Article Preview
Top1. Introduction
Nowadays in industry, the development of future generations of manufacturing systems is not a trivial activity since several new constraints and rules should be satisfied in addition to their productions that should not be stopped but kept optimal as much as possible according to user requirements (Khalgui, 2008; Li, Liu, Hanisch, & Zhou, 2012). A production line of a system is classically composed of a set of machines allowing productions of jobs. Each job is performed by a subset of machines to apply well-defined operations. We characterize each job by a production cycle which is a used parameter to define the whole production system’s performance. This parameter defines in fact the productivity rate of such a job and should be kept optimal as much as possible. Each operation to be applied by a machine on a job is characterized by a worst case release time. The new generations of manufacturing systems address new criterion such as flexibility and agility (Khalgui, Mosbahi, & Zhiwu Li, 2011). In fact, these systems should be changed and adapted when hardware as well as software faults occur, or when new business requirements impose automatic run-time reconfigurations to be applied (Shahdi & Sirouspour, 2012; Jiménezj & Zhong, 2012; Habibi & Novinzadeh, 2012). We define a reconfiguration scenario in the current study as any operation allowing (i) additions-removals of jobs and/or machines to/from production lines under constraints, (ii) modifications of production cycles of jobs, and (iii) modifications of release times of operations to be applied by machines on jobs. Two policies exist today to reconfigure manufacturing systems: static reconfigurations to be applied off-line before the whole system’s cold start, and dynamic reconfigurations to be applied at run-time. Two solutions can be followed in the second policy: manual reconfigurations to be dynamically applied by users, and automatic reconfigurations that should be handled by intelligent software agents which control the system’s adaptation (Khalgui & Hanisch, 2011). We are interested in the current paper in automatic reconfigurations of manufacturing systems to add-remove new jobs and/or machines to-from production lines. We assume three constraints to be satisfied before and after any reconfiguration scenario: the first deals with the performance of the system that should be kept optimal, the second constraint suppose that each scenario should not increase the energy consumption by adding machines or jobs. The third new ecologic constraint supposes that the quantity of CO2 emissions should not be increased after any scenario. These constraints are very important and requested in industry since important questions dealing with energy and CO2 are considered. To meet all these requirements, we propose an agent-based architecture where an intelligent software agent is deployed to evaluate the whole architecture after any automatic reconfiguration scenario in order to verify the assumed constraints. In the case of violation, the agent proposes three technical solutions: the first is to modify release times of operations, the second is to modify the system’s performance by reconfiguring production cycles of different jobs, and the third is to remove some jobs or machines. These solutions are non-binding if users can dynamically judge their limitations at run-time according to production strategies. In this case, they should provide their own manual solutions. We assume in the current paper two examples: the first is a classic formal drilling platform similar to FESTO benchmark production system that we studied several times before in Khalgui and Thramboulidis (2008) and Khalgui (2010) and the second is the footwear production system of the research institution ITIA-CNR in Italy (Khalgui & Carpanzano, 2008), which will be assumed as a real industrial case study. The next section describes related works to prove the current paper’s originality. Section 3 proposes the Manufacturing Model to be assumed in this paper. We present the first case study to be analyzed in Section 4, and propose the intelligent agent-based architecture for ecologic systems in Section 5. The Algorithm encoding intelligent agents is described in Section 6, and the footwear factory is analyzed in Section 7 to apply the proposed contributions. We conclude the paper in Section 8.