To develop a high-accuracy fatigue life prediction model for AISI 316 and AISI 430 under combined axial and internal pressure multiaxial proportional and non-proportional loading conditions, the Bi-LSTM and Bi-GRU deep learning algorithms were employed. In addition, two conventional machine learning algorithms, namely random forest (RF) and support vector regression (SVR), were utilized for comparison. The obtained results indicate that the von Mises equivalent stress approach, as well as the RF and SVR models, failed to accurately predict fatigue life under such complex loading conditions. In contrast, the Bi-LSTM model achieved higher prediction accuracy for AISI 316, with almost all predicted data points aligning closely along the diagonal line, except for one outlier. On the other hand, the Bi-GRU model exhibited better predictive performance for AISI 430. Furthermore, 50 independent holdout splits were conducted for both Bi-LSTM and Bi-GRU models to analyze the RMSE distributions of the testing subsets. The results were found to be consistent with the predictive outcomes, confirming the reliability and stability of the developed deep learning models.

The surface defects of continuous casting slab have an extremely negative impact on the quality and productivity of steel. Detection of surface defects poses significant challenges in real-time production line, such as excessive reliance on human intervention, low efficiency, and limited measurement accuracy. To tackle these challenges, this paper proposes a lightweight deep learning model called CCSNet, which can achieve efficient feature fusion while reducing computational costs. The new model solves the balance between accuracy and real-time performance of slab surface defect detection. Firstly, real-time photos of the slab surface were obtained using the visual inspection platform installed on the continuous casting production line. From these photos, a dataset containing five distinct kinds of defects was produced. Next, the model composition is introduced, and the feature information required for detection is generated by training the dataset. Finally, the suggested model’s validity is demonstrated by model testing and comparison with other models. The CCSNet achieves an average accuracy of 90.1% on the self-made dataset, with the model’s weight size being only 6 MB. The detection performance surpasses other traditional models, which can work for real-time defect detection under actual conditions.

Accurate prediction of phosphorus content in molten steel at the end of the electric arc furnace (EAF) smelting process is crucial for optimizing smelting efficiency and ensuring product quality. Inaccurate predictions can result in increased resource consumption and compromised product quality. This study proposes a novel FA-MM-TabNet model, which integrates Firefly Algorithm (FA) optimization, metallurgical mechanisms (MM), and the deep learning-based TabNet algorithm to predict endpoint phosphorus content in EAF operations. The model was trained and tested on actual production datasets and demonstrated superior performance compared to competing models, achieving a mean absolute error of 0.0024, a root mean square error of 0.0034, and a hit rate of 90.82% within a ±0.005% error margin. Furthermore, an interpretability analysis using explainable artificial intelligence methods, including model-specific and SHAP analyses, confirmed the model’s alignment with metallurgical principles. This enhances the model’s reliability and applicability in industrial settings, contributing to more efficient steelmaking practices.

In the production of high-speed wire rod, the precise measurement and grading of fine pearlite content in the metallographic structure of wire rod are critical for product quality control and process optimization. However, traditional manual grid counting methods suffer from low efficiency and high labor intensity. To address this, this paper proposes a deep learning-based detection method: first segmenting the lamellar pearlite and ferrite with lower content, then calculating the fine pearlite content through region-wise subtraction operations. The core of this method lies in addressing the complex structures of fine pearlite, lamellar pearlite, and ferrite in metallographic structures, which are interwoven with blurred boundaries. By improving the U-Net model (incorporating a VGG16 backbone network, a Focal Loss function, and attention modules), the method achieves over 95% segmentation accuracy. The fine pearlite content detected based on the segmentation results exhibits high precision, and its grading results are completely consistent with the manual grading results in steel plants, indicating the method’s practical applicability potential in fine pearlite content measurement and grading for steel mills.

Avoiding the contamination of tramp elements in steel requires the non-ferrous materials mixed in steel scrap to be identified. For this to be possible, the types of recovered steel scrap used in the finished product must be known. Since the thickness and diameter of steel are important sources of information for identifying the steel type, in this study, the aim is to employ an image analysis to detect the thickness or diameter of steel without taking measurements. A deep-learning-based image analysis technique based on a pyramid scene parsing network was used for semantic segmentation. It was found that the thickness or diameter of steel in heavy steel scrap could be effectively classified even in cases where the thickness or diameter of the cross-section of steel could not be observed. In the developed model, the best F-score was around 0.5 for three classes of thickness or diameter: less than 3 mm, 3 to 6 mm, and 6 mm or more. According to our results, the F-score for the class of less than 3 mm class was more than 0.9. The results suggest that the developed model relies mainly on the features of deformation. While the model does not require the cross-section of steel to predict the thickness, it does refer to the scale of images. This study reveals both the potential of image analysis techniques in developing a network model for steel scrap and the challenges associated with the procedures for image acquisition and annotation.

