【作者】 李俊基；

【导师】 许镇琳；

【作者基本信息】 天津大学 ， 控制理论与控制工程， 2005， 硕士

【摘要】 热风回流炉是表面贴装工艺中的关键设备,回流炉温度的稳定性直接影响着产品的质量与成品率。在热风回流炉的温度控制过程中,被控参数具有时变、非线性、不确定因素等因素。为了提高系统的自适应能力和抗干扰能力,本文使用了一种新型的控制器——模糊神经网络控制器,与常规PID控制器并联,形成并联复合控制,实现了对热风回流炉温度的精确控制,有效地克服了采用单纯的PID控制带来的响应慢、超调大、控制稳定性差等缺点。在设计的过程中,将模糊理论的知识表达容易和神经网络学习能力强这两种优势结合起来,设计了一种具有模糊结构的等价神经网络,构成一个新的网络结构(Fuzzy Neural Networks Control FNNC),解决了传统模糊控制由于隶属函数和模糊规则选取不当造成的控制缺陷。它不仅具有清晰的空间结构,而且具有良好的自学能力和非线性逼近能力。考虑到热风回流炉温度控制的实际特点,使用了一种用于被控对象输出预测的神经网络预测器(Neural Networks Predictor NNP),预测器通过对网络的学习,预测被控对象的未来输出,使控制器预先感知系统输出状态的变化趋势,从而做出相应的调整。利用Matlab仿真软件,建立Matlab仿真模型,分别对常规PID控制、模糊神经网络复合控制(PID+FNNC)、带有神经网络预测器的模糊神经网络复合控制(PID+FNNC+NNP)进行对比仿真实验,仿真实验结果表明:带有神经网络预测器的模糊神经网络复合控制(PID+FNNC+NNP)的控制效果在三者中最优。更多还原

【Abstract】 Sirocco-circumfluence soldering system is the important equipment ofSurface Mount Technology (SMT). The temperature dynamic-characteristic of thesystem has influenced the quality of production directly. In production, theprocedure of the system has many characters such as time variable, nonlinear andindefinite. In the paper, a Fuzzy Neural Networks Controller (FNNC) connectedparallel with general PID controller has been used to improve the adaptive andanti-jamming ability, avoid the general PID control disfigurement, e.g. slowlyresponse, bad control stability and overrun, and get good effects.Fuzzy logic has virtue of expressing knowledge easily, neural networks hasgood self-learning ability. A kind of new controller——Fuzzy Neural NetworksController (FNNC) has been designed by combining the virtue of these twotechniques. In traditional fuzzy control, unsuitable selected membership function andfuzzy rules will induce the control disfigurement, but the Fuzzy Neural NetworksController has avoided it. It has clear structure, good self-learning and nonlinear mapability. Considered the characteristic of Sirocco-circumfluence soldering system, aNeural Networks Predictor(NNP) has been designed to predict the output ofcontrolled system,then the controller will forecast the change trend of controlledsystem ,and adjust the system in advance. At last, establish the model of Sirocco-circumfluence soldering system forsimulation with Matlab, simulation is done for general PID control system, FuzzyNeural Network Control system compound contro(lPID+FNNC) and Fuzzy NeuralNetwork Control system compound control equipped Neural Networks Predictor(PID+FNNC+NNP)with Matlab. The result approved the Fuzzy Neural NetworkControl system compound control equipped Neural Networks Predictor(PID+FNNC+NNP)has the best control effect.更多还原