Common FDS input file errors and how to fix them
fdsrun editorial team · · 7 min read
Missing &TAIL, overlapping meshes, devices outside the domain, references to undefined SURF: errors that stop FDS or spoil the results.
Most failed FDS runs fail not on physics but on the input file: a missing slash, a mesh exported twice, or a device that ended up outside the domain after a geometry change. Some of these errors stop FDS immediately; others let it compute for hours and spoil the results. Here are the most common ones, from the most serious down.
Errors that stop FDS
Unclosed namelist group (missing "/")
Every group starts with &NAME and ends with a slash /. A missing slash — often after manual editing or with an odd number of apostrophes in a text — makes FDS read the following lines as part of the same group.
&SURF ID='BURNER', HRRPUA=500.0, COLOR='RED' ← missing /
&OBST XB=0,1,0,1,0,0.5, SURF_ID='BURNER' /No &HEAD with CHID, or invalid characters
CHID names every output file. Use only letters, digits, _ and -, no spaces or dots: &HEAD CHID='hall_variant1' /.
No &TAIL at the end of the file
The last line of the file should be &TAIL /. Some export tools, or joining files together, lose it.
Reference to an undefined ID
SURF_ID, MATL_ID, RAMP_ID, PROP_ID, CTRL_ID — if they point to a name that was never defined (or differs in letter case), FDS stops right after the start. The typical story: a material renamed in one place and the old name left in another.
A mesh without valid IJK or XB
Every &MESH needs IJK (cell counts in three directions) and XB (bounds in metres). Coordinates in the wrong order in XB (for example x2 < x1) give a mesh with a negative size.
MPI_PROCESS order
Process numbers must be consecutive and ascending, starting from 0. A gap (for example 0, 1, 3) or a descending order is a start-up error in MPI mode.
Errors that spoil the results
No T_END
Without &TIME T_END=… FDS defaults to 1 second of simulation. The run "succeeds", but the result is useless.
Overlapping meshes
FDS allows meshes to overlap — the one listed first then takes precedence. In engineering models, however, overlap is almost always a mistake (a mesh exported twice, or shifted by one cell) that adds cells for nothing and disturbs the data exchange between meshes. Meshes should abut face to face.
Mismatched cells at mesh interfaces
At the boundary between two meshes, cells should be the same size or in an integer ratio (1:2, possibly 1:3). Non-integer ratios and large resolution jumps are a common source of velocity errors at interfaces (more: FDS cell size).
Devices outside the domain
A &DEVC with XYZ outside every mesh measures nothing. It happens after mesh bounds change while the devices stay in their old places — and it only shows up when you analyse the results.
Walls without TMP_INNER at an unusual temperature
If the air in the model starts at a different temperature from ambient (&INIT) and the walls have no TMP_INNER, their interior starts at TMPA. Sometimes intended, often an oversight.
Small things worth fixing
- Duplicate &OBST — identical obstructions can be removed without changing the model.
- Zero-thickness obstructions — FDS snaps them to the nearest cell face; for a thin partition, check that it ended up where it should.
- Obstructions outside the meshes — they take no part in the calculation (FDS cuts them off), but they clutter the file.
- No DT_RESTART — by default FDS writes no restart files; after a machine failure you start from zero.
&DUMP DT_RESTART=300 /solves it.
How to check quickly
Some errors show up when FDS runs with T_END=0 (set-up only, no computing), and the geometry is worth a look in Smokeview. Faster: upload the file to the free fdsrun tool — it checks every problem described above in a few seconds, gives the line number and suggests a fix.
More guides
Sources: FDS User Guide (NIST SP 1019), sections 5.6, 6.2, 6.3, chapter 7; fdsrun experience from analysing users' files. FDS and Smokeview are software of the National Institute of Standards and Technology; NIST is not affiliated with fdsrun.