Continuous casting is a critical and complex process in steel manufacturing, in which product quality is highly sensitive to process parameter variations. Early identification of anomalous parameters and timely process optimization are, therefore, essential for reducing quality defects and production losses. However, existing data-driven approaches often face limitations in practical industrial applications, particularly in achieving high accuracy under strict response time requirements. In this study, a data-driven framework is proposed to address the above limitations by identifying key anomalous parameters and inferring feasible parameter adjustments under operational constraints, with interpretability analysis incorporated to enhance consistency with metallurgical knowledge. The proposed framework was trained and evaluated on a real-world continuous casting dataset comprising 191,254 records, where the data were temporally split, and the final 25 pct was reserved as an independent test set. Experimental results demonstrate strong performance in both anomaly identification and process optimization, achieving a Top-10 anomaly identification rate of 96.64 pct and an optimization hit rate within ±5 pct of 91.54 pct, demonstrating its effectiveness and practical applicability.

Continuous casting is a critical process in steel production, in which molten steel is solidified into cast products through continuous cooling. Solidification control significantly affects microstructural evolution, defect formation, and the mechanical performance of final products. To achieve such precise control, numerical analysis based on heat transfer and solidification models is widely used to calculate the temperature field in continuous casting. However, this approach struggles to strike a balance between computational accuracy and efficiency. In this study, molten steel flow and mold–shell interfacial heat transfer under various processes were investigated using numerical simulations. Based on simulation data, deep learning-based surrogate models were developed to predict molten steel velocity and interfacial heat flux. The surrogates were integrated into a finite-difference heat transfer and solidification model, forming a hybrid data-driven and mechanistic model for high-accuracy, real-time analysis of slab continuous casting. Experimental results showed that the proposed model reduced the absolute deviation of the surface temperature at the caster exit from 24 to 7 °C, with the relative error decreasing from 2.9 to 0.8

Online solution annealing is suggested as a low‐cost process for hot‐rolled austenitic stainless steel, utilizing residual rolling heat to soften work‐hardened microstructures. However, the coil rolled using the new process exhibits inferior pickling efficiency compared to conventional processing due to variations in oxide layer morphology and chromium‐depleted layer (CDL) formation. The effects of final rolling temperature, air‐cooling time, and coiling temperature on the microstructure of the oxide layer and CDL in 304 austenitic stainless steel are investigated through hot‐rolling experiments. Based on these investigations, the online solution annealing process parameters are optimized to improve the surface quality of the strip after pickling. Higher final rolling temperatures (1020–1060 °C), air‐cooling times of 5–10 s, and lower coiling temperatures (below 500 °C) result in an oxide layer structure that is easier to pickle. Ultimately, an industrial hot‐rolling experiment incorporating online solution annealing is conducted, followed by pickling and cold‐rolling processes. In addition to mechanical properties and resistance to intergranular corrosion, the surface quality of the product is not significantly different from that of the cold‐rolled coil produced by the conventional process.

Hot-rolled steel is widely used in the construction, ships, and home appliances due to its excellent mechanical properties and processability. With the rapid development of the manufacturing industry, surface quality has become a key element to measure product quality. Compared with carbon steel, austenitic stainless steel forms a dense Cr2O3 film, which endows it with excellent corrosion resistance. As a result, it is widely used in surface engineering applications, where stringent surface quality requirements are often imposed. Oxidation throughout the production process of austenitic stainless steel strips (reheating, hot rolling, coiling, and annealing) is influenced by the interaction of alloy elements and various process parameters. The structure of the oxide layer and the chromium-depleted layer determines the surface quality after pickling. Therefore, the evolutionary pattern of the surface oxidation state during process oxidation is reviewed, which is necessary to improve the surface quality of hot-rolled austenitic stainless steel and the competitiveness of the industry.

