CEN4000-N

Advanced Geotechnics

Finite Element Modelling

Dr Wengui Huang

1
School of Computing, Engineering & Digital Technologies 1
 What is numerical modelling?

Mesh size matters?

Boundary conditions
Affected by the
number of nodes

2
School of Computing, Engineering & Digital Technologies
 What is numerical modelling?

Depend on
constitutive models

Affected by the
number of nodes
3
School of Computing, Engineering & Digital Technologies
 What is numerical modelling?

An example

4
School of Computing, Engineering & Digital Technologies
 Why numerical modelling is useful?

Realistically modelling of:

 Geometry
 Soil stratigraphy
 Soil behaviours (e.g., nonlinear, stress dependant)
 Boundary conditions
 Stage by stage constructions
 Soil-structure interaction
……

5
School of Computing, Engineering & Digital Technologies
Practical Considerations in FEM
Many slides are from
Prof. Harry Tan
(National University of Singapore)

6
 Contents
– Plane strain, Axi-symmetry, 3D model
– Model boundaries
• General considerations
• Excavations
• Shallow foundations
• Embankments
• Tunnels
– Meshing

7
 Plane strain
Considerations:
– One dimension is relatively long
– Similar geometry and stress or loading conditions
in any cross section  ‘long’ dimension

Consequences:
– No strain  ‘long’ dimension (stress can change!)
– No shear stress and arching  ‘long’ dimension
– Model represents 1 length unit  ‘long’ dimension

8
 Plane strain
Examples:

9
 Plane strain
NOT a plane-strain situation:

45 m
30 m
45 m
8m

10
 Axi-symmetry
Considerations:
– Geometry is circular
– Similar geometry and stress or loading conditions
in any cross section that includes the central axis

Consequences:
– Stress and strain  central axis are radial
– Model represents 1 radian around central axis

11
 Axi-symmetry
Examples:

NOT possible with gravity!

12
 Axi-symmetry
NOT an axi-symmetric situation:

Gravity!

13
 3D models
Considerations:
– Do I really need a 3D model?
– If I still use a 2D model:
• Would this give conservative or optimistic results?
• How large is the error?

Consequences:
– More difficult modelling and interpretation of
results
– Longer calculation times
– Generally less accurate results (due to coarser
meshes) 14
3D models

15
3D models

16
3D models

17
 Model boundaries
General considerations

– Type of analysis: Deformation, stability, dynamics,

… (flow).
– Type of behaviour: Drained or undrained.
– Is the situation (fully!) symmetric? Can we model
only half the problem?
– Boundaries should not influence results.
– Changes in stress and strain at boundaries should
be insignificant (except for symmetry boundaries).
– What is the consequence of taking boundaries
closer or further away?
18
 Model boundaries
Stability analysis:

– Mechanism must fit in model

– Only plastic deformation is relevant
– Stress state may not be disturbed by boundaries
(arching!)
– Model can generally be smaller than for
deformation analysis

19
 Model boundaries
Deformation analysis:
Deformations may still occur at a large distance from
the action, especially for undrained analysis
(conservation of volume!).
– Both elastic and plastic displacements are of
significant influence.
– Model should generally be larger than for stability
analysis

drained

undrained
 Model boundaries
Dynamic analysis:

– Vibrations may occur at very large distance from the action.

– Even very small displacements (vibrations) are of influence.
– Even if measures are taken to avoid spurious reflections at
boundaries, it is better to take boundaries far away
(considering wave speed and duration of analysis).
– Model should generally be larger than for deformation
analysis

~
~ ~ ~ ~
 Model boundaries
Stability analysis

Drained
deformation analysis

Undrained
deformation analysis
~
~ ~ ~ ~

Dynamic analysis
 Model boundaries –
Shallow
foundations
a w a



initial after loading

stress
a
distribution

w
(0.1 to 0.2) accepted

Suggestions: Stability analysis: a  2w

Drained deformation: a  3w
Undrained deformation: a  4w
 Model boundaries – Shallow
foundations
Take account of the following:

– For deformation analysis:

• When using Mohr-Coulomb, use different layers with
increasing stiffness; bottom layer with height w
should have large small-strain stiffness.
• When using Hardening-Soil, use bottom layer with
height w with large small-strain stiffness for E urref.
• Best results using HSsmall model.
– For horizontal loading components: Increased
width in loading direction.
Model boundaries – Embankments

a w a
h

Similarity with shallow footings

Suggestions: Stability analysis: a  2w

Drained deformation: a  3w
Undrained deformation: a  4w
Model boundaries – Embankments

Take account of the following:

– Embankments are considered to follow similar rules

as shallow foundations with the same base width w
 Model boundaries –
Excavations
a w a

l d

General Stability analysis: a  l and a  2d

suggestions: Structural forces analysis: a  2d (Vermeer & Wehnert, 2005)
Deformation analysis: a = 2d to 3d or 2w to 3w (max. of d and w)
(Meiβner, 2002)
Deformation analysis: a  2d ; bottom layer d with small-strain
stiffness for Eurref (Vermeer & Wehnert, 2005)
25
Model boundaries – Excavations

