You are on page 1of 13

SEISMIC BEHAVIOUR OF HISTORIC MASONRY BUILDINGS

K. Meskouris; C. Butenweg; M. Mistler; W. Kuhlmann


Chair of Structural Statics and Dynamics
RWTH Aachen University, D-52056 Aachen, Germany

1. SUMMARY
There is always the danger of historic masonry buildings in seismically active regions being severely damaged
by earthquakes, since they certainly have not been explicitly designed by the original builders to withstand
seismic effects, at least not in a “scientific” way from today’s point of view. The assessment of their seismic
safety is an important first step in planning the appropriate interventions for improving their pertinent resis-
tance.
This paper presents a procedure for assessing the seismic safety of historic masonry buildings based on meas-
urements of their natural frequencies and numerical simulations. The modelling of the brittle nonlinear behav-
iour of masonry is carried out on the micro- and macro-level. On the micro-level, the units, the mortar and the
unit-mortar interface are modelled separately to obtain the homogenized material parameters and the failure
criterion. These calculated parameters serve as input values for the nonlinear macro-model which is then used
for simulating the nonlinear structural behaviour of the masonry building in question on the system level under
earthquake loading.
As an example, a recently completed investigation of the seismic behaviour of the Aachen Cathedral is given,
this being the first German cultural monument to be included in the UNESCO cultural heritage list in 1978. Its
construction goes back to the 9th century AD and it is considered as one of the finest examples of religious
architecture in Central Europe. The investigation is based on measurements of the natural frequencies at dif-
ferent positions and numerical simulations using a detailed finite element model of the Cathedral.

2. INTRODUCTION
Masonry, as the oldest building material, has been widely used within living memory because of the ease in
assembling its constituent parts “by hand”. The techniques are essentially the same as the ones developed some
thousand years ago. Apart from its aesthetic appearance, structural advantages of masonry such as its durabil-
ity with low maintenance costs explain why many famous masonry structures in Europe like Romanesque
structures and ancient cathedrals have successfully withstood the test of time.
In terms of their load-carrying behaviour, the study of existing masonry structures is complicated in that there
exist many variations of materials and building techniques. The former vary heavily, so that detailed informa-
tion about a specific structure is usually not available. Most of the masonry structures were built based on em-
pirical data and experience passed on from masters to apprentices, without any mathematical analysis, but with
great practical skill. Often the structural form suggests a rudimentary grasp of the basic nature of the acting
forces, but even today our knowledge about masonry load-behaviour patterns is not as well-developed as with
other materials. Many calculation methods for capacity assessments hardly ever consider the complex behav-
iour of masonry as a “composite” material. Since this paper intends to contribute to an improved capacity as-
sessment of masonry buildings, in particular historical structures, a brief introduction to some historical facts is
given first. It is followed by a description of masonry models considering three different levels of modelling
and after that the description of an integrated concept is presented, which supports all steps from the determi-
nation of the material parameters to the simulation of the whole structure. Finally, the investigation of the
seismic behaviour of the Aachen Cathedral is presented as a practical application.
3. MASONRY COMPONENTS AND TYPES
Building with masonry is certainly one of the most important achievements of early civilizations. Bricks, the
oldest man-made construction material, have been in use since 4000 AD in Mesopotamia. Later civilizations
have developed various techniques and layouts for masonry walls. Especially the Romans have added impor-
tant enhancements to masonry building techniques and opened the way to vast new application areas.
The oldest and most basic type of masonry is stone masonry, consisting of entire or cleaved stone blocks. Since
this type of masonry can be erected without chiseling work, it has been the standard type for masonry walls in
rural regions for a long time. Early structures have been erected without any interlocking by grout, while later
clay or lime mortar has been used. Adobe bricks have also been a very common building material, at least in
the early stages of the Greek civilization..
In spite of the large number of adobe buildings, the characteristic masonry technique of this time, in which the
Mycenic culture was at its apex in Greece, was Cyclopic masonry. It consists of more or less accurately
trimmed natural stones that have been placed without mortar and with joints as small as possible. Due to im-
provements in stone processing techniques, this later evolved to a polygonal pattern of joints (von Wölfel [17]).
The Greeks also have developed the dressed stone masonry, which became the standard way for constructing
monumental buildings in both the late Greek civilization as well as in the Roman Empire. Stones were con-
nected without mortar, but with iron or bronze anchors for mechanical interlocking. At the beginning it was
only possible to handle stones of moderate hardness, but with wider availability of harder materials for chisels,
the Romans have also used harder natural stones, such as granite or basalt.
Brick masonry has been employed in Greece since the 7th century AD. Primarily employed for urban habita-
tions, brick masonry was typically used with lime mortar. Several standardized formats of bricks have been
developed as well as various masonry types. The construction form using ‘opus caementitium’ or ‘Roman Con-
crete’ evolved as a mixture of sand, lime, bricks and added stone material and it is very similar to modern
concrete. Knowledge of the hydraulic properties of added puzzolan has allowed the Romans to construct ma-
sonry water pipelines or even underwater buildings, but with the decline of the Roman Empire this knowledge
disappeared in almost all of Europe until the discovery of ‘Trass’ as a new puzzolanic binder in the middle of
the 17th century.
The Romans have often combined the new technique of ‘opus caementitium’ with conventional masonry con-
struction. Working with the Roman concrete has speeded up the construction process, with, however some
esthetic demands sacrificed along the line. To counteract this, an outer (visible) layer from natural stones was
often used, while poured material served for the inner, invisible parts of the walls. Depending on the material
used for the outer layer of the masonry wall, it was called ‘opus incertum’ (dressed stones) (Figure 1a) or ‘opus
implectum’ (quarrystones) (Figure 1b). Also, special shapes such as the ‘opus mixtum’, a mixture of dressed
stones and bricks, were developed (van Wölfel [18]).

