With the advancement of carbon peak and carbon neutrality policy,higher demands have been placed on the blast furnace ironmaking process,which constitutes a primary energy consumption segment within the iron and steel industry. Achieving intelligent sensing of key indicators,diagnosing furnace conditions,and optimizing control of operational parameters in the blast furnace ironmaking process is of paramount significance for promoting its safe,green,and low-carbon development. Firstly,taking intelligent sensing and prediction of key state indicators in blast furnaces as a starting point,providing a comprehensive review of sensing and prediction methods for three critical indicators:gas utilization rate,molten iron silicon content,and permeability index. Secondly,an analysis of the current research status of blast furnace condition monitoring and diagnosis is conducted from two perspectives:expert system and data-driven approaches. Subsequently,advancements in optimization and control of blast furnace operation parameters are reviewed from three angles:expert system and expert experience extraction,multi-objective optimization,and data-driven predictive control. Finally, by analyzing the strengths and weaknesses of various models and algorithms,the current challenges and development directions for intelligent sensing,furnace condition diagnosis,and operation optimization of blast furnaces are proposed.

The interference of the surface medium of seamless steel pipes and the high leakage rate of manual inspection result in the accuracy of surface defect detection of steel pipes not meeting the requirements of online detection. The insufficient depth of field and uneven grayscale of imaging detected by two-dimensional cameras result in low defect recognition rate,false detection of defects,and missed detection. Therefore,a 3D point cloud based steel pipe outer surface detection system was proposed to solve the problems of small imaging depth,low defect recognition rate,and interference defects on the steel pipe surface. Deep learning algorithms and 3D size measurement technology are applied to defect detection of seamless steel pipes. This defect detection system is used to quantify the size of defects and detect them more accurately. The on site defect recognition rate of this system can reach over 90% ,and the detection speed is fast. In addition,this system has multiple functions such as periodic defect alarm,steel pipe surface defect statistical report printing,defect grading,etc. ,making the surface defect detection system multifunctional and intelligent,which has positive significance in reducing manual labor intensity and improving the quality of seamless steel pipes.

In the context of the sintering blending process in steel production,challenges arise due to significant fluctuations in iron ore powder prices,the complexity of sintering raw material information, and the impact of various factors on sintering ore blending. Traditional genetic algorithm (GA) can easily fall into local optima. To address this issue,this study proposed a mathematical model based on an improved GA aimed at optimizing the sintering ore blending process to tackle the challenges posed by these influences on the cost of sintering materials. The model automatically adjusts the size of operators during the operational process based on the specific problem environment,effectively avoiding the premature convergence issue encountered by traditional GA. This ensures that the algorithm ultimately outputs a globally optimal solution when optimizing the sintering modeling. Starting with iron ore powder,the system utilizes technologies such as Python,MySQL and PyQt5 to construct an integrated sintering ore blending model. Through analysis and processing of backend data,the system ultimately generates optimized sintering ore blending solutions.

To address the challenge posed by the contradiction between the large-scale mass production in steelmaking and the demand for variety and specification diversity in subsequent processes,a study on the steelmaking flow balance problem was conducted. An optimization method based on intelligent decision-making was proposed to enhance the overall production process忆s continuity and stability. The approach involved analyzing the process rules and technical indicators that needed optimization to achieve flow balance in steelmaking. To accurately represent the problem structure,a mixed integer programming model was developed. Recognizing the need for efficient solutions,especially for large-scale problems,an enhanced tabu search algorithm was introduced. This algorithm incorporates multiple neighborhood structures and disturbance factors to prevent it from getting trapped in local optima. Numerical experiments were conducted on practical instances of varying scales,and the results indicate significant improvements in solution quality and convergence of the proposed enhanced tabusearch algorithm compared to the standard solver CPLEX and conventional tabu search algorithms. Furthermore,it outperforms manual decision method and can basically meet the needs of flow balance intelligent decision-making.

The hot metal pretreatment desulfurization process in modern converter steelmaking process can improve the efficiency of impurity removal,reduce the burden of converter blowing,and shorten smelting time.It is a necessary process for smelting variety steel and clean steel.The pretreatment process of molten iron undergoes complex high-temperature physical and chemical reactions,which is a black box process,making precise control of the smelting process very difficult.Establishing an accurate process control model is the core of achieving precise control of the hot metal pretreatment process,which is of great significance for enterprises to reduce steel production costs,promote digital and green transformation.This paper summarizes the modeling principles,characteristic and research progress of various models,such as mechanism model,statistical regression model,expert system and machine learning model,established by domestic and foreign researchers for hot metal pretreatment desulfurization process.Based on the different uses of the model,the development process of application in practice and prospects of the hot metal pretreatment desulfurization model were proposed,focusing on the prediction of endpoint sulfur content and desulfurization rate,prediction and optimization of smelting process parameters,and prediction of desulfurization agent consumption and utilization rate.

