Rough Strip Footing on a Cohesive Frictionless Material 3 - 1

3 Rough Strip Footing on a Cohesive Frictionless Material

3.1 Problem Statement
The prediction of collapse loads under steady plastic-owconditions can be difcult for a numerical
model to simulate accurately (Sloan and Randolph 1982). As a two-dimensional example of a
steady-ow problem, we consider the determination of the bearing capacity of a strip footing on a
cohesive frictionless material (Tresca model). The value of the bearing capacity is obtained when
steady plastic ow has developed underneath the footing, providing a measure of the ability of the
code to model this condition.
The strip footing is located on an elasto-plastic material with the following properties:
shear modulus (G) 0.1 GPa
bulk modulus (K) 0.2 GPa
cohesion (c) 0.1 MPa
friction angle () 0

dilation angle () 0

3.2 Closed-Form Solution

The bearing capacity obtained as part of the solution to the Prandtls wedge problem is given by
Terzaghi and Peck (1967) as
q = (2 +)c (3.1)
in which q is the average footing pressure at failure, and c is the cohesion of the material. The
corresponding failure mechanism is illustrated in Figure 3.1:
FLAC
3D
Version 5.01
3 - 2 Verication Problems
Collapse Load
q=(2 + )c
Figure 3.1 Prandtl mechanism for a strip footing
3.3 FLAC
3D
Model
For this problem, half-symmetry and plane-strain conditions are assumed in the numerical simula-
tion. The domain used for the analysis is sketched in Figure 3.2, together with its dimensions:
X
y
z
rough footing
1
m
1
0
m
20 m
a
Figure 3.2 Domain for FLAC
3D
simulation half symmetry
FLAC
3D
Version 5.01
Rough Strip Footing on a Cohesive Frictionless Material 3 - 3
A system of coordinate axes is selected as indicated on the gure. The area representing the strip
footing has a half-width a, the far x-boundary is at a distance of 20 m from the y-axis of symmetry,
and the far z-boundary is located 10 m below the footing. The thickness of the domain is selected
as 1 m.
The boundary conditions applied to this domain are sketched in Figure 3.3. The displacement of the
rough footing is restricted in the y-direction, and a velocity is applied to the model in the negative
z-direction to simulate the footing load. In the data le prandtl.f3dat, the rightmost gridpoint of
the footing is free in the x-direction. This condition can be justied because the physical constraint
exactly at the edge is ambiguous, and can be chosen arbitrarily. Releasing the constraint leads to a
more uniform stress distribution under the footing, but does not affect the limit load.
Figure 3.3 Boundary conditions for FLAC
3D
analysis half symmetry
The domain is discretized into one layer of 520 zones organized in a graded pattern, as represented
in Figure 3.4, with the grading arranged to increase denition (zone limits) in the areas of high
strain gradient. The area representing the footing overlaps six zones, and a velocity of magnitude
0.5 10
5
m/step is applied at the contact nodes for a total of 25,000 calculation steps.
When a velocity is applied to gridpoints to simulate a footing load, the average footing pressure is
found by assuming that the footing width is represented by a velocity that varies from the value at
the last gridpoint to zero at the next gridpoint. The half-width, a, is then
a = A(x
l
+x
l+1
) (3.2)
FLAC
3D
Version 5.01
3 - 4 Verication Problems
where x
l
is the x-location of the last applied gridpoint velocity, x
l+1
is the x-location of the gridpoint
adjacent to x
l
, and A accounts for the variation. If a linear variation is assumed, then A = 0.5, and
a = 3.5 m for this problem. The effect of this assumption is discussed in Section 3.4.
The data le prandtl.f3dat, usedtogenerate the numerical solution, is presentedinSection3.6. The
FISH function p load computes the numerical value of the normalized average footing pressure,
p/c, and the corresponding relative error.
The FLAC
3D
model takes approximately 17 seconds to run 25,000 steps on a 2.7 GHz Intel i7
computer.
