Le Laboratoire d’Automatique de l’EPFL a le plaisir de vous inviter aux séminaires selon la liste ci-après. Une mise à jour régulière des informations concernant ces séminaires est disponible à l’adresse sur cette page. En particulier, il est conseillé aux visiteurs externes de vérifier que les séminaires soient dispensés comme prévu ci-dessous.
Where: Salle de séminaire LA-EPFL, ME C2 405 (2è étage), 1015 Lausanne
When: Friday at 10h15
Adaptive Extremum-Seeking control as a tool for real-time optimization
01.07.2005 – Prof. Martin Guay – Queen’s University, Canada
In this presentation, we discuss the application of adaptive extremum-seeking control to the solution of real-time optimization problems. Two specific problems are discussed. First, the technique proposed is applied to the solution of optimal periodic control problems. In this application, the extremum-seeking task consists in steering the system to an orbit of the control system that optimizes a user-defined cost-functional. Second, we will discuss the application of the techniques for the solution of nonlinear model predictive control problems. It is shown that the resulting technique addresses the gap that exists between computational requirements and stability conditions in nonlinear model predictive control.
Nonlinear PI Control
16.06.2005 – Prof. Z.P. Jiang – Polytechnic University at Brooklyn, New York
In this talk we first motivate the importance of considering the problem of nonlinear PI control. Then, we present a methodology which we have developed in recent years, called nonlinear small-gain control method. Next, we show how to systematically apply the small-gain techniques to design robust nonlinear PI
controllers for a larger class of uncertain nonlinear systems with a triangular structure. It will be explained why the studied class encompasses the classes of nonlinear systems considered in the books of Krstic-Kanellakopoulos-Kokotovic (1995) and Marino-Tomei (1995). A practical example of electric fan speed control illustrates the presented nonlinear PI design scheme.
Predictive Control Applied to a Heat-Exchanger Network
03.06.2005 – Prof. J. L. Marchetti – Institute of Technological Development for the Chemical Industry (INTEC), Universidad Nacional del Litoral, Argentina
A new strategy for on-line optimisation and control of heat-exchanger networks (HENs) through a two-level structure will be presented. The low level is for a constrained model predictive control (MPC), and the high level is for a supervisor on-line optimiser. The proposed MPC algorithm uses the traditional step-response representation to perform the output predictions, and to account for the constraints. On the other hand, the supervisor program delivers key desired manipulated-variable positions to MPC, leading to minimum utility consumption. The coordination between the supervisor program and MPC is achieved through the definition of an extended cost-function that provides the ability of driving the system to optimal conditions. The proposed method was successfully tested by nonlinear simulation of a typical HEN system of the process industry.
Periodic Control: Theory and Applications
27.05.2005 – Prof. S. Bittanti – Dipartimento Di Elettronica e Informazione, Polytecnico di Milano, Italy
In control engineering, the typical design approach is based on a common paradigm, that the controller be linear and time invariant. This is the subject of thousands of papers, textbooks and monographs. However, there are problems for which a time-invariant controller may lead both to poor performances and to other problems that cannot be solved by time-invariant methods. Controllers with a periodically varying control law offer more degrees of freedom that can be usefully exploited; they may be considered as a bridge between time-invariant and generic time-varying controllers. The seminar will start by presenting some challenging engineering problems, so as to motivate the use of periodic controllers. Some main results achieved over the last 3 decades will then be overviewed. A few hints towards future research lines will conclude the lecture.
A Deterministic Global Optimisation Algorithm for Optimal Control Problems: Theory and Applications
20.05.2005 – Prof. C. S. Adjiman – Centre for Process Systems Engineering, Department of Chemical Engineering, Imperial College London, UK
A deterministic spatial branch and bound global optimisation algorithm is presented for systems with an initial value problem for a set of first-order, typically nonlinear, differential equations in the constraints. Upper bounds on the global minimum are obtained using the sequential approach for the local solution of the dynamic optimisation problem. The solution of a convex relaxation of the problem provides lower bounds. Well-known convex underestimation techniques are used for the relaxation of the algebraic functions. The concept of differential inequalities is used for the development of parameter independent as well as parameter dependent bounds on the dynamic system. Three convex relaxation procedures are proposed for the parameter dependent solution of the initial value problem. The global optimisation algorithm is illustrated by applying it to several case studies. The derivation of the underestimators is demonstrated step by step on small examples.
