Template-driven discrete element simulation
Demulate is a Windows desktop workbench for the discrete element method: pick a template, adjust parameters, run the solver in an isolated process, and inspect the results — no scripting required.
How it works
Five workspaces, one case at a time — the flow mirrors how a lab test is set up.
1 · Pick a template
Each template is a complete, physically meaningful scenario with a parameter panel generated from its schema — adding a template never means writing UI code.
2 · Adjust parameters
Every parameter carries its declared unit and validated range. Values that cannot build a scene are rejected before a run starts, not after five minutes of solving.
3 · Run
The YADE solver runs in an isolated process with live progress. Stopping is safe; nothing silently half-finishes.
4 · Inspect results
Results land in the open .dres format (HDF5). Analysis cards compute the numbers — angle of repose, stress–strain curves, shear envelopes — and every value carries its provenance.
5 · Export
Export the command stream as a YADE Python script, save the tuned case as a .demcase file, or fit a strength envelope (c, φ) across several runs with Mohr circles.
6 · Invert (where measured)
One template ships an inverse solve: back out particle friction from a target angle of repose — with its measured anchor curve and caveats stated right on the panel.
Why a workbench instead of scripts
The alternative is not wrong — it is slower in exactly the places a workbench can be honest about.
| The usual route | Demulate | |
|---|---|---|
| Setup | Write and debug Python scripts against the solver API, or click through a generic CAD-style modeler. | Here: Pick a template — the parameter panel is generated from the template schema, so every knob the UI offers exists in the solver. |
| Feedback | Errors surface mid-solve, five minutes after launch. | Here: Values that cannot build a scene are rejected before the run — units, ranges and cross-field rules are checked up front. |
| Solver | Physics modified in-house, or a black box you cannot inspect. | Here: YADE runs in an isolated process and the app never modifies its physics; the engine plan is published with acceptance criteria. |
| Results | Opaque dumps you parse by hand. | Here: The open, versioned .dres format (HDF5) plus 7 analysis cards — and every value traces back to the frames it was computed from. |
| Claims | Trust the vendor. | Here: A public ledger ships with the app: 35 verified, 6 planned, 7 excluded with written reasons. |
Before you trust a number
Every figure below came out of a real solve — and the app is built so you can check that yourself.
Estimate before you commit
The estimate bar states the particle count and the expected runtime before you press Run — including what a larger grain diameter would buy you.
Rejected before the run
Parameters carry declared units and validated ranges. A configuration that cannot build a scene never reaches the solver.
Provenance or nothing
No value on a result card is hand-waved: each one traces back to the queried frames it came from. The software never invents an estimate outside the solver.
Gates that name themselves
When something would be slow, the UI says so and quantifies it — the fibre gate names itself and states what switching it off would save, instead of just reporting "too slow".
Measured on the core
Every figure below came out of a real solve or a measured check — the test conditions sit right next to each number.
Who it is for
Labs calibrating micro-parameters
Angle of repose, shear envelopes and triaxial curves are exactly what the five templates produce — with presets to start from.
Teams with existing DEM data
The box-lift template mirrors the EDEM Calibration Kit standard case, giving the most direct cross-validation channel against commercial results.
Engineers studying granular scenarios
Slope runout, pile penetration behaviour, shear strength — run the scenario, read the cards, export the case for a colleague.
Students learning DEM
No scripting barrier: pick a template, watch the physics respond to a knob, then export the YADE script to see what ran underneath.
An honest capability ledger
The solver capabilities are tracked in a ledger that ships with the app — 35 verified, 6 planned, 7 excluded with written reasons. What the UI offers is exactly what the ledger says.
Open the capability map →Where things stand
Labeled honestly as shipped or planned. No promises without dates set.
- Desktop workbench with 5 templates, 13 preset groups and 8 case recipes
- Capability ledger: 35 verified / 6 planned / 7 excluded
- Open .dres result format with a version contract
- YADE script export and .demcase case files
- Portable Windows package, ~344 MB
- Widening the verified capability set (the ledger grows before the UI does)
- More inverse solves with measured anchor curves and stated caveats
- Result viewing in the browser (upload a .dres, read-only)
- A template-distribution channel — only if a real need shows up
- In-house GPU kernel, gated on statistical equivalence (stage 3)
Powered by YADE
The solver is YADE (OpenDEM), a widely used open-source DEM engine, executed in an isolated process. Demulate never modifies YADE physics — and the staged plan is to first vendor the exact solver build, then grow an in-house GPU kernel that must reproduce it statistically before replacing it.
Read about the engine stance →Questions you may have
Do I need to write scripts?
No. You pick a template and work in a parameter panel. When you want to see what ran underneath — or continue in YADE directly — one click exports the command stream as a YADE Python script.
What does the solver run on?
YADE runs inside WSL2 (Ubuntu 22.04) — one apt install away on any Windows 10/11 machine. The interface runs natively on Windows; browsing templates and reading past results work without WSL, running a new simulation connects to the WSL engine.
Is it free?
Yes — no license, no subscription, no feature gates. The download page states exactly what ships.
What is a .dres file?
The open result format: HDF5 with particle trajectories, time series and metadata, behind a versioned format contract — old results still open in new versions. Copy the whole case folder to a colleague and it opens on double-click.
How do I know the numbers are right?
Three ways: the capability ledger ships with the app and states what is verified; every analysis-card value carries provenance down to the frames it was computed from; and the pre-run checks refuse configurations that cannot build a scene.
The inverse solve gives a friction value — can I trust it directly?
Treat it as a curve inversion, not an analytical solution. The shipped anchor curve was measured at specific settings (damping 0.2, wall friction equal to particle friction); change either and the curve no longer applies — the template states this on the panel, with the measured caveat that the absolute angle sits about 10° below real sand because the model has no rolling friction.
Is my data sent anywhere?
No. The workbench and the solver both run locally — there is no account, no telemetry and no cloud component. Case files are plain folders on your disk.
When can I download it?
The app is pre-release: the package builds and runs, the first public distribution is being prepared. The download page renders that state honestly — no fake buttons.