IIAc1u 3.66
,rp, ,WDVFD mRQVXOWLQJ EURXSl ,QFg
6WHS ,errr
8G,/G,rp, ,:S.:,. U0
=RQH
mRORUE\: 8QLIRUP
=21O
,WDVFD mRQVXOWLQJ EURXSl ,QFg
0LQQHDSROLVl 01 86U
Figure 3.4 FLAC
3D
grid vertical plane view
3.4 Results and Discussion
The load-displacement curve corresponding to the numerical simulation is presented in Figure 3.5,
in which p load is the normalized average footing pressure, p/c, and c disp is the magnitude of
the normalized vertical displacement, u
z
/a, at the center of the footing. The numerical value of the
bearing capacity, q, is 523.0 kPa, and the relative error is 1.73% when compared to the analytical
value of 514.2 kPa.
The apparent width of the footing is taken to be 3 m, plus half the zone width adjacent to the footing
edge (since forces are exerted on the footing by this zone, we assume that the forces are divided
equally between left and right gridpoints).
FLAC
3D
Version 5.01
Rough Strip Footing on a Cohesive Frictionless Material 3 - 5
Note that the error in the bearing capacity is related to the indeterminacy in the apparent width
of the footing. The mechanism illustrated in Figure 3.1 implies a velocity singularity at the ends
of the footing. In a numerical simulation, this singularity is spread over the width of one zone.
The apparent position of the velocity jump within that zone depends on the exact geometry of the
velocity eld that develops. In deriving q, it is assumed that the jump occurs half a zone width from
the end of the controlled boundary segment (A = 0.5 in Eq. (3.2)); note that if a variation factor
of A = 0.63 is assumed, the error reduces to less than 0.1%. For ner grids, the indeterminacy in
footing width decreases, and the match to the exact solution improves.
0.20 0.40 0.60 0.80 1.00 1.20
0.:0
1.00
1.:0
2.00
2.:0
1.00
1.:0
4.00
4.:0
:.00
/RDG [tnAct
'
L
V
S
O
D
F
H
P
H
Q
W
IIAc1u 3.66
,rp, ,WDVFD HRQVXOWLQJ UURXSl ,QFg
6WHS ,errr
8G,/G,rp, ,:S.:,. (0
+LVWRU\
p SBORDG CN,6+o
, SBVRO CN,6+o
YVg . FBGLVS CN,6+o
,WDVFD HRQVXOWLQJ UURXSl ,QFg
0LQQHDSROLVl 01 86(
Figure 3.5 Load-displacement curve
Velocity contours and velocity vectors at the end of the run are presented in Figure 3.6, showing
good agreement with the mechanism in Figure 3.1.
FLAC
3D
Version 5.01
3 - 6 Verication Problems
IIAc1u 3.66
,rp, ,WDVFD 6RQVXOWLQJ UURXSl ,QFg
6WHS ,errr
8G,/G,rp, ,:S.:,. 70
9HORFLW\
0D[LPXP: eg99preHir5
6FDOH: p8.8,e
&RQWRXU2I9HORFLW\
eg99pr+ir5
egerrr+ir5
egrrrr+ir5
Sgerrr+ir5
Sgrrrr+ir5
.gerrr+ir5
.grrrr+ir5
,gerrr+ir5
,grrrr+ir5
pgerrr+ir5
pgrrrr+ir5
egrrrr+ir9
rgrrrr+urr
,WDVFD 6RQVXOWLQJ UURXSl ,QFg
0LQQHDSROLVl 01 867
Figure 3.6 Velocity eld at collapse load
The same problem was run again using FLAC
3D
s nodal mixed discretization (NMD) feature (le
prandtl-nmd.f3dat). For this model, an all-tet grid (le prandtl nmd.Flac3D, which is imported
into FLAC
3D
not listed here) was used. The tet grid has gridpoints identical to the hex grid. With
NMD, the numerical value of the bearing capacity, q, is 534.2 kPa, and the relative error is 3.9%
when compared to the analytical value of 514.2 kPa. As stated previously, the error in the bearing
capacity is related to the indeterminacy in the apparent width of the footing. Figures 3.7 and 3.8
show the NMD results, and correspond directly with Figures 3.5 and 3.6 (results with an all-hex
grid). As can be seen, the results are very similar.
