The Performance Forecast Of IF Steel Based On Neural Network

Interstitial free steel(IF steel) with its excellent mechanical properties of materials has been widely used as automotive steel plate. Now that microstructure is the direct factor in determining the mechanical properties of the material, therefore, an appropriate model need to be selected to predict the mechanical properties of IF steel which is of great significance for the development and design of steel grade.In order to establish the forecasting model of the relationship between the microstructure and mechanical properties of IF steel, firstly, the quantitative microstructure of IF steel is used as the input variables of the forecasting model in this paper, and feature selection method based on mutual information is used to analyze the correlation between input variables and output variables(mechanical properties of IF steel) which can select the reasonable input feature for the forecasting model, through the the simulation we can known: The yield strength of IF steel and ferrite grain s size, ferrite grain s shape factor, second phase particle s size, average spacing are related, so we select these four parameters as the input variables of the forecasting model of yield strength of IF steel; The tensile strength of IF steel and ferrite grain s size are related, so we select the ferrite grain s size as the input variables of the forecasting model of tensile strength of IF steel; The elongation of IF steel and ferrite grain s size, ferrite grain s uniformity, second phase particle s size, average spacing are related, so we select these four parameters as the input variables of the forecasting model of elongation of IF steel; The correlation between the value of r of IF steel and ferrite grain s size, texture {111}<112>>{111}<110>>{001}<110>>{112}<110>>{554}<225> is larger, so we select these six parameters as the input variables of the forecasting model of value of r of IF steel; The value of n of IF steel and ferrite grain s size are related, so we select ferrite grain s size as the input variables of the forecasting model of value of n of IF steel.After the input variables of the forecasting model is determined, then, the structure of adaptive neural-fuzzy inference system(ANFIS) is identified by subtractive clustering algorithm and the network parameters are trained by hybrid learning algorithm; Finally, the ANFIS network and BP network are respectively used to establish the forecasting model of relationship between microstructure and mechanical properties of IF steel, the model is simulated and verified. The result of the simulation proves that the forecasting model based on ANFIS network in terms of speed of convergence and accuracy of modeling is superior to the traditional BP neural network.