# Research on the evaluation model of pilots’ in-flight stress state based on flight critical events

## TL;DR

This study develops a model to assess fighter pilots' stress during flights using physiological data, aiming to improve mission performance and safety.

## Contribution

A novel stress state assessment model for pilots using real in-flight physiological data and machine learning.

## Key findings

- Screened ten ECG and five respiratory indicators sensitive to pilot stress levels.
- Achieved 95.56% accuracy in stress state classification using an ensemble learning model.

## Abstract

Stress refers to the non - specific response that occurs in an organism when it is subjected to stressful stimuli. When fighter pilots execute high-stress missions, they are likely to experience stress responses. These responses may have a negative impact on the cognitive decision-making processes of the pilots, subsequently undermining the mission effectiveness. However, to date, there is a dearth of methodologies for evaluating the in-flight stress states of pilots. The aim of this research is to establish an assessment model by using the real physiological data of pilots in flight to predict their stress state during flight.

This study utilized the in - flight physiological data of 32 pilots aged between 22 and 30 years old. The data included electrocardiogram (ECG) data, respiratory data, and human body acceleration data. By analyzing the differences in the physiological characteristics of pilots under various stress conditions, sensitive physiological indicators corresponding to different stress states were screened out. Using electrocardiogram (ECG) and respiratory indices that are sensitive to stress changes as model inputs, a stress state assessment model was constructed via machine learning techniques. The performance of the model was evaluated using accuracy, sensitivity, specificity, and the F1 - score.

Screen out ten electrocardiogram (ECG) indicators that are sensitive to the stress level of pilots, which are Mean HR, MeanRR, SDNN, RMSSD, pNN50, pNN20, LF/HF, LFn, HFn and SD1/SD2. Screen out five respiratory indicators that are sensitive to the stress level of pilots, which are Mean Rsp, EB1, EB2, EB3 and EB4. The binary classification assessment model of stress state, which was constructed using the ensemble learning approach, achieved an accuracy of 95.56% in the five - fold cross - validation.

This study effectively accomplished the assessment of pilots’ stress states based on physiological characteristics. It is conducive to optimizing the stress training strategies for pilots and enhancing their capabilities to cope with and manage stress. This research has certain theoretical significance and engineering application value for improving the performance of flight missions and ensuring flight safety.