Home > Research > Publications & Outputs > Automated data inspection in jet engines

Electronic data

  • 2020Harjitphd

    Final published version, 5.62 MB, PDF document

    Available under license: CC BY-NC-ND: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Text available via DOI:

View graph of relations

Automated data inspection in jet engines

Research output: ThesisDoctoral Thesis

Published
Publication date2020
Number of pages217
QualificationPhD
Awarding Institution
Supervisors/Advisors
Publisher
  • Lancaster University
<mark>Original language</mark>English

Abstract

Rolls Royce accumulate a large amount of sensor data throughout the testing and
deployment of their engines. The availability of this rich source of data offers exciting opportunities to automate the monitoring and testing of the engines. In this thesis we have developed statistical models to make meaningful insights from engine test data.

We have built a classification model to identify different types of engine running
in Pass-Off tests. The labels can be used for post-analysis and highlight problematic engine tests. The model has been applied to two different types of engines, in which it gives close to perfect classification accuracy. We have also created an unsupervised approach when there are no defined classes of engine running. These models have been incorporated into Rolls Royce systems.

Early warnings for potential issues can enable relatively cheap maintenance to
be performed and reduce the risk of irreparable engine damage. We have therefore developed an outlier detection model to identify abnormal temperature behaviour. The capabilities of the model are shown theoretically and tested on experimental and real data.

Lastly, in a test decisions are made by engineers to ensure the engine complies
with certain standards. To support the engineers we have developed a predictive
model to identify segments of the engine test that should be retested. The model is tested against the current decision making of the engineers, and gives good predictive performance. The model highlights the possibility of automating the decision making process within a test.