【作者】 管宇；

【导师】 张迎春；

【作者基本信息】 哈尔滨工业大学 ， 控制科学与工程， 2012， 博士

【摘要】 姿态控制系统是保障卫星在轨正常工作的核心部分，由于长期运行于恶劣的空间环境，尤其是敏感器、执行机构等部件长期不断地在执行各类控制操作，极易出现故障，进而会导致不可修复的严重后果，因此卫星姿态控制系统的故障诊断和容错控制技术一直是航天领域研究热点和难点问题。同时，由于机构运动、燃料消耗及空间环境影响等将导致在轨卫星的系统模型参数发生未知变化，传统上基于精确模型的姿态容错控制技术具有一定的局限性。本论文在国家自然科学基金项目(课题编号：61104026)资助下，面向我国重大设施及装备的高可靠性及长寿命战略发展需求，从理论和应用两方面对具有不确定模型影响的卫星姿态容错控制问题进行了深入的研究，取得如下研究成果。考虑存在各种噪声、未知干扰和模型不确定性，引入鲁棒H∞控制理论，提出了一种面向线性不确定系统的完整性被动容错控制设计方法。该方法利用线性不确定系统模型，将执行机构/姿态敏感器失效情况下被动容错控制器设计转换为相应的线性矩阵不等式的可行性问题，并进一步给出了系统具有完整性的充分条件和鲁棒H∞容错控制器的设计方法。在对卫星姿态动力学模型线性化基础上，采用所提方法设计了执行机构/姿态敏感器故障状态下的卫星姿态鲁棒H∞容错控制器，数学仿真分析表明该控制器能够保证执行机构/姿态敏感器突变或缓变故障时卫星姿态控制系统的控制性能，具有对已知故障良好的容错能力。针对存在未知干扰输入的Lipschitz非线性系统，融合基于模型和数据驱动的控制技术，提出了一种基于迭代学习未知输入观测器(IL-UIO)的主动容错控制方法。首先，采用对模型不确定和干扰解耦的原理，设计了一种非线性未知输入观测器，增强了系统故障观测的鲁棒性；在此基础上，利用被控系统的I/O数据，采用迭代学习利用上一步的故障观测信息来估计当前故障信息，跟踪和检测系统由于故障引起的改变并进而实现对故障的在线重构。采用Lyapunov稳定性理论证明了IL-UIO的鲁棒稳定和速度误差的一致有界性。将所提方法应用于面向执行机构故障的卫星姿态容错控制器设计，数学仿真结果表明IL-UIO能够有效跟随过去的故障系统，确保一定系统性能的前提下实现对飞轮故障的容错，由于采用了基于模型的观测器与数据驱动方法相结合，二者相互渗透、优势互补，即可增强系统残差对未知干扰输入的鲁棒性，又提高了系统对数据的使用质量。由于卫星姿态敏感器故障会污染到下一时刻其它健康敏感器的测量信息，影响对真正故障源的准确判断，对此，考虑卫星动力学模型参数发生的未知变化，基于PCA和SIM识别技术，提出了一种基于子空间数据驱动的卫星姿态敏感器故障检测、隔离和辨识(FDI)方法。该方法首先采用等价空间法对测量数据进行预处理，构建残差信号；然后采用子空间法辨识出增广观测矩阵，进而实现敏感器故障鲁棒检测；并通过线性变换对等价空间进行降阶，减小算法在线计算量。最后，将上述方法应用于卫星姿控系统的敏感器故障问题，数学仿真结果表明，不依赖故障先验知识，仅利用输入输出观测数据，就能够在线完成未知系统模型情况下的敏感器故障检测、隔离与辨识，由于采用了故障降阶检测处理，有效减少了计算量，适宜于卫星在轨应用。状态监测和系统重构是卫星长期自主运行的保障，也是开发星上健康管理系统的关键技术之一，首先，给出了一种含有状态监测和系统重构环节的健康管理系统容错控制体系结构；针对存在未知干扰输入的非线性系统，采用基于数据驱动的模式识别方法，设计了一组基于支持向量机(SVM)的非线性故障估计器，利用正常操作条件下和各种故障状态下的历史数据实现了故障信息的近似检测和隔离估计；进而利用检测的系统信息和辨识出的系统参数，设计了学习观测器，通过学习算法提供用于控制器重构的信息，修改反馈控制律，实现了对故障的自适应故障补偿，并理论证明了故障估计误差有界。最后，以卫星对地观测过程中自主运行任务为例进行了仿真验证，结果表明在执行器发生未预知故障时仍可以实现系统状态监测和故障重构。更多还原

