Professional Documents
Culture Documents
Four approaches to
determine masonry
strength domain
Gabriele Milani MEng, PhD
Associate Professor, Department of Architecture, Built Environment and
Construction Engineering A.B.C., Technical University of Milan, Milan, Italy
Four models to determine the homogenised strength domain of running bond masonry in-plane loaded are
compared. The first is a lower bound, where the elementary cell is subdivided into a few rectangular sub-domains
and the micro-stress field is expanded using polynomial expressions. The second is again a lower bound, where joints
are reduced to interfaces and bricks are subdivided into constant stress triangular elements. The third procedure is a
compatible identification (kinematic approach), where joints are reduced to interfaces and bricks are assumed
infinitely resistant. The last model is again a kinematic procedure based on the so-called method of cells. The
representative element of volume is subdivided into six rectangular sub-cells with pre-assigned polynomial fields of
periodic velocity. The first and latter models have the advantage that the reduction of joints to interfaces is not
required. The second approach, albeit reduces joints to interfaces, still allows considering failure inside bricks.
The third approach is the most straightforward, but is reliable only for thin joints and strong blocks. Some illustrative
examples regarding the determination of masonry homogenised strength domains are discussed, focusing on
pros and cons of the models, role played by joint thickness, constituent materials failure surfaces and numerical
efficiency.
1.
Introduction
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
The major limitation of homogenisation is related to nonlinear finite-element (FE) computations, because a continuous
interaction between meso- and macro-scale is needed in the
non-linear range. The practical translation of this issue is a
huge computational effort, since the field problem has to be
solved numerically for each loading step, in all Gauss points.
For the above reasons, limit analysis combined with homogenisation technique still seems to be one of the most powerful
and direct structural analysis methods to provide reliable
and fast results at collapse. Such an approach requires only
a reduced number of material parameters and allows the
avoidance of independent modelling of units and mortar. In
addition, it provides limit multipliers of loads, failure mechanisms and, at least on critical sections, the stress distribution
at collapse. It belongs to the so-called direct methods,
because the results are obtained in a single step solving a suitable optimisation problem. Different homogenisation models
have been recently proposed in the technical literature for the
evaluation of homogenised strength domains for masonry
walls in-plane loaded (de Buhan and de Felice, 1997; Milani
et al., 2006a, 2006b).
The paper is aimed at critically reviewing some of the most
effective models already available, with a comparison of their
numerical performance within some applications of technical
interest.
In particular, four different strategies for the evaluation of the
homogenised strength domain of running bond masonry inplane loaded are discussed and critically compared. Under the
assumption of mortar and bricks obeying a rigid plastic behaviour with associated flow rule, and within basic assumptions of
the homogenisation theory (such as periodicity and antiperiodicity of the local velocity field and micro-stress, respectively), an estimation of masonry macroscopic strength domain
is possible by means of the application of both the classic
upper and lower bound theorems of limit analysis on the
chosen elementary cell (Suquet, 1983). In particular, the lower
bound approach requires the imposition of micro-stress equilibrium, admissibility as well as anti-periodicity and allows
obtaining lower bound estimates of the actual homogenised
failure surface by means of the constrained maximisation of
the homogenised internal actions. The upper bound approach,
dually, requires dealing with kinematically admissible velocity
fields (i.e. obeying associated flow rules), with periodicity conditions applied at the boundary of the elementary cell, allowing upper bound estimates of the actual homogenised failure
surface by means of the constrained minimisation of the total
2.
100
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
1
dY
A Y
1
dY
, .
A Y
E , .
1:
3.
where u
u Ey uper ; uper on @ Y
n; anti periodic on @Y
per
3:
8 8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
< >
<
f hom j
>
>
>
>
>
> >
>
>
>
>
>
>
>
>
>
>
>
>
>
>
: >
:
, .
1
A
9
>
a >
>
>
>
>
>
>
>
>
=
b >
dY
Y
div 0
nint 0
n;
y [ f m ;
anti-periodic
on
8y [ Y m y [ f b ;
@Y
8y [ Y b
c >
>
>
>
>
>
>
d >
>
>
>
;
e
Here, is the jump of micro-stresses across any discontinuity surface of normal n int. Conditions 3a and 3d are derived
from periodicity, condition 3b imposes the micro-equilibrium
and condition 3e represents the yield criteria for the components (brick and mortar).