With the development of the steel industry,the steel manufacturing mode is gradually shifting from manual production to unmanned and intelligent production. At present,the working intensity of the converter steelmaking process is high,the equipment operation is cumbersome,and the working environment after the furnace is relatively harsh. With the goal of one click steelmaking and safe steelmaking,the converter steelmaking system was improved and optimized to establish an intelligent steelmaking system. For the transformation of the automation system,firstly increase the setting of the steel tapping curve and improve the functional design of the alloy chute. Then,add security chain programs,ladle car detection devices,and converter tilting detection devices to ensure the safety of the steelmaking process. At the same time,machine vision assisted system and L2 models are used to monitor and calibrate the steel tapping process,ensuring that the molten steel and slag do not overflow. Through practical application in two 200 t converters of HBIS Tangsteel Company,it has been verified that the system promotes standardized production of converter steelmaking,reduces labor intensity of workers,and ensures security and stability of tapping process.

A disk pelletizing automatic control system based on flow rate target was proposed to address the difficulties in accurately controlling pellet size and outdated production methods in the production process of pellet ore. Firstly,relying on image recognition algorithms,basic information such as particle size,distribution,and number of pellets are extracted from actual production process pellet images. Secondly,the system analyzes and calculates the received pellet information to obtain the current production status and particle size change trend of the pellets. Finally,the system establishes a flow target setting value based on the trend of pellet particle size change and production status,and adjusts the valve opening adjustment cycle appropriately according to the pellet production status to adapt to production. The system was actually applied to the control of a disk pelletizer in a domestic steel plant. The practical results show that in terms of the difference between the average particle size and the target particle size,the automatic control mode reduces the error by 32. 54% compared to traditional manual control mode. Moreover,under different pelletizing conditions,the proposed control system exhibits stable fluctuations in flow during actual production. Compared to manual control mode,the root mean square error of pellet size is reduced by 6. 59% ,which has good stability. It has a positive effect on improving the production efficiency and qualification rate of pellets,and reducing manual labor intensity.

The blast furnace gas flow can characterize the operating state of a blast furnace,while the cross-temperature measurement reflects the distribution state of the blast furnace gas flow. This paper proposed a dynamic modeling method for multivariate correlation of blast furnace cross-temperature measurement based on periodic registration and seasonal and trend decomposition using loess (STL),which can improve the accurate estimation of gas flow. Firstly,the periodic partitioning and periodic registration among multiple variables are performed by sliding window,which is helpful to achieve precise multivariate correlation. Next,the RobustSTL is introduced to retain key information, extract global changes,and enhance the accuracy of the online estimation model. Then,the gated recurrent unit (GRU) is employed to establish a multi-step prediction model for the multivariate correlation of cross-temperature measurement. Finally, experimental verification is conducted using the cross-temperature measurement dataset, and the results show that the proposed predictive model achieves significant performance improvement.

By analyzing the problems existing in the existing single technology of converter smelting detection,the online quantitative evaluation technology of slag foam degree and the online prediction technology of process feature points of flue gas analysis are developed on the basis of summary to solve the problem of converter black box and realize converter smelting visualization. Based on the identification results of the detection technology, the intelligent intervention technology of powder spraying and slag suppression at the tapping port and the intelligent intervention technology of the oxygen gun are used to achieve stable control of the converter blowing process. After the application of technology,the risk of slag overflow during the converter smelting process is significantly reduced, preventing explosive splashing in the middle and later stages,and reducing the loss of metal materials. After the control of the converter smelting process is stabilized,it is beneficial to improve the end-point temperature and carbon prediction hit rate.

The smelting intensity (SI) affects the physical and chemical reactions within the blast furnace,causing the relationship between the gas utilization rate (GUR) and the blast supply parameters undergoes variations with changes in SI. Disregarding the SI means neglecting the dynamic correlation between GUR and blast supply parameters,resulting in adverse effects on predicting GUR using blast supply parameters. This paper introduces a GUR prediction model that takes into account the classification of SI. Firstly,the impact of SI on the state parameters of blast furnaces is evaluated from the perspective of molten iron smelting mechanisms. Then,a weighted kernel fuzzy c-means clustering method (WKFCM) based on state parameters is proposed to classify the SI. Subsequently, supervised principal component analysis (SPCA) is employed to reduce the dimensionality of the input data and a support vector regression (SVR) method is used to predict the development trend of GUR. Finally,the model is applied to predict real GUR data under different SI. Analysis of actual production data indicates that the prediction method considering SI classification is more suitable for forecasting GUR time series in the complex production environment of blast furnaces.

