Professional Documents
Culture Documents
net/publication/215469225
CITATIONS READS
0 666
2 authors, including:
Tammam Bakeer
Technische Universität Dresden
44 PUBLICATIONS 113 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Master thesis: ENGINEERING MODELS FOR THE IN-PLANE LOAD CARRYING CAPACITY ASSESSMENT OF CONFINED MASONRY: NUMERICAL AND ANALYTICAL MODELS View
project
CASA MILÀ The influence of the construction of high speed train tunnel on the works of Antoni Gaudì, Part 2 View project
All content following this page was uploaded by Tammam Bakeer on 15 March 2016.
Dr.-Ing. Tammam Bakeer, Technische Universität Dresden, Faculty of Architecture, Chair of Structure Design,
D-01062 Dresden, Germany, E-Mail: tammam.bakeer@tu-dresden.de
Dipl.-Ing. Peter Schöps, Jäger Ingenieure GmbH, Wichernstr. 12 * D-01445, Germany, Radebeul, Email:
p.schoeps@jaeger-ingenieure.de
(b) Simplified micro-modelling through the adoption of (b) Determining of the contact point in the isoparametric
"geometrically expanded" masonry units with a single coordinates of the master segment.
"averaged" interface representing the mortar and the two (c) Computing the projection distance in the local coordi-
mortar/unit interfaces. This model requires the material nate system which is embedded in the master segment.
model of the expanded unit and masonry joints.
(d) When the projection distance found to be negative,
- Macro-modelling or single phase material model, the its absolute value indicates the depth of the penetration.
quasi-periodic nature of masonry has prompted to investi- The slave nodal force is calculated according to the fol-
gate the use of homogenization techniques, where all ma- lowing equation
sonry components are smeared by an equivalent homoge-
nized continuum. One-phase material models have been f
{s
= K
{c
⋅ δ{
treated masonry as an ideal homogeneous material with (1)
constitutive equations that differ from those of the compo- contact force contact stiffness penetration depth
nents.
(e) Distributing the contact force to the master segment
2 Methods of modelling nodes. Each master node gets a fraction of the slave
force based on the contact point location by using the
2.1 Discrete Modelling isoparametric shape functions
Several discrete modelling methods are available in litera-
ture (Rigid Bodies Spring Method, Discontinuous Deforma-
tion Analysis, Non-Smooth Contact Dynamics, Modified
Discrete Element Method, Combined Finite-Discrete Ele-
ments and Applied Element Method), Bakeer [4]. The mod-
performed using parallel 40 Intel Itanium processors “SGI
segment normal
Altix 4700” at the centre of High Performance Computing of
TU-Dresden. The calculation for the model was carried out
projection distance
3 (-1,-1) 4 (1,-1)
(a)
(b)
In ANSYS most materials basically can better simulate the After the damage is initiated, the stress is linearly scaled
hardening behaviour. However, the concrete material in down until the crack width reaches the critical distance
ANSYS can be used with Solid element 65 to represent wc at which the interface failure is complete. Tiebreak con-
brittle softening. ANSYS has user-programmable features
tact in LS-DYNA uses penalty method. This produces some
that offer the possibility to implement user materials. To
relative displacement between the surfaces before the dam-
consider the softening behaviour of masonry constituents,
age of the contact which results in deponding, due to which,
2D material model has been implemented Figure 4. Figure 5 the yield criterion is possible to be achieved. This brings out
shows the calculation of element length to consider the an unrealistic behaviour because the tractions between the
possible orientation of crack. The material model has been surfaces are suddenly jumped down. Further options to
tested on shear wall to examine the damage patterns and prevent this behaviour are available in LS-DYNA by increas-
the load deflection curve, Figure 6. ing the stiffness scale factor of the contact. However, care
should be taken with higher penalty stiffness, which results
LS-DYNA comprises material models that cover a wide in high frequency modes and therefore instability in the
range of masonry constituents, but such materials have solution.