A dual kinematic definition of f hom, also due to Suquet
(1983), can be derived through the support function hom(D)
as follows
9
8
>
>
: D hom D 8D
>
>
>
>
>
>
>
>
=
<
1
f hom j hom D inf PvjD
v n n vdS
>
>
>
v
2 @Y
>
>
>
>
>
>
>
>
>
>
>
;
: >
:
Pv Y d dY S v; ndS
8
>
>
>
>
>
<
4:
In Milani et al. (2006a, 2006b), a model based on the assumption of equilibrated polynomial stress fields inside the unit cell
is presented. Essentially, the approach relies on a rough subdivision of the unit cell by means of rectangular sub-domains,
where a polynomial interpolation of the stress field is a priori
given. Equilibrium inside each element and at the interface
between contiguous elements, as well as anti-periodicity conditions are imposed. In this way, the micro-stress field is antiperiodic and auto-equilibrated, fully complying constraints
3bd.
As shown in Figure 1, one-fourth of the REV is subdivided into nine geometrical elementary entities (subdomains), so that the whole cell is sub-divided into 36
sub-domains.
For each sub-domain, polynomial distributions of degree (m)
are a priori assumed for the stress components. The generic
ijth component can be written as follows
5:
ij X yS Tij ;
y [ Yk
where X(y) = [1 y1 y2 y21 y1 y2 y22]; Sij = [Sij (1) Sij (2) Sij (3) Sij (4)
m 1 m 2=2
Sij (5) Sij (6) ] is a vector of length N
representing the unknown stress parameters; and Yk represents
the kth sub-domain.
The imposition of equilibrium with zero body forces inside
every sub-domain, the continuity of the stress vector on interfaces and anti-periodicity of n allows for a strong reduction of
the total number of independent stress parameters.
In particular, equilibrium has to be imposed everywhere inside
each sub-domain, that is, ij,j (x, y) = 0, i = 1, 2 8(x, y) [ subdomain. Since ij (x, y) is a polynomial expression of degree
(m), a linear combination of its derivatives (div ) is a polynomial of degree (m 1). This leads to write (2N) linear
independent equations in the stress coefficients, where N =
((m 1)2/2) + (3(m 1)/2) + 1 = (m(m + 1)/2).
A further reduction of the total unknowns is obtained a
priori imposing the continuity of the stress vector on internal
101
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
1/4 of
elementary cell
eh
y
ev
interfaces (ij(k)nj + ij(r)nj = 0; i = 1, 2) for every (k) and (r) contiguous sub-domains with a common interface of normal
n (Figure 2(a)). Being n polynomial expressions of degree (m)
in the abscissa s of the interface, other 2N equations (where
k
r
N = m + 1) in S
and S for each kr interface can be
written.
Finally, anti-periodicity of n on V requires 2N additional
equations per pair of external faces (m)(n) (Figure 2(b)), that
is, it should be imposed that stress vectors n are opposite on
opposite sides of V
6:
m
n
m
n
X ij yS n1 X ij yS n2
where n (m) and n (n) are oriented vectors of the external faces of
the paired sub-domains (m)(n).
Again n on V is a polynomial expression of degree (m) in
the abscissa s of the external edge of the sub-domain.
(kr) Interface
y2
Elementary cell
(m)n1
n2
(r)
n1
(k)
y1
(q)
(n)n2
2
4 3
3
(qk) Interface
(a)
(b)
(c)
102
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
8:
7:
k 1; . . . ; kmax
k yS;
k X
maxfg
8
1X
>
k yS dY
>
n
X
>
>
Y
>
Y
k
>
>
>
>
>
j
>
y ; nodal point
>
>
>
>
>
>
k yj S
>
j X
>
>
>
>
<
j [ f k ; j 1; . . . ; rq; k 1; . . . ; 4k max
>
>
1
>
>
>
n;1 cos 1 cos2 sin 1 cos2
>
>
2
>
>
>
>
>
1
>
>
n;2 cos 1 cos2 sin 1 cos2
>
>
>
2
>
>
>
>
>
: n 1 cos cos2 sin cos2 tan2
;3
2
4.