The smooth condition of the blast furnace is important for the production and quality of the hot metal. The stability of the slag crust indicates the stability of the conditions in the blast furnace, and the temperature in the cooling stave can describe the stability of the slag crust. For predicting the conditions of the blast furnace according to the dynamic features of the temperature in the cooling stave,this study presented an intelligent method for predicting the conditions of the blast furnace based on the information granules of the temperature in the cooling stave. Firstly,the Spearman correlation analysis method is employed to select the parameters that affect the dynamic features of the temperature in the cooling stave. Secondly,the information granulation method is used to extract the dynamic features of the selected parameters and represent the data in a granular form. Then,the prediction model of the information granule of temperature in the cooling stave is built based on the support vector regression with the inputs of information granules of the selected parameters,realizing the prediction of the temperature in the cooling stave. Finally,based on the predicted information granules,the conditions of the blast furnace can be recognized using a condition prediction method. Experiments conducted using actual steel enterprise data show that the presented method can predict the conditions in the blast furnace and provides powerful guidance for operators to make a proper burden distribution decision-making strategy.

The preparation of wheel production plans is one of the core tasks in the management of transit materials production plans. A reasonable production plan helps regulate the production pace of each order,ensuring smooth and orderly wheel production. Based on the scheduling of transit materials wheel production plans,taking into account factors such as rolling,heat treatment,outsourcing
rough machining,precision machining,and inspection capabilities,as well as the logistics cycle between insulation covers, tool switching, and sequential operations. By establishing a mathematical model and employing a multi-objective combinatorial optimization algorithm (MOCOA),intelligent scheduling of wheel plans is achieved,effectively addressing the production planning and scheduling issues arising from small batch sizes,multiple specifications and personalized requirements in Masteels transit materials wheel production plans. The results of the system operation demonstrate that by repeatedly balancing the contractual delivery dates of wheel orders,unit capacity,material inventory,process paths,logistics cycles,equipment utilization rate,availability time of production lines and contract manufactory capabilities throughout the entire factory. The system generates interlinked production plans for each process unit,implementing a single plan creation process,which was previously taking two days,has been shortened to within 2 h,effectively enhancing the efficiency of wheel production scheduling. The reduction has led to a 20% decrease in wheel manufacturing lead times and a 20% reduction in inventory occupancy,aligning with the smart factory production requirements aimed at cost reduction and efficiency enhancement.

To address the challenges of low model accuracy and poor robustness in traditional single models for blast furnace molten iron temperature prediction,a combined model was proposed. The model integrates the intrinsic computing expressive empirical mode decomposition with adaptive noise(ICEEMDAN), kernel principal component analysis ( KPCA) and relevance vector machine
(RVM) to achieve precise and stable predictions of molten iron temperature. Firstly,ICEEMDAN is employed to decompose the time series of molten iron temperature,yielding several intrinsic mode functions. Secondly,KPCA is applied to reduce the dimensionality of the multidimensional key variables in the steel production process and extract the main features of these key variables. Lastly,RVM is utilized to predict the molten iron temperature time series based on the reduced variables,resulting in the cumulative prediction results of molten iron temperature. The results demonstrate that compared to the traditional complete ensemble empirical mode decomposition with adaptive noise ( CEEMDAN) model,the new model exhibits a 13. 0% reduction in root mean square error (RMSE) and a 10. 9% increase in training speed,enabling a better understanding of the dynamic changes in molten
iron temperature. Compared to traditional single models like RVM,the new model reduces the mean absolute error (MAE) by 2. 47 and the training time by 0. 463 s,which has the advantages of higher model accuracy and faster speed. Thus,the proposed model provides theoretical support for real-time control of blast furnace temperature,which is of practical significance in ensuring the stability of blast furnace smelting and accelerating the metallurgical process towards intelligent automation.

Given the complex production process of traditional converter steelmaking,the difficulty of high-temperature molten steel detection and manual operation. An multi-task parallel architecture converter steelmaking process control system is developed to improve the production efficiency of converter steelmaking. The system includes six functional modules,namely process tracking module,human machine interface (HMI) module,data communication module,model calculation module,data management module and data validity judgment module. It realizes the automatic steelmaking process without human intervention in the converter steelmaking process. The problems of large error in data detection and excessive dependence on manual experience are solved. The system adopts multi-process structure,and the new mode of one task one process is used in the process,which reduces the
coupling between the function modules. The practical application shows that the system is simple in operation,strong in stability,and the interaction is good.

