【作者】 张晓燕；

【导师】 吴光辉；

【作者基本信息】 西北工业大学 ， 飞行器设计， 2016， 博士

【摘要】 航空安全是航空工业发展的重中之重。人机工效是改善和提高航空安全的重要途径之一,飞行员认知与操纵特性是飞机驾驶舱人机工效设计和评估的基础,是“以人为中心”的驾驶舱设计的基本依据。掌握飞行员的认知和操纵规律能够提高驾驶舱人机工效的设计水平,减少由飞行员失误引起的航空事故,提高航空安全。本文首先分析了飞行员认知与操纵特性的研究进展,在此基础上确定本文的研究重点为飞行员动态视觉特性和操纵特性。根据研究目的设计并开展了不同于传统飞行仿真平台上的进近着陆滑行阶段的飞行员实验。实验选择20名不同飞行经验的航线飞行员在东方飞行培训中心的飞行模拟器上进行了3个不同能见度,2条不同难度滑行路线等6个场景的实验,采集了飞行员主观问卷数据、视频数据、眼动数据等；建立了实验数据的分析方法；对不同场景下、不同经验飞行员的注视模式进行了分析。为论文后续的飞行员特性研究和建模仿真提供数据支撑和验证依据。针对飞行员动态视觉特性,本文基于信息论,以任务难度与飞行员的注视频率为变量建立了视觉绩效指数模型,能够表征飞行员在执行任务过程中所需识别和编码的信息量。通过飞行员地面滑行实验数据验证了飞行员视觉绩效指数模型的正确性。本模型综合考虑任务客观情况以及飞行员注视频率,能够定量的表征飞行员的注视行为,与视觉熵结合可以完整表征动态视觉过程。以飞行员的个人能力与任务客观特性为变量建立任务难度模型,模型综合考虑飞行员的知识背景和经验水平等个人能力,以及影响任务的外界环境和任务本身的复杂度等任务的客观特性,能够定量评价和预测不同飞行员完成不同滑行任务时的任务难度。针对飞行员操纵特性,本文基于驾驶舱环境下飞行员操纵特性实验,分析了操纵目标特征对飞行员操纵特性的影响,建立了能够表征飞行员操纵特性的Fitts定律,提出了驾驶舱人机界面设计中关键操纵设备设计与布置的建议；根据飞行员定位运动实际触点的分布规律,建立了无故障宽度的计算模型,提出了理论上操作失误率为零的按钮设计宽度,揭示了操纵失误产生的根本原因；以信息加工阶段论为基础建立了飞行员信息处理时间预测模型,结合飞行员的认知和操纵特点,提出了飞行员的视觉搜索策略和目标编码定位策略,利用该模型可预测飞行员对驾驶舱显示信息的处理时间和信息识别编码过程。针对飞行员认知与操纵特性建模与验证,本文基于人的认知特性规律,提出了建立飞行员仿真模型的方法,以ACT-R模型架构为基础,建立了基于任务的飞行员仿真模型架构；将飞行员的任务背景知识、动态视觉特性以及操纵规律完整的表征在飞行员仿真模型的陈述性知识模块、感知与作动模块、产生式系统中,完成了基于任务的飞行员认知与操纵特性建模；提出了分场景分群体分层次的飞行员仿真模型验证方法,结合地面滑行实验数据验证了仿真模型的正确性。利用飞行员仿真模型对韩亚空难进行了多可能事件的仿真,仿真结果与韩亚航空公司的事故分析结论一致,进一步表明了飞行员仿真模型的正确性和有效性。更多还原

【Abstract】 Safety is actually most important in civil aviation. Ergonomics is a best way to improve the aviation safety. Pilot cognition and motion performance as the primary of "human-centered design" is also the criterion what the ergonomic design and evaluation of flight deck must accords to. The Pilot cognition and motion performance mastered can enhance the ergonomic design naturally, the aviation accident induced by pilot error can be reduced consequently, and therefore the aviation safety can be improved significantly.This paper firstly analyzes the research advancement of the pilot’s cognitive and motion performance, and puts forward the research focus is the pilot’s dynamic visual performance and motion performance.To support the research and validation of pilot performance, this paper has designed and carried out multi-scenario testing by pilots in the flight simulator in Shanghai Eastern Flight Training CO., LTD.. The experiment which is different from the traditional testing executed in flight simulation equipment covers 3 different visibility and 2 different difficult taxi routes executed by 20 pilots with different flight experience in flight simulator. The data includes video, eye movement, and questionnaires and so on. The paper gives the method of how to analyze the huge data both subjective and objective and also analyzes the eye movement mode of different scenarios and pilots.For pilot dynamic visual performance, this paper has built the Index model of dynamic Visual Performance (IPv). The model based on the information theory is the function of task difficulty and pilot fixation rate, and has been verified by the data from the taxiing experiment. IPv actually represents the information pilot needs to identify and encode during the task, therefore it is a quantitative model to describe the pilot fixation performance and if combined with visual entropy which is the only synthetic index representing the saccadic performance can represent the whole process of visual performance. The second work here is to build the pilot task difficulty model. The task difficulty model considers both the pilot capability which means the knowledge and the experience and the objective task performance such as external environment and task difficulty. The model can predict and evaluate the difficulty of different taxi tasks by different pilots in different scenarios.With regard to the pilot motion performance in flight deck, there are three models built. Firstly, the effect of targets characteristics such as size, arrange position, shape and density on pilot motion performance has been studied first based on the pilot test in the flight deck, and Fitts’law described the mathematical pilot motion regularity has been established. According to the research conclusions derived some suggestions for equipments design such as what size should the button is, or which position is best for the emergency operation buttons are also proposed. Secondly, no error width model has been established through the distribution of pilot movement end-points. The model discovers the essence of pilot motion error and gives the no error design width theoretically. The predictive time modeling for pilot information processing has been built finally based on the information theory covers the pilot cognition and motion performance, such as pilot’s visual scan strategy and target encoded and oriented strategy. The model can evaluate and predict the display information process time and information identify and encode strategy by pilot in flight deck.Finally the pilot cognition and motion performance model is built and verified. The paper proposes the strategy for pilot modeling. And based on ACT-R cognition architecture, the pilot’s knowledge about flight, their dynamic visual performance and motion performance could all be represented in the perception and motion module, the declarative knowledge module, the production system of model respectively and finally the pilot cognition and motion performance model is completed. The multi-scene multi-group multi-hierarchy validation method has been proposed and finally verifies and validates the pilot cognition and motion performance model successfully. The validation method solves the problem of how to validate the cognition model because of the randomness and complication of human being and is also an effective exploration. And finally, The pilot performance model built here is used to simulate the Asiana accident and to find the causes the multi-event simulation is executed. The simulation has the same conclusions as the NTSB. The simulation conclusions suggest that again the model is validity and effective.更多还原