Aide a la prise de decision en temps reel dans un contexte de production adaptative.

摘要

The market dynamics have evolved and manufacturing facilities have followed this trend to stay competitive. The classic factory has been defined by its manufactured goods. The value of these goods has been measured primarily by their material components. But under the market pressure and its increasing dynamism, factories wishing to stay competitive are becoming modern service centers. It has resulted in management problems for which many companies are not yet prepared. Today, the economic efficiency of value creation is not a property of the products but rather of the process. It means the decisive potentials of companies are to be found not so much in their production capability but in their process capability. Indeed, increasing globalization is necessarily leading towards more anonymous products out of long supply chains. Any enterprise wishing to stand out from the competition in the future needs a strategy which offers the customer an additional added value, such as, for example, high flexibility, short delivery times, high delivery reliability, and wide range of variants. These properties are created by the processes. The requirement for process capability gives rise in turn to the requirement that all value-adding processes be geared to the process result and thus to the customer. A necessary condition of process transparency is the ability to map the company's value stream in real time. Processes must be able to adapt to environmental changing conditions, react to unforeseen events and to solve these difficulties by collaborating. Under these conditions they can be called adaptive processes.;The works developed during this thesis are based on four published, accepted or submitted papers to specialized papers: (1) "Manufacturing execution system -- a literature review", published in Production Planning & Control, Taylor & Francis. This paper is a MES literature review and demonstrates that decision-making support is still missing in traditional ERP-MES couple. (2) "Engineering Change Order Processing in ERP Systems: An Integrated Reactive Model", accepted by European Journal of Industrial Engineering, Inderscience Publisher. This paper proposes a generic manufacturing execution platform integrating ERP and MES to a real-time decision-making nucleus combining a genetic algorithm and a simulator. An Engineering Change Order processing scenario, based on a real scenario in the aerospace industry, is used to validate the proposed platform is validated applied on a case study of a given order of technical change in the aviation industry. (3) "Development and integration of a reactive real-time decision support system in the aluminum industry", published in Engineering Applications of Artificial Intelligence, Elsevier. This paper proposes a more flexible and robust of our manufacturing execution platform by adding an integration server. The platform is evaluated on a case study of alloy production scheduling in the aluminum industry. (4) "An Improved Genetic Algorithm Approach for On-line Optimization Problems", submitted to Production Planning & Control, Taylor & Francis. This paper aims at enhancing standard genetic algorithm to reduce the execution time and improve the solution quality when the objective function is expensive to evaluate. More precisely, we propose to add a cache memory to avoid re-testing an already tested solution and a predator mechanism to relax evaluation. These mechanisms reduce the run time and improve the solution quality.;This thesis focuses on scheduling process in manufacturing environments. The main objective is to implement real time decision-making support components as well as feedback loop mechanisms integrating optimization and simulation techniques in ERP and MES applications allowing connecting the shop floor to the rest of the enterprise. The proposed platform responds in real time to various events occurring on the shop floor and may be extended beyond the scheduling issue.