ArcelorMittal (Luxembourg)

Abstract Despite advances in joint modelling, the ongoing demand for more efficient and sustainable steel structures calls for further improvements in the design of connections. This paper presents part of a broader research effort aimed at enhancing the design efficiency of steel joints by realistically capturing their behavior and enabling the development of more accurate and reliable design criteria. A novel application of Physics‐Informed Neural Networks (PINNs) is proposed to model the moment–rotation behaviour of welded beam‐to‐column joints. PINNs are trained on a parametric dataset generated from validated FEM models, embedding analytical expressions for initial stiffness as a physics constraint. The results show excellent agreement with FEM simulations in the elastic range and promising accuracy beyond it. By replacing FEM for large‐scale data generation, PINNs unlock the stochastic characterisation of joint behaviour within the probabilistic framework of EN 1990, supporting Monte Carlo simulations and the derivation of statistically robust resistance criteria. Future developments will extend the approach to other joint configurations and integrate the results into macroelements for system‐level digital workflows, supporting the advancement of the Direct Design Method.

THz pulsed imaging (TPI) has emerged as an essential nondestructive evaluation (NDE) technique, particularly valued for its ability to deliver detailed stratigraphic information and conduct precise structural analyses across diverse materials. Despite its advantages, achieving accurate identification of genuine peaks and suppression of false signals is often faced with challenges like low signal-to-noise ratio (SNR) and overlapping echoes in cases involving thin layers. Although sophisticated signal-processing approaches, including sparse deconvolution (SD) and orthogonal matching pursuit (OMP), have demonstrated potential in overcoming these limitations, their success heavily depends on meticulous parameter selection, typically requiring extensive user experience. This work introduces an innovative technique aimed at improving both the accuracy and ease of parameter tuning in THz TPI. Specifically, it involves applying SD directly to the impulse response after it has undergone time-domain filtering instead of raw ones. The practicality and adaptability of our proposed approach are validated through experiments. The reconstructed impulse response clearly illustrates the versatility and enhanced precision offered by our method.

In the hot-rolling process of steel strip, cooling generates a surface oxide known as mill scale. If mill scale is not ultimately adequately removed, it compromises the adhesion of subsequent coatings, accelerating corrosion. Typically, mill scale removal is achieved via acid pickling, which, while effective, produces toxic waste and risks surface damage if excessively applied. Hence, accurately determining mill-scale thickness is important for pickling optimization, ensuring steel durability and minimizing waste. In this study, we employ THz time-of-flight tomography, a nondestructive testing method, to measure mill-scale thickness. Due to the sometimes optically thin mill-scale layers at terahertz frequencies, a back-propagation neural network is used to processes for fast, precise thickness estimation. This study explores two neural network strategies: a regression model and a classification model. Both utilize THz data to predict mill-scale thickness, typically ranging between 5 to 15 μm. The regression approach achieves a root mean square error (RMSE) of approximately 1.6 μm, while the classification approach sorts samples into three thickness categories with 87 % accuracy on testing data.

Abstract In 2020, the construction sector contributed to 37 % of global CO2e emissions. While efforts have primarily targeted reducing environmental impacts related to building operations through enhanced energy efficiency, the significance of addressing embodied impacts (arising from materials/products) has recently garnered attention. Underground car parks (UCPs) mitigate the scarcity of parking space in densely urbanized regions, though they pose unique design and construction challenges not encountered in above‐ground structures. This paper seeks to compare the environmental performance of retaining walls within the excavation pit of an UCP associated with an office building in Berlin, Germany, over a service life of 50 years through a life cycle assessment (LCA). Four structural alternatives are evaluated: (a) permanent steel sheet pile, (b) temporary steel sheet pile with a permanent reinforced concrete wall, (c) permanent concrete secant pile, and (d) permanent concrete diaphragm wall. Utilizing the life‐cycle framework outlined by EN 15798, the LCA encompasses the product stage, construction process, repair, end‐of‐life, and benefits and loads beyond the system boundary. Predominantly based on environmental product declarations in accordance with EN 15804+A2, the LCA meets ISO 14044 requirements and has successfully undergone critical review by an independent panel comprising three German experts.

AHSS grades present a good resistance/formability compromise to produce structural and safety parts for automotive. In order to improve their resistance to corrosion, these steels can be coated with a zinc layer by hot dip galvanization. After cold rolling, the steel strip is annealed under controlled atmosphere and temperature according to a well-defined thermal cycle to obtain the targeted microstructure, then it is immersed in a liquid zinc bath at 460 °C. The annealing performed under reducing atmosphere composed of N2 + 5% H2 induces a significant absorption of atomic hydrogen in the bulk. At the exit of the galvanizing bath, the zinc layer in surface of the strip constitutes a barrier to the hydrogen desorption during the cooling and the storage of the coil at room temperature before cutting and stamping by the customer. This hydrogen absorbed during the process can be harmful for the zinc coated steel grade by inducing a deterioration of its formability or delayed fracture phenomena after cold stamping. It is then important to control this absorbed hydrogen and to be able to predict the content introduced in the steel as a function of the annealing parameters and the evolution of microstructure during this annealing. A numerical model was developed by FEM simulation to predict the hydrogen content absorbed in the steel during the galvanization annealing as a function of annealing parameters and microstructure. This physical model uses Sievert and Fick’s laws to estimate the solubility and the diffusivity of hydrogen in the steel as a function of the annealing temperature. In addition, the influence of trapping by microstructural defects in the different phases during the annealing is described in the model by McNabb and Foster’s approach.