A Markov decision process model case for optimal maintenance of serially dependent power system components

Abstract

Purpose – Using a case study for electrical power equipment, the purpose of this paper is to investigate the importance of dependence between series-connected system components in maintenance decisions. Design/methodology/approach – A continuous-time Markov decision model is formulated to find a minimum cost maintenance policy for a circuit breaker as an independent component while considering a downstream transformer as a dependent component. Maintenance of the dependent component is included implicitly in terms of the costs associated with certain state-action pairs. For policy and cost comparisons, a separate model is also formulated that considers only the circuit breaker as the independent component. After uniformizing the continuous-time models to discrete time, standard methods are used to solve for the average-cost-optimal policies of each model. Findings – The optimal maintenance policy and its cost differ significantly depending on whether or not the dependent component is considered. Research limitations/implications – Data used are from manufacturer databases; additional model validation could be conducted if applied to an electric utility asset fleet within their generation, transmission, and/or distribution system. This model and methodology are already being applied in other contexts such as industrial machinery and equipment, jet engines, amusement park rides, etc. Practical implications – The outcome of this model can be utilized by asset and operations managers to make maintenance decisions based on prediction rather than more traditional time- or condition-based maintenance methodologies. This model is being developed for use as a module in a larger maintenance information system, specifically linking condition monitor data from the field to a predictive maintenance model. Similar methods are being applied to other applications outside the electrical equipment case detailed herein. Originality/value – This model provides a structured approach for managers to decide how to best allocate their resources across a network of inter-connected equipment. Work in this area has not fully considered the importance of dependency on systems maintenance, particularly in applications with highly variable repair and replacement costs.

Designing predictive maintenance systems using decision tree-based machine learning techniques

“…Literature review and framework Madu (2000) outlines the importance of businesses and organizations creating a competitive advantage by effective maintenance practices. Others have detailed the importance of using statistical analysis to determine maintenance decisions (Lawrence et al, 1995), and numerous industry applications exist for systems such as repairable and non-repairable systems (Singh Jolly and Jit Singh, 2014;Settanni et al, 2016), jet engines (Settanni et al, 2015), electrical equipment (Bumblauskas et al, 2012(Bumblauskas et al, , 2017aBumblauskas, 2015;Bumblauskas et al, 2015;Chan and Asgarpoor, 2006;Yam et al, 2001), gas distribution equipment (Pievatolo and Ruggeri, 2004), agricultural equipment (Bumblauskas et al, 2015), healthcare (Eker et al, 2012), and building systems (Wong and Li, 2008). Today, post-industrial revolution, we seemingly use terms such as industry 4.0 and industry 5.0 interchangeably to denote the use of cyber-physical systems (CPS), which is well documented (Bumblauskas et al, 2017a;Herterich et al, 2015;Lee et al, 2015aLee et al, , 2015b, industrial CPS (ICPS), safety (Wang et al, 2016), ML, and AI to represent work such as that detailed in this article.…”

Contribution of the three-dimensional model to the reliability allocation of multiphase systems

“…This optimization is carried out either by the cost reduction of dependability attributes (e.g. reliability) or by its increase under the cost constraints (Ray et al, 1980;Elena and Richard, 2013;Bumblauskas, 2015).…”

Applying Forgotten Lessons in Field Reliability Data Analysis to Performance-Based Support Contracts

“…In this sector, engineering asset and operations managers analyze maintenance data from fleets of similar assets to make operating decisions about resource allocation decisions across a network of equipment -see e.g., Bumblauskas (2015). This research can be useful to asset and operations managers as it presents data analytics that are mostly overlooked in the literature such as MCF and power law NHPP, with few exceptions including Bumblauskas et al (2012).…”