Influences of Random Imperfection Distribution On The Compressive Properties of Interlocking Block Wall
Influences of Random Imperfection Distribution On The Compressive Properties of Interlocking Block Wall
Influences of Random Imperfection Distribution On The Compressive Properties of Interlocking Block Wall
Structures
journal homepage: www.elsevier.com/locate/structures
A R T I C L E I N F O A B S T R A C T
Keywords: Block imperfections exist inevitably owing to manufacturing and construction quality control. For dry-stacked
Interlocking blocks interlocking block structures, imperfections result in small gaps randomly distributed between blocks, which
Imperfection affect the compressive strength of the wall. In this study, stochastic analysis is conducted to predict the
Monte-Carlo simulations
compressive properties of interlocking block walls with spatially varying randomly distributed block imperfec
Spatial variability
Seating effect
tions. Monte-Carlo simulation is conducted in the analysis. The number of block imperfections is assumed to
follow the Binominal distribution in massive block production process; the imperfection sizes are assumed to
follow the truncated normal distribution. Based on these hypotheses, the damage development mechanism and
load-path of interlocking block walls with different imperfection distributions are investigated. It is found that
the compressive strength of walls containing blocks with mixed imperfection levels is lower; for walls with a
higher number of imperfections, a larger coefficient of variation of imperfections leads to a significant decrease
in compressive strength, while the seating effect at the initial stage of compression is reduced. The results provide
a guidance to the quality control of mortar-less interlocking block structures.
* Corresponding author.
E-mail address: [email protected] (X. Zhang).
https://doi.org/10.1016/j.istruc.2023.104875
Received 21 February 2023; Received in revised form 6 June 2023; Accepted 10 July 2023
Available online 25 July 2023
2352-0124/© 2023 The Author(s). Published by Elsevier Ltd on behalf of Institution of Structural Engineers. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
T. Shi et al. Structures 56 (2023) 104875
when the wall was subjected to gradually increased compressive load. wall performance and failure mode. Non-spatial simulation was found to
Al-Fakih et al. [11] conducted compressive tests on dry-stacked inter overestimate the mean of wall compressive strength compared with
locking masonry prisms, wallets and panels. It was found that the im spatial simulation. Some previous studies also investigated the influence
perfections on the contacting interfaces considerably reduced the of workmanship quality on the strength of masonry structures [37,38].
contact stiffness of the bed joints. Dorji et al. [12] found that compres For instance, Stewart and Lawrence [38] studied the effect of work
sive cracks in ungrouted dry-stacked masonry would not be transmitted manship quality, discretising of masonry unit thickness and wall width
to the bottom masonry units due to the mortarless feature and the lower on the reliability index. It was concluded that the structural reliabilities
overall stiffness of the dry-stacked masonry comparing to mortar- are very susceptible to the workmanship quality, discretising of masonry
bonded masonries. Zahra et al. [13] carried out compressive tests on unit thickness and wall width. Martínez et al. [16] discussed the impacts
dry-stacked interlocking masonry panels and found that grouting could of random variation in the rough surfaces of block. Shi et al. [19]
significantly enhance the compressive strength of mortarless masonry investigated that the compressive behavior of interlocking prisms
and reduce the seating effect. Many numerical studies have also been comprised of interlocking block with non-spatially varied imperfections.
carried out based on both the simplified linear analysis and detailed No seating effect was observed in the numerical simulation. Ngapeya
nonlinear finite-element modellings to analyse the compressive capacity et al. [9] estimated the load bearing capacity considering the height
and failure mechanisms of interlocking prisms, wallets and walls imperfection of each block. It was found that the height imperfection of
[14–21]. The influences of some design parameters, such as the block individual block plays a critical role in the failure mechanism and load
material strength, prism/wallet height and block geometry on the bearing performance of a dry-stacked block wall subjected to compres
compressive capacity and deformation capacity of dry-stacked inter sive load. This is because the height imperfection controls the actual
locking prisms and wallets have also been studied. contacting area between the neighbouring masonry courses, and thus it
Imperfection unavoidably exists in dry-stacked interlocking blocks dominates the load percolation system in the wall. Gooch et al. [39]
resulted from block manufacturing error/tolerance and the quality of established finite-element models of masonry structures, and found the
workmanship. For conventional clay blocks or concrete masonry units numerical modellings tend to overestimate the elastic stiffness of labo
bonded with mortar, these imperfections could be mitigated by the ratory structures. Therefore, it is necessary to properly understand and
mortar layer, which nevertheless does not exist in mortar-less/dry- interpret the numerical modelling results. Till now, only very few
stacking construction of interlocking blocks. It has been found that im studies investigated the structural performance of dry-stacked inter
perfections will not only affect the local strength and stiffness of inter locking structures with consideration of the spatial variation of imper
locking blocks [19] but also influence the overall behaviour of structures fections. Critical structure performance factors such as the stiffness and
made of interlocking blocks [22,23]. Therefore, it is necessary to prop compressive strength of dry-stacked interlocking block wall are not
erly evaluate the influence of imperfection on interlocking block struc understood well yet.
tures constructed with the mortar-less method. This study presents numerical investigations on the compressive
Previous laboratory tests found that imperfections of block units behaviour of dry-stacked interlocking block wallets using stochastic
could result in seating effect when masonry piers/walls are subjected to analysis considering spatial variability of block imperfection. Monte-
compressive load [24–27]. Some researchers took block imperfection Carlo simulations are carried out to examine the influence of different
into considerations when numerically modelling the compressive number of block imperfections, imperfection locations in a wallet and
behaviour of masonry structures. For instance, Shi et al. [19] employed varying imperfection sizes. The quality of interlocking block walls is
different modelling strategies to consider the gaps at block joints asso categorized according to the number of imperfect blocks, while the co
ciated with the block imperfections for interlocking block prism. It was efficient of variation (COV) of the block imperfection is employed to
found that such imperfections would strongly influence the compressive represent dispersion of imperfections. Stress–strain curves and failure
behavior of interlocking block prisms. Nevertheless, previous studies progression are examined. The variability of load carrying capacity,
only considered non-spatial variability of imperfections in which the stiffness and damage mode are studied for interlocking block walls with
same imperfection was assumed throughout the block pier/wall. This different qualities. It should be noted that this study does not intend to
simplification apparently would not properly represent the real condi research on stochastic methods, but uses existing method to investigate
tion of block structures, where the size of imperfection of each block unit the influences of random interlocking block imperfections on masonry
varies; the locations of these imperfections also vary across the structure wall load carrying capacities. The novelty and significance of this study
[28]. Therefore, a proper study that considers the random spatial vari reside in its pioneering stochastic analysis of the compressive properties
ations of imperfection of each block and location within a structure of dry-stacked interlocking block walls considering spatially varying
should be performed. block imperfections. This research innovatively models these imper
Stochastic analysis has been commonly used to investigate the in fections using Binominal and truncated normal distributions for the
fluences of random variables on structural performance of masonry number and sizes of imperfections, respectively. The examination of the
structures [29–31]. For example, the influence of material uncertainties influence of imperfection distributions on the damage development
on the axial compressive strength of masonry wallets was investigated mechanism and load-path of interlocking block walls fills a gap in the
through probability analysis based on Monte-Carlo simulations [32]. existing literature. This approach not only enhances the accuracy of
Stewart and Lawrence [33] generated a probabilistic model to assess the compressive strength prediction but also furnishes critical insights for
structural reliability of masonry walls subjected to concentrically improving quality control in mortar-less interlocking block structures.
