Students projects (Summer 2006)
Attitude determination and control of Cubesat
The LMTS (Microsystems for Space Technologies Laboratory) in partnership with the Space Center are starting to work on the first EPFL satellite, to be launched in early 2007. The Cubesat format (1 kg cube with a 1 litter volume) is chosen and several laboratories participate in developping and building of this sattelite. The LA is in charge of the attitude determination and control of Cubesat. The attitude control is a challenging problem due to power and mass limitation of the actuators and sensors (max. 0.1 W and 50 g). The following steps can be considered in this project:
1. Choosing the sensors and actuators for attidude determination and control.
2. Dynamic modeling of Cubesat.
3. Determination of Cubesat attitude.
4. Optimal controller design.
The designed controller will be tested in simulation using the dynamic model of Cubesat.
For more information see the official site of EPFL Cubesat.
Modeling and control of the Hydropter
The Hydropter is a sailing boat using hydrofoils (lifting surfaces operating in liquids). The hydrofoils generate enough lift to carry the boat and balance the forces produced by the sails at a level of drag far inferior to those obtained with classical hulls : the boat flies on the water. The speeds reached by the hydropter exceed largely those obtained with standard sailing boats. The main drawback of this type of system is the lack of stability with respect to waves and the difficult steering.
The goal of this project is to study the dynamical behavior of the hydropter using a model based on rigid bodies and approximate aerodynamical / hydrodynamical interractions. The controllability of the machine will be studied as a function of the boat parameters. The resonance modes due to waves (and their damping) will be studied. The control problem will be tackled by active and passive approaches, by manipulating the lift forces of the empennage (active control) and the angles of attack (AoA) of the hydrofoils (active/passive control). Eventually, the control will be optimized to maximize the performance and to minimize the fatigue of the structure.
Amortissement de vibrations fil en électroérosion
L’électroérosion est un procédé d’usinage pour matériaux conducteurs de toute dureté, servant à réaliser des découpes avec une précision du micromètre. Par ses propriétés, l’électroérosion est une des technologies clefs pour la réalisation de moules et d’outils de découpe. Il existe deux processus distincts : l’usinage par fil et l’enfonçage. Dans l’usinage par fil, un fil est tenu immergé dans un liquide diélectrique entre deux guides, séparés de 5mm à 500 mm. Suite à l’application d’une tension électrique entre le fil électrode et la pièce à usiner, une étincelle de haute densité d’énergie éclate. Ainsi la matière de la pièce à usiner est fondue, et par l’arrêt busque de l’étincelle les particules sont éjectées dans le diélectrique.
Les applications de pointe de l’électroérosion par fil nécessitent l’emploi de fils de plus en plus fins, c’est-à-dire de diamètres inférieurs à 50 micromètres. La fréquence propre de ces nouveaux fils peut entrer dans la bande passante des entraînements des guides. Ainsi, il devient envisageable de réaliser un amortissement actif des oscillations qui sont induites par le processus d’enlèvement de matière ainsi que par le conditionnement du diélectrique.
L’objectif du projet consiste à concevoir un algorithme d’amortissement actif des vibrations du fil en se servant des signaux d’usinage comme entrées et des commandes de vitesse d’axe comme sorties. Le procédé proposé doit s’adapter aux changements de fréquence propre qui résultent de différentes distances entre-guides, matériaux du fil et précharges.
Ce projet est proposé en collaboration avec Charmilles Technologies SA.
Modeling and control of a closed-loop current sensor
Measurement of electrical current is of great importance in many industrial fields, especially when a precise and accurate measurement has to be obtained. In these cases, the magnetic field created by the current is typically measured by a Hall-effect sensor. The precision can be further improved by means of a feedback closed-loop system. The idea is to generate a magnetic field in the opposite direction of that of the current to be measured such that the error between the two fields becomes zero. This way, it will be possible to amplify the error with a high-gain power amplifier to make more precise measurements.
In this project, a high-precision sensor based on the “flux-gate” technology with application in medical photography will be studied. The Hall sensor is replaced by a complex sensor containing two separate flux-gate systems for high and low frequencies. The objective of the project is to model the different parts of the closed-loop system and construct a complete model for simulation. This model will be used to design an optimal controller for noise disturbance rejection and high-precision current measurement.
This project is proposed in collaboration with LEM .
Control of the chaotic movement of a magnetic pendulum
In the context of building a demonstrator for the automatic control lab, a magnetic-chaotic pendulum is proposed to be controlled. The proposed setup is constituted of a permanent magnet pendulum balancing over an angular-controlled base with 4 magnets on it. A DC motor, is controlling the angular position of the base-magnets.
This magnetic pendulum is passively following a chaotic trajectory (resulting from the interaction between the magnets), which we propose to reject using an elaborate nonlinear controller, acting on the base-magnet angular position.
This challenging project includes part or all of the following points:
-Mathematical modelisation of the dynamic of the system Inclusion of the actuation in the dynamics
-Design of a linearized control law
-Design of global controller rejecting the chaotic trajectory and stabilising the pendulum, either on a contolled circle, or on stable point.
-Realization of a prototype, including the position sensor of the pendulum to close the loop.
