Clustering of maintenance work data for failure mode discrimination

Abstract

A fast and efficient method to discriminate failure modes from maintenance work orders will facilitate and motivate proactive maintenance development. This paper aims to propose a faster and as efficient clustering methodology that differs from previous text mining attempts. Text mining attempts are very dependent on correctly classifying text but the method proposed here is text independent. It is based on time to repair (TTR), time before failure (TBF) and other available identifiers. Using K-means as the clustering algorithm, the processing speed was greatly reduced. Singularity of discriminated failure modes were as good as previous text mining attempts.