Professional Documents
Culture Documents
28asce 29mt.1943 5533.0003113
28asce 29mt.1943 5533.0003113
Abstract: A computational modeling procedure for generating a random aggregate structure (RAS) for recycled concrete aggregate (RCA)
Downloaded from ascelibrary.org by University of Birmingham on 04/10/20. Copyright ASCE. For personal use only; all rights reserved.
materials is proposed in this paper. The two-dimensional aggregate generation was based on a convex hull algorithm to randomize arbitrary
planar shapes of RCA particles. During the generation of RAS, randomizations of the maximum aggregate size and spatial distribution of
RCA particles were considered. Finite-element models for RCA systems were generated based on an extensive image analysis procedure.
Numerical analyses were performed for 12 different RCA systems including crushed-shape and rounded-shape aggregates by varying the
morphological parameters of the RCA particles. The mechanical performance of RCA systems with different particle shapes, maximum
aggregate sizes, adhered mortar content levels, and aggregate ratios were used as factors to evaluate the mechanical performance of RCA
systems under simulated compressive loading. Numerical results were compared with experimental studies from a large statistical database.
The numerical simulation results for the mechanical properties of RCA systems showed represented experimental behavior such as the
simulated elastic moduli and compressive strengths were within one and two standard deviations, respectively, of experimental results from
a robust data set obtained through a detailed literature investigation. The proposed aggregate generation approach and numerical simulation
procedure can be used by researchers to better understand how aggregate morphological properties influence the mechanical behavior of
concrete made with recycled aggregates. DOI: 10.1061/(ASCE)MT.1943-5533.0003113. © 2020 American Society of Civil Engineers.
generate round and angular aggregate shapes. These shapes were Realistically, the aggregates in a concrete system are randomly
identified based on the number of edges that is required to form distributed throughout a contained concrete boundary, varying with
a complete polygon to generate the aggregate. In another study, sizes, shapes, angularities, adhered mortar contents, and other ag-
such polygonal shapes had been generated by using Cartesian co- gregate physical properties. It is understandable that such mesoscale
ordinates or polar coordinates depending on the aggregate shape aggregate representations can be highly challenging to generate and
(i.e., angular or round, respectively) that needs to be formed in thus, very little research has been done using authentic RCA gen-
the RAS (Wang et al. 1999). eration. Therefore, to generate an efficient aggregate model, a way
There have been other attempts reported for generating ran- needs to be implemented such that aggregate shape and distribution
dom aggregate structures by using spherical or elliptical shapes are realistically similar to an actual concrete system.
instead of roundlike aggregates (Bažant et al. 1990; Schorn and In this study, the authors propose a new methodology to generate
Rode 1991; Schlangen and Van Mier 1992; Zhou and Xia 2011). a two-dimensional RAS through a robust computational algorithm
Such shapes are made only for the simplicity in idealizing the that has the ability to manipulate the geometric properties of RCA
aggregates, and the simulated response of systems with elliptical particles. The computational approach accounts for factors such as
or spherical aggregates has shown different performance trends in aggregate gradation, maximum aggregate size, aggregate shape,
comparison with those systems using angular-shaped particles in aggregate angularity, aggregate ratio, adhered mortar content level,
concrete (Bažant et al. 1990; Masad 2013; Wang et al. 2015; Chen adhered mortar location, and spatial distribution of the aggregates at
et al. 2018; Thirumalaiselvi et al. 2019). For example, a numerical a mesoscopic level. Subsequently, the computationally generated
study was conducted by Wang et al. (2015) for polygonal and RAS is mapped onto a two-dimensional finite-element model for
spherical-shaped aggregate systems. It was reported that the corners numerical computations through an image analysis procedure. The
of the polygonal aggregates initiated localized stresses upon load- meshed geometry is then used to analyze and interpret the compres-
ing under tension due to the existing sharp-edged geometry, and sive response with varying aggregate properties to evaluate the sen-
the failure was propagated with the increasing strain levels. How- sitivity of RAC systems to changes in aggregate properties. This
ever, the response of spherical particles showed evenly distributed aggregate generation technique can be used as a powerful tool to
stresses around the aggregates caused by the smooth geometry of understand and better predict the mechanical properties of RCA sys-
the particles. Consequently, this smooth stress distribution around tems through studying the stochastic distribution of RCA morphol-
the spherical aggregate resulted a delay in the fracture event and ogies by developing a large-scale database consisting of numerical
increased the predicted tensile strength. material model geometries, and eventually, it will advance the un-
In consideration of a mesoscopic model for a natural aggregate derstanding of the use of recycled concrete as aggregates appropri-
concrete system, it can be modeled as a two-phased system consist- ately in concrete applications.
