For a given mechanical equipment, knowing its modular topology has the advantage of facilitating its maintenance. Indeed, during a maintenance problem, we will not act on the whole product except on the failed module (product subsystem) and we would also gain time to detect, diagnose and compensate for the observed failure. On the other hand, the clustering algorithm, which has served as a reference for several works has several limits. It generates much more complex and more expensive modules in terms of coupling costs, which could require more resources, more intervention time and more maintenance work. This has worse consequences for product maintenance, because the more complex the product modules are, the more expensive the maintenance is. We therefore propose an improved clustering algorithm which has the advantage of reducing maintenance costs by reducing the coupling and decoupling costs (Disassembly and reassembly costs) of the modules, generated by the reference algorithm for good maintainability (dis-assemblability). The application is made on a soy roaster. The approach followed in the proposed algorithm consists first of all in defining a DSM (Design Structure Matrix) which will make it possible to define the correction coefficients of the coupling cost, then in formulating an objective function to reduce the coupling costs, and finally to take into account the integrating elements to reduce the size of the modules. The result achieved is the proposal for a modular topology (modular architecture) leading to a significant reduction in maintenance costs. The developed algorithm also allows an economy of scale in reducing the complexity of the modules, promoting good maintainability.

To define the reliability network of a system (machine), we start with a set of components arranged in an appropriate topology (series, parallel, or parallel-series), choose the best terms of the ratio performance / cost, and gather by links with the aim to combine them. This process requires a long time and effort, given the very large number of possible combinations, which becomes tedious for the analyst. For this reason, it is essential to use an appropriate optimization approach when designing any product. However, before trying to optimize, it is necessary to have a reliability assessment method. The objective of this paper is to display a meta-heuristic method, which is sustained on the genetic algorithm (GA) to improve the machines reliability. To achieve this objective, a methodology that consists of presenting the functionalities of genetic algorithms is developed. The result achieved is the proposal of a reliability network for the optimal solution.

A Linear transport problem can be defined as the action of transporting products from "m origins" (or units) to "n destinations" (or customers) at the lowest cost. So the solution to a transportation problem is to organize the transportation in such a way as to minimize its cost. The objective of this paper is to determine the quantity sent from each source (origin) to each destination while minimizing transport costs. Achieving this objective requires a methodology which consists in deploying an algorithm whose purpose is the search for an optimal solution, based on an initial solution. The application is made on a factory producing mechanical parts.