Design and flight testing actuator failure accommodation controllers on WVU YF-22 research UAVs

This dissertation describes the design, development, and flight testing of a Neural Network (NN) based Fault Tolerant Flight Control System (FTFCS) with the ability to accommodate for actuator failures. The goal of this research was to demonstrate the ability of a specific set of control laws to maintain aircraft handling qualities in the presence of failures in the actuator channels. In this study, two-failure scenarios have been investigated: aileron failure (locking of the right aileron at a trim position) and elevator failure (locking of the right elevator at a trim position). A fleet of WVU YF-22 research aircraft test-beds were manufactured and instrumented for developing and testing of flight control software. An on-board payload with a PC-104 format computer system, sensors, and custom made circuit boards were designed and developed for these aircraft test-beds. The fault tolerant flight control systems for this study were designed to recover the aircraft with damaged actuators. On-board real-time data acquisition and control software was developed to achieve the Actuator Failure Accommodation (AFA) flight demonstration.; For the purposes of this research, control laws were required to be adaptive to changing aircraft dynamics during a failure scenario. On-line learning NNs---with their non-linearity and learning abilities---were used in the design of the on-board aircraft control scheme. The on-line training reduced the criticality of an extensive on-line Parameter IDentification (PID) during the failure and gives an on-board flight controller the capability to adjust to maintain the best possible flight performance during an unexpected failure.; This document will outline and describe the design and building of the flight controller, aircraft test-beds, on-board payload systems, and software in detail. Flight test results will be presented and documented to demonstrate the performance of a NN based FTFCS under failure conditions.