AeroCRAFT
AI-agent direction for additive-manufacturing process simulation
Product overview
An agentic workflow for metal additive manufacturing: import, slice, plan the toolpath, simulate the heat, and export to the solvers and robots downstream.
Status
In development
Evidence boundary
Technical work in progress — not a finished product demo.
Capabilities
- Process path
- Computed thermal response
- Computed material response
- Research collaboration
From CAD to toolpath
Slice a real part into typed layer polylines and see every laser-on move separated from every dry travel hop before anything runs.
What the clip shows
A bracket sliced into typed polylines - outer wall, inner wall, infill, contour and cladding, each in its own colour - which then resolve into one continuous red deposition path with the green travel hops between islands called out.
Illustrative animation. The figures shown inside this clip are not measured values and are not quotable.
Predict defects before printing
Run a recipe through the transient solver and get a verdict, pinned hotspots and a ranked fix list before any wire is burned.
What the clip shows
One calibration cube built with three different energy recipes. The melt pool travels through the voxel grid, the energy-per-unit-length dial swings out of the fusion window at both ends, defect hotspots pin themselves in place, and the build is scored with the penalties broken out.
Illustrative animation. The figures shown inside this clip are not measured values and are not quotable.
One file, the solvers you run
Toolpath and thermal mesh go into a single VMAP file and come out for LS-DYNA, MSC Marc and Abaqus, with an ABB or KUKA robot program alongside.
What the clip shows
A checked toolpath collapsing into one VMAP file that carries both the path and the thermal mesh, with speed, power, laser state and bead width riding inside as named state variables, then fanning out to LS-DYNA, MSC Marc and Abaqus and to a robot program that advances point for point with the 3D path. VMAP is a neutral container, so other solvers and robots are possible in principle; these are the ones the workflow has been taken through.
Illustrative animation. The figures shown inside this clip are not measured values and are not quotable.