a) b) c)
Figure 1: ‘opus incertum’ (dressed stones), ‘opus implectum’ (quarrystones), multi-layer assembly
This multi-layer form has been the standard type for masonry walls in Europe’s middle ages over many centu-
ries. The outer layer consisted of quarrystones, dressed stones or bricks up to 50 cm thick. The inner part con-
sisted of poured or loose stonewall masonry. Sometimes a third layer was provided to take care of aesthetic
aspects (Figure 1c). For the inner layer that was not exposed to the elements, inferior quality materials were
used, e.g. gypsum mortar in northern Europe.

4. MATERIAL MODELS OF MASONRY


Engineers usually describe a structure by means of computer models which are mostly geared to a linear analy-
sis. For masonry structures, this does not describe the structural behaviour correctly and should be employed
only in a preliminary analysis, if realistic results are to be expected. For more detailed analyses, the use of
nonlinear material models is necessary in order to mirror the nonlinear anisotropic behaviour and the post-
peak load-carrying capacity which may be the key to why masonry buildings have withstood seismic loading in
the past. Unfortunately, the definition of the required material parameters is very complex and furthermore the
numerical models must be able to simulate all pertinent failure modes due to different kinds of loading.
Considered from the macroscopic point of view, the following failure modes may occur:
Shear failure: This failure mode is usually characterized by cracks in the bed and head faces. In the case of
low-strength bricks, the cracks typically bisect the stone. This failure mode exhibits a ductile behaviour com-
pared to the other modes (Figure 2a).
Friction failure: This failure mode corresponds to horizontal sliding, which is apt to occur when low vertical
dead loads are combined with high horizontal loads (Figure 2b).
Bending failure: This mode occurs in slender panels with high compression areas at the support corners. The
higher the vertical loads, the more brittle the failure mode (Figure 2c).

a) b) c)
Figure 2: Macroscopic failure modes of a masonry panel:
shear failure (a), friction failure (b), failure due to bending (c)
Since the direction of the horizontal loading changes during an earthquake, these modes will generally occur
in both directions and alternate during the loading process.
Considered from another point of view, the assembly of bricks and mortar is subjected to biaxial stress states.
For an ultimate limit state design and for assessing the capacity of the structure, an in-plane failure criterion
for masonry is required; however, failure modes and strength characteristics of masonry are highly sensitive to
the orientation of the principal stresses with respect to the joint plane. This anisotropy is mainly due to the
relative weakness of the mortar compared to the units and it may be even more prominent if the units exhibit
anisotropic strength properties because of perforations. Thus, to define masonry failure, a three-dimensional
surface in terms of the two principal stresses and their respective orientation to the bed face is required.

