
D3T4H2S
A data-driven digital twin for improved hydrogen storage vessels towards challenges for the energy transition
COFUND-LEAP-RE-D3T4H2S

Europe Horizon – LEAP-RE program:COFUND-LEAP-RE-D3T4H2S
Financing contract for project execution no. 11/2024.
Total Funding amount: 425 000 Euro. Implementation period: 1 September 2023 - 14 February 2026.
UDJG Funding amount: 55 000 Euro.
- UGAL Team:                                                                                                     International partners:
-
1. Viorel MINZU - Leader UGAL team
cv
                      - S VERTICAL: Medium Entreprise, France (COORDONATOR)
-
2. Eugen RUSU - Senior Scientist
cv
                              - University of South Africa, UNISA
-
3. Ana CHIROSCA - Postdoctoral researcher
cv
        - International University of Rabat, Moroco
- 4. Magduța CHIVU – Financiar Responsible                                                  - École Nationale Supérieure de Techniques Avancées Bretagne, France                                                                                                                                 - University of Hassan II Casablanca (UH2C)
- UGAL specific research objectives:
- - Development of Machine Learning models for composite materials used in the manufacture of hydrogen tanks that also allow the analysis of defects in the structure. - Using AI techniques for the optimal design of composite hydrogen tanks.
Project Objectives
The green hydrogen market will likely grow significantly over the next few years because there is more demand for clean energy sources, and the government is doing more to build a sustainable environment. Hydrogen could be a vital part of a sustainable energy system in the future because it can help get carbon out of the transportation sector. Material science and artificial intelligence (AI) discoveries lead to much new science and technology, such as green hydrogen technologies. These technologies try to meet the challenge of reducing carbon dioxide emissions to help with climate change and the energy crisis. So, it is clear that AI is one way to make the environment more sustainable. The project tackles a global challenge in reaching affordable and clean energy targets by addressing the design of a small-scale proof-of-concept storage vessel in collaboration with SVERTICAL (an industrial partner). The complexity of the topic is strongly related to many aspects involving material sciences, structural thermomechanics, and structure design. The holistic treatment of such aspects exceeds the possibilities of a single research project. Thus, attention will mainly focus on the elaboration of a hybrid carbon fiber reinforced polyamide 12 doped with carbon nanotubes for designing ultralight cryogenic composite vessels (ULCCVs) and on the multi-scale and multi-physical study of its long-term behavior at cryogenic temperatures, its permeability performances, and the issue of damage initiation and propagation. The applicants ambition for the mainstream is to develop ULCCV base materials that provide structural integrity and adequate microcrack resistance against harsh chemicals, thermal, and mechanical loads, as well as the development of an expert tool for better life-cycle management and accurate predictive maintenance, allowing it to reduce costs and maintain a competitive advantage for hydrogen applications. The latter makes sense in digitalization by developing digital and/or hybrid twins dedicated to real-time prediction and correction.
RESULTS OBTAINED IN PHASE 1 - 2024
In the project implementation plan, in the first phase (March - December 2024), we are committed to achieving the following results:
   1. Machine-learning (ML) models characterizing the thermo-mechanical behavior of composite materials In activities 1.2 and 1.3, machine-learning (ML) models characterizing the thermo-mechanical behavior of composite materials were developed and validated. These models are described in the study entitled: "Machine Learning models' construction for the load behavior of composite materials in the undamaged zone - Report of UGAL for the project Leap-Re D3T4H2S." [link] This study has been presented in online meetings and submitted to the consortium coordinator, Mourad Nachtane, from S VERTICAL company in order to make it accessible to all D3T4H2S project partners. The developed Machine Learning models have been realized and validated at computer level using MATLAB system. The above study calls upon the multiple programs contained in the archive and well annotated for use. These programs also confer a way of checking ML models. The summarized scientific report addressed to the UEFISCDI is presented in the document "SCIENTIFIC REPORT 2024 annual stage". [link]    2. Publication in a national/international journal/conference volumeProcesses 2024, 12, 991.
https://https://doi.org/10.3390/pr12050991;
Impact Factor 2.8; (WOS SCIE Q2); Citescore: 5.1 (Q2)
Special Issue: Industrial Process Operation State Sensing and Performance Optimization
Published: 13 May 2024
Energies 2024, 17, 5820;
https://https://doi.org/10.3390/en17235820;
Impact Factor 3 (WOS SCIE Q2); Citescore: 6.2 (Q2).
