[[000-publish/Projects|◀️Back]] ![[Logo_FESR.png]] ### Project Information Title: _High Performance Computing infrastructure for Numerical Analysis of Metal Additive Manufacturing Systems Using Artificial Intelligence_ Acronym: AI4AM Project type: **ERDF 2021-2027**. Start date: 01/02/2025 End date: 31/12/2027 ### Funding This project is funded by the European Union through the European Regional Development Fund (ERDF) under the 2021–2027 programming period. The initiative benefits from financial support aimed at fostering applied research, technological innovation, and knowledge transfer in the field of additive manufacturing. The EU contribution enables the implementation of experimental development activities, the establishment of advanced computing infrastructure, and collaboration between academic and industrial partners. ### Abstract Additive Manufacturing (AM), also known as 3D printing, is a manufacturing process that creates objects from three-dimensional (3D) digital models by adding material layer by layer, as opposed to subtractive manufacturing methodologies such as traditional machining. The main advantages of AM for the industry include its ability to create complex ultra-lightweight structures, integrate functions, use multiple materials for the same part during a single production run, and optimize the weight and shape of the fabricated parts. The adoption of AM at a large scale is considered one of the key drivers of the **Fourth Industrial Revolution**, which is an ongoing automation effort that will radically innovate industrial processes using advanced smart technologies. By streamlining processes, eliminating the need for tooling and handling and reducing material-related energy demands, this technology can help the transition to a more sustainable economy and holds immense promise for stimulating economic growth and fostering a culture of innovation and entrepreneurship for local businesses. Although the technology has been established for almost three decades, its optimisation is still performed mainly by trial and error experimentation, which remains a lengthy and expensive process. Numerical modeling and computer simulations can play an important role to overcome several shortcomings that can impact the quality of the as-built parts, including porosity defects, keyhole effects, balling phenomenon, cracks and residual stresses, surface roughness and irregularities, material properties and powder contamination. They can assist to predict mechanical properties, thermal behavior, and microstructure evolution, helping to identify suitable materials and optimize process parameters. In this project we propose **machine learning-assisted numerical finite element simulations** of metal AM systems with minimal trial and error experiments that are suitable for a parallel implementation on modern **petascale and pre-exascale computer systems**. We will explore complementary cooperation models where the training data are produced by parallel finite element methods developed in our research, in order to realize the huge improvement in computational speed that has been identified as being necessary to enable a seamless transition to AM-enabled manufacturing processes. Our proposal contributes to the establishment of a new **high-performance computing (HPC) infrastructure** for the **Artificial Intelligence Laboratory (AI-Lab)** at the **Free University of Bozen/Bolzano (UniBZ)**, with the goal of aggregating existing academic expertise at UniBZ in AI methodologies and of bridging the gap between academic excellence and local needs. Furthermore, the infrastructure will not only be needed for carrying out the current project but also enhance the competitiveness of UniBZ in applying for and pursuing future scientific initiatives in various fields. By expanding our capabilities and resources, we aim to position UniBZ as a leader in AI research and innovation, fostering collaborations and driving important benefits to both academia and industry. ### Objectives The project aims to develop and implement an innovative infrastructure and methodological framework to advance the state of the art in additive manufacturing (AM). Specifically, the objectives are: - To establish a new **high-performance computing infrastructure** at the Free University of Bozen-Bolzano (UniBZ) dedicated to artificial intelligence and numerical simulation of metal additive manufacturing processes. - To design and develop **parallel finite element simulation models**, including adaptive mesh refinement and Krylov subspace solvers, suitable for execution on petascale and pre-exascale computing systems. - To integrate **machine learning and neural network techniques** into simulation workflows to enable accurate prediction of material properties, temperature distributions, and microstructural evolution during AM processes. - To reduce the reliance on costly and time-consuming trial-and-error experimentation by leveraging advanced modeling and data-driven approaches. - To foster knowledge transfer, industrial collaborations, and local innovation capacity in the field of smart manufacturing and Industry 4.0 technologies. ### Expected Results The project is expected to achieve the following outcomes: - Creation and operational deployment of a **Dell PowerEdge R6625 HPC server**, enabling parallel simulations and AI model training. - Development of a suite of **open-source numerical codes** for modeling and simulating laser powder bed