Documente Academic
Documente Profesional
Documente Cultură
net/publication/245395510
CITATIONS READS
4 149
2 authors:
Some of the authors of this publication are also working on these related projects:
Strength and ductility of high-strength concrete column confined with high-strength steel ties View project
All content following this page was uploaded by Hooshang Dabbagh on 25 April 2016.
Abstract: This paper reports the development of the cracked membrane model using a
fixed-crack approach for the finite element modelling of orthogonally reinforced
structural elements under plane stress. The model falls into the broad category of
“compression field models” combining elemental components from the modified
compression field theory with those of the tension chord model. The tension stiffening
effect is developed using a steel-concrete bond relationship with its basis in limit
analysis. With this approach the tension stiffening and tension softening components
of the concrete tension resistance mechanism are decoupled allowing for rational
models for each component. With adoption of a fixed crack formulation aggregate
interlock along the cracks is explicitly modelled allowing for modelling of reinforced
concrete members containing no or low quantities of steel reinforcement where slip
along cracks significantly affects the results. For example, the modelling of shear
critical beams containing low shear reinforcement volumes. The model has been
incorporated into the finite element program for the analysis of reinforced concrete
structures with verification of the finite element model presented.
Key words: finite element, reinforced concrete, cracked membrane, aggregate interlock.
A Smeared – Fixed Crack Model for FE Analysis of RC Membranes Incorporating Aggregate Interlock
CMM maintain equilibrium at the crack faces. As where θr is the angle between a vector normal to the
equilibrium is fully satisfied at the cracks, and not cracks and the global X-axis (–π/2 < θr ≤ π/2); σx, σy, and
expressed in terms of average stresses across an τxy are the in-plane normal and shear stresses in the
element, the link to limit analysis is maintained and the global XY-coordinate system, respectively; σcn and σct
coupling between principal stress and principal strain are the concrete stresses normal and parallel to the
angles removed. Foster and Marti (2002, 2003) direction of cracking, respectively; τcnt is the corresponding
implemented the CMM as a finite element procedure shear stress; ρx and ρy are the steel reinforcement ratios in
based on a smeared-rotating crack approach for the the global X- and Y-directions and σsx, and σsy are the
analysis of orthogonally RC members in plane stress stresses in the X- and Y-reinforcement, respectively.
such as deep beams and walls. The subsequent
experimental verification gave good results but 2.1. Tension Stiffening
underscored the difficulty noted by Vecchio (2000) of Tension stiffening is a direct resultant of transfer of
using a rotating crack approach for the analysis of lightly forces between the reinforcing steel and the concrete via
reinforced RC elements failing in shear. bond. If sufficient force transfer occurs such that the
In this paper, a FE smeared-fixed crack model is stress in the concrete exceeds the tensile strength of the
developed for the analysis of RC structures based on the concrete, the concrete cracks. Using the stepped-rigid
CMM. The link to limit analysis is maintained with plastic model of Marti et al. (1998) and Kaufmann and
the development of an aggregate interlock model for the Marti (1998), shown in Figure 2(c), it follows that the
calculation of shear on the crack faces. The model is mean stress in the concrete between cracks before
verified using experimental data from tests undertaken yielding of the reinforcing steel is
by Vecchio and Collins (1982), Bresler and Scordelis
(1963) and Leonhardt and Walther (1966). τ b λ x fct τ b λ y fct
σ ctsx = ; σ ctsy = (2)
τb0 2 τb0 2
2. CRACKED MEMBRANE MODEL
For equilibrium at the cracks for the plane stress element where σctsx and σctsy are the average stresses and λx fct
shown in Figure 1, it can be shown that and λy fct are the maximum stresses in the concrete
between cracks due to tension stiffening in the X- and
σ x = σ cn cos2 θr + σ ct sin 2 θr + τ cnt sin (2θr ) + ρ x σ sx (1a)
Y-directions, respectively; fct is the tensile strength of
σ y = σ cn sin 2 θr + σ ct cos2 θr − τ cnt sin (2θr ) + ρ y σ sy (1b)
the concrete and τb is the bond shear stress. After
yielding of the reinforcing steel, τb0 in Eqn 2 is replaced
τ xy = 0.5 (σ cn − σ ct ) sin (2θr ) − τ cnt cos (2θr ) with τb1 (Figure 2(c)).
