Soft Sensors Based on Adaptive Stacked Polymorphic Model for Silicon Content Prediction in Ironmaking Process

In blast furnace (BF) ironmaking process, the silicon content of hot metal is regarded as an important index to evaluate the stability of blast furnace and the quality of pig iron. In reality, the reactions inside BF with the features of dynamic, nonlinear and hysteretic impede accurate mathematical modeling. Data-driven models predict silicon content variation tendency through the analysis of historical data, which have achieved beneficial results. However, existing data-driven models are inefficient to handle time series specifically. Therefore, this paper proposes a recurrent neural network (RNN) which is apt at processing sequential data for silicon content prediction. A Long short-term memory (LSTM) unit is introduced to solve long-term dependency problems in conventional recurrent neural networks. Experiment results indicate that LSTM-RNN shows stronger adaptability and higher prediction accuracy at 86%-94% than that of other data-driven models, which is of great significance for controling the silicon content of hot metal.

Silicon content ([Si]) of the molten metal is an important index reflecting the product quality and thermal status of the whole blast furnace (BF) ironmaking process. Since the direct online measure on this index is difficult and larger time lag exists in the offline assay procedure, quality modeling is required to achieve online estimation of [Si], which is an open problem for realizing BF automation. Focusing on this practical problem, this paper proposes a data-driven dynamic modeling method for [Si] prediction using extreme learning machine (ELM) with the help of principle component analysis (PCA). First, data-driven PCA is introduced to pick out the most pivotal variables from multitudinous factors that influence [Si] to serve as the secondary variables of modeling. Second, since this BF metallurgical process is nonlinearity dynamic system with severe time-varying characteristic, dynamic ELM modeling technology with good generalization performance and strong nonlinear mapping capability is proposed by applying the self-feedback structure on traditional ELM. The self-feedback connection enables ELM to overcome the static mapping limitation of its feedforward network structure so as can cope with dynamic time-series prediction problems very well. At last, industrial experiments and compared studies demonstrate that the constructed model has a better modeling and estimating accuracy as well as a faster learning speed when compared with different modeling method and different model structure.

Environmental protection and carbon emission reduction pressures are increasingly becoming a key factor affecting the sustainability of China’s steel companies at this stage. The key to energy saving and emission reduction in the steel industry lies in the innovation of ironmaking technology. It is inevitable to innovate and improve the traditional blast furnace ironmaking process and further strengthen the development of new ironmaking processes. This paper briefly reviews the representative achievements of iron and steel energy saving and emission reduction and new ironmaking process development in recent years, such as blast furnace utilization of alternative fuel, oxygen blast furnace ironmaking process, HIsarna smelting reduction ironmaking technology, COURSE50 technology and RHF iron nugget technology. It is expected to provide reference for achievement of energy saving, emission reduction and sustainable development of China’s steel industry.

Blast furnace (BF) system plays an indispensable role in the iron making process of iron and steel industry. It has been a crucial issue to maintain the stable and smooth operation state of BF. Real-time BF monitoring is the prerequisite of BF control, as the basis of prompt adjustments of operation parameters. Apart from the abnormal states, most of the time, BF is in the stable and normal state. During the stable state, BF is operated and running in different status, which are regarded as the sub-modes of the stable operation state. This paper proposes a novel idea for real-time BF monitoring, based on the temporal BF sub-mode recognition. The BF sub-modes are precalculated based on BF working state clustering adopting mean shift clustering algorithm (MSCA). The temporal BF sub-mode is sequentially recognized according to the real-time BF working state, adopting the minimum Euclidean distance algorithm (MEDA). The real-time BF stable state monitoring is implemented based on the obtained temporal BF sub-mode recognition results. The proposed method is carried out on two practical BFs located in the south of China. The silicon content in molten iron (which is used as the BF state indicator) estimation results obtained adopting the proposed method, are with a mean absolute percentage error (MAPE) of 15.89% and a root mean square error (RMSE) of 0.0854, compared with the measurement results.

Intelligent Multivariable Modeling of Blast Furnace Molten Iron Quality Based on Dynamic AGA-ANN and PCA

Fault diagnosis plays a vital role in ensuring the operation safety of blast furnaces and improving the quality of molten iron in the ironmaking and steelmaking industry. The blast furnace ironmaking process (BFIP) is intrinsically nonlinear. To address the nonlinearity issue of BFIP, a novel fault diagnosis approach that combines the randomized method, local structure information, and Fisher discriminant analysis is proposed in this paper. Using a randomized feature map, the process data is first mapped onto a randomized explicit low‐dimensional feature space. Compared to kernel methods, explicit low‐dimensional random Fourier features considerably reduce the computational cost, particularly for real‐time fault diagnosis for a large training dataset or a large‐scale process. Additionally, the local structure information contained in the randomized low‐dimensional feature space is extracted. The fault diagnosis performance is improved through the exploration of the local structure of random Fourier features. Finally, the blast furnace iron‐marking process state is determined using Bayesian inference. Case studies on a real‐world BFIP are carried out to demonstrate the superior performance of the proposed method in comparison with other related methods.

