Strong multiphysics coupling
Mass, momentum, energy, species and additional transport equations are assembled and advanced in a unified implicit framework.
NeuralFlow multiphysics CFD
A highly validated CFD platform for difficult propulsion and aerothermal problems
NeuralFlow is LIKUA's fully implicit finite-volume solver for compressible, thermal, reacting and particle-laden flows. It is developed around accuracy, validation evidence and robust convergence for difficult engineering problems, while its modular numerical architecture supports rapid customization of physical models, boundary conditions, source terms and analysis workflows.
Validated numerical methods for difficult flow regimes, stiff multiphysics and large three-dimensional simulations.
High accuracyNumerical methods developed for engineering fidelity across difficult flow regimes.
Validation evidenceAnalytical, experimental and engineering reference cases documented across the solver.
Difficult coupled physicsStrongly coupled flow, thermal, reacting and dispersed-phase formulations.
Rapid customizationFlexible physics, boundary conditions, source models and workflow extensions for specialized applications.
What makes NeuralFlow different
Coupled physical models within the same solver framework
NeuralFlow is designed around equation-level coupling. Turbulence, species, combustion, dispersed phases, thermal solids and other source-driven models can participate directly in the implicit solution rather than being treated only through weak sequential updates.
Robust stiff-flow treatment
Pressure waves, shocks, recirculation, high heat release, strong source terms and particle interactions are handled with methods developed for demanding propulsion flows.
Engineering-focused convergence
Continuity acceleration, near-wall solution acceleration, multigrid and flexible linear solvers target practical convergence on difficult meshes and operating conditions.
Core flow and time integration
One formulation across flow regimes and time scales
The same core architecture supports steady and transient simulations from incompressible and low-Mach internal flow to shock-dominated high-speed aerodynamics.
Compressible and incompressible flow
- Incompressible solver capability for low-speed engineering applications
- Low-Mach, transonic, supersonic and hypersonic regimes
- Time-derivative and low-Mach preconditioning
- AUSM+up and HLLC flux formulations
- Second-order and bounded-central discretizations
Accurate implicit dual-time stepping
- Fully implicit physical-time advancement
- Inner pseudo-time convergence for each physical step
- Strongly coupled transient multiphysics
- Stable treatment of stiff and highly dynamic flows
Scale-resolving simulations
- Hybrid RANS and scale-resolving turbulence options
- Separated and highly unsteady flows
- Resolved pressure and coherent-flow structures
- Transient spectral and modal post-processing
Time-resolved solver evidence
The fully implicit transient formulation advances resolved vortical structures while preserving time-accurate pressure histories for frequency-domain analysis. The paired views show temporal and spectral evidence from the same workflow.
GERANS turbulence and aerothermal modelling
High-speed boundary layers and wall heating
NeuralFlow uses the GERANS turbulence framework, including its Realizable SST-kω formulation and scale-resolving extensions, for complex separated and compressible turbulent flows.
Its y⁺-independent wall treatment supports turbulent heat transfer, conjugate heat transfer and advanced thermal-wall models over changing near-wall resolution.
Aerothermal heatingSupersonic and hypersonic wall heat flux, shock-boundary-layer interaction and re-entry-relevant conditions.
y⁺-independent wall functionsConsistent near-wall momentum and thermal treatment over broad mesh-resolution ranges.
Conjugate heat transferStrongly coupled fluid-solid thermal loading for hot structures and cooling systems.
Thermal protection responseOne-dimensional charring, pyrolysis, blowing and material-regression modelling.
Eulerian dispersed phase
Particle transport with evolving interfacial area
The Eulerian particle solver is integrated within the strongly coupled framework for metallized propulsion products, dust-laden flows, granular systems and thermally interacting dispersed phases.
Interfacial-area evolution made visible
The transported interfacial-area field links local dispersed-phase morphology to gas-particle momentum and heat exchange. The solution resolves changing effective particle scale instead of assuming a fixed uniform diameter.
It shows breakup near the lip, coalescence in the turbulent recirculation region and accumulation driven by stagnation and secondary flow.
Conservative particle equations
- Particle mass, momentum and energy transport
- Gas-particle momentum and heat exchange
- Dilute and dense dispersed environments
- Gravity and application-specific source terms
Interfacial Area Transport Equation
- Transport of interfacial area density
- Evolution of effective particle or droplet scale
- Consistent coupling with phase exchange models
- Suitable for changing dispersed-phase morphology
Breakup and coalescence
- Particle and droplet breakup modelling
- Coalescence source terms
- Size-distribution evolution through IATE
- Applications in nozzles, plumes and gas generators
Coupled particles in a launch environment
The dispersed phase is transported with the surrounding compressible gas, including momentum and thermal exchange. The launch-plume result demonstrates the coupled response under a strongly expanding, high-temperature flow.
Linear and nonlinear solver technology
Scalable solution of large, fully coupled systems
The implicit formulation produces large sparse systems containing the interaction of flow and additional physics. NeuralFlow combines GPU-capable sparse linear algebra, Krylov methods, algebraic multigrid and geometric multigrid technologies to make these systems practical on engineering meshes.
Krylov and preconditioned iterative methodsFlexible linear-solver selection for different equation structures and operating regimes.
Algebraic multigrid and block preconditioningAMG, block-Jacobi and domain-decomposition approaches for coupled sparse systems.
Geometric multigrid and nonlinear FASV-, W- and F-cycle strategies with coarse-grid correction for nonlinear acceleration.
Full Multigrid initializationCoarse-to-fine nested iteration for rapid generation of physically useful initial flow fields.
