The present status of maintenance strategies and the impact of maintenance on reliability

TLDR

Although more complex, probabilistic models better describe actual processes and enable optimization for maximal reliability or minimal costs. The paper reviews common power‑system maintenance strategies and argues that only mathematically linked models can quantify their impact on reliability. Maintenance is categorized by replacement versus condition improvement, fixed‑interval versus as‑needed schedules, and heuristic versus deterministic or probabilistic mathematical models. The review finds that fixed‑interval maintenance dominates utility practice, with emerging reliability‑centered approaches gaining traction, yet mathematical model‑based methods remain largely unused.

Abstract

In this paper, the most frequently used power system maintenance strategies are reviewed. Distinction is made between strategies where maintenance consists of replacement by a new (or "good as new") component and where it is represented by a less costly activity resulting in a limited improvement of the component's condition. Methods are also divided into categories where maintenance is performed at fixed intervals and where it is carried out as needed. A further distinction is made between heuristic methods and those based on mathematical models; the models themselves can be deterministic or probabilistic. From a review of present maintenance policies in electric utilities, it is concluded that maintenance at fixed intervals is the most frequently used approach, often augmented by additional corrections. Newer "as needed"-type methods, such as reliability-centered maintenance (RCM), are increasingly considered for application in North America, but methods based on mathematical models are hardly ever used or even considered. Yet only mathematical approaches where component deterioration and condition improvement by maintenance are quantitatively linked can determine the effect of maintenance on reliability. Although more complex, probabilistic models have advantages over deterministic ones: they are capable of describing actual processes more realistically, and also facilitate optimization for maximal reliability or minimal costs.