Section A5: Hydrogen Energy
Published: 21 November 2024
RESULTS OBTAINED IN PHASE 2 - 2025
In the project implementation plan, in the second stage (January - December 2025), we are committed to achieving the following results:
   1. Publication in a national/international journal/conference volumeProcesses 2025, 13, 602.
https://https://doi.org/10.3390/pr13030602;
Impact Factor 2.8; (WOS SCIE Q2); Citescore: 5.1 (Q2)
Special Issue: Application of Artificial Intelligence in Industrial Process Modelling and Optimization
Published: 20 February 2025
Abstract: This article addresses the complex behavior of composite laminates under varied layer orientations during tensile tests, focusing on carbon fiber and epoxy matrix composites. Data characterizing the mechanical load behavior is obtained using twelve composite laminates with different layer orientations and the DIGIMAT-VA software. First, this data was used to elaborate a complex comparative analysis of composite laminates from the perspective of materials science. Composite laminates belong to three classes: unidirectional, off-axis-oriented, and symmetrically balanced laminates, each having a specific behavior. From the perspective of designing a new material, a prediction model that is faster than the finite element analysis is needed to apply this comparative analysis's conclusions. As a novelty, this paper introduces several machine learning prediction models for composite laminates with 16 layers arranged in different orientations. The Regression Neural Network model performs best, effectively replacing expensive tensile test simulations and ensuring good statistics (RMSE=34.385, R2=1, MAE=19.829). The simulation time decreases from 34.5 sec (in the case of finite element) to 0.6 sec. The prediction model returns the stress-strain characteristic of the elastic zone given the new layer orientations. These models were implemented in the MATLAB system, and their running proved good models' generalization power and accuracy. Even specimens with randomly oriented layers were successfully tested.
YEAR 2024
- Minzu, V.; Arama, I.; Rusu, E., 2024, Machine Learning Algorithms that Emulate Controllers based on Particle Swarm Optimization - An Application to a Photobioreactor for Algae Growth Processes 2024, 12, 991. https://doi.org/10.3390/pr12050991; (Q2, IF=2.8);
- Chirosca, A.-M.; Rusu, E.; Minzu, V., 2024, Green Hydrogen—Production and Storage Methods: Current Status and Future Directions Energies 2024, 17, 5820. https://https://doi.org/10.3390/en17235820; (Q2, IF=3.0);
-
A. Works published in international journals
YEAR 2025
- Brayek, B.E.B.; Sayed, S.; Mînzu, V, 2025, Machine Learning Predictions for the Comparative Mechanical Analysis of Composite Laminates with Various Fibers Processes 2025, 13, 602. https://https://doi.org/10.3390/pr13030602; (Q2, IF=2.8);
-
A. Works published in international journals
D3T4H2S - Workshops
D3T4H2S - Certificates
- Viorel Minzu and Eugen Rusu, “Certificate of acceptance for the manuscript (processes-2999578) titled: Machine Learning Algorithms that Emulate Controllers based on Particle Swarm Optimization - An Application to a Photobioreactor for Algae Growth”
- Ana Maria Chirosca, Eugen Rusu and Viorel Minzu, “Certificate of publication for the article titled: Green Hydrogen—Production and Storage Methods: Current Status and Future Directions”
- Baha Eddine Ben Brayek, Sirine Saye, Viorel Minzu and Mostapha Tarfaoui, “Certificate of publication for the article titled: Machine Learning Predictions for the Comparative Mechanical Analysis of Composite Laminates with Various Fibers”