compressive loading considering variations in mortar type, live-to-dead By bolstering the understanding of the impacts of block imperfections on
load ratio, block material compressive strength and tributary area. The structural performance, this study contributes significantly to the sci
accuracy of design code, i.e., the Australian masonry design code entific robustness of construction methods involving dry-stacked inter
AS3700-2001 [34], was examined based on the reliability analysis. Ex locking blocks.
istence of spatial variability of block joint flexural bond-wrench
strengths was also studied using stochastic analysis. For example, the 2. Numerical model
influence of spatial correlation among unit block joint flexural bond-
wrench strengths was experimentally studied by Heffler et al. [35]. It 2.1. Block configuration
was found that the flexural bond-wrench strength of each block unit is
statistically independent of its neighbouring unit. Li et al. [36] found Fig. 1 illustrates the interlocking blocks considered in this study.
that there are obvious differences between spatial and non-spatial Inspired by the osteomorphic block [40], the studied block achieves
analysis of the flexural bond-wrench strength of block wall in terms of topological interlocking through concavoconvex surfaces (i.e., the
2
T. Shi et al. Structures 56 (2023) 104875
tenons and mortises on the blocks). Thus, compared with other common 2.2.2. Boundary condition and contact properties
interlocking masonry units that achieve interlocking through special For the interlocking block wall, the bottom surface of the wall is fully
connectors, this type of block can be assembled into complex structures fixed, and the lateral and rotational movements (out-of-plane) at the top
with single or multiple layers [41]. Previous study [41] showed that of the wall are restrained while its vertical degree of freedom is enabled
when structures made of this type of block are damaged, the damage for the application of the vertical loading. Displacement-controlled
tends to be confined to individual blocks, rather than leading to global method is used to load the interlocking block wall till failure.
failure as in solid structures. Also, because the interlocking mechanism A hard contact is employed to model the contact behavior between
between these blocks is achieved through tenons and mortises distrib adjacent blocks in the normal direction that the normal stress is trans
uted in various positions on the blocks, stress concentration at joints can ferred through the two surfaces. The contact surfaces will separate from
be largely avoided [40,41]. each other when subjected to a normal tensile force; hence no tensile
As shown in Fig. 1, the block has a width of 200 mm, a height of 180 stress will be transferred across the contact surfaces. The contact
mm and a thickness of 100 mm. Asymmetric interlocking shear keys are behavior in the tangential direction of the interface is simulated using a
designed for the block, where on the frontal side there is a wide shear finite sliding model following the Coulomb friction law. The tangential
key with a dimension of 70 mm width × 30 mm height × 35 mm, and on movement initiates when the surface traction stress at the connecting
the rear side there are two smaller keys with geometric dimensions of 35 interface reaches the threshold shear strength τ, which is governed by
mm width × 30 mm height × 35 mm. The configuration of interlocking the normal contact pressure p and the coefficient of friction μ (τ = μ p).
blocks is designed to provide improved shear resistance which differs The coefficient of friction between the contact surfaces of neighbouring
from other existing interlocking blocks whose interlocking keys are interlocking blocks is taken as 0.3 following previous studies [16,45].
usually only for self-alignment. The shear keys have inclined surfaces,
which make the block assembly easier and allows the blocks to slide on
2.3. Model validation
each other under high lateral force. The material of blocks is a mixture
composed of sand, cement, gravel, and fly ash, therefore it has properties
To validate the suitability and accuracy of the above numerical
similar to those of concrete.
modelling method, it is used to model the compressive behavior of
interlocking block prisms in the authors’ previous study [19]. As illus
2.2. Finite element model
trated in Fig. 2a, the specimen is composed of two full-sized interlocking
blocks and two half blocks at both ends. The compressive load is applied
2.2.1. Model detail
using displacement-controlled loading method. The interlocking block
A detailed three dimensional finite element model of interlocking
prism is numerically modelled using the above method in Section 2.2.
wall with dimension 1200 mm × 800 mm × 100 mm (height × width ×
Both perfect contact and imperfect contact are considered to simulate
thickness) is constructed using the commercial software Abaqus [42]. As
block imperfections at the joints. There is no gap between blocks at the
mentioned above, interlocking blocks are modelled using solid element
joint for the perfect model. For the imperfect model, gaps between
C3D8R with the Concrete Damage Plasticity (CDP) material model, in
blocks are modelled. Each surface imperfection height was measured
which both block crushing in compression and cracking in tension can
using a height dial indicator (Fig. 2b), and the measured gap sizes are
be considered. The material properties of the interlocking block are
applied to the numerical model with imperfect contact as shown in
shown in Table 1, where E0 and v denote the Young’s modulus and the
Fig. 2c. An average gap width is 0.29 mm with a standard deviation of
Poisson’s ratio. More detailed modelling of the interlocking block
0.16.
properties and the corresponding verification can be referred to [43].
Fig. 3 compares the compressive force versus axial displacement
The compressive strength of block material is 17.46 MPa [43]. The
curves from the laboratory test and the numerical simulations. As
tensile strength is assumed as ft = 0.1 fc, which is a relationship
shown, in the laboratory test the force increases gradually with the
commonly employed for concrete material [16,44].
imposed displacement because of seating effect. As gaps gradually close,
the force increases quicker until reaching the peak load of 112.8 kN and
Table 1
Material properties of interlocking block.
Mass density (kg/m3) Elasticity Plasticity
Initial Young’s modulus, E0 (GPa) Poisson’s ratio υ Dilatation angle ψ (◦ ) Eccentricity Biaxial stress ratio fbo /fco K Viscosity
Parameter
Note: K is the ratio between the second stress invariant on the tensile meridian and compressive meridian at initial yield.