4.1. MODELLING LEVELS

Masonry can be modelled based on several different approaches (Figure 3), with each method having its own
merits and drawbacks:
• Modelling each brick, mortar face and interface separately
A large number of micro-models were developed for simulating monotonic as well as cyclic loading in the
past. Normally, each component is endowed with its own nonlinear behaviour and the bond between the
brick and the mortar is considered by using special joint or contact elements which are able to simulate the
gaps. A modelling prerequisite is the exact knowledge of the actual geometry of the whole wall.
• Modelling the bricks and the interface jointly
Since the real brick-mortar joint behaviour is often not known, discrete joint elements of zero thickness
with smeared characteristics may be used, so that grout elements do not need to be generated. The
dimensions of the bricks have to be expanded, which has a bearing on their stiffness, and the accuracy of
the model is lower. Brick failure due to transverse tensile stresses cannot be simulated in a 2-D model, but
the global behaviour is described satisfactorily.
• Using a smeared model with homogenized properties
Here the masonry is modelled as an anisotropic continuum. This model is computationally less expensive
and allows the structural behaviour of large substructures subject to dynamic loading to be investigated. In
order to use this method, the effective stiffness of the masonry and the smeared nonlinear characteristics
must be known, requiring the definition of stress-strain relationships for monotonic and cyclic loading. In
addition, the global failure criterion has to be defined.
• Using macro-elements to simulate additional components such as walls and panels
Such elements are not in general use because their input variables must be calibrated on the basis of shear
panel tests in which each element has different characteristics.

brick
head joint

bead joint

brick Mortar brick Smeared joint Smeared, homogenous Macro-element


Joint
brick-mortar

Figure 3: Different approaches for modelling masonry


In conclusion, even though the accuracy of the first approaches mentioned above is much higher than of the
latter, they are much more expensive in terms of computational costs and the corresponding high number of
degrees of freedom limits the applicability. Only for small structure may an accurate model be used with confi-
dence, while whole buildings are almost impossible to simulate by means of such micro-models. Smeared mod-
els as well as macro-element based models are less accurate by better suited for this purpose; however, they still
require additional effort for their calibration which is mostly based on (expensive) experimental results.

4.2. INTEGRATED CONCEPT BY HOMOGENISATION

As already mentioned, the problems in using smeared material models lie not only in the inherent complexity
of their nonlinear behaviour, but also in the necessity of their calibration, meaning the determination of some
material parameters by testing. The experimental costs can be reduced if the effective, nonlinear material prop-
erties can be derived from the characteristic values of the constituents (which are known for a wide variety of
brick and mortar types) in order to use them as input variables of the smeared model.
In the following, an integrated concept is described which supports all steps, beginning from the definition of
the material parameters to the simulation of the whole structure, by means of an homogenisation technique. It
is based on modelling the realistic behaviour by means of a small cell for deriving the necessary input data for
a smeared model. It is a general procedure for numerically formulating the failure criterion of masonry derived
directly from the material parameters of the constituents. Since the procedure is independent of the used mate-
rial law of each constituent, it can be applied to all kinds of masonry.

Homogenisation
There are several homogenization techniques available, which may be broadly classified in two different
groups, namely “two-step” and “single-step” methods. In the first type, the procedure for estimating the effec-
tive elastic stiffness is divided into two steps: First, brick units and head joints are homogenized horizontally
and then this layer and the continuous bed joints are homogenized vertically or vice-versa. The drawbacks of
the two-step homogenization are that results depend on the order of the steps and sometimes differ markedly.
A single-step approach was used e.g. by Anthoine [3] for deriving numerically the overall elastic constants by
applying the homogenisation theory of periodic media to masonry. In the following, the essential steps are
explained and extended to the inelastic range.
ν2

h
ν1

l d

Figure 4: Basic cell for running bond masonry


Masonry may be considered as a periodic composite continuum. The periodicity may be characterized by a
frame of reference of two skew vectors v1 and v2 (Figure 4). As a consequence, it is enough to define the me-
chanical properties of the medium on a small domain (unit cell) which can be moved about without rotation
both in the horizontal and in the vertical direction replicating the masonry pattern. The most natural cell type
is the parallelogram (or rectangle for stack bond), the sides of which are the vectors of the reference frame. The
boundary of this cell consists of two pairs of identical sides obtained through a translation along v1 or v2. In
order to identify the macroscopic behaviour a part of the wall this cell is subjected to a homogeneous stress
state. Here, all cells of the wall are loaded uniformly and thus deform in the same way. In particular, two adja-
cent cells must still fit together in their deformed state. In mechanical terms, this means that when passing
from one cell to the next the stress vector s ⋅ n must be continuous and therefore opposite on opposite sides
and that the strains must be compatible. To ensure the latter it is necessary that deformed opposite sides fit
together. The rotational component vanishes because of the compatibility condition. This strain-periodic dis-
placement field may always be written in the following form:
u = E ⋅ x + u p (x) (1)
where E is the mean strain tensor 〈e(u )〉 . The antisymmetric part of E corresponds to a rigid rotation of the cell
and that’s why only the symmetric part of E must be considered. up is a displacement field which assumes
equal values on opposite sides of the boundary. Since u is strain-periodic, the term u − 〈e(u )〉 ⋅ x is also peri-
odic. The easiest way to derive E is from Eq. (1). For the displacements of two opposite points the following
holds:
u p ( x + v1 ) = u p (x ) ⇒ u ( x + v1 ) − u (x ) = E v1 (2)

