For a dual redundant-control system, which is typical for short-haul aircraft, if a failure is detected in a control sensor, the engine control is transferred to a safety mode and an advisory is issued for immediate maintenance action to replace the failed sensor. The safety mode typically results in severely degraded engine performance. The goal of the High Reliability Engine Control (HREC) program was to demonstrate that the neural-network-based sensor validation technology can safely operate an engine by using the nominal closed-loop control during and after sensor failures. With this technology, engine performance could be maintained, and the sensor could be replaced as a conveniently scheduled maintenance action.
The neural network architecture used here for the sensor validation is the Auto-Associative Neural Network (AANN). This feed-forward network architecture has output data that reproduce the network input data (see the following figure). AANN consists of two layers, the mapping layer and the demapping layer, which are interconnected through the bottleneck layer. The mapping layer compresses the data into a reduced-order representation, eliminating redundancies and extracting the key features (principal components) in the data. Reducing the number of dimensions is the key characteristic of this architecture. The demapping layer recovers the encoded information from the principal components. In addition, a fault detection logic identifies the unique sensor failure based on the pattern of the error vector of the input and output of the AANN.
Auto-associative neural network schematic (7-10-4-10-7 network) for turbofan engine estimates.
|
Variable |
Engine parameter |
|
DPBLD |
Delta pressure between compressor discharge and bypass duct |
|
P25 |
Low-pressure compressor inlet pressure |
|
P3 |
Compressor discharge pressure, burner inlet |
|
T3 |
Temperature at burner inlet |
|
WF |
Fuel flow |
|
XNL |
Fan rotor speed |
|
XNH |
Core rotor speed |
In March 1998 the High Reliability Engine Control program successfully demonstrated the neural-network-based sensor failure detection and accommodation algorithm on a real-time simulation of the Allison's AE3007 engine and its Full Authority Digital Engine Controller (FADEC, ref. 1). Successful accommodation of faults for low- and high-speed rotor speed sensors was demonstrated in the closed-loop engine simulation (see the following flow diagram).
Closed-loop control with sensor validation.
|
Variable |
Engine parameter |
|
ALT |
Aircraft altitude |
|
CVGMA |
Compressor variable geometry command |
|
CVGFB |
Compressor variable geometry feedback |
|
ITT |
Interstage turbine temperature |
|
MMVFB |
Main metering valve feedback signal |
|
PLA |
Power lever angle (thrust command) |
|
P2 |
Fan inlet pressure |
|
T2 |
Fan inlet temperature |
|
T2P5 |
Compressor inlet temperature |
|
WFMA |
Main metering valve command |
|
XM |
Aircraft Mach number |
|
XNL |
Fan rotor speed |
|
XNH |
Core rotor speed |
Simulation results show that the neural-network-based sensor validation algorithm can identify both hard and soft sensor failures and can estimate how long the engine control will operate without degradation (ref. 2).
The High Reliability Engine Control program is a collaborative effort between NASA, the Allison Engine Company, and Scientific Monitoring Inc., a small business. In this program, NASA provided the sensor failure detection technology which included the neural network modeling, training, and detection algorithm. Allison Engine Company provided the expertise in AE3007 engine model and control, and Scientific Monitoring Inc. did the system integration and implementation. This program gave these research collaborators an opportunity to further develop and transfer NASA-developed advanced technology for sensor failure detection, isolation, and accommodation to aircraft engine industry.
Lewis contact: Dr. Ten-Huei Guo, (216) 433-3734, Ten-Huei.Guo@grc.nasa.gov
Author: Dr. Ten-Huei Guo
Headquarters program office: OAT
Program/Project: HREC
[Table of Contents] [Index of Authors and Contacts] [Glenn Research & Technology Reports for Other Years] About the name change from Lewis to Glenn] [Word Search of Glenn's WWW sites]
Last updated June 16, 1999,
Responsible NASA Official:
Gynelle.C.Steele@nasa.gov
216-433-8258
Point of contact for NASA Glenn's Research & Technology reports:
Cynthia.L.Dreibelbis@nasa.gov
216-433-2912
SGT, Inc.
Web page curator:
Nancy.L.Obryan@nasa.gov
216-433-5793
Wyle Information Systems, LLC
