Periodic Reporting for period 1 - MFLOPS (Multiphase Flow Optimisation Strategies with Industrial Applications)
Okres sprawozdawczy: 2023-01-01 do 2024-12-31
MFLOPS develops optimisation strategies to address real-world industrial applications aligning with Horizon Europe work programme objectives for Climate, Energy and Mobility and addressing the United Nations goals for ‘Affordable and green energy for all’.
The Doctoral Candidates develop accurate simulation methods and tools for multiphase flows, couple them with efficient optimisation methods and demonstrate their capabilities for the effective design of hydraulic turbines (Kaplan, Pelton), marine propellers and marine vessel hydrodynamics, fuel cells, high-pressure injectors for E-fuels, cooling systems for compression-ignition engines and immersed Battery Thermal Management System (BTMS).
The research topics are divided into three Work Packages:
WP 1: Physical models and optimisation methods for cavitation, erosion and noise, aiming to assist in the design of hydraulic turbines (Kaplan and Pelton) and (composite material marine propellers.
WP 2: Physical models and optimisation methods for immiscible liquid flows and apply them to the design of low-temperature PEM gas diffusion layers and flow passages of the graphite plates utilised in fuel cells and power/drag-optimised merchant ships and their propeller integration.
WP 3: Physical models and optimisation methods for flows involving boiling or supercritical phase-change and heat transfer and aiming to the design of high-pressure fuel injectors utilising e-fuels, submerged BTMSs for EVs and hydrogen electrolysers.
The Doctoral Candidates develop accurate simulation methods and tools for multiphase flows, couple them with efficient optimisation methods and demonstrate their capabilities for the effective design of hydraulic turbines (Kaplan, Pelton), marine propellers and marine vessel hydrodynamics, fuel cells, high-pressure injectors for E-fuels, cooling systems for compression-ignition engines and immersed Battery Thermal Management System (BTMS).
The research topics are divided into three Work Packages:
WP 1: Physical models and optimisation methods for cavitation, erosion and noise, aiming to assist in the design of hydraulic turbines (Kaplan and Pelton) and (composite material marine propellers.
WP 2: Physical models and optimisation methods for immiscible liquid flows and apply them to the design of low-temperature PEM gas diffusion layers and flow passages of the graphite plates utilised in fuel cells and power/drag-optimised merchant ships and their propeller integration.
WP 3: Physical models and optimisation methods for flows involving boiling or supercritical phase-change and heat transfer and aiming to the design of high-pressure fuel injectors utilising e-fuels, submerged BTMSs for EVs and hydrogen electrolysers.
WP1
DC1 developed and validated a continuous adjoint method for isothermal cavitating flows using the GPU-enabled PUMA solver. The model accurately predicts vapor formation in hydrofoils and hydraulic turbines. Sensitivity derivatives obtained with the adjoint method matched finite difference results, confirming correctness. Optimization studies successfully maximized lift and minimized cavitation.
DC2 extended PUMA framework to simulate two-phase cavitating flows by implementing a volume fraction equation and three cavitation models, calibrated with experimental data. A GPU-parallelized erosion estimation algorithm was also developed. The solver captures complex cavitation phenomena like sigma break curves and rope cavitation, showing potential for design-phase predictions.
DC3 initiated development of a coupled RANS-FEM FSI solver for cavitating composite propellers, with early LES results obtained. DC11 introduced an innovative incremental SVD-based model reduction for SPH simulations, validated on 2D/3D cases with minimal error and overhead, marking a first in SPH and enabling efficient adjoint-based optimization for Pelton turbines.
DC11 implemented an incremental SVD-based model reduction technique in the SPH-ALE solver ASPHODEL. The method was validated on 2D/3D test cases, showing negligible error and minimal overhead, representing a novel application within SPH frameworks.
WP2
DC5 developed a primal/adjoint diffuse-interface model within the FreSCo+ solver for optimizing free-surface flows around vessels, enabling shape optimization with potential for unsteady studies.
DC6 worked on ship hull optimization, integrating adjoint methods with machine learning (CVAE and neural networks) to predict propulsion effects and reduce computational costs; 42% of training data has been generated and preliminary tests conducted.