By writing down this equation for another pair of points related by the vector v2 one gets a linear system giving
E and thus the symmetric part of E in terms of v1, v2 and of the displacements u at four particular points of the
cell boundary. This way of deriving E from u remains valid even if the basic cell includes holes and/or special
media (interfaces, joints, bars, etc.), see Anthoine [3]. Concerning the mean stress tensor, it should by defini-
tion coincide with the stress tensor Σ , applied to the specimen. One obtains a linear system for Σ by formulat-
ing the equilibrium at each side of the basic cell. It is also valid even if the basic cell includes holes and/or
joints as opposed to the well-known definition of the average of the stress tensor of the cell, namely,
1
S ∫S
〈s 〉 = s ds (3)

which may also be applied, but is valid only if the basic cell is exclusively composed of two-dimensional con-
tinua without holes, otherwise it must be specifically modified.

Loading
The macroscopic loading which may consist of permissible combinations of macroscopic strain E and stress Σ ,
is imposed through displacements and/or forces at master points, e.g. the edge nodes of the basic cell P1 and/or
P2. In order to obtain the in-plane failure surface, the load path is defined by a constant ratio of the principal
stresses, expressed by a straight line through the origin in the principal stress space. The end point defines the
strength of the homogenized material for this load condition. The sum of all points yields the failure criterion
of the homogenized material expressed by the principal stresses.
For following these radial load paths even if the calculation is displacement-controlled in order to reach the
descending branch of the macroscopic Σ -E relationship, the following procedure is suggested: The macro-
scopic stress Σ and the strain E are described in the rotated reference system (I, II) combined with the re-
quested condition of Σ1 / Σ 2 = a = const, − ∞ < α < ∞ as:
Σ I 0   E11 E12 
S =
α ⋅ Σ I  E E22 
E = .
0  21
Introducing the energy S o E gives:
S o E = Σ I ⋅ E11 + Σ II ⋅ E22 = Σ I (E11 + α ⋅ E22 ) (4)
*
where the expression (E11 + aE22) can be regarded as a virtual displacement E to be imposed. The condition
now reads:
( 1 + α )E * = E11 + αE22 , α ≠ −1
E * = 0,5 ⋅ E11 − 0,5 ⋅ E22 , α = −1 . (5)
Since the failure criterion depends on the direction to the bed joint angle θ, the expression becomes:
( ) ( )
(1 + α ) E * = E xx cos 2 θ + α sin 2 θ + E xy sin 2θ (1 − α ) + E yy sin 2 θ + α cos 2 θ , α ≠ −1 ,

E* = 0 ,5 ⋅ (E xx − E yy )cos 2θ + E xy sin 2θ , α = −1 . (6)

wherein the macroscopic strain E can again be expressed by the displacements of P1 and P2.

Material models
In a plane masonry wall the thickness is much smaller than its length and height. The numerical simulations
may thus be carried out in two-dimensional space under the generalized plane strain assumption with a con-
stant strain e33 perpendicular to the x1 x2-plane. Any two plane sections perpendicular to x3 still remain plane,
but may translate proportionally to their relative distance. Since no load is applied on the external faces of the
considered masonry wall, the value of e33 is obtained by setting the resultant force acting perpendicular to the
x1, x2-plane equal to zero, see Anthoine and Pegon [5]. Another possibility would be the assumption of a plane
stress state, which however does not yield better results – quite the contrary (Anthoine [4]).
The homogenisation procedure is independent from the adopted failure criteria of each single component.
Nevertheless, it is very important to assign the correct failure conditions when choosing a material model for
each continuum. Failure criteria may be set up as single- or multiparameter models. Since the behaviour of
many non-metallic materials such as rocks and concrete is characterized by their hydrostatic pressure depend-
ency, the two-parameter models deem to be the simplest type of appropriate models. In the approach used a
hyperbolic yield function was employed based on the Mohr-Coulomb model. It is rounded in the meridional
and in the octahedral plane. The resulting yield stress is continuous and differentiable for all stress states and
the Mohr-Coulomb yield surface can be modelled as closely as desired. The yield function F is described in
Abbo [1]. A perfectly elastoplastic material behaviour gives an upper limit of the strength of the material. Sof-
tening effects should be considered in order to evaluate the post-peak behaviour; the Mazar damage model does
that quite efficiently (Mazar [14]).