3
T. Shi et al. Structures 56 (2023) 104875
then plummets due to block damage. The numerical model with perfect
contact predicts an ultimate load capacity of 113.8 kN. However, it
could not model the seating effect because there are no pre-existing
gaps. Moreover, an initial stiffness of 71.18 kN/mm is predicted,
which is significantly larger than that of the lab test (61.85 kN/mm). The
numerical model with imperfect contact predicts an ultimate capacity of
105.0 kN indicating less than 7% difference comparing to the lab testing
result, while the predicted initial stiffness of 54.49 kN/mm by the nu
merical model with imperfect contact is much similar to that of the
laboratory test.
It demonstrates that the numerical model considering block imper
fection is crucial for proper estimation of the stiffness of mortarless
interlocking block wall under compressive loading.
Fig. 4 compares the failure modes of interlocking block prisms from
laboratory test and numerical simulations. In the laboratory test, wing
cracks are initiated at the corner of the interlocking block key and
extend both up and down. Due to the non-perfect contact conditions
between neighbouring blocks, asymmetric failure mode is observed. The
numerical model with imperfect contact predicts similar damage pattern
where cracks initiate on the keys of the second block, and then extend
vertically. The model with perfect contact predicts typical X-shaped
Fig. 3. Comparison of compressive force versus axial displacement curves from failure because of the damages of shear keys due to symmetry, which
laboratory test and numerical modelling. could reflect the damage pattern of interlocking block but not able to
fully represent the influence of imperfect blocks. Through the above
comparison, it can be found that the developed numerical modelling
4
T. Shi et al. Structures 56 (2023) 104875
method could reasonably closely represent the behavior of mortarless are pressed in mould using servo-controlled high-pressurized block
interlocking block wall with consideration of block imperfection under machine, imperfections could occur to the produced blocks with varia
compressive loading. tion to block heights, i.e., total block height, and shear key height, and
inclined surfaces as shown in Fig. 1. Fig. 6 shows the set-up of inter
2.4. Size effect locking block wall with spatial imperfections. Because of the asymmetric
shear key layout on each interlocking block unit, the wide shear key and
The modelled wallet size (height 1200 mm, width 800 mm and the small keys on the frontal and rear surfaces are shown respectively in
thickness 100 mm) follows the recommendation of BS1052-1 [46] for Fig. 6a and Fig. 6b. Considering the shape and geometry of the shear
the determination of compressive strength of the interlocking block keys, imperfections could exist on the flat surfaces or the inclined sur
wall. To examine the potential size effect and to quantify its influence on faces of the shear keys, as seen in Fig. 6c and 6d. It is to note that height
the modelled compressive strength of the mortar-less interlocking block imperfections on the flat surfaces are most common. This is because
wall, a group of masonry walls with different dimensions are numeri during the manufacturing process of the interlocking blocks, the weight
cally modelled, i.e. 600 mm × 400 mm × 100 mm, 1200 mm × 800 mm of the material added into the mould and the pressure applied to
× 100 mm, and 2400 mm × 1600 mm × 100 mm (as shown in Fig. 5a). compress the material remain constant for each block. However, due to
For easy engineering application, axial stress and strain are used to variations in the moisture content of the block material, the height of the
quantify the performance of the interlocking block wall in lieu of axial produced blocks may vary slightly, resulting in imperfections on the flat
compressive load and displacement. The axial stress is the averaged surface.
compressive stress in the cross section which is calculated by dividing As shown in Fig. 6c, the imperfection height h means the roughness
the measured axial compressive load by the cross-sectional area of the height normal to the surface of interlocking block. It is to note that the
interlocking block wall; and the strain is calculated by dividing the sizes of the above imperfections are normally small, since blocks with
vertical displacement with the initial height of the interlocking block large imperfections will be filtered out in the quality control process,
wall. Fig. 5b compares the maximum compressive strengths and the and blocks with large imperfections caused during transportation will
initial stiffness, from which it can be found that varying the width and also be identified in construction and usually discarded because they
height of the wall model by four times leads to a maximum variation in will not stack properly.
the compressive strength by 3.40%, and a maximum variation in the
stiffness by 0.23%. It demonstrates that the influence of model size on 3.2. Stochastic analysis with Monte-Carlo simulation
the compressive loading capacity is insignificant and the recommended
wallet dimension by BS1052-1 is appropriate for predicting the The spatial variability of imperfection for interlocking block wall is
compressive strength of mortar-less interlocking block wall. generated using the above-mentioned numerical method in stochastic
analysis based on the Monte-Carlo simulations. Since blocks are manu
3. Probabilistic models factured in large quantities, binominal distribution is employed to
describe the existence of imperfection on each block unit. The proba
Compared to conventional masonry where mortar bonds blocks bility of imperfection for the binominal distribution, p, depends on the
together leading to negligible influence of block imperfection, dry- quality control in block manufacture. Following the recommendation of
stacked interlocking blocks are constructed without mortar. Unavoid block manufacturer, the probability of imperfection existence during
able imperfections in block could potentially result in significant local manufacturing is stratified into four groups: high quality, high-medium
stress concentration and relative movements of adjacent blocks and thus quality, medium–low quality, and low quality, which correspond to
influence the mechanical performance of the masonry structure. This 0–25%, 25–50%, 50–75%, and 75–100% possibility of each block unit
paper focuses on quantifying the influence of geometric imperfections having imperfections. p is assumed to follow uniform distribution within
on the compressive properties of interlocking block walls, where other each group. It is worth noting that Monte-Carlo simulations usually use
uncertainties such as block material strength, are not explicitly inves simple random sampling, which entails a large number of simulations to
tigated. To study the influence of random imperfection of interlocking achieve convergence, therefore is computationally costly. The employed
block on the compressive strength and axial stiffness of the interlocking stratified sampling method could help to achieve an improved conver
block wall, the above-selected 1200 mm × 800 mm × 100 mm inter gence rate. The stratified sampling method is operated by subdividing
locking block wall with spatial block/joint imperfection is modelled. the sample space into smaller regions and sampling within these regions.