Silicon content plays an important role in reflecting the thermal state of a blast furnace (BF). The commonly used methods to predict silicon content at present are univariate and multivariate predicted method. The univariate predicted method only uses the information of silicon content in hot metal and can not reflect the complexity of blast furnace ironmaking process. The multivariate predicted method takes advantage of the rich information in blast furnace, different variables usually have a strong correlation with each other, which is not only useless for prediction but also brings in a lot of noise. So it always can not obtain the special satisfactory results. In this article, we have collected the real data from a medium-size BF and done three experiments based on support vector regression. These experiments can divide into two categories, univariate predicted method (UPM) and multivariate predicted method (MPM), the latter includes local variables prediction (LVP) and whole variables prediction (WVP). Multivariate correlation analysis and Spearman's rank correlation coefficient are applied to select variables which have low correlation for LVP. Phase space reconstruction is used to obtain the input and output of UPM. The experiment results show that LVP is better than the UPM and WVP. This also powerfully illustrates that choosing more variables can indeed bring in more useful information for prediction, especially when the correlation is eliminated through the selection of variables, the prediction result will be better.

Blast furnace (BF) is the predominating process for hot metal production, featured by high energy consumption and CO2 emission. A CFD based BF process model is further developed and validated at various levels. It is then used to study different strategies for improving BF process efficiency. This includes optimization of burden distribution and inner profile, the implementation of burden hot charge and developing oxygen BF ironmaking process. The applications prove that the model can be used not only for a better understanding but also for design and control of the BF ironmaking process.

The main production route for cast iron and steel is through the blast furnace. The silicon content in cast iron is an important indicator of the thermal condition of a blast furnace. High silicon contents indicate an increase in the furnace's thermal input and, in some cases, may indicate an excess of coke in the reactor. As coke costs predominate in the production of cast iron, tighter control of the silicon content therefore has economic advantages. The main objective of this article was to design an artificial neural network to predict the silicon content in hot metal, varying the number of neurons in the hidden layer by 10, 20, 25, 30, 40, 50, 75, 100, 125 , 150, 170 and 200 neurons. In general, all neural networks showed excellent results, with the network with 30 neurons showing the best results among the 12 modeled networks. The validation of the models was confirmed using the Mean Square Error (MSE) and Pearson's correlation coefficient. The cross-validation technique was used to re-evaluate the performance of neural networks. In short, neural networks can be used in practical operations due to the excellent correlations between the real values ​​and those calculated by the neural network.

Hybrid prediction method of blast furnace gas generation considering multi-operation modes in steel plants near the city

In steel production, the blast furnace is a critical element. In this process, precisely controlling the temperature of the molten iron is indispensable for attaining efficient operations and high-grade products. This temperature is often indirectly reflected by the silicon content in the hot metal. However, due to the dynamic nature and inherent delays of the ironmaking process, real-time prediction of silicon content remains a significant challenge, and traditional methods often suffer from insufficient prediction accuracy. This study presents a novel Multi-Scale Fusion Convolutional Neural Network (MSF-CNN) to accurately predict the silicon content during the blast furnace smelting process, addressing the limitations of existing data-driven approaches. The proposed MSF-CNN model extracts temporal features at two distinct scales. The first scale utilizes a Convolutional Block Attention Module, which captures local temporal dependencies by focusing on the most relevant features across adjacent time steps. The second scale employs a Multi-Head Self-Attention mechanism to model long-term temporal dependencies, overcoming the inherent delay issues in the blast furnace process. By combining these two scales, the model effectively captures both short-term and long-term temporal dependencies, thereby enhancing prediction accuracy and real-time applicability. Validation using real blast furnace data demonstrates that MSF-CNN outperforms recurrent neural network models such as Long Short-Term Memory (LSTM) and the Gated Recurrent Unit (GRU). Compared with LSTM and the GRU, MSF-CNN reduces the Root Mean Square Error (RMSE) by approximately 22% and 21%, respectively, and improves the Hit Rate (HR) by over 3.5% and 4%, highlighting its superiority in capturing complex temporal dependencies. These results indicate that the MSF-CNN adapts better to the blast furnace’s dynamic variations and inherent delays, achieving significant improvements in prediction precision and robustness compared to state-of-the-art recurrent models.