4.3. NONLINEAR SMEARED MODELS FOR MASONRY

Hypoelastic model using the principle of equivalent uniaxial strains


This nonlinear model is based on the biaxial orthotropic hypoelastic concrete model of Darwin and Pecknold
[6] as it has been modified for application to masonry by Vratsanou [19]. It uses the principle of equivalent
uniaxial strains as a simplification of the complex biaxial material behaviour. The constitutive relation is de-
scribed for each principal stress direction by means of the uniaxial stress-strain relation and the equivalent
uniaxial strains are fictitious strains in the principal stress directions. The principal stresses correspond to the
stresses of a fictitious uniaxial state. An important advantage is that the uniaxial stress-strain relationships and
other required material characteristics can be derived by means of uniaxial tests. The failure criterion used is
the one derived by the experimental tests of Page [15, 16]. For simplicity, a constant angle of θ = 45° is as-
sumed.
The monotonic stress-strain relationship (Figure 5a) consists of three ranges. The tension region is defined by
a straight line OT ending at the point of the maximum tensile strength T (εt, ft ). The compression region con-
sists of the increasing range OC, defined by the compression strength C (εc, -fc), and the decreasing range CU
which describes the softening effect by a linear behaviour until U (εf, σf). In the numerical application the
residual compression strength is set equal to 1/1000 · fc in order to avoid computational problems.
−σ −σ

E0 E0
-f c -fc
C C A
B

σf σf U
U S

ε f = C ε εc
εf = C ε εc
σf = Cσ (-f c)
σf= C σ (-fc)

O O E0
ε εt −ε P
−ε
εc εf εt εc ε P εf
T ft ft
T

a) b)
Figure 5: Stress-strain relationships for monotonic (a) and cyclic loading (b)
Concerning the stress-strain relationships for cyclic loading it is important to distinguish between the unload-
ing paths before and after exceeding the compression strength (Figure 5b). In the first case the unloading path
is a straight line defined by the elastic modulus E0 and tensile stresses are still possible. In the second case the
unloading path doesn’t reach the tensile region. After exceeding the maximum tensile strength ft cracks occur
perpendicular to the principal stress direction. Based on the smeared crack model a smeared crack width is
then calculated. This model was implemented into the finite element program ANSYS [2].
Smeared continuum model based on a nonsmooth multisurface plasticity theory
This smeared elasto-plastic continuum model is based on the nonsmooth multisurface plasticity theory and
includes anisotropic elastic and inelastic behaviour. The fundamental notion of the plasticity theory is the exis-
tence of a yield function f as boundary to the elastic domain. For masonry, the failure criterion proposed by
Ganz [9] or by Mann/Müller [12] can be used. The surface (Figure 6) itself is composed of several single fail-
ure criteria. Each criterion describes a different failure mode of masonry, either component or bond failure.
t xy
sy
sx

- fmy

- fmx

Figure 6: Failure criterion of unreinforced masonry after Ganz (1980)


It is assumed that the total strain vector can be divided into an elastic and a plastic component. Yielding can
only occur if the stresses satisfy the general yield criterion
f (σ , κ ) = 0 (7)
where the yield stress is a function (hardening law) of the scalar κ which is introduced as a measure for the
amount of hardening or softening. It influences the expansion and/or the reduction of the initial yield surface
dependent on the load path, as well as on the translation of the yield criterion in the stress space. The introduc-
tion of separate softening functions for each strength parameter makes it possible to formulate an orthotropic
softening model that depends on the failure mode. The yield domain, its hardening and softening laws are
defined based on experimental results. Thus, it becomes possible to simulate masonry-specific failure and dam-
age mechanisms. For the implementation of the model at the integration point level the return mapping proce-
dure is used: First, an elastic predictor step calculates a “trial” stress state, and if this state is not inside the
yield domain, a local iteration projects the stress state on the actual yield surface. The big advantage of the
multi-surface plasticity is that it allows different failure mechanisms of masonry to be considered, which may
also act simultaneously.

