روش های جدید برای به حداقل رساندن هزینه تعمیر و نگهداری پیشگیرانه سیستم های سری موازی با استفاده از بهینه سازی کلونی مورچه ها
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|22379||2005||9 صفحه PDF||سفارش دهید||5305 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Reliability Engineering & System Safety, Volume 89, Issue 3, September 2005, Pages 346–354
This article is based on a previous study made by Bris, Châtelet and Yalaoui [Bris R, Chatelet E, Yalaoui F. New method to minimise the preventive maintenance cost of series–parallel systems. Reliab Eng Syst Saf 2003;82:247–55]. They use genetic algorithm to minimize preventive maintenance cost problem for the series–parallel systems. We propose to improve their results developing a new method based on another technique, the Ant Colony Optimization (ACO). The resolution consists in determining the solution vector of system component inspection periods, TP. Those calculations were applied within the programming tool Matlab. Thus, highly interesting results and improvements of previous studies were obtained.
The availability and reliability are good evaluations of a system performance. Their values depend on the system structure as well as on the component availability and reliability. These values decrease as the component ages increase, i.e. their serving times are influenced by their interactions with each others, applied maintenance policy and their environments. The approach proposed in this article considers mainly the maintenance policy influence. The other influences can be taken into account with the developed method (see Section 3.5). Among the different types of maintenance policy, we suggest to study the preventive maintenance (PM), widely applied in large systems such as transport systems, production systems, etc. Preventive maintenance consists of a set of technical, administrative and management actions to decrease the component ages in order to improve the availability (and the reliability) of a system (reduction of probability failure or the degradation level of a system's component). These actions can be characterized by their effects on the component age: the component becomes ‘as good as new’, the component age is reduced, or the state of the component is lightly affected only to ensure its necessary operating conditions, the component remaining appears to be ‘as bad as old’. PM policy has been the subject of many studies in recent years. Those studies take into consideration several criterions like cost, economic life, risk or a combination of the above stated criteria. Ushakov  tried to optimize preliminary planned PM considering binary-state system reliability. Barrow and Hunter  demonstrated that the cost of scheduled replacement is lower than the unscheduled one. Legat et al.  determined the optimal interval for preventive maintenance/replacement using either an age-based or a diagnostic based renewal strategy. Wang et al.  represented an algorithm of decision making about replacement scheduling of a system's main component by pursuing a maximal system profit. Bris et al.  tried to optimize for each component of a system, the maintenance policy minimizing the cost function, with respect to the availability constraint (A(t)≥A0, A0 being a fixed lower limit) on a given mission time TM. They have calculated T0 and Tp, with T0 being the first inspection time vector and TP, the vector of system component inspection periods. T0 was generated based on the time dependent Birnbaum importance factor (definition see in , for example). Genetic algorithm (GA)  was used to calculate this last vector. GA are used more often to optimize system reliability , , , ,  and  and in maintenance strategy. A genetic algorithm was used in Levitin and Lisnianski  to calculate replacement policies leading to minimal cost plan of PM actions during multi-state system lifetime. Tsai et al.  used GA to decide optimal activities-combination which maximizes system unit cost life. In this paper, we propose a new meta-heuristic to solve the problem of TP calculation treated in . Algorithms based on ant colony optimization (ACO) replace the GA in their search for best calculation of TP. A comparison is established between results obtained by those algorithms via cost and time evaluation. Ant Colony Optimization (ACO) is a population-based general search technique for the solution of difficult combinatorial problems. This method is inspired by the pheromone trail laying behavior of real ant colonies. Ant Colony Optimization takes elements from real ant behavior to solve more complex problems than those solved by real ants , ,  and . In ACO, artificial ants are stochastic solution of construction procedures. They build probabilistically solutions by taking into account (artificial) pheromone trails which change dynamically at run time to reflect the agents acquired search experience. Solution construction is biased by pheromone trails which change at run-time, heuristic information on the problem instance and the ants private memory . ACO was initially used for the resolution of traveling salesman problem TSP . Lately, ACO algorithms have become interesting for problems which cannot be easily solved by classical techniques. Gravel et al.  have demonstrated their performance by finding solution for real industrial scheduling problem. They were applied to problems with a very short path where costs change dynamically as for example in telecommunication networks routing problems. Liang et al.  used an ant colony meta-heuristic optimization method to solve the redundancy allocation problem (RAP), a comparison between GA and ACO performance was established. And recently, Nahas et al.  used ant system to optimize the reliability of a series system with multiple choices and budget constraints. Simulations, in , have shown that the proposed ant system is efficient with respect to the quality of solutions and the computing time. In this proposed research, ACO are used as a technique to optimize the determination of preventive maintenance periods leading to maintenance cost minimization of series–parallel systems. Once applied in , this technique gives interesting results. This introduction is followed by four sections which present successively detailed explanations of the ACO algorithms (course of the solution, stop and evaluation criteria choice of the starting point, transition and updating rules), results and a conclusion.
نتیجه گیری انگلیسی
In this paper, we try to elaborate a new effective technique to assess quasi-optimal maintenance periods. Three algorithms have replaced genetic algorithms in their search with good results. Subsequently, we point out that ant colony algorithms were very performing compared to genetic algorithms at least for the studied system under the same conditions as in . Moreover, obtained results encourage us to study larger and more complex systems using ACO (other reliability properties, system structures, cost functions). This conclusion is reinforced by the fact that similar effects have been obtained by  which established a comparison between GA and ACO performances. Metaheuristics are non typical, thus the critical issue in ACO algorithms (it is the case in most of metaheuristic algorithms) is the ‘proper’ parameters' determination which must be fixed before running those algorithms. A ‘proper’ determination of those leads to a reasonable solution. A more precise determination of those parameters could be the subject of further studies. Also, discretizing the range of possible values instead of randomly generating times in interval (LB,UB) according to a uniform distribution seems to be interesting to test. In this paper we establish a comparison between a classic genetic algorithm and the proposed ant colony algorithms. It seems interesting to establish this same comparison between a more evolved genetic algorithm e.g hybrid GAs not only with ant colony algorithms but with local research algorithms too. Finaly, an hybrid algorithm formed by a genetic algorithm and ant colony algorithms would be interesting to test. In this case, genetic algorithms should determine colony's initial parameters, the latter seeks intervention dates to make out the preventive maintenance or vice versa.