Blast furnace iron-making process (BFIP) is one of the most critical procedures in the iron and steel industry where timely detection and accurate classification of faults have always been of core focus. However, the coupling effects of system’s nonlinear and nonstationary characteristics often cause process consistent underlying information to be buried, allowing accurate extraction to be a significant challenge. This also complicates the development of BFIP fault diagnosis model. Therefore, we propose a novel data-driven joint fault diagnosis strategy that employs regularized mutual kernel analytic stationary subspace analysis (RMK-ASSA) and deep broad stationary kernel network (DBSKNet) to eliminate this interference. To develop this method, we first construct an RMK-ASSA approach to address the poor modeling accuracy caused by standard analytic stationary subspace analysis (ASSA)’s inability to handle complex process nonlinearity. Global and local kernels are utilized to account for multiple nonlinearities in BFIP data. The weight of different nonlinear data is calculated by regularized principal component analysis, and the main information is imported into ASSA to obtain more robust and accurate modeling results by eliminating the interference of redundant noise. Subsequently, we design a DBSKNet-based classifier to implement the fault diagnosis task. This network further considers the nonlinearity by boosting kernel structure in depth and width while distinguishing the respective contributions of different kernels to fault diagnosis results. Finally, a double-layer loop parameter optimization algorithm is used for optimizing. Simulated cases and practical BFIP tests validate that RMK-ASSA eliminates the negative impact caused by nonstationary data and that the proposed joint fault diagnosis strategy outperforms other methods. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —BFIP’s nonlinear and nonstationary coupling properties pose unique challenges in eliminating distractions, constructing fault classifiers and accurately detecting process anomalies. To tackle these challenges, this paper proposes a joint fault diagnosis strategy based on RMK-ASSA and DBSKNet. RMK-ASSA effectively estimates nonlinear consistent features, while DBSKNet mines rich deep nonlinear information, accurately distinguishing variations in BFIP data under different working conditions. Experimental results demonstrate that this data-driven strategy can perform high-quality fault diagnosis, enabling field engineers to execute operations efficiently.

Slag flow behaviour is critically important in the lower zone of the ironmaking blast furnace, and is closely related to the selection of charged raw materials, coke bed permeability, process stability and hot metal quality. To better understand the effect of slag properties on flow behaviour in the coke bed, a numerical approach was applied to characterize the slag flow through funnel analogues. These analogues were used to represent molten slag flow through the inter-particle voids of a coke packed bed. A critical funnel neck size, through which no slag flowed was experimentally established and confirmed by numerical modelling. The influence of slag wettability on the occurrence of blockage was also determined via numerical modelling. An increase in either contact angle or surface tension can make the occurrence of blockage easier. For a constant neck size, the relationship between surface tension and contact angle is non-linear. The status of the remaining slag in the funnel corresponding to different slag wettabilities was differentiated in terms of the blockage in the upper part and hanging in the lower part of the funnel. Modelling was also undertaken of slag flow through the inter-particle void between spherical particles to evaluate empirical correlations for predicting the remaining slag in the packed bed. These results show that the numerical approach is very useful in providing some level of guidance to help understand and predict the slag flow behaviour in the blast furnace ironmaking process.

An actual direction in the development of blast-furnace smelting technology is the implementation of stable operating modes of blast furnaces in terms of the parameters of the used charge materials and hot combined blast. For this purpose, complex indicators characterizing the thermal and reduction modes of blast furnace smelting, and integral indicators of iron ore raw materials (IORM) quality are being studied. The object of the study is the blast furnace process under the conditions of PJSC NLMK. The database of the main indicators of the operation of blast furnaces No. 4, 6 and 7 for the period 2013&ndash;2018 was analyzed. The furnaces operate under raw material conditions, with similar combined blast parameters and coke consumption. When the consumption of natural gas, pulverized coal and process oxygen changes, the values of the degree of compensation (the ratio of the consumption of reducing additives to the consumption of process oxygen) are in the range of 0.9&ndash;1.1. This provides a constant level of hearth heating, stability of the gas-dynamic mode of melting and coke consumption. At the same time, the different level of development of the processes of indirect (Ri) and direct (rd) reduction in blast furnaces No. 4, 6 and 7 makes it possible to perform correct studies of the features of blast furnace smelting in a wide range of Ri and rd values. It has been established that under the conditions of stable operation of blast furnaces on raw materials from sinter and pellets using combined blast, the type of dependence of the sulfur and silicon content in pig iron on the values of the integrated iron ore raw materials quality index is determined by degree development level of direct reduction processes rd. It has been established that the mode of blast furnace smelting with natural gas, pulverized coal and process oxygen, characterized by the level of rd values less than 20&ndash;25 %, is distinguished by the presence of a pronounced extreme dependence of coke consumption and productivity on the parameters of the combined blast.

In this paper, the overall and zonal thermal balances of the blast-furnace smelting were analyzed. Under the current blast furnace smelting conditions, the specific consumption of coke is determined by the development of the heat transfer processes occurring in the lower zone of the furnace. In this case, the heat losses found by using the heat balance of the lower zone of the furnace sufficiently reflect such technological factors as specific coke consumption and silicon content in cast iron. It was shown that the existing methods of analytical calculation of the heat losses occurring during smelting of cast iron in blast furnaces do not adequately reflect the actual melting process. It was established that the analysis of the heat balance of the lower stage of heat transfer enables determining the heat losses in this zone of the furnace and estimating the correction factors when the blast furnace smelting conditions change. The control of heat losses is necessary for estimating the condition of the blast furnace masonry, rational gas distribution, and also, for adjustment of coke consumption.