4.4. EXAMPLES

Verification of the homogenisation


For the verification of the homogenisation tool the tests carried out by Page [15] were numerically modelled.
The basic cell of the running bond pattern is a parallelogram (Figure 7). A refined approximation of the ma-
sonry wall is obtained by repeatedly duplicating the cell. The nodes on the boundary must still coincide after
applying the loads.

Figure 7: Check of compatibility


The elastic properties of masonry are numerically calculated for uniaxial loading, e.g. depending on the orien-
tation angle ? between the bed joint plane and the direction of the acting force. Assuming YOUNG’s moduli of
2 and 10 kN/mm² for mortar and brick, respectively, the elastic modulus of the composite is as shown in Figure
8. Additionally, the applied nodal forces for uniaxial loading are also shown for θ = 70°.

8000
E-modulus [N/mm²]

7500

7000

6500

6000
0 15 30 45 60 75 90
Angle [°]

Figure 8: E-modulus depending on the orientation angle θ and nodal forces for the case of θ = 70°
The maximum of the principal stress is used as failure criterion. The obtained surfaces in terms of the two
principal stresses and their respective orientation to the bed joints are shown in Figure 9, here for the two di-
rections of 22.5° and 45°. The squares denote the results of the experiments of Page, the circles are the results
of the current simulation. Not only do the shapes of the failure criterion correspond very well with the experi-
mental results of Page, but also the failure mode agrees with the test results. Although the simplest type of
pressure dependent failure model is used, a very good agreement is reached.
1 1

-11 -9 -7 -5 -3 -1 -1 1 -11 -9 -7 -5 -3 -1 -1 1

-3 -3

-5 -5

-7 -7

-9
-9 PAGE

-11 continuum model


-11

a) b)
Figure 9: Simulation and experimental results for directions of 22.5° (a) and 45° (b)

Verification of the macro model


For the verification of the smeared hypoelastic model the experimental data of a shear wall test after Lurati und
Thürlimann [11] were used. The wall stands on a rigid concrete foundation. First, a vertical dead load of
419 kN is applied, distributed by means of a reinforced concrete beam. Afterwards, the drift d is increased in a
displacement-controlled experiment. Figure 10 shows the FE-discretization with plane shell elements. At the
foundation, the horizontal DOF’s are constrained, whereas vertical spring elements allow a gap to form. The
agreement between the experimental data and the numerical results in terms of the pushover curve is very
good.

400
p 350

d 300
[kN][kN]

250
force
Schubkraft

200
Horizontal

Plane 150
Scheibenelemente
elements
100
Numerical results
numerische Ergebnisse
50
Y Experimental results
experimentelle Ergebnisse
Z X 0
0 2 4 6 8 10 12 14
Horizontal displacement
Horizontalverschiebung [mm]
[mm]
Spring elements

Figure 10: Finite element model and pushover curve

5. SEISMIC BEHAVIOUR OF THE AACHEN CATHEDRAL


After more than 1200 years of permanent use, the static system of the Aachen Cathedral shows considerable
damage of the masonry, the medieval anchoring system, the roof construction, the vaults and the pillars.
Therefore, rehabilitation measures of the cathedral's load bearing system were carried out in the 20th century.
Due to the location of the cathedral in a seismic active region, an assessment of its seismic behaviour, as pre-
sented in the following, became necessary. The investigations were based on a detailed finite element model of
the cathedral which was calibrated by measurements of its natural frequencies. Numerical simulations were
carried out by using nonlinear time history analyses.

