【作者】 魏玲；

【导师】 薛定宇；

【作者基本信息】 东北大学 ， 模式识别与智能系统， 2008， 博士

【摘要】 网络控制系统(Networked Control Systems, NCS)是指在某个区域现场传感器、控制器以及执行器和通信网络的集合,用以提供设备之间的数据传输,使该区域内不同地点的用户实现资源共享和协调操作,是一种全分布式、网络化的实时反馈系统。网络控制系统因其具有高诊断能力、安装与维护简便、能有效地减少系统的重量和体积、增加系统的灵活性和可靠性等诸多优点,正成为控制领域研究的新热点,也是未来工业控制系统发展的新方向。但是将网络引入控制系统后,由于带宽有限、信道分时复用、数据传输路径不确定等因素的存在会导致信息传输中的等待、阻塞、时序错乱及丢包等现象的发生,不可避免地产生了网络诱导延时。这些受网络通信协议、网络节点驱动方式、数据采样技术等因素影响的网络诱导延时,无论是固定的还是时变的,也无论是有界的还是无界的,都会不同程度地影响控制系统性能,甚至导致系统失稳,给网络控制系统的分析和设计带来很大困难。虽然对于延时系统的分析、建模与稳定性研究近年来已经取得了很大进展,但是由于网络控制系统自身的特点和复杂性,已有的控制理论与方法必须要重新分析和设计才能应用于网络控制系统中。具体而言,就是要在网络控制系统的分析和设计的过程中要考虑控制与通信耦合的情形。因此,要使网络控制系统能够稳定运行并且具有良好的动态性能,就必须建立网络控制系统的数学模型,同时发展与网络控制系统相适应的分析和设计方法,所以本课题的研究具有积极重要的理论研究意义和实际应用价值。本文的主要目的是研究网络诱导延时对网络控制系统分析、建模以及稳定性的影响,并进一步探索有效的解决方法,对于网络控制系统的理论研究和实际应用提供了有意借鉴。具体研究成果如下：首先,分析了近年来网络控制系统研究中的一些典型研究方法。一方面,这些研究方法中的某些方法已在本论文的研究过程中有所体现,另一方面,对于今后网络控制系统的研究工作有所借鉴。其次,较为详细地分析了网络控制系统研究中的基本问题,从中挖掘出本文所感兴趣的研究方向,也为以后的研究工作设定了目标,同时也希望这些基本研究问题能够引起更多同行的重视,继续对网络控制系统进行理论和实际应用研究,进而形成一套完整的适用于网络控制系统自身特点的理论和方法。第三,研究了网络控制系统中的网络诱导延时问题,这是网络控制系统研究中的一个基本而又非常重要的问题。具体分析了延时的组成、对于系统稳定性和性能指标的影响、网络传输数据的特点和影响延时的因素等,并给出了网络协议、节点驱动方式和采样周期的选择标准,在此基础上对具有定常延时且延时小于一个采样周期的网络控制系统的稳定性进行了分析和设计。第四,研究了网络控制系统的建模问题。首先给出了单输入单输出网络控制系统连续数学模型,在此基础上,建立了带有系统噪声和观测噪声的多传感器多执行器网络控制系统连续数学模型,得到一个存在多个延时的微分方程；建立了短延时、单包传输和多包传输且无数据包丢失情况下的离散数学模型；建立了基于增广状态的任意有界延时的多输入多输出闭环网络控制系统离散数学模型；建立了网络控制系统在四种不同节点驱动方式下的开环离散数学模型和统一离散数学模型,模型拓展了原有延时小于一个采样周期的假设条件,并考虑了系统噪声和观测噪声以及控制器的动态特性；将大于一个采样周期的随机时变网络诱导延时分解成恒定延时(采样周期的整数倍)和随机时变延时(小于一个采样周期)两部分来研究,从而建立了具有结构不确定性的网络控制系统离散时间模型。第五,提出了利用区间代数理论来对具有时变采样周期和时变延时的网络控制系统进行建模和稳定性研究。该方法把原始时变的闭环系统模型稳定性分析问题简化为一个时不变矩阵来研究,得出的结论虽然有些保守,但其分析简单方便。第六,研究了具有数据包丢失的网络控制系统建模和稳定性问题,主要从理论上推导出了使系统渐近稳定的条件,所得结论还有待于在工程实践中加以检验。首先,在给出了单包传输有数据包丢失的网络控制系统离散数学模型的基础上,建立了多包传输有数据包丢失的网络控制系统综合被控对象离散数学模型和闭环系统离散数学模型。接着,考虑网络诱导延时与数据包丢失同时存在的情况,分别建立了单包传输和多包传输的网络控制系统连续数学模型。在此基础上,对网络信息单包传输时设计了延时无界和延时有界情况下的网络控制系统渐近稳定条件；在多包传输时设计了延时变化率在小于1和小于0情况下的网络控制系统渐进稳定条件。最后,通过在节点的发送端或接收端设置队列型缓冲区的办法将随机延时确定化,并针对状态不可测的系统设计了三种延时状态观测器,从理论上证明了只要原系统可观,则观测误差渐近收敛。更多还原

