The Classification And Optimization Simulation Of Non-asbestos Gasket Formula

Artificial Neural Networks (ANN) is a new algorithm which widely used in pattern recognition, optimization and many others fields, especially in performance prediction, formula optimization of kinds of material. The non-asbestos gasket is a new sealing material, whose formula and performance is always nonlinear relationship, and formula always determines its performance. Therefore, how to predict performance of non-asbestos gasket and how to optimize their formula? This is what we should do in this thesis.The classification of non-asbestos gasket is studied in the first of the paper. The "supervised" Probabilistic neural networks(PNN) and Learning vector quantization(LVQ),"unsupervised" Self organized mapping(SOM) of classification method are all used in the paper, which provides a certain guide of classification of non-asbestos gasket.The performance of non-asbestos gasket is usually obtained by repeated experiments, and its result always inaccurate because some reasons in the experiments, so we have to do it again and again and it will waste much of our time to complete the experiments. A model between formula and performance of non-asbestos gasket is established via BP Neural Networks. The formula and performance is respectively as input and output in the network. Due to the result of BP is not good, some improvement algorithm are proposed to predict performance of non-asbestos gasket. After compare all these algorithms, it is concluded that Levenberg-Marquart algorithm is a best algorithm among all these algorithms, whose number of the iteration is small. Scaled conjugate gradient algorithm and Poweii-Beale reset conjugate gradient algorithm are also can precisely predict performance of non-asbestos gasket. The model of regression analysis is established between the density and formula, the loss on ignition and formula and the intensity and formula, which provide a theory for formula optimization in the paper. Finally, fmincon and fgoalattain function in MATLAB, LINGO and genetic algorithm are all used to optimize formula in order to obtain optimal formula of non-asbestos gasket on the condition of our desirable performance. And the result of LINGO and genetic algorithm is best among these methods.Non-asbestos gasket is simulated in the paper, predict performance and optimize formula of it is proved correct, and it provides relevant theoretical basis and direction for the future research of non-asbestos gasket.