Research On Recognition Of Converter Blowing State And Batching Optimization Based On Flame Image

Since the 13th Five-Year Plan,my country's steel industry has gradually entered a stage of transition from high-volume development to high-quality development.At present,80% of my country's steel output comes from converter steelmaking.Among them,the oxygen topblowing converter is a widely used in China.In the converter steelmaking process,the most important link is blowing,that is,blowing high-purity oxygen into the furnace through the oxygen lance on the top.The conversion process is divided into three periods: the first,middle and last periods.The corresponding relationship between the bath temperature and carbon content and the blowing period is an important basis for judging whether the furnace can obtain qualified molten steel.When the corresponding relationship does not meet the standard,it is necessary to adjust the ingredients.Due to the lack of dynamic monitoring equipment on site,the converter is still controlled by manual experience in production.However,the operating conditions of the converter are complex,this method lacks objectivity,and it is difficult to produce high-quality steel.Therefore,the dynamic monitoring of converter operation study is a difficult problem that needs to be solved urgently in the field of steelmaking.In this context,this paper mainly studies the problem of converter blowing state recognition and batching optimization based on flame images.The main work is as follows:(1)Aiming at the subjective problem of manually judging the blowing period,collect the flame images of the converter furnace mouth,and establish a converter blowing state recognition model based on the DenseNet deep network.The original network is tailored to improve the calculation speed.The loss function of the classification layer is introduced into the Center function,and finally the converter is classified during the conversion period,and the average accuracy can reach about 91.6%.(2)The composition of molten steel during converter operation is the basis for judging the quality of molten steel,but it can only be tested after the oxygen blowing is completed in field production,which has a hysteresis.To solve this problem,the sampling data is extended,combined with the flame image,and the support vector machine with differential evolution algorithm to optimize the parameters is used to establish the molten steel carbon content and temperature detection model.The two feature vectors of the local binary mode and the directional gradient histogram of the fusion image are used as input to detect the molten steel and carbon content.The detection hit rate of carbon content and molten steel can reach about91% and 94%,respectively.(3)The operation process of the converter is complicated.When the quality of the molten steel does not meet the standard,the operator needs to adjust the ingredients,mainly the increase or decrease of the oxygen blowing amount and the cooling amount.Using fuzzy theory,an oxygen blowing optimization model based on expert experience is established to obtain the optimized compensation amount of oxygen blowing,and at the same time,the optimized amount of coolant is obtained through the law of temperature rise of molten steel decarburization in the converter.In this study,the oxygen top-blowing converter of a steel plant is taken as the research object,the molten steel quality target is considered,combined with the expert experience,the flame image is used to identify the blowing state and the carbon content and temperature of the molten steel,and proposed a method of batching optimization in the carbon-drawing stage of the blowing process.The experimental results are within the empirical setting interval,which can provide production guidance for the converter steelmaking site.