Description du projet
L’IA pour les carburants synthétiques neutres en carbone dérivés de l’hydrogène
Les politiques de l’UE prévoient la transition du secteur des transports des combustibles fossiles vers les énergies propres. Les carburants synthétiques neutres en carbone dérivés de l’hydrogène et produits à partir de sources d’énergie renouvelables présentent une empreinte CO2 sur le cycle de vie inférieure à celle des véhicules électriques et sont adaptés aux moteurs à combustion. Cependant, les méthodes expérimentales existantes de pulvérisation de carburant empêchent une large utilisation des e-carburants. Le projet AI-FIE, financé par l’UE, est une bourse du programme Actions Marie Skłodowska-Curie visant à développer un algorithme d’apprentissage profond et d’IA fondé sur les données pour prédire la structure de la pulvérisation résolue dans l’espace et dans le temps et les paramètres critiques pour la conception du moteur. Le projet innovant basera l’apprentissage sur la base de données expérimentale la plus vaste disponible publiquement pour les pulvérisations de carburant de l’ECN (Engine Combustion Network).
Objectif
Current EU policies mandate the gradual disengagement of the transport sector from fossil fuels. In order for such a transition to become a reality, hydrogen-derived carbon-neutral synthetic fuels produced using renewable energy sources (e-fuels), have overall less life-cycle CO2 footprint than their counterpart electric vehicles while they are suitable for use over the wide range of combustion engines. However, today’s fuel spray experimental methods are compromised by the long time needed for the characterisation of the effect of new fuel molecules; similarly, relevant predictive models that can address in detail the effect of the wide range of fuel chemical composition at time scales relevant to industry are not available. The main objective of the proposed MSCA fellowship is the development of a data-driven deep learning (DL) Artificial Intelligence (AI) algorithm able to predict the spatially and temporally resolved spray structure, as well as critical air / fuel mixture parameters for engine design. Training of the AI model will be based on the largest publicly available experimental database for fuel sprays of the Engine Combustion Network; this covers a wide range of injector configurations, air thermodynamic conditions and liquid fuels. The training matrix of the AI algorithm will be complemented by relevant computational fluid dynamics simulations for operating conditions and fuel composition for which experimentation is not possible. For this purpose, a state-of-the-art CFD model of the compressible Navier-Stokes and energy conservation equations employing elaborate real-fuel thermodynamic closures based on the PC-SAFT equation of state will be employed. The project innovative nature spans across diverse research aspects with emphasis on renewable alternatives of Diesel and gasoline. As such, it is expected to assist EU energy, marine, aviation and automotive industries to meet the goals imposed regarding the utilisation of renewable fuels.
Champ scientifique
- engineering and technologyenvironmental engineeringenergy and fuelsliquid fuels
- engineering and technologyenvironmental engineeringenergy and fuelsrenewable energy
- engineering and technologymechanical engineeringvehicle engineeringautomotive engineering
- engineering and technologyenvironmental engineeringenergy and fuelssynthetic fuels
- natural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learning
Programme(s)
Régime de financement
MSCA-IF-GF - Global FellowshipsCoordinateur
EC1V 0HB London
Royaume-Uni