Prediction of compressive strenght of high alumina refractories neural refractories by artifical neural network modeling

Abstract

An artificial neural network (ANNs) with modular neural network (MNN) has been created in order to predict the compressive strength of high alumina refractory bricks. The parameters used as inputs for modeling include chemical composition (SiO2%,Al2O3%,TiO2%, Fe2O3%, CaO%, MgO%, Na2O% and K2O%), sintering temperature, brick volume, bulk density and apparent porosity. The output parameter of the artificial neural network is compressive strength. A sigmoid function was used as the transfer function in the model. The feedback of the errors was performed by using back propagation algorithms (BPA). The utility of the model is in the potential ability to predict the compressive strength of high alumina bricks.The optimal result was obtained after 65500 iterations with an average error of 3.07 %.The model has proven that artificial neural networks may be used to aid manufacturing and designing of the refractory brick with properly selected variables.