Mechanical properties of steels depend on their internal microstructures. However, current mechanical property prediction methods mainly rely on composition and process data, making it difficult to fully reflect the effects of microstructure on mechanical properties. Thus, the MDCAFF-Net (multimodal dynamic cross-attention feature fusion network) was proposed. The MDCAFF-Net employed a dynamic cross-attention feature fusion module (DCAFFM) and a multiscale feature attention fusion module (MSFAFM) to improve the prediction accuracy of the model. Compared to the bagging regressor (BR), random forest (RF), and deep neural network (DNN), the MDCAFF-Net showed lower mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean squared error (RMSE) and higher R 2 values for yield strength (YS), tensile strength (TS), and elongation (EL). Furthermore, compared to the fully connected layer data concatenation method (FCLDCM), the MDCAFF-Net model for YS, TS, and EL showed improvements of 13.27%, 11.9%, and 25.38% in R 2, decreases of 59.42%, 49.21%, and 79.69% in MAE, lower MAPE by 60.58%, 48.53%, and 79.33%, and a reduction in RMSE by 55.24%, 46.51%, and 60.92%, respectively. As a result, the proposed MDCAFF-Net achieves precise predictions of mechanical properties based on multimodal data.

The evolution of microstructure during hot-rolling production constitutes a pivotal factor in determining the quality and properties of steel. However, the unique characteristic of steel materials lies in their intrinsic microstructural sensitivity to composition and processing conditions. Therefore, precisely establishing quantitative relationships between composition, processing, structure, and properties (CPSP) in the hot-rolling production process of steel, and thoroughly investigating the key characteristic parameters influencing performance, are of paramount importance for enhancing the properties of steel products. Based on this, this paper proposes a multimodal data fusion approach. By mining multi-source information, such as microstructural images, composition, and processing conditions, to capture richer features, it employs Gradient-weighted Class Activation Mapping (Grad-CAM) to visually analyze the correlation between microstructural characteristics (size, phase composition, etc.) and yield strength. The results demonstrate that this method achieves significantly higher computational accuracy than purely numerical models, while also revealing the crucial mechanism by which the bainite phase and its fine grains influence yield strength. This provides novel insights for accurately and comprehensively elucidating the complex nonlinear relationships between composition, microstructure, processing, and properties.

The flexible optimization control of the cooling path after hot rolling of strip steel has a significant influence on the phase transformation microstructure, which in turn determines the stability of mechanical properties and final product quality. To further analyze the complex coupling relationship between cooling path parameters and mechanical properties, a physically guided multi-objective optimization strategy is proposed for post-rolling cooling path control, combining continuous cooling transformation (CCT) diagram to achieve precise regulation of phase transformation microstructure and volume fraction. First, based on experimental data and machine learning algorithms, predictive models for the relationships among chemical composition, physical metallurgical parameters, and phase transformation temperatures are established, enabling accurate prediction of the CCT diagram. Subsequently, multi-objective optimization is applied to determine optimal cooling paths, with results compared against actual industrial production data. Finally, the optimization outcomes are validated through metallographic analysis and mechanical property testing, while the grain refinement strengthening theory is employed to analyze how CCT diagram-based cooling path optimization affects material properties. This approach achieves the goal of enhancing mechanical properties in hot rolled structural steels through controlled cooling path processes.

The high-temperature compressive deformation behavior of medium manganese steel using a four-roll reversible rolling mill is investigated, revealing the effects of different Mn contents on the thermal deformation behavior of oxidation products in the alloy. It is found that within the experimental temperature range, the higher the deformation temperature, the better the plasticity of the oxidation products. It was observed that increasing the Mn content refines the grains, enhances the deformation ability of the oxidation products, and improves the flatness of the interfaces. Since (Fe, Mn)O has a similar crystal structure to FeO, the addition of Mn refines the grains of (Fe, Mn)O, causing the deformation to be distributed across more grains under the same deformation amount, and thereby improving its plasticity. At the interface between Fe-Mn alloy oxidation products and the matrix, there exists a spinel-phase solid solution, which can deform together with the oxidation products and the matrix at high temperatures. It was found that with increasing the Mn content, the size and number of pores between the spinel phases increased. First-principles simulation calculations were used to verify this, showing that Mn promotes the generation of vacancies. The greater number of pores in the spinel phase can effectively relieve the compressive stress caused by rolling deformation, thereby improving the deformation capability of the oxidation products at the interface.