DC9 implemented a numerical model for water transport in PEM fuel cells, incorporating a dynamic contact angle model for better simulation of immiscible fluid behavior, with optimization work planned next.
DC12 focused on boiling bubble detachment; after a literature review, work is underway to validate a custom OpenFOAM solver against experiments.
DC13 is optimizing electrolyser designs, validating the VOF method for hydrogen bubble dynamics and analyzing forces influencing detachment; conceptual designs for improved channel geometries have been developed.
WP3
DC8 has developed and validated a compressible adjoint solver for cavitating flows based on real-fluid equations of state, achieving stable numerical convergence on simplified injector geometries. The solver extends OpenFOAM’s capabilities, representing a significant advancement as it enables optimization workflows for multiphase compressible flows at all Mach numbers, a first in the literature.
DC10 successfully derived and implemented adjoint equations for multiphase flows with heat transfer and phase change, solving shape optimization problems for EV battery thermal management systems (BTMS). This work extends current methods by including conjugate heat transfer (CHT) and phase change in multiphase adjoint optimization—an innovative contribution not previously available in literature.
DC14 focused on simulating dual-fuel e-fuel injection. He developed a multicomponent solver that treats gaseous and liquid phases distinctly, includes compressibility and velocity equilibrium, and incorporates real equations of state. The solver has been validated against experimental data, achieving the first major project milestone: incorporation of dual e-fuel mixtures into nozzle design.
DC4 advanced the prediction of physical properties for gaseous-liquid e-fuel mixtures. His work includes developing machine learning workflows for property modeling, creating software interfaces (MATLAB-ASPEN PLUS-EXCEL), and preparing optimized property tables for CFD simulations.
DC7 investigated bubble dynamics critical to electrolyser performance. He conducted modeling work on ultrasound-induced bubble detachment, using OpenFOAM and FORESTFV, and simulated interactions between bubbles under acoustic forces.
DC1 developed and validated a continuous adjoint method for isothermal cavitating flows using the GPU-enabled PUMA solver. The model accurately predicts vapor formation in hydrofoils and hydraulic turbines. Sensitivity derivatives obtained with the adjoint method matched finite difference results, confirming correctness. Optimization studies successfully maximized lift and minimized cavitation.
DC2 extended PUMA framework to simulate two-phase cavitating flows by implementing a volume fraction equation and three cavitation models, calibrated with experimental data. A GPU-parallelized erosion estimation algorithm was also developed. The solver captures complex cavitation phenomena like sigma break curves and rope cavitation, showing potential for design-phase predictions.
DC3 initiated development of a coupled RANS-FEM FSI solver for cavitating composite propellers, with early LES results obtained. DC11 introduced an innovative incremental SVD-based model reduction for SPH simulations, validated on 2D/3D cases with minimal error and overhead, marking a first in SPH and enabling efficient adjoint-based optimization for Pelton turbines.
DC11 implemented an incremental SVD-based model reduction technique in the SPH-ALE solver ASPHODEL. The method was validated on 2D/3D test cases, showing negligible error and minimal overhead, representing a novel application within SPH frameworks.
WP2
DC5 developed a primal/adjoint diffuse-interface model within the FreSCo+ solver for optimizing free-surface flows around vessels, enabling shape optimization with potential for unsteady studies.
DC6 worked on ship hull optimization, integrating adjoint methods with machine learning (CVAE and neural networks) to predict propulsion effects and reduce computational costs; 42% of training data has been generated and preliminary tests conducted.
DC9 implemented a numerical model for water transport in PEM fuel cells, incorporating a dynamic contact angle model for better simulation of immiscible fluid behavior, with optimization work planned next.
DC12 focused on boiling bubble detachment; after a literature review, work is underway to validate a custom OpenFOAM solver against experiments.
DC13 is optimizing electrolyser designs, validating the VOF method for hydrogen bubble dynamics and analyzing forces influencing detachment; conceptual designs for improved channel geometries have been developed.
WP3
DC8 has developed and validated a compressible adjoint solver for cavitating flows based on real-fluid equations of state, achieving stable numerical convergence on simplified injector geometries. The solver extends OpenFOAM’s capabilities, representing a significant advancement as it enables optimization workflows for multiphase compressible flows at all Mach numbers, a first in the literature.
