Skip to content

Outdoor+ Solver Troubleshooting

Field notes for runs of the Outdoor+ (urbanMicroclimateFoam) case that mesh fine but stop during the solve. For installation problems see Setup Troubleshooting; for the workflow itself see First Steps.

Where to look first

Everything the solver prints lands in the case folder (the Working Directory of the Outdoor+ Case component, one sub-folder per case name):

  • the solve log — the last Solving for solid region <name> line before the error tells you which region diverged (buildings, terrain, or a vegetation region);
  • checkMesh_<region>.txt — one per meshed region, written at the end of meshing;
  • the *_dump fields (Ts_dump, pc_dump, ws_dump, Crel_dump) written into the last time folder when the solid solver gives up — load them in ParaView to see where the field ran away.

This is going to crash (Ts)! … Cannot decrease time step further

The solid regions (buildings, terrain) are solved with an adaptive sub-time-step inside each hourly fluid step. After every solid iteration the solver checks that the solid temperature Ts stays within 0 K < Ts < 423.15 K (150 °C) and that the capillary pressure pc stays negative. If not, it prints This is going to crash (Ts)! (or (pc)!), reverts the fields, halves the time step and retries. When the time step reaches minDeltaT (the written case uses 1e-6 s) it aborts with Cannot decrease time step further.

The message says nothing about the cause, so always read the log: the last Solving for solid region ... line names the region (buildings or terrain) that diverged. Known causes, in the order to check:

1. A light hygroscopic wall material

Materials with a moisture model and a very low density/conductivity — CalciumSilicate (ρ 270 kg/m³, λ 0.06 W/mK) and Hamstad5Insulation (ρ 130) — diverge in the coupled heat-and-moisture solve as soon as the air field has converged and the surface starts receiving solar radiation (seen on an internal case, 2026-10). It is not a sufficient cause on its own: a flat-terrain courtyard case with a CalciumSilicate inner layer ran through sunrise without a single time-step halving.

Integer-wired Material inputs

The Material input of Building Material, Terrain Surface Material and Soil Material expects a material name. Older documents wire a Number Slider into it; the plugin then resolves the number by position in the dropdown list, so slider 1 silently means CalciumSilicate. Prefer the dropdown (right-click the input) or a panel with the name.

index name ρ (kg/m³) λ (W/mK) moisture model
0 AsphaltConcrete 2100 0.75 yes
1 CalciumSilicate 270 0.06 yes, strongly hygroscopic
2 Hamstad5Brick 1600 0.68 yes (plugin default)
3 Hamstad5Insulation 130 0.04 yes, strongly hygroscopic
4 Hamstad5Mortar 1900 0.82 yes
5 HamstadBrick 2005 0.5 yes
6 HamstadCase2 1600 0.68 yes
7 HamstadConcrete 2400 1.6 yes
8 HamstadPlaster 790 0.2 yes
9 Impermeable 2400 1.0 none — heat conduction and storage only
10 PorousAsphalt 2000 0.7 yes
11 SabaBrick 1608 0.73 yes
12 SabaBrickMod 1608 0.73 yes
13 Savonnieres (buildings) / Soil (terrain) 1830 / 1150 1.2 / 1.5 yes
14 Soil (buildings) / VanGenuchten (terrain) 1150 / 1500 1.5 / 1.0 yes

Remedy: use Impermeable for both wall layers (stable, models heat storage and conduction, no moisture in the wall), or a heavy hygroscopic material such as Hamstad5Brick or HamstadConcrete. To bisect, set every material (walls, terrain surface, soil) to Impermeable once; if that run finishes, bring the moisture models back one region at a time.

2. Inverted cells from the wall extrusion

The solid buildings and terrain are made by extruding the surface inwards (two layers, 0.1 m each by default, 10 cells per layer at expansion ratio 1.2). Wherever the surface has a feature narrower than roughly twice the wall thickness — a 20 cm ledge, a fin, a thin parapet, sliver triangles between unwelded faces — the offset prisms cross and checkMesh_buildings.txt reports negative cell volumes. The solver then fails on the very first solid step, at t of a few seconds, with the same message as above.

The plugin reads the checkMesh logs and raises a warning on Run when it finds negative volumes (since 1.14). Remedy: remove or thicken the thin features in Rhino, pass the façades through Gmsh Mesh, and re-mesh.

Small determinants are expected

checkMesh -allGeometry always reports many cells with small determinant (< 0.001) in the building and terrain regions. The thinnest extruded cell is about 4 mm thick under a face that is 1–3 m wide; that aspect ratio gives tiny determinants by construction and does not affect the one-dimensional conduction through the wall. Only negative volumes, incorrectly oriented faces or severely non-orthogonal faces in a solid region are a problem.

3. An air field that has not converged

The air region is solved to steady state inside every hourly step, from whatever field it has. The first ~100 iterations of a cold start are a violent transient — the log shows Min/max T: pinned to 273.15 / 373.15 K and very large |U| — that has died out by iteration ~100 and converged by ~500. Max Fluid Iterations below 300 freezes that transient into hour 1 and hands the solid regions a garbage boundary condition. Keep it at 300 or more; 500–1000 for the first hour is cheap insurance.

4. Time-step controls

Only after the three points above: Simulation Settings → Min Delta T (1e-6 s in the written case; a wired value overrides it) and Initial Solid Time Step Factor reduced, to give the solid solver more room. This delays a material-driven divergence rather than curing it.

This is going to crash (pc)!

Same mechanism, but the moisture equation overshoots pc ≥ 0 (saturation). Seen with the Soil and Hamstad models from a cold start; cases written by plugin 1.10 or later carry a floor on the moisture capacity (minCrel) that removed it in all tested cases. If it still appears, switch the affected region to Impermeable and report the case.

Need help?

Open a thread in the Eddy3D Discussions and attach the .gh, the tail of the solve log (from the last Solving for solid region line) and the checkMesh_*.txt files.