(1c)
The tension stiffening factors as obtained from
Figure 2(b) are given by
∆σ cx srm
(a)
σy
(c) λx = = (3a)
τxy
σcn cos θr ρxσsx fct srmx 0 cos | θr |
∆σ cy srm
σx τcnt cos θr λy = = (3b)
+ fct srmy 0 sin | θr |
1
σx σx
τxy τcnt sin θr
where ∆σcx and ∆σcy are the X- and Y-component
σct sin θr
σy stresses of the tension stiffening stress; srm is the crack
(b) (d) spacing measured normal to the cracks and srmxo and
ρy σsy
srmyo are the crack spacings for uniaxial tension in the
σct cos θr + σcn sin θr X- and Y-directions, respectively, determined using the
t n stepped rigid plastic bond-slip relationship (Figure 2(c))
θr θr τcnt sin θr
of Marti et al. (1998) and are given by
τcnt cos θr
τxy
σy
srmx 0 =
fct Ø (1 − ρ x )
; srmy 0 =
( )
fct Ø 1 − ρ y (4)
1 2τb0 ρx 2τb0 ρy
Figure 1. Orthogonally reinforced membrane subject to plane stress: where Ø is the diameter of the reinforcing bars and
(a) Applied stresses, (b) Axis notation, (c) and (d) Stresses at a ρx and ρy are the reinforcement ratios in the X- and
crack (Foster and Marti, 2003) Y-directions, respectively.
rm
τb
s
rm
s
rm
s
srmy
θr τb1
srmx
X
δy δ
λx fct
Tension stiffening
srmx srmx stresses
Figure 2. Tension stiffening stresses: (a) In the material axis directions, (b) In orthogonal tension chords of the cracked membrane,
(c) Bond-stress versus slip relationship (Foster and Marti, 2002)
From the Mohr’s circles plotted in Figure 3(a), for the τxy
cot θc =
stress between the cracks not to exceed the tensile τxy tan θr + fct (1 − λx ) (7)
strength of the concrete it can be shown that
fct ( λ x − 1) + |τ xy | cot | θr | ( λ x − 1) − | τxy | tan | θr | (6b) b = srmx 0 sin θr + srmy 0 cos θr (9b)
λy =
fct ( λ x − 1) − | τ xy | tan | θr |
c = 2 srmx
2
( )
0 + srmy 0 sin θr cos θr − 2 srmx 0 srmy 0 (9c)
2
y
λx fct
τ between
θr = 60º
(σcx, τxy) cracks 150
θr = 30º
fct 0.2
srm0 (mm)
Figure 3. CMM: (a) Concrete stresses at and between the cracks, (b) Polar representation of Eqn 8 for panel with τb0/2 = fct = 2.9 MPa,
Øx = Øy =16 mm, ρx = 0.0125 and ρy = 0.0250
A Smeared – Fixed Crack Model for FE Analysis of RC Membranes Incorporating Aggregate Interlock
The second solution to Eqn 5 is obtained by changing • For θr < θλx = λy and η < srmx0/srmy0, srm is
the sign ahead of the square root term in Eqn 8. The calculated by Eqn 8 for η = srmx0/srmy0, λx is
crack spacing given by Eqn 8 is plotted in Figure 3(b) calculated by Eqn 3a and λy by Eqn 6b.
for various values of η for a panel with τb0/2 = • For θr > θλx = λy and η < srmy0/srmx0, srm is
fct = 2.9 MPa, Øx = Øy =16 mm, ρx = 0.0125 and calculated by Eqn 8 for η = srmy0/srmx0, λy is
ρy = 0.0250. For the solution presented in Figure 3(b), calculated by Eqn 3b and λx by Eqn 6a.
however, it is seen that a difficulty arises at the Using this methodology the implied boundary
boundary for θr = 0 where for a solution to exist λx = 0, requirements of uniaxial tension are satisfied whilst
λy = 0 and srm = 0. This is different to the known maintaining the limit conditions dictated by Mohr’s
boundary condition for uniaxial tension in the stress circle.
