Study On Temperature Forecast And Control Model Of RH Vacuum Refining

RH vacuum refining can produce high-quality steel. It is one of the most effectivemeans of producing ultra-clean steel at present. As the connection process between theconverter and continuous casting, temperature regulation by RH refining process is thekey to a good match in steelmaking process. Good control of the end of the RH refiningprocessing temperature, which is beneficial to the stability of continuous castingpouring temperature, can improve the quality of steel.In this paper, the changes of temperature of molten steel during vacuum refiningprocess of250ton RH-MFB. The results showed that:(1) Carbon and oxygen reaction and the heat that was absorbed by gas influencedthe temperature of molten steel in vacuum process. And decarburization reaction wasthe important factor. The degree of influence was proportional to the amount ofdecarburization. As less heat was taken away by gas which only reduced thetemperature of molten steel0.5℃, so it could be ignored.(2) The effect of blowing oxygen on molten steel temperature could be mainlymanifested in two aspects: first, the dissolved oxygen into the molten steel released theheat during blowing oxygen; second, the oxygen-aluminum reaction heat raised thetemperature of molten steel.(3) When the alloy is added into the molten steel, melting transition endothermicand oxidation exothermic of the alloy could occur. The temperature change of moltensteel mainly depended on the variety and the amount of alloy elements.(4) The effect of vacuum chamber and RH ladle of molten steel temperature wasvacuum chamber and ladle refractory material heat storage and their radiation heatrelease.In this paper, a dynamic neural network temperature prediction model based on thefield data has been developed using the neural network method. Combining theprediction model with the RH-MFB operating process, a refining temperature forecast and control model of the process also has been developed. The test results of Modelshowed that:(1) The heats which the error were within5℃between the target temperature andpredicted temperature calculated by the NARX neural network model reached89.5%.(2) The developed model for temperature prediction could accurately forecast andwas used to control the terminal temperature of RH-MFB refining, and also couldprovide the guidance for the calculation of alloy addition amount.