Open AccessData driven sintering moisture control model based on neural network
Author(s) -
Runbin Shi,
Mengyao Cui,
Yufei Wang,
Shiyuan Liu,
Zihui Li
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1941/1/012056
Subject(s) - nonlinear autoregressive exogenous model , artificial neural network , water content , sintering , moisture , process engineering , computer science , environmental science , artificial intelligence , engineering , materials science , composite material , geotechnical engineering
Sintering production is an important part of ironmaking industry. The moisture content of sinter material will affect the output and quality of sinter. But at present, most sintering plants still use manual water to control the moisture content of sintering material, which leads to a great fluctuation of moisture content, and it is difficult to improve the sintering efficiency. Only stable control in the best range can stabilize the working condition of sintering machine. Therefore, this paper studies the water control scheme and designs the control system based on the existing data, and uses the established data driven moisture control model to control the appropriate amount of water, and compares and analyzes the advantages of the control ability of different models compared with manual water. Firstly, we eliminated the invalid data, found that five materials were significantly correlated with water content, and then established a multiple linear regression model to further study the influence of the proportion of different materials on water content. Secondly, we established a data-driven sintering moisture automatic control model. The neural network model is used, the input variable is the material usage, water addition and moisture measurement data, and the output variable is water addition. The prediction results obtained by simple BP neural network and NARX neural network with feedback are compared and analyzed. It is found that NARX neural network has smaller prediction mean square error, more stable prediction results. Thirdly, we compared the data-driven NARX model prediction results with the results of manually adding water to control the moisture content of sintering, and made quantitative analysis with standard deviation and coefficient of variation. It was found that the standard deviation and coefficient of variation of NARX model prediction results were smaller, so NARX neural network model was more stable and excellent. If this model can be popularized in industrial production, the output and quality of steel will be improved, the production cost will be reduced, the labor resources will be saved, and the enterprise will get more benefits.