ing of coarse aggregate and mortar matrix, which are separated by
interfacial transition zone (ITZ) layer around the coarse aggregate
boundaries. Unlike virgin aggregates, RCA particles are produced Generation of RCA Mesostructure and
by crushing demolished concrete and processed as coarse aggre- Finite-Element Models
gates to reuse in new concrete mixtures. Due to the crushing oper-
The proposed methodology is divided into two stages. In the first
ation, a recycled aggregate consists of two main material phases
stage, the concrete geometry is generated for given aggregate char-
with natural aggregate and adhered mortar attached together. There-
acteristics using a source code developed in MATLAB version
fore, a concrete system with RCA particles can be much more
R2018a. The second stage is focused on generating a finite-element
complex than a natural aggregate system due to the existence of mesh in the commercially available finite-element software DIANA
additional material phases present in the concrete, such as adhered version 10.2 (DianaFEA 2018) through coded scripts written in the
mortar and the bond zone (i.e., old ITZ) between the natural aggre- Python-based programming language.
gate and adhered mortar phases (Otsuki et al. 2003; Liu et al. 2011;
Xiao et al. 2013a). Consequently, computational mesoscopic mod-
els on RCA systems have not been widely studied due to this com- Aggregate Generation
plexity of the material heterogeneity and the geometric irregularities
of the RCA particles. Defining Aggregate Gradation for Recycled
Among the available studies on RCA systems in the literature, Concrete Aggregates
Xiao et al. (2012) examined the mechanical performance and Generally, aggregates (i.e., both fine and coarse) occupy between
damage criteria through a lattice model with vertical and tilted 60% and 75% of the total volume of concrete (Kosmatka et al.
bar elements. Numerical and experimental results concluded that 2011), and a significant portion of the volume (i.e., 40%–50%) is
both responses showed similar trends up to the applied peak load occupied by coarse aggregates (Wang et al. 1999). The mechanical
under compression and tension. Another computational technique performance of an RCA system is heavily governed by (1) the
dmax
consecutively, such that a target aggregate area is obtained for each
successive sieve size pair. Discretization of each sieve aperture
where Pðds Þ = cumulative percentage passing a sieve with aperture
size pair over the selected gradation is significant when aggregate
size of ds ; dmax = maximum aggregate size in the selected aggregate
packing density of the concrete structure is considered. When the
size range; and n = gradation exponent that varies from 0.33 to 0.50
amount of discretization in each aperture size pair is decreased, a
depending on the coarseness of the gradation for a desired maxi-
densified aggregate packing system can be achieved. Conversely, as
mum particle density (Andreasen 1930).
the amount of discretization gets higher in each sieve aperture size
It was found that the maximum particle density in concrete was
pair, a poor packing density in the concrete system would be ob-
achieved when a number between 0.40 and 0.45 was used as the
served. However, the proposed aggregate generation is controlled
gradation exponent n (Hummel 1959; Kennedy et al. 1994). There-
by the aggregate ratio of the concrete system, and the discretization
fore, a value of 0.40 was used as the Fuller curve exponent for the
of sieve aperture size pairs is exceedingly efficient in filling the
randomly generated aggregates shown in this paper. Fig. 1 shows
concrete area with planar shapes for high aggregate ratio systems.
representative aggregate gradations for three types of coarseness
After determining the size distribution of aggregates that needs
levels, where the range that covers coarse aggregates is depicted
to be generated, the aggregate geometry is built within a specified
within 4.75- and 32-mm sieve aperture sizes.
concrete boundary using the convex hull algorithm based on the
The mesostructure of RCA systems were randomly generated
work developed by Barber et al. (1996), and this algorithm is de-
two-dimensionally, and thus RCA particles were considered as pla-
scriptively elaborated in the next subsection.
