Reinforcement and Deep Reinforcement Learning-based Solutions for Machine Maintenance Planning, Scheduling Policies, and Optimization
​
This research work presents a literature review on the applications of reinforcement and deep reinforcement learning for maintenance planning and optimization.
​
To capture the common ideas without losing touch on the uniqueness of each publication, taxonomies used to categorize the systems were developed, and the reviewed publications were highlighted, classified, and summarized based on these taxonomies. The adopted methodologies, findings and well-defined interpretations of the reviewed studies were summarize using graphical and tabular representations to maximize the utility of the work for both researchers and practitioners.
This work also highlights the research gaps, key insights from the literature and areas of future work.
PROJECT INFO
PRODUCT
A journal paper - Review Article
TIMELINE
3.5 months
ROLE
Author and Lead researcher
Contributions of this work
-
This paper presents a systematic and integrative review, that focuses on methodologies, findings, and well-defined interpretations of the reviewed studies while finding common ideas and concepts, identifying methodological problems, and pointing out areas of research gaps. It also draws insights from existing literature and defines some areas of future work.
-
This paper is structured in a way that it gives new researchers looking to apply reinforcement learning (RL) or deep RL for maintenance planning and optimization a general overview and understanding of the underlying concepts, helping them to see the common, well-explored practices and approaches in the literature..
-
It presents tables and figures that help to make quick deductions and see relationships between defined categories.
-
This paper helps researchers understand the maintenance planning problem and RL and DRL-based solutions. It also references other resources that can help to gain a deeper understanding of the core concepts such as different RL and DRL algorithms.
-
For experts and practitioners, this paper presents a quick glance through literature, maps what has been achieved so far, highlights the problems that have been discussed and the existing solutions, and helps them to deduce new areas of research that can be explored in the future work section.
​
Publication Sections Summary
-
Section 1 introduces the maintenance planning (MP) and scheduling problem. The essence, evolution and previous efforts made in developing optimal maintenance policies. It also highlights the impact of predictive maintenance models on single-unit maintenance planning, the need for RL/DRL for maintenance planning and its suitability to solve the MP problem, it also shows an increasing trend in RL/DRL based applications for maintenance planning and optimization.
-
The RL solution is introduced and discussed in Section 2.
-
Sections 3,4 contain a qualitative review, analysis, comparisons, collection of common ideas, and insights from the literature on the proposed RL and DRL-based maintenance planning solutions. Specifically, Section 3 discusses the maintenance planning problem formulation process. It summarizes and classifies reviewed publications in terms of the factors considered in the problem formulation process. It also contains figures and tables to compare and show relationships between different subsets of data within the domain of this literature review.
-
Section 4 reviews the formulation of the maintenance planning problem into the RL framework, state definitions, reward formulations, and environment. It also contains tables and figures that describe and show relationships between defined entities, and discusses the RL and DRL and hybrid RL and DRL-based algorithms that have been used in literature to solve the optimization problem.
-
Sections 5 and 6 present key insights of the review analysis, implementation details and challenges, areas of future work, and conclusions respectively.
Required Skills
-
Research, Technical Writing, Qualitative research