Periodic Reporting for period 2 - AI-FIE (Artificial Intelligence for optimisation of Fuel Injection Equipment suitable for carbon-neutral synthetic fuels)
Période du rapport: 2024-01-01 au 2024-12-31
Decarbonizing the transport sector is vital for addressing climate change. While electric vehicles contribute to reducing Green House Gas (GHG) emissions, they are not entirely zero-emission due to the embedded emissions in battery production and the use of non-renewable electricity. Hydrogen-derived CO2-neutral synthetic fuels (e-fuels) present a promising alternative, as they can be used in existing internal combustion engines (ICE) and jet engines, avoiding the need for new infrastructure. This project supports the European Green Deal's goal of achieving net-zero GHG emissions by 2050.
Overall Objectives:
1.Modeling Fuel Properties: Develop a computational framework using PC-SAFT Equations of State (EoS) to predict the physical and transport properties of various conventional and e-fuels under different pressures and temperatures. This includes evaluating the performance of fuels like oxymethylene dimethyl ethers (OME3–4), gasoline, diesel, aviation fuel, and alcohol blends.
2.Creating the Specialized CFD Tool: Develop a specialized CFD code to generate training data for the AI, validated against experimental data, to accurately simulate complex flow dynamics within fuel injection systems.
3.AI Algorithm Development: Create a deep learning AI algorithm using deep autoencoders and neural networks to predict in-nozzle flow and spray characteristics of vaporizing liquid fuels. This algorithm aims to accelerate the simulation process by 3-4 orders of magnitude faster than current experimental and CFD methods, considering e-fuel composition, fuel injection equipment (FIE) design, and varying P-T conditions in combustion systems.
4.Application to E-Fuels: Apply the AI-driven methods to design and optimize new synthetic e-fuels, leveraging the specialized CFD code to simulate a broad range of scenarios and capture intricate behaviors of e-fuels in injection systems.
Conclusions of the Action: The project successfully developed a predictive thermodynamic model and a specialized CFD code, significantly reducing the need for physical experimentation. These tools were validated against experimental data, demonstrating high fidelity. However, certain aspects, such as integrating a neural network to connect the latent space of the encoder-decoder with the fuel composition, remain unfinished. The overall progress supports the rapid development and optimization of e-fuels, contributing to the reduction of GHG emissions and advancing the EU's decarbonization goals.
Overall Objectives:
1.Modeling Fuel Properties: Develop a computational framework using PC-SAFT Equations of State (EoS) to predict the physical and transport properties of various conventional and e-fuels under different pressures and temperatures. This includes evaluating the performance of fuels like oxymethylene dimethyl ethers (OME3–4), gasoline, diesel, aviation fuel, and alcohol blends.
2.Creating the Specialized CFD Tool: Develop a specialized CFD code to generate training data for the AI, validated against experimental data, to accurately simulate complex flow dynamics within fuel injection systems.
3.AI Algorithm Development: Create a deep learning AI algorithm using deep autoencoders and neural networks to predict in-nozzle flow and spray characteristics of vaporizing liquid fuels. This algorithm aims to accelerate the simulation process by 3-4 orders of magnitude faster than current experimental and CFD methods, considering e-fuel composition, fuel injection equipment (FIE) design, and varying P-T conditions in combustion systems.
4.Application to E-Fuels: Apply the AI-driven methods to design and optimize new synthetic e-fuels, leveraging the specialized CFD code to simulate a broad range of scenarios and capture intricate behaviors of e-fuels in injection systems.
Conclusions of the Action: The project successfully developed a predictive thermodynamic model and a specialized CFD code, significantly reducing the need for physical experimentation. These tools were validated against experimental data, demonstrating high fidelity. However, certain aspects, such as integrating a neural network to connect the latent space of the encoder-decoder with the fuel composition, remain unfinished. The overall progress supports the rapid development and optimization of e-fuels, contributing to the reduction of GHG emissions and advancing the EU's decarbonization goals.
Work Performed:
• Thermodynamic Model: A predictive thermodynamic model was developed to calculate the properties of various conventional and e-fuels. This model, grounded in PC-SAFT and VLE calculations, was validated against a broad set of experimental and computational data.
• CFD Code Development: An innovative CFD code was created to generate data for AI training. This code integrates the thermodynamic model and accurately simulates multiphase flows within fuel injection systems.
• Development of AI Algorithm: The goal was to develop a deep learning AI algorithm to streamline the development of e-fuels and fuel injection systems by predicting nozzle flow dynamics and spray patterns faster than current methods. Work was done in this direction
by training an encoder-decoder with simulations of supercritical injections of multiple fuel mixtures.
Main Results Achieved:
• Predictive Thermodynamic Model: Successfully developed and validated a model that accurately predicts the properties of various fuel types.
• CFD Code: Developed and validated a CFD code that reduces the need for physical experimentation by accurately replicating complex in-nozzle flow phenomena and spray dynamics. Presented in a symposium and under review for journal publication.
• AI Integration: Made progress in the development of the encoder-decoder models for efficient simulation of fuel injectors.
Overview of the Results for the Final Period:
• Thermodynamic Model and CFD Code: The model and code continued to provide a robust foundation for AI training and simulation, leading to significant advancements in understanding and predicting fuel properties and behaviors.