In material design, the establishment of process–structure–property relationship is crucial for analyzing and controlling material microstructures. For the establishment of process–structure–property relationship, a central problem is the analysis, characterization, and control of microstructures, since microstructures are highly sensitive to material processing and critically affect material’s properties. Therefore, accurately estimating the morphology of material microstructures plays a significant role in understanding the process–structure–property relationship. In this paper, we propose a deep-learning framework for estimating material microstructures under specific process conditions. The framework utilizes two deep learning networks: vector quantized variational autoencoder (VQVAE) and pixel convolutional neural network (PixelCNN). The framework can predict material microstructures from the transformation behavior given by some physical models. In this sense, the framework is consistent with the physical knowledge accumulated in the field of material science. Importantly, our study demonstrates qualitative and quantitative evidence that incorporating physical models enhances the accuracy of microstructure prediction by deep learning models. These results highlight the importance of appropriately integrating field-specific knowledge when applying data-driven frameworks to materials design. Consequently, our results provide a basis for integrating data-driven methods with the accumulated knowledge in the field. This integration holds great potential for advancing material design through deep learning.

Recently, remarkable advances have been made in statistical analyses based on deep-learning techniques. Applied studies of deep learning have been reported in various industrial fields, including the iron and steel-making industries. The production of iron and steel requires a variety of processes, such as the processing of ingredients, iron-making, casting, and rolling. Consequently, the data acquired from them are diverse, and various tasks exist that can be assisted by deep-learning algorithms. Hence, providing a summary of the application is helpful for researchers specializing in information science to grasp the current trend of applied studies on deep learning techniques and for researchers specializing in each field of the iron and steel-making industry to understand what types of deep learning techniques are being utilized in other specialized fields. Therefore, in this study, we summarize the current studies on the application of deep learning in the iron- and steel-making fields by organizing them into several categories of processes and analytical methodologies. Furthermore, based on the results, we discuss future perspectives on the development of deep-learning techniques in this field.

A rolling force prediction method based on transfer learning and Inception-LSTM neural network is proposed to address the problem of low efficiency in predicting rolling forces for individual rolling mill stands due to the complex rolling conditions and distribution differences in the collected process data. The Inception-LSTM neural network combines the spatial feature extraction of the Inception model and the time sequence modeling capability of the LSTM network to comprehensively capture the features in the rolling process, thus establishing a baseline prediction model. Then, the transfer learning method is employed to transfer part of the parameters and structure of the baseline prediction model to the new prediction model. Simultaneously, the model is fine-tuned to establish a new transferred model for rolling force prediction, which is compared and analyzed against the neural network prediction model without using transfer learning. Experimental results show that the model built with transfer learning is higher fitting accuracy than the model built directly for rolling force prediction, and the training time of the model is significantly reduced. It can be used for steel shape and thickness control and digital twin simulation of rolling process.

Accurate measurement of the position and shape of the blast furnace (BF) burden surface (BS) is crucial for automated and intelligent BF control. The detection method based on key points of BS can improve the accuracy of radar BS detection in challenging environments with strong interference, ensuring long-term detection accuracy stability across long maintenance intervals. However, transferring a successfully applied detection model from the original BF to a new target BF requires sufficient sample collection in the target BF and model retraining. In practical scenario, data collection is often constrained by plenty of factors like time, space, and personnel. Thus, reducing the number of samples needed for retraining, or implementing model transfer based on small samples, can significantly boost BS detection efficiency. This paper proposes a Conv-Adapter Transfer Net (CATNet) designed for the transfer problem of key point detection models in BFBS and small sample scenarios. The CATNet adopts a newly designed parallel Conv-Adapter transfer module, a small-scale deep learning network with a minimal parameter quantity. Compared to fine-tuning the entire backbone model with a large number of parameters, retraining this transfer module requires far less training data. Therefore, CATNet is more suitable for model transfer in small sample scenarios. The experimental results demonstrate that the CATNet can exhibit strong transfer capability and higher prediction accuracy than the commonly used pretrain-finetune method in small sample scenarios, thereby enhancing the efficiency of model transfer across different BFs.