In the process of blast furnace smelting,under the influence of dynamic changes of working conditions and complex factors at the production site,the fluctuation of differential pressure has a certain time lag,and it is still difficult to realize the accurate forecast of differential pressure based on real-time online data. To address this problem,based on the actual smelting process of the blast furnace,which has the characteristics of multivariate and time-dependent time series data,the volatility analysis and decision tree feature importance analysis methods that can effectively reflect the degree of fluctuation of the production process parameters are adopted,and different subsets of the model input features are selected,so as to establish the temporal pressure difference prediction model based on the long short-term memory (LSTM). The comparison results of the two methods show that the LSTM prediction model based on volatility analysis to determine the input features has an improved hit rate of 0. 761% within the prediction error range[ - 5, + 5] kPa. The feature selection method based on the volatility analysis of production parameters can effectively improve the prediction accuracy of the LSTM model,and verify the validity of the input feature selection method of the temporal differential pressure prediction model under the condition of oxygen-enriched blast furnace.

The final judgment of hot rolled product quality is carried out in the manufacturing execution system (MES),and the sampling method is used for inspection,which is difficult to meet the needs of customers for production process control. In order to accurately characterize the influence of process parameters on product quality,the quality process judgment and grading method provided in this paper adds the head and tail removal rule on the basis of the traditional judgment based on manual empirical rules,and then comprehensively evaluates the judgment results after sub item grading.In the actual application of the field,compared with the previous method based on manual empirical rules,this method provides more comprehensive,refined and accurate data support for MES quality final judgment,reduces the outflow rate of quality defected coiles,improves customer satisfaction,and enhances enterprise competitiveness.

The permeability index of blast furnace is an important parameter that can quickly,intuitively,and comprehensively reflect the condition of the blast furnace. Accurately predicting the permeability index of blast furnaces can detect and avoid abnormal furnace conditions such as pipeline, suspended material,collapse,and gas loss as early as possible (about 10 min in advance). This article proposes a blast furnace permeability index prediction model that combines kernel principal component analysis ( KPCA), convolutional neural network ( CNN), and long short-term memory (LSTM). Firstly,KPCA is used to reduce the dimensionality of the original high-dimensional input variables,followed by CNN to capture the features of the data,and finally,LSTM is used to predict
the permeability index of the blast furnace. The results show that the constructed KPCA-CNN-LSTM blast furnace permeability index prediction model significantly reduces prediction errors and improves prediction accuracy compared to before dimensionality reduction. It is beneficial for blast furnace operators to quickly grasp the instantaneous changes in furnace conditions and take effective measures to restore smooth operation of the blast furnace.

The precise control of the end-point carbon content and temperature of Consteel electric arc furnace is extremely important to ensure the high-quality liquid steel. Different machine learning models including prediction models and hyper-parameter optimization models were used to fit the actual production data from the smelting process of Consteel electric arc furnace in a domestic steel plant. Based on the prediction results of machine learning models,the end-point carbon content and temperature prediction model of the Consteel electric arc furnace based on Bayesian optimization algorithm (BOA) and gradient boosting decision tree (GBDT) was established. Through the verification by using the collected data,this model achieves better prediction performance for the end-point carbon content and temperature of Consteel electric arc furnace smelting,which can provide certain guidance for the actual production process.

Continuous casting slabs are raw materials of steel production.The defect of slab will lead to quality defect of final steel product.The low and high frequency data collected from continuous casting process on site were studied.Cleaning methods of complex process industrial data and feature extraction methods of high frequency industrial data were proposed.Based on machine learning theory,four kinds of slab surface defect prediction models,namely classification and regression tree (CART),AdaBoost,random forest (RF),and optimal classification tree (OCT) were established.Feature selection were carried out using Relief and RF model.The prediction accuracy of different models was compared and analyzed through a large number of experiments.The experimental results show that the RF model gives the best prediction accuracy.The top 10 features,such as liquidus temperature,tundish (TD) lower limit temperature and TD target temperature,which play a key role in slab surface defects are found out.The method in this paper can be extended to industrial data analysis and utilization modeling in other scenarios,which has important reference value for using industrial data to improve product quality.

