【作者】 崔平；

【导师】 谢剑英； 翁正新；

【作者基本信息】 上海交通大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 动态系统的安全性、可靠性和对环境的污染问题日益受到人们的重视,这些要求不仅仅局限于核反应堆、化工过程、飞行器等安全性很重要的系统,而且已经扩展到诸如自主车辆、快速轨道交通等新型系统。故障诊断与容错控制是提高自动化系统可靠性与安全性的有效途径。因此深入研究故障诊断与容错控制技术,不但具有重要的理论意义,而且具有很大的实际应用价值。由于建模不确定性和未知扰动的广泛存在,近年来鲁棒故障诊断也受到了广泛关注。线性参数变化(LPV)系统是一类特殊的线性系统,其状态空间矩阵(系统状态空间描述中的系数矩阵,简称“状态空间矩阵”)是某些时变参数向量的函数。当这些时变参数沿某给定的参数轨迹变化时,LPV系统就退化为一般的线性时变(LTV)系统;当这些参数为某固定值时,系统退化为线性定常(LTI)系统。而且从实际应用的角度来看,通过沿参数轨迹进行线性化可以将一大类非线性系统转化为LPV系统。因此,有必要研究LPV系统的鲁棒故障诊断与容错控制。目前关于LPV系统的鲁棒故障诊断与容错控制已经引起了广泛关注,成为控制界的热门研究课题之一。本文重点研究了一类LPV系统的鲁棒故障估计、时滞LPV系统的鲁棒故障估计、LPV系统的鲁棒控制器与鲁棒故障估计器的联合设计、以及基于H∞增益调度技术的主动容错控制问题。本文的创新点主要有以下几点:(1)针对一类LPV系统,提出了一种鲁棒H∞故障估计器的设计方法。这类系统的特点是其状态空间矩阵仿射依赖于时变参数向量。本文首先将鲁棒故障估计问题转化为鲁棒H∞控制问题。然后基于增益调度思想和LPV系统的有界实引理,以线性矩阵不等式的形式给出了LPV故障估计器存在的充分必要条件,并给出了自调度LPV故障估计器的详细构造方法,所构造的LPV故障估计器与被检测对象具有相同的参数依赖关系。最后通过计算机仿真验证了该方法的有效性。(2)针对一类时滞LPV系统,提出了一种鲁棒故障估计器的设计方法。这类系统的特点是其状态空间矩阵仿射依赖于一组可实时测量的时变参数,系统中存在未知的状态时滞,且时滞的变化率有界。基于H∞控制理论和时滞LPV系统的有界实引理,以线性矩阵不等式为工具,推导了自调度LPV故障估计器存在的充分条件,并给出了一种与被检测对象具有相同参数依赖关系的鲁棒LPV故障估计器的构造方法,所构造的故障估计器能够有效、快速地跟踪故障信号,并且对状态时滞和外界不确定扰动具有良好的鲁棒性。(3)针对一类仿射参数依赖的不确定LPV系统,提出了一种鲁棒控制器/故障估计器的联合设计方法,即同时设计鲁棒控制器和故障估计器。首先通过引入故障估计误差,将联合设计问题转化为H∞控制问题,再引入虚拟性能块,将鲁棒控制问题转化为鲁棒稳定性问题。然后基于定标H∞理论,给出了LPV系统的定标有界实引理,并在此基础上导出了LPV联合“控制器/故障估计器”存在的充分条件。当条件满足时,给出了该联合“控制器/故障估计器”的构造方法。所构造的LPV联合“控制器/故障估计器”与对象具有相同的参数依赖关系,且能够同时给出鲁棒控制信号和故障估计信号。最后针对一个具有调节器故障的不确定系统进行了仿真研究,验证了方法的有效性。(4)基于增益调度H∞设计策略提出了一种新的主动容错控制方案。首先针对同时出现传感器故障、部件故障和调节器故障的情形,建立了故障系统的LPV模型。然后在假设故障对系统矩阵的影响可以表示为仿射参数依赖的前提下,基于增益调度H∞理论设计了一种可重构的鲁棒LPV控制器,该控制器为故障影响因子的函数,即控制器的调度变量为故障影响因子,可以利用故障检测与分离机构所产生的残差来对故障影响因子进行在线估计。为了验证所提出的主动容错策略的有效性,针对两级倒立摆系统中的电机测速回路故障问题,设计了主动容错控制器。更多还原

