【作者】 刘璐杨；

【导师】 靖固；

【作者基本信息】 哈尔滨理工大学 ， 计算机系统结构， 2011， 硕士

【摘要】 模糊控制、神经网络控制都是先进控制技术的控制方法,在工业过程控制中获得广泛的应用。实际工业过程往往具有非线性、不确定性、难以建立精确的数学模型等特点,使得单一的一种控制方法难以达到理想的控制效果。如果能够结合两种控制方法,各取所长,优势互补,设计一个通用的模糊神经网络控制芯片对工业生产过程进行合理、有效地控制,就可以缩短系统开发周期,降低成本。模糊神经网络的研究主要包括模糊神经网络理论的研究、模糊神经网络应用的研究和模糊神经网络实现技术的研究。本文主要侧重的是模糊神经网络实现技术方面的研究,利用FPGA嵌入式系统的应用开发技术实现模糊神经网络控制器的研究与设计,并将其封装成为一个专用的IP核,供其他的控制系统使用。首先对模糊神经网络的控制原理和设计中使用的算法进行了深入地研究与分析;其次,利用MATLAB设计多变量模糊神经网络控制器,针对特定的被控对象进行仿真实验,并获得比较理想的控制效果;然后,研究基于FPGA的多变量模糊神经网络控制算法的实现,对控制器进行分层设计。系统的设计模块主要包括模糊化模块、控制规则模块、权值-参数计算模块、去模糊化模块和耦合处理模块等。在设计过程中遵循自顶向下的设计原则,使用Altera公司的软件QuartusII 8.1对各个模块设计进行优化处理,最后进行整个系统的设计综合。两个仿真实验结果表明,基于FPGA的模糊神经网络控制器比MATLAB设计的模糊神经网络控制器性能优良,在利用较少硬件资源的前提下,不仅可以提高控制器的运行速度,还可以改善传统控制器的控制性能。更多还原

【Abstract】 Fuzzy control and neural network control, which are the control method of advanced control technology, access to a wide application in the industrial process control. The actual industrial process often has many characteristics involving nonlinear, uncertain and difficult to establish accurate mathematical model, which can not achieve desired control effect by using a single control method. If we can combine the two control methods with each other in their respective merits. It is designed the general control chip of fuzzy neural network and carried out the effective and reasonable control in the process of industrial production.The control method can cut down the developing period of system and reduce the developing cost of system.The research of fuzzy neural network mainly includes the theoretical study of fuzzy neural network, the application research of fuzzy neural network and the achievement technology research of fuzzy neural network. It is emphasized on the implementation technical research of fuzzy neural network, innovatively used of the application and development technology of FPGA-based embedded system to accomplish the research and design of fuzzy neural network controller, and packaged the controller into a dedicated IP core for the other control system.First and foremost, the paper introduced the prorouthly research and analysis of control theory and design algorithms of fuzzy neural network. In addition, it is designed the multivariable fuzzy neural network controller in MATLAB, whose simulation experiment of the particular charged object is carried out, and it is obtained to the ideal control effect. Last but not least, it is investigated the implementation of the algorithm of multivariable fuzzy neural network control based on FPGA, and it is carried out the hierarchal design of the controller, then it is carried on the simulation experiment with the software of Altera’s QuartusII 8.1. The results of the two simulation experiments show that the performance of the FPGA-based fuzzy neural network controller is more excellent than that of the fuzzy neural network controller designed in MATLAB. The design module of the system includes fuzzification module, control rule module, defuzzification module, coupling treatment module and weight parameter module.It is followed with the principle of top-down design in the design process, and has been optimized the each module design, then it is carried out the whole system design and synthesis. The design can not only raise the speed of the controller, under the premise of the use of fewer hardware resources, but also improve the control performance of the traditional controller.更多还原