A completely new Engineering Workflow

Agent driven CFD. A complete, self contained simulation platform where engineers and AI agents work on the same goal.

Get notified at releaseExplore

About Flux

Flux Engineering is a startup building CFD software by engineers, for engineers. One application takes you from CAD import through meshing and solving to post-processing.

One wing, from CAD to results
  1. A swept flying wing with winglets, as a smooth grey CAD solid.

    Geometry

    Build or import the CAD and heal it.

  2. Structured mesh around the root section of the wing: an O-grid hugging the airfoil, refined towards the wall, inside a rectangular block grid.

    Meshing

    Structured, boundary-layer resolved meshes.

  3. Mach number around the wing section at mid-span, highest over the front of the upper surface, with the wake behind the trailing edge.

    Solver

    SU2 or XLB, on your own machine.

  4. Pressure coefficient on the upper surface of the wing, lowest along the leading edge and near the winglets.

    Analysis

    Surface pressure, fields, streamlines and forces.

Current software: accessible but expensive, or cheap but hard to implement.

Flux: accessible, AI driven and cheap.

The CFD market keeps growing, yet advanced simulation has priced out talented teams, startups and researchers for too long, and the software has barely changed in years. We use recent advances in AI to make the slow engineering workflow fast.

Flux will be free for non-commercial use, with no artificial limits. For commercial clients we offer a modular version for secure enterprise deployment that ensures data sovereignty.

Agents, with the engineer in charge

Engineers test a few configurations because every simulation takes hand work. In Flux, an agent will drive the whole workflow, from geometry to results, and test a large range of configurations in a short time. The decisions stay with you, made on data.

  1. You set the goal

    State the objective, the constraints and the range to test. For example: the highest maximum lift for a high-lift system, over flap and slat deflections, gaps and overlaps.

  2. You approve the plan

    The agent proposes the configurations to run and the compute they need. Nothing starts before you approve it, and the limits you set are hard limits.

  3. The agent runs the campaign

    Geometry, mesh, solve and extracted results for every configuration: many fast runs first, the promising flap and slat settings verified at high fidelity. Every step is recorded.

  4. You decide

    You review the aggregated results and choose the configuration. Agent output is evidence, never sign-off.

Agentic workflow: the engineer states the design space, an agent runs eight low-fidelity simulations, the three most promising go to high-fidelity simulation, and the engineer reviews the strongest result.

Capabilities

Meshing for CFD

3D fluid mesh around the letters pySMESH, cut at mid depth, with the boundary-layer prism cells coloured from red at the wall to yellow at the edge of the layer stack.

3D boundary-layer prism meshes with pySMESH. Structured meshes, anisotropic adaptation and quad surface meshes. SALOME SMESH and Open CASCADE are built into Flux.

Adjoint shape optimization

Pressure coefficient on the surface of a flying wing, from low pressure in blue to high pressure in red.

Deform the shape and see the predicted change in drag or lift at once. A full solve then verifies it. Built on the SU2 discrete adjoint. One adjoint solve gives the sensitivity of the objective to every surface point, however many design variables there are.

From CAD to results in one app

Import and repair CAD, mesh, solve and post-process in one Windows application. It runs locally, with no server and no cloud.

Solvers

Streamlines over a flying wing from the front, coloured by local speed, with the tip vortices trailing from both winglets.

SU2 for steady and unsteady flow: RANS, URANS, DES and LES. XLB lattice Boltzmann on the GPU. Every SU2 solver and setting is available. Defaults are carefully set for maximum accuracy and speed. Pick the model the flow needs, as on the wing below.

One wing, two flows

The same flying wing, solved two ways. RANS represents the attached flow, LES the separated flow. Switch between them and turn the wing around.

Loading the solution

Speed / free stream
0.71.01.3
Drag to orbit, scroll to zoom.Swipe sideways to orbit, pinch to zoom.

Contact Us

[email protected]
Get notified at releaseFollow on LinkedIn