Expert system dedicated to condition-based maintenance based on a knowledge graph approach: Application to an aeronautic system

Abstract

Condition Based Maintenance (CBM) has become the focus of many research topics over the past decades. This is mostly related to the development of new machine learning algorithms and the ever increasing capacity to collect data allowing failures to be detected and the system’s remaining lifetime to be estimated while requiring few or no expert knowledge. However, current machine learning based CBM solutions have limitations. They require an extensive and relevant data set to train on and are performed at the component level, not system-wide. Conversely, Expert Systems (ES) do not have these restrictions but should be used on systems with available expert knowledge and are currently suffering from efficiency, scalability and applicability limits. In this paper, an ES solution for CBM based on an heterogeneous information network will be presented to address the efficiency, scalability and applicability issues of modern ES. An application to an aircraft system will be used as case study to illustrate the process and performance of this solution for anomaly detection and diagnostics.