In so doing, the produced samples could more effectively fill the sample
3.1. Block and joint imperfection space and therefore reduce the variance of computed statistical esti
mators. Previous study demonstrated this sampling method could pro
Considering the manufacturing process of interlocking blocks which vide fast converging analysis for Monte-Carlo simulation with
5
T. Shi et al. Structures 56 (2023) 104875
satisfactory accuracy [47]. The imperfection size of each block unit is compression method till the failure of the wall. The peak
simulated following the truncated normal distribution with a mean of compressive force and stiffness of the wall are determined
0.3 mm, which was determined through statistical analysis on the through the axial load versus axial displacement curve simulated
different batches of block specimens [43]. The truncated normal dis by the numerical analysis. The compressive strength of the
tribution rather than the normal distribution is used herein to avoid a interlocking block wall is determined with the peak compressive
negative joint gap value. A python-based program is written to auto force.
matically generate interlocking block wall models in batches in Abaqus. (5) Convergence analysis is performed, where the accumulated mean
Despite the complex configuration of blocks due to the random imper compressive stress and accumulated compressive stress COV for
fections, in the numerical modelling all the initial locations of the blocks the interlocking wall with different numbers of simulations are
are set at the proper positions as an intact wall without considering the examined to determine the termination of numerical simulations.
imperfection. The generation of spatial variability model can be Repeat the above steps 1 to 4 until the statistical convergence is
expressed as follows: reached, in which the COV varies between 0.1 and 0.5. A COV of
0.1 represents the imperfection size has slight dispersion, while a
(1) For a 1200 mm × 800 mm × 100 mm interlocking block wall, it COV of 0.5 represents the imperfection size has large dispersion.
comprises of 40 blocks (as shown in Fig. 7). Each block is labelled (6) The average compressive stress is calculated by the ratio of the
with a number from 1 to 40 in the sequence from the bottom left total force on the interlocking wall and the total sectional area of
to the top right in layers. the block wall. The average strain is measured by the ratio of the
(2) Depending on the location of the imperfections in one block, the compressed displacement at the top surface of the interlocking
imperfections of a block are classified into four types, i.e., height wall and the height of the interlocking wall. The data are counted
imperfection on the small keys, height imperfection on the large into different groups with different COVs and/or different range
keys, slide (side) imperfection on the front sliding surfaces, and of defect rates (p) and then used for construction of the histo
slide imperfection on the rear sliding surfaces. The existence of grams of compressive strength, equivalent stiffness and linear
imperfection is randomly determined following the Binominal stiffness. The probability density function is generated and
distribution with a possibility of p. The block quality is classified compared with specific probability density functions to deter
into four groups, i.e. high quality, high-medium quality, medi mine the type of distribution.
um–low quality, and low quality, depending on the value of p.
(3) Once the location of imperfection in a block is determined, the 3.3. Convergence study
imperfection size on the surface of this block is assumed to be
uniform over the entire surface for simplicity. The imperfection Convergence test of Monte-Carlo simulations is performed with the
size follows the truncated normal probability distribution with a numerical model for interlocking block walls of different qualities. Fig. 8
mean hm = 0.3 mm and a specific standard deviation. The char presents the variations of the mean compressive strength of interlocking
acteristic value of imperfection height is estimated using h = hm block walls and its corresponding COV acquired from the Monte-Carlo
± 1.645σ. Since there is very limited data on the standard devi simulations. It is found that after 60 random cases, the variations on
ation of imperfection size σ, the COV is assumed to vary between the mean compressive strength and the corresponding COV are very
0.1 and 0.5 [48]. small, indicating the simulations converge. Therefore, 60 simulations
(4) Each numerical model is established with the random spatial are needed to achieve the converged results. It is to note that the number
imperfection distribution, after which numerical modelling is of simulations to achieve convergence is small because the stratified
carried out by loading the wall using displacement-controlled sampling method is employed, which helps to unconditionally reduce
6
T. Shi et al. Structures 56 (2023) 104875
7
T. Shi et al. Structures 56 (2023) 104875
compressive strength is achieved), when crack initiates on the wall. As wall, the compressive load that the wall could withstand increases lin
can be seen in Fig. 10a, initial cracks are developed on the interlocking early to about half of the maximum compressive strength, which is lower
blocks with large size imperfections. This is because under compressive than that of the high-quality interlocking wall. Being similar to the high-
loading, smaller joint gaps due to block imperfections would close quality wall, cracks initiate around the joints with larger size imper
earlier than the larger gaps, which enables load transfer at these joints fections. It can be seen from Fig. 10a that for the high-quality and high-
and hence more uniform stress distribution. But it also leads to non- medium quality wall, because relatively less imperfections exist and
uniform joint gaps with large block imperfections, which causes stress smaller imperfections would close under the imposed compressive
concentration and hence crack initiation. For the high-medium quality loading first, the compressive force can then be transferred through
8
T. Shi et al. Structures 56 (2023) 104875
Fig. 10. Initial crack and load path at the stage of OA.
multiple load paths over the interlocking blocks. For the low-quality initial seating movement to close the gap between blocks. This is
wall, because of the widely spread imperfections on blocks, under the because the low quality wall is associated with wider spread imperfec
initial compressive loading, less contacts are developed between blocks tions. Under compressive loading, the gaps in block wall close relative
resulting in seating effect (as highlighted in yellow). As a result, with the uniformly, whereas the medium–low and high-medium quality walls
increased compressive loading, only less number of load paths are have less number of imperfections. Under compressive loading, closing
developed. Similar to the medium–low quality wall, initial cracks are of the gaps is not uniform and causes stress concentration, hence more
developed around the load path where blocks are bearing high damage to the blocks. Therefore, their compressive loading capacity is
compressive stress. As shown, the low-quality block wall has higher smaller than that of the low-quality wall.
compressive strength than the medium–low and high-medium quality
wall although its equivalent stiffness is small owing to a relatively large
9
T. Shi et al. Structures 56 (2023) 104875
As the imposed compressive load continues to increase, these cracks Fig. 12 displays the compressive stress–strain curves for interlocking
extend, and more cracks are developed until the maximum compressive block walls with varying quality and imperfections. Engineering stress
strengths of each wall are reached. Fig. 11 shows the eventual damage and strain are used, which are calculated based on the net cross section
contours of the four interlocking block walls of different qualities. area and the wall height, respectively, offering a clear visual on how the
Because the distribution and size of block imperfections are random, the strength and deformability of the wall are affected by imperfections.
