【作者】 任波；

【导师】 李炜；

【作者基本信息】 兰州理工大学 ， 控制理论与控制工程， 2012， 硕士

【摘要】 现代控制系统不断朝着规模集成化、结构复杂化和功能智能化方向发展，众多数量的传感器、执行器以及系统元部件协同运行，不可避免地会增加故障发生的几率。为减少因故障造成的不利影响，提高系统可靠性和安全性变得尤为重要。作为解决途径之一的故障诊断与容错控制，应对此类问题的各类新方法的研究已成为该领域的热点和难点。由于实际系统中或多或少存在非线性，加之系统建模误差和各种外界干扰，使得对实际系统的故障诊断与容错控制研究变得更加复杂和困难。基于此，本文针对该技术领域内存在的一些问题，结合逆系统方法，围绕非线性系统故障调节中的故障诊断观测器及基本控制器设计问题开展了以下研究工作：其一，为了提高已有成果中非线性系统执行器跳变故障估计的收敛过程不理想问题，文中通过引入比例环节，改进了一类纯积分型故障自适应算法，设计出的PI自适应故障诊断观测器提高了故障估计的快速性和平稳性，文中还从理论上对诊断观测器的状态和故障估计的收敛性进行了证明。其二，针对非线性系统执行器时变故障难以诊断估计问题，文中通过引入PD型迭代学习策略，设计了一种迭代学习故障诊断观测器实现了对时变故障的准确、快速估计，亦从理论上对该诊断观测器的状态和故障估计的收敛性做了证明。其三，依据以上设计的诊断观测器，考虑系统外界干扰和参数的不确定性影响，在逆系统方法框架下将内模控制引入，最终在执行器故障快速估计的基础上，实现了对执行器跳变及时变故障的有效补偿和对系统不确定性的鲁棒容错控制。其四，针对非线性系统的执行器故障诊断观测器存在的故障自适应律切换问题，设计一种扩张状态故障诊断观测器。文中将执行器故障和系统干扰扩展成一阶状态重构了一个增广系统对故障进行检测和估计，避免故障自适应律的切换问题；并结合逆系统方法依据故障估计值补偿故障对系统的影响，在此基础上还为补偿后的伪线性系统设计了鲁棒保性能最优控制器。针对上述方法采用典型算例进行有效性仿真研究,结果表明文中所给出的不同执行器故障估计算法对相应的故障估计快速、精确；基于逆系统方法的内模及保性能控制器使系统在正常和故障情形下均有良好的动态、静态及鲁棒性能。更多还原

【Abstract】 Modern control system are becoming more and more integrated, complex andsophisticated in their demand for performance, reliability and increasing autonomy. Itis inevitable for lots of sensors, actuators, and system components coordinatedoperation that the fault occurs. In order to reduce bad influence caused by failure, it isbecoming particularity important to improve system reliability and security. Themethod of fault diagnosis and fault-tolerant control was considered as one of mainsolutions, which has been a hotspot and difficulty for these problems.However, the existence of the nonlinearity in the practical plant and uncertaintyand noise of the plant models make it more and more complex and difficult. Base onthis statement, the fault accommodation approaches for nonlinear system are studied,which main works include the design of fault diagnosis observer and basic controller.The main research results are given as follows：1. In order to enhance the convergence of fault estimation of nonlinear actuatorfailure system in the existing research results, a pure integral style fault adaptivealgorithm was improved by introducing a proportional term; an adaptive observer wasdesigned to estimate the actuator fault, so that the fastness and stability of faultestimation were improved, and it was proved that the convergence of the faultdiagnosis observer could be achived in theory.2. An iterative learning observer of fault diagnosis for the weakness of actutotfault research in linear system was proposed by introducing a PD style iterativelearing strategy, which could estimatie the fault fastly and exactly. And it is alsoproved that the convergence of the fault diagnosis observer could be achived intheory.3. According to the above designed fault diagnosis observer, the fastness ofactuator fault estimation was realized, so that the research on fault conpense and therobust of uncertainty could be carried out by the guidance of inverse system methodand introducing internal model control for the existence of parameter uncertainty anddisturbances of the plant model.4. An extended state observer was designed for the switched problem of faultadaptive law in the study of nonlinear system actuator fault. An augment system wasconstruted by making the unknown system actuator fault and disturbance as an extended system state for fault detection and fault estimation to avoid the switchedproblem of fault adaptive law; and combined with inverse system method and thevalue of fault estimation to compensate the fault influence. And a robust guaranteedcost optimal controller for the compensated perseo linear system was proposed basedon the extended state observer.Finally, according to the above method for typical example of the effectivenessof simulation research, the results showed that different actuator fault estimationalgorithms given by the paper for the corresponding fault were fast, accurate; underthe situation of nomal and fault the system had a good dynamic, static, and robustperformance based on the internal model and the guaranteed cost controller of theinverse system method.更多还原