Skip to main content

Fuzzy algorithms for the control of multi-input multi-output processes

Objective

FAMIMO aims at designing fuzzy systems for the reliable control of high-dimensional complex processes. This will entail the development of methods and software environments for an easy design of general, robust and friendly-to-use fuzzy systems, and the testing of these environments together with a comparison of different fuzzy control schemes on two multivariable processes taken as benchmark: a biotechnological fermentation process and an automobile power-train management.
A main result should be a software library of several fuzzy control schemes together with some well-analysed forms of these schemes hybridisations which could be better suited for particular control complex applications.
The members of the project are: IRIDIA from Brussels, University of Delft, AICIA from Seville, University of Lund and SIEMENS automotive from Toulouse with, as sub-contractor, the LAAS laboratory from Toulouse. During a first part of the project, the partners will develop in a Matlab/Simulink software form, some methods they have preferentially studied. Reinforced interactions will, in a following phase, lead to a mutually enrichment of the methods by their complementary aspects. Then all resulting methods will be tested on two multivariable nonlinear processes (benchmarks) which both have strong industrial relevance despite the difficulty to obtain a precise mathematical description. At this level of the project, we will naturally be provided with quantitative measures to assess the originality and quality of the proposed methods (and therefore a sort of final assessment of FAMIMO outcomes). Well-argued statements about the industrial potentials of the most satisfactory fuzzy algorithms for complex MIMO processes should naturally conclude the project.

Fuzzy systems are becoming more and more accepted, both in industry and in the academic control community, as an emerging tool for the facilitated design of non linear feedback complex control systems. Right now and as a main result of the explosive Japanese scientific and industrial efforts dedicated to the development of fuzzy controllers, they have been designed for the control of very low dimensional processes or for the control of multi-variable but weakly coupled processes. However real challenges addressed to traditional control schemes appear for the control of high dimensional, varying on different time scales, strongly coupled, multi-objectives (often antagonist), multi-constrains, and hard to model processes. There is a real need for new analysis and design strategies that could be applied to facilitate the construction of fuzzy controllers, to reduce the time-consuming parameter tuning while emancipating this construction from an explicit mathematical knowledge of the process, to assure robustness towards imperfect modelling and external disturbances, and to obtain high performance and reliability. FAMIMO will move in two directions: first enrich fuzzy systems methodology thanks to a cross-fertilisation with control theory, but also enrich the set of applied nonlinear control tools with advanced fuzzy system schemes which have been empirically shown to solve challenging control problems. FAMIMO will also aim at making available analytic and computerised tools for assessing the stability of both the control and learning algorithms integrated in the future industrial exploitation of these methods.

Both experimental platforms, although quite different, are highly nonlinear and coupled control problems so that it should be obvious that the control softwares to be developed within the framework of FAMIMO have potential utility going largely beyond the two specific applications and aim at addressing most MIMO processes in general. Important aspects of these two benchmarks like "safety enhancement", "energy consumption saving"environmental safeguard" are of great interest for a lot of other process industries. In addition, each academic partner has numerous industrial collaborations which should draw a lot of profits from the three years developments and the results to be achieved by FAMIMO. The software to be delivered during the execution of the project should be generic enough to be easily adapted and used for a large diversity of MIMO processes.

Funding Scheme

CSC - Cost-sharing contracts

Coordinator

Universite Libre de Bruxelles
Address
Av. F. Roosevelt 50
1050 Bruxelles
Belgium

Participants (4)

Asociacion de Investigacion y Cooperacion Industrial de Andalucia
Spain
Address
Avda. Reina Mercedes S/n, Edificio Escuela De Ingenieros
41012 Sevilla
DELFT UNIVERSITY OF TECHNOLOGY
Netherlands
Address
Julianalaan 134
Delft
Lund Institute of Technology
Sweden
Address

Lund
SIEMENS VDO AUTOMOTIVE
France
Address
Avenue Paul Ourlia, 1
31036 Toulouse