PhD: University of Poitiers, France

Contributed by: Guillaume Mercère, guillaume.mercere@univ-poitiers.fr

Experimental modeling of heat exchangers: a comparison of LPV and 2D model identification

Due to the increasing exploitation of heat exchangers in different industrial frameworks, it is necessary to have access to reliable and relevant models of such processes. The determination of such efficient models is the main objective of this PhD program. Although some interesting results can be found in the heat exchanger modeling literature, several important practical and theoretical problems remain to be solved as far as heat exchanger system identification is concerned. One of them has to do with the selection of a proper model structure. Indeed, looking closer at the literature, most of the developments dedicated to heat exchanger identification deal with linear time-invariant (LTI) models or black-box artificial neural networks. The different experiments carried out by the researchers involved in this PhD program have shown that using only LTI black-box models leads to inaccurate models because of the non-linearities, as well as some varying parameters governing the behavior of heat exchangers. In order to circumvent this difficulty, two complementary approaches could be considered.

First, the linear-parameter varying (LPV) model structure can be considered. Indeed, when thermodynamics or heat transfer laws are studied, it can be shown that important physical parameters such as the heat transfer coefficients, are directly related to specific state vector components such as the inlet mass flow rates. Thus, a model structure taking account of this dependency explicitly could improve the accuracy of the final model. Such a structure is called linear-parameter-varying or LPV. LPV models are linear systems depending on time-varying parameters. These parameters are dependent on time-varying exogenous signals (called the scheduling parameters) that are assumed to be measured in real life. By this way, the model structure is close to the standard LTI one but with a structural flexibility able to picture time-varying, even non-linear behaviors.

Second, a particular attention could be paid to 2-D models. 2-D systems refer to those described by two independent variables, whether they are time/space, space/time, time/time, or space/space. Discrete 2-D systems can be represented by either transfer function models using the 2-D z-transform, or by using state space models. A number of 2-D state space models can be found in the literature, such as those proposed by Roesser, Attasi, Fornasini and Marchesini. It is now well-accepted that the Roesser model can represent most 2-D causal systems of interest. In addition, it has been shown that some partial differential equation models can be formulated as a Roesser state space model. That is the main reason why this PhD program should mainly focus on this specific Roesser model.

The successful PhD candidate will be employed by the University of Poitiers.

France, at the Laboratoire d'Informatique et d'Automatique pour les Systèmes (http://lias-lab.fr/).

The position will be available from October 2012. Most of the PhD program will be realized at the University of Poitiers. However, because the test beds used in this PhD are available at the University of Valenciennes, several stays at the TEMPO laboratory (http://www.univ-valenciennes.fr/TEMPO/) will be required in order to carry out experiments required for the validation of the theoretical developments suggested by the PhD student. As far as 2-D systems are concerned, a collaboration with Prof. J. Ramos from Nova Southeastern University, Fort Lauderdale, Florida, USA, will be required for the good achievement of this task. Prof. J.

Ramos is indeed one of the most productive researchers as far as 2-D system identification is concerned.

Please send your application (and any inquiries) via e-mail to Guillaume Mercère (guillaume.mercere@univ-poitiers.fr). Applications should include a CV (in English or in French) with photo, a letter of motivation (in English and in French), university transcripts (grades record in English or in French), and possibly a list of publications.

Please describe your qualifications with respect to the particular requirements of the project as defined above.

Requirements:

University Masters Degree in a technical field.

Excellent skills in Mathematics and Automatic Control.

Excellent English skills, verbally and in writing.

Good French skills, verbally and in writing.

Ability to cooperate and work with a team of researchers.