Research On Stability Evaluation Method Of Blast Furnace Condition

The energy consumption of blast furnaces accounts for 70%of the steel industry.Stable blast furnace condition is the key to reducing costs in the steel industry.Production practice shows that the blast furnace is stable and straightforward in order to achieve high output,low energy consumption,good production indicators and good economic indicators and economic benefits.The influencing factors of blast furnace ironmaking production are numerous and complex,and the operation often relies on personal experience and lacks analysis and consolidation of long-term data.The blast furnace ironmaking expert system that was developed in the 1950s,due to the huge difference in equipment conditions between the blast furnace raw fuels,requires a large number of sensors to collect data during operation,and cannot be applied on a large scale in production.Based on the above problems,this paper analyzes the relationship between blast furnace production data and key indicator parameters through modern calculation and data analysis methods,and establishes a blast furnace condition stability evaluation system to quantitatively judge and evaluate the blast furnace production status.This paper takes a large domestic blast furnace as the research object,and uses data analysis software to clean,integrate,convert,discretize and reduce the original data.Data visualization tools have been developed using the Python language to implement data scatter plots,data fitting,correlation analysis and more.Based on theoretical knowledge and production experience,a static evaluation model of blast furnace stability was established.The output and fuel ratio were selected as technical indicators and economic indicators of blast furnace production to evaluate the model.In order to solve the problem that the static model model can’t self-learn,the paper based on the data mining algorithm,through k-means clustering,Apriori algorithm,BP neural network and other algorithms to analyze the blast furnace data,compare the effects of different processing methods,and realize the characterization of blast furnace process parameters.Results parameters,evaluation of the effect of blast furnace stability.The main conclusions are as follows:(1)Using data analysis,decision tree and principal component analysis to reduce dimensionality of data.Through the correlation analysis to find the redundant data in the original blast furnace data,the decision tree finds the important parameters affecting the furnace condition.The two methods can integrate the parameters from 559 to 425,and then use the principal component analysis method to reduce the parameters.Up to 54 and retain 80%of the information.(2)Based on theoretical knowledge and production experience,the artificial scoring model of blast furnace stability was established.The output and fuel ratio were selected as the technical indicators and economic indicators of blast furnace production to evaluate the model,and the accuracy and recall rate of the model prediction model of the model were obtained.And the accuracy is 89.4%,74.0%,and 72.9%,respectively.(3)In order to improve the recall rate and accuracy of the model,the correlation analysis is used to optimize the parameters and weights,remove the result indicators in the parameters,and retain the parameters that can reflect the operating state of the blast furnace.The accuracy,recall and accuracy of the improved model condition judgment model were 95.0%,86.5%and 88.1%,respectively.(4)The k-means cluster is used to define the blast furnace data label,and the k value is determined to be 3 or 4 according to the elbow method and the contour coefficient method.(5)By k-means clustering the blast furnace data,each attribute is clustered into 8 categories,and then the blast furnace data is classified by Apriori algorithm.The Apriori algorithm can identify normal furnace conditions to a certain extent,but the algorithm is sensitive to abnormal points and cannot identify abnormal furnace conditions.(6)BP neural network algorithm is used to establish the relationship between process parameters and result parameters,and compare the effects of different tag definition methods and PCA dimensionality reduction on algorithm results.PCA dimensionality reduction has no obvious effect on BP neural network algorithm,but can significantly reduce the computation time of the algorithm.The study found that the training model has the highest prediction accuracy by more than 90%through the k-means poly 3 definition label.