Comprehensive clustering approach for managing maintenance in large fleet of assets

Abstract

The maintenance management of large fleets of assets which include several technical solutions operating in different operational contexts has been a recurrent research topic in the literature. Current approaches to establishing fleet maintenance plans are primarily criticality-based, considering failures consequences and assets reliability; the reliability model is often supported by the idea of pooling data from similar pieces of equipment. In spite of the capability to reduce the population offered by data-pooling, its criteria may still lead to a quite large number of segments. Therefore, it results in an equally large amount of maintenance plans along with their inherent operational and administrative difficulties. It is the purpose of the paper to introduce a novel and comprehensive approach; it integrates statistical methods and clustering algorithms to render a fleet segmentation which allows better customization of maintenance plans involving fewer efforts. The approach is summarized in a decision chart which collects the logic behind the use of every algorithm, tool and technique.