X-direction where λx =1, λy = 0 and srm = srmx0. The For the case of θr = θc the solution to Eqn 5 simplifies
difficulty arises from an over-calculation of the bond- to the Vecchio and Collins (1986) equation
stress in the Y-direction reinforcement and used in −1
Eqn 4. This occurs as a result of the assumption that the cos | θ c | sin | θc |
srm = + (11)
limiting bond-stress is attained along the length of the srmx 0 s rmy 0
bar (that is τb = τb0). However, for the special case of
θr = 0 the force in the Y-direction reinforcement is zero Examination of Eqn 4 shows that θr = θc occurs for
and, thus, in the Y-direction τb = 0. Similar comments θr = 0, θr = π/2, λx = λy =1 and as the shear stress to
apply for the case of θr = π/2. concrete tension strength ratio η → ∝.
For θr = 0, solving Eqn 8 for η with srm = srmx0 gives The average stresses due to any set of applied
η = srmx0/srmy0 and, similarly, for θr = π/2 with boundary tractions are attained with substitution of Eqns
srm = srmy0 gives η = srmy0/srmx0. In general, for the 2 and 3 into Eqn 1 and writing in terms of the principal
solutions of Eqn 5 to satisfy the boundary limitations 12-axes
η = srmx0/srmy0
m
sr
θλx = λy
2.2. Aggregate Interlock and Dowel Action
Considering the maximum principal stress as the criterion
srmx0 for cracking, the shear stress and strain at the crack
are zero at the point of cracking. As loading continues
Figure 4. Boundary limit conditions for the calculation of srm for and if the direction of the crack is fixed, the principle
τb = τb0 in both X- and Y-steel strains and stresses are not coincident and shear forces are
(a) (b)
s rm
X
/2
θr σcnn
uc τcnt
Y
νc
wcr
t(ν
)
n( ν
)
δs n
t δs
w cr
Figure 5. (a) Displacements and strains, (b) Slip and crack opening displacement and conjugate stresses
where g is a function of δs and wcr. For a uniform crack τ co = − fcp 24 ...... for unreinforced concrete (17a)
spacing, srm, and noting that the component of shear
displacement of concrete between cracks is small τ co = 0 ...... for reinforced concrete (17b)
compared to the slip at the crack, the shear and normal
strains at the cracks are For reinforced concrete, Eqn 16 is modified to
include dowel action:
ε cnn = wcr / srm ; γ cnt = δ s / srm (15)
α 1.18 fcp0.75δ s . α fcp0 5
τ cnt = ≤ (18)
The aggregate interlock model adopted in this study 0.1 + 1.6 wcr + 3wcr
2 0.31 + 24 wcr ( a + 16 )
is developed from the experimental work of Walraven
(1980). Using a regression analysis it was found for
unreinforced concrete that α = 1 + 13.5 ρ 3 fcp (19)
3
τcnt +τo = 0.75
fcc 1.5 –20%
2 ds 0.1 + 1.6w + 3w2
1.0
1 R2 = 0.97
0.5
0 0.0
0 0.5 1 1.5 2 2.5 3 0.01 0.02 0.03 0.04 0.05 0.06 0.07
wcr (mm)
r / 3 fcp
Figure 6. Comparisons of: (a) Modified Walraven (1980), (b) Dowel shear factor with test data (fcc = cube strength)
A Smeared – Fixed Crack Model for FE Analysis of RC Membranes Incorporating Aggregate Interlock
30 –sc
fcp = 40 MPa
r = 0.01 bfcp
25
fcp
wcr = 0.2 mm
20
b>1
tcnt (MPa)
bfcp
15 0.5 mm
b=1
10 1.0 mm
b<1
2.0 mm
5 becp ecp becp –εc
Figure 7. Shear stress versus slip given by Eqn 18 for fcp = 40 MPa
and ρ = 0.01 a biaxial strength criterion (see Foster and Marti 2003).
