Abstract: In order to improve production efficiency and reduce production cost, the decarbonization process of RH refining furnace in JISCO was studied based on the full-mixing model.The carbon component prediction model was developed and programmed with C# language.The online operation of the model was realized and the carbon content in the actual production process was predicted.The comparison with the actual production data shows: for the end carbon mass fractionprediction, the hit rate is 100% for the error within ±5×10-6, and the hit rate is 80% for the error within ±3×10-6;for the process carbon mass fractionprediction, more than 90% of the predictioniswithin theerrors of±10×10-6.The influences of initial carbon, oxygen and aluminium contents, as well as vacuum time, on decarbonization are discussed.The following conclusions are obtained:in order to obtain molten steel with a lower terminal carbon content, the initial carbon content must be controlled at a low level and the amount of oxygen blowing is preferably controlled to be greater than the theoretical amount of (200～300)×10-6,at the same time, the treatment time should be controlled within the corresponding interval.

Key words: RH refining, carbon content, predictionmodel, influence factors