Retour au site du LAAS-CNRS

Abstract

This paper presents a method for enabling the relational learning or inductive logic programming (ILP) frame- work to deal with quantitative information from experimental data in systems biology. The study of systems biology through ILP aims at improving the understanding of the physiological state of the cell and the interpre- tation of the interactions between metabolites and signaling networks. A logical model of the glycolysis and pentose phosphate pathways of E. Coli is proposed to support our method description. We explain our original approach to building a symbolic model applied to kinetics based on Michaelis-Menten equation, starting with the discretization of the changes in concentration of some of the metabolites over time into relevant levels. We can then use them in our ILP-based model. Logical formulae on concentrations of some metabolites, which could not be measured during the dynamic state, are produced through logical abduction. Finally, as this re- sults in a large number of hypotheses, they are ranked with an expectation maximization algorithm working on binary decision diagrams.

Mots-Clés / Keywords

Mots-Clés / Keywords

Abstract

The paper deals with fault-isolation time problem of the adaptive observers based scheme. The Fault Detection and Isolation (FDI) method is applied to an alcoholic fermentation process using an Estimator based on the approach of the Model Reference (EMR). We propose in this work a new structure of the estimator gains in order to improve the convergence speed of the EMR observer and by the way the time needed to isolate the fault.

Résumé

Dans cet article, nous présentons une approche pour la sélection des paramètres pertinents, basée sur la notion de conflit, dans un processus de détection de pannes. La sélection des paramètres est réalisée en deux étapes : La première consiste à mailler l'espace des paramètres pour déterminer les pavés qui décrivent la zone correspondant au bon fonctionnement du système. La deuxième étape consiste à utiliser les caractéristiques de la zone de bon fonctionnement et des données acquises sur le système en panne pour calculer le conflit entre paramètres et déterminer les plus pertinents. Des résultats expérimentaux sont présentés afin d'illustrer cette approche.

Mots-Clés / Keywords

Abstract

Dynamic models most of time involve differential equations, which are time-local. Such models can also be considered globally, that is in the sense of trajectories in the state space-time. Up to adapted concepts, such a different interpretation reveals itself more flexible, namely because it allows to use various operatorial transformations whose time-local equivalent in general cannot exist and from which can result some remarkable properties. Namely, we introduce a principle of parametrizing for dynamic equations by means of such transformations. We then consider an example of bioreactor model for which we highlight how suitable timenonlocal transformations can sometimes be used to efficiently solve some nonlinear control problems.

Mots-Clés / Keywords