Research On Formulation Design In Ceramic Products Base With Back Propagation Neural Network

With the rapid development of abrasive industry, it’s called the teeth of Industry, theceramic abrasives which have the advantages of stable in chemistry and excellent mechanicalproperty take up the important position. Formula design is critical. Although the traditionalmethod including orthogonal experiment method and return law analysis deeply depEnd Subson a basic formula and then adjust the content of each component. The method waste plentyof time,money and physical resources. In addition, the factor of influnce among eachcomponent which keep a complex relationship with the mechanical property of ceramicabrasives is not considered.So the traditional method is hard to meet this need. In contrast,BP(Back Propagation)neural network has unique ability of approximating nonlinear functionand generalization, it can find out the complicated relationships between input and output．Soit provides a reasonable and effective method for the prediction of mechanical property of theceramic abrasives．On the basis of study in the theory of Back Propagation neural network and the datas ofthe content of each component, the corresponding hardness, this paper builds the mold ofBack Propagation neural network in predicting hardness of the ceramic abrasives．This paperdevelops the corresponding predicting software using VC++6.0which strictly follow thestandard. The predicting BP neural network mold gets the reasonable parameters from theinflunces of comparing with dferent study rates, dferent nerve cell numbers of hidden layerand dferent study frequency in system average error. The practical operation illustrates thepredicting software runs well. Using this software can traint by sample datas, then save theweights, finally predict the mechanical property. The predicting results illustrate: the averageerror between predicted output value and the measured output value is within1.52％．Thisshows that the BP neural network’s struction is reasonable,fast and accurate. It is a fast andaccurate way to predict the mechanical property of ceramic abrasives.