الگوریتمی دقیق برای برنامه ریزی نگهداری پیشگیرانه در سیستم های سری و موازی
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|22539||2009||9 صفحه PDF||سفارش دهید||6760 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Reliability Engineering & System Safety, Volume 94, Issue 10, October 2009, Pages 1517–1525
Reliability is a meaningful parameter in assessing the performance of systems such as chemical processing facilities, power plant, aircrafts, ships, etc. In the literature, reliability optimization is widely considered during the system design phase and it is carried out by an opportune selection of both system components and redundancy. On the other hand, the problem of maintaining a required level of reliability by an opportune maintenance policy has been poorly examined. The paper tackles this problem for a system whose major components can be maintained only during a planned system downtime. An exact algorithm is proposed in order to single out the set of components that must be maintained to guarantee a required reliability level up to the next planned stop with the minimum cost. In order to verify the algorithm effectiveness, it has been applied to a complex real case regarding ship maintenance.
In the last years, the interest of researchers working on the maintenance field has been focused on preventive maintenance policies for multi-component systems. Cho and Parlar  give the following definition about the multi-unit maintenance models: “Multi-component maintenance models are concerned with optimal maintenance policies for a system consisting of several units of machines or many pieces of equipment, which may or may not depend on each other (economically/stochastically/structurally)”. In particular, the economic dependency implies that costs can be saved whenever some components are jointly maintained rather than separately. In fact, if different maintenance actions require a system stop, then a simultaneous intervention of several maintenance crews can significantly reduce the non-operating time. Moreover, the component maintenance often requires a preparatory or set-up work that involves system unavailability costs whenever the system cannot be used during maintenance. Set-up costs can be appropriately saved grouping maintenance actions. Those authors present a wide overview about the multi-unit system maintenance models developed up to 1991 while to date overviews are reported in  and . A classification due to Dekker et al.  regards the planning aspect: stationary or dynamic. In stationary models, a long-term stable situation is considered and an infinite planning horizon is usually assumed. This kind of models provides static rules for maintenance, which do not change over the planning horizon. They generate, for example, long-term maintenance frequencies for groups of activities or control limits for carrying out maintenance, depending on the components’ state. With regard to the dynamic models, short-term information such as an unexpected component deterioration or downtime opportunities can be taken into account. Such models generate dynamic decisions that may change over the planning horizon. The present paper proposes a preventive maintenance policy considering a stationary model; therefore, the literature review will just mention such class of models. Models may differ for the considered objective function, for the eventual constraints and for the resolution technique. Bris et al.  develop availability and cost models for systems with periodically inspected and maintained components. For each system component, the research aims to optimize the maintenance policy, minimizing the cost function and respecting the availability constraint. For solving the problem, the authors propose a genetic algorithm, whose structure includes the first inspection time and durations between two maintenance interventions for each component. The availability assessment method is based on a simulation program. A genetic algorithm approach is also employed by Tsai et al.  to individuate the optimal activities combination by maximizing the unit-cost life at each preventive maintenance stage. Actions that can be performed at each stage are preventive maintenance activities that change the system reliability to a newer state, and preventive replacement that restores the reliability curve to a new one. The preventive maintenance scheduling is stopped when the unit-cost life with maintenance is smaller than the discarded life. Levitin and Lisnianski  generalize a preventive maintenance optimization problem to multi-state systems having a range of performance levels. The possible preventive maintenance actions are characterized by their ability to affect the effective component age. A universal generating function technique and a genetic algorithm are used to solve the problem with respect to system performance by the minimum maintenance cost. In the aforementioned papers, the failure event can also be accepted within a maintenance policy aiming at global cost minimization. If a preventive maintenance is approached, failure can eventually bring to an occasional use of opportunistic policies, but that characterizes non-stationary models. For example, Wildeman et al.  deal with the problem of grouping the maintenance activities in order to save set-up costs, since the execution of a group of maintenance activities requires a single set-up. The authors propose