been developed basically for concrete and soils. They rep-
1
0
-7 -6 -5 -4 -3 -2 -1 0 1
-1
-2
σ 1 in N/mm ²
σ 2 + σ 2 − (α ⋅ f )2 TT − Region
-3
1 2 t t
F = σ 1α k fk − σ 2α t ft − α k fkα t ft TC − Region -4
σ 1 + σ 2 − σ 1 ⋅ σ 2 − α k ⋅ fk CC − Region
2 2
-5
-6
-7
σ1 in N/mm²
Figure 5 Possible crack orientation and calculation of the correspond element length
Figure 6 Load-deflection curve and damage pattern at load of 0.55 N/mm² (0.5 σ0) without defining contact on the head
joints (variation coefficient of tensile strength 15%)
Beattie et al. [6], Beattie et al. [7] and Burnett et al. [9] have Several interface models have been developed in literature
developed a discrete crack model in LS-DYNA for modelling to model masonry mortar joints. Interface element also has
masonry joints within a project to study the performance of been developed in LS-DYNA in ver. 971, and has been
masonry parapet walls that subjected to vehicle impact. The used for problems of dynamic delamination modelling, Ian-
yield surface in the proposed model is concave. For this nucci [11] and modelling of damage in composite materials,
reason the return mapping has been used parallel to shear Jiang et al.[14]. Interface elements are available to be used
axis in compression region instead of radial return mapping, with the following cohesive material models in LS-DYNA:
but in such case the assumption of the plastic potential
theory (Huber-von Mises theory) is not valid. (1) Elastic cohesive material model,
(2) Tvergaard and Hutchinson cohesive material The adoption of non-smooth yield surface needs further
model, and treatments for corners which increase the complexity of
implementation, and increase the computation time. The
(3) General cohesive material model. complexity is going to be worse especially for explicit
To simulate the dynamic events after the failure of the inter- solvers like LS-DYNA, where the material subroutine has to
face elements, it would be possible to replace the surfaces be called in time steps smaller than those in implicit solvers.
which linked by the failed interface elements with frictional An interface model has been implemented into LS-DYNA
contact model. The deletion of the interface element after that based on smoothed yield surface (Bakeer [4]), It re-
the failure does not bring any loss in the mass if the thick- duces the computation time and avoids the treatments of
ness of the interface element has been set to zero. In such corners, Figure 7. The proposed model is multi yield surface
way, the inherent difficulties associated with large displace-
but does not need any further treatment of the transition
ment after the failure of the interface element also are
points.
avoided. In LS-DYNA ‘contact eroding single surface’ offers
a possibility to detect the contact on the eroded surfaces
after the failure of the interface element.
23
τ
µ = µ0 + (µr − µ 0) ⋅ (1 − Ω (κ))
Shear Stress
F2 = τ − c ⋅ Ω(κ) + µ ⋅σ
arctan( µ ) F12
Region 3 Region 2
c2
F3 =τ 2 − Ff2 ⋅ Fc F1 =τ 2 + (1 −α) ⋅σ 2 + 2µ ⋅ c ⋅ Ω⋅σ − c 2 ⋅ Ω2
ft 2
c ⋅ Ω (κ)
Softening
σ + L(κc )
2
Fc =1 − arctan( µr )
X (κc ) − L(κc ) Region 1
B
Ff = c ⋅ Ω(κ) + µ ⋅σ c
-fc ⋅ Ωc (κc ) -L(κc ) f t ⋅ Ω(κ ) µ
σ
normal stress
A C
f t 2 −α
A= Ω
2 1 −α
c 2 −α
B= Ω
2 1 −α
c α2
C= Ω
Region 2 4µ 1 −α
F12
F23
Figure 7 The smooth yield surface of the cohesive interface model, Bakeer [4].
0,8
the tensile part governed by elliptic yield surface, Figure 9.
0,7
Several options have been provided for the definition of the
0,6
strength-crack width relationships: linear, bilinear, exponen-
0,5
tial, and Hordijk functions.
0,4
The implemented subroutine has been examined by three 0,3
point flexural test, and DIN shear test. The simulation results spröde
0,2 duktil
of the three point bending test Figure 10 and the DIN shear
0,1
test have been demonstrated good conformity with tests
0
results. The influences of the crack energy and element size
0 50 100 150 200
have been studied. Figure 8 shows the results for brittle and Elementlänge [mm]
ductile behaviours.
