Cryogenic jets atomisation
High-fidelity simulation of flashing, cavitating cryogenic jets and spray formation, enabling accurate prediction of dispersion, phase change, and safety-critical behaviour in energy systems
Technical capabilities in MPflow
MPflow provides advanced capabilities for modelling cryogenic jet atomisation under extreme thermodynamic conditions:
- Dedicated solvers for compressible, multiphase, and phase-change flows
- Modelling of flash boiling and cavitating jets under rapid depressurisation
- Captures metastable thermodynamic states and non-equilibrium phase change
- Eulerian–Lagrangian framework for dense-to-dilute spray transition
- Advanced treatment of:bubble nucleation and growth, internal and external flashing and evaporation and heat transfer in cryogenic conditions
- Fully coupled RANS and LES turbulence modelling
- Validated across cryogenic fluids, including LNG and hydrogen


Our CFD approach
- Simulation of pressurised cryogenic jets expanding into low-pressure environment
- Resolution of internal nozzle flow and onset of flashing
- Modelling of bubble nucleation within the liquid core and its impact on breakup
- Captures multiple flow regimes: bubbly, slug, and annular structures
- LES-based prediction of: Transient jet instability and breakup, spray dispersion and vapour cloud formation
- Accounts for:
- Superheat and subcooling effect
- Ambient conditions and thermodynamic paths (isothermal / isobaric)
- Geometry and nozzle L/D ratio
Physics-based ML acceleration
- ML models trained on high-fidelity CFD and experimental cryogenic dataset
- Surrogate prediction of: flashing onset and intensity, spray structure and dispersion, vapour cloud evolution
- Embedded ML acceleration for:
- Phase-change source terms
- Turbulence modelling (RANS/LES)
- Enables rapid evaluation across: Different cryogenic fluids (LNG, LH₂, etc.), Pressure and temperature conditions, Leak sizes and geometries
- Preserves thermodynamic consistency and physical constraints
Why ML-CFD matters for Cryogenic Jets Atomisation
- Simulation time reduced from ~8 hours to seconds with deep neural network deployment for flashing cryogenic releases
- 3–5× acceleration with on-the-fly ML-enhanced RANS & LES turbulence modelling
- High accuracy (<5–8% deviation) in spray structure, vapour cloud evolution, and dispersion behaviour
- Instant evaluation of safety-critical scenarios, including accidental releases and leak conditions
- Scalable across fluids and operating conditions, including LNG and hydrogen systems
- Enables rapid risk assessment and integration into digital twins


