Mariusz Wermiński

Mr. Mariusz Wermiński

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 April 2023 in Materials Research Proceedings
Verified
Conference Paper
Open access
PDF file
MWMariusz Wermiński
Mariusz Wermiński
KPK. Perzyński
K. Perzyński
MSM. Sitko
M. Sitko
LML. Madej
L. Madej
Evaluation of data transfer influence in coupled Monte Carlo finite element model on microstructure evolution predictions WERMIŃSKI Mariusz, PERZYŃSKI Konra

1 January 2025 in Materials Research Proceedings
Verified
Conference Paper
Open access
PDF file
MWMariusz Wermiński
Mariusz Wermiński
MSM. Sitko
M. Sitko
LML. Madej
L. Madej
Changes in the microstructure that occur during the continuous cooling of steels significantly influence their properties. Accurate prediction of these changes is essential for understanding material behavior during complicated thermo-mechanical cycles. Full-field methods for modeling microstructure evolution during phase transformations (PT) enable detailed analyses of material behavior, offering valuable insights for result interpretation. This study uses the cellular automata (CA) full-field method to model the ferrite phase growth during the austenite-to-ferrite transformation in a modular manner, with a focus on how the discretization of the computational domain affects grain growth behavior. Three approaches to mitigate the artificial influence of the CA regular grid onto the shape of grain during the growth process are proposed. These approaches use different-sized Moore neighborhoods and include weights calculated from the Gaussian function. They are also assessed both qualitatively and quantitatively, and the outcome is presented in this work.

9 July 2025 in Materials
Verified
Journal Article
Open access
PDF file
MWMariusz Wermiński
Mariusz Wermiński
Mateusz SitkoMateusz Sitko
Mateusz Sitko
Lukasz MadejLukasz Madej
Lukasz Madej
The prediction of microstructure evolution during thermal processing plays a crucial role in tailoring the mechanical properties of metallic components. Therefore, this work presents a comprehensive, multiscale modelling approach to simulating phase transformations in large-diameter steel rings during cooling. A finite-element-based thermal model was first used to simulate transient temperature distributions in a large-diameter ring under different cooling conditions, including air and water quenching. These thermal histories were subsequently employed in two complementary phase transformation models of different levels of complexity. The Avrami model provides a first-order approximation of the evolution of phase volume fractions, while a complex full-field cellular automata approach explicitly simulates the nucleation and growth of ferrite grains at the microstructural level, incorporating local kinetics and microstructural heterogeneities. The results highlight the sensitivity of final grain morphology to local cooling rates within the ring and initial austenite grain sizes. Simulations demonstrated the formation of heterogeneous microstructures, particularly pronounced in the ring’s surface region, due to sharp thermal gradients. This approach offers valuable insights for optimising heat treatment conditions to obtain high-quality large-diameter ring products.

1 January 2025 in Computer Methods in Materials Science
Verified
Journal Article
Open access
MWMariusz Wermiński
Mariusz Wermiński
Mateusz SitkoMateusz Sitko
Mateusz Sitko
Lukasz MadejLukasz Madej
Lukasz Madej
Understanding the local evolution of phase transformations in steels, particularly the γ (austenite) → α (ferrite) transformation, is crucial for controlling the microstructure and properties of steel components. Over recent decades, significant progress has been made in the numerical modeling of this complex phenomenon. This development has been driven by both scientific curiosity and industrial needs, especially in processes such as hot rolling, forging, thermal treatment, etc. The developed models have evolved from simple solutions based on local equilibrium to more complex approaches that consider local heterogeneities. Modern computational approaches, such as Phase-Field (PF), Level-Set (LS), Cellular Automata (CA), Monte Carlo (MC) or Vertex based simulations, allow for the precise reproduction of microstructural evolution considering local instabilities. They also enable the analysis of phase boundary motion in an explicit manner. These techniques also allow for direct integration with thermodynamic data and mechanical models, thereby better capturing the physical mechanisms of phase transformations, such as chemical composition, diffusion resistance, or the influence of deformation. An overview of the state of the art in this area is presented within the paper. The model’s concepts, assumptions, fundamental equations, advantages, limitations, and potential practical applications are summarized. Special attention is given to modeling the γ → α transformation by the Cellular Automata method. The importance of incorporating phenomena such as diffusion, nucleation, and growth is emphasized. The need for consistency between experimental results and simulations is also highlighted.

1 January 2026 in Computer Methods in Materials Science
Verified
Journal Article
Open access
MKMaciej Krzywda
Maciej Krzywda
MWMariusz Wermiński
Mariusz Wermiński
Szymon LukasikSzymon Lukasik
Szymon Lukasik
Amir H GandomiAmir H Gandomi
Amir H Gandomi
We propose Concat-Tree Genetic Programming (CTGP) to automatically discover graph neural network (GNN) architectures for binary classification of DFT-derived materials properties in the JARVIS-DFT database. CTGP is inspired by Cartesian genetic programming (CGP) and classical tree-based genetic programming (GP): genotypes are binary concat trees that decode into explicit, heterogeneous sequential GNN stacks, evolved by mutation-only elitist search. Seven tasks span electronic structure (metal vs. insulator), magnetism, photovoltaic efficiency proxies, topological band character, and thermoelectric transport. Using fixed JARVIS train/validation/test splits and published binarization thresholds, we compare four search configurations – (1 + 1) and (1 + 4) regimes, each with and without a learned surrogate filter – with MLP, RF, DT, and SVM baselines and fair GNN references across four representation levels (L0–L3).L0 uses atomic number only, L1 adds learned elemental descriptors, L2 adds bond-distance summaries, and L3 adds bond-angle statistics. Each configuration uses ten independent random seeds per task. The search yields compact architectures with tens to hundreds of thousands of parameters and millisecond-scale model-forward inference. Accuracy is reported with model complexity, search cost, test AUC, macro-F1, and property-aware analysis of discovered operator sequences. Fixed GNN baselines and random search often match or exceed CTGP on test accuracy, while CTGP remains competitive on selected electronic-structure tasks and provides architecturally transparent stacks for exploratory, property-related interpretation, without claiming causal explainability. On benchmarks with a stable signal above majority-class performance, these surrogates may support rapid screening of large preprocessed graph libraries at low inference cost.

11 January 2025 in arXiv
Verified
Preprint
Open access
PDF file
MKMaciej Krzywda
Maciej Krzywda
MWMariusz Wermiński
Mariusz Wermiński
SŁSzymon Łukasik
Szymon Łukasik
AGAmir H. Gandomi
Amir H. Gandomi
This paper presents the application of Kolmogorov-Arnold Networks (KAN) in classifying metal surface defects. Specifically, steel surfaces are analyzed to detect defects such as cracks, inclusions, patches, pitted surfaces, and scratches. Drawing on the Kolmogorov-Arnold theorem, KAN provides a novel approach compared to conventional multilayer perceptrons (MLPs), facilitating more efficient function approximation by utilizing spline functions. The results show that KAN networks can achieve better accuracy than convolutional neural networks (CNNs) with fewer parameters, resulting in faster convergence and improved performance in image classification.