Sami Othman

Failure diagnosis and nonlinear observer. Application to a hydraulic process

This paper deals with the problem of fault detection and isolation for nonlinear systems. The proposed approach is based on nonlinear disturbance decoupling techniques and nonlinear observers. First, we give sufficient conditions and a design procedure for the synthesis of residual generators. Next, we characterize a class of nonlinear systems for which a high gain observer can be designed. Finally, the two above developments are illustrated through a real application dealing with the detection and isolation of three failures in a hydraulic process. The design of the residual generators is shown and their performances are discussed.

Nonlinear observers for parameter estimation in bioprocesses

In this paper, we propose simple nonlinear observer-based estimators for the on-line estimation of kinetic rates in fermentation processes. These estimators, designed in continuous and discrete-time, do not assume or require any kind of specific relationships between the reaction rates and the state variables. Moreover, they are very easy to implement and in particular to calibrate. Indeed, the gain of the proposed estimators does not require the resolution of any dynamical system and it is explicitly given. Furthermore, due to the particular structure of this gain, the tuning of the estimators is reduced to the calibration of a single parameter and this whatever the number of considered components and reactions. Numerical simulations and real-life experiments have been carried out and corresponding results are given in order to highlight the performance of the proposed estimators. In this paper, we propose simple nonlinear observer-based estimators for the on-line estimation of kinetic rates in fermentation processes. These estimators, designed in continuous and discrete-time, do not assume or require any kind of specific relationships between the reaction rates and the state variables. Moreover, they are very easy to implement and in particular to calibrate. Indeed, the gain of the proposed estimators does not require the resolution of any dynamical system and it is explicitly given. Furthermore, due to the particular structure of this gain, the tuning of the estimators is reduced to the calibration of a single parameter and this whatever the number of considered components and reactions. Numerical simulations and real-life experiments have been carried out and corresponding results are given in order to highlight the performance of the proposed estimators.

Continuous-discrete observer for crystal size distribution of batch crystallization process

A continuous-discrete observer was designed to estimate the CSD (Crystal Size Distribution) in batch crystallization processes. The observer is based on the discretization of PBE (population balance equations), it describes the evolution of the CSD with the finite difference method. Since the discretization leads a high order model, a reduced model was then constructed in order to simplify the observer synthesis and to reduce the computation time. The observer’s output allows to calculate the moments of the CSD, which may be used for control purposes or process supervision.

On the use of nonlinear receding-horizon observers in batch terpolymerization processes with partially unmodelled dynamics: Formulation and experimental validation

Abstract In this paper, a nonlinear receding-horizon observer is proposed for state reconstruction in batch terpolymerization reactors. It is shown that when constrained receding-horizon estimation is used together with a dedicated crossing singularity heuristic, state reconstruction is possible even in presence of measurement noise and up-to 10% error on the r.h.s of the ODEs describing the systems dynamics. The efficiency and the real-time implementability of the overall scheme is shown through illustrative scenarios including both simulation and experimental validation.

Observer design based on immersion technics and canonical form

Abstract A crucial problem in immersion based observer design is the case where the dimension of the immersed system is greater than that of the original system, and the analytical expression of the inverse of the immersion is unknown. In this paper, we have proposed a method for constructing observers for such autonomous nonlinear systems.

Multivariable model predictive control of wind turbines in presence of actuator fault

Wind energy represents a promising renewable resource of electricity. Due to the higher complexity of modern wind turbines, advanced control becomes necessary. Multi input multi output control is known to be a good tool to account for coupling between the input and output variables, and is thus preferred in this work over decoupled single input single output controllers. Moreover, in order to account for the constraints on the inputs and outputs, model predictive control (MPC) is chosen. MPC also allows overcoming errors in the model parameters, which might occur due to faults in the actuators or sensors. A horizontal wind turbine is considered in this work, by manipulating the pitch angle and generator torque, in order to maintain producing the nominal power while minimizing loads on the generator system in the full load regime, i.e. at high rated wind speed. By tuning the weighting matrices of the controller, different weights can be attributed to the inputs and outputs. The controller was found to overcome faults in the actuator.

ON-LINE MONITORING AND CONTROL OF NON-LINEAR SYSTEMS VIA CONTINUOUS-DISCRETE OBSERVERS

In this paper, a continuous-discrete observer is proposed for a class of nonlinear systems. One of the main interests of this observer is that it can be used to accurately estimate certain parameters for which one usually requires a model. It has been applied to a solution copolymerization reactor, and validated in simulation for the case of a semi-batch vinyl acetate (VAc) and butyl acrylate (BuA) reaction. The output of the observer is then used to control the cumulative copolymer composition.