FLAC
3D
Version 5.01
Rough Strip Footing on a Cohesive Frictionless Material 3 - 7
0.20 0.40 0.60 0.80 1.00 1.20
1.00
1.:0
2.00
2.:0
1.00
1.:0
4.00
4.:0
:.00
/RDG [tnAct
'
L
V
S
O
D
F
H
P
H
Q
W
IIAc1u 3.66
,rp, ,WDVFD HRQVXOWLQJ UURXSl ,QFg
6WHS ,errr
8G,/G,rp, ,:Se:S. (0
+LVWRU\
p SBORDG CN,6+o
, SBVRO CN,6+o
YVg . FBGLVS CN,6+o
,WDVFD HRQVXOWLQJ UURXSl ,QFg
0LQQHDSROLVl 01 86(
Figure 3.7 Load-displacement curve (NMD results)
IIAc1u 3.66
,rp, ,WDVFD 9RQVXOWLQJ UURXSl ,QFg
6WHS ,errr
prG,eG,rp, ,3SS3.. 30
9HORFLW\
0D[LPXP3 eg.r/:eHir5
6FDOH3 ,pr:S:
&RQWRXU2I9HORFLW\
eg.r/4+ir5
egrrrr+ir5
Sgerrr+ir5
Sgrrrr+ir5
.gerrr+ir5
.grrrr+ir5
,gerrr+ir5
,grrrr+ir5
pgerrr+ir5
pgrrrr+ir5
egrrrr+ir/
rgrrrr+urr
,WDVFD 9RQVXOWLQJ UURXSl ,QFg
0LQQHDSROLVl 01 86-
Figure 3.8 Velocity eld at collapse load (NMD results)
FLAC
3D
Version 5.01
3 - 8 Verication Problems
3.5 Reference
Sloan, S. W., and M. F. Randolph. Numerical Prediction of Collapse Loads Using Finite Element
Methods, Int. J. Num. & Analy. Methods in Geomech., 6, 47-76 (1982).
Terzaghi, K., and R. B. Peck. Soil Mechanics in Engineering Practice, 2nd Ed. New York: John
Wiley and Sons (1967).
FLAC
3D
Version 5.01
Rough Strip Footing on a Cohesive Frictionless Material 3 - 9
3.6 Listing of Data Files
The project le for this problem is located in the folder mechanical\prandtl.
Example 3.1 PRANDTL.F3DAT
new
set fish autocreate off
set fish safe on
set deterministic off
;---------------------------------------------------------------------
; 2D rough strip footing on
; Tresca material (Prandtls wedge problem)
; -associated plastic flow-
;---------------------------------------------------------------------
gen zone brick size 6 1 20 p1=3.0,0.0,0.0 ratio 0.9 1.0 0.97
gen zone brick size 20 1 20 p0=3.0,0.0,0.0 p1=20.0,0.0,0.0 ...
ratio 1.08 1.0 0.97
ini zp mul 0.5
model mech mohr
prop bul 2.e8 shea 1.e8 cohesion 1.e5
prop friction 0. dilation 0. tension 1.e10
fix x range x -.1 .1
fix x y z range z -.1 .1
fix x y z range x 19.9 20.1
fix y
fix x y z range x -.1 3.1 z 9.9 10.1
free x range x 2.9 3.1 z 9.9 10.1
ini zvel -0.5e-5 range x -.1 3.1 z 9.9 10.1
def p_cons
array LoadPoints(50) ; Must increase if grid zones are increased
global pdis1 = gp_near(0.,0.,10.)
global pdis2 = gp_near(0.,1.,10.)