To completely cure the internal shrinkage cavity defects of continuous casting billets, the hot core heavy reduction rolling process (HHR2) is proposed. Through pilot-scale tests and numerical simulations, its mechanism of action and optimization approaches were deeply analyzed. The traditional process still cannot completely close the 4 mm void when the single-pass reduction rate is ≥ 22.2

The steel industry is a large and complex modern process industry, and continuous casting is one of its most important production links. It is a complex process due to the physical, mechanical, and chemical components involved. In recent years, machine learning techniques have become an indispensable part of the monitoring of complex production processes. However, the characteristic of the class imbalance problem in continuous casting industrial data has influenced the application of machine learning techniques. To overcome this limitation, a contrastive learning pretext task called similarity discrimination, and a two-stream neural network for predicting multiple defects in continuous casting have been proposed. The network effectively combines metallurgical knowledge with data-driven models. It was trained and evaluated on an industrial dataset. The two-stream neural network achieved an accuracy of 0.886 and a binary accuracy of 0.896 for predicting multiple quality defects in continuous casting. The comparative experiment results showed that it is at least 45.5 % higher than other machine learning methods in recall. To analyze the relationship between model components and model performance, ablation studies were performed. The contrastive learning pretext task and the new neural network architecture increased the accuracy and recall of the model by 7.1 % and 18.6 %, respectively. Furthermore, the interpretable machine learning technique was introduced to ensure the interpretability of the neural network. It enhanced the credibility of the machine learning systems, which helped users trust the model and predictions.

After hot rolling, 304 austenitic stainless steel requires a solution annealing treatment to prevent intergranular corrosion and eliminate work hardening effects. Compared to traditional offline processes, on-line solution annealing offers advantages in terms of cost and time savings. However, both recrystallization behavior and M(23)C(6 )carbide precipitation behavior are significantly influenced by the cooling process after rolling, which poses conflicting requirements. This study investigates the precipitation behavior of M23C6 carbides and the recrystallization softening behavior during the continuous cooling process of hot-rolled samples. The kinetics equations are derived using the Scheil's additivity rule. The temperature profiles in different regions of the plate are studied using finite element analysis. A practical approach for online solution annealing is proposed and applied in industrial testing.

Some heavy reduction technologies near the solidification end have been gradually applied to casting slabs or blooms in recent years to eliminate internal shrinkage cavities. These technologies can be roughly classified as single-pass reduction type and multi-pass reduction type. To date, the conclusions about the influence of reduction type on void closure are inconsistent. Herein, based on the simulation experiments, the influence of the reduction type on the closure of the shrinkage cavity in bloom during hot-core heavy reduction rolling (HHR 2 ) was investigated with the physical metallurgical model and numerical simulation. HHR 2 processes with single-pass type and double-pass type were designed to evaluate void closure behavior, and the total reduction amount of different reduction types was 60 mm in both cases. Results show that the single-pass reduction type of HHR 2 is more beneficial to the closure of the shrinkage cavity in bloom than the double-pass reduction type. Furthermore, the difference between the two reduction types was pointed out, which mainly includes three aspects: the geometry of the deformation zone, inter-pass static recrystallization and temperature of bloom. Finally, a quantitative analysis of the effect of the three abovementioned factors on void closure was performed. The obtained results show that the geometry of the deformation zone is the main factor, which leads to different degrees of void closure in different reduction types.

薄规格带钢轧后的层流冷却过程运行速度高、温降快,为了满足冷却过程实时性的要求,在线温度模型要有较高的计算效率.文章基于有限差分法建立显式格式和隐式格式的差分温度模型,分析了不同差分格式以及网格划分方法对模型计算效率的影响,提出了一种具有快速响应特性的在线温度控制模型.在保证计算精度不变的条件下,采用隐式差分格式进行较大的时间步长划分.使用自然对数的网格划分方法,有利于提高温度模型的计算效率,实现在线控制的快速响应.

The hot-core heavy reduction rolling (HHR2) technology is a solidification end reduction technique to enhance the quality of steel. In the current study, the Cellular Automata (CA) calculation is applied to investigate the casting process before HHR2 and the HHR2 dynamic recrystallization (DRX) process. The experimental data and calculation data are compared to verify the accuracy of the method and models. Afterward, based on the actual casting process, the solidification structure in the thickness direction of a billet is calculated, and the influence on the HHR2 grain size of various conditions, such as rolling speed, deformation temperature, and reduction strain, is simulated. Furthermore, based on the fixed HHR2 process, the influence of different superheats, casting speeds, and cooling conditions on the core-surface temperature schedule and the grain size before and after HHR2 deformation is calculated. Finally, the process optimization trend to produce high-carbon bearing steel billet is obtained.

