Gas Flow Distribution Identification Method Of Blast Furnace Based On Intelligent Algorithm

Gas flow distribution patterns of blast furnace plays a decisive role in the productionprocess, good gas flow distribution can not only guarantee the stability of furnacecondition, extend the service life of blast furnace, but also can increase the indirectreduction efficiency inside the furnace, improve the utilization rate of gas, reduce the cokeratio, achieve a goal of saving energy and reducing consumption, high yield and goodquality. To distinguish gas flow distribution patterns currently has a great guidingsignificance to the operation of distributing materials, however, because of the complexenvironment of blast furnace, makes certain difficulty to the distribution of gas flowidentification.For blast furnace gas flow distribution in the direct measurement approach is notfeasible in the industry, and indirect measurement is difficult.This paper studied themutual restriction relations between the cross temperature measuring data and top videoimage data and put forward a kind method of the center of the gas flow distribution, basedon the cross temperature measurement edge gas flow distribution and the camera at topvideo image respectively to identify for consolidation.Secondly, using the ellipse parameter to input into the genetic algorithm to optimizeBP neural network algorithm, and with the blast furnace expert evaluation result as theoutput of the images, the network was trained, concluded that normal development center,development center development and owe three center identification model of gas flowdistribution. At the same time, inputting the edge three temperature data into the hybridneural network model, the network was trained to get the edges of normal development,edges through development and underdeveloped identification model of three kinds ofedge gas flow distribution. In the end, the center of the gas flow distribution identificationmodel and edge gas flow distribution model were made the fusion recognition, gotten ninedifferent distribution patterns of gas flow as a whole. Finally, using the furnace top camera images to get the isotherm diagram, threedimensional histogram, false color map reference images, other more intuitive imageguidance was drawn for the production process.According to the proposed recognition model of the distribution of gas flow thesimulation results has carried on the simulation software and showed that the proposedmethod has better recognition performance, recognition rate of93%, in the iron and steelsmelting process can meet the detection for gas flow distribution pattern. The gottenindustry reference images can be more intuitive to response the concrete smeltingcondition, give the actual working condition of smelting blast furnace expert a strongerbasis.