LythosFEA — reference

Command line

Command	What it does
lythos gui	the browser interface (port 8777)
lythos examples -o models	write the worked examples as model files
lythos mesh model.json	mesh statistics, without analysing
lythos run model.json -o out	analyse, write an HTML report and JSON
lythos import drawing.dxf -o model.json [--plot p.png] [--keep-coordinates]	a model from DXF
lythos import --sample -o model.json	import the drawing that ships with the package

From a clone every command also runs as python main.py ….

DXF import

Which drawing layer becomes what is decided by its name, matched case-insensitively and ignoring hyphens, underscores and spaces — SU-SEVIYESI and su seviyesi both read as the water table.

Drawing layer contains	becomes
soil, clay, sand, rock, fill, zemin, kil, kum, tabaka, …	a soil layer
wall, pile, sheet, diaphragm, perde, kazık, …	a wall or pile row
water, phreatic, gwl, su seviyesi, …	the water table
load, surcharge, yük, …	a line load
anchor, strut, prop, ankraj, …	an anchor
text, dim, hatch, grid, defpoints, ölçü, …	ignored
anything else	soil if the outline is closed, a structure if not

Soil regions are recovered as the faces of the planar arrangement the lines make, so both usual CAD conventions work: each stratum as its own closed polyline, or one outer boundary plus the lines dividing it.

The construction sequence

A number at the end of a layer name is the step at which that thing happens:

Layer	Meaning
WALL-1	the wall is built at step 1
EXC-2, KAZI-2	this region is dug out at step 2
ANCHOR-3	the anchor is stressed to its lock-off load at step 3
SURCHARGE-1	the load is applied at step 1
SOIL-CLAY	no number: present from the start, never removed

Excavation regions can be drawn either way: as closed outlines of each lift, or as the dig levels drawn from the wall outwards. Drawing units are read from $INSUNITS (a drawing in millimetres arrives in metres). Survey coordinates are moved to the origin and the shift is reported; --keep-coordinates turns that off.

Validation

Every claim below is a test in tests/:

Check	Reference	Lythos
Mohr-Coulomb failure deviator, plane strain	σ1 = σ3Kp + 2c√Kp	within 0.5 %
Active earth pressure, zero lateral strain	Kaσv + 2c√Ka	exact
Geostatic stress and settlement under self weight	γz, γH²/2E	exact
Patch test, linear displacement field	constant stress	exact to 1e-12
Cantilever tip deflection	PL³/3EI + PL/GA	within 0.2 %
Slender beam, h/L = 1/1000	no shear locking	within 2 %
2:1 slope factor of safety	1.377, independent Bishop search	1.381 on a 471-element mesh
Algorithmic tangent	numerical derivative	within 4e-4 of E, everywhere

pytest                    # the quick suite
pytest -m slow            # the full analyses as well

Full table: docs/validation.md; formulation: docs/theory.md.

Limitations

Worth knowing before trusting a number to a design:

• Drained or total-stress (undrained) analysis only; there is no consolidation, no transient flow and no coupled pore pressure. Undrained behaviour is modelled by giving a layer phi = 0 and c = su.
• Pore pressure is hydrostatic below the phreatic surface. There is no seepage analysis, though the horizontal gradient of an inclined phreatic surface is carried as a seepage body force.
• Small strain throughout; no updated-mesh or large-displacement option.
• Elastic-perfectly plastic soil. There is no hardening model, so settlement predictions under working loads are only as good as the single stiffness chosen for the stress range that matters.
• A row of piles is smeared into an equivalent plate, the usual plane-strain idealisation; it says nothing about arching between the piles.
• Quadratic triangles are much better than linear ones under constant-volume plastic flow, but not immune to volumetric locking. Undrained (phi = 0) collapse loads are slightly on the high side; refine, and treat an undrained factor of safety from a coarse mesh with particular suspicion.
• Strength reduction reports the factor at which equilibrium is lost. Like any such analysis it is sensitive to how the failure criterion is judged, so the displacement-versus-factor curve is reported alongside it and should be looked at.