Abstract:Prediction model of molten slag activity was developed based on GA-BP neural network,which had good prediction precision proved by verifying the activity value of different slag system in the condition of different temperatures. On this basis,thermodynamic model of deoxidation of austenite and ferrite stainless steel was also built up. And the model was showed that the higher the Cr mass fraction,the more difficult the deoxidation. And the ability decreased with the increase of the Ni mass fraction on the condition of the deoxidation of austenite stainless steel. The decrease of the activity of deoxidation product in molten slag was beneficial to the reduction of dissolved oxygen mass fraction in steel in any case.