Settlement behind a sheetpile wall

0.2
0.1

relative displacement dv/dh

0
-0.1 0 0.5 1 1.5 2 2.5
-0.2
-0.3
-0.4
Clay
-0.5
Sand low stiff
-0.6 Sand medium stiff
-0.7 Sand Higher friction
-0.8
relative distance x/h

Motivation: Settlement trough behind the wall from FE analyses

(K.J. Bakker, PAO course Damwanden) 24
Model boundaries – Excavations
Take account of the following:

1. Suggested model depth requires that large small-strain

stiffness is used below the excavation.
2. When using Hardening-Soil, use bottom layer with height d
with large small-strain stiffness for Eur .
ref

3. Using a ‘normal’ stiffness for Eurref will result in unrealistic

heave of excavation bottom (and wall); a lower model depth
should then be considered (but 2 is preferred).
4. Ignoring small-strain stiffness will generally result in a too
wide settlement trough behind the wall, regardless the model
width.
5. For a < 3d significant settlements may be expected at upper
model corners. This is even more pronounced for undrained
behaviour.
Model boundaries – Excavations

Considering the wall:

1. Unrealistic heave of excavation bottom gives unrealistic

heave of wall > use large stiffness below excavation
2. For a < 2d vertical model boundaries influence wall
displacements
3. Model depth and width seem to have little influence on
the wall forces (bending moments)
(Vermeer & Wehnert, 2005)
 Model boundaries – Tunnels
w D w w w

D D

a a TBM or
NATM excavation

Suggestions: Face stability: a  ½D ; w  2D (Ruse, 2003)

Structural forces, probably w  2D
Drained deformation: w  3D
Undrained deformation: w  4D
Model boundaries – Tunnels
Take account of the following:

1. Large unloading and small strain stiffness below the tunnel

2. Suggested model depth requires that large small-strain
stiffness is used below the tunnel (MC: different sub-layers;
HS: large Eurref).
3. Ignoring small-strain stiffness will result in unrealistic heave of
tunnel; a lower model depth should then be considered (but 2
is preferred).
4. Ignoring small-strain stiffness will generally result in a too
wide settlement trough above the tunnel, regardless the
model width.
5. For w < 3D significant settlements may be expected at the
upper model corners. This is even more pronounced for
undrained behaviour.
6. For deep tunnels the overburden may be modelled as load,
provided that at least a height w above the tunnel is included
in the model. 29
 Meshing
Type of element:
 Meshing
Type of element:

• Two types of volume elements are available

in Plaxis V21: node (u , u ) x y

stress point (, )

x
x
x x x
y-axis
x x xx
x x x x
x x
x-axis

6-node triangle 15-node triangle

The 15-node triangle is the default element. It provides a fourth order interpolation for displacements
and the numerical integration involves twelve Gauss points (stress points).
 Meshing
Type of element:

• 3D element (only one type): 3

12
1
9
8
6
 1
4 7
10
2
15 2
 3

14

5
4 6 11

13
5
32
(quadratic interpolation)
 Meshing
General considerations:

• Fine meshes required near loads and structures (also at

tunnel faces; Ruse, 2003).
• Coarser meshes may be used towards the model
boundaries.
• Better to use larger models with relatively large elements
(coarse mesh) near the boundary than to use smaller
models.

Suggestion:

• Use local element size factors to make meshes fine near

loads and structures and coarse at model boundaries (local
element size factor may be larger than 1.0!).
 Meshing

Using local refinement:

37
Structural elements in Plaxis 2D
and soil-structure interaction

38
 PLAXIS 2D Tutorial 06: Dry excavation using a tie back
wall
• Plate – Diaphragm wall
• Node to node anchor – used to model anchor
• Interface – used to model soil-structure interaction
• Line load – used to model surcharge
• Stage construction – used to model stage by stage
construction procedures

https://communities.bentley.com/products/geotech-analysis/w/wiki/45558/plaxis-2d-tutorial-06-dry-excavation-
using-a-tie-back-wall
39
PLAXIS 2D Tutorial 04: Settlements due to tunnel
construction
• Plate – building foundation slab and tunnel lining
• embedded beam rows – pile foundation
• Interface – used to model soil-structure interaction

https://communities.bentley.com/products/geotech-analysis/w/wiki/45560/plaxis-2d-tutorial-04-settlements-due-to-
tunnel-construction 40
 Plates and shells

• 3 or 5 noded line elements

• 3 degrees of freedom per node
• Elastic or elastoplastic behaviour
• To model walls, floors, tunnels
Input parameters for plates
EI  E  h  b
3

• Flexural rigidity 12 (b=1 m)

• Normal stiffness EA  E  h  b (b=1 m)