Iterative Feedback Tuning: from the bare essentials to some recent developments
07.04.2005 – Prof. M. Gevers – Center for Systems Engineering and Applied Mechanics (CESAME), Louvain-La-Neuve, Belgium
Iterative Feedback Tuning (IFT) is a model-free controller tuning method, in which the controller parameters are iteratively updated in order to reach the minimum of a LQG-like control performance criterion. The minimization is performed using a gradient descent scheme. The gradient is estimated from data collected on the closed-loop system, using a special experiment. The methodology is still undergoing developments and extensions. This talk will present the basics of the methodology, some industrial applications, and some extensions and recent developments.
Powerline Communications for In-home Networks and Internet Access
18.03.2005 – Prof. H.A. Latchmann – Electrical & Computer Engineering, University of Florida, USA
This presentation will discuss physical and medium access control protocols required in supporting a new generation of high speed networking using existing electrical wiring within typical homes. The unique power-line channel characteristics of Power-Line Communication (PLC) in-home LANs and Broadband Power-Line (BPL) Access Networks require sophisticated modulation, channel access and error control coding schemes in order to achieve effective data rates in the 100 Mbps range. The presentation will describe the HomePlug 1.0 protocol and ongoing work on the HomePlug AV protocol. An overview of BPL access will also be provided together with a discussion of the implications for the co-existence of in-home and access networks.
Commande prédictive et observation d’un lit mobile simulé
05.11.2004 – Prof. J.P. Corriou – ENSIC, Nancy, France
Le lit mobile simulé est un procédé de séparation continu par chromatographie qui repose sur la comparaison avec un lit mobile vrai. Toutefois, dans le lit mobile simulé, la phase solide est fixe et son mouvement est simulé par des commutations cycliques de vannes qui entraînent un fonctionnement non stationnaire. Son opération dépend étroitement d’un modèle de simulation performant. Une commande prédictive multivariable de ce lit mobile simulé sera présentée. En outre, un développement d’observateur permettant d’estimer les profils
dans le lit et les puretés sera explicité.
Measurement-based Run-to-run Optimization of Batch Processes: Application to Industrial Acrylamide Copolymerization
29.10.2004 – M.G. François – LA EPFL
Measurement-based optimization via NCO-tracking is a method that enforces the Necessary Conditions of Optimaltity using the available measurements and feedback control. This presentation will focus on meeting both the terminal constraints and sensitivities parts of the NCO. It will be shown through a variationnal analysis of the NCO that the terminal parts of the NCO can be met using batch-end measurements. Two run-to-run optimization algorithms are proposed, and both are shown to be globally convergent for a class of nonlinear systems. Finally, the proposed methodologys implemented in a 1-ton industrial acrylamide copolymerization reactor leading to a batch time reduction of 30% within three batches.
Optimisation basée sur les mesures du processus d’éléctro-érosion
19.11.2004 – M. P. Huguenin – LA EPFL
L’usinage par électro-érosion par enfonçage a des caractéristiques qui évoluent dans le temps, au fur et à mesure que l’électrode entre dans la pièce. Il est possible de montrer que la poursuite de la vitesse maximum d’usinage conduit à une stratégie optimale en temps. L’électro-érosion au fil, par contre, ne varie pas dans le temps. Mais les phénomènes aléatoires sous-jacents, qui se reflètent dans les mesures, rendent la tâche de l’optimisation très ardue. Les algorithmes classiques sont très peu robustes aux bruits de mesure, ce qui nuit à la qualité de leur convergence. Des modifications simples des algorithmes Golden Ratio Search et du Simplex de Nelder-Mead seront présentées permettant d’améliorer considérablement les résultats de l’optimisation, tout en limitant la nécessité de filtrage. Des résultats en simulation et en pratique seront présentés.