Extensibility and engineering workflow
Built to adapt to specialized applications
NeuralFlow is not limited to fixed built-in models. Its architecture supports case-specific development, external workflow integration and user-defined extensions.
Python UDF support
- User-defined modelling and case-specific extensions
- Python-driven workflow automation
- Custom engineering logic without rebuilding the full solver
- Integration with analysis and optimization pipelines
Geometry and mesh support
- 2D, axisymmetric and fully 3D problems
- Tetrahedral, hexahedral, prism and mixed meshes
- True polyhedral unstructured control volumes
- CGNS, Fluent and HDF5-oriented workflows
Parallel and integrated execution
- Distributed-memory MPI parallelization
- Optimized data packing and inter-rank communication
- Restartable long-running HPC simulations
- Coupling with thermal, structural, optimization and trajectory tools
Advanced numerical capabilities
Robust methods for difficult coupled simulations
NeuralFlow combines specialized transport formulations, convergence acceleration and scalable linear algebra for problems that are difficult to solve reliably with conventional workflows.
Hyperbolically regularized particle solver
- Conservative Eulerian particle mass, momentum and energy transport
- Special hyperbolic regularization for robust particle-wave propagation
- Strong gas-particle momentum and heat-transfer coupling
- Interfacial Area Transport, breakup and coalescence
Species transport
- Conservative multi-species transport
- Molecular and turbulent diffusion
- Composition-dependent thermodynamic properties
- Strong coupling with flow, energy and reacting-source terms
GPU-supported linear solvers
- GPU-enabled sparse vector and matrix pathways
- CUDA-supported sparse linear algebra
- Krylov, block-preconditioned and multigrid solution options
- Parallel CPU and GPU execution paths for large systems
Continuity acceleration
- Accelerated propagation of pressure and mass-flow corrections
- Improved global conservation convergence
- Designed for chambers, ducts, nozzles and recirculating systems
- Compatible with the fully implicit coupled formulation
Boundary-layer acceleration
- Strong near-wall convergence acceleration
- Robust performance on highly stretched cells
- Designed for separated compressible boundary layers
- Effective for shock-boundary-layer interaction and aerothermal flows
Highly validated engineering solver
- Validated against analytical, experimental and engineering reference cases
- Applied to challenging propulsion, aerothermal and multiphysics problems
- Conservative residual and Jacobian architecture
- Direct control of numerical methods and convergence evidence
Application areas
Designed around difficult aerospace and industrial systems
NeuralFlow is especially valuable where aerodynamic performance, thermal loading, transient response and additional physics must be assessed together. Propulsion systems are shown with representative cases; broader flow and thermal domains are summarized below.
Propulsion systems
01
Solid rocket motors
Internal ballistics, combustion chambers, nozzles, pressure oscillations, particles and thermal loads.
Representative case: rocket-nozzle internal expansion, plume structure and surrounding-flow interaction.
02
Ramjets and ducted rockets
Supersonic intakes, combustors, fuel-rich gas generators, boron particles and aero-propulsive integration.
Representative case: boron-particle transport in a ducted-rocket fuel-gas-generator application.
Additional application domains
High-speed aerodynamics
Shock systems, separation, drag, aerothermal heating and high-Mach operating envelopes.
Nozzles and plumes
Internal expansion, jet structure, launch interaction and condensed-phase exhaust products.
Thermal protection and cooling
Conjugate heat transfer, charring materials, hot walls and cooling-flow design.
Rotating and industrial systems
Multiple Reference Frames, turbomachinery, fans and coupled process-flow systems.
Particle-laden and granular flows
Dispersed solid phases in propulsion and process flows, including particle inertia, gas-particle momentum and heat exchange, loading effects and interfacial-area evolution.
Droplet and spray transport
Injected droplets and sprays in combustors, ducts and cooling flows, including transport, gas-droplet momentum and heat exchange, breakup and coalescence.
GUIX-H integration
An integrated engineering environment around NeuralFlow
GUIX-H connects case definition, solver controls, execution, convergence monitoring and post-processing in one engineering workflow. Its direct communication with NeuralFlow supports interactive simulation control and real-time access to solver and result data.
Explore GUIX-HIntegrated case setupBoundary conditions, physical models, numerics and solution controls managed through a consistent interface.
Real-time solver communicationDirect monitoring and interaction while simulations are running.
Engineering post-processingFields, monitors, plots, streamlines and technical result inspection within the same environment.
Reusable workflowsRestart, automation and repeatable analysis processes for development and production work.
Related LIKUA capabilities
Continue through the connected simulation platform
NeuralFlow works within a broader software and engineering environment covering user interaction, learned solver technology, hydrodynamics and scalable computing.
Development status
Mature capabilities separated from active R&D
Production capabilities are presented separately from models still undergoing implementation, verification or extended validation.
Available solver capabilities
- Highly validated fully implicit and strongly coupled flow solution
- GERANS turbulence and scale-resolving options
- Accurate implicit dual-time transient simulation
- Continuity and boundary-layer solution acceleration
- Species transport and reacting-flow coupling
- Hyperbolically regularized Eulerian particle solver
- IATE, breakup and coalescence
- Aerothermal, CHT and charring-wall modelling
- GPU-supported linear algebra, AMG and geometric multigrid
- Python UDF support and GUIX-H integration
Extended modelling roadmap
- Additional detailed turbulent-combustion frameworks
- VOF interface-capturing methods
- Expanded LES capabilities
- Detailed participating-media radiation models
Technical contact
Discuss a CFD application or validation requirement
Share the geometry, operating conditions, physical models and engineering decision you need to support. LIKUA will define a technically defensible NeuralFlow workflow and the evidence required to trust the result.
Contact