The influence of mechanical twinning on mechanical performances of advanced steels remains a central issue in alloy design. In the present work, we systematically investigated the effects of mechanical twinning on tensile ductility and impact toughness in Fe-0.6C-0.5Si-24Mn-0/3/5Al (wt.%) steels. We hence conclude that mechanical twinning is indispensable for achieving outstanding tensile ductility but not a prerequisite for attaining high impact toughness. Notably, high impact toughness can instead be realized through alternative toughening mechanisms, particularly in alloys with elevated stacking fault energy. These findings challenge the prevailing paradigm linking twinning to both ductility and toughness, and offer a new framework for designing highstrength steels with an optimized balance of properties for demanding applications.

Accurate determination of the relationship among composition, process, and properties is crucial for predicting the yield strength of HRB400E rebar and enhancing the stability of its yield strength. Four yield strength prediction models of HRB400E rebar built using the random forest (RF) algorithm are compared, and a reverse process design is conducted based on the optimal model. The first model is an RF model driven solely by industrial big data, the second is an RF model with optimized hyperparameters (optimized RF model), the third combines physical metallurgy (PM) with industrial big data (PM-RF model), and the fourth is a dual-driven model of optimized PM and industrial big data (optimized PM-RF model). In the establishment of the optimized RF and optimized PM-RF models, a dynamic hyperparameter optimization algorithm was introduced, employing the Optuna optimization framework to optimize the curve parameters of an improved particle swarm optimization (PSO) algorithm with sigmoid-like inertial weight (Optuna-S-PSO). During the establishment of the PM-RF and optimized PM-RF models, the newly introduced input parameters, including ferrite grain size ( d_ ) and the fraction of precipitated phases in ferrite ( f_p- ), were calculated using the PM model. The results demonstrate that the application of the Optuna-S-PSO algorithm, along with the inclusion of PM parameters, significantly improves the models’ prediction accuracy. Among these, the optimized PM-RF model exhibited the highest yield strength prediction accuracy, with the coefficient of determination, root mean square error, and mean absolute error values of 0.856, 4.00 MPa, and 3.31 MPa, respectively. Based on this model, the SHapley Additive exPlanation (SHAP) method was used to comprehensively analyze the effects of composition, rolling parameters, and microstructure on the material’s yield strength. Ultimately, the yield strength fluctuation range of multi-specification HRB400E rebar is effectively reduced by reversely designing the rolling speed based on the optimized PM-RF model.

The model-based correlation between the chemical composition, process parameters, and mechanical property of the steel lies at the heart of the design of rolling process optimization. Yet, the hot rolling process is characterized by tightly coupling, many variables, and nonlinearity. The complicated link between the chemical composition, process factors, and mechanical properties of the high strength steel makes it difficult to construct a mathematical equation. On the basis of industry data for hot rolling, thermodynamic methods were applied to compute the effective Ti concentration here. Random forest was used to create the corresponding relationship model of the chemical composition, process parameters, and mechanical property for the high strength steel, obtaining a high level of mechanical property prediction precision. The root mean squared error for predicting yield strength is 21.07 MPa, for predicting tensile strength it is 19.12 MPa, and for predicting elongation it is 2.18%. Using the same chemical composition billet in conjunction with multi-objective evolutionary algorithm based on decomposition algorithm algorithm and taking into account the limits of the process circumstances, the best designs for the hot rolling process of different strength level steels are accomplished. The viability of process optimization is determined by industrial tests and theoretical analysis of the strength increment.

The pursuit of high-quality steel production necessitates precise control of operational parameters, particularly in the pretreatment of hot metal (PHM), where traditional methods may lead to inefficiencies. Consequently, the hierarchical fusion learning architecture (HFLA), an innovative system leveraging statistical theory, machine learning, and intelligent optimization techniques, was presented for accurate sulfur content prediction in PHM. By employing a strategy-driven fusion approach, HFLA enhances feature extraction via stacked kernels, incorporating Tikhonov regularization to guide the meta-learner predictions. Validation results from steel mill production demonstrate an impressive 4.3