nar shapes such that each aggregate would have a specific area
within a concrete boundary. The concrete boundary was predeter- Aggregate Generation Algorithm
mined based on an assumed width (W) and height (H), which can The aggregate particles in a concrete system need to be popu-
be defined by the user, and aggregates were generated within this lated as a randomly generated pattern, such that the physical char-
two-dimensional boundary. Representative aggregate areas that are acteristics of the aggregate and the aggregate distribution can be
passed through two consecutive sieve sizes are defined based on the realistically generated within a predetermined concrete boundary
relationship provided in Eq. (2) (e.g., in this paper a 100 × 100 mm section was simulated). Physi-
cal properties of aggregates in the concrete system such as angu-
larity, shape, and size were generated randomly using the convex
100 hull algorithm (Mora and Kwan 2000; Wang et al. 2015; Ma et al.
n = 0.40 2016). A uniformly distributed random number generation was
adopted to locate the vertices of each recurring shape because the
n = 0.50
aggregates were dispersed across the concrete boundary area with-
Total percentage passing, P(ds ) [%]
Once an RCA particle is generated within the specified two- consisting of a 200 × 200 pixel resolution. The dimensions of the
dimensional space (i.e., circular or squared space), the geometric image were set at 200 pixels to reduce computational time con-
coordinates of the aggregate shape need to be translated to a ran- strains and related each pixel to a size of 0.5 mm square. At a later
dom location within the concrete boundary. The translated location stage, the pixels of the raster image are converted to a mesh gen-
of the coordinates is based on a uniformly distributed random num- eration to build the finite-element models. The image is then sub-
ber generation technique that depends on the size of the concrete jected to a pixel color scan in which the color intensities of each
specimen in which the aggregates are being generated. In this pa- pixel are identified in a columnwise array.
per, the aggregate translation was performed through a uniformly The pixel color intensity is measured by a scale of red, green,
distributed random number generator within 0 and 100 mm along and blue (RGB) values for each color that varies from 0 to 255.
x- and y-axes with respect to the origin located at the lower-left By scanning the image pixels in a columnwise direction, the pixel
bottom corner of the squared concrete boundary. However, in this color transition from one material phase to the other can be filtered,
process, some random aggregate translations can be overlapped where eventually, the ITZ regions can be easily distinguished. For
with each other and may be unrealistically closer to the concrete example, the color transition from red to blue and red to green, or
boundary edge. To address the former issue, an overlapping con- vice versa, controls the material region corresponding to the new
dition was implemented by enforcing a logical array to the trans- ITZ, whereas the color transition from green to blue, or vice versa,
lated aggregate shapes until the overlapped area was equaled to controls the material region corresponding to the old ITZ. Logical
zero. To control the aggregates being translated near the concrete operators were then used to identify the color transitions for old ITZ
boundary edge, border effects were accounted such that not a single regions and new ITZ regions, where white and black were assigned
particle was allowed to come closer to the concrete surface than a to those material regions, respectively. The technique that was used
prescribed distance. In this study, the minimum boundary gap be- in the image processing work adopted through raster images was
tween the RCA particle and the nearest concrete edge was taken as effective to identify the ITZ regions in the RCA aggregate structure.
The RGB values for those five types of colored pixels are provided
in Table 1.
After the aggregate image is processed, the refined RCA sys-
tem now consists of five unique color-coded pixels representing
Sieve aperture dimension
the material phases. The pixel color information of the image was
stored in an n × 3 matrix, where n is the total number of pixels
(i.e., 200 × 200 ¼ 40,000) in the refined image, and the number 3
represents the RGB value intensity entries. Fig. 4 depicts the differ-
ence between the initial image and the processed image after the
material phases included in the concrete system. Subsequently, this
colored pixel image was used to generate the finite-element mesh
to numerically simulate different RCA systems in a finite-element
framework.
Numerous sophisticated simulation platforms are available to
(a) (b)
simulate the performance of concrete materials. The goal of this
paper is to use the proposed methodology for simulating RCA sys-
Aggregate phase tem geometries and understand how the uniaxial compressive re-
sponse is affected by variations of a range of aggregate geometries.
