Abstract: In order to improve the RDI_+3.15 of vanadium-titanium sinter,BP neural network algorithm was applied to the prediction of low temperature reduction degradation of vanadium-titanium sinter. The samples of the indicator data were divided into input samples and output samples,the input samples includes carbon,basicity,w （Mg O） and FMG ore,and the output sample was low temperature reduction degradation of vanadium-titanium sinter. The relationship between input samples and output samples was explored by using BP neural network algorithm. The results show that BP neural network model is suitable for studying on the low temperature reduction disintegration properties of sinter. It can predict the output samples effectively according to the input samples,and the average relative error is 5.7%,meeting the requirement of prediction precision in engineering practice,which provides guidance for the production of vanadium-titanium sinter.