During the continuous casting process,the instantaneous abnormal mold level fluctuation has great detrimental effects on slab quality,thus the mold level fluctuation is a key parameter for continuous casting process of high-quality steel.In the present study,the slab continuous casting process data for low carbon steel,medium carbon steel,hypo-peritectic steel and peritectic steel were collected,the fast Fourier transform (FFT) and continuous wavelet transform (CWT) were used to analyze the data characteristics,and then the influence of process parameters on the instantaneous abnormal mold level fluctuation was studied.The results of FFT analysis indicate that bulging has no significant effect on the instantaneous abnormal mold level fluctuation.The time-frequency characteristics of instantaneous abnormal mold level fluctuation and stopper-rod position were analyzed by CWT,and the results show that under different steel grades and casting speeds,before the instantaneous abnormal mold level fluctuation occurrence,the CWT coefficients in high-frequency region of stopper-rod position shows a linear increase trend.Therefore,by conducting CWT analysis of the high-frequency region of the stopper-rod position,it is possible to predict the instantaneous abnormal　mold level fluctuation.

The prediction of sulfur content of liquid steel in ladle furnace (LF) is of great significance for the precise control of the composition of refined molten steel and the improvement of product quality.In view of the complex mechanism,multi-variable,and nonlinear characteristics of LF,a prediction method of sulfur content in LF was proposed based on autoencoder-back propagation neural network (AE-BPNN).Firstly,the effects of noise and missing values are eliminated through AE network for data reduction and feature extraction.Then,the BPNN is used for predicting the sulfur content of liquid steel in LF.The actual on-site data validation show that the E_RMS,E_MA and the correlation coefficient R² are respectively up to 1.403,1.083 and 0.824,which has good prediction performance.

The accurate prediction of the Linz Donawitz converter gas (LDG) holder level in iron and steel enterprises can provide an important basis for gas system scheduling.Given the impact of LDG recovery,scheduling workers are particularly concerned about the problem of exceeding the upper limit of the holder level.Based on a large amount of on-site actual data,a cost-sensitive learning-based support vector machine (SVM) method for predicting the level of LDG holder was proposed,which can improve the prediction accuracy of the situation when the level exceeds the upper limit.This method takes the recovery and consumption flow of LDG as inputs,and the future holder level value as output.By using the KKT equation,the original constraint conditions are transformed into equation constraints,different costs are set for the gas holder level exceeding the limit and false alarms.Finally,by minimizing the model′s false alarm error,the original prediction problem is transformed into a series of linear equations and solved.A simulation experiment was conducted on data　from a domestic steel plant,and the results showed that the proposed method can effectively reduce the false alarm rate of converter gas holder level exceeding the limit to 0.16%,providing more rapid and effective guidance for scheduling workers to develop reasonable scheduling strategies.

During the traditional hot strip rolling production process,due to the rapid oxidation of the strip steel,the significant measurement deviation arises in the temperature measurement at finishing rolling entrances due to the obstruction and interference of the oxidized iron scale on the surface of the strip steel,which becomes an important influencing factor for the calculation and setting of disturbance rolling parameters and model self-learning regulation. A finishing rolling entry temperature prediction model was established based on machine learning neural networks,which integrates range analysis methods to determine data features and screens data based on mechanism and equipment conditions. The model predicts the rough rolling outlet temperature and calculates the temperature drop of the mechanism model to obtain the predicted value of rolling temperature,which aims at correcting the disturbance of finishing rolling entry temperature measurement caused by iron oxide scale on the surface of strip steel. Through continuous production data analysis and comparison,the temperature deviation decreases from ±9. 15 ℃ to ±5. 33 ℃. The model evaluation indicator R² has been increased from 0. 41 to 0. 84. The average value of the finishing rolling entry temperature difference of samples with sharp pits in the strip steel temperature decreases from 48. 45 to 11. 02. After performance evaluation,it's believed that the prediction model has high accuracy and strong generalization.

The blast furnace smelting is carried out in a completely closed and high-pressure environment,it is impossible to observe the internal operating conditions of the blast furnace and the shape of the material surface,making it difficult to accurately judge the furnace condition,and the utilization rate of the material surface data resources is not high,which affects the operators adjustment of the distribution system of the furnace top. In order to improve the data utilization rate and improve the quality and accuracy of point cloud data,a double-sided filter was proposed to preprocess the three dimensional point cloud data of the original blast material surface in the paper. Poisson reconstruction algorithm is used to reconstruct the filtered point cloud data,build a multi-scale feature coding network,and repair the missing 3D point cloud material surface. Poisson surface reconstruction can retain the detail characteristics of the surface and smooth the surface,which provides an important basis for quickly judging the type of the surface. By extracting the point cloud feature information of different scales,3D point cloud feature enhancement and multi-level expression are realized. Experiments show that the proposed method has small error in point cloud missing prediction and complete shape of point cloud,which provides a fast,efficient and practical solution for processing point cloud data with missing material surface.