【Abstract】 Networked Control Systems (NCS) are the combinations of sensors, controllers, actuators in some area and the communication networks, providing the data transmission among the equipments and making different users in different places realize resource share and coordinated operations. Networked control systems are the entirely distributed, networked, real-time feedback control systems. Because of the stronger diagnosis capability, the simple installation and maintenance, reducing the weight and the volume effectively, increasing the agility and the reliability of systems, and so on, networked control systems are becoming the new hot research points and new trends of industrial control systems. When networks are introduced into the control systems, networked-induced delays are produced inevitably by waiting, collision and congestion, making data transmission be in disorder and resulting in data dropout, because of bandwidth constraint, time-sharing multiplexing, uncertainty in transmission path, and so on. The network-induced delays, which are affected by network protocols, driven modes of nodes, data sampling methods, and some other factors, whether it is constant, time-varying, bounded or unbounded, can influence the performances of control systems. Sometimes they can cause system’s instability, which brings difficulties to the analysis and designs of NCS. Although the researches have obtained many results in the analysis, modeling, and stability of delay systems, because of the own characteristics and complexities of NCS, the existed control theory and methodology can only be used to NCS after being reanalyzed and redesigned. Specifically, the case coupling of control and communication must be taken into account during the analysis and designs of NCS. It is necessary to establish the mathematic models so as to make NCS be run steadily and have well dynamic performances. At the same time, it is necessary to develop the methodologies of analysis and designs corresponding to NCS. So this dissertation has very important theoretical significance and practical application worth.The main purpose of this dissertation is to study the influences of network-induced delay upon the analysis, modeling and stability of NCS, and to explore effective solution methods, so as to provide meaningful use for reference. The following outlines the research results.Firstly, some typical research methods of NCS in recent years are analyzed. Some of them are listed in this dissertation. Also they can be used for future research work of NCS. Secondly, some basic issues of NCS are analyzed in detail. Some interesting research aspects are dug and can become the future study trends. At the same time, hoping that these issues can arouse the recognition of researchers who can continue to study the theory and practical applications of NCS, the theories and methodologies corresponding to the characteristics of NCS can be made perfectly.Thirdly, network-induced delay of NCS is studied. This is a basic and important issue of NCS. The analysis and research of the delay’s composition, the influences to the stability and performance indexes of NCS, the characteristics of data transmission in networks and the factors of influencing on delay are investigated. The choice standards of network protocols, driven modes of nodes and sampling period are discussed. Based on the above discussions, the analysis and designs of stability of NCS with deterministic and constant delay which is less than one sampling period are studied.Fourthly, the modeling of NCS is presented. Based on the continuous models of single-input-single-output NCS, the general continuous models of multi-input-multi-output (MIMO) NCS with system noise and observation noise are built. The discrete models of short delays, single-packet transmission and multiple-packet transmission with no packets dropout are established. The discrete models of MIMO NCS with any bounded delays based on the augmented states are constructed. The open-loop mathematic models and unified mathematic model of NCS with long time delay are developed when the system is driven by anyone of the four different driven modes. The random time-varying delay is divided into two parts which are constant delay and time-varying delay, and then, discrete uncertain model of NCS is built.Fifthly, based on the theory of interval algebra, modeling and stability of NCS with time-varying sampling period and time-varying delay are studied. This method simplifies the stability analysis of original time-varying closed-loop system model which is changed into a time-independent one. The deduced conclusion is simple and effective although it is conservative.Sixthly, modeling and stability of NCS with data-packet dropout are studied. The asymptotic stability criteria of NCS are derived theoretically. The achieved conclusions need to be verified in the engineering applications. First of all, based on the discrete model of single-packet transmission with data-packet dropout of NCS, the discrete models of associated plant and closed-loop system considering multiple-packet transmission with data-packet dropout are set up. And then, considering the network-induced delay and data-packet dropout synthetically, asymptotic stability conditions of NCS with unbounded delays and bounded delays based on single-packet transmission are derived. Based on multiple-packet transmission, asymptotic stability conditions with time-delay variate rate which is less than one or less than zero are deduced.Finally, the method which can change the random delay into the deterministic delay by means of installing queue buffer at the sending node or receiving node is studied. Three kinds of state observers of NCS whose states are unmeasured are designed. The fact that observer errors are asymptotically convergent is proved theoretically only if original system is observable.更多还原