DC10 successfully derived and implemented adjoint equations for multiphase flows with heat transfer and phase change, solving shape optimization problems for EV battery thermal management systems (BTMS). This work extends current methods by including conjugate heat transfer (CHT) and phase change in multiphase adjoint optimization—an innovative contribution not previously available in literature.
DC14 focused on simulating dual-fuel e-fuel injection. He developed a multicomponent solver that treats gaseous and liquid phases distinctly, includes compressibility and velocity equilibrium, and incorporates real equations of state. The solver has been validated against experimental data, achieving the first major project milestone: incorporation of dual e-fuel mixtures into nozzle design.
DC4 advanced the prediction of physical properties for gaseous-liquid e-fuel mixtures. His work includes developing machine learning workflows for property modeling, creating software interfaces (MATLAB-ASPEN PLUS-EXCEL), and preparing optimized property tables for CFD simulations.
DC7 investigated bubble dynamics critical to electrolyser performance. He conducted modeling work on ultrasound-induced bubble detachment, using OpenFOAM and FORESTFV, and simulated interactions between bubbles under acoustic forces.
The findings beyond the state of the art in this milestone include key advancements:
1. Cavitating Flows and Compressible Adjoints: A continuous adjoint method for fully compressible, two-phase cavitating flows was developed in OpenFOAM, improving optimization for cavitation problems across all Mach numbers.
2. Marine Free-Surface Flows: A diffuse interface primal/adjoint model was added to the FreSCo+ solver, enabling adjoint computations for shape optimization in ship design, such as resistance reduction.
3. Ship Hull Optimization: A hybrid ML and adjoint-based framework using neural networks and Conditional Variational Autoencoders (CVAE) was introduced, reducing computational costs and enabling optimization under new conditions.
4. PEM Fuel Cells: A dynamic contact angle model using an algebraic Volume-of-Fluid (VoF) method improved water dynamics simulations for optimized fuel cell water management.
5. Boiling Bubble Dynamics and Electrolyser Optimization: New solvers for bubble coalescence and detachment advanced hydrogen production and boiling heat transfer simulations.
6. Fuel Mixture Modeling and Injector Design: A tabulated approach incorporating real equations of state (EoS) optimized dual-fuel injector design and was validated for real-world applications.
7. Fuel Properties for CFD Simulations: A thermodynamic closure model was created to optimize fuel properties for high-compressibility flows, enhancing CFD simulation accuracy.
8. Bubble Detachment in Electrolysers: New techniques for predicting hydrogen bubble detachment improved flow path optimization in bipolar plates for better hydrogen dispersion.
1. Cavitating Flows and Compressible Adjoints: A continuous adjoint method for fully compressible, two-phase cavitating flows was developed in OpenFOAM, improving optimization for cavitation problems across all Mach numbers.
2. Marine Free-Surface Flows: A diffuse interface primal/adjoint model was added to the FreSCo+ solver, enabling adjoint computations for shape optimization in ship design, such as resistance reduction.
3. Ship Hull Optimization: A hybrid ML and adjoint-based framework using neural networks and Conditional Variational Autoencoders (CVAE) was introduced, reducing computational costs and enabling optimization under new conditions.
4. PEM Fuel Cells: A dynamic contact angle model using an algebraic Volume-of-Fluid (VoF) method improved water dynamics simulations for optimized fuel cell water management.
5. Boiling Bubble Dynamics and Electrolyser Optimization: New solvers for bubble coalescence and detachment advanced hydrogen production and boiling heat transfer simulations.
6. Fuel Mixture Modeling and Injector Design: A tabulated approach incorporating real equations of state (EoS) optimized dual-fuel injector design and was validated for real-world applications.
7. Fuel Properties for CFD Simulations: A thermodynamic closure model was created to optimize fuel properties for high-compressibility flows, enhancing CFD simulation accuracy.
8. Bubble Detachment in Electrolysers: New techniques for predicting hydrogen bubble detachment improved flow path optimization in bipolar plates for better hydrogen dispersion.