The Design And Implementation Of An Automatic Evaluation System For Pilot Training On Flight Simulator

Simulator training is the best even the only way for airline pilot training, due to its advantages ofcontrollable, no damage, economic and reliable. Traditionally, simulator training evaluation mainlyrelies on the field observations of pilot manipulation by coach, which has strong subjectivity, heavyworkload and difficulty in establishing training standards. Therefore, it has important practicalsignificance to research and design automatic evaluation system for pilot training on flight simulatorto train qualified pilots in order to adapt to the rapid development of China civil aviation system.Firstly, the thesis discusses the functional requirements and performance criteria of the automaticevaluation system for pilot training on flight simulator. And combined with database, networkcommunications and data mining techniques, it provides the overall design and software design.Aiming at the problem of difficulty in obtaining flight rules and long time spent by humans, the thesismainly studies a rule extraction approach based on parameter reduction and immune algorithm. Basedon this, it realizes the functions of automatic extraction of flight rules from the large number of flightdata and construction of scoring standard library, which provides the basis for the realization ofautomatic scoring. Then the thesis analyzes flight rules according to the quaternion type in scoringstandard library and compares them with flight data to realize identification and automatic scoring ofpilot training. In view of that training course performance is evaluated on the basis of phase,subjectand task hierarchy, the thesis adopts AHP to construct performance evaluation model and achievesobjective evaluation of the training course by comprehensive evaluation of all subjects in a training.Finally, according to analysis of pilot’s history performance and the main factors,the thesis studies adecision tree algorithm based on C4.5to realize the prediction of pilot performance trend, in order tofacilitate the scientific formation of training plans for pilots.Based on the detailed design of overall scheme and each functional module of the system, thethesis realizes the automatic evaluation system for pilot training on flight simulator. The systemrequires about4minutes to evaluate4hours’ flight data, about7seconds to display profile and thecorrect rate of pilot performance predicted reaching87.5%, which meet the performance criteria ofthe system.