مقاله انگلیسی رایگان در مورد استراتژی تعمیر و نگهداری پویا – الزویر 2020

Maintenance grouping methods such as the rolling horizon approach are effective in reducing maintenance costs of multi-component systems. Despite the theoretical advancements of this approach, it still faces three challenges. First, the extensively adopted minimal repair assumption upon failures limits its application. Second, opportunistic maintenance upon corrective maintenance is overlooked, unable to fully take advantage of economic dependence. Third, maintenance plans are not based on actual maintenance history and health information, which may increase failure risks. To address these challenges, this paper formulates a novel dynamic planning framework that captures economic dependence in both preventive and opportunistic replacement. Unlike conventional approaches that restrict all maintenance activities into a finite planning horizon, our proposal focuses on activity-to-activity scheduling without specifying the horizon. As such, the subsequent maintenance schedule is dynamically updated once a system maintenance is executed. A flexible dynamic programming algorithm is developed to optimize the maintenance grouping, and the strategy framework is further extended to condition-based maintenance scenarios. The effectiveness and generality of the proposed maintenance strategy are demonstrated by numerical experiments.

Introduction

Diverse industrial systems, such as smart grids, wind farms and high-speed trains are subject to multiple interdependencies existing among components or sub-systems [1]. Typically, there are three categories of dependencies, i.e. economic dependence [2], stochastic dependence [3], and structural dependence [4, 5]. Among them, the economic dependence attracts the most notable attention due to its significant impact on system operations & maintenance costs [6]. Such dependence allows to share set-up and downtime costs when multiple components are maintained simultaneously, so that maintenance resources can be significantly harnessed [7]. Group maintenance [2, 8-10] and opportunistic maintenance (OM) [11-14] are two representative maintenance policies taking advantage of the economic dependence. The former specifies a pre-determined schedule for inspections or preventive maintenance (PM), while the latter provides PM opportunities for other components when a component undergoes preventive or corrective maintenance (CM). Notably, OM of multi-component systems is generally scheduled based on operational age and/or the reliability level of components. This triggers tremendous operational states and brings difficulties for the analytical modelling [11, 15, 16]. Consequently, many OM policies are optimized via simulations [11, 17, 18], which is trivial and time-consuming. In this regard, recently a few researches employed (deep) reinforcement learning (RL) methods to address this problem [19-22].