5.1. STATIC SYSTEM

Currently the cathedral consists essentially of the West Tower, the Octagon and the Choir Hall (Figure 11a, b).
The most important aspect of the static system in terms of a horizontal earthquake excitation is the common
lateral load carrying system of the Choir Hall and the Octagon. Due to its slender columns and the glassed
surfaces, the Choir Hall is not able to carry horizontal loads to the foundation. Therefore several anchor sys-
tems (Figure 12) have been installed over the last centuries to ensure the structural integrity of the cathedral
under horizontal loads.
Nikolas Chapel
Chapel of Charles
and Hubert

Choir Hall

Mathias Chapel
Octagon
West Tower
Anna Chapel
Hungarian Chapel

a) b)
Figure 11: Side view (a) and plan view (b) of the Aachen Cathedral
Historic anchors: Originally, five iron anchors were installed for the connection of the Choir Hall to the Oc-
tagon. These anchors are now supposed to be inactive due to corrosion and inappropriate repair works during
the last centuries.
Pirlet-anchor: The Pirlet-anchor, which was installed in the 1920's, ensures nowadays the integrity of the
Choir Hall. The main anchor in longitudinal direction consists of four prestressed steel L-profiles. At the east
end of the Choir Hall, the anchor is connected to several circularly aligned steel anchors which carry the an-
chor forces to the outer columns. Above the vault of the Octagon, the main anchor is supported by a space truss
system.
Further concrete and steel anchors: In addition to the Pirlet-anchor a concrete anchor surrounding the Choir
Hall has been introduced. Furthermore, new steel anchors have been added in recent years to the outside of the
Choir Hall at approximately the height of the historic anchors. Around the Octagon, a new steel anchor is
being installed right now.
Pirlet anchor (longitudinal) Star-shaped end
Pirlet anchorage system

Concrete
anchor
around the
Choir Hall

Anchors
around the
Octagon Anchors around the
Transverse anchors Choir Hall (historic
in the Choir Hall and new)

Figure 12: Sketch of the anchor system

5.2. FINITE ELEMENT MODEL

The development of the numerical model started with the generation of a geometry model based on the infor-
mation taken from construction plans, photographic material and measurements on site by using the finite
element pre- and postprocessor FEMAP [8]. In the next step, the geometry model was meshed and exported to
ANSYS [2]. Figure 13a shows the finite element model of the cathedral with a total number of 20130 ele-
ments. The boundary conditions of the model were introduced as rigidly clamped at the bottom and do not
include any soil structure interaction effects. The different historic materials, which were considered in the
numerical model, are shown in Figure 13b. The precise values of the material properties of the single stones
types can be taken from Kuhlmann et al. [10].

Herzogenrathian sandstone
Bluestone
Calcareous tufa face wall
Carolingian masonry
Luxembourgian sandstone face wall
a) b)
Figure 13: Finite element model (a) and historic material distribution (b)
An important aspect of the modelling was the realistic representation of the column cross-sections because
most of the columns consist of two cross-sectional parts with different material properties. As an example, the
cross-section of the slender columns of the Choir Hall is shown in Figure 14a. The core of the columns consists
of a conglomerate of mortar, sand and chippings with low strength and the outer part is made up of sandstone
(Herzogenrathian sandstone) with higher strength. In the finite element model, the columns were idealized by
special beam elements with a subdivision of the cross-section into different cells (Figure 14). For each cell the
material properties were defined separately and a rigid bond was assumed between the cells. In case of the
columns, the core was simply assumed to be an elastoplastic material and a smeared nonlinear material model
was used for the cells of the outer part.
Smeared linear elastic material behaviour

Nonlinear stress-strain relation

Exemplar of a real
column cross section
Inner part: conglomerate of
mortar, sand, stone chippings
Outer part: blocks of granit

Figure 14: Column cross-section of the Choir Hall and model idealization

5.3. RESULTS

Natural frequencies and mode shapes


The first natural frequency (Figure 15a) corresponds to a translational mode of the Choir Hall in which the end
of the hall vibrates perpendicularly to the longitudinal axis of the cathedral. This natural mode results from the
weakness of the Choir Hall due to its slender columns and glassed surfaces. The second natural frequency
(Figure 15b) corresponds to a translational mode in the longitudinal direction of the cathedral with the whole
cathedral participating in the movement.
(a) (b)
Figure 15: (a) First eigenmode (plan), (b) Second eigenmode (elevation)

Stresses in the anchorage system


The behaviour of the cathedral under earthquake load was studied with nonlinear time history analyses. For
this purpose, time histories compatible to the response spectrum in the German code E DIN 4149 [7] were
generated with the programs ESYN and BASKOR (Meskouris [13]). One of the resulting synthetic accelero-
grams with a duration of 10 seconds is shown in Figure 16a. In Figure 16b the stress in the main Pirlet-anchor
over time are shown. The maximum and minimum stresses of 16.9 and 11.8 MN/m² correspond to about 10%
of the Pirlet-anchor capacity because the anchor has been designed very conservatively. Since the stresses in
the other anchors are also lower than the maximum permissible stresses the structural safety of the cathedral’s
lateral load carrying system is ensured.
(x10**4)

