See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/368690910

Stochastic Load-carrying Capacity Assessment of Brick Masonry Arch Bridges

Conference Paper · July 2023

DOI: 10.1201/9781003323020-171

CITATION READS

1 106

3 authors:

Bowen Liu Imrose B. Muhit

University of Leeds Teesside University
13 PUBLICATIONS 54 CITATIONS 36 PUBLICATIONS 205 CITATIONS

SEE PROFILE SEE PROFILE

Vasilis Sarhosis
University of Leeds
207 PUBLICATIONS 4,090 CITATIONS

SEE PROFILE

All content following this page was uploaded by Imrose B. Muhit on 11 July 2023.

The user has requested enhancement of the downloaded file.

 Life-Cycle of Structures and Infrastructure Systems – Biondini & Frangopol (Eds)
© 2023 The Author(s), ISBN 978-1-003-32302-0
Open Access: www.taylorfrancis.com, CC BY-NC-ND 4.0 license

Stochastic load-carrying capacity assessment of brick masonry

arch bridges

B. Liu
School of Civil Engineering, University of Leeds, Leeds, UK

I.B. Muhit
School of Civil Engineering, University of Leeds, Leeds, UK
School of Computing, Engineering & Digital Technologies, Teesside University, Middlesbrough, UK

V. Sarhosis
School of Civil Engineering, University of Leeds, Leeds, UK

ABSTRACT: Masonry arch bridges form an integral part of our infrastructure network, and
their safety is important for the functioning of our society. Now, assessing the structural perform-
ance of ageing masonry infrastructure is a complex task. Existing ageing masonry arch bridges
are characterized by inherent variability and may have stochastic material properties even in the
same bridge. Hence, a realistic methodology for structural assessment of masonry arch bridges is
crucial to protect the ageing masonry bridges and utilize these resources efficiently. In this study,
stochastic-based assessment on the load-carrying capacity of masonry arch bridges has been per-
formed, which introduces material variabilities into a two-dimensional structural analysis model
based on the Discrete Element Method (DEM). Over 100 probabilistic analyses have been devel-
oped to assess the ultimate load that a bridge can carry when subjected to monotonic loading at
a quarter span. The bond properties of mortar joints, including cohesion, tensile strength, and
friction angle, were considered as stochastic variables following a normal distribution. The com-
putational results were compared against the experimental results obtained from the literature.
From the results analysis, it was shown that the computational model considering the random
variability of bond strength properties can better predict the load-carrying capacity of the
masonry arch bridge than the deterministic one. The bond strength at the unit-to-mortar inter-
face significantly affected the ultimate strength of the masonry arch bridge.

1 INTRODUCTION

Masonry arch bridges form a significant part of the UK and European infrastructure network. It
is estimated that there are approximately 200,000 masonry arch bridges and culverts in Europe
and 40,000 masonry arch bridges in the UK. Most of these bridges were constructed between the
17th and 19th centuries and are still in service today (Forgács et al. 2021). After being used for
over a hundred years, the significant structural deterioration caused by several environmental
impacts may result in a reduction in the durability and load-bearing capacity of the bridges. On
the other hand, most of these old masonry arch bridges support traffic loads much higher than
those originally designed to carry, which can pose a serious threat to the safety of the bridges.
Thus, there is an urgent need to assess the in-service behavior and predict the load-bearing cap-
acity of masonry arch bridges to inform their repair, maintenance, and rehabilitation strategies.
Assessing the structural performance of ageing masonry arch bridges is challenging due to the
inherent variability in their material properties. So far, several computational approaches, ranging
from simple to high-fidelity, have been developed to predict the ultimate load-carrying capacity of

