware redundancy in a case of major actuator failures. The nonlinear law is achieved using an input-output linearization by state feedback. These failures correspond to a blocking or complete loss of an actuator which could lead to a large loss in the nominal performances. This method is based on the computation of new reference inputs in order to drive the system in an optimal operating order (optimal trim point for linear systems) with respect to desirable performances and to their degrees of priority. For a plant without actuator redundancies, it becomes of prime importance because it is impossible to maintain it at some acceptable level of performances in the presence of major actuator failures. The effectiveness of this approach is evaluated through the application to a nonlinear pilot plant, a three-tank system.

Brief paper: Observability analysis for structured bilinear systems: A graph-theoretic approach

This paper is devoted to the generic observability analysis for structured bilinear systems using a graph-theoretic approach. On the basis of a digraph representation, we express in graphic terms the necessary and sufficient conditions for the generic observability of structured bilinear systems. These conditions have an intuitive interpretation and are easy to check by hand for small systems and by means of well-known combinatorial techniques for large-scale systems.

Brief paper: State and input observability recovering by additional sensor implementation: A graph-theoretic approach

This paper deals with the problem of additional sensor location in order to recover the state and input observability for structured linear systems. The proposed method is based on a graph-theoretic approach and assumes only the knowledge of the systems structure. It allows one to provide the minimal number of the required sensors and either their pertinent location or a necessary and sufficient condition which allows one to check if a sensor location is adequate or not. We obtain a sensor placement procedure based on classical and well-known graph theory algorithms, which have polynomial complexity orders.

DECENTRALIZED AND AUTONOMOUS DESIGN FOR FDI/FTC OF NETWORKED CONTROL SYSTEMS

Abstract In this paper, the fault tolerant control problem is addressed in a networked framework. An isolation filter together with a fault compensation mechanism are proposed for FDI/FTC. Several design procedures are studied. First, in the centralized architecture, the inputs and outputs information used for FDI & FTC are collected and processed in a central node. Considering that sub-systems exchange information over the network, a decentralized architecture is then proposed for FDI/FTC. The autonomous capability of the decentralized design is also considered. But, whatever the design architecture is, fault detectability and isolability, are dependant on the available information. A method based on system structural analysis is then proposed, to provide fault detectability and fault isolability conditions, thus making possible fault tolerant control. Finally, we give an algorithm which allows to distribute a system into subsystems such that the problem of autonomous FDI is solvable for each of these subsystems.

Brief paper: Observability of switching structured linear systems with unknown input. A graph-theoretic approach

This paper deals with the observability of the discrete mode, the internal state and the input of switching structured linear systems with unknown input. The proposed method, based on a graph-theoretic approach, assumes only the knowledge of the systems structure. We express, in graphical terms, necessary and sufficient conditions for the generic observability of the discrete mode, the continuous internal state and the input of a switching structured linear system. These conditions can be implemented by classical graph theory algorithms based on finding particular paths and cycles in a digraph.

Fault tolerant control in dynamic systems using convex optimization

In this paper a method for fault diagnosis and control reconfiguration in dynamic systems is presented. The proposed approach is composed of two steps. In the event of a fault, the first step is the detection of the failed component, and the estimation of the fault parameters. The second step consists of the design of a new control structure which implicitly reconfigures the control law using the online estimates of the altered system dynamics. The application to a winding machine is presented to illustrate the proposed approach.