Semester

Spring

Date of Graduation

2008

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Mechanical and Aerospace Engineering

Committee Chair

Mario Perhinschi.

Abstract

Pilot fatigue has been proven to be the cause of many aviation accidents. Fatigue introduces error into the pilot's inputs, which can potentially lead to accidents. To date, fatigue has been widely researched through physiological variables and sleep studies. Often, systems monitoring physiological variables would require constant physical contact with the pilot during flight. This arrangement could be cumbersome to pilots, and may hinder their flying ability even more. These systems will also add unnecessary weight to the aircraft, which could lead to increases in fuel consumption. Sleep studies have been investigated in an attempt to determine causes of pilot fatigue based on the amount and quality of sleep they have received pre-flight, but they only serve for fatigue prevention purposes.;The main objective of this research effort is to show that separation between 'rested' and 'tired' pilot conditions can be put into evidence using parameters based on aircraft state and control variables and to design a fatigue detection scheme to determine the 'on-line' state of the pilot for a set of typical maneuvers.;Five pilots were instructed to fly a 6 degrees-of-freedom flight simulator through a given flight scenario under 'rested' and 'tired' conditions. State and control variables such as aircraft roll rate, angle of attack, elevator deflection, and others were recorded during flight. The desired values of these variables were determined depending on what maneuver the pilot was trying to accomplish. Steady state flight conditions and doublet inputs in still air and turbulence were considered in this study. Tracking errors were defined as the difference between the actual variable value and the desired value. Standard deviation and mean of the tracking errors were considered as candidate fatigue detectors and their performance was analyzed. The most promising detectors were then used to define composite detection parameters as weighted sums.;Two detection schemes were designed to determine the 'rested' or 'tired' state of the pilot based on comparing the composite parameter values to a threshold. The first scheme used heuristic and binary logic to define a series of rules hard coded through 'if else' statements capable of determining the pilot's condition. The second detection scheme relied on fuzzy logic to make a 'rested' or 'tired' determination. Results showed that both schemes were capable of correctly classifying the condition of the pilot for many maneuvers. The detection schemes performed the best for the maneuvers performed in still air, but the detection rate was reduced when severe turbulence was present. A third approach of fatigue detection was investigated through implementation of a fuzzy neural network, and positive preliminary results deemed this method worthy of further exploration.;The analysis in this study presented compelling evidence that fatigue detection can be accomplished through the monitoring of aircraft state variables. Further research into using these detection schemes in conjunction with a flight compensation system may prove to be a viable, cost-effective intervention for reducing the number of accidents attributed to pilot fatigue.

Download

Share

COinS