For tension-compression stresses β is a disturbance
factor obtained from the modified compression field
where ρ is the reinforcement ratio perpendicular to model of Vecchio and Collins (1986).
crack face, a is the maximum size of the aggregate (in The stress versus strain relationship adopted for
mm) and α is the dowel shear factor (α =1 for concrete in compression is taken as a scale of the
unreinforced concrete). The limiting value for τcnt on the Thornfeldt et al. (1987) uniaxial compression base
right hand side of Eqn 18 is a modified version of that curve with the calibrated parameters of Collins and
used by Vecchio and Collins (1986) adding the dowel Porasz (1989). Figure 8 shows the stress versus strain
shear component. In Figure 6(a), the model represented relationship used in this study for various scaling
by Eqn 16 is compared with the test data for factors, β.
unreinforced concrete. A good correlation is observed.
In Figure 6(b), the proposed dowel shear factor is 2.4. Reinforcing Steel
compared with the Walraven data. The data scatter is A tri-linear stress-strain model is adopted to model
not unreasonable considering the simplicity of the the reinforcing steel with the properties as defined in
model but more work is needed to isolate the key Figure 9.
parameters. The resulting stress versus slip model is
plotted in Figure 7 for fcp = 40 MPa and ρ = 0.01. 3. CONSTITUTIVE RELATIONSHIPS
The secant shear stiffness due to aggregate interlock A modified form of the equivalent uniaxial strain
is calculated as model (Darwin and Pecknold 1977) is used to obtain
the biaxial stress-strain relationships for uncracked
Gcr = τ cnt γ cnt = srm τ cnt δ s (20) concrete in tension-compression and for concrete in
2.3. Concrete
Concrete is a quasi-brittle material that softens both in σs
compression and tension leading to strain localization.
fu
For tension, the bilinear stress-strain model of Petersson
fw Eu
(1981) is used (refer to Foster and Marti 2003). 1
The behaviour of concrete in compression is a Ew
function of the biaxial state of stress. The concrete fsy 1
compression strength is given by Ed
1
fc* = β fcp (21) Es
1 εs
where fcp is the compression strength of the in-place
εy εw εu
concrete. For biaxial compression stresses the factor β
in Eqn 21 is a confinement factor and is obtained from Figure 9. Tri-linear stress-strain model for reinforcing steel
( Ectsx + ρ x E sx )
compression-compression; the details of which are
0 0
given in Foster and Marti (2003). The equivalent
uniaxial strains in the principal 1–2 coordinate system [ D] cts + [ D] s = 0 (Ectsy + ρ y E sy ) 0 (29)
(written as ε1u, ε2u) are the strains that would exist in 0 0 0
each direction when the stress is zero in the other. For
clarity, the subscript ‘u’ is dropped from the notation in Finally, the element stiffness matrix is obtained in the
the formulation that follows. usual manner
Relating the average stresses and strains of the
[k ] = t ∫A [ B] [ D] xy [ B] dA
T
cracked concrete in the familiar manner (30)
{σc1 , σc 2 , τc12 }T = [ D] c12 {εc1 , εc 2 , γ c12 }T (22) where t is the element thickness and [B] is the strain
displacement matrix.
where [D]c12 is the constitutive matrix of cracked Further details of the constitutive relationship used,
concrete as given by including the compression modelling are given in Foster
and Marti (2002, 2003).