【Abstract】 There is an increasing demand for dynamic systems to become safer, more reliable and less polluting to the environment. These requirements extend beyond normally accepted safety-critical systems of nuclear reactors, chemical plants or aircraft, to new systems such as autonomous vehicles or fast rail systems. An effective way to improve the safety and dependability in automated systems is to introduce fault diagnosis and fault-tolerant control. Hence, the study on fault diagnosis and fault-tolerant control technology has both theoretical and practical importance. Due to the existence of modeling uncertainties and unknown disturbance, robust fault diagnosis has received more and more attention in recent years.Linear parameter-varying (LPV) systems are special linear plants whose state-space matrices (i.e., the coefficient matrices of the state-space model) are functions of some vector of varying parameters. An LPV system can be reduced to a linear time-varying (LTV) system for a given parameter trajectory, and it can also be transformed into a linear time-invariant (LTI) system on a constant trajectory. From a practical point of view, a large class of nonlinear systems can be reduced to LPV systems by linearization along the trajectories of the parameters. So it is necessary to study the robust fault diagnosis and fault tolerant control for LPV systems. At present, it has drawn wide attention, and has become one of the main topics in control domain.This dissertation focuses on the robust fault estimation for a class of LPV systems, the robust fault estimation for a class of LPV time-delay systems, the integrated design of robust controller and fault estimator for a class of uncertain LPV systems, and the active fault-tolerant control based on gain-scheduled H∞design strategy. The main contributions of this dissertation can be summarized as follows: (1) For a class of LPV plants which depend affinely on the vector of time-varying parameters, a robust H∞fault estimator design method is proposed. First, the robust fault estimation problem is transformed into a robust H∞control problem. Then, based on gain-scheduled techniques and bounded real lemma for LPV systems, necessary and sufficient conditions for the existence of LPV fault estimators are presented in terms of linear matrix inequalities (LMIs). When these conditions hold, a method for constructing the self-scheduled LPV estimators is proposed, and the resulting LPV estimators have the same parameter dependence as the plants. Finally, the effectiveness of the proposed method is demonstrated through an example.(2) For a class of LPV time-delay plants where the state-space matrices depend affinely on time-varying parameters that can be measured in real-time and the time-delay is unknown but with bounded variation rates, a method for designing the robust fault estimator is presented. Based on H∞control theory and bounded real lemma for LPV time-delay systems, sufficient conditions for the existence of the self-scheduled LPV fault estimators are developed in terms of LMIs that can be solved via efficient interior-point algorithms. As these conditions hold, a robust fault estimator which has the same parameter dependence as the plant is constructed, the resulting estimator can follow the fault swiftly and effectively and with robustness to time-delay and exogenous disturbance.(3) An integrated methd for designing robust controller and fault estimator for a class of uncertain LPV plants which depend affinely on a vector of time-varying parameters is proposed. First, the integrated design problem is reduced to an H∞control problem by introducing the concept of fault estimation error, the resulting robust control problem can be further transformed into a robust stability problem by inserting some fictitious performance blocks. Then, based on scaled H∞theory, a scaled bounded real lemma for LPV systems is proposed. On the basis of this lemma, sufficient conditions for the existence of an LPV integrated“controller/fault estimator”are developed. As these conditions hold, an algorithm for constructing the LPV integrated“controller/fault estimator”is presented. The resulting LPV“controller/fault estimator”has the same parameter dependence as the plants, and can generate both control signals and fault estimations. Finally, to demonstrate the effectiveness of the proposed method, an uncertain system with actuator faults is investigated.(4) Based on the gain-scheduled H∞design strategy, a novel active fault-tolerant control scheme is proposed. An LPV model for systems with all possible sensor, component and actuator faults is presented. Under the assumption that the effects of faults on the state-space matrices of systems can be of affine parameter dependence, a reconfigurable robust LPV controller is developed based on gain-scheduled H∞theory. The resulting controller is a function of the fault effect factors which can be estimated on-line from the residual vector of the fault detection and isolation (FDI) mechanism. To demonstrate the effectiveness of the proposed method, an active fault tolerant controller is designed for a double inverted pendulum system with a fault in the motor tachometer loop.更多还原