RCA systems of 100 × 100 mm samples were simulated under
Adhered mortar phase uniaxial compression through finite-element modeling. The meso-
scopic modeling geometry of the RCA systems were based on the
Random points pixel color identification that was obtained from the refined RAS
image. The pixel dimension of the raster image was 200 × 200,
(c) where each pixel was represented by a quadrilateral element in the
finite-element model. The size of each element was selected to be
Fig. 2. Generation of RCA particles using random points allocated
0.5 mm, such that the size effects for those individual material
within the specified space through convex hull algorithm for different
phases were reasonable and computationally efficient.
aggregate shapes and angularity: (a) roundlike aggregate with low an-
In total, 12 RCA systems were simulated, which varied in terms
gularity using 15 random points; (b) crushedlike aggregate with high
of aggregate properties for RCA shape, RCA ratio, adhered mortar
angularity using eight random points; and (c) material legend.
content, and number of attached locations of adhered mortar to the
Fig. 3. Flowchart for generation of the random aggregate structure for RCA systems.
Table 1. RGB intensity values assigned to each material in the RCA RCA used in the generation (i.e., C for crushed and R for rounded),
system 12 represents the maximum RCA size in millimeters, L denotes the
Material Color RGB value level of adhered mortar content of the system (i.e., N for no adhered
mortar content, L for low, M for moderate, and H for high), the
Aggregate Blue [0, 0, 255]
Adhered mortar Green [0, 255, 0] single digit represents whether the RCA system was made using
Cement mortar matrix Red [255, 0, 0] none, one, or two adhered mortar locations, and the last decimal
Old interfacial transition zone White [255, 255, 255] value indicates the RCA ratio of the concrete system.
New interfacial transition zone Black [0, 0, 0] Two adhered mortar locations were used when the adhered mor-
tar content was moderate or high in quantity so that the adhered
mortar generation is evenly distributed on the natural aggregate.
natural aggregate in the system. The adhered mortar content is cat- Adhered mortar locations that had three or more locations were
egorized as low, moderate, and high, which corresponded to ad- not incorporated in the aggregate generation algorithm due to the
hered mortar contents of 1%–20%, 21%–50%, and 51%–100%, significant computational time that would have been required to
respectively. A shorthand notation for identifying the simulation generate aggregates. Realistically, there is a possibility of adhered
models is used as C-12L1-0.50, where C is the shape of the mortar phases surrounding the natural aggregate on all sides, but
Mortar matrix
New ITZ
Old ITZ
(a) (b) (c)
Fig. 4. Material assignment based on the pixel identification through image analysis: (a) original raster image obtained from RAS; (b) processed
Downloaded from ascelibrary.org by University of Birmingham on 04/10/20. Copyright ASCE. For personal use only; all rights reserved.
image after color refinement of the material regions; and (c) material legend.
the aggregate generation method presented in this paper only con- assigned an elastic modulus, compressive strength, tensile strength,
siders one or two surface locations. The main focus of the proposed and Poisson’s ratio as a result of such aforementioned parametric
method was to implement a novel methodology to capture funda- studies. Representative values for aggregate, adhered mortar, and
mental RCA aggregate property variations (i.e., shape, size, grada- mortar matrix elastic moduli were used as input variables based
tion, adhered mortar content, aggregate ratio, among others), and to experimental studies performed on the material properties in RCA
understand how those factors would affect the mechanical behavior systems (Xiao et al. 2013a, b, c).
of numerically simulated RCA systems. The simulation metrics are Representative material properties for interfacial transition
given in Table 2. zones were obtained through the nanoindentation results from
The RCA system is a heterogeneous material that requires prop- prior research (Zhu et al. 2004; Mondal et al. 2009; Wang et al.
erty assignments for each material phase in concrete mesoscale 2009). Poisson’s ratios were defined as 0.20 for both the interfacial
modeling. Regardless of the fact that mechanical properties of each transition zones according to experimental work by Ramesh et al.
component in the RCA system may vary due to several reasons (1996). The tensile fracture energy due to the crack-opening mode
such as aggregate source, age of the parent concrete, mixture pro- of a material for the aggregate phase was obtained from the avail-
portions of the concrete, or chemical influence from the additives, able literature (Winkler 1974), and this parameter was assumed
average mechanical properties were assigned for the current mate- based on proportional scaling or obtained from the literature (Zhu
rial phases in this study by considering parametric analysis results et al. 2004; Mondal et al. 2009; Wang et al. 2009; Xiao et al.
that were used in a prior research (Xiao et al. 2013b). Similarly, 2013a, b, c) for other material phases in the RCA system. A con-
a factorial design analysis on the influence of variability of material stant shear retention factor of 1% was used for each material be-
properties in RCA components studied by the authors showed cause the shear stiffness decreases with the crack initiation in the
that the mechanical properties of RCA systems were governed by composite material. Table 3 summarizes the material properties that
the relative stiffness difference of the major material components were used in the numerical analysis for different RCA systems.