Ec1 ν12 Ec1 0
[ D]c12 =
1 ν E
Ec 2 0 4. NUMERICAL EXAMPLES
1 − ν12 ν21 0 (23)
21 c 2
0 (1 − ν12 ν21 ) Gc12 The model described has been incorporated into a finite
element program (RECAP) for the analysis of RC
From Eqn 22, the shear strain is written as (Foster and Gilbert 1990; Foster 1992). Four node
isoparametric concrete membrane elements were
γ c12 = γ cr12 + γ cs12 (24) developed using the material and constitutive
relationships discussed above. The FE results are
where the subscript ‘cs’ refers to the strain in the solid compared with tests on the RC panel PV19 tested by
concrete between cracks and the subscript ‘cr’ refers to Vecchio and Collins (1982), beams 0A1, 0A2 and 0A3
the strain in the fracture (crack) zones. The secant shear by Bresler and Scordelis (1963) and the deep beam
stiffness is then given by DWT2 of Leonhardt and Walther (1966) together with
the results of the rotating crack CMM:FE model of
Gcr12 + Gcs12 Foster and Marti (2002, 2003). For all analyses: v = 0.2,
Gc12 = (25)
Gcr12 ⋅ Gcs12 λ =1.0, τb0/2 = τb1 = fct and Gf = 75 N/m.
A Smeared – Fixed Crack Model for FE Analysis of RC Membranes Incorporating Aggregate Interlock
Figure 10. Panel PV19: dimensions, reinforcement details and shear stress versus shear strain
(2003) was 3.34 MPa, compared the experimental value 4.2. Bresler and Scordelis (1963) Beams OA1,
of 4.0 MPa. For the fixed crack model, the peak shear OA2 and OA3
stress calculated was 4.20 MPa (26 percent higher than Although the Bresler and Scordelis (1963) beams have
that obtained by the rotating crack model). For both been modelled by many with good accuracy between
analyses, the element failed by crushing of the concrete numerical and experimental results (for example
after yielding of the Y-direction reinforcing steel but Vecchio 1989; Foster and Gilbert 1990) Vecchio
before yielding of the X-direction steel, as observed in (2000a) highlighted a significant flaw in previous
the experiment. As is expected, both the rotating and methodologies where tension stiffening and tension
fixed crack models provide identical results to the point softening were typically incorporated using a single
of yielding of the Y-steel after which the solutions model. In the case of Vecchio (2000a) for concrete
diverge as the friction along the crack becomes members with low quantities of reinforcement, a check
significant. of stresses along a crack determined that slip at the crack
P
b = 305 mm
465
560
Figure 11. Bresler and Scordelis (1963) beams OA1, OA2 and OA3
would occur at loads significantly below those of the Conversely, the failure loads calculated by the fixed
experiment unless a significant residual tensile stress crack implementation compare well with the test results.
was carried across the cracks. Vecchio later developed The increase in failure load for the fixed crack model
the disturbed stress field model to account for the over that of the rotating crack model was between 21
identified weakness (Vecchio 2000b). and 38 percent. Figure 13 shows the principal tensile
As for Vecchio (2000a), in the CMM:FE rotating strain contours for beam OA1 after failure. For the fixed
crack implementation of Foster and Marti (2003) it was crack model, the contours show the initiation of the
recognised that lacking from the implementation was localised shear crack while for the rotating crack model
the allowance for shear transfer parallel to the crack the shear mechanism is clearly identified. In all cases,
faces via aggregate interlock and dowel action leading for both the rotating and fixed models, failure was by
to an under-estimation of the capacity of lightly beam shear, as observed in the experiments. For the
reinforced members. Examples of beams with low fixed crack model, the response beyond failure for
volumes of web reinforcement are beams OA1, OA2 these tests was difficult to capture using conventional
and OA3 of Bresler and Scordelis (1963). The beams arc length or displacement control due to bifurcations
were modelled using the rotating and fixed crack caused by snap-back leading to instabilities in the
implementations of the CMM:FE model using 4-node solution routine. While special techniques may be used
membrane elements. The beam dimensions, details and to capture this response, this is not the focus of this paper.