Hot‐core Heavy Reduction Rolling (HHR 2 ) is a new technology that improves the core quality of blooms by using a heavy reduction roller at the solidification end. Dynamic recrystallization (DRX) is expected to optimize the microstructure in this process. But the difference of DRX characteristic between HHR 2 and conventional hot rolling (CHR) is an unclear problem. Herein, the differences of DRX behavior between two rolling processes are studied. If two rolling processes are carried out under the same deformation temperature and strain rate, the work hardening in CHR samples is greater than that in HHR 2 . Moreover, DRX occurrence in HHR 2 lags behind the CHR process due to lower growth rate of dislocation density achieved in HHR 2 . By using DRX models, the DRX area and DRX volume fractions in bloom are calculated. The great temperature gradient and low strain rate decrease the DRX critical strain in the HHR 2 process. The DRX region is mainly around the core of the HHR 2 bloom, where the DRX volume fraction achieves 41.8%. The maximum value of DRX volume fraction in CHR bloom only achieves 20.1% because the great strain rate and uniform temperature lead to a low level of deformation permeability.

热冲压成形汽车零部件的室温组织为全马氏体组织,虽然强度高,但延展性差.为此,提出了一种采用热轧后直接淬火获得马氏体组织,随后在冲压工序进行回火以提高冲压件延展性的温冲压成形工艺.采用热轧实验机和MMS-200热力模拟实验机模拟温冲压成形过程,并对实验钢力学性能和组织结构进行了分析.结果表明:随温冲压成形温度的升高及保温时间的延长,实验钢成形后抗拉强度和维氏硬度值不断下降,伸长率呈先上升后下降再上升的趋势.随成形温度的增加,实验钢组织由马氏体不断转变为回火马氏体、回火屈氏体和回火索氏体.在350℃C保温120～180 s,实验钢成形后力学性能最佳,抗拉强度超过1 500 MPa,伸长率大于8％,硬度值在425HV～440HV之间.冲压成形温度越高,对冲压设备所需求的力能参数越低.

Softening of work-hardened metals due to static recovery and recrystallization is an important issue during continuous cooling after hot deformation, which will affect the properties of metals. A mechanical test method called "twined double-pass mechanical tests", has been proposed to measure the softening fraction of a deformed metal during continuous cooling, and then both non-isothermal and isothermal softening data of 304 austenitic stainless steel have been measured with the mechanical tests. An attempt has been made to build the relationship between non-isothermal softening behavior and isothermal softening kinetic data. On the assumption that the softening fraction only depends on the softening fraction X and temperature T, the softening curve during continuous cooling has been predicted from the isothermal softening kinetic data with numerical method. A good agreement was observed between the predicted curve and those values measured with twined double-pass mechanical tests. The Scheil's additivity rule was extended from the prediction of incubation period for pearlite transformation to the whole range of softening fraction of work-hardened metals. The relationship between non-isothermal and isothermal softening reactions has also been investigated with Scheil's additivity rule, and the additivity of the softening process of work-hardened metals during continuous cooling has been verified.

Hot-core heavy reduction rolling is an innovative technology for continuous casting billet at the end of solidification. The deformation characteristics of GCr15 bearing billet were investigated over the temperature range of 1000–1300 °C with the strain rate of 0.001–10 s−1 by the Gleeble 3800 thermo-mechanical simulator. Firstly, the true stress–strain data are calibrated by the friction correction method to acquire accurate parameters near the solidus. Then, the Laasraoui-type constitutive model and dynamic austenite grain size model are established by the thermal compression results and corrosion experimental results, and the predicted values of the experimental interval are obtained by using the models. Meanwhile, the accuracy of models is verified by comparing the predicted curves with the experimental data. Last but not the least, the dynamic parameters between solidus and liquidus are predicted by the two models, and the difficulty of experimental collecting near the melting point is solved.

The cracks formed in the hot deformation are typical defects caused by plastic capacity limitations. For tested steels, the fracture behavior is associated with temperature, strain rate, and stress state. In this study, a highly crack-sensitive steel with Nb micro-alloying is selected as the research material to establish the fracture criterion considering the above factors comprehensively. First, compression and tension experiments are performed, and various specimen geometries are designed to produce different stress states in the tests at different temperatures and strain rates. Second, the critical stroke of these tests is extracted, and corresponding calculations are completed using the finite element method (FEM). The loading paths of stress states are independent of temperature and strain rate; however, they differ when distinct specimen shapes are considered in the tests. Third, a high-temperature Mohr-Coulomb criterion, which is applicable at elevated temperatures, is established by considering the effects of temperature, strain rate, and stress states on fracture. Finally, some examples of compressive and tensile tests are employed for verifying the formulated fracture criterion. By comparing the FEM and experimental results, it is found that the fracture criterion could accurately describe the fracture behavior during hot deformation with distinct stress paths.