Blast furnace radar burden line extraction currently commonly used neural network plus energy center of gravity method of two-step extraction of burden line method,there is a mixture of network model and mechanism model step-by-step computation,susceptible to the influence of the special environment of strong noise problem. In this paper,an improved BS-TransUNet algorithm for blast furnace burden line extraction based on semantic segmentation was proposed. Firstly,to address the problems of periodic morphology and particle size variation of blast furnace burden surface and signal to noise ratio attenuation,the atrous spatial pyramid pooling (ASPP) module is introduced between convolution neural network (CNN) and Transformer modules to obtain fine-grained features of the burden surface. Then,the coordinate attention (CA) module is integrated after each up-sampling to filter out the background noise more comprehensively and inhibit the extraction of ineffective high-frequency texture features. Finally,the jump link is replaced with the BiFusion module to further improve the segmentation performance. The experimental results show that the improved algorithm improves the mean intersection over union (MIoU) and F1 scores by 1. 77% and 1. 46% ,respectively,the mean pixel accuracy (MPA) by 1. 97% on the blast furnace radar burden surface dataset, and the F1 score can reach 86. 18% . Compared to the conventional two-step extraction of burden line method,the one-step method with end-to-end split burden line in the harsh environment of a blast furnace provides improved accuracy and stability of burden line acquisition.

The converter heat loss rate is one of the important parameters that affect the prediction accuracy of material consumption.Based on the calculation of heat loss rate,the historical production data of 1 900 heats of a 150 t converter in a steel plant was applied to accurately predict the converter heat loss rate using machine learning algorithm.The prediction results show that light gradient boosting machine (LightGBM) algorithm has the highest prediction accuracy compared with support vector regression (SVR) and random forest (RF) algorithms.Considering the influence of the last furnace,after adding the final smelting temperature variable of the last furnace,the determination coefficient R² of LightGBM algorithm increases from 0.89 to 0.93,and within the range of ±0.005 and ±0.01,the prediction hit rate of heat loss rate increased from 85%,89% to 90%,93% respectively.In addition,the prediction accuracy of the model can be further improved by optimizing the internal parameters of the algorithm.For the LightGBM algorithm,the determination coefficient R² and root mean square error (RMSE) further reach 0.94 and 0.009,and the prediction hit rate of the heat loss rate further increases to 91% and 94% within the range of ±0.005 and ±0.01.Based on historical data of converter smelting,intelligent algorithms can be used to predict the heat loss rate of the converter,which can provide support for the prediction of the material consumption of the converter.

Plate hot rolling is a typical process industry,which goes through continuous casting,heating,descaling,rolling,cooling,coiling and other technological processes in turn. Because of the large range and fast speed of temperature change in the cooling process,the cooling process has the greatest influence on the microstructure and properties of steel plate,and the final cooling temperature is a
key control parameter in the cooling process. In order to improve the accuracy of the final cooling temperature prediction,the LightGBM (light gradient boosting machine) model was used for regression prediction of the final cooling temperature. The size of plate,chemical composition and upstream and downstream process parameters are used as inputs of the model,and the final cooling temperature is used as output of the model. Bayesian optimization method is used to complete the super-parameter optimization of the model. In addition,Shapley additive explanation (SHAP) method is used to test the influence of input parameters on the predicted parameters. The results show that Bayesian optimized LightGBM (BO-LightGBM) model achieves lower error in both training set and test set,95% of the absolute error of predicted data is controlled at ±10 ℃,and the time consumption of the model is reduced by 97% compared with other ensemble learning models,the prediction accuracy and prediction efficiency of hot rolling process temperature of plate are improved.

Blast furnace permeability index is an important index to reflect the indirect reduction degree of charge and furnace condition,which is affected by blast furnace operations in multiple time scales. The existing research analysis,modeling and prediction of the development trend of the permeability index are mostly based on the same time-scale and the prediction step is short,so the prediction results are difficult to guide the on-site judgment. Therefore,this paper proposed a multi-step prediction model of blast furnace permeability index based on multi-time scale. Firstly,the time domain characteristics of the influence of blast furnace operations on the permeability index on multi-time scale are calculated through mechanism and data analysis,and the multi-time scale effects of different operations on the permeability index in different time scales are analyzed in combination with the frequency domain characteristics. Then,according to the characteristics of blast furnace operation affecting the development of permeability index in different time scales,a single-step prediction model based on support vector machine is established. Finally,a multi-step prediction model of permeability index based on recursive strategy is established on the basis of single-step prediction model. The experimental results show that this method can effectively predict the future development trend of permeability index and is convenient for on-site decision-making.

