基于多源生理数据融合的飞行员工作负荷评价方法研究  

Study on Evaluating Pilot Workload Based on Multi-source Physiological Data Fusion

作　　者：江佳运 孙有朝[1] 张夏 宴传奇 JIANG Jia-yun;SUN You-chao;ZHAN Xia;YAN Chuan-qi(College of Civil Aviation,Nanjing University of Aeronautics&Astronautics,Nanjing 211106,China)

机构地区：[1]南京航空航天大学民航学院,南京211106

出　　处：《人类工效学》2023年第3期1-10,共10页Chinese Journal of Ergonomics

基　　金：国家自然科学基金委员会-中国民用航空局民航联合研究基金项目资助(U2033202,U1333119);国家自然科学基金项目资助(52172387)。

摘　　要：目的通过分析多源异构生理数据,提出评价飞行员工作负荷的机器学习方法,从而优化飞机驾驶舱人机工效学设计。方法在Kamov Ka-52飞行模拟器中进行12组飞行测试,每组测试包含五种类型的正常或异常事件。在飞行过程中收集了心电(ECG)、肌电(EMG)、脉搏和呼吸数据。首先对飞行过程进行定性分析,以显示飞行条件和工作负荷的总体趋势。然后,通过主成分分析(PCA)评估不同飞行阶段的工作量,并选择有效的生理参数进行降维并融合。最后,三种不同的机器学习模型,即基于支持向量机的(PWE-SVM)、基于遗传算法优化的支持向量机(PWE-GASVM)和基于反向传播神经网络的飞行员工作负荷评估模型(PWE-BP)被提出来量化飞行中的工作负荷。结果需要高水平操作技能和注意力集中的告警、地形跟随和着陆过程的飞行员工作负荷也相对较高。肌电和心电参数较脉搏和呼吸参数可以更有效的反应工作负荷。PWE-GASVM在飞行员工作负荷评估上表现最好。结论提出的飞行员工作负荷评估方法有效可行。Objective This paper proposes multiple machine learning methods of evaluating the pilot workload by analyzing the multi-source heterogeneous physiological data,thereby supporting the optimization of ergonomics design in aircraft flight decks.Methods Twelve sets of flight tests were carried out in a Kamov Ka-52 flight simulator and each test contained five types of normal or abnormal events.Electrocardiogram(ECG),Electromyography(EMG),pulse and respiratory signals were collected during the flight.A qualitative analysis was carried out at first to show the overall trend of the flight conditions and workload.Then,the workload in different flight stages was evaluated by Principal Component Analysis(PCA)and effective physiological parameters were selected for dimension reduction.Finally,three different machine learning models,namely the Pilot Workload Evaluation model with Support Vector Machines(PWE-SVM),the Pilot Workload Evaluation model with SVM optimized by Genetic Algorithms(PWE-GASVM)and the Pilot Workload Evaluation model with Back Propagation neural networks(PWE-BP),were proposed to quantify the in-flight workload.Results The results show that the events of alerting,terrain following and landing,which requires a high level of operating skills and concentration of attention,trigger high pilot workload.ECG and EMG parameters are more indicative of pilot workload than respiration and pulse parameters.Particularly,PWE-GASVM performs best on pilot workload evaluation by using the small-scale dataset.Conclusion The proposed pilot workload evaluation method is effective and feasible.

关 键 词：航空交通工程 人因工程 交通安全 飞行员工作负荷 生理参数 ECG EMG SVM 

分 类 号：TB472[一般工业技术—工业设计] R135.99[医药卫生—劳动卫生]