摘要
智能控制技术在解决现代复杂被控对象的控制上比传统控制技术有明显的优越性,而基于神经网络的控制由于其所特有的并行处理、自适应、自组织和自学习能力,在自动控制领域中有广泛的应用前景,其中常规PID控制具有算法简单,可以改善系统的动态特性和稳态特性的优点,因而被广泛的应用于拉力控制系统中。但是在拉丝机进行卷取拉力控制时,由于卷取中卷径的变化以及其它的一些不可检测的因素的影响,相应的带来了系统参数的变化,传统PID控制器的参数往往是针对一种情况进行整定,因此很难保持拉力的恒定。针对这种情况,本文设计出一种模糊神经网络自适应PID控制器,根据卷取拉力控制中的实时变化,利用模糊神经网络控制器对PID控制器的参数进行在线自调整,使其能够适应被控制过程中对象的变化,从而实现恒拉力卷取控制。
本文较为详细地阐述了模糊神经网络控制原理。模糊神经网络控制系统主要由模糊控制系统和神经网络控制系统两者结合起来,取长补短,将神经网络的学习能力引用到模糊控制系统中去,充分利用其自组织、自学习能力,实现模糊规则的自寻优及隶属函数的自调整,从而克服神经网络结构难以确定以及模糊控制无自学习能力的缺点,能适应一些对控制要求较高的系统。
该系统在实验室仿真成功,系统工作稳定,操作方便,能够获得满意的性能指标,具有较好的快速跟随性,且稳态精度高,调节时间有了明显改善,提高了系统的抗干扰能力,全面的改善了系统的动态性能。经过一段时间的实际应用,效果良好,明显取得了比常规PID控制器在拉力控制中更好的控制特性。
The technology of intelligent control has more advantage than tradition technology in controlling complex objects of modern industrial control. And the intelligent control as an effective way has great value in the auto control application potentially due to its ability of parallel processing, adapting, self-organizing and self-study. And because the algorithm of the conventional PID controller is simple and can improve the systematic dynamic characteristic and the steady characteristic, it is widely applied to tension control system. But while carrying on the roll tension control, the systematic parameter is changed according to the change of the diameter of the roller and the influence of some other detectable factors. The parameter of conventional PID controller is often exactly designed to a kind of situation and is invariable. So the conventional PID controller is very difficult to keep tension well. Then we design a kind of fuzzy neural network adaptive PID controller. In view of the real change of roll tension control, we use the fuzzy neural network controller to change the parameter of PID controller in time in order to enable it to adapt for the process with the change of object. Then we can realize the permanent tension control.
In this article,we elaborate the fuzzy neural network control principle in detail. The fuzzy neural network control system mainly includes two parts: the fuzzy control system and the neural network system. We can adapt the learning capability of the neural network to the fuzzy control system and use it by the organization and the learning capability to realize seeking superior function of the fuzzy rule and adjust the subordination. The fuzzy neural network control system can overcome difficulty of the define of the neural network structure as well as the shortcoming of the learning capability of the fuzzy control. It can apply to more complicated system.
We simulate this system in laboratory successfully. The system work stably and operate easily. It can obtain the satisfied performance and high stability, follows the source well. The time of control improve obviously. It improves the system anti jamming ability and comprehensibly improves the dynamic performance of system. After a period of practical application, it has obviously obtained the better control characteristic of tension control compared to the conventional PID controller.
