Aerobase Innovations AB is part of GENMAT, a Horizon Europe research and innovation action (Project ID 101295332, topic HORIZON-CL4-INDUSTRY-2025-01-DIGITAL-61). The project runs from 2026 to 2029, brings together 14 beneficiaries, and is coordinated by Politecnico di Torino. GENMAT is building Holistic-M3FM, a multi-scale, multi-modal foundation model for materials science. Aerobase is the consortium’s finite element specialist, an SME partner whose job is to connect AI-driven material discovery to structural deployment, so that a model’s prediction becomes a part that an engineer can actually load.
A new material still takes 10 to 20 years to reach the market. GENMAT is built to compress that.
The Challenge: Compressed Timelines and Multimodal Data Ecosystems
Moving a material from first synthesis to a qualified product still takes 10 to 20 years, a pace the energy and mobility transitions can’t wait for. The evidence gathered along the way is scattered: chemistry, spectra, micrographs, process logs, and sensor streams sit in different formats and rarely connect. Today’s machine-learning tools often make this worse. Most are trained for a single task, behave as black boxes, and ignore the physics that governs how a material deforms and fails; few carry over from the atomic scale to a full structure. GENMAT addresses both problems together, with a FAIR data layer that links the modalities and a single model that maintains physically consistent predictions across scales.
One model spanning every scale, from atomic chemistry to in-service performance.
Key Innovations by Aerobase in the GENMAT Project
1. Physics-informed foundation modeling. Aerobase helps build and fine-tune Holistic-M3FM with physics-informed neural operators (PINO) and physics-informed neural networks (PINN). These embed conservation laws, structural constraints, and failure criteria within the model, so its output remains physically valid rather than merely plausible-looking. Uncertainty quantification marks where confidence drops, which is exactly what you want to know the moment a prediction feeds a safety margin.
From scattered data to a designable material: the predictive engine Aerobase helps train.
2. From prediction to structural mechanics. Aerobase converts model output into finite-element load cases and compares them against how composites actually respond. The work covers anisotropy, viscoelastic–viscoplastic behavior, crack propagation, creep lifetime and damage evolution across scales. For the project’s self-sensing composites (Use Case 5, structural health monitoring), it reaches into the electromechanical coupling behind piezoresistive response, tying ΔR/R signals to the onset of damage, and into the microstructural durability limits of low-microplastic coatings. The heavy pretraining and simulation runs use the consortium’s EuroHPC allocation on CINECA’s Leonardo.
Commercialization of Innovation: SafeLight and Phases
The methods Aerobase develops in GENMAT flow back into our own software. SafeLight, our failure and buckling library, gains sharper treatment of post-buckling anisotropy, stress tri-axiality, and multi-scale failure evolution. Phases, which predict microstructure and phase evolution, pick up the model’s microstructure tracking. Every validated material card and constitutive routine the project produces widens what these tools can simulate for automotive and aerospace customers and shortens the path from a candidate chemistry to a load-bearing design.
← Closing the loop: an AI target becomes a made, tested, & recyclable part.
Funding
GENMAT is funded by the European Union under Horizon Europe (GenAI4EU; call HORIZON-CL4-2025-01), Grant Agreement No. 101295332, and runs from 2026 to 2029. Details are on the CORDIS project page.
The Consortium
The GENMAT consortium brings together 14 partners from 7 countries, each contributing its expertise to the project. Politecnico di Torino (Italy) coordinates the work, together with partners Brunel University of London, Brunel Composites Centre (BCC), TWI, Stockholm University, Aerobase Innovations AB, Entelea, Talos Analytics, Cubic Snail, IRES – Innovation in Research and Engineering Solutions, Bar-Ilan University, NVIDIA (Mellanox Technologies), Technovative Solutions, CINECA, and Digital Security Lab Ukraine.
Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the granting authority. Neither the European Union nor the granting authority can be held responsible for them.
Here are the 3 sister projects: EarthGenerator, AgriScienceFM, EVELYN SimuLingua
Connect 🔗contribute.
Are you interested in learning about GENMAT and participating in its various phases? Connect with us on LinkedIn and share your thoughts.