| How to deeply mine and analyze the smelting process data of the converter,and then develop a very effective endpoint dynamic control technology is a key factor that restricts the intelligent control of the converter.Therefore,based on the evolution mechanism of the flame spectrum at the converter mouth,Firstly,a fundamental experimental simulation study was conducted on the spectral radiation characteristics of CO gas basic combustion flames and the evolution law of furnace mouth flame spectral characteristics during different smelting periods under laboratory conditions.Then,based on historical furnace mouth flame spectral data and smelting process data in the actual smelting site,artificial intelligence algorithms were combined to establish corresponding endpoint dynamic control models.The main research process and results are shown below.1)The spectral line of CO basic combustion diffusion flame radiation spectrum is mainly composed of continuous band spectral lines,and the spectral line features a double peak overall.When the wavelength value is greater than 550 nm,the radiant intensity of the flame spectrum shows an exponential decay trend.The increase of calcium oxide and iron oxide dust particles will inhibit the combustion of CO gas diffusion flame,and then reduce the radiant intensity of CO gas combustion flame spectrum.2)In the early and middle stages of converter blowing,the temperature value of molten steel in the furnace is the main factor determining the evolution of flame spectral information at the furnace mouth.As the temperature of the molten metal in the furnace increases from 1250 ℃ to 1400 ℃,the continuous band spectral characteristics gradually become prominent.When the temperature of the molten metal in the furnace reaches above 1450 ℃,the alkali metal potassium ions and sodium ions in the spectral line characteristics begin to exhibit strong spectral peak characteristics at the wavelength value of 590 nm and around 770 nm.3)In the later stage of smelting,when the carbon content of the molten metal in the furnace decreases to 0.6%,the decarbonization rate of the molten steel in the furnace is mainly determined by the mass transfer ability of the molten steel.At this time,the spectral characteristics characterized by the flame spectral information at the furnace mouth show a regular change with the change of the carbon content and temperature value of the molten metal in the furnace.4)In the actual converter blowing process,the flame radiation spectrum data of the converter mouth exhibits different time-domain and frequency-domain characteristics during different smelting periods,and the evolution law of spectral characteristics is mutually confirmed by simulation experimental results.And the flame spectral information collected in the later stage of smelting is highly correlated with the composition and temperature values of molten steel in the furnace,which can provide more accurate sample data support for establishing a smelting endpoint control model.5)The application of segmented least squares fitting algorithm and wavelet analysis algorithm for feature value extraction can achieve perfect characterization of the full spectrum information of furnace mouth flame in the later stage of smelting.By utilizing static smelting models and functional modeling methods,effective big data sample sets can be constructed by matching the furnace flame spectrum big data with smelting process data one by one in time series.6)The SVM algorithm and the SVR algorithm are coupled to build a continuous prediction model of carbon temperature in the later period of steelmaking based on the intelligent mining of the flame spectrum information of the converter mouth.After training and testing,the average absolute error of carbon content prediction of the model is basically within 0.3%-2%,and the average absolute error of temperature prediction is basically within 1%-9%.7)Integrating this model into the smelting module of the automatic steelmaking system for industrial testing,after tracking and statistics of smelting data within four months,it can be found that after incorporating the steelmaking dynamic prediction model designed in this article,more than 90% of carbon pulling was achieved at once,and the double hit rate of endpoint carbon temperature reached about 85%.Figure 91;Table 45;Reference 136... |
PDF Full Text Request