时变丢包率下网络化系统故障检测

发布时间：2018-08-26 08:05

【摘要】：网络化系统是由一个用于数据传输的通信网络将在不同地点的执行器组、控制器组、传感器组以及被控对象连接起来而构成的间歇式闭环反馈系统,进而完成系统所要求的控制任务。但由于网络的通讯能力有限,传输信号在基于数字通讯的信道中进行传输不可避免会发生丢包等问题,这会使网络化系统的性能变差,甚至带来系统失稳等问题。这些动态给系统的分析与综合带来了新的挑战。本文主要研究了时变丢包率下,网络化系统H_∞故障检测滤波器的设计问题。文中将数据丢失现象描述为随机理论中的Bernoulli随机过程,基于以上状态空间模型设计了系统的H_∞故障检测滤波器。本文将丢包率的变化描述为在几个区间内跳变,同时这种跳变又分别服从Markov和semi-Markov随机过程。对处于不同的丢包率区间的情况下的系统采用不同增益的滤波器,这将大大降低设计的局限性。文章还考虑了建模误差或者其他因素导致的模型不确定性情况下的鲁棒控制问题,并基于范数有界的不确定性设计了故障检测滤波器。论文的主要内容及研究成果将由以下部分构成:第2章介绍了两种随机过程,马尔可夫链(MC)和半马尔科夫链(SMC)。同时为了引入半马尔科夫链,又介绍了嵌入式马尔科夫链(EMC)与马尔可夫更新链(MRC),并进行了区分。考虑到后续章节大量的定理证明,在本章同时介绍了Shur补引理与H_∞控制理论。为了分析系统性能,本章还介绍了均方意义下的随机稳定性,σ均方稳定性,非线性系统随机稳定性,以及无穷级数收敛性定理。为了后面两章的故障检测滤波器工作做准备,本章研究了具有传输数据随机丢失情况下,系统的H_∞滤波器的设计与仿真问题。首先将通信网络中的数据随机丢失现象建模为Bernoulli随机分布过程,然后证明了系统在Lyapunov意义下的随机稳定性和H_∞性能。本章基于范数有界不确定性进行了鲁棒性的分析。最后通过一个数值模拟算例的仿真,证明了理论的正确性和所设计滤波器的有效性。第3章针对丢包率可变且变化过程服从Markov随机过程情况下,探讨了H_∞故障检测滤波器设计问题。通过故障检测滤波器作为残差信号发生器,故障检测问题转化为H_∞滤波问题。本章考虑的系统假设在滤波器和系统之间的信道存在数据丢失。以线性矩阵不等式(LMI)形式给出了保证系统均方意义下随机稳定性,且具有H_∞指标γ的定理。考虑到实际物理模型参数具有不确定性,本章基于这样的范数有界不确定性进行了鲁棒性的分析。最后通过隧道二极管物理模型验证了故障检测滤波器设计方法的有效性和考虑不确定性和变化丢包率的必要性。第4章探究了利用semi-Markov链来描述时变丢包率的网络化系统的H_∞故障检测滤波器设计问题。提出了一种更为广泛的Lyapunov函数,其不仅依赖于系统的当前模态,而且依赖于当前模态的驻留时间。每个模态驻留时间的概率密度函数的参数或者类型都可以通过半马核的方法处理。同样考虑到实际物理模型参数具有不确定性,本章基于这样的范数有界不确定性进行了semi-Markov系统的鲁棒性的分析。在仿真部分,选取小车倒立摆模型验证了本章所提出理论方法的有效性。同时阐述了引入semi-Markov随机过程的优越性和考虑丢包率变化的必要性。
[Abstract]:A networked system is an intermittent closed-loop feedback system consisting of an actuator group, a controller group, a sensor group and a controlled object connected by a communication network used for data transmission, thus completing the control tasks required by the system. Packet loss will inevitably occur in the transmission of a message in the channel, which will deteriorate the performance of a networked system and even lead to instability of the system. These dynamics bring new challenges to the analysis and synthesis of the system. This paper mainly studies the design of H_uuuuuuuuuuu Data loss is described as Bernoulli stochastic process in stochastic theory. Based on the above state space model, the H_ fault detection filter is designed. In this paper, the change of packet loss rate is described as hopping in several intervals, and the hopping is subjected to Markov and semi-Markov stochastic processes respectively. This paper also considers the problem of robust control in the case of model uncertainty caused by modeling error or other factors, and designs a fault detection filter based on norm-bounded uncertainty. In chapter 2, two kinds of stochastic processes, Markov Chain (MC) and Semi-Markov Chain (SMC), are introduced. In order to introduce Semi-Markov Chain, embedded Markov Chain (EMC) and Markov Renewal Chain (MRC) are introduced and distinguished. Shur complementary lemma and H_ control theory are The design and Simulation of H_uuuuuuuuuuuuuuuuuuuuuuuuuuuu The validity of the theory and the effectiveness of the designed filter are proved by a numerical simulation example. Chapter 3 discusses the design of H_uuuuuuuuuuuuuuuuuuuu The problem is converted to H_ The robustness of the sample norm bounded uncertainties is analyzed. Finally, the validity of the fault detection filter design method and the necessity of considering uncertainties and varying packet loss rate are verified by the tunnel diode physical model. Chapter 4 explores the H_uuuuuuuuuuu A wider Lyapunov function is proposed, which depends not only on the current mode of the system, but also on the dwell time of the current mode. The parameters or types of the probability density function of each mode dwell time can be dealt with by the hemi-horse kernel method. Based on the norm bounded uncertainties, the robustness of semi-Markov system is analyzed. In the simulation part, a car inverted pendulum model is selected to verify the validity of the proposed theoretical method.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP277

【相似文献】