crack pattern are asymmetric and random. Block cracks widely spread The stress–strain curves for high to low-quality walls, with a coeffi
among the walls. But some common cracking similarities exist in these cient of variation (COV) of 0.1, reveal significant differences. High-
four quality interlocking walls. Cracks initiate from the shear keys with quality walls exhibit a linear elastic behaviour up to 75% of the
vertical cracks indicating tensile damage [19]. Compressive damages maximum compressive capacity, followed by a non-linear increase and
are then developed in the cracked blocks as the wall is subjected to an eventual decrease, indicative of considerable damage. High-medium
further increased compressive loading. It indicates that increasing the quality walls demonstrate a similar pattern, but the linear elastic stage is
post-cracking strength of interlocking blocks would improve the me shorter, only lasting to about 50% of compressive capacity. This is
chanical performance of interlocking block wall under compressive because in the high-medium quality wall, there are more imperfections,
loading. In the meanwhile, some differences can be observed on these resulting in a smaller compressive stress distribution area and a more
four different quality interlocking block walls: for the high-quality and uneven distribution pattern on the cross-section of the wall. This, in
high-medium quality walls, since the majority of blocks are intact thus in turn, leads to the earlier arrival of the elastoplastic stage in the high-
full contact for the compressive load to transfer through, the load dis medium quality wall. As block quality deteriorates, the influence of
tributes relatively uniformly and thus cracks are also more uniform, block imperfection variation on the initial modulus becomes increas
which leads to a higher compressive capacity and stiffness than those of ingly evident due to the larger number of imperfections. For low-quality
the walls of the other qualities. For the medium–low quality wall with walls, pronounced seating effects are observed in their stress–strain
more imperfected blocks, load is redistributed after the initial cracking curves, which is a result of the gradual closure of the gaps due to im
of blocks around the load path. More blocks are thus to bear the perfections under compressive loading.
compressive force. More cracks can be observed on the failed walls when While the compressive strength of the wall decreases as the block
the peak compressive capacity of the wall is reached. For the low-quality quality lowers from high to high-medium and then to medium–low, an
wall with the majority of blocks having imperfections, joint openings increase is observed in the walls with low-quality blocks compared to
due to block imperfection actually would close under the initial the walls with medium–low quality blocks despite a considerable
compressive loading, leading to a more uniform distribution of stress on decrease in initial equivalent stiffness when the COV is 0.1 and 0.3. This
the entire block wall. As a result, on the final damage contour of the low- counterintuitive result can be attributed to the relatively uniform
quality wall, cracks are distributed widely and more uniformly as imperfection distribution in those low-quality blocks with a relatively
compared with the medium–low quality wall. small imperfection variation, causing an apparent seating effect, i.e.,
initial slippage at low compressive load. Upon the closure of imperfec
5. Results and analysis tion gaps caused by compression, a near-linear behaviour similar to
high-quality walls is observed due to the increased contact area between
The influence of random imperfections of interlocking blocks on the blocks, suggesting that uniform imperfection distribution can compen
compressive behavior of interlocking block wall is presented and ana sate for the lack of block quality to some extent.
lysed in this section. The compressive stress–strain curves are firstly In summary, the probability of block imperfection negatively im
presented. The compressive strength and axial stiffness of interlocking pacts wall quality and initial compressive stiffness, particularly for the
block wall are summarized and analysed. first three quality levels. However, for low-quality blocks with relatively
low block imperfection variation, as gaps close after the strong seating
effect, the wall behaves similarly to a high-quality block wall. Thus, the
10
T. Shi et al. Structures 56 (2023) 104875
Fig. 11. Damage contours of interlocking block wall at peak compressive loads.
distribution of block imperfections plays a significant role in deter compressive strength. Conversely, a larger COV, indicating non-uniform
mining the compressive strength and deformability of interlocking block imperfection distributions, can result in more local damage due to stress
walls. concentrations, thereby reducing the wall’s compressive loading
capacity.
5.2. Compressive strength and probability distribution Fig. 14 presents the histograms of wall compressive strengths, which
are modeled using different probability distribution functions. These
The compressive strengths derived from the stress–strain curves for histograms demonstrate that the increase of COV of imperfection size
the interlocking block walls with varying quality and imperfections are scatters the compressive strength of the interlocking block wall, with a
summarized in Table 2. The table lists mean compressive strength, heavier tail for COV = 0.5, indicating a greater probability of wall failure
standard deviation σ, and the associated coefficient of variation (COVf) before reaching the predicted compressive strength. Particularly for
for each quality category. Fig. 13 depicts the predicted compressive low-quality walls, an increase in COV significantly reduces the wall
strength from the spatial analysis. strength, as explained above.
High-quality walls with fewer imperfect blocks exhibit the highest In summary, these results highlight the role of COV in block imper
mean compressive strength. The COV of block imperfection sizes only fections in determining wall compressive strength, suggesting that
slightly affects the mean compressive strengths of high and high- neglecting this factor may overestimate the compressive capacity,
medium quality walls due to the limited number of block imperfec especially for low-quality walls.
tions. However, for medium–low and low-quality walls, an increase in
COV results in a decline in compressive strength, with low-quality walls 5.3. Equivalent stiffness and probability distribution
seeing a significant drop from 4.00 MPa to 2.72 MPa as COV increases
from 0.1 to 0.5. This is because in low-quality walls, most gaps close To quantify wall stiffness, the equivalent stiffness is introduced,
under compressive loading. A smaller COV results in more uniform gap defined as the slope of the secant line from the origin to the peak
closures and better inter-block contact and hence leads to relatively high compressive force on the modelled stress–strain curves (Fig. 15). Fig. 16
11
T. Shi et al. Structures 56 (2023) 104875
Fig. 12. Stress–strain curves of interlocking block walls with spatially varying imperfections.
12
T. Shi et al. Structures 56 (2023) 104875
Table 2
Summary of wall compressive strengths from Monte-Carlo analysis.
Imperfection High-quality wall High-medium quality wall
COV fMax. fMin (MPa) fMean σ COVf Distribution fMax. fMin. fMean σ COVf Distribution
(MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa)
0.1 4.96 3.94 4.47 0.26 0.058 Normal 2.83 3.84 3.38 0.22 0.065 Loglogistic
0.3 5.12 3.88 4.55 0.29 0.064 Loglogistic 3.86 2.63 3.32 0.27 0.080 Weibull
0.5 5.17 3.64 4.40 0.36 0.081 Normal 2.47 3.84 3.27 0.29 0.090 Normal
COV fMax. fMin. fMean σ COVf Distribution fMax. fMin. fMean σ COVf Distribution
(MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa) (MPa)
0.1 3.16 2.44 2.81 0.18 0.060 Normal 4.65 3.36 4.00 0.34 0.085 Normal
0.3 3.28 2.36 2.79 0.20 0.070 Loglogistic 4.04 2.70 3.33 0.32 0.095 Gamma
0.5 3.41 2.17 2.70 0.23 0.086 Loglogistic 3.12 2.17 2.72 0.22 0.079 Smallest
Extreme
and Table 3 illustrate the distributions and statistics of equivalent decreases in peak load displacements despite minor variations in peak
stiffness for different imperfection size COVs. compressive load. With a wider and non-uniform distribution of block
As expected, the mean equivalent stiffness decreases with increasing imperfections in the wall, local damage from stress concentration will
imperfection quantity, attributable to a corresponding increase in peak occur, reducing the wall displacement at failure. Conversely, low-
compressive load displacement. This highlights the potential for over quality walls exhibit a decrease in equivalent stiffness with increasing
estimation of dry-stacked block wall stiffness in numerical models COV, which is a result of the significant drop of the wall’s compressive
neglecting existing imperfections [16,19]. strength with the increase of the COV of block imperfection sizes, as
Interestingly, as COV increases from 0.1 to 0.5, the equivalent stiff explained in Section 5.2.
