【作者】 费有文；

【导师】 王国峰；

【作者基本信息】 大连海事大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 本文以东北制药总厂的磷霉素钠生产重大隐患和危险源的识别和控制项目为背景，通过运用人工神经网络的BP网络模型，实现了动态液位的测量，并对生产过程中液位变化规律进行了总结，对生产重大隐患和危险源进行了分级研究和应用。最后对生产的过程控制中就几种基于神经网络的控制方案作了一下探讨和研究。 文中的BP网络模型都是由三层构成：输入层、隐含层、输出层。节点单元以及各层间的连接强度决定了BP网络的执行情况。文中讨论了网络模型中隐含层神经单元个数的选取问题，输入矢量的标准化处理问题，以及网络连接权值的初值选取问题等，同时还对基本的BP算法进行了改进、优化。设计了基于BP神经网络的动态液位测量模型和生产重大隐患和危险源的识别模型。系统构造于DOS之上，采用C语言实现。 动态液位网络测量模型反应出了影响实际液位变化的因素及其相关程度，为生产的事故隐患提供了简易判定依据，更重要的是为生产工艺的改进打下了基础。生产重大隐患和危险源的网络识别模型对生产中的重大隐患和危险源进行了分级，较好地服务于安全控制和过程控制，保证了生产的安全、高效运行。更多还原

【Abstract】 This paper based the project of recognition and control of major incipient danger & hazards in manufacture of Fosfomycin sodium in Northeast General Pharmaceutical Factory, solved the problem of the measurement of dynamic liquid level, made a summary of the variational rule of liquid level ,graded the incipient danger & hazards by using BP networks of artificial neural networks. At the last part of this paper, it provided several control scheme based on NN only for researches.The network consists of three layers: the input layer, the hidden layer, and the output layer. The nodes and the connectivity between the layers determine the performance of the BP neural network. Discussed are the problem of determining the number of hidden layer’s neural units, standardizing of input vectors and choosing the initial connecting weights etc. Also this paper improved and optimize the basic BP arithmetic, and designed the measurement module of dynamic liquid level and the recognition module of incipient danger & hazards based BP NN. The system constructed on DOS with C program.The measurement network module of dynamic liquid level reflected the variational factors and their degrees, produced a simple way of incipient accident recognition, important of all , it made for bettering craft . The recognition network module of incipient danger & hazards divided them into several graduations, assured of safety-control & process-control.更多还原