DOI: 10.1201/9781003323020-171

1392
masonry arch bridges (D’Altri et al. 2019). Within these numerical methods, a large number of
input parameters are required to describe the mechanical behavior of materials and define the
interactions between different elements. These parameters need to be carefully selected so that the
model can accurately reproduce the behavior of real structures (Sarhosis & Sheng 2014).
Now, in order to assess the influence of material variability on the ultimate load-carrying
capacity of masonry structures, high fidelity numerical modelling approaches coupling with
Monte-Carlo simulations have been used (Li et al. 2014, Muhit et al. 2022). For example, Sar-
hosis et al. (2019) proposed a methodology which considers the spatial variability of masonry
materials. The proposed method was integrated into a two-dimensional (2D) discrete element
modelling approach to predict the stochastic strength of masonry walls containing openings.
The study also carried out a series of sensitivity analyses to determine the factors affecting the
load-carrying capacity of masonry walls. Later, the approach was adopted and incorporated
within the three-dimensional (3D) Discrete Element Method (DEM) software for assessing
the mechanical behavior of other masonry structures (Pulatsu et al. 2022, Gonen et al. 2022).
Following the research carried out by Sarhosis et al. (2019), this paper presents a stochastic-
based assessment of the load-carrying capacity of masonry arch bridges subjected to static load.
More than 100 realizations were carried out on a masonry arch bridge model developed based on
the 2D DEM. Computational results were compared against the experimental findings obtained
from the literature. The stochastic parameters considered in the current study included the friction
angle, tensile strength, and cohesion at the unit-to-mortar interface, representing the bond proper-
ties of mortar joints. Moreover, sensitivity analysis was carried out to investigate the influence of
these mortar bond properties on the load-carrying capacity of the masonry arch bridge.

2 METHODOLOTY

The heterogeneity and inherent variability in material properties are the most remarkable charac-
teristics of ageing masonry arch bridges, which considerably affect the structural behavior and
load-carrying capacity of the bridge. However, most of the computational models in previous
research have failed to consider material variabilities. In other words, the same (uniform) mater-
ial parameters were usually assigned throughout the masonry domain by engineers/researchers
(Sarhosis et al. 2019). This study has adopted a probabilistic approach to investigate the influence
of random material properties on the mechanical behavior of masonry arch bridges. The frame-
work of the methodology adopted in this study is summarized in Figure 1.

Figure 1. Framework for introducing material variabilities in computational models.

1393
 Overall, the methodology involved introducing material variabilities into the computa-
tional model. As the first step, the critical parameters for the probabilistic analysis need to
be identified, and the mean values and probability distributions of these parameters can be
obtained from either small-scale laboratory/in-situ tests or the literature. Having assigned
deterministic and stochastic parameters to the computational model, simulations and sensi-
tivity analyses can then be executed. It is worth noting that the parameters used for the
probabilistic analysis should be selected based on the characteristics of the structure. For
example, as the low bond strength of mortar joints has been found to be one of the domin-
ant factors affecting the mechanical behavior of historical masonry arch bridges (Sarhosis
et al. 2019), the bond properties of the unit-to-mortar interface, including friction angle,
tensile strength, and cohesion, have been selected as random variables for probabilistic ana-
lysis in the study.

3 EXPERIMENTAL TESTING OF A MASONRY ARCH WITH BACKFILL

Full-scale static tests on brickwork masonry arch bridges were carried out by Augusthus-
Nelson et al. (2018, 2020) at the University of Salford. Figure 2 shows the experimental setup
and dimensions for the test chamber and the bridge model. The test was carried out under
plain-strain conditions with two stiff walled chambers to restrict any out-of-plane movement
of the fill and arch barrel. Class A Engineering bricks and a lime mortar with a mix ratio of
1:2:9 (cement:lime:sand) were used to construct the bridge model to produce the low bond
strength characteristic found in real masonry arch bridges.
The experimental model mainly contained an arch barrel constructed with a header bond
configuration. The span and thickness of the arch barrel were equal to 3 m and 0.215 m,
respectively, and the span-to-rise ratio of the arch was 4:1. The cohesionless MOT type
I limestone was selected as the backfill material, which was placed into the test chamber and
compacted layer by layer until it reached the 0.3 m over the crown of the arch. During the
test, a monotonic load was applied by a hydraulic actuator via a load-spreading beam to the
level surface of the backfill above the quarter of the arch barrel.

