【作者】 张希平；

【导师】 冯冬青；

【作者基本信息】 郑州大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 本文以玻璃窑炉的温度控制系统为研究对象，探讨了工业过程控制中非线性时变复杂系统的智能控制算法的应用，并且利用MATLAB进行了仿真实验。 在玻璃制品生产过程中，窑炉温度是系统运行中一个主要的监控参数，温度过高或过低都会对产品质量带来不利影响，维持温度稳定是保证产品质量的必要条件。本文首先介绍了玻璃窑炉的结构、工艺流程以及生产对温度控制系统的要求，分析了窑炉温度系统的动态特性。窑炉温度系统是一个自平衡过程，其纯滞后时间和时间常数都比较大，干扰因素较多，对象参数具有非线性和时变性等特点，过程的不确定性较大，因此属于复杂过程系统。控制对象可以被描述为一个单输入单输出时变系统，其传递函数为带纯滞后的一阶惯性环节，滞后时间、时间常数和增益系数等参数在反应过程中随窑炉温度和运行工况的不同不断变化。 玻璃窑炉温度系统现有的控制方法主要是传统的PID控制。虽然这种方法在工作点附近的小范围内能够取得较好的控制效果，但在系统运行工况有较大波动时，由于纯滞后及参数变化的影响，控制效果明显变差，有较大的局限性。因此，人们一直在设法改进对这类系统的控制。对于大纯滞后系统的控制问题，自1957年史密斯提出预估补偿控制以来，出现了多种控制方法，但到目前尚未完全解决。本文在回顾了大滞后过程的各种控制方法及近年来的进展后，基于智能控制理论设计了玻璃窑炉的智能控制方案。 智能控制理论和计算机技术的发展为玻璃窑炉这类变参数大滞后过程的控制提供了新的方法和手段。本文基于智能控制的理论基础，详细讨论了玻璃窑炉的智能控制算法。重点研究了模糊控制、神经网络控制、遗传算法并将三者结合用于玻璃窑炉的温度控制。本文所作的研究工作主要有以下内容： 1．将智能控制技术和参数辨识技术相结合用于玻璃窑炉的温度控制。利用智能控制技术解决时变增益和大惯性过程的控制问题，利用在线参数辨识和最优预测解决时变大纯滞后过程的输出预测问题。 2．用遗传算法离线优化模糊控制器参数，以解决过去在模糊控制器设计中存在的主观性过强的问题，并使模糊神经网络控制器获得性能较好的初始参数。 3．采用新型参数辨识方法，将遗传算法用于过程参数辨识，在线估计系统的时变参数和纯滞后时间，实现过程参数和纯滞后时间的动态跟踪。 4．对遗传算法进行改进，在优化搜索过程中采用双变异算子。即优化过程的前期使用均匀变异算子，后期使用自适应变异算子，在发挥遗传算法全局优化能力的同时提高算法的局部搜索能力。 摘要 5.将模糊推理和神经网络结合，以多层前馈神经网络结构实现基于神经网络的自适应模糊控制。利用两者智能特点的互补性弥补各自性能上的不足，使控制器既具有模糊推理非线性控制的优点，又具有一定的在线自学习能力，以提高控制的智能程度。 本文基于实际生产过程，以解决实际工作中关心的复杂过程的控制问题为目的，就如何将先进的控制算法应用于复杂过程的控制进行了尝试。对所设计的智能控制系统进行了仿真试验，结果表明控制效果较好，控制器具有较好的调节能力、在线自学习能力和鲁棒性。更多还原

【Abstract】 This thesis takes temperature system of glass furnace as research object, discusses the application of intelligent control to nonlinear time-vary complicated system in industrial process, and uses simulation experiments by MATLAB verifying the intelligent control algorithms.The temperature of glass furnace is a main parameter in glass production process; unduly high or low are detrimental to products quality. Maintaining temperature on its set point is the key factor for assuring high glass products quality. The thesis firstly introduces the structure, technological process and temperature control requirement of glass furnace, and then presents a dynamic characteristic analysis of it. The controlled object is a self-balance process with features of large time lag and inertia; all the parameters are changed with the temperature and operating mode of furnace. Therefore, the furnace temperature is difficult to be controlled with the above uncertain factors.At present, the mainly method used in glass furnace temperature control is PID control. Although PID control can achieve satisfied result in small range near temperature set point, it can’t control temperature to its set point rapidly and smoothly at the time of the work condition change by a wide margin due to the influence of time lag and parameters varying. So PID control is subjected to comparatively big limitation. For this reason, researchers try to improve the control of this kind of systems continuously. For the large time lag and inertia system, many control methods have been put forward since the Smith predicator control rose in 1957, but the problem has not been solved completely. So it is of theoretical and practical significance to conduct researches in this respect. After review of control methods for large time lag process and their improvements in recent years, the thesis design an intelligent control scheme based on intelligent control theory.The advances in intelligent control theory and computer technology provide new methods and means for the control of large time lag and parameters varying system like glass furnace. Based on the foundation of intelligent control theory, the thesis expounds the intelligent control algorithms for glass furnace temperature in detail. The research emphases are fuzzy logical control, neural network control, genetic algorithm and the application of combination of the three aspects in glass furnace temperature control.The particular research work of this thesis are as the follows:1. Combine intelligent control technology with parameter identification technology to control glass furnace temperature. Use intelligent control technology to solve the control problem of process with time varying gain and large inertia; Use parameter on-line identification technology and optimal predication to solve the output predication problem of process with large time varying time lag.2. Use genetic algorithm to optimise the parameters of fuzzy logical controller off-line to solve the problem of excessive dependence of experience on designing of fuzzy logical controller and to enable the fuzzy neural network controller has relatively good initial parameters.3. Use new type of optimisation method梘enetic algorithm as parameter identification method to estimate the time varying parameter and time lag of the controlled process and realize the tracking to the dynamic parameters of process.4. Improve genetic algorithm by using double mutation operators in optimum process. In the earlier optimum stage a uniform mutation operator is used and in the later optimum stage an adaptive mutation operator is used, thus enhance the partial search ability of genetic algorithm besides its general search ability.5. Implement adaptive fuzzy logical control based on feed forward neural network through the combination of fuzzy logical inference and neural network. The combination remedies the defects of each in performance and makes the controller possess not only the advantage of non-linear control of fuzzy logical更多还原