fusion processes, integrating adaptive refinement and advanced solvers. - Implementation of **physics-informed neural networks** for accelerating simulation workflows and improving predictive capabilities. - Production of high-quality datasets and simulation results that will support optimization of additive manufacturing parameters and improve the reliability of AM processes. - Strengthening UniBZ’s capacity to participate in international research projects, attract funding, and act as a regional hub for advanced manufacturing innovation. - Contribution to environmental sustainability through reduced material waste, energy efficiency, and optimized design processes. ### Principal Investigator - [Carpentieri, Bruno; Faculty of Engineering (Free University of Bozen-Bolzano), Tenured associate professor)](https://boris.unibz.it/converis/mypages/browse/Person/23700705) ### Team Members - [Calvanese, Diego; Faculty of Engineering (Free University of Bozen-Bolzano), Full Professor)](https://www.unibz.it/it/faculties/engineering/academic-staff/person/3562-diego-calvanese) - [Novotny, Ladislav; Faculty of Engineering (Free University of Bozen-Bolzano), Research Assistant)](https://www.unibz.it/it/faculties/engineering/academic-staff/person/45047-ladislav-novotny) ### UniBZ main research areas - Data-driven Artificial Intelligence (D2AI) (Faculty of Engineering) - Energy Resources and Energy Efficiency (Faculty of Engineering) - Industrial Engineering and Automation (IEA) (Faculty of Engineering)[[000-publish/Projects|◀️Back]] ### Scientific Reports - [Rapporto di avanzamento progetto - 15.02.2025 – 31.10.2025](https://www.inf.unibz.it/~bcarpentieri/Report_Year_1_AI4AM.pdf) ### Publications 1. Adaptive Parallel Methods for Polynomial Equations with Unknown Multiplicity - M. Shams, B. Carpentieri - Algorithms 19 (1), 21, 2025. DOI: https://doi.org/10.3390/a19010021 2. Convergence-Enhanced and ANN-Accelerated Solvers for Absolute Value Problems - M. Shams, B. Carpentieri - Axioms 14 (12), 880, 2025. DOI: https://doi.org/10.3390/axioms14120880 3. Efficient Hybrid ANN-Accelerated Two-Stage Implicit Schemes for Fractional Differential Equations - M. Shams, B. Carpentieri - Mathematics 13 (23), 3774, 2025. DOI: https://doi.org/10.3390/math13233774. 4. A New Class of Dimension Expanded Preconditioners for Efficient Solution of Saddle Point Linear Systems - H. Aslani, D. Khojasteh Salkuyeh, W.H. Luo, B. Carpentieri - Mathematical Methods in the Applied Sciences, 1–14, 2025. DOI: https://doi.org/10.1002/mma.70100. 5. Efficient Hybrid Parallel Scheme for Caputo Time-Fractional PDEs on Multicore Architectures - M. Shams, B. Carpentieri - Fractal and Fractional 9, 607, 2025. DOI: https://doi.org/10.3390/fractalfract9090607. 6. Pressure-induced tetragonal distortion in niobium: Insights from synchrotron X-ray diffraction - M. Masoumi, B.R.C. Saraiva, P.W.C. Sarvezuk, L. Novotný, T.C. Andrade, H.F.G. de Abreu, M. Béreš - Materials Today Communications, 44, 111948, ISSN 2352-4928, 2025. DOI: [https://doi.org/10.1016/j.mtcomm.2025.111948](https://doi.org/10.1016/j.mtcomm.2025.111948 "Persistent link using digital object identifier") 7. A class of high-order fractional parallel iterative methods for nonlinear engineering problems: Convergence, stability, and neural network-based acceleration - M. Shams, N. Kausar, B. Carpentieri - Chaos, Solitons & Fractals, 199(2), 116646, ISSN 0960-0779, 2025. DOI: https://doi.org/10.1016/j.chaos.2025.116646 8. Chaos-Enhanced Fractional-Order Iterative Methods for the Stable and Efficient Solution of Nonlinear Engineering Problems - M. Shams, B. Carpentieri - Algorithms, 18(7), 389, 2025. DOI: https://doi.org/10.3390/a18070389 9. Efficient families of higher-order Caputo-type numerical schemes for solving fractional order differential equations - M. Shams, B. Carpentieri - Alexandria Engineering Journal, 124, 337-361, ISSN 1110-0168, 2025. DOI: https://doi.org/10.1016/j.aej.2025.02.111 10. A new collocation method for solving some first-order differential models of biological systems - M.A. Rufai, B. Carpentieri, A.A. Alimi, M.A. Babalola - Scientific African, 28, e02735, ISSN 2468-2276, 2025. DOI: https://doi.org/10.1016/j.sciaf.2025.e02735 11. A High-Order Fractional Parallel Scheme for Efficient Eigenvalue Computation - M. Shams, B. Carpentieri - Fractal and Fractional, 9(5), 313, 2025. DOI: https://doi.org/10.3390/fractalfract9050313 12. Application of Variable Step‐Size Hybrid Methods for Solving Third‐Order Lane‐Emden Equations - M.A. Rufai, T. Tran, B. Carpentieri, H. Ramos - Mathematical Methods in the Applied Sciences, 2025. DOI: https://doi.org/10.1002/mma.11147 ### Conferences - SIMAI (Società Italiana di Matematica Applicata e Industriale) Conference (https://simai2025.cimne.com/) - SISSA Trieste (Italy), 1-5 September 2025 - Invitation to the minisymposium "MS031 Preconditioning Techniques for Large-Scale Scientific Applications". Title of the talk: _Preconditioning and Iterative Methods for Solving Large and Ill-Conditioned PageRank Problems_ ### Related Projects - [[000-publish/Projects/SmartPrint|SmartPrint]] - [[PREDICT]] - [[HiTech Manufacturing]] ### Some images ![[FESR_1.png]] **Figure 1.** Finite element model of the Laser Powder Bed Fusion (LPBF) process: (a) meshed cuboidal domain with 34 layers, (b) temperature distribution during progressive element activation showing the localized heat input, and (c) thermal field evolution illustrating steep gradients near the melt pool.