Key Engineering Materials Vol. 1050

Abstract: Kolmogorov-Arnold networks (KANs) have emerged as a promising counterpart to multi-layer perceptrons (MLPs) which offer a more interpretable functionality for different machine learning(ML) applications. Their main difference lies in the definition of KAN layers, using learnable activa-tion functions, which has made these networks optimal for physics-based applications. In this work,we focus on analyzing the performance of KANs in capturing the physics of the hot rolling process,which is an integral part of steel manufacturing industry. Initially, we introduce non-dimensional pa-rameters to encapsulate geometrical factors in the process. We perform space-filling sampling in thespace spanned by these parameters. The sampled points yield the necessary parameters for the finite el-ement (FE) simulations, forming the ground truth (GT) data for the network. A closed-form analyticalmodel for spread is considered from previous studies in the literature, and its predictive performanceis assessed against the FE results. In defining the input space for the network, different alternatives arecompared and it was seen that input space containing the non-dimensional features and the predictionsof the analytical model reduced overfitting and better generalization. The effect of KAN hyperparam-eters are evaluated, and the network with tuned parameters demonstrate optimal performance on thetest set. Lastly, after applying symbolification for this network, a closed-form expression is obtainedthat captures the discrepancy between the analytical model and the GT results, and its performance istested against test set data.

Abstract: The mechanical joining of continuous fiber-reinforced thermoplastics (cFRTP) and metal sheets represents a promising approach for manufacturing hybrid lightweight structures. To reduce the time and cost associated with extensive experimental investigations, numerical modeling strategies are increasingly applied. In this numerical study, a further step in the modelling strategy for the direct pin-pressing (DPP) process of cFRTP and metal sheets is presented. The study focuses on modeling and simulating the occurring deformation mechanisms of decomposition, compaction, and separation of individual rovings on the mesoscale to analyze the resulting material structure. For this purpose, two simplified models were derived. The textile architecture is represented based on micrographs of cross-sections and discretized using the finite element method. The deformation of individual rovings during joining leads to a deformation of their initial elliptical cross section. To capture this level of resolution, both a cohesive zone and a pure contact approach are applied within the rovings. The highly viscous thermoplastic melt is modeled as a fluid employing the Arbitrary Lagrange–Eulerian (ALE) method. Matrix and roving meshes are coupled to account for fluid–structure interaction (FSI) during process. The study shows that coupling of matrix and rovings is necessary to obtain more accurate predictions of the deformation behaviour. Furthermore, the cohesive zone approach is better suited to simulate the emerging deformation mechanisms.

Abstract: Roll forming pipes for hydrogen infrastructure poses particular challenges for process design, especially with regard to geometric accuracy and the avoidance of forming defects that could compromise the integrity of the pipelines. Geometric accuracy is crucial to ensure uniform pressure distribution within the pipe. Conventional trial-and-error approaches to developing roll flower designs are time-consuming and cost-intensive, especially when working with high-strength steel grades. This work presents an integrated methodology for roll forming of monolithic sheet by incorporating real-world machine stiffness and experimental anisotropy. A finite element model was developed for S235 and S355 steels, validated through three-point bending and Digital Image Correlation (DIC). While database-derived models (JMatPro) underestimated yield stress by up to 30%, the experimental model precisely predicted strain distributions (error < 2%). A central novelty is the integration of in-situ 3D laser scans of the roll forming mill under load, allowing the simulation to account for elastic machine deflection. This enables the prediction of process-induced residual stresses, which are critical for the long-term integrity of pipelines against hydrogen-induced cracking.

Abstract: Laser surface texturing (LST) is an effective technique for tailoring the surface properties of Ti6Al4V alloy, widely employed in biomedical applications where surface topography plays a key role in osseointegration and functional performance. Nevertheless, the strong nonlinear relationship between laser process parameters and resulting surface roughness still limits predictive control of laser-textured surfaces. This work presents an experimental study aimed at investigating the influence of laser surface texturing parameters on the surface morphology of Ti6Al4V. Key process variables, including laser power, scanning speed, pulse frequency, pulse duration and overlap percentage are systematically varied using a fiber laser system. The textured surfaces are characterized through three-dimensional surface roughness parameters, namely Sa, Sz, Sku, Svk, and Ssk, providing a detailed quantitative description of surface topography relevant for biomedical applications. The resulting experimental dataset represents a fundamental basis for the subsequent development of artificial intelligence models, based on neural networks, for predicting surface roughness parameters as a function of laser processing conditions. The proposed approach supports data-driven optimization of laser surface texturing processes within intelligent and sustainable manufacturing frameworks.