GUY NORRIS / LOS ANGELES
Objective is to find means for stabilised landing of combat-damaged or faulty aircraft
NASA is preparing to test an intelligent flight control system (IFCS) that it hopes will ultimately be capable of providing stabilised flight control and landing for aircraft damaged by combat or suffering system failures.
"It's a pretty big technical leap," says NASA Dryden Flight Research Center IFCS project manager John Carter, who adds that initial flight tests will be made using NASA's Boeing F-15B advanced control technology for integrated vehicles (ACTIVE) testbed. Using specially developed neural network software and an advanced flight control computer, IFCS builds on initial neural net flight tests made by NASA on the ACTIVE F-15 in 1999.
Two significant technological developments form the core of the IFCS. The first is the neural network itself, and the system's autonomous pattern recognition or "dynamic cell structure" feature which forms the self-learning, intelligent element of the system. The other is a set of algorithms that have been developed to identify specific "stability derivatives" or flight parameters. "We have simplified the number of these from around 26 to 12," adds Carter who says initial tests will be performed within a limited flight envelope.
The system will be hosted on a specially developed flight control computer - the airborne research testing system, which piggybacks on the aircraft FCS. First tests, starting next month, will evaluate updated instrumentation on the aircraft and collect flight data, which will be used to verify the parameter identification software in a set of ground tests.
Flight tests of "Generation 1" IFCS software is provisionally set for January 2003. The aircraft's "pre-trained" neural network will assume the aircraft is flying with the two-dimensional thrust-vectoring nozzles originally fitted to the ACTIVE aircraft. However, these have been replaced with conventional round nozzles that give significantly different flying qualities. The parameter identification software should spot the difference and work with the IFCS to adjust flight controls to compensate.
Tests of a more advanced Generation 2 IFCS, which would use the neural net implicitly, are planned for late 2003. Further evaluation of a more sophisticated system on a US Air Force Boeing C-17 is planned for around 2005. "Unlike the single channel F-15 system, which is prototypical, the C-17 will have a production-like version with four-channels and will cover the full envelope including take-off and landing," says Carter.
Source: Flight International