ness of high-quality, high-medium-quality and medium–low-quality These results underscore the strong influence of imperfection size
walls increases. This counterintuitive result could be due to significant distribution (COV) on the equivalent stiffness of interlocking block
13
T. Shi et al. Structures 56 (2023) 104875
Fig. 13. Distribution of wall compressive strengths corresponding to different COVs of block imperfections.
walls. It is noted that low-quality walls with COV = 0.1 have higher 5.4. Linear stiffness and probability distribution
equivalent stiffness than medium–low quality walls. This is also due to
the significantly higher compressive strength of low-quality walls with For high-quality and high-medium quality walls, where seating effect
low COV of block imperfection sizes compared to that of the medium is negligible, the linear stiffness is defined as the slope of the secant line
–low-quality walls. from the origin to the softening stage’s starting point. For walls of
Fig. 17 presents equivalent stiffness histograms and best-fit proba medium–low and low-quality blocks, the seating effect is pronounced.
bility distribution functions, indicating a gradual dispersion of equiva Therefore, the linear stiffness is determined as the slope of the
lent stiffness with increasing COV. For high-quality walls, there is a load–displacement curves’ linear portion to reflect the wall’s post-
larger possibility of high equivalent stiffness with the increase of the seating and prior-to-non-linear behavior.
COV of block imperfection sizes, while for low-quality walls, increasing As presented in Table 4 and Fig. 18, a correlation exists between the
COV significantly reduces the equivalent stiffness, which conform to the wall’s quality and its linear stiffness. High quality, high-medium and
above analyses. It is noteworthy that the histograms for low-quality medium–low quality walls have mean linear stiffness values of 511kN/
walls (Fig. 17d) are higher and narrower, which suggests less vari mm, 320kN/mm, and 119kN/mm (COV = 0.1), respectively, implying
ability in equivalent stiffness compared to other walls of different that decreasing wall quality (i.e., increasing imperfections and reducing
qualities. This is a consequence of the simultaneous decrease in contact areas) significantly diminishes linear stiffness. The mean linear
compressive strength and the corresponding displacement in low- stiffness for these wall qualities sees a gradual increase as the COV of
quality walls as the COV of block imperfection sizes increases. block imperfection sizes increases. The larger variation of imperfection
14
T. Shi et al. Structures 56 (2023) 104875
Fig. 16. Distribution of wall equivalent stiffness corresponding to the different COVs of block imperfection size distributions.
Table 3
Summary of wall equivalent compressive stiffness from Monte-Carlo analysis.
Imperfection High-quality wall High-medium quality wall
COV EMax. (kN/ EMin (kN/ EMean (kN/ σ (kN/ COVE Distribution EMax. (kN/ EMin. (kN/ EMean (kN/ σ (kN/ COVE Distribution
mm) mm) mm) mm) mm) mm) mm) mm)
0.1 342.64 185.64 247.94 40.45 0.163 Loglogistic 206.20 114.20 148.31 23.30 0.157 Gamma
0.3 442.85 190.09 265.00 51.09 0.193 Loglogistic 211.30 101.80 151.70 28.68 0.189 Loglogistic
0.5 460.04 170.72 291.38 66.70 0.229 Loglogistic 262.00 108.30 163.77 35.10 0.214 Gamma
COV EMax. (kN/ EMin. (kN/ EMean (kN/ σ (kN/ COVE Distribution EMax. (kN/ EMin. (kN/ EMean (kN/ σ (kN/ COVE Distribution
mm) mm) mm) mm) mm) mm) mm) mm)
0.1 107.73 72.05 87.75 9.25 0.105 Normal 103.60 80.71 92.90 5.49 0.059 Normal
0.3 123.20 73.79 92.21 11.65 0.126 Loglogistic 93.34 67.43 80.47 5.76 0.072 Normal
0.5 147.86 68.49 97.68 15.39 0.16 Loglogistic 90.93 60.12 72.61 7.04 0.097 Normal
15
T. Shi et al. Structures 56 (2023) 104875
Table 4
Summary of wall linear stiffness from Monte-Carlo analysis.
Imperfection High-quality wall High-medium quality wall
COV EMax. (kN/ EMin (kN/ EMean (kN/ σE (kN/ COVE Distribution EMax. (kN/ EMin (kN/ EMean (kN/ σE (kN/ COVE Distribution
mm) mm). mm) mm) mm) mm). mm) mm).
0.1 575.04 439.63 511.44 28.34 0.055 Normal 439.79 212.16 320.30 48.57 0.152 Normal
0.3 590.62 439.95 521.99 29.75 0.057 Normal 415.77 216.33 324.09 49.42 0.153 Normal
0.5 604.49 474.61 533.24 30.11 0.056 Normal 418.84 258.70 334.76 40.41 0.121 Normal
COV EMax. (kN/ EMin (kN/ EMean (kN/ σE (kN/ COVE Distribution EMax. (kN/ EMin (kN/ EMean (kN/ σE (kN/ COVE Distribution
mm) mm). mm) mm) mm) mm). mm) mm).
0.1 158.78 79.97 119.35 19.59 0.164 Loglogistic 350 121.79 231.04 51.35 0.22 Lognormal
0.3 184.63 90.99 126.03 20.83 0.165 Loglogistic 181.37 93.85 132.87 20.24 0.15 Normal
0.5 217.41 104.82 148.51 24.66 0.166 Gamma 142.68 81.37 113.53 15.01 0.13 Normal
quality walls may even exhibit higher compressive strength than walls 7. Conclusion
made of blocks with mixed qualities due to the more uniform
compression of gaps if the imperfections are of similar sized, however, it This study performs stochastic analysis to investigate the compres
would have significant seating effect even if the wall is subjected to sive properties of dry-stacked interlocking block walls with spatially
relatively low compressive loading, causing the subsidence of the whole varying block imperfections. A detailed numerical model of interlocking
structure. More generally, the number and severity of localised block block walls with block imperfection is generated. The number of block
damage and failures increase with decreasing block quality, leading to a imperfections is assumed to follow the Binominal distribution in massive
lower compressive strength and axial stiffness. This signifies that to block production process. The imperfection sizes are assumed to follow
maintain the structural performance and compressive load-bearing ca the truncated normal distribution. the imperfection size on the surface of
pacity of interlocking block walls, it is necessary to minimize the this block is assumed to be uniform over the entire surface for simplicity.
quantity of imperfect blocks and control the sizes of imperfections as The influences of imperfection on the compressive strength, equivalent
much as possible during production. stiffness, linear stiffness of interlocking block wall are studied. The
following conclusions are derived from this study:
16
T. Shi et al. Structures 56 (2023) 104875
Fig. 18. Distribution of wall linear stiffness corresponding to different COVs of block imperfection size distributions.