引文
[1]陈伯时,电力拖动自动控制系统,机械工业出版社,1992。
[2]陈永红,等,关于一种变参数PID控制器的分析与研究,微计算机信息,2005.10。。
[3]陈振翼,张力控制系统,纺织工业出版社,1998。
[4]李士勇,模糊控制.神经控制和智能控制论[M],哈尔滨工业大学出版社.1998。
[5]方千山,基于模糊盒树法的自适应PID,仪器仪表学报,2002,25(5)。
[6]代军,苏娅芳,张雪,PID算法及其演变,辽宁省交通高等专科学校学报,2005. 5。
[7]蔡自兴,智能控制-基础与应用,国防工业出版社,1998。
[8]常满波,胡鹏飞,基于MATLAB的模糊PID控制器设计与仿真研究,机电车传动,2002.5
[9]李友善,等,模糊控制理论及其在过程控制中的应用,国防工业出版社,1993。
[10]李祖枢,涂亚庆,仿人智能控制[M],国防工业出版社,2003。
[11]刘增良,刘有才,模糊逻辑与神经网络理论研究与探索,北京航空航天大学出版社,1996。
[12]焦李成,神经网络计算,西安电子科技大学出版社,1993
[13]陆奎,汤培榕,基于T-S模型的模糊PID控制系统设计,江南大学学报,2003,2(6)。
[14]马翔,王动伯,一种智能自整定PID控制器的设计与阶梯信号的应用,福州大学学报,2005,51(4)。
[15]斯华龄[美],电脑人脑化:神经网络—第六代计算机(普及本),北京大学出版社,1992。
[16]宋银宾,电机拖动基础,冶金工业出版社,1984.6。
[17]孙增圻,张再兴,邓志东,智能控制理论与技术,清华大学出版社,1997。
[18]谭永红,神经网络自适应PID控制及其应用,模式识别与人工智能,1993。
[19]陶永华,等,新型PID控制及其应用,机械工业出版社,1998。
[20]腾表芳,秦春林,党建武.神经元自适应PID控制,兰州铁道学院学报,2005,22(1)。
[21]肖冰,王印松,杨光军,一种基于专家调节增益的单神经元PID控制[J].控制理论与应用,2005,22(4)。
[22]杨智,工业自整定PID调节器关键设计技术综述,化工自动化及仪表.2000.27(2)。
[23]袁震东,自适应控制理论及其应用,华东师范大学出版社,1986。
[24]夏红,赏星耀,宋建成,PID参数自整定方法综述[J],浙江科技学院学报,2005,15(4)。
[25]薪蕃,神经计算智能基础原理方法,西南交通大学出版社,2000。
[26]赵明旺,现代控制理论基础[M].冶金工业出版社.1995。
[27]张吉礼,模糊神经网络控制原理与工程应用,哈尔滨工业大学出版社,2004。
[28]张志涌,精通MATLAB6.5版,北京航空航天大学出版社,2003。
[29] Coloni A,Dorigo M and Maniezzo V.Ant system:optimization by a colony of cooperatingagent[J].IEEE Trans on Systems,Man and Cybernetics-part B:Cy bernetics,1996,26(1).
[30] Colorni A,Dorigo M and Maniezzo V.Distributed optimization by ant colonids[A].In:Proc.of I st European Conf.Artificial Life [C].Pars,France:Elsevier,1991.
[31] D. Knittel. Modeling and Robust Control of Winding System s for Elastic W ebs [ J ]. IEEE transactions on Control System s Technology, 2001, 10 (3).
[32] Grossberg S. Some Networks that Can Lean, Remenber, and Reproduce any Number of Compialted Space-time Patterns,stud. Appl. Math. 1970,(49).
[33] Hecht-Nielsen, R., The Theory of Backpropagation Neural Network, In Review, 1988.
[34] Holland J H. Adaptation in Natural and Artificial Systems, Ann Arbor: The University of Michigan Press, 1975.
[35] Lee C C.Fuzzy logic in control system:fumy logic[J1.IEEE Trmm Syst,1990,20(2).
[36] Nielson H R.Theory of the Backpropagation Neural Network[J].IEEE UCNN.1989.
[37] Miller G A.Control of Tension in Textile Filament W inding.M echatronies,1995(2).
[38] Nikiforuk P N。Gupta M M.Adaptive Tracking of S1SO Nonlinear Systems Using Multi—layered Neural[ Networks.Proc of American Control Conference,1992.
[39] Stein R B, Leung K V, Mangeron D, Oguztoreli M N. Improved Neuronal Model s for Studying Neural Networks, Kybernetik, 1974,(15).
[40] Sejnowski T. Higher Order Boltzmann Machines. In: Denker J ed. AIP Conf Prpoceeding 151:Neural Networks for Comptuing, New Youk: American Institute of Physics, 1986.
[41] SIEMENS, SIMOVERT MASIER DRIVE矢量大全。