【Abstract】 Attitude control system is the core subsystem to maintain regular satellite operation.Due to long-term in-orbit work in the harsh environment of space, the components suchas actuators and sensors in the attitude control system continue to perform in-orbitcontrol operations, therefore are prone to faults. Due to the severe consequence of faultirreparability, the satellite attitude control system fault diagnosis and fault tolerant controltechnology has always been a strong concern in the field of aerospace industry.Meanwhile, conventional fault-tolerant attitude control techniques based on precisemodels have limitations to deal with the unknown changes of on-orbit satellite systemmodel parameters caused by mechanical motion, fuel consumption and spaceenvironment. Oriented from the high reliability and long-life requirement for majorfacilities and equipment, and funded by the National Natural Science Foundation ofChina (61104026), the thesis performs in-depth study on fault-tolerant control issues ofsatellite attitude control system with model uncertainties in both theoretical and appliedaspects. The research mainly involves the following points.Considering the existence of noises, unknown disturbances and model uncertainties,an integrity passive fault-tolerant control design method based on robust H∞theory isproposed to deal with the linear uncertain system control problem. It uses linear uncertainsystem model to transform the passive fault-tolerant control problem into the feasibilityof corresponding linear matrix inequalities in the case of actuator/sensor failure. Inaddition, it gives the sufficient condition for system completeness, and fault-tolerantcontroller design method using robust H∞theory. Based on linearized model of satelliteattitude dynamics model, the fault-tolerant robust H∞controller is designed in the case ofactuator/attitude sensor fault states. Mathematical simulation analysis shows that thecontroller can guarantee control performance of satellite attitude control system againstfaults of sudden or slow actuator/attitude sensor changes, therefore have good fault-tolerant performance against known faults.To deal with Lipschitz nonlinear system with unknown disturbance input, an activefault-tolerant control method based on iterative learning-unknown input observer (IL-UIO) is proposed. First, using the principle of decoupling model uncertainties anddisturbance, a nonlinear unknown input observer is designed to guarantee robustness forfault observation. Then based on I/O data of the controlled system, the current faultinformation is estimated with last step fault observation using IL. The estimation is usedto track and detect fault caused system changes, and implement online faultreconstruction. The robust stability and uniform boundness of IL-UIO are proven usingLyapunov stability theory. The proposed method is applied to the design of fault-tolerant satellite attitude controller with actuator faults. Mathematical simulation results showthat IL-UIO can effectively track past fault system and realize tolerance against flywheelfaults with certain guaranteed system performance. The combination and fusion ofmodel-based observer and data driven approach are adopted here to reinforce robustnessof system residual against unknown disturbance input and improve system data usequality.Satellite attitude sensor faults may pollute measurement information of otherhealthy sensors at next time step, and therefore affect accurate determination of the realfault source. To address the problem and unknown changes of satellite dynamics modelparameters, a fault detection, isolation and identification (FDI) method is proposed,which is based on subspace-aided data driven design using PCA and SIM identificationtechniques. First the equivalent space method is used to preprocess measured data andconstruct residual signal. Then the subspace method is employed to identify augmentedobservation matrix to realize robust detection of sensor faults. The order of the equivalentspace is reduced further with linear transformation to reduce online computation. Finally,the proposed method is applied to satellite attitude control problem with sensor faults.Mathematical simulation results show that it can achieve online FDI with unknownsystem model by only using input-output observation, independent on prior faultknowledge. The adoption of reduced-order fault detection and processing effectivelyreduces computation, and makes the method suitable for online satellite applications.State monitoring and system reconfiguration safeguard long-term autonomoussatellite running, and are also key techniques in the development of onboard healthmanagement system. First, a fault-tolerant control architecture of health managementsystem containing state monitoring and system reconfiguration is given. Consideringnonlinear system with unknown disturbance input, the data driven based patternrecognition method is used to design a set of nonlinear fault estimators based on supportvector machine (SVM). It employs historical data in normal operation and faults states ofdifferent kind to achieve approximate detection and isolated estimation of faultinformation. The information and identified system parameters are then used to designlearning observer to realized adaptive fault compensation. The learning algorithm here isexploited to supply controller reconfiguration information and modify feedback controllaw. Theoretical analysis proves error boundness of fault estimation. Finally, simulationdemonstration is performed under the background of autonomous running task of smallsatellite earth observation. Results show system state monitoring and systemreconfiguration can be achieved in the case of unknown actuator faults.更多还原