Furnace Condition And Trend Forecasting To Blast Furnace Ironmaking Process

The blast furnace is the main production equipment in the iron-steel industry which isthe very pillar industry in our country. How to ensure the stable operation of the furnace andthe efficient production is a quite important goal. Blast furnace iron-making (BFiron-making)has switched its production control mode from the traditional operatorexperience to data with the rapid development of the computer technology. But thecomplexity of the blast furnace smelting process brings lots of difficulties for its control.Chief difficulties are as follows:1) There are high noises existing in the collected data due ofthe complex environment in process of blast furnace smelting. Foreman can acquire themeasured values of parameters, but the real values is unknown, just depend on the expertexperience to get the estimated values. It is not conducive to the judgment of actual conditionof the blast furnace;2) It is the blast furnace internal temperature and pressure that makes itunable to obtain the current furnace directly. So operator gets a condition judgment of thefurnace only by relying on the other related parameters;3) The complexity of blast furnaceproduction makes data abnormal or lack, which is not conducive to the blast furnaceoperation and post study;4) As a multi-sensor system, the problem of strong coupling, notunified data sampling period and time lag between different data has lead extremedifficulties to the production operation optimization;5) In order to judge the current furnacecondition and its future development well, operators have to refer varieties of parameters,curves and images. Worse yet, dispersion of data is not conducive to the blast furnaceoperation. This paper combines blast furnace state parameter information like Si content,iron temperature, feeding speed with the blast furnace control parameter information like airvelocity, air temperature, spray coal, oxygen enrichment and so on to reflect the blast furnacecondition information indirectly. Then the real-time monitoring of blast furnace is achievedby designing an integrated monitoring system of blast furnace smelting process to assist blast furnace operators with furnace condition judgment and operation. The main researchcontents are as follows:（1） Modeling for blast furnace data based on one variable time series regression isquite different to the conventional. There are high noises existing in the collected blastfurnace data, its true value and noise variance are both unknown. The modeling and analysisare used for data values compensation and prediction. Then the noisy AR model without biasparameter estimation provides the basis for the evaluation of information and data filtering.（2） From the perspective of actual production data of blast furnace, this paperexpounds problems existing in data and makes an analysis of problems that traditional outlierdetection and repair method of abnormal value is not applicable to the complex blast furnacesystem. The proposed new method combining multi-scale with PaTa Criterion for outlierdetection and the application of AR model for data repair and prediction contribute to notonly ensuring the integrity and validity of data, but also providing the data for later research.（3） As a complex multi-sensor system, the blast furnace produces multi-sourceinformation. Analysis and research are conducted to solve the problem that unit andsampling period of different data are different. It is more beneficial for later research toeliminate the influence of dimension and frequency. Meanwhile, we also take information ofthe molten iron parameters into fusion. The fusion information is more comprehensive thanthe single parameter and more accurately to reflect the changes of furnace heat.（4） Data of Blast furnace is recorded in different forms, operators have to refer avariety of data to make a judgment on the current and future development of furnace. Stateparameter comprehensive judgment map and operating parameters comprehensive judgmentmap are designed by eliminating the adverse effects of different data with the above methodto let operators get a clear grasp of the parameters variation on the map.（5） On the basis of the above research, the comprehensive monitoring system ofblast furnace smelting process is designed. With the aid of the system, operators do not need to contact algorithm and program, they can simply run the program directly by clicking themouse to get the processed data and judge the furnace condition more intuitively and clearly.