In situ observation of Fe-Cr-Ni alloys at 1100 °C under Ar using Laser confocal microscopy (LSCM), combined with electron probe microanalyzer (EPMA) and X-ray diffractometer (XPS) analysis, presents preferential selective oxidation of Cr to form Cr2O3. Oxide nucleation initiates at grain boundaries and evolves into spot-like and island-like structures with extended oxidation time. Thermodynamics and elemental diffusion kinetics analysis suggest that the sequence of Gibbs free energy for oxide formation is Cr2O3, Fe3O4, Fe2O3, and NiO. Moreover, the faster diffusion rates at grain boundaries compared to the bulk matrix drive preferential Cr oxidation and accumulation at these sites. First-principles calculations comparing Cr-doped γ-Fe (111) with undoped γ-Fe (111) show reduced adsorption energy and density of states for the doped surface, with the B-H site identified as the optimal Cr adsorption location. Cr doping also enhances O2 molecule dissociation efficiency, and hybridization of Cr p-orbitals with O d-orbitals provides evidence for Cr increased tendency toward selective oxidation.

5.5Ni cryogenic steel was developed through a microalloying design. A lamellar treatment, added between traditional quenching and tempering (QT) processes, is referred to as QLT process. By adjusting the reversed transformation austenite/ferrite phase content in the lamellar microstructure, a fibrous texture was achieved in 5.5Ni cryogenic steel. This adjustment promotes the redistribution of C, Mn, and Ni elements within the microstructure during the subsequent tempering process. This reduces the resistance to deformation of the microstructure by external forces and produces a higher number of high-angle grain boundaries. As the lamellarization temperature decreases, the formation of fiber structures reduces the tendency for martensite variants to form in the microstructure, encouraging the formation of reversed transformation austenite. The self-regulation of martensite variant stress, the blocking of cracks by high-angle grain boundaries, and the deflection effect of reversed transformation austenite on cracks enhance the impact toughness of 5.5Ni cryogenic steel. Additionally, during uniaxial tensile testing, when shear stress is parallel to the habit plane of the slip system, the material exhibits easier sliding and greater plastic deformation. Furthermore, the special textures (Brass, Goss, E-type and copper) in the microstructure promote uniform plastic deformation. Therefore, when shear stress aligns with both the slip plane and the special texture, the plasticity of the material is maximized. Using this process, 5.5Ni cryogenic steel with a yield strength of 658 MPa, elongation of 23.90

In this study, Cu-4Ti-1Ni alloy was fabricated through a "pre-aging-cold rolling-re-aging" process, and the microstructure evolution and the mechanism of mechanical and electrical property regulation during the multi-stage process were systematically investigated. Transmission electron microscopy (TEM) analysis revealed that during the 450 °C pre-aging for 4 h, in addition to the retention of the primary NiTi phase, the alloy matrix also precipitated nanoscale β′-Cu4Ti phases, which were uniformly distributed and provided stable nucleation cores and strain energy storage for subsequent deformation and re-precipitation processes. The high-density dislocations introduced by cold rolling interacted with the fine β′-Cu4Ti precipitates and NiTi phases formed during pre-aging, creating a dislocation entanglement network and promoting re-precipitation at the interfaces. After the final aging, the alloy achieved a multi-level microstructure consisting of nanoscale β′-Cu4Ti phases (with an average size of approximately 4.96 nm) uniformly distributed in the Cu matrix and microscale NiTi phases. This multi-level microstructure not only significantly enhanced the alloy’s strength to 1168 MPa through the synergistic effect of precipitation strengthening and dislocation pinning, but also reduced the Ti solute concentration in the matrix, increasing the conductivity to 15.07

This study conducted multiple cold rolling and aging treatments on Cu-3Ti alloy, systematically exploring the “primary cold rolling—aging + secondary cold rolling—aging” process, and deeply analyzed the influence of aging parameters at different stages on the mechanical properties, electrical conductivity, and microstructure evolution of the alloy. Through tensile tests, electrical conductivity measurements, and transmission electron microscopy (TEM) observations, the formation behavior of precipitates at each aging stage and their regulatory effects on the comprehensive performance were analyzed. The results show that the optimal performance is achieved when the first aging is carried out at 450 °C for 4 h, with a tensile strength of 997 MPa and an electrical conductivity of 18.7

