redundancy allocation problem

In the present study, the redundancy allocation problem (RAP) of series-parallel system has been investigated to maximize the system's availability. To achieve the research objective, budget, weight and volume constraints, and the maximum and minimum number of elements assigned to each subsystem have been considered. The main innovation of this research is to consider the failure and repair rates of components with non-exponential distribution function in the process of optimization. The parameters affecting the under-study system in this paper make it impossible to calculate the availability using mathematical relations. Therefore, the present study has used simulation method to calculate system availability. Since the simulation has no optimization capability, this research tries to represent the results of the simulation as a mathematical function, which explains the way decision variables affect the system's availability. Further, due to the high degree of difficulty of developed mathematical function, the genetic metaheuristic algorithm was used to solve it. Finally, the efficiency of the genetic algorithm was measured against particle swarm algorithm and simulated annealing algorithm. To compare fairly, the parameters affecting the algorithms are adjusted using the Taguchi method and the algorithms are in their best practice. The computational results prove the high ability of the genetic algorithm in optimizing the concerned problem.

This research was carried out with the aim of simultaneously examining the two categories of the most commonly encountered issues in the field of production and operations including the redundancy allocation and the buffers allocation. The study sought to optimize goals such as accessibility, system costs, and buffer capacity and for this purpose, variables such as the optimum capacity of buffers between machines, the number of high-reliability machines and their allocation, and the appropriate time schedule for maintenance and repair were investigated.

Considering the categorization of emergency and preventive failures for machinery, taking into account the cost of any failure for machinery, and considering the non-exponential and increasing distribution function for a variety of failures, it is very difficult to obtain and calculate mathematical functions related to the objectives of Availability and Cost explicitly. Therefore, a combination of simulation, experimental design, and neural network approach was used to estimate these two objective functions. In order to solve the proposed model, the NSGA-II algorithm was coded in MATLAB. Also, in order to analyze the efficiency of the suggested Algorithm, the MOPSO Algorithm was used and the Algorithms were compared with each other based on the performance measures of the algorithms.

After applying the numerical example with the approach used, the results of the research indicate the validity of the proposed methodology for the problem under study.

Based on the set of solutions obtained from the algorithms used, different combinations of variables (including the number of machines per station, buffer capacity and duration of repairs) can be used to achieve the appropriate level of objectives.