0,8 1920

1840
0,6
Acceleration [m/s²]

1760
0,4
1680
0,2 1600

0 N/m2 1520

-0,2 0 2 4 6 8 10 1440

-0,4 1360

1280
-0,6
1200
-0,8 1120
Time [s] .5
1.46
2.42
3.38
4.34
5.3
6.26
7.22
8.18
9.14
10.1

TIME

(a) (b)
Figure 16: Synthetic accelerogram (a) and time history of the stresses in the Pirlet-anchor (b)

6. SUMMARY
The paper presents a procedure for assessing the seismic safety of historic masonry buildings. An integrated
concept which supports all steps starting with the determination of the material parameters till the simulation
of the whole structure is described and the practical use of the developed methods is demonstrated by means of
the investigation of the seismic behaviour of the Aachen Cathedral.

7. REFERENCES
[1] Abbo, A.J. and Sloan, S.W.: A smooth hyperbolic approximation to the Mohr-Coulomb yield crite-
rion. Computers and Structure 54, 3, p. 427-441 (1995).
[2] ANSYS, www.ansys.com, FE-Software, SAS IP Inc. (2003).
[3] Anthoine, A., Derivation of the in-plane elastic characteristics of masonry through homogenization
theory. International Journal of Solid Structures, 32(2), p. 137-163 (1995).
[4] Anthoine, A.: Homogenization of periodic masonry: Plane stress, generalized plane strain of 3d mod-
elling? Communications in numerical methods in engineering, Vol. 13, 319 – 326 (1997).
[5] Anthoine, A., Pegon, P.: Numerical analysis and modelling of damage and softening of brick ma-
sonry. Chapter 7 of: Numerical Analysis of modelling of composite materials. Blachu Academic &
Professional, Chapman & Hall, Glasgow (1996).
[6] Darwin, D.; Pecknold, D.A.: Inelastic model for cyclic biaxial loading of reinforced concrete. Univer-
sity of Illinois at Urbana-Champigaign, Urbana/Illinois (1974).
[7] E DIN 4149, Normenausschuss Bauwesen (NABau) im DIN Deutsches Institut für Normung e.V.,
Entwurf Oktober (2002).
[8] FEMAP: Finite Element Modeling and postprocessing, Version 8.2 (2002).
[9] Ganz, H. R.: Mauerwerkscheiben unter Normalkraft und Schub. ETH Zürich, Institut für Baustatik
und Konstruktion. Birkhäuser Verlag Basel (1985).
[10] Kuhlmann, W., Butenweg, C., López, M., Fernández, S.: Seismic Vulnerability Assessment of the
Historic Aachen Cathedral, Conference proceedings of the 13th World Conference on Earthquake En-
gineering, Vancouver (2004).
[11] Lurati, F., Thürlimann, B., Versuche an Mauerwerkswänden aus Zementstein, Institut für Baustatik
und Konstruktion der ETH Zürich, Bericht Nr. 8401-3, Birkhäuser Verlag Basel (1990).
[12] Mann, W., Müller, H.: Schubtragfähigkeit von Mauerwerk. Berlin, Ernst & Sohn (1978).
[13] Meskouris, K., Hinzen, K.-G.: Bauwerke und Erdbeben, Vieweg Verlag (2003).
[14] Mazar, J.: A Description of Micro- and Macroscale Damage of Concrete Structures. Engineering
Fracture Mechanics, Vol.25, No. 5/6, p. 729 – 737(1986).
[15] Page, A.W.: The biaxial compressive strength of brick masonry. In: Proc. Intsn. Civ. Engrs., Part 2,
71, p. 893-906 (1981).
[16] Page, A.W.: The strength of brick masonry under biaxial compression-tension. In: Int. J. Masonry
Constr., 3(1), p. 26-31 (1983).
[17] von Wölfel, W.: Mauerwerkstechnik in der griechischen Antike, das Mauerwerk (2000).
[18] von Wölfel, W.: Bautechnik im Römischen Reich, das Mauerwerk (2000).
[19] Vratsanou, V.: Das nichtlineare Verhalten unbewehrter Mauerwerksscheiben unter Erdbebenbean-
spruchung. -Hilfsmittel zur Bestimmung der q-Faktoren-. Diss. Universität Karlsruhe (1992).

You might also like