results are presented in Figure 11 and the FE meshing
used for the analyses is shown in Figure 12. Both the 4.3. Leonhardt and Walther (1966) Deep Beam
fixed and rotating versions of the model compare well DWT2
with the load deflection response to the point of failure. Leonhardt and Walther (1966) tested nine deep beams to
The rotating crack model, however, consistently failed study non-flexural behaviour of reinforced concrete
at loads lower than those observed in the experiment. members. In this study, the two span deep-beam DWT2
ρx = 0.0 P/2 L
C
ρy = 0.0
All load and reaction
plates 25 mm thick
418 steel plate
143
ρx = 0.089
ρy = 0.0
158 125 1613 150
Symmetry about
centreline
418
143
ρx = 0.111
ρy = 0.0
158 125 2070 150
ρy = 0.0
418
143
Beam OA3
A Smeared – Fixed Crack Model for FE Analysis of RC Membranes Incorporating Aggregate Interlock
(a)
0.002
(b)
Rotating Crack Model
Beam OA1
Princ. Tensile Strains
(Max = 0.020)
Figure 13. Principal tensile strain contours after failure in Beam OA1: (a) Fixed crack model (P = 193 kN),
(b) Rotating crack model (P = 208 kN)
concrete:
R = 0.094P 480 480 480 360
fcp = 27.1 MPa fct = 2.5 MPa S = 0.250P
Ec = 27.1 GPa ecp = 0.0020 T = 0.156P
R ST T SR L
C A
P/2 P/2
steel:
Es = 206 GPa Ew = 5.0 GPa 3 Midspan
support
esy = 0.002 esw = 0.05 thickening
region 1:
1490 1∅6
rx = 0.0195 ry = 0.0020 1∅6
∅x = 8.0 mm ∅y = 5.0 mm 1∅6
region 2: 1∅6
2∅6
rx = 0.0195 ry = 0.0020
∅x = 6.0 mm ∅y = 5.0 mm 110
region 3: 120 2 4∅8 100
1 A
rx = 0.0019 ry = 0.0020 Section A-A
2∅5 e/w
∅x = 5.0 mm ∅y = 5.0 mm 160 1280 160
Figure 14. Details of Leonhardt and Walther (1966) beam DWT2 and FE mesh
(shown in Figure 14) is analysed using the finite of 953 kN. An inspection of the displaced shape at
element mesh shown with one half the beam modelled failure revealed that after yielding of the reinforcement a
with symmetry. shear slip occurs in the elements adjacent to the central
The results of the analysis of beam DWT2 are plotted support thickening. For the rotating crack model,
in Figure 15 for the load versus midspan displacement however, no load can be transmitted to the support via
and it is seen that the rotating CMM:FE formulation shear along the cracked surface. In the fixed CMM:FE
(without residual concrete tension strength after cracking) formulation the stiffness results are more consistent
under-calculates both the stiffness and the failure load of with the measured data with a portion of the load being
the member after cracking. The model failure load was transferred through aggregate interlock and dowel
985 kN, 80 percent of the experimental failure load of action. The failure load was 1090 kN (89 percent of the
1230 kN and is consistent with the plastic collapse load reported failure load).
1400 Collins, M.P., and Porasz, A. (1989). “Shear strength for high
Exp. Failure Load = 1230 kN strength concrete”, Bulletin No. 193, Design Aspects of High
1200 Strength Concrete, Comité Euro International du Béton (CEB), pp.
75–83.
1000 953 kN Darwin, D. and Pecknold, D.A. (1977). “Non-linear stress-strain law
for concrete”, Journal of the Engineering Mechanics Division,
Load (kN)
A Smeared – Fixed Crack Model for FE Analysis of RC Membranes Incorporating Aggregate Interlock