ABSTRACT
Maintenance resource prediction is difficult because it is influenced by a multitude of factors such as the rate of system deterioration, operational effectiveness, and the efficiency of the asset management program. Work orders (WO) can provide useful insights to account for such factors, but they come in unstructured non-standardized forms that is difficult to analyze. This paper proposes an approach that leverages semantic networks for WO text analytics to support predictive maintenance. The proposed system pulls the basic facility attributes from BIM and historical work orders. This is primarily done by transferring the text of WOs into semantic networks. Using network-based machine learning, clusters of WO typical concepts are extracted. The clusters are categorized to act as prototypes for future WOs. Along with BIM features, these clusters help match the scope of work to previous similar WOs, which are used to estimate the required resources for a new WO.