• Element thickness d  h  12 EA
EI

h h
b

b = 1 m in plane strain
b = 1 meter in axisymmetry
b
 Plate weights
• Compensate for overlap:
w  ( concrete   soil )  dreal
• For soil weight use:
γunsat above phreatic level
γsat below phreatic level
 Plate weights for tunnels
dreal
rinside r
routside
lining soil

• Overlap is only for half the lining thickness


w   concrete  dreal  soil  1 2 dreal 
r  1 2 rinside  routside 
Other structure/wall types

Sheet pile

45
Other structure/wall types

Contiguous pile wall

46
Gaba, A., Hardy, S., Doughty, L., Powrie, W., & Selemetas, D. (2017). Guidance on embedded retaining wall design.
London, UK: Ciria. Page 85. 47
Gaba, A., Hardy, S., Doughty, L., Powrie, W., & Selemetas, D. (2017). Guidance on embedded retaining wall design.
London, UK: Ciria. Page 46. 48
Gaba, A., Hardy, S., Doughty, L., Powrie, W., & Selemetas, D. (2017). Guidance on embedded retaining wall design.
London, UK: Ciria. Page 87. 49
 Interfaces
• Soil-structure interaction
– Wall friction
– Slip and gapping between soil and structure
• Soil material properties
– Taken from soil using reduction factor Rinter
Cinter = Rinter * Csoil
tan()inter = Rinter * tan()soil
– Individual material set for interface
 Interfaces
Suggestions for Rinter:
– Interaction sand/steel = Rinter ≈ 0.6 – 0.7
– Interaction clay/steel = Rinter ≈ 0.5
– Interaction sand/concrete = Rinter ≈ 1.0 – 0.8
– Interaction clay/concrete = Rinter ≈ 1.0 – 0.7
– Interaction soil/geogrid = Rinter≈ 1.0
(interface may not be required)
– Interaction soil/geotextile = Rinter≈ 0.9 – 0.5 (foil,
textile)
 Interfaces
• Try to omit stress oscillations at
corners of stiff structures

Inflexible corner
points, causing bad
stress results

Flexible corner
points with
improved stress
results
 Boundary Conditions
Force BCs

Force BCs Displacements BCs

Displacements BCs

Mixed BCs
Plaxis analysis settings

54
Initial stress Calculation type at Initial Phase in Plaxis:

55
Calculation type
Short/long-term

Time dependant

Factor of safety for ULS

56
Pore pressure calculation type

Phreatic: u = γwz
Steady state groundwater flow

57
Other settings

Practice with a few examples

in the Plaxis tutorial to gain
deeper understanding.

58
Stage by Stage Constuction

59
Top-down construction

60
Top-down construction
Top-down is defined by the use of the permanent internal structure as the temporary propping to the
retaining wall, cast in a top-down sequence. The higher-level slabs are cast before the lower-level
slabs to act as horizontal frames for wall support as the excavation progresses.

Openings would be needed. How to consider this in numerical modelling?

61
Bottom-up construction

62
Bottom-up construction

A bottom-up construction sequence is defined by the construction of the

permanent works from the lowest level upwards, casting the foundation slab
before the internal walls and slabs above.

63
A design example
The total depth of the basement excavation is 16
m and the construction sequence is fully top-down
with support provided to the wall by three levels of
permanent slab (denoted P1, P2 and P3). For
each excavation stage, formation level will be 0.5
m below the soffit of the permanent slab.

In the permanent case, a 1.0 m thick basement

slab will also support the wall. The basement is 80
m long and 40 m wide.

The stratigraphy at the site comprises 4.0 m of Top-down construction?

made ground, overlying 7.0 m of gravel which in
turn overlies a clay stratum with a thickness of 50 Bottom-up construction?
m. Chalk underlies the clay.

64
Advantage of numerical analyses:
• Soil-structure interaction
• Stage construction can be simulated
• Considering variable boundary conditions
and also variable ground conditions.

65
But it can be very complicated. A good modelling
requires:

1) A good understanding of the problem

• Soil mechanics/geotechnics, etc.
• Material properties, boundary conditions, etc.

2) A good understanding of the numerical tool

• constitutive models
• assumptions and limitations

3) A good understanding on what is the most important

and what can be simplified.

Useful reference:
O’Brien, A. S., Higgs, A. G. (2020). The management of
advanced numerical modelling in geotechnical modelling
(CIRIA guide C791)
66
Please watch this video (compulsory)
https://www.issmge.org/education/recorded-webinars/an-introduction-to-numerical-modelling-in-geotechnical-engineering

67
 Summary
– Plane strain, Axi-symmetry, 3D model
– Model boundaries
– Meshing
– Structural elements in Plaxis 2D
– Plaxis analysis settings
– Stage construction

68
 Thank you
Contact Details:
Dr Wengui Huang
Email: [email protected];
Office: IC1.41 Stephenson Building (appointment first)
Academic support hub (IC1.43a): Every Thursday from 1pm to 2pm
You are very welcome to attend my weekly Academic Support Hub!

69
School of Computing, Engineering & Digital Technologies 69