Figure 2. Masonry arch bridge model tested in the laboratory (Augusthus-Nelson et al. 2018).

4 DISCRETE ELEMENT MODELLING AND THE COMPUTATIONAL MODEL

Past experience has demonstrated that cracking in masonry arch bridges occurs at the unit-
to-mortar interfaces due to the low tensile and shear resistance of the mortar joints. This
characteristic makes the discontinuum modelling approach more appropriate compared
with other approaches (e.g., homogeneous ones) in simulating the cracking behavior and
failure mechanism of masonry arch bridges. In this study, the 2D commercial software, Uni-
versal Distinct Element Code (UDEC), based on the DEM, was used to develop the numer-
ical model (Itasca 2011).
Developing a model in UDEC starts with creating a single block covering the domain to be
analyzed. Then, the original block generated is discretized into several small blocks in accord-
ance with the geometry feature of the target structures. In the numerical model shown in
Figure 3, the arch barrel was assembled by several individual bricks, which were represented
by independent deformable blocks with a linear elastic behavior. Mortar joints were

1394
represented by zero-thickness interface elements between blocks as shown in Figure 4a. The
mechanical behavior of joints in the normal and shear directions (see Figure 4b and c) are sub-
jected to the following equations:

where ǻın and ǻIJs are the change in normal and shear stress, respectively; Kn and Ks stand
for the stiffness in the normal and shear directions, respectively; and ǻun and ǻus are the
change of displacement in the normal and shear directions, respectively.
The numerical model developed in the present work had the same geometry and dimen-
sions as the large-scale bridge tested in the laboratory. Table 1 lists the material and mech-
anical properties of masonry units, backfill, and backfill-to-arch ring interfaces in the
numerical model. These parameters were kept constant (deterministic) throughout the sto-
chastic analysis. On the other hand, material properties of mortar joints, including the ten-
sile strength (Jten), cohesion (Jcoh), and friction angle (Jfric) were considered stochastic to
isolate the influence of remaining parameters (i.e., backfill properties and backfill-to-arch
ring interface properties) on the load-carrying capacity of the masonry arch bridge and
highlight the effect of bond properties with inherent variability found in most real masonry
arch bridges. The mean and coefficient of variations (COV) of the stochastic parameters
are listed in Table 2 which were obtained from material characterization tests in the labora-
tory. Jcoh and Jfric were assumed to be normally distributed (Gaussian distribution) while
Jten was assumed fully correlated to Jcoh according to the following equation (Milani &
Lourenco 2013):

In terms of loading and boundary conditions of the computational model, the blocks repre-
senting the tank walls, basement, and abutments were considered fixed. A monotonic load
was applied by controlling a block with the same width as the load-spreading beam to move
downwards at a constant velocity. It is worth noting that adaptive damping was assigned to
the model to obtain a convergent static solution. In this way, the large displacement of blocks
was allowed with the sequential contact detection and update.

Figure 3. 2D numerical model of the masonry arch bridge based on the DEM showing meshing at the
backfill and arch barrel.

5 RESULTS AND DISCUSSION

Figure 5a shows a typical failure mechanism of the masonry arch bridge obtained from the
computational model. From all simulations, the same four-hinge failure mechanism was
observed, with high compressive stresses concentrated at the hinge locations, which was con-
sistent with the experimental observation (see Figure 5b). In particular, with the downward
movement of the load platen, a tensile crack initiated at the intrados of the arch underneath
the load platen, forming one hinge at the quarter of the arch. The other three hinges occurred

1395
Figure 4. (a) Contacts between blocks; Mechanical behavior of the zero-thickness interface at the (b)
normal and (c) shear direction.

Table 1. Material and mechanical properties of masonry units, backfill and backfill-to-arch interfaces
adopted in the numerical model (assumed).
Unit Young’s Poisson’s Friction Tensile
Material weight modulus ratio angle strength Cohesion

Units kN/m3 MPa - degree MPa MPa

Backfill 20 300 0.3 45 0 0.0005
Bricks 22 9 0.2 - - -
Backfill-to-arch - - - 35 0 0
interface

Table 2. Statistics of the bond properties of unit-to-mortar interfaces.