Cu-Ti alloys are an attractive alternative to Cu-Be alloys. However, Cu-Ti alloys usually have low electrical conductivity compared with conventional Cu-Be alloys after conventional processing. In this study, a new processing scheme combining discontinuous precipitation (DP) and cold rolling deformation was adopted to investigate the synergistic optimization of mechanical properties and electrical conductivity in Cu-4Ti-1Ni alloy. The over-aging process was carried out at 400-550 degrees C for 1-36 h to obtain different aging microstructures. After that, 80 % cold rolling deformation was performed. The results indicated that the proper combination of properties was obtained after quenching from the over-aging process at 500 degrees C for 24 h, as the hardness was 180.9 HV and the electrical conductivity was 23 %IACS. Then, the sample was further cold rolled. After this process, its hardness reached 380.5 HV and the tensile strength was 1162 MPa, while the electrical conductivity decreased slightly to 21 %IACS after cold rolling, so an excellent combination of mechanical strength and electrical conductivity was achieved. The results of microstructure analysis indicated that the lamellar Cu3Ti precipitates formed during the DP process could decrease the lattice strain in the matrix and then improve the electrical conductivity. Furthermore, the addition of the Ni element could enhance the precipitation of stable NiTi intermetallic compounds, and then promote the DP behavior. The strengthening mechanisms of the cold rolled Cu-4Ti-1Ni alloy were dislocation strengthening and DP strengthening, and the NiTi and Cu4Ti phases also made some contributions.

Precise control of bainitic content and morphology is important for the development of high strength steels with good combination of strength, toughness, and ductility. However, due to elusive mechanisms controlling the bainitic morphology, classical theories of physical metallurgy are inadequate to provide a complete prediction of the bainitic transformation. Here, we propose a machine learning approach to predict both fractions of phase transformation products and their morphologies for different compositions and process parameters after hotrolling of steel. The modelling strategy is firstly to transform reheating and deformation parameters into physical factors such as parent austenite grain size and stored deformation energy, which can be more directly related to the phase transformation behavior. Secondly, a stacked machine learning model to classify phase transformation products and predict their components under different rolling and cooling conditions was developed using gradient boosting tree classification and support vector machine algorithms. The model is capable of discriminating the type of bainite and the phase fractions for different compositions and process parameters. The model is lastly applied to designing a few lean steel alloys and to optimizing their processing routes, which is validated through analyses of the final microstructure and mechanical properties.

In hot rolling of steels, both microstructural evolution and work roll/steel interfacial state are critical for high quality products. Unfortunately, they are taking place in forms of black boxes because they cannot be readily detected during productions. Therefore, accurate modelling evolutionary behaviours of microstructures and surface scales during hot rolling and changes of as-rolled mechanical properties has become significantly important. In this paper, typical semi-empirical models developed since the 1970s and data-to-data models by artificial neural networks are briefly reviewed for their advantages and disadvantages. Then, physical metallurgy guided machine learning is discussed for its superiority in logicalities and accuracies. For the newest development, industrial foundation models (IFM) are proposed to integrate different processes in hot rolling and accelerated cooling, by which recrystallizations of austenite grains, strain induced precipitations, mechanical loading, and changes of interfacial friction coefficients during hot rolling can be simultaneously worked out, and dynamic continuous cooling transformation diagrams are instantly generated to account for variations of mechanical properties based on deep learning and heterogeneous data. Finally, typical applications to hot strip/plate lines for high efficiency production and stable control of mechanical properties are described.

The Fe–Mn damping alloys possess considerable damping capacity, but their yield strength is rather low. The 800 MPa Fe–Mn alloy with expected damping capacity was designed by the combination of grain refinement and ε-martensite introduction. The yield strength can be greatly raised to around 700 MPa by refining grain size from 88.4 to 1.8 μm. Although there exist numerous stacking faults in the fine-grained alloy, the damping capacity is strongly deteriorated due to the suppression of thermally activated ε-martensite. We demonstrate that the stacking faults cannot provide effective contribution to damping capacity and hence introduce a considerable volume fraction of stress/strain-induced ε-martensite to raise damping sources, including ε-martensite and γ/ε interfaces, etc., by a small pre-strain. From this, the damping capacity can be improved, and the yield strength can be further enhanced from nearly 700 MPa to around 800 MPa. Thus, the combination of high yield strength and good damping capacity is realized.

