There appears to be a limit to how big of a cell size change FDS will tolerate. If the central column is finer by a factor of up to 19 it runs, at 20 it fails the alignment test.

Here is the FDS code that checks:

` MM=>MESHES(NOM)`

ALIGNED = .TRUE.

IF ( (ABS(IOR)==2 .OR. ABS(IOR)==3) .AND. MM%DX(IIO_MIN)<=M%DX(I) .AND. &

ABS( ((IIO_MAX-IIO_MIN+1)*MM%DX(IIO_MIN)-M%DX(I)) / MM%DX(IIO_MIN))>0.05 ) ALIGNED = .FALSE.

IF ( (ABS(IOR)==1 .OR. ABS(IOR)==3) .AND. MM%DY(JJO_MIN)<=M%DY(J) .AND. &

ABS( ((JJO_MAX-JJO_MIN+1)*MM%DY(JJO_MIN)-M%DY(J)) / MM%DY(JJO_MIN))>0.05 ) ALIGNED = .FALSE.

IF ( (ABS(IOR)==1 .OR. ABS(IOR)==2) .AND. MM%DZ(KKO_MIN)<=M%DZ(K) .AND. &

ABS( ((KKO_MAX-KKO_MIN+1)*MM%DZ(KKO_MIN)-M%DZ(K)) / MM%DZ(KKO_MIN))>0.05 ) ALIGNED = .FALSE.

IF (.NOT.ALIGNED) THEN

WRITE(LU_ERR,''(A,I3,A,I3)'') ''ERROR: MESH '',NM,'' is out of alignment with MESH '',NOM

PROCESS_STOP_STATUS = SETUP_STOP

IERR = 1

RETURN

ENDIF

I didn''t bother to track down all those variables, but that tolerance of 0.05 matches pretty well with the 1/20th I mentioned above. It looks like you can''t have a cell size difference of more than a factor of 19 in any direction.