Primitives and booleans
Start from boxes, cylinders, and other primitives, then combine them with boolean union, cut, and intersection, plus transforms, to build the model geometry directly in the app.
Build the geometry, mesh it, solve the coupled thermo-hydro-mechanical physics, and read the 3-D result — without leaving a single desktop application. An AI surrogate sits alongside the solver to make design exploration real-time.
Geometry, meshing, the coupled solver, the AI surrogate, and the result viewer share one model. Nothing is exported to an external pre-processor and re-imported.
Build and edit solids on an owned CAD kernel — primitives, booleans, transforms, and face-level boundary conditions.
Learn more →Automatic tetrahedral meshing at coarse, medium, or fine resolution — no manual element work.
Learn more →A coupled thermo-hydro-mechanical finite-element engine that resolves heat, flow, and ground deformation together.
Learn more →A validated Gaussian-process surrogate of the coupled solver — design exploration roughly 985,000× faster.
Learn more →An in-app 3-D viewer with colour-mapped temperature fields, the pipe loop, and the model domain.
Learn more →Build the model from scratch on an owned CAD kernel — no external pre-processor in the loop.
Start from boxes, cylinders, and other primitives, then combine them with boolean union, cut, and intersection, plus transforms, to build the model geometry directly in the app.
Pick individual faces of a solid and assign temperatures or other boundary conditions to them, so the boundary setup follows the geometry instead of a separate mesh-side workflow.
Save a model and reopen it later. The geometry flows straight into meshing and the coupled solve as one reproducible pipeline — single piles to whole pile fields.
A built-in mesher turns the geometry into an analysis-ready tetrahedral mesh, automatically.
Mesh the model directly from the built geometry — no hand-built elements and no external pre-processor between geometry and solve.
Choose a resolution to trade speed against detail. A design mesh resolves the trends quickly; a fine mesh sharpens the field where it matters.
The mesh hands off to the coupled engine as part of one reproducible pipeline, so geometry, mesh, and solve stay self-consistent.
A custom Fortran finite-element solver resolves heat, pore-water/gas flow, and ground deformation together, with an embedded pipe-network model for the circulating fluid.
A monolithic Newton–Raphson scheme solves the thermo-hydro-mechanical fields together, capturing the coupling that simpler models leave out.
Seventeen soil and rock constitutive laws, on original, dependency-free kernels — single piles, borehole loops, and whole pile fields.
An embedded pipe-network model carries the circulating fluid through the soil — the pipe↔pile↔soil heat exchange resolved inside the same solve.
A multi-output surrogate, trained on real coupled-FEM runs, returns the full loop performance — temperature, heat, and pumping cost, with a confidence band — roughly 985,000× faster than a full solve.
Outlet temperature, heat delivered, and loop pressure drop in one shot — validated to R² 0.9997 against the held-out solver, with a mean error of 0.009 °C across 200 held-out runs.
Global sensitivity (Sobol) and confidence intervals, instantly: which inputs drive the outlet temperature, and which drive heat and pressure drop.
Map thousands of candidate designs to a heat-versus-efficiency frontier, or turn a plain-English brief into a sized, simulated design.
Read the result without exporting it. An in-app 3-D viewer renders the colour-mapped field straight from the solve.
The temperature field renders as a colour-mapped 3-D view, with a colour bar, right where you ran the solve.
See the soil field, the pipe loop, and the model domain together — self-consistent IDs straight from the same solve, no re-import.
Rotate the result to look at the field from any angle — interactive 3-D results, in the same window as geometry, mesh, and solve.
A standard Windows desktop is all it takes. Terra Multiphysics is a self-contained application — its own geometry builder, mesher, and 3-D post-processor, with a native solver binary.
Everything runs locally on your machine. No cloud account, no cluster, and no internet connection required to model, mesh, solve, or view results.
Project data never leaves your computer. The optional natural-language copilot is the only feature that can call an external service, and only if you switch it on with your own API key.
Geometry, meshing, the coupled solver, and the post-processor ship as one installer — no external pre-processor and no dependency chase.
We're onboarding a small group of early-access design partners now.
Request access