Hot-core heavy reduction rolling (HHR2) is an innovative technology that uses a temperature gradient to improve the core quality of continuously cast steel at the end of solidification. In the current study, the influence of the temperature on the quality of continuously cast steel rolled with HHR2 is investigated through experiments and simulation. Four HHR2 processes with different surface deformation temperatures (800, 850, 950, and 1100 °C) are carried out. The results show that the surface deformation temperature of 1100 °C led to the best density quality of the steel compared with several other processes. Moreover, based on the cross-section temperature field and the influence of temperature on deformation resistance, an effective strain evaluation method of the billet center is proposed to analyze HHR2 billet deformed quality. Furthermore, the HHR2 processes and traditional hot rolling process are compared. The results show that the 1100 °C HHR2 process has a higher temperature gradient than traditional hot rolling with the same surface temperature, which enhances both the stress triaxiality ratio and effective strain to obtain a higher core quality billet.

高温黏塑性本构模型是连铸坯近凝固终点压下工艺数值模拟的基础,但该条件下的应力应变数据极为缺乏,严重限制了连铸新工艺的开发.利用热模拟实验对比研究了车轴钢在近凝固终点压下和常规热变形工艺下的流变行为.结合动态回复和动态再结晶理论构建了近凝固终点压下工艺下的本构模型.结果表明:在近凝固终点压下工艺下,类铸态组织奥氏体晶粒粗大,流变应力明显低于常规热变形工艺下的流变应力;同一变形量下,动态再结晶体积分数较大.本文构建的本构模型对不同变形条件下的应力预测值与实验值吻合较好,平均相对误差约为2.62％.

为了研究连铸坯热芯大压下轧制工艺对铸坯内部缩孔和表面开裂的影响,以EH47船板钢为研究对象,利用数值模拟方法,对比研究了热芯大压下轧制和常规热轧的工艺效果.结果表明,热芯大压下铸坯厚向温度梯度更大,芯部应变水平和厚向变形均匀性提高;当压下量为50 mm时,热芯大压下轧后残余孔隙体积比常规工艺小18.4％;由于热芯大压下铸坯表面的应变速率和变形温度较低,导致其侧表面和角部的开裂风险大于常规轧制,但不会造成表面裂纹缺陷.

连铸坯直接轧制技术作为一种变革性的绿色钢铁生产流程,目前主要用于超薄带和线棒材生产,近年来国内外逐步开始了中厚板直接轧制工艺的探索性工作.直接轧制工艺与常规热轧工艺相比,具有不同的温度履历和物理冶金学过程.选取Nb-Ti微合金钢为研究对象,从产品组织与性能的角度,探讨中厚板直接轧制工艺的可行性.采用炼钢-连铸-轧制中试试验,对比研究了直接轧制工艺及常规热轧工艺下中厚板产品的组织和性能,并基于动态再结晶模型,探讨了直接轧制工艺下试验钢的组织细化机制.研究结果表明,直接轧制工艺下,虽然连铸坯轧前未经过γ-α-γ相变过程,仍保留铸态粗大的奥氏体晶粒,但轧制过程中较大的芯表温差有利于变形向芯部渗透,芯部再结晶进行得更加充分,可以用形变再结晶机制代替常规热轧工艺的相变机制细化成品芯部组织,获得与常规热轧工艺相近甚至更优的显微组织与力学性能.

The hot-core heavy reduction rolling (HHR2) process makes use of the reverse temperature gradient of casting steel to eliminate the inner shrinkage and porosity defects during continuous casting. As the key connecting process between continuous casting and rolling, the charging process affects the microstructure evolution of workpiece rolled by HHR2 process. The direct rolling and hot-charging experiments were carried out for the workpieces rolled by HHR2 process, to investigate the effects of charging process on the microstructure and mechanical properties of hot-rolled steel plate. For comparison, the conventional cold-charging experiments were also performed on the workpiece without HHR2. The results show that the similar microstructures and mechanical properties can be obtained from the hot-rolled plates produced from the workpieces rolled by HHR2, which were followed by direct rolling or hot-charging process, as well as that of hot-rolled plate produced from the workpiece without HHR2 via cold-charging process. Moreover, the microstructure and mechanical properties at the core of the hot-rolled plate via direct rolling process are better than those without HHR2. It indicates that the γ-α-γ reciprocating phase transformation, which occurs during the cold-charging process, can be replaced by deformation recrystallization that occurs during the HHR2 process, to refine austenite grains.