global p_sol = (2. + pi)
global setCohes = z_prop(zone_head,cohesion)
local pnt = gp_head
local n = 0
local xnext = 20.0
loop while pnt # null
if gp_zpos(pnt) > 9.9
if gp_xpos(pnt) < 3.1
n = n + 1
LoadPoints(n) = pnt ; ... save addresses
endif
if gp_xpos(pnt) > 3.1 ; ... get 1st gp outside footing
xnext = min(xnext,gp_xpos(pnt))
FLAC
3D
Version 5.01
3 - 10 Verication Problems
endif
endif
pnt = gp_next(pnt)
endLoop
global nLoads = n
global EffectiveWidth = (3.0 + xnext) / 2.0
end
@p_cons
;---------------------------------------------------------------------
; p_load : average footing pressure / c
; c_disp : magnitude of vertical displacement at footing center / a
;---------------------------------------------------------------------
def p_load
local pload = 0.0
local n
loop n (1,nLoads)
local pnt = LoadPoints(n)
pload = pload + gp_zfunbal(pnt)
endLoop
global actLoad = pload / (EffectiveWidth * setCohes)
global c_disp = -(gp_zdisp(pdis1) + gp_zdisp(pdis2)) / 2.0
p_load = actLoad
end
def p_err
p_err = 100. * (actLoad - p_sol) / p_sol
end
hist nstep 50
hist add fish @p_load
hist add fish @p_sol
hist add fish @c_disp
hist add unbal
cyc 25000
list @p_err
save pran3
return
FLAC
3D
Version 5.01
Rough Strip Footing on a Cohesive Frictionless Material 3 - 11
Example 3.2 PRANDTL-NMD.F3DAT
new
set fish autocreate off
set fish safe on
set deterministic off
;---------------------------------------------------------------------
; 2D rough strip footing on
; Tresca material (Prandtls wedge problem)
; -associated plastic flow-
; Nodal Mixed Discretization
;---------------------------------------------------------------------
impgrid prandtl_nmd
set nmd on
model mech mohr overlay 1
prop bul 2.e8 shea 1.e8 cohesion 1.e5
prop friction 0. dilation 0. tension 1.e10
fix x range x -.1 .1
fix x y z range z -.1 .1
fix x y z range x 19.9 20.1
fix y
fix x y z range x -.1 3.1 z 9.9 10.1
free x range x 2.9 3.1 z 9.9 10.1
ini zvel -0.5e-5 range x -.1 3.1 z 9.9 10.1
def p_cons
array LoadPoints(50) ; Must increase if grid zones are increased
global pdis1 = gp_near(0.,0.,10.)
global pdis2 = gp_near(0.,1.,10.)
global p_sol = (2. + pi)
global setCohes = z_prop(zone_head,cohesion)
local pnt = gp_head
local n = 0
local xnext = 20.0
loop while pnt # null
if gp_zpos(pnt) > 9.9
if gp_xpos(pnt) < 3.1
n = n + 1
LoadPoints(n) = pnt ; ... save addresses
endif
if gp_xpos(pnt) > 3.1 ; ... get 1st gp outside footing
xnext = min(xnext,gp_xpos(pnt))
endif
endif
pnt = gp_next(pnt)
FLAC
3D
Version 5.01
3 - 12 Verication Problems
end_loop
global nLoads = n
global EffectiveWidth = (3.0 + xnext) / 2.0
end
@p_cons
;---------------------------------------------------------------------
; p_load : average footing pressure / c
; c_disp : magnitude of vertical displacement at footing center / a
;---------------------------------------------------------------------
def p_load
local pload = 0.0
local n
loop n (1,nLoads)
local pnt = LoadPoints(n)
pload = pload + gp_zfunbal(pnt)
end_loop
global actLoad = pload / (EffectiveWidth * setCohes)
global c_disp = -(gp_zdisp(pdis1) + gp_zdisp(pdis2)) / 2.0
p_load = actLoad
end
def p_err
p_err = 100. * (actLoad - p_sol) / p_sol
end
hist n 50
hist add fish @p_load
hist add fish @p_sol
hist add fish @c_disp
hist add unbal
; plo set pers off
cyc 25000
list @p_err
save pran-nmd
return
FLAC
3D
Version 5.01