Phase transformations during heat treatment are crucial in regulating the microstructural properties of maraging stainless steel (MSS). In this study, the evolution of chi phase in MSS during annealing and aging process was studied by HRTEM. The nucleation behavior of the chi phase in unannealed delta was observed, with an orientation relationship (330)chi // (110)delta, [113]chi // [011]delta. During annealing, the chi phase exhibited rapid growth, driven by the diffusion of Mo and Cr elements. Subsequently, in the aging process, the Laves phase was observed to nucleate at the incoherent interface between the chi phase and matrix, showing a tendency to grow toward the chi phase, which was an unprecedented phenomenon in steel materials. A schematic diagram illustrating the evolution mechanism of the chi phase in MSS was established. Additionally, the mechanical properties of the material were evaluated, and nanoindentation measurements showed that the high nanoindentation hardness of the chi phase, dispersed within the matrix, improved both the hardness and yield strength of the material.

A low-Ni steel has been developed to overcome the ductile-to-brittle transition of body-centered cubic (BCC) structures at -196 degrees C. The design of the Cr + Mo microalloy composition and simple ultra-fast cooling-lamellarizing-tempering (UFC-LT) process was used to achieve low-cost preparation. Its exceptional cryogenic impact toughness stems from its heterogeneous structure, a ferrite/martensite (F/M) dual-phase lamellar structure and dispersed reversed austenite (RA). The 5.5 % Ni steel with this microstructure exhibits a cryogenic impact energy of about 220 J at -196 degrees C, attributed to martensitic variant selection and a multiphase coordinated deformation mechanism. The lamellarizing treatment promotes the self-plasticity modulation of the martensitic transition by refining the Bain groups and by introducing FM to induce certain variants in the surrounding grains. Among them, high-frequency variants, in particular, create high-energy grain boundaries with a variety of misorientations, which improve cracks deflection and propagation resistance. In addition, the F/M heterogeneous structure suppresses strain localization during impact and promotes hetero deformation-induced (HDI) hardening for improved fracture resistance. Meanwhile, the synergistic plastic deformation of RA and F/M heterogeneous structure delays toughness fracture through the crack blunting and transformation-induced plasticity (TRIP) effect. This strategy can potentially replace 9 % Ni steel for cryogenic applications.

Ti micro-alloyed steels have been found widespread application due to their significant cost-effectiveness and exceptional property characteristics, such as high strength, good toughness and corrosion resistance. However, the microstructural evolution of Ti micro-alloyed steel during hot rolling is highly complex and cannot be directly measured, often being regarded as a black-box process. Therefore, a thorough investigation into the microstructural evolution behavior during the hot rolling process of Ti micro-alloyed steels is crucial for achieving microstructural control and enhancing product quality. In the presented work, a data-driven and mechanistic hybrid model for predicting and optimizing the microstructure and property of hot-rolled Ti microalloyed steel was proposed. Firstly, a model framework for simulating the microstructural evolution during the hot rolling process of Ti micro-alloyed steels was established based on the principles of physical metallurgy. Combining literature data and intelligent optimization algorithms, a self-learning method for the model parameters is implemented. Then, a machine learning algorithm was used to establish the linkage of microstructure with mechanical property, through which the microstructure and mechanical property could be calculated based on the chemical composition and process parameters of Ti micro-alloyed steel. The framework demonstrates calculation accuracy of 95.0 % for yield strength (YS), 94.0 % for tensile strength (TS) and 95.5 % for elongation (EL). Furthermore, by combining with a multi-objective optimization algorithm, an intelligent optimization method for the hot rolling process of Ti micro-alloyed steel was developed. Based on this approach, the Mn content in 510 L steel was successfully reduced by approximately 36 % of that in the traditional process through optimized hot rolling parameters.

Affected by the modes of bainite transformation, its microstructure, mechanical properties and its coordinated deformability with ferrite are changed. So that the deformation behavior and the mechanism of ferrite/bainite dual-phase steels are different at different deformation stages. In this study, SEM, EBSD and EMPA are used to investigate the effect of different intragranular and grain boundary bainite microstructures on the deformation behavior of ferrite/bainite dual-phase steels. The results show that: bainite transformation mode determines the morphology of carbide in the bainite, which in turn affects the bainite hardness. The bainite with high hardness can hinder the expansion and merging of microcracks, delay the occurrence of plastic instability in experimental steels, and improve the deformation property of the experimental steel at the later stage of tensile deformation. The bainite with low hardness has good coordination of deformation with the ferrite matrix, and the low-angle grain boundaries are conducive to the diffusion of stress to ferrite, which improves the deformation property of the experimental steel at the initial stage of tensile deformation. The bainite microstructure obtained through microstructure control has the advantages of the above two types of bainite and shows good deformation properties throughout the deformation process.