EXPERT SUPERVISION OF FUZZY LEARNING-SYSTEMS FOR FAULT-TOLERANT AIRCRAFT CONTROL

Actuator sensor or other aircraft subsystem failures, or structural fa ilures that result from, for example, battle damage can cause catastro phes that may lead to loss of the aircraft. While experienced pilots c an often compensate for failures, in certain emergency situations ther e is the need for computer-assisted or fully computer-automated reconf iguration of the aircraft control laws to save the aircraft. In this p aper, we begin by showing that the fuzzy model reference learning cont roller (FMRLC) [1]-[5] can be used to reconfigure the nominal controll er in an F-16 aircraft to compensate for various actuator failures wit hout using explicit failure information (e.g., the time of occurrence of the failure or its magnitude). Next, we show that the performance o f the FMRLC can be significantly enhanced by exploiting failure detect ion and identification (FDI) information to achieve a ''performance ad aptive'' system that seeks an appropriate performance level depending on the type of failure that occurred We develop an expert supervision strategy for the FMRLC that uses only information about the time al wh ich a failure occurs and show that it achieves higher performance cont rol reconfiguration than an unsupervised FMRLC. In addition, we show t hat similar performance can be achieved if we only use estimates of th e failure time and magnitude obtained from a fuzzy estimator We close our study with a brief assessment of the advantages and disadvantages of the approaches used in this paper.