Figure 8 Influence of crack energy and element size
τ fvk0-µ·σ
f Linear
Bilinear
Elliptic part
Exponential
Softening αf Hordijk
fvk0
µS·σ
w
ft -σ wc
Figure 9 The yield surface and the possible functions for strength-crack width relationship
300
Versuch Spaltöffnung
250
FEM linear
150
FEM Hordijk
100
50
0
0 0,2 0,4 0,6 0,8 1
-50
Weg in mm
Figure 10 (Left) Simulated bending test with a notched prism for an inverse calculation of the crack energy by using user-
implemented Interface-Elements, (Right) Comparison of FEM and test for tensile strength of 0.78 N/mm² and
crack energy of 10 N/m for different softening functions (PP4 AAC)
5 Loading regime
High
The type of loading has a great influence on the perform- pressure
ance of the FEM solver. Explicit based integration solvers
are highly recommended to solve short duration loading Figure 13 Simulation the dynamic response of the
problems, whereas implicit based solvers are recommended structure under high pressure
to solve long term loading problems. LS-DYNA is basically
has powerful explicit solver, therefore, it has widely em-
ployed to simulate high dynamic events like impact, explo-
sion, and earthquake actions. On other hand, ANSYS has 6 Confined masonry
powerful implicit solver with which the static and long term Confined masonry differs from both reinforced masonry and
loading problems can be well solved. infill masonry. The most essential difference in comparison
with infill masonry consists in the fact that masonry carries a
portion of the vertical load. The sequence in the erection of
the structural members is therefore an important factor for
confined masonry
In the following the typical numerical crack patterns have
been simulated in ANSYS (Schöps et al. [24]). Compared
with unconfined masonry the cracks are found out clearly
wide and also the stress distribution in the masonry is more
homogeneous. In Figure 14, it can be seen that the first
cracks goes diagonally through the masonry and both re-
sulting wall halves are held together by the frame. With the
most unfavourable estimation that both wall halves take the
same portion of the shear load a shear action arises for the
40%g
frame by the half height of the external shear load.
(b)
The joint failure could be examined only numerically, be-
cause the bond strength of the used AAC-unit is greater
Figure 12 Simulation of masonry structure under
than the tensile strength of the units. The failure type varies
earthquake actions, Bakeer [4]. depending on the unit geometry, the vertical load, and the
The combination of explicit and implicit solvers is common in relation of tensile bond strength to the initial shear strength.
both FEM packages when several loading regimes are be- The gaping shown on the right in Figure 14 must not lead to
ing applied in sequence. For example, it is essential, before failure in the case of confined masonry, but rather to an
imposing the earthquake action, or wind action on the additional load of the frame.
model, to initialize the stresses and deformation state in the
structure which can be developed from gravity loads. The
application of gravity loads immediately together with earth-
Figure 14 Left: numerical crack pattern for masonry wall (monotonic static loading)
Right: Joint failure for shorter masonry unit without unit failure (normal stress, compression is positive)
7 Modelling of reinforced masonry meshes. In LS-DYNA [18] the following methods can be
employed for modelling the reinforcement of masonry:
Various modelling approaches have been proposed for
masonry. Nevertheless, the inclusion of the reinforcement (1) Truss elements tied to solids by one dimensional con-
into the model is still challenging, and fraught with difficul- tact
ties, consequently, reinforced masonry is still lacking in The 1D Contact was originally developed to offer bond
literature. In analogue with modelling strategies that have slip failure for modelling reinforced concrete. In addi-
been proposed for reinforced concrete models, the following tion, it is possible to employ this feature for reinforced
modelling strategies can be employed for the reinforced masonry. The principle of this contact model is to allow
masonry: discrete modelling and smeared modelling. the sliding of reinforcement nodes along masonry
7.1 Discrete modelling nodes, where the sliding initiates after the rebar
debonds. The bond model is assumed to be elastic
In discrete modelling, the reinforcement can be modelled by perfectly plastic.
means of bar elements, and masonry can be modelled us-
ing solid elements (2D or 3D). The nodes of reinforcement (2) Constrained Lagrange in solid
bars must be merged with masonry elements through the This constrained method has been developed for
shared nodes, Figure 15-a. modelling the fluid structure interaction and frequently
used to embed the reinforcement rebar inside concrete
The restriction to create shared nodes might result in some
element, Abu Odeh [1]. The reinforcement mesh main-
inflexibility in mesh generation. However, it is not quite accu-
rate to apply full bonding between the reinforcement and tained to be fixed within the solid elements. However,
masonry mesh. The bonding model can be represented by the bond slip failure has not been considered in this
dummy spring elements that connect the duplicated nodes formulation. For masonry, the reinforcement can be
from reinforcement and masonry. The spring element has treated as a slave material that is linked to the master
no dimension and serves only as a breakable linkage be- material of masonry by means of ‘constrained La-
tween reinforcement and masonry. Therefore, the failure grange in solid’. Both masonry and reinforcement
model is the most important part of spring element. mesh must be Lagrangian.
