Research on the evaluation of the manual manipulation level of a student pilot in a typical scene

Abstract

The traditional evaluation method of student pilots ' manual manipulation level relies too much on teachers' subjective judgment. A quantitative evaluation method based on flight data is proposed to compensate for the shortcoming of the traditional method. According to the theory of the core competence of a pilot, the evaluation index system of manual manipulation level of student pilots in a typical scene is constructed. The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) categorizes pilot manipulation levels into four grades. The improved support vector machine (SVM) model with particle swarm optimization (PSO) is trained using index data and rank labels as inputs. The data is split into training and verification sets with the cross-validation method. The training data for Guanghan Airport are taken as an example to verify the model's reliability. The model's prediction accuracy is compared with that of K-nearest neighbor and random forest models. The results show that the PSO-SVM model has higher prediction accuracy. The research can provide a theoretical reference for quantitatively evaluating manual manipulation levels and improving flight training quality.

Keywords

• Flight training
• Manual manipulation
• Typical scene
• Machine learning
• Evaluation model