Skip to content
SHD Sim
Menu
PricingAbout
Sign in

Changelog

Releases

Newest first.

  1. 0.3.5

    SHD Sim 0.3.5

    Several analysis types were silently running the wrong solver. Solid-only thermal ran laplacianFoam and ignored anisotropic conductivity; spray had droplet break-up hardcoded off; boiling and fire ran a solver neither of them should have; premixed combustion's ignition tuning was dead code; N-phase cases carried a phase list that was never built; and Pedestrian Wind Comfort ran a single direction relabelled, not a sweep. If you have results from any of those, re-run them — the set-up you entered was correct, the solver behind it was not.

    Separately, 2-D cases were meshed 32 cells thick in the direction they call empty, which poisons force coefficients and, after refinement, produced meshes that would not run at all and said nothing about why.

    This release also adds acoustics as a fifth physics family, coupled fluid–structure simulation via preCICE, and SHD Cloud compute. Windows.

    Fixed — results that were wrong

    • Solid-only thermal ran the wrong solver. It used laplacianFoam, and the anisotropic conductivity you could enter was never read.
    • Spray break-up never happened. breakupModel was hardcoded to none on every leaf that offered one.
    • Boiling and Fire ran a solver that models neither.
    • Premixed combustion did not ignite. The ignition tuning was dead code.
    • N-phase cases had no real phase list on the three leaves that promised one.
    • Pedestrian Wind Comfort was one direction, not a sweep.
    • 2-D cases were not 2-D. A snapped 2-D case came out 32 cells thick across the direction declared empty. The empty patches then distort force coefficients, and refining such a case produced an unrunnable mesh with no diagnostic. Both are fixed, and a collapsing time step now says so.
    • Aerofoil layers. Asking for a wall-resolved mesh produced a worse one than the default; the near-wall layer was sized by an unclamped turbulent correlation.
    • Flat solid-result colouring and mislabelled vector components.
    • Material browser was broken across every family; 2-D patch sync and a stale result-control crash are fixed with it.
    • Kill left parallel ranks orphaned.

    Added — physics

    • Acoustics, the fifth family: thermoviscous ducts and BEM radiation, with the air region meshed for you.
    • Coupled fluid–structure simulation through preCICE — a coupled study is its own family, with its own panes and runner. The two components are fetched at install like every other backend, and it runs on native Windows with no WSL.
    • Solid families brought up to the fluids side: nonlinear static past the elastic limit, harmonic response, transient dynamics, fatigue life, eddy currents and thermal radiation — each verified against the solver rather than assumed.
    • Aeroacoustics: Curle dipole sources and Proudman acoustic power.
    • Irregular waves: JONSWAP spectral discretisation.
    • Elmer and code_aster convergence now drawn in the Residuals pane.

    Added — SHD Cloud compute

    Paid tiers can now run meshing and solving remotely and have results delivered back into the app.

    • The app estimates the run before you commit and shows the Compute Hours it expects to use, from mesh size, step count and cores.
    • When the job finishes it reports what it actually cost, measured from real solver runtime — not from the estimate.
    • Jobs that fail or are cancelled are never charged, and the app says so.
    • Every job carries a wall-clock ceiling, so a hung run cannot bill indefinitely.
    • Billing starts when the solver starts, not when the worker picks up the job: fetching the backend image is our cost, not yours.

    Changed — licensing

    The app now checks your licence every time it starts. Previously it only asked once the cached check-in window had lapsed, which could be a week away.

    • A licence that has been withdrawn, or a machine deactivated from your account, now returns the app to the Free tier promptly. It used to display a warning and carry on with paid entitlements.
    • Floating seats have a much shorter offline grace than named seats. An offline floating client holds a seat the pool has already released, so the window is now one day rather than a fortnight.
    • Nothing about this touches your work. Cases open, results stay, and nothing becomes read-only.

    Validation

    Published benchmarks now cover all four backends rather than the fluids alone, including first Tier C cases for structural, electromagnetics and acoustics. The cylinder and NACA 0012 studies were re-run and written up, including a grid-convergence study extrapolating to within 0.8 % of Ladson.

    Notes

    No solver backend version changed in this release.