• AI Framework: Progress in the AI framework included the development of encoder-decoder models that compress high-dimensional flowfield data into a lower-dimensional latent space, although further improvements are needed to preserve finer details and maintain
intensity consistency.
• Publications and Dissemination: Key findings were published in scientific journals and presented at international conferences, ensuring broad dissemination of the project's results.
Exploitation and Dissemination of Results:
• Publications: Research findings were published in peer-reviewed journals and conference proceedings, with all publications acknowledging EU funding support.
• Conferences and Outreach: Results were presented at major conferences, including ILASS Europe 2023 and the International Symposium on Cavitation (CAV2024), facilitating knowledge exchange and engagement with the scientific community.
• Project Website: Activities and results were communicated through a project website and advertised via the researcher's LinkedIn profile, ensuring visibility and adherence to EU guidelines on open access.
• Thermodynamic Model: A predictive thermodynamic model was developed to calculate the properties of various conventional and e-fuels. This model, grounded in PC-SAFT and VLE calculations, was validated against a broad set of experimental and computational data.
• CFD Code Development: An innovative CFD code was created to generate data for AI training. This code integrates the thermodynamic model and accurately simulates multiphase flows within fuel injection systems.
• Development of AI Algorithm: The goal was to develop a deep learning AI algorithm to streamline the development of e-fuels and fuel injection systems by predicting nozzle flow dynamics and spray patterns faster than current methods. Work was done in this direction
by training an encoder-decoder with simulations of supercritical injections of multiple fuel mixtures.
Main Results Achieved:
• Predictive Thermodynamic Model: Successfully developed and validated a model that accurately predicts the properties of various fuel types.
• CFD Code: Developed and validated a CFD code that reduces the need for physical experimentation by accurately replicating complex in-nozzle flow phenomena and spray dynamics. Presented in a symposium and under review for journal publication.
• AI Integration: Made progress in the development of the encoder-decoder models for efficient simulation of fuel injectors.
Overview of the Results for the Final Period:
• Thermodynamic Model and CFD Code: The model and code continued to provide a robust foundation for AI training and simulation, leading to significant advancements in understanding and predicting fuel properties and behaviors.
• AI Framework: Progress in the AI framework included the development of encoder-decoder models that compress high-dimensional flowfield data into a lower-dimensional latent space, although further improvements are needed to preserve finer details and maintain
intensity consistency.
• Publications and Dissemination: Key findings were published in scientific journals and presented at international conferences, ensuring broad dissemination of the project's results.
Exploitation and Dissemination of Results:
• Publications: Research findings were published in peer-reviewed journals and conference proceedings, with all publications acknowledging EU funding support.
• Conferences and Outreach: Results were presented at major conferences, including ILASS Europe 2023 and the International Symposium on Cavitation (CAV2024), facilitating knowledge exchange and engagement with the scientific community.
• Project Website: Activities and results were communicated through a project website and advertised via the researcher's LinkedIn profile, ensuring visibility and adherence to EU guidelines on open access.
Progress Beyond the State of the Art:
• Thermodynamic Model: The development of a predictive thermodynamic model based on PC-SAFT and VLE calculations for calculating properties of various conventional and e-fuels. This model has been validated against a broad set of experimental and computational
data, providing a reliable tool for fuel property prediction. The model generates essential data for AI training, enhancing the AI framework's predictive capabilities.
• CFD Code: An innovative CFD code was created that integrates the thermodynamic model and accurately simulates multiphase flows within fuel injection systems. This code has shown high fidelity in replicating complex in-nozzle flow phenomena and spray dynamics. By
generating comprehensive data for AI training, the CFD code supports the development of more accurate and efficient AI algorithms.
Potential Impacts:
• Socio-Economic Impact: The project's outcomes will significantly reduce the need for physical experimentation in fuel development, thereby conserving resources and reducing environmental impact. This efficiency can lead to cost savings in fuel research and
development.
• Wider Societal Implications: By accelerating the development of e-fuels and improving fuel injection systems, the project supports the transition to carbon-neutral fuels, contributing to global efforts in combating climate change. The dissemination of findings through
publications and conferences ensures broad accessibility and encourages further research and innovation in the field.
• Thermodynamic Model: The development of a predictive thermodynamic model based on PC-SAFT and VLE calculations for calculating properties of various conventional and e-fuels. This model has been validated against a broad set of experimental and computational
data, providing a reliable tool for fuel property prediction. The model generates essential data for AI training, enhancing the AI framework's predictive capabilities.
• CFD Code: An innovative CFD code was created that integrates the thermodynamic model and accurately simulates multiphase flows within fuel injection systems. This code has shown high fidelity in replicating complex in-nozzle flow phenomena and spray dynamics. By
generating comprehensive data for AI training, the CFD code supports the development of more accurate and efficient AI algorithms.
Potential Impacts:
• Socio-Economic Impact: The project's outcomes will significantly reduce the need for physical experimentation in fuel development, thereby conserving resources and reducing environmental impact. This efficiency can lead to cost savings in fuel research and
development.
• Wider Societal Implications: By accelerating the development of e-fuels and improving fuel injection systems, the project supports the transition to carbon-neutral fuels, contributing to global efforts in combating climate change. The dissemination of findings through
publications and conferences ensures broad accessibility and encourages further research and innovation in the field.