NASA Logo

NTRS

NTRS - NASA Technical Reports Server

Back to Results
A function approximation approach to anomaly detection in propulsion system test dataGround test data from propulsion systems such as the Space Shuttle Main Engine (SSME) can be automatically screened for anomalies by a neural network. The neural network screens data after being trained with nominal data only. Given the values of 14 measurements reflecting external influences on the SSME at a given time, the neural network predicts the expected nominal value of a desired engine parameter at that time. We compared the ability of three different function-approximation techniques to perform this nominal value prediction: a novel neural network architecture based on Gaussian bar basis functions, a conventional back propagation neural network, and linear regression. These three techniques were tested with real data from six SSME ground tests containing two anomalies. The basis function network trained more rapidly than back propagation. It yielded nominal predictions with, a tight enough confidence interval to distinguish anomalous deviations from the nominal fluctuations in an engine parameter. Since the function-approximation approach requires nominal training data only, it is capable of detecting unknown classes of anomalies for which training data is not available.
Document ID
19930065674
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Whitehead, Bruce A.
(Tennessee Univ. Tullahoma, United States)
Hoyt, W. A.
(ERC, Inc. Tullahoma, TN, United States)
Date Acquired
August 16, 2013
Publication Date
June 1, 1993
Subject Category
Spacecraft Propulsion And Power
Report/Patent Number
AIAA PAPER 93-1776
Meeting Information
Meeting: AIAA, SAE, ASME, and ASEE, Joint Propulsion Conference and Exhibit
Location: Monterey, CA
Country: United States
Start Date: June 28, 1993
End Date: June 30, 1993
Sponsors: SAE, ASME, ASEE, AIAA
Accession Number
93A49671
Funding Number(s)
CONTRACT_GRANT: NAS3-39184
Distribution Limits
Public
Copyright
Other

Available Downloads

There are no available downloads for this record.
No Preview Available