The Research And Application Of Fuzzy Mathematics And Neural Network In The Design Of The Cigarette Formulating

Cigarette formulating is a typical uncertain and nonlinear problem because it involves complex data and techniques. Its primary tasks are: The first is grouping effectively the single-material tobacco; the second is building the rational mathematic model of the chemistry index of the cigarette formulating and the chemistry index of smoke; the third is obtaining the best proportion of the tobacco compounding by building the optimized model and solving the problem restrictedly according the cost-least principle. So in this dissertation it is considered that using the integrated application of fuzzy clustering and BP neural network and nonlinear constrained optimization to design the cigarette formulation.The dissertation consists of five chapters.The first chapter puts forward the background of this subject, the research content, the research meaning, the technical course and the point of innovation.Chapter two particularly expatiate the traditional procession of the cigarette classification and its abuse, puts forward the fuzzy clustering method based on the fuzzy equivalence matrix, researches the method of constructing the fuzzy similar matrix and ascertaining the coefficients, ascertains the fuzzy similar matrix fits the cigarette formulating, sequentially groups the single-material tobacco effectively.The third chapter researches the application of the BP network in the forecast of the chemistry attribution of the cigarette group, analyses the reasons for the slow convergence in the application of solving the complex problems of BP arithmetic according to the arithmetic principle of BP network, and compares the astringency of several improved arithmetic from many aspects including activation function, the node-number of concealed layer, weight, study rate and so on, lastly combines the problem of the cigarette formulating and puts forward the topological structure of the BP network and settings of the diversified parameters, solves the hard problem of the sensory-quality assessment and smoke measuring in tobacco product.Chapter four begins from the chemistry property of the single-material tobacco, uses the nonlinear constrained optimization and seeks the proportion of multifarious single-material tobacco to composing the cigarette formulating. Using this method to design the cigarette formulating can bring the better economic benefit and social benefit.Chapter five introduces the integrated application of the three mathematic methods including fuzzy clustering, BP network and nonlinear constrained optimization in the design of the cigarette formulating, explains the design process and the flow of the system in detail, and summarizes the whole dissertation.