Property Symbol Mean value COV Distribution

Unit-to-mortar interface friction angle Jfric 35 degrees 0.06 Normal

Cohesion at the unit-to-mortar interface Jcoh 0.02 MPa 0.2 Normal
Tensile strength at the unit-to-mortar interface Jten Jten = Jcoh/1.4 - -

at approximately the two edges and three-quarters of the arch barrel. Furthermore, the inde-
pendence of the failure mechanism from the unit-to-mortar bond characteristics suggested
that the failure mechanism of the masonry arch bridge was governed by the boundary condi-
tions and bond configurations of the arch barrel, rather than by the material properties of
mortar joints.

Figure 5. Failure mechanism and deformation of the masonry arch bridge obtained from the (a) com-
putational (magnitude factor is 5) and (b) experimental model.

Figure 6 compares the numerical results against the experimental load-displacement rela-
tionships of the masonry arch bridge. The mean values for Jfric and Jcoh were equal to 35
degrees and 0.02 MPa, respectively, and the respective COV of these two parameters were
equal to 6% and 20%. As the load platen moved downwards, the load carried by the bridge
increased until it reached a peak. The fluctuations and reductions observed in the load-
displacement curves were related to the crack initiation and propagation process, as well as

1396
the stiffness degradation of the bridge under loading. The results also illustrated the influence
of the stochastic bond strength properties on the load-carrying capacity of the masonry arch
bridge. In principle, a relatively large range of ultimate loads, from 124.33 kN to 273.65 kN,
was obtained from the computational model with stochastic bond strength properties. The
experimental model had an ultimate load of 142.7 kN, which fell within the range obtained
from the computational model and was closer to the lower limit of the range.
Figure 7 shows the sensitivity analysis results of the ultimate load on the bond strength of
the mortar joint. In numerical simulations, the mean values of Jcoh varied from 0.015 MPa to
0.04 MPa with a COV of 20%. The values of tensile strength and cohesion were fully correl-
ated according to Equation 3. For all the simulations, the mean value and COV for Jfric
remained at 35 degrees and 6%, respectively. In Figure 7, the red lines are trend curves for the
maximum, average, and minimum loads obtained from the simulations. It was calculated that
the standard deviations for the results of four groups with different mean values of bond
strength were all equal to approximately 40%, which highlighted the influence of stochastic
bond properties of the unit-to-mortar interface on the ultimate load of the masonry arch
bridge. With the stochastic properties of mortar joints, a range of peak loads was obtained
from the computational model with different mean values of cohesion and tensile strength.
Nevertheless, a significant positive correlation between the bond strength and ultimate loads
was observed. With the increase in the mean bond strength, the average load-carrying capacity
of the bridge increased, suggesting that the bonding strength of the mortar joints is one of the
dominant factors affecting the overall strength of the masonry arch bridge.

Figure 6. Experimental (red line) against numerical (grey lines) load-displacement curves.

Figure 7. Sensitivity of the ultimate load on the bond strength of mortar joints (the mean value of Jcoh
at the unit-to-mortar interface varied from 0.015 MPa to 0.04 MPa, with COV remaining at 20%).