1. Typical crack patterns on imperfect block units are presented, which being 4.00 MPa, 3.38 MPa and 2.81 MPa, respectively. This phe
demonstrates strong influence of imperfection on the crack patterns. nomenon is attributed to the more uniform closure of gaps under
It is also found that block imperfections could strongly influence the compressive loading in the low-quality walls. In contrast, medium-
crack initiation and development, as well as the load path of dry- quality walls suffer from non-uniform deformation and stress con
stacked interlocking block walls. centration, leading to lower compressive strengths.
2. The quantity and distribution of imperfect blocks greatly impact an 3. The COV of block imperfection size has a marginal influence on the
interlocking block wall’s mean compressive strength. A high-quality mean compressive strengths of high-quality and high-medium
wall, with fewer imperfect blocks, shows the highest mean quality walls. Specifically, when COV increases from 0.1 to 0.5, the
compressive strength of 4.47 MPa at a COV of 0.1. Interestingly, a mean compressive strengths decrease slightly from 4.47 MPa to 4.40
low-quality wall with more imperfect blocks could still outperform MPa for high-quality walls, and from 3.38 MPa to 3.27 MPa for high-
high-medium and medium–low quality walls in terms of compressive medium quality walls. In contrast, the mean compressive strength of
strength, especially when COV = 0.1, with the compressive strengths
17
T. Shi et al. Structures 56 (2023) 104875
low-quality walls is more sensitive to the COV, significantly dropping [9] Ngapeya GGC, Waldmann D, Scholzen F. Impact of the height imperfections of
masonry blocks on the load bearing capacity of dry-stack masonry walls. Constr
from 4.00 MPa to 2.72 MPa as COV increases from 0.1 to 0.5.
Build Mater 2018;165:898–913.
4. Initial seating is obvious for medium–low and low-quality block [10] Jaafar MS, Thanoon WA, Najm AMS, Abdulkadir MR, Abang Ali AA. Strength
walls under compressive load because of gap closure between correlation between individual block, prism and basic wall panel for load bearing
imperfect blocks, and the seating displacement is prominent in the interlocking mortarless hollow block masonry. Constr Build Mater 2006;20(7):
492–8.
walls with relatively low-quality blocks. [11] Al-Fakih A, Mohammed BS, Al-Shugaa MA, Al-Osta MA. Experimental
5. The mean equivalent stiffnesses increase with the rise of COV of investigation of dry-bed joints in rubberized concrete interlocking masonry. J Build
imperfection size distribution from 0.1 to 0.5 for high-quality, high- Eng 2022;58:105048.
[12] Dorji S, Derakhshan H, Thambiratnam DP, Zahra T, Mohyeddin A. Behaviour and
medium quality and medium–low quality walls. This is attributed to material properties of versaloc semi-interlocking mortarless masonry. Mater Struct
the significant decrease in displacements at the peak compressive 2023;56(1).
loads, while the peak compressive load remains relatively un [13] Zahra T, Dorji J, Thamboo J, Cameron N, Asad M, Kasinski W, et al. Behaviour of
reinforced mortarless concrete masonry panels under axial compression: An
changed. However, for low-quality walls, the mean equivalent stiff experimental and analytical study. Constr Build Mater 2023;377:131097.
ness demonstrates a decline with the increasing COV of block [14] Zahra T, Dhanasekar M. Characterisation and strategies for mitigation of the
imperfection sizes due to the lower compressive strength, decreasing contact surface unevenness in dry-stack masonry. Constr Build Mater 2018;169:
612–28.
significantly from 92.9kN/mm to 72.6kN/mm as the COV of [15] Thanoon WAM, Alwathaf AH, Noorzaei J, Jaafar MS, Abdulkadir MR. Finite
imperfection size increases from 0.1 to 0.5. element analysis of interlocking mortarless hollow block masonry prism. Comput
6. The mean linear stiffnesses of high-quality, high-medium and Struct 2008;86(6):520–8.
[16] Martínez M, Atamturktur S, Ross B, Thompson J. Assessing the Compressive
medium–low quality interlocking block walls undergo a significant
Behavior of Dry-Stacked Concrete Masonry with Experimentally Informed
decrease with the degradation of block quality. Specifically, the Numerical Models. J Struct Eng 2018;144(7).
mean linear stiffnesses are found to be 511kN/mm, 320kN/mm, and [17] Fundi, S.I., J.W. Kaluli, and J. Kinuthia, Finite element modelling of interlocking
119kN/mm for high-quality, high-medium and medium–low quality stabilized laterite soil block walls. SN Applied Sciences, 2021. 3(2): p. 1-11.
[18] Martínez M, Atamturktur S. Experimental and numerical evaluation of reinforced
walls respectively when COV = 0.1. This correlation is attributable to dry-stacked concrete masonry walls. J Build Eng 2019;22:181–91.
the influences of the number of imperfections exerting on the contact [19] Shi T, et al. Experimental and numerical investigation on the compressive
areas. properties of interlocking blocks. Eng Struct 2021;228:111561.
[20] Huamani K, Enciso R, Gonzales M, Zavaleta D, Aguilar R. Experimental and
numerical evaluation of a stackable compressed earth block masonry system:
In conclusion, this study investigated the influence of block imper Characterization at cyclic shear loads. J Build Eng 2022;60:105139.
fections on the compressive properties of dry-stacked interlocking block [21] Al-Fakih A, Al-Osta MA. Finite Element Analysis of Rubberized Concrete
Interlocking Masonry under Vertical Loading. Materials (Basel) 2022;15(8):2858.
walls. The findings demonstrate the significant impact of imperfections [22] Xie G, Zhang X, Hao H, Bi K, Lin Y. Response of reinforced mortar-less interlocking
on crack patterns, load path, and compressive strength. The study also brick wall under seismic loading. Bull Earthq Eng 2022;20(11):6129–65.
highlights the importance of imperfection size and quantity in deter [23] Xie G, Zhang X, Hao H, Shi T, Cui L, Thomas J. Behaviour of reinforced mortarless
interlocking brick wall under cyclic loading. Eng Struct 2023;283:115890.
mining the properties of the interlocking block walls. The results suggest [24] Andreev K, Sinnema S, Rekik A, Allaoui S, Blond E, Gasser A. Compressive
that low-quality walls may have higher compressive strength than high- behaviour of dry joints in refractory ceramic masonry. Constr Build Mater 2012;34:
medium and medium–low quality walls due to the presence of gaps 402–8.
