The ability to detect and diagnose engine faults, failures, or even shutdowns in flight has clear implications for aviation safety. One objective of the work presented here is to investigate mature data-driven techniques for the detection and diagnosis of real-time engine shutdowns in flight. This investigation is enabled by the use of FOQA datasets and heuristics provided by operational experts for the establishment of a baseline for ground truth. IMS and Orca were the algorithms used to identify such scenarios in this study, via the implementation of novel feature selection techniques based upon detection capability and implementation of isolation and localization-based diagnostics.
Future work will focus on capturing a wider variety of simulated faults using data generated by jet engine simulators such as EFS and C-MAPSS. This work will also involve the additional objective of prognostics and the investigation of a broader suite of data-driven algorithms to include those that are novel. We will begin our study by leveraging on previous work from the FOQA dataset by adapting the existing codebase to the EFS dataset. EFS is more suitable for a prognostic studies rather than real-time scenarios due to its reliance on flight cycles as the primary temporal indicator. This will be followed by a re-adaptation of the codebase to C-MAPSS, from which simulated real-time data can be generated and the objectives of anomaly detection and diagnosis can be investigated as well. Thus, anomaly detection, diagnosis, and prognosis can all be investigated with the repository of data that can be generated by these two simulators.
Nothing to see here at the moment. Check back later.