A Model-Based Anomaly Detection Approach for Analyzing Streaming Aircraft Engine Measurement Data

Simon, Donald L.; Rinehart, Aidan W.

This paper presents a model-based anomaly detection architecture designed for analyzing streaming transient aircraft engine measurement data. The technique calculates and monitors residuals between sensed engine outputs and model predicted outputs for anomaly detection purposes. Pivotal to the performance of this technique is the ability to construct a model that accurately reflects the nominal operating performance of the engine. The dynamic model applied in the architecture is a piecewise linear design comprising steady-state trim points and dynamic state space matrices. A simple curve-fitting technique for updating the model trim point information based on steadystate information extracted from available nominal engine measurement data is presented. Results from the application of the model-based approach for processing actual engine test data are shown. These include both nominal fault-free test case data and seeded fault test case data. The results indicate that the updates applied to improve the model trim point information also improve anomaly detection performance. Recommendations for follow-on enhancements to the technique are also presented and discussed.

Document ID

20140010036

Acquisition Source

Glenn Research Center

Document Type

Conference Paper

Authors

Date Acquired

July 23, 2014

Publication Date

June 16, 2014

Subject Category

Report/Patent Number

Meeting Information

Meeting: 2014 ASME Turbo Expo

Location: Dusseldorf

Country: Germany

Start Date: June 16, 2014

End Date: June 20, 2014

Sponsors: American Society of Mechanical Engineers

Funding Number(s)

Distribution Limits

Public

Public Use Permitted.

Keywords

Available Downloads

Name

Type

20140010036.pdf

STI

No Preview Available

NTRS

NTRS - NASA Technical Reports Server

Available Downloads

Related Records