CIVL 4750 Numerical Solu5ons To Geotechnical Problems: Lecture 7: 2D FEM For Stress-Strain Analysis
CIVL 4750 Numerical Solu5ons To Geotechnical Problems: Lecture 7: 2D FEM For Stress-Strain Analysis
CIVL 4750 Numerical Solu5ons To Geotechnical Problems: Lecture 7: 2D FEM For Stress-Strain Analysis
Jidong Zhao
Last & This Lecture
• Last lecture
– ConGnuum mechanics
– ConsGtuGve models for soils
• This lecture
– RealisGc simulaGon of geotechnical problem
• Stress-strain, force-seQlement, excavaGon etc
– Basic procedures and issues
– Using 2D FEM as an effecGve tool / PLAXIS
d
f Basic unknowns in sGffness method of
Downwards structural analysis are nodal displacements.
Load applied
displacement of The final problem wriQen as linear
to the Spring
the Spring equaGons in terms of nodal displacements.
• CompaGbility
Spring 1: k
1
d1 f1 • Material behavior
Spring 2: k
2
d2 f2
ta
Node 1 Node 2
tb f1
f2
• Matrix form
⎡ 2k −k ⎤ ⎡ d1 ⎤ ⎡ f1 ⎤
⎢ −k k ⎥ ⎢ d ⎥ = ⎢ f ⎥
⎣ ⎦ ⎣ 2⎦ ⎣ 2⎦
Mesh
Domain
x
CIVL 4750 | Lecture 7 | HKUST 15
FEM Shape Func5on
The elements can be rectangle, straight-
sided and curved side quadrilaterals.
In all cases, the local (or natural)
coordinate system is defined such that
the coordinates vary from -1 on one side
to +1 at the other, taking zero over the
quadrilateral medians.
– IteraGve method
• Gauss-Seidel or Jacobi iteraGons
• Accuracy depending on number of iteraGons
• Efficient, but convergence not guaranteed
– Nonlinear method
• Newton-Raphson iteraGon
• For nonlinear consGtuGve relaGons (elastoplasGc)
• Non-convergence problem
• Improved by reducing step size of load, refining mesh and others
CIVL 4750 | Lecture 7 | HKUST 25
Sample FE Formula5on
σ
b At each node, there should be
equilibrium of nodal forces
1 3 1 5
2 2 3 4
(a) (b)
2 2
1 3 5 5 6 7
3
4 1 4
(c) (d)
• There are only enough nodes to crudely approximate the
end nodes for a simply supported beam
• Five nodes are required to capture the second mode as
shown in (c)
• Seven nodes are required to capture the third mode in (d)
• more elements? Or elements of more nodes?
CIVL 4750 | Lecture 7 | HKUST 28
Guidelines for Meshing
• Use as uniform a mesh paQern as possible.
• When making transiGons from coarse‑ to fine‑mesh regions,
do not change the dimensions of adjacent elements by more
than a factor of 2. If necessary, make the transiGon over a
series of elements to maintain this factor.
• Use quadrilaterals wherever possible. Use triangular
elements only for transiGons or when required by the
geometry.
• The aspect raGo (length to width raGo) should be kept as
close to unity as possible. Aspect raGos of up to 5.0 are
permissible but should be kept lower than 3.0, if possible.
• Do not use elements of extremely acute or obtuse angles.
• Poisson's raGo < 0.5. For materials approaching Poisson's
raGo of 0.5 special elements are required to prevent locking.
CIVL 4750 | Lecture 7 | HKUST 29
Convergence of Solu5ons
• Equilibrium Errors
– Source 1: yielding and the subsequent stress correcGon always gives rise to
out‑of‑balance loads.
– Source 2: Due to sharp changes in sGffness or inappropriate Gme.
– These errors should be <5% or preferably 1%.
• Numerical Conversion
– The numerical answer is not dependent on the increment size, both mesh and loading.
– Refine the mesh and see if the answer changes.
– Increase the number of iteraGons unGl the final soluGon doesn't change much.
– Divide load into smaller steps.
– Replace stress‑control loading with displacement controlled loading.
– Higher order elements can also be used to provide a beQer soluGon.
• Check Ques5ons
– Does the deformaGon paQern look reasonable?
– Do the stress paths (you need to calculate and plot) behave in an acceptable manner?
– Are principal stress/strain direcGons consistent with applied loads?
– Is the distribuGon of pore pressures sensible?
– Is the distribuGon of the strains credible?
• Consolida5on Analysis