Section: New Results

Modeling, observation and control: systems modeled by ordinary differential equations

Nonlinear system identification

Participants : Pierre-Alexandre Bliman, Boyi Ni, Michel Sorine, Qinghua Zhang.

In the framework of the joint Franco-Chinese ANR-NSFC EBONSI project, in collaboration with the Laboratory of Industrial Process Monitoring and Optimization of Peking University, and with Centre de Recherche en Automatique de Nancy (CRAN), the topics studied this year on nonlinear system identification are mainly on extended Hammerstein system identification with hysteresis nonlinearity and on continuous time block-oriented nonlinear system identification.

Motivated by the modeling of control valves with significant stiction, we have studied extended Hammerstein systems composed of a hysteresis nonlinearity followed by a linear dynamic subsystem. The joint characterization of the control valve and of the controlled process is formulated as the identification of an extended Hammerstein system. A point-slope based hysteresis model is used to describe the input hysteresis nonlinearity of the control valve. An iterative algorithm is proposed to solve the identification problem. The basic idea is to separate the ascent and descent paths of the input hysteresis nonlinearity subject to oscillatory excitations. Industrial examples are tested to verify the effectiveness of the proposed identification algorithm for characterizing complex behavior of control valve stiction in practice. This work has been presented at the 16th IFAC Symposium on System Identification [66] .

A Hammerstein-Wiener system is composed of a dynamic linear subsystem preceded and followed by two static nonlinearities. Typically, the nonlinearities of such a system are caused by actuator and sensor distortions. The identification of such systems with a continuous time model has been studied this year in collaboration with colleagues of CRAN. Based on previously developed simplified refined instrumental variable method, and by making use of an adaptive observer for data filtering, a combined approach, referred to as Kalman pre-filtered instrumental variable based method, is developed. By taking advantages of the two aforementioned methods, the new method is faster and has a naturally stabilizing Kalman filter that does not color white noises. This work has been presented at the 16th IFAC Symposium on System Identification [62] .

Model-based fault diagnosis

Participants : Abdouramane Moussa Ali, Qinghua Zhang.

The increasing requirements for higher performance, efficiency, reliability and safety of modern engineering systems call for continuous research investigations in the field of fault detection and isolation. This year we have studied algebro-differential systems through an adaptive observer based approach, and linear time varying systems through a Kalman filter based statistical testing approach.

In the framework of the MODIPRO project funded by Paris Region ASTech, the monitoring of the air conditioning system of an aircraft has been studied this year. Part of this system is modeled by nonlinear algebro-differential equations. A method for fault diagnosis of such systems has been developed in our study. Through a particular discretization method and under realistic assumptions, the considered continuous time DAE model is transformed to an explicit state space model in discrete time. An adaptive observer is then applied to the discretized system for monitoring faults possibly affecting the system and represented by changes in model parameters. This work will be presented at the 5th IFAC Symposium on System Structure and Control [61] .

While the theory of fault diagnosis has been mostly developed for linear time invariant (LTI) systems, in many industrial applications it is important to take into account the nonlinear behavior of the monitored systems. One possible approach is to linearize a nonlinear system all along its state trajectory, resulting in linear time varying (LTV) or linear parameter varying (LPV) models. In collaboration with Michèle Basseville of IRISA (Institut de Recherche en Informatique et Systèmes Aléatoires), fault diagnosis for stochastic LTV systems has been studied this year. By applying the Kalman filter in a particular manner avoiding the difficulty related to unknown faults possibly affecting the system, the problem of fault diagnosis in a dynamic LTV system is transformed into a hypothesis testing problem in a simple linear regression model. Generalized likelihood ratio (GLR) tests are then applied to the resulting hypothesis testing problem. This work has been presented at the 16th IFAC Symposium on System Identification [67] .