Ten years at Laboratoire d’automatique : On the synergy between optimization and control
3.12.2004 – Dr. B. Srinivasan – LA EPFL
in room CM 010 map
Automatic control and optimization of dynamic systems are two main areas addressed by the systems community. Techniques proposed for the optimization are typically model-based, while those proposed for control are typically based only on current measurements. A synergetic use of model and measurements for control and optimization has been my major research interest at LA. An original research approach termed “Measurement-based optimization via NCO tracking” has been initiated, where the measurements are used directly in the optimization framework to compensate uncertainty and process variations. In the domain of control, the scope has been broadened towards information-based control where control structures are proposed based on model information and includes various past and current measurements. The talk will be followed by a series of presentations on this subject by those whom I have had the pleasure to work with.
Compiler for fixed-point filters
10.12.2004 – Dr. O. Chételat – CSEM, Neuchâtel
The translation of a transfer function to its fixed-point realization is intrinsically difficult, because there is an infinite number of solutions and the optimum criterion is not well defined. One would like a fixed-point filter to require as little as possible computing power while at the same time having minimum numerical noise, no saturation, and minimum alteration of frequency response. A moderately-complex fixed-point filter is already a difficult, tedious, time consuming and error prone process when done by hand. We developed a compiler that can translate any transfer function to its C code fixed-point realization.
The realization may not be optimal in all cases, but provides a good general tradeoff, sufficient in most situations. The theory justification and some implementation aspects will be presented.
Metabolic Pathways and Macroscopic Reaction Schemes: Relationship, Model Reduction, and Generation
17.12.2004 – Dr A. Vande Wouwer – Faculté Polytechnique de Mons, Belgique
In this study, a class of dynamic models based on metabolic reaction pathways is analysed, showing that systems with complex intracellular reaction networks can be represented by macroscopic reactions relating extracellular components only. Based on rigorous assumptions, the model reduction procedure is systematic and allows equivalent input-output representations of the system to be derived. The procedure is illustrated with a few examples, and a comparison is made with another recently published method for generating and evaluating macroscopic reaction schemes.
Robustification de lois de commande prédictive par la paramétrisation de Youla. Application à la commande en position d’un moteur asynchrone
Observateurs des systèmes non linéaires et application aux procédés de polymérisation
21.01.2005 – Prof. H. Hammouri et Prof. G. Fevotte – Université Claude Bernard, Lyon, France
Le filtre de Kalman étendu a fait l’objet de plusieurs travaux théoriques concernant sa convergence. Plusieurs applications ont montré son efficacité. Hormis le cas linéaire, la convergence du filtre de Kalman étendu ne peut être garantie que localement et sous certaines hypothèses. De plus, plusieurs difficultés numériques apparaissent lors du réglage des gains de cet observateur. L’objet de cette conférence consiste à analyser pour quelle classe de systèmes non linéaires on peut construire un observateur dont la convergence est globale et dont le réglage du gain est simple à mettre en oeuvre. La deuxième partie de cet exposé est consacrée aux applications à l’estimation en ligne dans les réacteurs de polymérisation. Nous parlerons aussi de la calorimétrie.
Model-based fault detection and diagnosis of technical systems
28.01.2005 – Prof. R. Isermann – Institut fuer Automatisierungstechnik, Darmstadt, Deutchland
For the improvement of reliability, safety and efficiency, advanced methods of supervision, fault detection and fault diagnosis become increasingly important for many technical processes. This holds especially for safety-related processes like aircrafts, trains, automobiles, power plants and chemical plants. The classical approaches include limit or trend checking of some measurable output variables. Because they do not give a deeper insight and usually do not allow a fault diagnosis, model-based methods of fault detection were developed by using input and output signals and applying dynamic process models. Also signal model approaches were developed. The goal is to generate several symptoms indicating the difference between nominal and faulty status. This contribution gives a short introduction into the field and shows some applications for an actuator, a passenger car and a combustion engine.