Project
Large-Eddy Simulations of flash-boiling ammonia spray in the context of automotive engines
Ammonia shows promise as a fuel for reducing carbon emissions in transportation and stationary combustion systems, such as gas turbines and furnaces. While efficient combustion of gaseous ammonia has been proven, it remains more difficult than burning traditional hydrocarbons. Typically, ammonia is stored as a liquid under pressure, requiring a vaporization step before combustion. This extra step often limits operational flexibility. Direct injection of liquid ammonia offers a promising alternative, bypassing the need for external vaporization. However, it adds complexity related to fuel evaporation and mixture formation processes. Under normal operating conditions, ammonia exists in a gaseous state, causing flash-boiling during injection from a high-pressure reservoir. Flash-boiling greatly improves atomization and aids mixture formation, making it essential for efficient ammonia spray combustion. Recent experiments have shown that nozzle geometry significantly affects flash-boiling atomization. However, the direct connection between internal nozzle flow features and the resulting spray structure remains unclear. Many computational efforts also neglect detailed nozzle flow modeling or rely on empirical spray data. This project develops a two-step modeling framework applied to three nozzle designs with different orifice lengths to address these gaps. The framework includes both internal nozzle flow and external spray modeling. Specifically, the external spray, simulated with an Euler-Lagrange Large-Eddy Simulation approach, requires high-performance computing because resolving transient flow and droplet dynamics demands fine spatial detail.
Project Details
Project term
April 17, 2024–June 29, 2025
Affiliations
RWTH Aachen University
Institute
Institute for Combustion Technology (ITV)
Principal Investigator
Methods
The initial step in the two-stage simulation framework involves modeling the internal nozzle flow with an extended Eulerian two-fluid model. This RANS-type model assumes mechanical equilibrium between the liquid and vapor phases. To accurately represent heat transfer-driven phase changes, thermal nonequilibrium is maintained, with the vapor assumed to be saturated. Simulations were carried out using Ansys Fluent 2024/R1, where the continuity and momentum equations are handled through the mixture model. The temperature equation was implemented via a user-defined scalar, and energy and mass source terms were added through user-defined functions. A steady coupled solver was used, and spatial discretization employed the QUICK scheme. The setup involved a two-dimensional axisymmetric domain with a boundary layer-resolving mesh. Boundary conditions included pressure inlet and outlet, with the injection temperature set as the inlet temperature.
The external spray was simulated using an Euler-Lagrange Large Eddy Simulation approach that accounts for both external and internal vaporization. The internal vaporization model considers vapor bubble growth within each droplet. The Lagrangian spray model was integrated into the in-house code CIAO, a finite difference Large Eddy Simulation software. All terms were discretized with second-order central differences, except for the convective term in the mixture fraction transport equation, which used a fifth-order WENO scheme. The ODE system of the Lagrangian spray model was integrated with a second-order Runge-Kutta method, with adaptive time stepping. The setup included a cubic domain spanning 200 orifice diameters in each direction, discretized with a uniform mesh spacing equal to the orifice diameter.
Results
The nozzle flow simulations show velocity increases significantly at the entrance and near the outlet. Liquid phase change starts upstream of the exit, and higher velocities at the entrance suggest a higher mass flow rate. Longer orifices, despite similar pressure drops, are expected to have more friction losses, confirmed by mass flow calculations. Extra acceleration near the outlet is likely due to phase expansion. As orifice length grows, the volume fraction decreases, supported by outlet data. Radial distribution shows smaller liquid fractions near the wall, linked to longer residence times. Liquid temperature trends follow volume fraction changes, with minor cooling effects compared to injection temperatures. The external spray simulations are compared with experimental shadowgraphy images. The experiments show typical flash atomization sprays with a strong spray opening near the nozzle, which diminishes further downstream. The simulations fail to accurately predict the spray appearance and instead display a spray that continuously opens. Spray angles were measured and compared. The experiments show an increasing trend with longer orifice length. While the simulations capture this trend, their values are underestimated. A parameter study indicated that both the decrease in mass flow rate and the increased pre-vaporization within the nozzle influence the growing spray angle.
Discussion
The two-step simulation approach used in this project shows promising results for linking the nozzle’s internal flow with the external spray in a flash-boiling ammonia spray. However, the framework needs improvements to enhance prediction accuracy. Bubble nucleation was modeled simply in the nozzle’s internal flow simulations and did not account for heterogeneous nucleation at the orifice wall. Future simulations should incorporate a wall nucleation model, which is likely an important mechanism for the flash-boiling flow inside a nozzle. Although the external spray captured the increasing spray angle, the overall spray appearance was incorrect. Since the straightening of the spray is closely related to droplet size, it is concluded that the spray breakup model predicts droplets that are too large. The current breakup model assumes the number of child droplets is proportional to the number of bubbles, which is a significant simplification. This should be refined with more physics-based approaches that consider different breakup regimes and the resulting droplet sizes. Using an improved breakup model is identified as a task for future work.
Additional Project Information
DFG classification: 404-03 Fluid Mechanics
Software: CIAO, ANSYS
Cluster: CLAIX