【作者】 赵伟志；

【导师】 吴舒辞；

【作者基本信息】 中南林业科技大学 ， 计算机应用技术， 2007， 硕士

【摘要】 胶粘剂是生产人造板的重要材料。胶粘剂生产过程属聚合反应,在影响胶粘剂聚合反应的诸多因素中,最重要的是生产过程的温度控制,温度控制的品质会直接影响产品质量。聚合反应过程既是化学反应过程,又是物理变化过程,聚合机理复杂,表现出非线性、时变、时滞、被控对象模型不确定等特点,对这类工业过程建立精确的数学模型非常困难,单纯的经典PID控制理论在聚合反应釜温度控制上难以取得好的控制效果。本文主要研究胶粘剂生产过程反应釜的温度控制,在分析该生产过程温度变化特点及控制难点,总结不同温度控制策略和不同控制算法具有不同控制精度原因的基础上,将模糊理论的知识表达与神经网络的自学习能力有机地结合起来,提出了一种模糊神经网络控制方法。该方法针对反应釜温度控制系统的特点,采用一种用于预测被控对象输出的神经网络预测器(NNP),预测器通过对网络的学习,使控制器预先感知系统输出状态的变化趋势,从而预测被控对象的末来输出,利用神经网络实现模糊控制器的功能,用反向传播学习算法(BP)来调整模糊神经网络的参数。在MATLAB语言的SIMULINK平台下分别对PID控制、常规模糊控制和模糊神经网络控制进行仿真比较。仿真结果表明：模糊神经网络控制具有更小的超调,无振荡,平稳性好,达到稳定状态的时间短,稳态误差小,其动态特性和静态特性均为最优,由此验证了模糊神经网络这种控制方法应用于胶粘剂生产过程这一复杂系统温度自动控制的可行性。为了进一步的研究,建立了相应的实验室,设计了基于PCI总线板卡的监控系统,该系统实现了数据的采集、分析和保存,控制算法实现,对执行机构的控制,流程显示,棒状图显示,报警日志,参数查询,手动/自动控制切换和系统设置等综合功能。更多还原

【Abstract】 Adhesive is an essential material in producing plywood. Its production belongs to polymerization process. Of all the factors that influence on the polymerization in adhesive production, the temperature control is the most importantance, and its quality can decide on the product quality.The process of polymerization is both the chemiscal reaction process and the physical change process, the mechanism of polymerization is very complex.The process of polymerization have much to characteristics such as nonlinear, time-varying, noise and delay And it is very difficult to establish the accurate mathematical model of the object. It is also very difficult to control the temperature of the industry process using the single classical PID control theory.This paper is devoted to controlling the temperature of reactors in the process adhesive production.On the base of analizing the temperature variational characteristics and controlling difficulties of reactors, and summarizing the reason why currently different controllers with different control precision, this paper present a kind of fuzzy neural network method, which comibines fuzzy knowledge representation with neural network self-learning ability.Due to the characteristics of system, the method use a Neural Networks Predictor(NNP)which makes the controller sence the change trend of its output states in advance through learning from the networks and predicts the next output of controlled system. The method utilizes the networks to implement the functions of the fuzzy controller and adopt back propagation algorithm (BP) to adjust the parameters of the fuzzy neural networks (FNN).The simulations have been done to compare with the PID, routine fuzzy and FNNC control separately by the SIMULINK platform of MATLAB. We can learn from the artificial result that FNNC has lighter exceeding, no shaking, fine station, short time to reach the stable state, little erro of the stable state. So its dynamic characteristic and static characteristic is the most superior. Therefore it has verified that the feasibility of using FNNC in the complex temperature control system such as the process of the adhesive’s production.For further research, a laboratory is established and the supervisory system based on PCI bus board is designed. The system implements the synthesis functions, including the data sampled, data analyzed and saved,control algorithm, controlling the executing outfit,flow displayed,bar-shaped drawing displayed,alarming logs,parameters queried,switching manual/auto control method etc..更多还原