[25] Lin K, Totoev Y, Liu H, Wei C. Experimental characteristics of dry stack masonry
between blocks. However, the number and severity of localized block
under compression and shear loading. Materials 2015;8(12):8731–44.
damage and failures increase as the quality of the block decreases, [26] Kaushik HB, Rai DC, Jain SK. Stress-strain characteristics of clay brick masonry
leading to lower compressive strengths. under uniaxial compression. J Mater Civ Eng 2007;19(9):728–39.
[27] Briga-Sá A, Silva RA, Gaibor N, Neiva V, Leitão D, Miranda T. Mechanical
Characterization of Masonry Built with iCEBs of Granite Residual Soils with
Cement-Lime Stabilization. Buildings 2022;12(9):1419.
Declaration of Competing Interest [28] Jaafar MS, Alwathaf AH, Thanoon WA, Noorzaei J, Abdulkadir MR. Behaviour of
interlocking mortarless block masonry. Proc Inst Civ Eng - Constr Mater 2006;159
(3):111–7.
The authors declare that they have no known competing financial [29] Chan C, Hover KC, Folliard KJ. Spatial variations in material properties of
interests or personal relationships that could have appeared to influence segmental retaining wall (SRW) units, part II: sampling considerations for
the work reported in this paper. absorption tests. J ASTM Int 2005;2(2):1–18.
[30] Lawrence, S. Random variations in brickwork properties. in Proc., 7th Int. Brick
Masonry Conf. 1985.
Acknowledgments [31] Stewart MG. Spatial variability of pitting corrosion and its influence on structural
fragility and reliability of RC beams in flexure. Struct Saf 2004;26(4):453–70.
[32] Zhu F, Zhou Q, Wang F, Yang Xu. Spatial variability and sensitivity analysis on the
The authors would like to acknowledge the financial support from compressive strength of hollow concrete block masonry wallettes. Constr Build
Australian Research Council for carrying out this study. Mater 2017;140:129–38.
[33] Stewart M, Lawrence S. Model error, structural reliability and partial safety factors
for structural masonry in compression. Masonry International 2007;20(3):107–16.
References [34] Standard, A., AS 3700: masonry structures. 2001, Sydney.
[35] Heffler L, et al. Statistical analysis and spatial correlation of flexural bond strength
for masonry walls. Masonry Int 2008;21(2):59–70.
[1] Zhang X, Hao H, Zheng J, Hernandez F. The mechanical performance of concrete
[36] Li J, Masia MJ, Stewart MG, Lawrence SJ. Spatial variability and stochastic
shear key for prefabricated structures. Adv Struct Eng 2021;24(2):291–306.
strength prediction of unreinforced masonry walls in vertical bending. Eng Struct
[2] Ramamurthy K, Kunhanandan Nambiar EK. Accelerated masonry construction
2014;59:787–97.
review and future prospects. Prog Struct Eng Mater 2004;6(1):1–9.
[37] Fyfe A, Middleton J, Pande G. Numerical evaluation of the influence of some
[3] Wang G, Li Y, Zheng N, Ingham JM. Testing and modelling the in-plane seismic
workmanship defects on the partial factor of safety (GAMMA M) for masonry.
response of clay brick masonry walls with boundary columns made of precast
Masonry Int 2000;13(2):48–53.
concrete interlocking blocks. Eng Struct 2017;131:513–29.
[38] Stewart MG, Lawrence S. Structural reliability of masonry walls in flexure.
[4] Kayaalp FB, Husem M. Enhancement of in-plane load-bearing capacity of masonry
Masonry Int 2002;15(2):48–52.
walls by using interlocking units. Earthquakes Struct 2022;22(5):475–85.
[39] Gooch LJ, Masia MJ, Stewart MG. Application of stochastic numerical analyses in
[5] Zhang X, Hao H, Li C, Do TV. Experimental study on the behavior of precast
the assessment of spatially variable unreinforced masonry walls subjected to in-
segmental column with domed shear key and unbonded Post-Tensioning tendon
plane shear loading. Eng Struct 2021;235:112095.
under impact loading. Eng Struct 2018;173:589–605.
[40] Dyskin A, Estrin Y, Pasternak E. Topological interlocking materials. Archit Mater
[6] Zhang X, Hao H, Li C. Multi-hazard resistance capacity of precast segmental
Nat Eng: Archimats 2019:23–49.
columns under impact and cyclic loading. Int J Protect Struct 2018;9(1):24–43.
[41] Dyskin AV, Estrin Y, Pasternak E, Khor HC, Kanel-Belov AJ. Fracture resistant
[7] Lourenço PB, Pina-Henriques J. Validation of analytical and continuum numerical
structures based on topological interlocking with non-planar contacts. Adv Eng
methods for estimating the compressive strength of masonry. Comput Struct 2006;
Mater 2003;5(3):116–9.
84(29–30):1977–89.
[42] ABAQUS, A., 6.14, Dassault Systèmes Simulia Corp. Provid. RI, USA, 2014.
[8] Sturm T, Ramos LF, Lourenço PB. Characterization of dry-stack interlocking
compressed earth blocks. Mater Struct 2015;48(9):3059–74.
18
T. Shi et al. Structures 56 (2023) 104875
[43] Shi T, et al. Experimental and numerical studies of the shear resistance capacities of [47] Shields MD, Teferra K, Hapij A, Daddazio RP. Refined stratified sampling for
interlocking blocks. J Build Eng 2021;44:103230. efficient Monte Carlo based uncertainty quantification. Reliab Eng Syst Saf 2015;
[44] Simulia, Abaqus Analysis User’s Guide, Version 6.14. 2014, Dassault Systemes 142:310–25.
Providence, RI. [48] Tabbakhha M, Modaressi-Farahmand-Razavi A. Analyzing the effect of
[45] Gorst N, et al. Friction in temporary works. Res Rep 2003:71. workmanship quality on performance of unreinforced masonry walls through
[46] EN, B., 1052-1 (1999) British Standard. Methods of test for masonry—Part 1: numerical methods. Comput Struct 2016;167:1–14.
Determination of compressive strength. CEN European Committee for Standardization.
Central Secretariat: rue de Stassart 36, B-1050 Brussels, September, 1998.
19