It is of great significance to clarify the phase distribution and lining erosion characteristics of blast furnace by the damage investigation,and the information on various phase parameters of the damage-investigation samples can be obtained rapidly and cost-effectively through the image-processing technology.The coke information of the residual iron section in the hearth,the three-dimensional structure of slag-iron-coke and the damage degree of the refractories were characterized quantitatively based on the damage-investigation work of a large blast furnace in China combining Photoshop (PS),Image-Pro Plus (IPP) software and industrial CT equipment.The process of image acquisition and binarization processing was firstly introduced,as well as "calculation method of pixel proportion" and the trend distribution of deadman voidage by grid partitioning were then characterized.At the same time,the correlation between two and three dimensional voidage was verified through detecting　the three-dimensional phase distribution and voxel ratio of slag-iron-coke in the deadman.Furthermore,the correction coefficient C was introduced based on the basis of "coke size calculated by equivalent circle" which reduced the influence of coke irregular morphology on particle size.The distribution,size,and depth of the refractory material′s cracks in surface macroscopic and internal microscopic were intuitively characterized by the binary images,and the number,volume,and porosity of pores in refractory materials can be also characterized by the three-dimensional structure,then determined the damage situation of refractory materials in the hearth quantitatively.

Belt conveyor is one of the most important equipment in the blast furnace delivery system.The failures of belt conveyor can lead to production losses and even safety accidents in severe cases.In order to discover the abnormalities of belt conveyor system in time,a fault monitoring system is designed for key components of the conveyor,including the motor,the reducer,the bearing of the balance weigh忆s bend pulley,the bearing of the tail pulley,etc. The hardware architecture addresses the challenges of wide distribution,numerous measuring points,and long distances in the star layout of the belt conveyor monitoring points. The use of integrated electronics piezo-electric (IEPE) interface sensors effectively reduces wiring costs while enhancing signal transmission distance and reliability.On the software side,an innovative approach is employed,involving decompose the feature of raw training data through variational mode decomposition before training the neural network. Experimental results show that,without increasing the complexity of the neural network,the software忆s accuracy in judgment has increased from 96. 5% to 99.3% ,while the false negative rate has decreased from 3.5% to 0. 7% . Additionally,training errors can converge rapidly,leading to a significant improvement in performance.

The content of FeO in the sintering process is an important reference index affecting the performance of sintering ore.Real-time observation and monitoring of changes in FeO content can reduce sintering energy consumption and improve sintering effect.Aiming at the current situation that the means of real-time observation of FeO content in enterprises is relatively single,the prediction model of FeO content in sinter based on bi-directional long short-term memory (BiLSTM) neural network algorithm was studied.The source of the data is partial process data generated in 2021 by the six-type sintering machine of Panzhihua Iron and Steel Co.,Ltd.After filtering,optimizing and other data processing,the BiLSTM neural network is selected for training,parameter adjustment,and combined with the on-site sintering process of the enterprise,which improves the prediction accuracy of the model.The accuracy rate basically realized the prediction of sintered FeO.The test results show that within the allowable range of enterprise error,the accuracy rate reaches 90.2%,so it can give　effective opinions on sinter production in the enterprise.

Aiming at the rescheduling problem after the failure of a smelting equipment in the steelmaking-continuous casting production process,in order to ensure the stability of production and reduce the variation of the rescheduling scheme compared with the initial scheduling scheme,a rescheduling model was proposed with the objective of minimizing the weighted variance of the start time,operation cycle and equipment assignment,which was established by mathematical planning method.By analyzing the production operation mode and production process of a steel plant,a rescheduling algorithm was proposed,which consists of an equipment assignment algorithm based on the "furnace-caster coordinating" scheduling strategy rule and a time adjustment algorithm based on the process flexible buffer control strategy rule.Taking the converter equipment failures or refining furnace equipment failures that often occur during the actual production of steelmaking and continuous casting in a large domestic steel plant as simulation samples,the experimental results indicate that the total weighted differences between the start time,operation cycle and equipment assignment before and after scheduling were 0.29,1.43,and 1.21,respectively,which can effectively maintain the consistency between the rescheduling plan and the initial scheduling scheme,and ensure the stability of production.And the solution time was less than 0.6 s,which can quickly provide better solutions to the rescheduling problem after smelting equipment failures.In the actual production process with frequent failures of smelting equipment,the orderly,stable,and efficient operation of steelmaking-continuous casting can be ensured.

