Assessment of inoperability from unplanned maintenance in building management

Abstract

Purpose Unplanned maintenance (UPM) decisions in building management are often made at the component level without explicitly considering component interdependencies and their role in cascading disruptions. This study aims to develop and validate a data-driven methodology to assess component inoperability and support maintenance prioritization in facility management (FM). Design/methodology/approach A three-phase methodology is proposed using historical work-order data from North American universities. Phase 1 assesses subsystem inoperability from UPM risk profiles and maps probabilities to five ordinal inoperability levels. Phase 2 derives component interdependencies from work-order descriptions using quantitative content analysis and social network analysis (in-strength and out-strength centralities). Phase 3 integrates subsystem inoperability and component centrality in a decision-making algorithm to assign component inoperability levels, followed by AutoML-based classification validation. Findings The methodology was demonstrated on HVAC-related UPM data, including 11,399 work orders (WOs) and 32 components across four subsystems. Distribution systems emerged as the most vulnerable subsystem, and components such as steel dampers, supply fans and air handlers were identified as critical in the interdependency network. The AutoML validation selected an Extra Trees classifier as the best-performing model and achieved 95.47% accuracy on unseen data, supporting the consistency of the assigned component inoperability levels. Originality/value This study provides a structured method that combines subsystem risk assessment, component interdependency analysis and an interpretable decision-making algorithm for component inoperability evaluation. The approach advances FM practice by enabling prioritization of UPM actions based on both subsystem risk and component network role rather than isolated work-order frequency alone.