Cloud
Browser access through SimuPort. Public simulators need no installation and no licence request.
Technologies and platforms
The same codebase reaches a project in several forms, depending on what the work requires.
Simulation suite
Geometry handling, meshing, solvers and post-processing in one environment, running on a workstation or on a project’s own servers.
Flowphys software capabilities
Our main product is the software Flowphys, a finite element multi-physics simulation software for general geometries in 1D, 2D, and 3D geometries. Some of the features of the Flowphys software are:
- 1D, 2D, 3D heat transfer, fluid dynamics (CFD), structural dynamics, electrostatics, acoustics
- 1D circuits (multi-phase pipe flow, electric circuits)
- Linear and non-linear, steady state & transient
- System-based Modeling, process simulation, process control algorithms
- Multi-phase, multi-component, incompressible & compressible flows with Newtonian & non-Newtonian fluids
- Computational chemistry equilibrium & reaction kinetics, corrosion, erosion, scaling
- Coupled analyses
- Shape, parameter, and process optimization
- Data-driven (AI/ML) models, AI enhanced physics models, physics informed AI models
Flux
An executable language for simulation — one a machine can write and our solvers run directly.
What it is
A FLUX script is a complete definition of a simulation: geometry, mesh, materials, boundary conditions, solvers, coupling and optimisation, in one readable file. Write it by hand, or let an agent generate it from a description and run it unattended.
What a script can do
- Define geometry, meshes, materials and boundary conditions
- Configure, run and monitor solvers
- Couple physics — phase-field with CFD, CFD with thermo-mechanics, 1D networks with 3D flow
- Run parametric and topology optimisation studies
- Generate batches for surrogate training, and build the surrogates
- Drive the self-improving agents on SimuPort
Why a language, not a menu
An interface records what you clicked. A script records what the model is.
That decides whether a simulation can be checked, repeated or handed to someone else — and whether it can scale. Two thousand melt-pool configurations is a loop in a script and an impossibility in a dialogue box.
Why FLUX, not generated Python
When an AI writes a simulation, the question isn't whether it runs. It's whether you can tell what it produced.
FLUX has a bounded grammar, so a script can be validated before a core-hour is spent — is the problem well posed, are the boundary conditions complete, is the mesh adequate. That check is what lets the agents refuse a bad request instead of returning a confident wrong answer.
And a FLUX script is short enough that someone will actually read it.
Built to be machine-written
FLUX is optimised for modern tokenizers, and our tools curate training files from simulation output automatically.
Efficient tokenisation leaves more of a model's context for the engineering problem rather than the syntax, and makes training more sample-efficient. Most simulation languages were designed before anyone needed a model to write them.
The loop
Simulations produce output → our tools curate it into training files → the model writes FLUX → FLUX runs the solvers.
Agentic framework
Describe a problem in words. The agents build the geometry, generate the mesh, apply boundary conditions, run and monitor the solve, train surrogates and optimise — iterating until the design meets the specification you gave.
The agents check their own work
In the run shown here, the Verifier rejected a mesh as critically under-refined in the longitudinal and outer transverse substrate regions, and reported it rather than solving on it. Feasibility checks run before and after every FEM and CFD solve. A result that fails them does not get returned.
The roster agents
- Geometry and meshing
- Geometry Planner · Implementer · Compiler · Meshing · Verifier
- Pipe networks
- Pipe Supervisor · Network Drafter · Topology Validator · Run Orchestrator · Parameter Advisor
- Across all simulations
- Boundary Conditions · Solver Input · Solver Runtime · Results Checker · Batch Run · Metamodel Trainer · Optimizer · Conclusions
Why it is reproducible
The agents write FLUX, the same language our solvers execute. Every model they produce is a complete, readable definition — open it, change one term, re-run it, hand it to a colleague and get the same result.
Three ways in
Geometry can come from conversation, from imported CAD, or from an image or sketch that our recognition turns into a 3D model or a pipe network.
Deployed on SimuPort today.
Solver library
The physics engines beneath all of the above. Every solver listed here is deployed and running on SimuPort today, covering fluids, structures, thermo-mechanics, materials, electromagnetics and 1D networks. The same solvers power the desktop suite, SimuPort and partner integrations. More solvers are in development and move to production as they are validated.
Fluids
Jet and impinging flows · Heat transfer · Particle tracking · Pipe and channel networks
Manufacturing
Computational welding mechanics · Melt pools · Metal forming
Structures
Structural dynamics · Topology optimisation
Materials and chemistry
Thermodynamics · Chemistry
Electromagnetics
Photonics
Programmatic access
Build on SimuPort. Bring your own solver, model or post-processing into the platform, and reach everyone using it.
What this is
SimuPort is a platform, not a closed product. Partners can extend it with their own solvers, material models, post-processing routines and analysis tools — running on our infrastructure, alongside our physics, callable from FLUX and from the agents.
Why partners do it
- Reach
- Your solver becomes available to anyone using SimuPort, without them installing anything.
- Coupling
- A plugin runs inside our multi-physics environment, so your model can couple to our CFD, thermo-mechanics or phase-field solvers rather than exchanging files with them.
- Automation
- Registered plugins are visible to the agents, so they can be meshed, run, batched and optimised through the same workflow as native solvers.
- Dissemination
- For a research partner, a plugin is a route to publishing a project's modelling contribution as something others can actually run.
In-house computing
Flowphys operates its own high-performance servers for simulation, surrogate training and hosting SimuPort.
Desktop and on-premises
The full suite on local workstations or a project's own servers, keeping data inside the organisation.
Air-gapped and off-grid
Operation on isolated infrastructure for consortia handling sensitive industrial data or with data sovereignty requirements.