The prediction model of end point temperature of molten steel refining in a 40 t ladle furnace has been developed bya modified artificial neural network calculation method. Compared with traditional Back-Propagation (BP) network calculation method

the modified artificial calculation method can increase prediction efficiency and precision. The examination in production situ showed that using modified BP artificial neural network calculation method

the heats percentage with ±5 ℃ error of prediction temperature of molten steel was 90%

while using traditional BP artificial neural network calculation method

that with ±5 ℃ error of prediction temperature only 77% .