Fig 5 - uploaded by Xiaoxian Huang
Content may be subject to copyright.
Source publication
Thermal state of iron ore sintering in iron and steel production cannot be revealed straightforward, which is unfavorable for field operations. In this paper, the soft-measuring models were established to extract the feature points through curve fitting method and evaluate the whole state via random forest algorithm. All the models proposed were va...
Context in source publication
Context 1
... ð Þ ; TS i f g ; i ¼ 1; 2; Á Á Á ; N, where N represents the number of samples, the six-dimensional data set x i1 , x i2 , . . ., x i6 indicate TRP, BRP, BTP and its corresponding exhaust gas temperature. TS refers to the thermal state of sintering process, TS 2 y 1 ; y 2 f g represent the label of normal state and fluctuating state respectively. Fig. 5 shows the sketch map of random forest algorithm, in which n refers to the number of decision trees in random ...
Similar publications
Coke oven gas (COG) injection is believed to improve the quality and yield of sinter in iron ore sintering process. A mathematical model is developed to simulate the sintering process with COG injection, particularly focusing on predicting the quality and yield of sinter. The model is validated by comparing the model predictions with sintering pot...
This research put forwards a stepwise parameter optimization approach based on maximum satisfaction aiming at the characteristics of multi-process control indices and cascading in iron ore sintering process. Given the sintering process involves multiple quality indices and the demand of reducing cost of first proportioning and the consumption of co...
Citations
... In view of this situation, the accurate BTP intelligent prediction has received many researchers' interests. 67 Some representative works are as follows. Liu et al. 68 used the gradient boosting decision tree (GBDT) algorithm and decision rules to predict BTP considering process knowledge and data characteristics dynamically. ...
Data-driven soft sensing modeling is becoming a powerful tool in the ironmaking process due to the rapid development of machine learning and data mining. Although various soft sensing techniques have been successfully used in both the sintering process and blast furnace, they have not been comprehensively reviewed. In this work, we provide an overview of recent advances on soft sensing in the ironmaking process, with a special focus on data-driven techniques. First, we present a general soft sensing development framework of the ironmaking process based on the mechanism analysis and process characteristics. Second, we provide a detailed taxonomy of current soft sensing methods categorized by their predictive tasks (i.e., quality indicators prediction, state parameters prediction, etc.). Finally, we outline several insightful and promising directions, such as self-supervised learning and digital twins in the ironmaking process, for future research.
... S INTERING, which is the primary modus to produce sinter ore with high iron content, has received widespread attention. Iron ore sintering includes many chemical and physical reactions [1]. The operating mode is the combustion status of the sintering material layer under a specific situation of the power apparatus. ...
The operating mode is an essential factor affecting product quality and yield of the sinter ore, which inspires the realization of operating mode recognition. Taking burn-through point (BTP) as the decision parameter of operating mode, an operating mode recognition method based on the fluctuation interval prediction is presented. Firstly, combining the principal component analysis and the fuzzy information granulation method, a fluctuation interval prediction model of the BTP is established through utilizing the Elman neural network. Then, the operating mode classification rules are built according to the data distribution of the BTP in the fluctuation interval. Finally, experiments are executed with the data collected from a factory. The results indicate that it can effectively predict the fluctuation interval of the BTP, and then successfully recognize the operating mode. The proposed method provides a valid reference to control the stable operation of the iron ore sintering process.
Sintering process, as a primary modus of the blast furnace ironmaking industry, has enormous economic value and environmental protection significance for the iron and steel enterprises. Recently, with the emergence of artificial intelligence and big data, data‐driven modeling methods in the sintering process have increasingly received the researchers’ attention. But now there is still no a systematic review on the data‐driven modeling approaches in the sintering process. Therefore, in this article, we conduct a comprehensive overview and prospects on the data‐driven models for the purpose of intelligent sintering. First, the mechanism and characteristics of the sintering process are introduced and analyzed elaborately. Second, the detailed research status of the sintering process is illustrated from four aspects: key parameters prediction, control, optimization, and others. Finally, several challenges and promising modeling methods such as deep learning in the sintering process are outlined and discussed for future research. This article is protected by copyright. All rights reserved.
Sintering is an important process in iron and steel manufacturing. FeO content of sinter is a key index for both sintering and blast furnace ironmaking, but it needs manual sampling and testing in practical production. In this research, an online measurement method based on machine-tailed infrared images and convolutional neural network (CNN) for the FeO content is proposed. Firstly, a target detection model of sinter bed cross-section (SBC) is established, to obtain the key frame image (KFI) of SBC in each unloading cycle. Then, the KFI is labeled with the offline chemical analysis of FeO content, and a modified CNN is proposed to build the online measurement model for FeO content. The experimental results show that the proportion of samples with detection error of FeO content within ±0.5% reaches 91.2% and the RMSE is less than 0.3%. It provides a novel measurement approach for online detection of FeO content.
Raceway is a key region in ironmaking blast furnace (BF). While the raceway depth is extremely difficult to measure, thermal images near tuyeres may be available. In this study, inspired by the concept of digital-twin, a soft-sensor approach is proposed to estimate the raceway depth from thermal images. This approach includes (1) The representative thermal images are generated through a raceway CFD model under industry-scale conditions of a specific BF; (2) A principal component analysis (PCA) method is used to reduce data dimension and extract key features from the thermal images of high dimension; (3) A model-learning tool, support vector machine (SVM) is developed to learn the underlying data-driven soft-sensor model between extracted features from PCA and raceway depth from CFD simulations. The result shows that the soft-sensor model can effectively capture the latent relationship between thermal images and the raceway depth, which can be used to estimate raceway depth in real time in practice.
A comprehensive prediction model of sinter quality based on machine learning algorithm was proposed. First of all, the mass historical data of actual sintering production was collected, cleaned and integrated. On this basis, the drum index and screening index of sinter were analysed by cluster analysis, and the quality of sinter was evaluated synthetically by clustering results and iron grade of sinter. Then, the important characteristic variables related to sinter quality index were screened by Recursive feature elimination, stability selection and random forest selection. The comprehensive classification model of sinter quality and the regression model of sinter’s total iron content were established by using various machine learning algorithms. The results show that the prediction accuracy of the classification model and regression model established by the extra tree is the best, and the application effect of the model is verified by using the testing set. The F1-score of the comprehensive quality index classification model is 0.92 and the R² of the total iron content regression model is 0.882. This model has good learning and generalization ability, and the accurate prediction of sinter quality index is realized.