Identification for state estimation
In “identification for state estimation” , a new cost function is minimized, which combines a classical output least squares criterion with an observability measure. The method has several advantages and drawbacks that have not been fully analyzed yet, and which call for further investigations:
a) The construction of the combined cost function, and in particular the selection of a weighting factor proceeds in a heuristic way. It would be desirable to more systematically build a combined cost function and there might be two possibilities to achieve this:
– the use of the L-curve approach for resolving the compromise ;
– the use of the results reported in “identification for optimization”  or in preferential estimation .
b) The identified model, with better observability properties, is biased. An idea for the compensation of the bias is to introduce some additional state variables called “garbage collectors”  to reduce the bias in a subset of states as it is done in preferential estimation.
Hence, the identification for estimation approach combined with preferential estimation, obtained from a) and b), would consist of two steps:
– Bias the identification model so as to increase a measure of observability for the preferred states,
– Compensate it later via preferential estimation using either off-line measurements (calibration) or infrequent on-line measurements.
The goal of the project is to investigate this two-step approach, and to report on successes and failures in the modification and extension of the concept of identification for state estimation. The approach will be illustrated in simulation by using a fermentation model.
 Ph. Bogaerts and A. Vande Wouwer. Parameter identification for state estimation – application to bioprocess software sensors. Chem. Eng. Sceince, 59:2465-2476, 2004.
 Per Christian Hansen, Rank-deficient and discrete ill-posed problems: numerical aspects of linear inversion, Society for Industrial and Applied Mathematics, Philadelphia, PA, 1999.
 B. Srinivasan and D. Bonvin. Interplay between identification and optimization in run-to-run optimization schemes. In American Control Conference, pages 2174-2179, Anchorage, Alaska, 2002.
 L. Bodizs, B. Srinivasan, and D, Bonvin. Preferential estimation via tuning of the Kalman filter. In Dynamics and Control of Process Systems – 7, Cambridge, Massachusetts, July 2004.
 P. de Valliere and D. Bonvin. Application of estimation techniques to batch reactors – II. Experimental studies in state and parameter estimation. Comp. Chem. Eng., 13:11-20, 1989.
Professor: Dominique Bonvin
Type of project: Master
Assistant: Levente Bodizs
Students: Stephanie Place and Quentin Trigallez
Parameter estimation of an industrial reactor
This project considers a chemical reaction system from the fine chemical industry. The main challenge of this reaction system is the presence of an intermediate unstable product that, above a certain temperature, decomposes under very large heat release. Thus, safe operation of the reactor must avoid accumulation of this hazardous product. Based on common chemical knowledge, an appropriate reaction model with a few model parameters has been obtained. In this project it will be important to estimate various parameters of the reaction system such as the reaction enthalpies, the kinetic reaction parameters, and the solubilities. Experimental data from several reaction calorimeter experiments are available from the Swiss company Lonza. The data includes on-line heat flow and infrared measurements. A new approach for evaluating both data sets simultaneously  will be studied. In this work, Matlab and gPROMS will be used for the parameter estimation.
 A. Zogg, U. Fischer and K. Hungerbühler, 2004, “A new approach for a combined evaluation of calorimetric and on-line infrared data to identify kinetic and thermodynamic parameters of a chemical reaction”, Chemometrics and Intelligent Laboratory Systems, 71, 165-176.
Implementation of a reference-tracking scheme for a batch distillation column
Batch distillation is utilized for separating mixtures of low quantity by exploiting a difference in boiling points of the mixture’s components. The objective is to attain a specified distillate purity at the end of a batch. A method to attain the desired purity is by on-line tracking of a purity reference of the accumulated distillate, which steers the system to the desired purity at final time. Unfortunately, on-line purity measurements of the distillate are only available with an important delay, thereby deteriorating tracking performance.
The main goals of this work are:
1) Conception of a reference-tracking scheme utilizing the delayed measurements (simulation with MATLAB).
2) Design of an observer for the distillate purity based on available temperature and flowrate measurements; utilization of the estimated purity for reference tracking (Simulation with MATLAB).
3) Implementation of the tracking schemes 1) and 2) on an existing batch distillation column with LABVIEW DSC.
This work is done in collaboration with the Ecole d’Ingénieurs de Fribourg (EIF) and requires working at EIF for the implementation part.
Professor: Dominique Bonvin
Type of project: Master
Assistant: Alejandro Marchetti
Student: Allouch Rabei
Modeling and identification of a double-axis positioning system
The dynamics of a double-axis positioning system using linear permanent-magnet synchronous motors depend on the axes position. A first study based on frequency-response identification of the system in different positions has shown this dependency. The first goal of this project is to model the system using a mass-spring-damper system, in order to find the influence of the position of the x-axis on the dynamics of the y-axis and vice versa. The second part of the project is to identify a parametric model for the system in different positions and to represent the model as a linear parameter varying system.
This project is defind in collaboration with ETEL.
Robust control of a flexible joint
Robust controllers are designed to stabilize and satisfy the performances for a set of models. This type of controllers can be applied to nonlinear systems linearized in different operating points, time-varying systems as well as systems with parametric uncertainty. An approach to robust control design is to cast the problem to a convex optimization problem which can be solved efficiently using the numerical methods.
In this project, the robust control of a flexible joint is considered. This system contains a main arm attached to a servomotor by two identical springs. The system parameters can be changed by adding an extra arm or changing the anchor points of the springs. The objective is to design a robust controller via convex optimization which stabilizes the all possible configurations of the flexible joint and gives the best possible performances.