TIME SERIES ANALYSIS OF CLOSED-LOOP PILOT VEHICLE DYNAMICS
Abstract
The off-line development of linear discrete autoregressive models for man-machine closed-loop flight performance is presented. The development includes single input, single output (single-channel) and multiple input, multiple output (multi-channel) closed-loop systems. Previous research is consolidated by extensive surveys of both single and multi-channel closed-loop modeling and is extended into a comprehensive joint autoregressive man-machine identification process. In single-channel closed-loop tracking tasks, "pilot" manual control is modeled in terms of linear discrete transfer functions which are parsimonious and guaranteed stable. The transfer functions are found by applying a modified superposition time series generation technique to relatively short data records, approximately 25 seconds long. The resulting model is then validated and analyzed. Results from a piloted laboratory simulation of single and double-integrator controlled elements (longitudinal axis) agree with previous research findings. The multi-channel identification process is developed in five steps. First, the mathematical "existence" conditions are derived for a joint innovations representation (JIR) and are related to experimental controls; second, the relationship is developed between the JIR and the associated physical model, the joint autoregressive representation (JAR), which incorporates Markov noise in each channel; third, a Levinson-based identification algorithm is derived called normalized predictive deconvolution (NPD) which identifies the JIR directly from the data sets; fourth, model order selection rules are presented; and fifth, a closed-loop analysis procedure is applied to case studies which provides mathematical insights useful in the evaluation of flying qualities, including pilot induced oscillations. The modeling process as presented is a powerful analysis tool for the flight control designer, test engineer, and handling qualities expert. These models, however, should not be used to associate model "parameters" with physiological constraints on the pilot. An applications data base is desired to explore and develop further the potential of the joint autoregressive modeling process.
Degree
Ph.D.
Subject Area
Aerospace materials
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.