飞机故障下的多层次驾驶员模型研究及故障容错飞行控制

The control authority of an aircraft is inevitably reduced when failure occurs. It is crucial to study the pilot behavior model and fault tolerant flight control(FTFC) of a post-failure aircraft in order to enhance its flight safety and survivability...Part one of this study will investigate the judgment, decision-making and control behavior characteristics of a pilot when an unpredictable structural,actuator or controller failure occurs to an aircraft during the flight.The methodology of the study in part one consists of three steps. First, simulation experiments of pilot in the loop are carried out on a flight simulator, and sudden benchmark failure cases are triggered. The pilot then reacts to the changes, and his/her adapting process is introduced. The experiment data are collected afterwards and further analyzed. Second, based on the pilot behavioral hierarchy of an automated flight human-machine system, system identification is applied to analyze the pilot behavior. Parameters of the pilot hierarchy model, which is derived for the purpose of decision-making and control when operating an aircraft, are identified using the intelligent decision-making theory and human-machine adaptive law. Third, we will validate the constructed pilot models by comparing their performance with that of a human pilot...Part two of this study focus on automated fault tolerant flight control. Global aerodynamic modeling and identification problem is addressed with sudden aircraft structural/actuator failures taken into account. More precisely, multivariate simplex B-splines (MVSB) will be studied, and the research is aimed at optimizing the spline model structure, enhancing its approximation accuracy, and reducing its computational complexity in time. Using this simplex B-spline model, an adaptive nonlinear dynamic inversion (ANDI) attitude/flight path controller will be designed. Besides, the sensor based backstepping (SBB) control law, which is essentially a nonlinear incremental control law and requires measurements from angular accelerometers, will be further studied.The SBB control method will be improved through solving the application related issues. The model based ANDI controller and improved SBB fault tolerant controller will be validated using F16 model and RECOVER Boeing 747 nonlinear aerodynamic model. .Finally, the automated controllers such as ANDI and SBB will be compared to the abovementioned pilot model to study their differences and similarities in terms of performance and model structures.

研究飞机故障情况下驾驶员行为建模和容错自主飞行控制对保障飞行安全提高生存能力有重要意义。本课题研究飞机在结构、舵机和控制系统故障下驾驶员判断、决策和控制行为特征，建立驾驶员多层次行为模型。首先在模拟器上进行驾驶员在环仿真试验，设置典型突发故障，激发驾驶员应对突变的过程，获取试验数据；其次基于人机系统中驾驶员多行为特征，确定智能决策与控制的多层次模型结构，采用智能决策理论和人机自适应规律辨识驾驶员模型参数；最后将所得驾驶员模型与飞机构成闭环系统检验模型的有效性。本课题还研究故障容错自主控制。研究基于Simplex样条的故障飞机全包线非线性动力学建模和辨识方法，重点优化模型结构选择、提高模型逼近精度和降低算法复杂度。设计基于模型的自适应非线性动态逆容错控制器，并拓展和完善一种基于角加速度传感器的增量型非线性控制算法。用F16和RECOVER模型验证控制器，并分析自主控制器与驾驶员模型的相似性。

本课题研究了飞机在遭受突发故障后的应急容错飞行控制问题，旨在提高现代飞机的战力和生存能力。飞机突发故障后呈现了强时变和强不确定性，本项目从故障后飞机的动力学操纵耦合特性的揭示以及故障后的自适应容错控制机理揭示两个方面展开了研究。.故障时变过程的动力学操纵耦合特性揭示方面，一方面提出了新颖的多变量simplex样条结构化飞机建模和辨识方法，解决了提高全局精度、模型简洁度、在线实时性和参数收敛一致性问题；另一方面围绕机器学习和深度学习等数据驱动类学习策略，提出了新颖的极限学习类支持向量机，解决了非线性全局模型结构优化和提升全局精度问题。在故障突发后的强时变过渡态容错自适应控制机理揭示方面，分别从全自动控制和驾驶员参与的人因智能启发式控制两个层次开展了容错控制策略研究。.全自动容错控制方面，首先，提出了基于样条高精度实时模型的非线性动态逆控制策略以及基于模型实时线性化的增量型模型预测控制策略，用模型辨识与自适应控制机制的组合来实现强故障容错控制，模型辨识单元负责故障信息感知。其次，提出了改进型的增量型反步法等不依赖精准模型的容错控制方法，解决了强时变过渡态模型更新精度、实时性和收敛一致性难以保障的问题。.在驾驶员智能容错方面，结合模拟器上的驾驶员在环仿真试验，开展了强故障突发后的驾驶员智能容错自适应行为特性分析、参数提取和机理建模研究。首先，开展了试验研究以获取优秀驾驶员的容错控制数据，针对故障强时变过程中驾驶员操纵行为的频谱特性分析问题，提出了通过小波分析来实现快时变系统的响应数据的短窗口频谱分析策略。其次，提出了包括智能感知、决策、执行三个层次的驾驶员多层次模型架构体系。其中，在Hess时变驾驶员模型结构的基础上，提出了改进的模型结构以及基于模糊逻辑调度策略的驾驶员智能容错变策略行为的建模方法。.针对典型机身/舵机突发故障，用F-16飞机以及F-18舰载机等模型数据，验证了本研究中模型辨识和容错控制策略的有效性。