Résumé -Développement de corrélations pour la prédiction des propriétés des fractions pétrolières par les algorithmes génétiques -Notre étude concerne la caractérisation des fractions pétrolières dont les propriétés thermodynamiques et physiques peuvent seulement être connues par une expérimentation lourde et coûteuse due à la multiplicité de leurs constituants. Après une introduction des éléments et des nouvelles tendances dans l'utilisation des techniques d'intelligence artificielle, cet article prouve que les algorithmes génétiques peuvent être appliqués à ce domaine du pétrole. Par conséquent, nous proposons une approche empirique pour estimer les propriétés critiques et le facteur acentrique des fractions pétrolières, basée sur leurs points d'ébullition et densité facilement accessibles. Les algorithmes génétiques nous fournissent aussi une forme appropriée de fonction pour la prédiction de ces propriétés. Des résultats très prometteurs sont obtenus et plusieurs perspectives méritant d'autres investigations sont soulignées. Abstract -Developing Correlations for Prediction of Petroleum Fraction Properties Using GeneticAlgorithms -This paper deals with the characterization of petroleum fractions whose thermo-physical behaviours can only be known through expensive measurement efforts due to the multiplicity of their constituents. After introducing the issue and new trends in the use of artificial intelligence techniques, this paper shows how genetic algorithms can be applied to this field. Hence, we propose an empirical approach for estimating petroleum fractions critical properties and acentric factor based on their boiling point and density that can be easily obtained. Genetic algorithms provide us with a proper function form for the prediction. Moreover, very promising results are obtained and several relevant issues that deserve further investigations are emphasized.

No abstract

Purpose The development of context-aware applications in ubiquitous environments depends not only on the user interactions but also on several context parameters. The handling of these parameters is a fundamental problem in these systems. The key purpose of this work is to enrich the unified modeling language (UML) class diagram with new constructs to provide a universal model capable of coping with the context-awareness concerns. Design/methodology/approach The authors provide a review of existing context handling approaches. Afterward, they relied on the UML extensibility mechanisms to propose a heavyweight extension for the UML class diagram. This generic approach allows describing the different context parameters since the modeling phase. Findings Existing solutions for context handling apply the contextual constraints on finished applications or tend to be dependent on a specific development process. This paper presents a solution based on UML, which allows dealing with context since the modeling phase, and independently of development processes. This proposal is implemented as an eclipse editor and illustrated through a case study in the healthcare field. Originality/value This paper addresses the problem of context handling, and it presents a review of the foremost existing solutions. The paper also presents a heavyweight extension for the UML class diagram, which consists in enriching it with additional constructs, capable of monitoring how applications are linked to context parameters and how the values of these parameters may affect the application behavior.

No abstract