In order to avoid the restriction of node sharing between (3) Constrained spotweld
reinforcement and masonry, an embedded formulation can The spotweld provides a breakable connection for the
be introduced. In the embedded formulation, the intersection nodal points of the nodal pairs. The failure force at
points of reinforcement bar with the segments of masonry which the spotweld is failing can be regarded as the
elements are first identified and then used to create the pull-out force of reinforcement.
nodal locations of the reinforcement elements, Figure 15-b.
(4) Discrete beam elements
Brookes et al. [8] have utilized the partially constrained spar
formulation to model reinforcement independently from with nonlinear plastic discrete beam material to simulate
masonry. The connection between the reinforcement and failure of the beams.
masonry meshes was achieved through a non-linear bond Although several methods are available in LS-DYNA to build
element. The arrangement of reinforcement is automated embedded discrete models of reinforcement concrete, care
without the need for topologically consistent element should be taken when applying these methods on reinforced
masonry by considering the correct bonding behaviour be- crete modelling of reinforced masonry in ANSYS and LS-
tween reinforcement and masonry. DYNA.
1
NODAL SOLUTION
JUN 3 2007
sheared node between STEP=1
19:01:41
SUB =1
reinforcement and unit TIME=.05
S1 (AVG)
DMX =.898E-04
SMN =-688.535
SMX =3830
reinforced bar element
MX
unit finite element
MN
Y
X
Z
points of compatible
displacements between
reinforcement and unit
[4] Bakeer, T.: Collapse Analysis of Masonry Structures [16] Liu, G.R.: Mesh Free Methods, Moving beyond the
under Earthquake Actions. PhD Thesis, Dresden Finite Element Method, CRC press: USA 2003
University of Technology. Publication Series of the
[17] Lourenço, P.B.: Computational Strategies for Ma- [22] Rots, J.G.: Smeared and discrete representations of
sonry Structures. PhD-Thesis, Delft University of localized fracture. International Journal of Fracture,
Technology. Delft University Press: Delft, The Neth- 51 (1991) 1, pp. 45-59
erlands 1996
[23] Schlegel, R.: Numerische Berechnung von
[18] LSTC: LS-DYNA ver. 971 Keyword User's Manual, Mauerwerkstrukturen in homogenen und diskreten
Livemore Software Technology Corporation: Califor- Modellierungsstrategien. Dissertation, Bauhaus-
nia 2007 Universität Weimar: Weimer 2004
[19] Mann, W. & Müller, H.: Failure of Shear-Stressed [24] Schöps, P.; Steinborn, Th. & Höveling, H.:
Masonry- an Enlarged Theory, Tests and Application Schubtragfähigkeit von Mauerwerk aus Porenbeton-
to Shear Walls. Proceedings of the British Ceramic Plansteinen und Porenbeton-Planelementen.
Society, 30 (1982) 1, pp. 223-235 Forschungsbericht. Technische Universität Dresden,
Fakultät Architektur, Lehrstuhl für Tragwerksplanung
[20] Massart, T.J.: Multi-scale modeling of damage in AIF-Nr.14642BG. 2009.
masonry structures. PhD-Thesis, Technische
Universiteit Eindhoven: Eindhoven, The Netherlands [25] Stoll, V.: The Main Cupola of the Dresden Frauen-
2003 kirche. In International Quality Network – Tradational
and Innovative Structures in Architecture – First
[21] Mistler, M.: Verformungsbasiertes seismisches report, Eds. W. Jäger; A. Lippert; L. Rietzschel; D.
Bemessungskozept für Mauerwerksbauten. Wendland, From Research to Practice in
Dissertation, Technischen Hochschule Aachen: Construction, Publication Series of the Chair of
Aechen, Deutschland 2006 Structural Design, Dresden University of Technology:
Dresden 2003, pp. 169-189