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
Elementary cell
a/2
(6)
(6)
(12)
(4)
(5)
(5)
(3)
(4)
(2)
(1)
(2) (1)
(3)
(9)
(11) (10)
Anti-periodicity of micro-stress
(8)
(1)
(2)
xx
(2)n
(1)
of elementary cell
y2
(3)
(8) (7)
(9)
b
1
4
P1
(7)
P2
(1)
(2)
(2)
(2)
(3)
(1)
(2)
n
n
(1)
yy
(2)
yy
Brick interface
(1)
(1)
xx
(1)
xx
(1)
Mortar interface
9:
1
1
2
1 2
xx xx m
1
1
2
2
yy yy m
10:
6
1
xx xx
1 6
104
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
11:
max
s:t:
8
P24 i
>
>
i1 xx Ai
>
>
>
>
2ab
>
>
>
>
P24 i
>
>
>
yy Ai
>
>
i1
>
>
2ab
>
>
>
>
P24 i
>
>
>
Ai
>
>
< i1
2ab
>
> AI X bI
>
>
eq
eq
>
>
>
> ap
>
Aeq X bap
>
eq
>
>
>
>
>
>
i
i
>
>
f iE xx ; yy ; i 0; i 1; . . . ; 24
>
>
>
>
>
>
>
i
>
: f iI i
0; i 1; . . . ; 32
I ; I
&
&
&
&
&
&
5.
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
Interfaces/thin joints
xy
Bricks/thick joints
xy
yy
xx
: Friction angle
2 : Compression linearised cap
2
c
fc : Compression strength
ft : Tensile strength
c : Cohesion
ft
fc
ft
fc
(a)
xx
(b)
ga
B
p
I
gb
ga
E11
E12 + E21
(a)
(b)
106
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
E22
(c)
P5 = [0 0 c]
P5 = [c/mf(1/f-f ) 0 0] P6 = [0 0 c]
P4 = [c/f-cf 0 0]
P4 = [c/mf(m-1/m) 0 c/mf]
01
xy: MPa
005
P1 = [0 c/f 0]
0
005
P3 = [c/f+c/m 0 0]
P2 = [c/f c/f 0]
yy: MPa
03
0
0
P1 = [0 c/f 0]
P3 = [c/f+c/m 0 0]
005
yy [MPa]
02
01
015
005
01
01
0
005
xy: MPa
01
xx: MPa
005
01
P2 = [c/f c/f 0] 01
015
xx: MPa
04
(a)
(b)
14:
tA vA tB vB dS
When two contiguous bricks A and B are considered, the velocity of a generic point P in a position [ I belonging, respectively, to A and B (where I indicates the common interface
between the two bricks) is
12:
A
A
vA vC M A
zz C
B
vB vC M Bzz CB
bv c vB vA
A
B
A
M Bzz
vC vC M A
zz C
CB
tA v dS
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
15:
w 1 ; 2
nlin
X
i1
@f
I
i 1 ; 2 I
@
16:
nlin
Pk Pk X
@f I
I
w 1 ; 2
;
i P1 k ; P2 k
@
i1
k 1; 2; 3
I
where i P1 k ; P2 k is the ith plastic multiplier rate of the inter
face I corresponding to Pk P1 k ; P2 k .
Internal power dissipated on the Ith interface, defined as the
product of the interface stress vector for the jump of velocities,
is evaluated by means of the following equation
17:
Iint
AI
wT dAI
nlin
X
AI i1
T
@
I
dAI
i 1 ; 2
@
nlin
3
X
1X
I
cIi
Pk ; Pk AI
3 i1 k1 i 1 2
w 1 ; 2 G I 1 ; 2 D
>
>
>
>
I
>
: G Pk D
nI
P
Iint T0 D
xD;
i Pk I 1
T1 D 1
nlin I
@
P
;
wPk
i P1 k ; P2 k
@
i1
min
I
Pk [ I
6.
