双网环境下工业过程运行反馈控制研究

发布时间：2018-03-20 17:40

本文选题：网络控制系统　切入点：工业过程　出处：《哈尔滨工业大学》2016年博士论文　论文类型：学位论文

【摘要】：随着全球化经济的发展,尤其是进入工业4.0时代,工业生产过程的规模越来越大、复杂程度越来越高。同时,随着国际化市场竞争越来越激烈,使得各个工业生产制造企业更加注重提高本企业的生产能耗效率、产品质量水准、提高产品性价比和竞争力。针对这种情况,并考虑设备层和操作层的网络传输,本文基于双层的工业生产过程结构给出了双网环境下的工业生产过程控制结构。在该结构中,设备层子系统主要考虑为非线性系统,利用神经网络函数构造指标预测函数,并结合非线性模型预测控制方法对操作层动态系统进行优化和补偿。本文以非线性控制方法为基本工具,结合Lyapunov函数设计、容错控制方法、随机系统稳定性理论等对双网环境下的非线性工业过程控制问题进行了初步研究。其主要工作可以概括如下:(1)在传统的设备层采样控制器设计过程中,主要利用采样信号和相应的控制器增益构建采样控制器,并通过零阶保持器,使得传输到执行器的控制信号为连续信号。然而由零阶保持器的特性可知,控制信号在一个完整的采样周期内为常值,在整个运行时间内的控制律则为相应的分段函数。在本文中,首先针对一类一般的非线性采样系统设计模糊状态估计器,其主要思想是利用模糊估计器中的估计状态和原系统中的采样状态共同构建控制器,在此框架下构建的控制律为连续信号,不仅利用了采样信号信息,并且利用了模糊估计器的状态信息,因此具有更好的鲁棒性,并且适用于大采样周期的情形。在针对标称系统控制设计的基础上,进一步考虑非线性严格反馈系统的H∞控制问题,通过结合上述的模糊状态估计器,预设H∞性能指标,模糊逻辑系统,以及反步法反馈技术设计自适应模糊控制器保证系统的稳定性。(2)针对双网环境下的工业过程,首先考虑非线性设备层子系统的控制问题。本文主要利用T-S模糊模型、神经网络函数或模糊逻辑系统对非线性系统(系统中未知非线性项)或相应的状态观测器建模,在此基础上基于操作层采样周期对向上传输通道的信号进行采样,并利用径向基神经网络函数构建指标预测函数,而针对设备层和操作层的网络传输通道,则考虑其存在双向随机丢包和网络诱导时滞,即设备层到操作层向上的传输通道和操作层到设备层向下的传输通道。进一步给出系统的预设性能函数,结合上述智能控制方法和模型预测控制策略设计混合智能控制优化方法保证闭环系统的稳定性。(3)若设备层系统存在随机故障和随机网络诱导时滞,则用两个独立的马尔科夫过程描述这两个随机过程,设计控制器保证离散随机系统的随机稳定性,在此基础上基于操作层采样周期对向上传输通道的信号进行采样,并利用径向基神经网络函数构建指标预测函数。给定操作层的设定值动态方程为一般情形下的非线性离散动态系统,将设定值和指标预测误差分别看做操作层系统的状态变量和干扰变量,基于各个变量的约束并将两个传输通道建模为单一传输通道设计补偿器,优化系统性能。(4)在前三个主要研究内容的基础上,考虑到某些复杂非线性工业过程系统模型很难获得,因此设计基于数据驱动的控制器和优化补偿器。通过优化两个目标代价函数:操作层跟踪误差代价函数和操作层指标预测误差函数,分别设计补偿器和优化算法,不仅保证系统设备层的跟踪性能,并可以基于预设的性能参数指标对操作层综合性能进行优化。最后给出本文的结论,总结了本文的主要研究内容和创新点,并对本文中的不足以及待解决的问题做了简要分析。
[Abstract]:With the development of economic globalization, especially in the industry of 4 times, the process of industrial production scale is growing, more and more complex. At the same time, along with the internationalization of the increasingly fierce market competition, the industrial manufacturing enterprises pay more attention to improve the energy efficiency of the production enterprises, product quality, improve product performance and competitiveness. In view of this situation, and consider the network transmission equipment layer and operation layer, the control structure of the industrial production process network under the environment of industrial production process is given based on the double structure. In this structure, the equipment subsystem mainly consider a nonlinear system, predictive function using neural network function index, and combined with the nonlinear model predictive control methods of optimization and compensation on the operation layer dynamic system. Based on the nonlinear control method as the basic tool, combined with the letter Lyapunov The number of design, fault tolerant control method, stochastic system stability theory conducted a preliminary study of the control problem of nonlinear industrial process network environment. The main work can be summarized as follows: (1) in the traditional equipment layer sampling controller design process, the main use of the sampling signal and the corresponding controller gain construct sampling controller and maintain device the zero order, the transmission control signal to the actuator. However, for the continuous signal by the zero order hold the characteristics of the control signal in a complete sampling period in constant value. The control law in the whole operation time is piecewise function accordingly. In this paper, firstly, for a class of general the sampling system design of fuzzy nonlinear state estimator, the main idea is to use fuzzy sampling state estimation of state estimator in the original system and the construction of common controller under this framework The control law is constructed for the continuous signal, not only the use of the sampling signal, and the use of state information of fuzzy rules, so it has better robustness, and is applicable to large sampling period. In nominal control design based on the system, to further examine the H control problem into nonlinear strict feedback systems, through the combination of fuzzy state estimator and the preset H 2 performance, fuzzy logic system, and adaptive backstepping feedback design technique of fuzzy controller to guarantee the stability of the system. (2) in the industrial process network environment, consider the problem of nonlinear control equipment level subsystem. This paper firstly uses T-S fuzzy model, neural network and fuzzy function the logic system of nonlinear system (unknown nonlinear system) or the corresponding state observer on the basis of modeling, based on the operation layer sampling week Period sampling to the signal transmission channel, the radial basis function neural network to construct the predictive index function, and the network transmission channel layer and operation layer of equipment, is considering the existence of two-way random packet loss and network induced delay, namely the equipment layer to the operation layer to the transmission channel and the operation layer to the device layer further down the transmission channel. This system combined with the preset performance function, control method and model predictive control strategy design of hybrid intelligent control optimization method to ensure the stability of the closed-loop system of the intelligent. (3) if the equipment layer systems with random failures and random network induced delay, with two independent Markov process to describe these two random the controller design process, guarantee the stochastic stability of discrete stochastic systems, based on the operation layer sampling period to the signal transmission channel. Like, the radial basis function neural network to construct the predictive index function. Given the operation layer set value dynamic equation is a nonlinear discrete dynamic system under normal circumstances, the setting value and index prediction error respectively as state variables and disturbance variables operating system layer, based on each variable constraint and the two transmission channel modeling for single channel design of compensator, optimize the performance of the system. (4) based on the three main research contents, considering that it is difficult to obtain some complex nonlinear industrial process system model, so the design and optimization of controller based on data driven compensator. Through optimizing the two target cost function: operation layer tracking error cost function and the operation layer index prediction error function, respectively compensator design and optimization algorithm, not only guarantee the tracking performance of the system equipment layer, and can be set up based on the performance of the pre The performance of operation layer is optimized by parameter index. Finally, the conclusion of this paper is given, and the main research contents and innovations of this paper are summarized, and the shortages and problems to be solved in this paper are briefly analyzed.

【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP273

【相似文献】