Commande et analyse des interactions des systèmes multivariables par bond- graphs

Abstract

In recent years, there has been considerable interest in developing process control strategies for multivariable control systems. Multivariable control involves the objective of maintaining several controlled variables at their independent set points. However, in these systems new characteristics due to interaction must be considered. Interaction results from process relationships that cause a manipulated variable to affect more than one controlled variable. Despite its importance, interaction analysis gives a full understanding of process behavior. There are two basic multivariable control approaches. The first is a straightforward extension of single-loop control to many controlled variables in a single process. This is termed multiloop control and has been applied with success for many decades. The second main category is coordinated or centralized control, in which a single control algorithm uses all measurements to calculate, all manipulated variables simultaneously. The traditional industrial control strategy for multivariable control problems is to use a multiloop control scheme consisting of several conventional PI or PID controllers. The design of the multiloop control system can be broadly divided into two stages as follows: 1. Selection of controlled and manipulated variables to be paired on one-to-one basis; and 2. Selection of control law and tuning controllers to provide an acceptable level of performance. vi In this approach, the key decision is to select the best pairing of controlled and manipulated variables.this work is related to modelling, analysis and control of the multivariable systems. For processes represented by bond-graph, the aim is to propose suitable tools for concurrent engineering of control systems: is it possible to design, for a given model, a control strategy by interaction analysisÀ The first chapter aims to recall the main results allowing defining the modes of representation of the multivariable systems, and different analytical methods for interaction analysis, based either on the transfer function, or based on the state space representation. In the second chapter, some existing results dealing with bond-graph formalism are presented. It is particularly focused on the methods used to emphasise the structure of a multivariable system from its bond-graph model. The third chapter id devoted to the study of square models: The numbers of input and output variables are identical. Structural analysis methods are first presented. They aim to characterise graphically on the open loop bond-graph model whether a stable decoupled model can be designed. Geometric methods are then presented to synthesise the symbolic expressions of control laws insuring both closed loop noninteraction and stability. These methods are finally implemented on distillation column, which the product composition is to control