108
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
relations hold
2
x2
1
u2 2W 1
hb
bb
U 2 U 1 x1
2
U1
;
bm
u1 2U 1
2
u1
20:
x1
;
bb
1
bm bb
2W 1 W 2
x1
C
x2 B
2
2 B
W 2C
@
A
bm
hb
0
u2
hb
hb
U 1 1 2b U 2
W 1 W 3 x2
x2
2
2
3
1
3
u1 u1
; u2 W 1
2hm
hm
hb
U 1 1 2b U 2
x2
2
4
1
4
3
u1 u1
; u2 u2
2hm
bb bm
hb
bb
x2
U 1 1 2b U 2
U 1 U 2 x1
x1
2
2
2
5
u1 U 1
bm hm
bm
hb
bb bm
bm
hb
W
W
x
W
x
h
2
2
1
1
m
2
x2
2
2
2
5
2 2
u2 W 3
bm hm
hm
1
0
U1 U2
hb
x2
U1
x1 B
2b
2 C
6
C
u1 2 B
U1
A
bb @
hm
W 2 W 3 2W 1 W 2
u2 W 1
jx1 j
bm
hb
x2
2
hm
u1
t3
u1
21:
x2
x bb =2
t1
t2
t1
t2
; u2 0; u1 u1 ; u2 W t1 1
hb
bm
t
t
x
b
=2
U
U
h
2
b
b
t3
1
U t1 2
x3
; u2 W t2
hm
hm
2
2U t1
t4
u1 ; u2
t5
u1 ; u2
t6
u1 ; u2
u1
u1
u1
t3
t4
t3
t5
t3
t6
t3
u2
x1 bb bm =2x2 hb =2 hm x1 bb =2
bm hm
x
x
h
=2
1
2
b
W t1
bm hm
W t1
109
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
bm
5
hm
x2
hb
x1
2
bb
bm
i
i
pl @v1 =@y1 @v2 =@y2 @v1 =@y2 @v2 =@y1 @f b;m =@,
i
i
where pl is the plastic strain rate field in the (i)th sub-cell;
(0) is the rate of the plastic multiplier, and f b,m is the (non)linear failure surface of either bricks (b) or mortar (m). Let the
failure surfaces of bricks and mortar be approximated by m
planes (see Figure 4), so that each strength criterion is defined
by a set of linear inequalities of the form fb,m;A in b in. As
i
pl varies at most linearly within each sub-cell, plastic admissibility is checked only at three of the corners. Hence, nine linear
equality constraints per sub-cell are introduced in the matrix
eq i
form as Aeq
U Ai 0, where U is an array collecting
Ui
the seven DOFs describing the microscopic
h velocity field
i (i.e.
i
iT iT iT T
U = {U1, U2, W1, W2, U1t, Ut2W1t}T); A B C
is an
array of 3m entries, collecting the rates of the plastic multii
pliers J at three of the corners of the rectangular sub-cell (J
eq
eq
= A, B, C); and AU(i)
, A(i)
are 9 7 and a 9 3m matrices,
respectively. The plastic admissibility conditions are then
assembled cell by cell into the following global system of
equality constraints
h
iT
h
iT
eq
Aeq
, 1 T . . . 6 T
where Aeq
U1 T . . . AU6 T
U
and A eq is a block matrix of dimension (6 9) (6 3m),
40
40
x2: mm
x2: mm
eq
Aeq
U U A 0
22:
20
0
20
0
20
20
100
0
50
100
x1: mm
Undeformed elementary cell (standard Italian brick,
joint 20 mm thick)
100
50
100
150
40
x2: mm
x2: mm
0
50
x1: mm
U1 0
40
20
0
20
50
0
50
x1: mm
U2 0
100
20
0
100
150
20
0
20
50
0
x1: mm
50
50
x1: mm
100
30
20
10
0
10
100
100
50
W1 0 W1 = W2
x2: mm
100
x2: mm
50
100
W3 0
50
50
x1: mm
100
150
All U1 = W1 / 2, U1 = 2W1,W1 = W2 W3 = 12 W1
110
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
40
x2: mm
x2: mm
40
20
0
20
0
20
20
50
0
50
100
x1: mm
Undeformed elementary cell (standard Italian brick,
joint 20 mm thick)
100
100
50
100
0
50
x1: mm
100
0
x1: mm
U1t 0
40
x2: mm
50
40
x2: mm
20
0
20
50
100
50
100
150
20
0
20
100
50
x1: mm
U2t 0
minimisation problem
eq
eq
eq
Aeq
A1 A2 A6
i
in
i iT i
iT i
bin B bin C
2
i i
i h
iT
iT
bin
01m bin
2
in
6
X
i h
i1
01m
iT
bin
iT
bin
min in
8 8
>
>
> D 1
>
>
> eq
> >
>
>
> AU U Aeq
>
>
0
<
< >
s:t:
s
>D 1
>
v n dS
>
>
> >
A
>
>
@Y
>
>
>
>
>
:
: >
0
a
b
c
d
7.
