Fields, meshing and simulation
Exact signed-distance fields on the true B-Rep surface today; meshes and simulation results traced to the faces that produced them, feeding back into the design, next.
The problem
Simulation today starts by throwing the design away: the model is exported, meshed in another tool, solved in a third, and the result comes back as a picture. Nothing links a hot spot to the face that caused it, so turning a result into a better design is manual, slow and easy to get wrong.
What Roshera does today: exact fields
Roshera computes the signed distance from any point to a solid, negative inside the material and positive outside, from the B-Rep's analytic surfaces. The sign comes from the surface normal at the nearest point, correct for every surface type, tori and NURBS included. Fields over a region are sampled with an adaptive octree that refines only where the surface can pass, and every sample records the face that gave its distance.
Used for wall-thickness and clearance checks tied to the exact part, for comparing two versions of a part (where material moved), and as the base for lattices and variable-thickness shells.
What we are building: the closed loop
- Mesh the certified part, with every element traced to its face and the mesh quality certified, like everything else in the kernel.
- Simulate it: stress, heat, flow, on the real geometry.
- Feed back: turn the result into a change the kernel can make and check, such as moving material, reshaping a hole or adding a fillet, and record it in the part's history like any other edit.
The animation on the home page shows the loop on a textbook case with real physics: a plate in tension with a hole has a peak stress three times nominal at the hole's edge (Kirsch); a slot aligned with the load lowers it to 1 + 2b/a (Inglis).
Where it stands
- TodayExact signed-distance fields, adaptive sampling, face provenance, mass properties.
- BuildingCertified meshing, simulation, results feeding design changes back into the timeline.
Talk to us
If simulation-driven design is your problem, or you want to train models on it: info@roshera.tech.