【作者】 吕明；

【导师】 吴晓蓓；

【作者基本信息】 南京理工大学 ， 控制科学与工程， 2007， 博士

【摘要】 本文研究了网络控制系统的故障检测与诊断，主要内容如下：(1)讨论了控制系统中故障的定义和分类问题。分别基于故障的发生部位、性质和作用类型，对常见故障进行了分类。基于故障的发生部位和故障的性质，对网络控制系统建立了故障的数学模型，并得出了发生执行器和传感器故障的统一数学模型。(2)研究了具有短时延和数据包丢失的网络控制系统的故障检测问题。针对传感器、控制器节点为时间驱动，执行器节点为事件驱动的短时延网络控制系统，建立了具有参数不确定性的离散时间系统模型。当不存在数据包丢失现象时，设计了故障观测器，利用Lyapunov稳定性定理，得出了观测器系统的稳定性条件。当存在数据包丢失现象时，在一定的丢包率情况下，将误差方程建模为异步动态系统，进行了故障检测的研究，并给出了系统指数稳定性的条件。(3)研究了具有长时延的网络控制系统的鲁棒故障检测与诊断问题。假定系统的各个节点均为时间驱动方式，在离散动态模型的基础上，分别按照系统具有输出时延、控制时延以及同时具有输出时延和控制时延来进行讨论。当系统具有输出时延时，建立了具有时延补偿功能的观测器，并将观测器误差方程建模为切换系统，得出了误差系统的稳定性条件。当系统具有控制时延时，考虑了传感器和执行器的采样时钟分别为同步和异步两种情况下的故障检测问题，设计了故障观测器，并证明了误差系统的稳定性。当系统同时存在输出时延和控制时延，且具有外加干扰输入时，研究了系统的鲁棒H_∞故障检测问题，并对鲁棒性和灵敏度进行了综合分析。在故障检测的基础上，采用方向性残差的方法，对故障进行了分离，实现了故障诊断。(4)研究了非线性网络控制系统的故障检测与诊断问题。针对被控对象为仿射非线性且具有输出时延的网络控制系统，利用模糊主导子系统的思想，设计了基于局部T-S模糊模型的时延补偿观测器，并给出了全局渐近稳定的充分条件。针对被控对象为一般非线性动态方程且具有干扰的网络控制系统，对其建立了一组局部模糊模型。考虑系统同时具有控制时延和输出时延的情况，利用模糊奇偶方程，构造了残差产生器，讨论了对系统进行故障检测和诊断的方法，并给出了奇偶向量的全解耦条件和求取方法。(5)研究了时延多包网络控制系统的鲁棒故障检测问题。当控制量采用多包传输时，对时延可以忽略以及存在控制时延或输出时延的情况分别进行了分析。当检测量采用多包传输时，分析了控制时延和输出时延同时存在的情况。对系统建立了基于增广模型的故障观测器，得出了具有统一形式的误差方程，并考虑数据包的不同传输方式，对系统进行了故障检测的研究。当系统具有外加干扰时，设计了H_∞故障观测器，保证了故障检测的鲁棒性。更多还原

【Abstract】 The fault detection and diagnosis (FDD) of the networked control systems (NCS) arestudied in this dissertation. The main research works are concluded as follows:(1) The definition and classification of the faults in control systems are discussed.The classification is based on the position, character and effect of the common faults,respectively. The faults of the NCS are modeled based on the position and the character ofthe faults, and the uniform model of the systems with actuator and sensor faults is given.(2) The fault detection of the NCS with short time delay and data packet dropoutis studied. The NCS with time-driven sensor node and controller node, event-drivenactuator node and short time delay is modeled as an uncertain discrete time system.Considering the NCS without data dropout, the fault observer is designed and the stabilitycondition is given based on Lyapunov stability theory. Considering the NCS with datadropout, under fixed packet dropout rate, the error equation is modeled as an asynchronousdynamical system, the fault detection is studied and the condition for exponential stabilityof the systems is given.(3) The robust fault detection and diagnosis of the NCS with long time delay arestudied. The sensor, controller and actuator nodes are all time-driven. Based on discretemodels, the problem is discussed supposing the NCS with output delay, control delay andboth output delay and control delay, respectively. For the NCS with output delay, theobserver with delay compensation is constructed. The error equation of the observer ismodeled as a switched system, and the stability condition of the error system is given. Forthe NCS with control delay, the fault detection problem is considered when the sensor andthe actuator nodes are synchronous or asynchronous. The fault observer is designed and thestability is proved. For the NCS with both output delay and control delay and disturbances,the robust H_∞fault detection is studied, and the robustness and sensitivity are analyzed.After the fault detection, the faults are isolated with directional residual method, and thefault diagnosis is accomplished.(4) The fault detection and diagnosis of the nonlinear networked control systemsare studied. For the NCS with affine nonlinear controlled object and output delay, using theconcept of the dominant subsystem in fuzzy models, the observer with delay compensationis designed based on local T-S fuzzy models, and the sufficient condition of globallyasymptotic stability is given. For the NCS with usual nonlinear controlled object and disturbances, the local fuzzy models are given. With both control delay and output delay,the residual generator is designed using fuzzy parity equation. The method of faultdetection and diagnosis is discussed, and the fully-decoupled condition and solving methodof the parity vectors are given.(5) The robust fault detection of the NCS with time delay and multi-packettransmission is studied. When the controller data is multi-packet transmission, the systemswithout delay and with control delay or output delay are analyzed, respectively. When thesensor data is multi-packet transmission, the system with both control delay and outputdelay is analyzed. The fault observer is designed based on the augment model, the uniformerror equation is given, and the fault detection is studied considering the differenttransmission manners of the data packet. When there are disturbances in the NCS, the H_∞fault observer is designed to ensure the robustness of the fault detection.更多还原