بهینه سازی تعمیر و نگهداری پیشگیرانه از طریق مدل شبیه سازی ترکیبی تعمیر و نگهداری ــ تولید

Maintenance problems are crucial aspect of nowadays industrial problems. However, the quest of the efficient periodicity of maintenance for all components of a system is far from an easy task to accomplish when considering all the antagonistic criteria of the maintenance and production views of a production system. Thus, the objective is to simultaneously ensure a low frequency of failures by an efficient periodic preventive maintenance and minimize the unavailability of the system due to preventive maintenance. This implies a minimum impact on the production. In this paper, several tools are combined to collaborate in order to optimize multi-component preventive maintenance problems. The structure of the maintenance-production system is modeled thanks to a framework inspired by our previous research projects. The dynamic aspects are modeled by a combination of timed petri-nets and PDEVS models and implemented in our VLE simulator. The parameters of the resulting simulation model are optimized via a Nelder–Mead (Simplex) Method.

The present economical context requires from companies that they practice an optimal exploitation of their production tools. In this purpose, every decision maker is asked to assure a maximum availability of these production tools at minimal cost (Percy and Kobbacy, 2000). The optimization consists in determining the best “parameters combination” which provides the best values of the technical and economical criteria (see for instance Rezg et al., 2005 and Boschian et al., 2009). However, in most cases, it appears to be very difficult to use analytical approaches without formulating restrictive hypotheses. In order to evaluate these performance criteria, simulation is the best adapted solution. In this paper, we suggest an approach integrating optimization and simulation. This approach consists in generating more and more efficient solutions with an optimization tool and to evaluate them via a simulation model until a halt criterion is satisfied. This approach has already been studied in the literature (see for instance Boschian et al., 2009 and Riane et al., 2009). This integration is illustrated in Fig. 1. In the following sections, according to Talbi (2002), this assembly of different units, at different levels of combinations is called a “hybrid model”. Our work aims to provide a framework to facilitate the optimization of production and maintenance through simulation. This paper focuses on the simulation aspect. We want to develop a generic modeling tool for simulation, easy to understand by decision makers. The objective is to facilitate the creation of simulation models by the use of constructs (elementary components). The remainder of this paper is organized as follows. The second section presents the maintenance problem; the third section introduces the simulation paradigms, formalisms and tools that constitute the bases of our framework; the fourth section depicts our modeling component; the fifth section describes an application of our optimization–simulation hybrid model. Finally several conclusions and perspectives are given.

Thanks to our VLE simulator, we have presented in this paper an hybrid method composed of the Nelder–Mead algorithm hybridized with a simulation multimodel. This multimodel is decomposed into several models implemented in the VLE simulator. This implementation is largely simplified by the extensions of the VLE simulator which provides several skeletons (similar to design patterns or constructs in other modeling tools/languages) to guide the implementation of the models. Our results have been experimentally validated by comparisons to two previously published approaches when disabling the production. However, contrary to these approaches, our model is able to integrate production aspects and to simulate sophisticated scheduling and maintenances strategies thanks to the integration of several extensions in our VLE simulator such as “Decision rules”. All possibilities of the simulation model are not used in this paper. Our next objective is to provide a new framework to optimize the combined scheduling of production and maintenance. Short term work will consist of the integration of more efficient maintenance strategies as well as sophisticated schedulers. We are also working on the building blocs (constructs) of our multimodels that will provide a complete GUI, easy to understand by decision makers.