1397
6 CONCLUSION

Ageing masonry arch bridges are characterized by inherent variability and may have sto-
chastic material properties even in the same bridge. Such random properties could result
in a considerable variation in the strength and stiffness of the masonry arch bridge. How-
ever, most of the previous studies have ignored this variation in material properties and
instead assigned the same (uniform) parameters through the masonry structure for simpli-
city, which can lead to an inaccurate evaluation of the ultimate strength of a real masonry
arch bridge.
This paper presents a stochastic-based assessment of the ultimate strength and mechanical
behavior of masonry arch bridges to address this issue. Stochastic bond properties of unit-to-
mortar interfaces were assigned to a 2D computational model developed based on the DEM,
and over 100 simulations were executed to predict the load-carrying capacity of the masonry
arch bridge. From the analyses of the results, it is shown that the computational model with
stochastic material properties can better predict the load-carrying capacity of a real masonry
arch bridge with inherent material variabilities. Moreover, the random bond properties did
not affect the failure mechanism of the masonry arch bridge investigated and all computa-
tional models had a four-hinged behavior. However, with the stochastic bond strength of
unit-to-mortar interfaces, the predicted ultimate loads of the bridge had a variation with
a COV of approximately 40%, which is expected as the mean value of bond strength increased,
the load-carrying capacity of the bridge increased significantly.
The study reported herein is the preliminary findings of the more sophisticated computa-
tional model being developed. Hence, such finding assumes and considers the simplest case,
while more complicated scenarios like spatially variable bond strength properties of masonry,
the correlation between unit-to-mortar interfaces in the same course and different courses,
complicated statistical relationships between strength properties, more sophisticated consider-
ations of the backfill materials, etc. can be considered in future studies.

ACKNOWLEDGEMENT

This work was funded by the EPSRC project “Exploiting the resilience of masonry arch
bridge infrastructure: a 3D multi-level modelling framework” (ref. EP/T001348/1). The finan-
cial contribution is very much appreciated.

REFERENCES

Augusthus-Nelson L. & Swift G. 2020. Experimental investigation of the residual behavior of damaged
masonry arch structures. Structures 27: 2500–12.
Augusthus-Nelson L., Swift G., Melbourne C., Smith C. & Gilbert M. 2018. Large-scale physical modelling
of soil-filled masonry arch bridges. International Journal of Physical Modelling in Geotechnics 18: 81–94.
D’Altri A.M., Sarhosis V., Milani G., Rots J., Cattari S., Lagomarsino S., Sacco E., Tralli A., Castellazzi G.
& de Miranda S. 2019. A review of numerical models for masonry structures. In B. Ghiassi & G. Milani
(eds), Numerical Modeling of Masonry and Historical Structures: 3–53. Woodhead Publishing.
Forgács T., Sarhosis V. & Ádány S. 2021. Shakedown and dynamic behavior of masonry arch railway
bridges. Engineering Structures 228: 111474.
Gonen, S., Pulatsu, B., Erdogmus, E., Lourenço, P.B. & Soyoz, S. 2022. Effects of spatial variability and
correlation in stochastic discontinuum analysis of unreinforced masonry walls. Construction and Build-
ing Materials 337: 127511.
ITASCA 2011. UDEC-universal distinct element code manual: theory and background. Itasca Consulting
Group, Minneapolis.
Li J., Masia, M.J., Stewart, M.G. & Lawrence, S.J. 2014. Spatial variability and stochastic strength
prediction of unreinforced masonry walls in vertical bending. Engineering Structures 59: 787–797.
Milani, G. & Lourenco, P.B. 2013. Simple Homogenized Model for the Nonlinear Analysis of FRP Strength-
ened Masonry Structures. II: Structural Applications. Journal of Engineering Mechanics 139(1): 77–93.

1398
 Muhit, I.B., Masia, M.J., Stewart, M.G. & Isfeld, A.C. 2022. Spatial Variability and Stochastic Finite
Element Model of Unreinforced Masonry Veneer Wall System Under Out-of-plane Loading. Engin-
eering Structures 267: 114674.
Pulatsu, B., Gonen, S., Parisi, F., Erdogmus, E., Tuncay, K., Funari, M.F. & Lourenço, P.B. 2022. Prob-
abilistic approach to assess URM walls with openings using discrete rigid block analysis (D-RBA).
Journal of Building Engineering 61: 105269.
Sarhosis V., Forgács T. & Lemos J.V. 2019. Stochastic strength prediction of masonry structure:
a methodological approach or a way forward? RILEM Technical Letters 4: 122–129.
Sarhosis V. & Sheng Y. 2013. Identification of material parameters for low bond strength masonry.
Engineering Structure 60:100–10.

1399

View publication stats