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 (2nd floor), 1015 Lausanne
When: Friday at 10.15am
26.02.2010 Prof. R. Ergon Telemark University College, Norway.
Process monitoring using methods from system identification and chemometrics.
On-line measurements of product qualities (primary outputs y) from industrial processes (chemical plants, food industry, etc.) are often not feasible. Instead, samples are taken at more or less regular intervals and brought to the laboratory for costly and time-consuming analyses. There is thus a need for primary output estimation at a high sampling rate, based on known inputs u and secondary process measurements z (flows, temperatures, etc.). In my talk I will discuss several related aspects.
For dynamical systems, we may use Kalman filtering and system identification methods, also when the primary samples are obtained at a very low and irregular sampling rate. Spectral secondary measurements (NIR, acoustics, etc.) must then be compressed into principal components, using principal component or least squares regression (PCR/PLSR).
PCR and PLSR may also be used directly for primary output estimation, using statistical limits for normal process operation. The Hotelling’s T2 statistic is then used to see if new z samples are acceptably close to the normal operating point within the projection space, while the squared prediction error SPE or Q statistic is used to detect abnormal deviations outside of the projection space. For this purpose new samples are split into z = zmodel + e, and for PCR this is unproblematic. The corresponding splitting in PLSR is much discussed at the time, and the issue is also complicated by sampling errors in y, which are often larger than errors in z.
The splitting of z in PLSR, i.e. the definition of zmodel and e, also affects the score-loading correspondence, and is thus of interest for fault diagnosis methods.
11.03.2010 Prof. P. Van den Hof Delft Center for Systems and Control, Delft University of Technology, The Netherlands.
Model-Based Control and Optimization Challenges in Reservoir Engineering
Due to urgent needs to increase efficiency in oil recovery from subsurface reservoirs, new technology is developed that allows more detailed sensing and actuation of multiphase flow properties in oil reservoirs. One of the examples is the controlled injection of water through injection wells with the purpose to displace the oil in an appropriate direction. This technology enables the application of model-based optimization and control techniques to optimize production over the entire production period of a reservoir, which can be around 25 years. Large-scale reservoir flow models are used for optimizing production settings, but suffer from high levels of uncertainty and limited validation options. One of the challenges is the development of reduced complexity models that deliver accurate long-term predictions, and at the same time are not more complex than can be warranted by the amount of data that is available. In this paper, an overview will be given of the problems and opportunities for model-based control and optimization in this field aiming at the development of a closed-loop reservoir management system.
19.03.2010 Prof. J.-P. Steyer Research Director, Laboratory of Environmental Biotechnology, (INRA Narbonne, France).
Lessons learnt from 15 years of ICA (instrumentation, control and
Anaerobic digestion (AD) plants are highly efficient wastewater treatment processes with inherent energy production. Despite these advantages, many industrial companies are still reluctant to use them because of their tendency to become instable when operating conditions change. There is therefore great potential for application of instrumentation, control and automation (ICA) in the field of anaerobic digestion. This seminar will discuss the requirements (in terms of on-line sensors needed, modeling efforts and mathematical complexity) but also the advantages and drawbacks of different control strategies that have been applied to AD high–rate processes over the last 15 years.
16.04.2010 Prof. K.J. Hunt Institute for Mechatronic Systems, Bern University of Applied Sciences, Switzerland.
Robotics-assisted Treadmill Technology: Feedback control methods for cardiopulmonary rehabilitation after neurological injury.
Rehabilitation robotics have been introduced in clinical practice to support therapeutic interventions in a wide range of neurological conditions. Several devices have been developed to provide robotics-assisted gait training for people who have suffered a spinal cord injury or a stroke. We have recently expanded the scope of application to encompass cardiopulmonary training and assessment. We developed a method of estimating the patient’s volitional effort during exercise, and a feedback control strategy, which allows pre-defined work rate profiles to be imposed during exercise. This structure has recently been augmented by a feedback control system, which automatically determines the work rate required to achieve a pre-defined target oxygen uptake (VO2) profile. These methods are successful in counteracting the strongly nonlinear response of VO2 during exercise. They can be applied during exercise training programs to optimize cardiopulmonary health, and during formal exercise tests to increase understanding and to improve estimates of important cardiorespiratory performance parameters. This talk will give an overview of this line of work, with a focus on feedback control aspects and the potential for clinical application.
27.04.2010 Prof. R. Sépulchre Systems and Modeling, Dpt. of EE and CS, University of Liege, Belgium.
Contraction measures for systems and control applications.
Contraction theory is the classical framework for the characterization and convergence analysis of fixed points in Banach spaces. But it is rarely exploited as such in systems and control theory, where the convergence analysis has been dominated by the construction of Lyapunov functions. The talk will emphasize the role of contraction theory in three applications of systems and control theory that have attracted considerable attention in the recent years: (i) energy-based control methods for electro-mechanical systems, (ii) synchronization theory, and (iii) consensus theory. In each of these areas, the specific role of contraction measures in contrast to the more conventional role of Lyapunov functions will be detailed. It will be shown that contraction theory is well adapted to engineering questions such as tracking or observer design for nonlinear systems and that the search of suitable contraction measures can serve as a fundamental guideline to the design of Lyapunov functions.
21.05.2010 Prof. F. Gallaire Laboratoire de mécanique des fluides et instabilités, Ecole Polytechnique Fédérale de Lausanne.
Application of optimal control theory for the re–stabilization of open flows: dream or reality?
Flow control refers to the ability to alter flows with the aim to achieve a desired effect; examples include drag reduction or noise attenuation among many other industrial applications. All these flow phenomena are associated with a strong flow unsteadiness that can be viewed as the nonlinear development of an initially linear instability of the underlying laminar flow. Feedback control appears as a natural candidate to quench these incipient instabilities at their earliest stages.
25.06.2010 Prof. B. Srinivasan Department of Chemical Engineering, Ecole Polytechnique de Montréal, Canada.
An Efficient Global Optimization Algorithm based on Multi-unit
Finding the global optimum of a nonlinear function is a challenging task that could involve a large number of functional evaluations. In this presentation, an algorithm that uses tools from the domain of extremum-seeking is shown to provide an efficient deterministic method for global optimization. Extremum-seeking schemes typically find the local optimum by controlling the gradient to zero. Here the multi-unit framework is used, where the gradient is estimated by finite difference for a given offset between the inputs. The gradient is pushed to zero by an integral controller. It is shown that, if the offset is reduced to zero, the system can be made to converge to the global optimum of nonlinear maps with constraints. Several illustrative examples are presented to show the capability of this methodology. In the examples, the proposed method is compared with other available methods of global optimization.