The Simulation Research Of The Andalusite Refractories' Erosion Resistance To Liquid Copper

Copper melting furnace materials for metallurgical industry faced harsh conditions and melting process requirements must have excellent physical and chemical properties, the most important property is the corrosion resistance, it directly determines the service life of the lining.In recent years, the andalusite with excellent corrosion resistance and volume stability has attracted wide attention.Therefore, exploring refractory materials based on andalusite to improve the corrosion resistance mechanism of lining material, is of great significance. Based on this, the construction of Artificial Neural Network BP network architecture would establish the relationship between the original mineral composition and production parameters with the corrosion resistance,promote research and development of materials, make the process efficient and diagiostnc.In this paper, the bauxite, andalusite, brown fused alumina, active ?-Al2O3 powder, SiO2 powder were used as raw material, boric acid as the flux, dextrin as binder, the effects of different andalusite particles additions on the properties of lining refractories of a melting copper induction furnace(8%, 16%, 25%, 33%, 42%, 51%)were discussed. X-ray diffraction(XRD) was used to test the phase composition change of the sample after joining andalusite. Scanning Electron Microscope(SEM) was used to observate and analyse the crystal morphology of sample after firing. The research would make use of neural network BP algorithm establishing models of predicting the corrosion resistance of refractories based on andalusite according to previous summary.The results show that:(1) With the increasing of andalusite content, compressive strength and bulk density trended similar to show two peaks. The changing trend of apparent porosity and firing linear changing rate was just the opposite.When the content of andalusite was 16%,corundum and vitreous phase made the structures become tight. When the content of andalusite was 51%, the increasing of content made the process of I mullitization and II mullitization generate a lot of mullite, mullite network structure could be strengthened.(2) When the content of andalusite was 25%, the specimen had the best corrosion resistance properties.The mullitization of andalusite formed network structure, mostly SiO2 liquid filled the pores, the remaining SiO2 liquid phase pushed into the crystalline surface continues to react with Al2O3 to complete mullite network structure.(3) The manufacturing processes of refractory material and their characteristics determine the refractory properties parameter can only be estimates based on data distribution, therefore, it is difficult to accurately predict with analysis formula.In this paper,the BP network models of the andalusite refractories' erosion resistance property to liquid copper illustrates the applicability of neural network for prediction of performance of refractories.The relationship between chemical composition and properties according to the models was completely consistent with the actual theory.These results showed that ANN method is a an effective tool to indicate the complex non-linear relationship between composition and properties of refractories based on andalusite.