a rolling horizon approach that takes the long-term tentative plan as the basis for the next adaptation according to the short-term information, as the component failure. This yields a dynamic grouping policy, which assists the maintenance manager during the planning job. After all, the most used measure in evaluating the production system performance essentially is its stationary availability, also defined as the expected percentage of time in which the system is working. Nevertheless, for some continuous operating systems (chemical processing facilities, power plants, aircrafts, ships, etc.) the failure event can be dangerous, too expensive or even disastrous. For these reasons, a high level of reliability is required. Reliability expresses the probability that the system operates without failure for a fixed period of time under some stated conditions. So, for these systems the reliability constitutes another meaningful parameter for performance appraisal. Cassady et al.  tackle the problem of singling out the set of elements on which to operate during a planned downtime between two missions aiming at maximizing system reliability for the next mission. Maintenance activities need to be completed within a stated time and a fixed cost. This decision-making process is referred to as “selective maintenance”. The problem is formulated by a mathematical programming model and two numerical examples with 10 and 12 components being, respectively, reported. The cost or time minimization with a reliability constraint is also considered. An extension of the previous model is proposed by the same authors . The components’ failure time is assumed to be a Weibull distribution while the decision-maker can adopt multiple maintenance actions: minimal repair on failed components, failed components’ replacement and functioning components’ replacement (preventive maintenance). Starting from a simple system, a simulation model is used to calculate the mission reliability. Rajagopalan and Cassady  approach the same problem of reliability maximization for a system constituted by a series arrangement of subsystems, each one containing a set of identical elements arranged in parallel, considering a constraint on maintenance time. All elements have a constant failure rate and therefore maintenance action reduces to the replacement of some failed elements. The decision variable is the number of failed elements to be replaced for each subsystem. The problem is formulated as a nonlinear knapsack problem and four improvements are proposed to speed up the total enumeration approach originally proposed by Rice et al. . The present paper is interested in solving maintenance problems by proposing a new exact algorithm capable of quickly solving big-size problems of different kinds. It starts from the Kettele  algorithm. In order to illustrate how it works and the relative efficiency, the following selective maintenance problem is tackled: given a system whose components can be maintained during a fixed planned period, it is aimed to individuate the set of components to be maintained so as a required reliability level is warranted up to the next stop with the minimum cost. The paper is organized as follows. The problem is mathematically formulated in the next section. In Section 3 the original Kettele's algorithm for the redundancy optimization of series systems and the new proposal for the series systems maintenance optimization are reported. Subsequently (Section 4), it is shown how the algorithm can be extended to series–parallel systems. Section 5 presents a case study solved by the proposed algorithm and final considerations about its performance. Conclusions and ideas for further developments conclude the paper.
نتیجه گیری انگلیسی
Maintaining a system at a given level of reliability by a convenient maintenance policy is a problem not frequently discussed in the literature. More often, the reliability is taken into account in system design, to suitably select system components and to appropriately use the redundancy tool. Usually, in the literature, the more tackled problem is the maximization of system availability with a minimum cost. Nevertheless, several systems are required to operate with a low failure probability. On these systems, maintenance concerning the major components can be often executed only during a priori fixed time windows for the system stops. The present paper starts from the foregoing considerations. At the increasing system complexity, the individuation of the elements on which it is opportune to quickly act becomes a problem that may not be faced by exhaustive enumeration or mathematical programming. Here, an exact algorithm has been proposed that starts from Kettele's algorithm for the redundancy optimization of a series system. The algorithm, adapted to the maintenance problem, has been extended to deal with series–parallel systems and operators have been set up to obtain a progressive drastic reduction of the solution space inside the optimal solution is sought. A complex real case has been solved by the algorithm in an extremely low time, thus proving its efficiency. Future developments could consider the use of a more complex cost model, the adaptation of the algorithm to the dual problem, that is the reliability maximization with a given budget, and the use of the algorithm to single out the non-dominated solutions in multi-objective maintenance problems.