Enhancing operational fault diagnosis by assessing multiple operational modes

Operating a modern technical system, such as a train or aircraft, calls for good organised engineering, operation and maintenance to keep the system in an optimal operational condition. Predictive maintenance is being studied and has as aim to identify errors early enough to still be able to propose a suitable solution before a real incident occurs. After all, technical problems in service may lead to delays or even interruptions of service due to extensive repair actions, such as the replacement of components. Often, predictive maintenance aims at recognising patterns in time series of monitored data and classifying these patterns as known conditions (faulty or correct). As such it provides a vital source of information for maintaining a healthy operational status. However, these approaches are still in their early phases and rely still heavily on skill and experience from the expert. In this paper, the use of self-organising maps for predictive maintenance is being discussed, applied to data of a jet engine. The aim of the study was to assess the usability of such approaches to real-life situations, assessing the learning and validation phases.