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
7.1
Frictional angle ()
Cohesion (c)
36
01 MPa
2c cos
fc
1 sin
2c cos
ft
1 sin
012
FEM solution
Model I-P4
01
Model I-P3
Model I-P2
v: MPa
008
Model I-P0
006
=0
h
004
002
0
0
005
01
015
02
h: MPa
112
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
025
03
= 45
014
012
= 45
h
v: MPa
01
008
Model IV (MoC)
Model I-P4
Model I-P3
Model I-P2
Model II and III
006
004
002
0
005
01
h: MPa
(a)
= 225
012
v = 225
01
v: MPa
008
Model IV (MoC)
Model I-P4
Model I-P3
Model I-P2
Model II and III
006
004
002
0
005
01
015
02
h: MPa
(b)
= 0
v: MPa
01
v
005
0
= 0
h
0
005
01
015
02
025
03
Optimisation time
min:s
035
h: MPa
(c)
00:12
01:28
06:33
09:33
01:44
01:33
02:19
113
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
x2: mm
2
0
2
10
50
0
x1: mm
10
Shear failure of bed joint
(a)
(b)
114
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
= 225
= 45
0
0
Model IV (MoC)
Model I-P4
Model I-P3
Model I-P2
02
04
02
06
= 45
08
v /fc
04
v
v /fc
Model IV (MoC)
Model I-P4
Model I-P3
Model I-P2
06
= 225
08
1
1
12
12
08
06
04
02
12
08
06
h/fc
h/fc
(a)
(b)
04
02
= 0
0
Model IV (MoC)
02
v /fc
04
Model I-P4
v
Model I-P3
= 0
Model I-P2
06
08
1
2
15
05
h/fc
(c)
7.2
115
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
Multi-linear
compression cap
MohrCoulomb failure criterion in plane strain
Compression cut-off
Rankine failure criterion
xy: MPa
fc
Tension cut-off
Rankine failure criterion
1
05
ft
Tresca strength cu
0
4
xx: MPa 1
3
0
yy: MPa
(a)
Failure surface 1
Failure surface 2
xy: MPa
1
05
0
0
2
xx: MPa
nn
Thick joint
plane stress
failure criterion
nt
yy: MPa
ft
nn: MPa
nn
nt
c
2
cu
fc
Failure surface 1
nt: MPa
(b)
Figure 15. (a) Plane stress failure criteria adopted for thick mortar
joints and (b) interface failure criteria adopted for joints reduced
to interfaces
116
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
Failure surface 2
cu: MPa
25c
2: deg
45
cu: MPa
25c
fc45: MPa
45
cu: MPa
02
cu: MPa
02
fcxx ( fcyy), masonry compressive strength along the x (y)-axis; ftxx ( ftyy), masonry tensile strength along x (y)-axis; cu, pure shear
strength; xx (yy), friction angle along the x (y) direction
Table 3. Mechanical properties adopted in the last set of numerical simulations
117
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
Line 2
yy = 0: MPa
05
45
045
yy
04
Line 1
xy: MPa
035
xx
03
025
02
D C
B
Lin
015
01
005
0
5
(a)
xx = 0: MPa
15
D
1
05
xy: MPa
xy: MPa
15
2
xx: MPa
(b)
D
E
C
05
B
0
5
A
4
0
6
A
5
yy: MPa
3
2
45: MPa
(c)
(d)
one for thick joints. As can be noted, there are some non-negligible differences among the results obtained with the first and
second failure surfaces considered in Figure 15, and also the
effect of the reduction of the joint to an interface modifies
rather visibly the failure surface, consider for instance
Figure 16(c) and the first failure surface.
8.
The most relevant outcome of the comparison is, however, the
inaccuracy of the macroscopic orthotropic models, especially
along section lines 1 and 3, even in comparison with the homogenised failure surfaces obtained assuming for joints the
second strength domain (i.e. with Tresca cut-off).
Conclusions
118
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
Point
Numerical
Experimental
A Line 1
xx = +100
B Line 1, 2, 3
xy = +100
F Line 1
xx = 100
119
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.
REFERENCES
120
Downloaded by [ Distributed Tech Svcs-Sci] on [19/09/16]. Copyright ICE Publishing, all rights reserved.