Reversible cold strip rolling mills are the exclusive equipment to produce strip steel products,and maintain its constant tension and ensure the steady-state accuracy of rolling speed are effective means to solve the quality problem of strip steel shape and thickness. To improve the tracking control precision of the speed and tension system of cold strip rolling mill,a extended state observers(ESOs)-based fixed-time prescribed performance control method was proposed. The ESOs are constructed to estimate the mismatched uncertainties of the system,and the observation values are introduced into the designed controllers for compensations,which improve the tracking control accuracy of the system effectively. The cold strip rolling mill speed and tension system controllers are designed based on prescribed performance function method and fixed-time control theory,which take into account the indicators of the system such as the convergence rate,overshoot and steady-state accuracy

The recognition of spray codes on steel plate surface is an important basis for realizing the tracking of steel plate materials in the cold zone.However,due to the difference of spray codes font shapes of different marking equipment,and the special circumstances such as manually handwritten steel plate numbers,the single model recognition method cannot have strong adaptability and ensure the accuracy of recognition.This article elaborated on the software process for implementing material tracking and analyzed the role and function of spray code recognition in it.A multi-level network spray codes recognition framework was designed,based on a lightweight object detection model,quickly locate and intercept the spray code area.The position of the steel plate number is accurately extracted by line segmentation model and prior knowledge,and the convolutional recurrent neural network (CRNN) model is used to achieve independent steel plate number recognition.The multi-level network model can achieve an overall steel plate spray codes recognition rate of over 99.2%,and　further improve the recognition accuracy by comparing it with the production plan.By interacting with the database to obtain steel plate details,it can provide effective data support for the material tracking process.

Hoist is the key equipment of dry quenching coke charging system,which has large load,frequent start and stop,unstable working condition,complex gear box structure and large reduction ratio.Therefore,the continuity and reliability of the hoist has always been the focus of production and equipment managers.An artificial intelligence algorithm based on multi-dimensional perception and neural network was proposed by the paper to identify the instantaneous fault of hoist equipment in real-time and monitor the hoist life cycle.By extracting and identifying the acoustic and vibration characteristics of the hoist gear box,combined with convolutional neural network (CNN) and long-short term memory (LSTM) neural network algorithm,extracting vibration space characteristics and time operating characteristics,judge and predict the operating state of equipment,so as to achieve real-time monitoring of equipment operating state,fault diagnosis and predictive maintenance.Engineering application practice shows that the method can effectively achieve 95.3% accuracy of fault identification,and can quantify the specific stage of the hoist′s operation life cycle,so as to remind maintenance personnel to respond to equipment faults in time,achieve predictive maintenance of hoist,reduce equipment failure rate,reduce maintenance costs,achieve continuous safe production,promote equipment professional management and smart factory construction.

Sensitive steel grades with billet cracks are prone to cracking at the corners during continuous casting. In order to ensure continuous production and solve product quality problems,a digital model and quality analysis algorithm for the casting billet have been established,forming a reliable and efficient analysis system. System achieves meter level tracking and positioning through online collection of high-frequency time series data and quality data,dynamically calculating the stability of process variables,an improved self organizing map (SOM) algorithm was developed based on random forest (RF) and Fisher忆s linear discrimination analysis (FDA). The algorithm establishes a factor model through variable screening,dimensionality reduction,and stability calculation,achieving high-dimensional data compression while preserving its spatial topology structure and projecting it onto a
two鄄dimensional plane for visualization,achieving dynamic tracking of corner crack risk rating and process production path. The accuracy of model prediction is maintained at over 90% ,achieving process optimization and online monitoring. Since the system was put into use,the incidence of corner cracks in typical steel grades has decreased from 35. 3% to 8. 3% .

In the hot rolling strip process,the coiling temperature is an important process parameter and the main control objective,which can to some extent determine the strip steel microstructure,and affect the mechanical properties and usability of the product.To improve the accuracy of hot strip coiling temperature,based on the actual production data of a hot strip rolling line,a data driven coiling temperature prediction model for hot strip rolling was established using random forest (RF) algorithm.The Bayesian optimization algorithm is used to determine the optimal hyper-parameter of the RF model,and the grid search is used to determine the Bayesian algorithm hyper-parameters.At the same time,an decision tree model (DT) optimized by Bayesian optimization algorithm,support vector regression model (SVR) and mechanism model based on classical heat transfer theory are used for comparison and verification.The model testing results show that over 97% of the prediction results of the RF model have a prediction error within -10-10 ℃ for sample points.Compared to on-site models,it can predict the coiling temperature and further improve the control accuracy of coiling temperature.