(i.e., natural aggregate, cement mortar matrix, and adhered mortar)
(Jayasuriya et al. 2018a). Due to such plausible material property Damage Constitutive Model
variabilities, a reliable set of material property data that fit the com-
ponents in RCA systems need to be used in the numerical sim- In computational modeling of concrete materials, damage and
ulations as input parameters. Therefore, each material phase was fracture phenomena have been modeled using elasticity, plasticity,
viscoelasticity, and fracture energy–based formulations (Ghosh and
Chaudhuri 2013; Wang et al. 2015; Chen et al. 2018). Concrete-
based materials are known to have a quasi-brittle response (de
Table 2. Simulation metrics of RCA systems prepared for the numerical Borst 2002), and failure occurs due to the initiation and propaga-
study tion of microcracks under applied stresses or deformations. In this
No. of study, the numerical simulations were subjected to compressive
Maximum Adhered adhered loading, and a total strain–based fixed crack constitutive material
Aggregate aggregate mortar level mortar RCA model was adopted with a parabolic softening curve in compres-
Simulation ID shape size (mm) [content (%)] locations ratio sion (Feenstra 1993).
C-12N0-0.50 Crushed 12 None (0%) 0 0.50
C-12L1-0.50 12 Low (8.60%) 1 0.50
C-25L1-0.50 25 Low (9.97%) 1 0.50
C-12L1-0.30 12 Low (6.50%) 1 0.30 Table 3. Material properties of the constituents in RCA system
C-12M2-0.50 12 Moderate (31.02%) 2 0.50 Elastic Compressive Tensile Fracture
C-12H2-0.50 12 High (78.43%) 2 0.50 modulus strength strength Poisson’s energy
Material phase (GPa) (MPa) (MPa) ratio (N/mm)
R-12N0-0.50 Rounded 12 None (0%) 0 0.50
R-12L1-0.50 12 Low (7.44%) 1 0.50 Aggregate 70.00 144.0 9.60 0.16 0.1628
R-25L1-0.50 25 Low (10.97%) 1 0.50 Adhered mortar 28.75 45.0 3.00 0.22 0.0509
R-12L1-0.30 12 Low (9.38%) 1 0.30 Mortar matrix 26.45 41.4 2.76 0.22 0.0468
R-12M2-0.50 12 Moderate (46.98%) 2 0.50 Old ITZ 17.50 36.0 2.40 0.20 0.0407
R-12H2-0.50 12 High (63.03%) 2 0.50 New ITZ 16.10 33.1 2.21 0.20 0.0371
tions, and the remaining nodal points were restrained only along the tified based on the assigned color intensity for each corresponding
global y-direction. element of the RCA system.
The analysis procedure of the finite-element models were based
Finite-Element Mesh Generation and Analysis Procedure on an incremental-iterative solution procedure followed by a quasi-
The preprocessing functionality in the DIANA interactive environ- Newton secant method with Broyden-Fletcher-Goldfarb-Shanno
ment (DianaIE) was used to mesh the randomly generated geom- (BFGS) iteration method. This iterative solution procedure was
etries through an automated Python script. Because the concrete chosen because it showed quicker and more stable convergence
compared with regular and modified Newton-Raphson methods.
A strain increment of 10 μϵ was applied to the specimen through
prescribed displacements where at times, 1-μϵ strain increments
were used to improve the convergence performance. Three different
convergence criteria were set in the analyses for force norm, dis-
placement norm, and energy norm with respective tolerances of
1%, 0.1%, and 0.01%.
Fig. 6. Representative finite-element mesh generations: (a) crushed RAC system; (b) rounded RAC system; and (c) material legend.
10
mortar content. The simulated finite-element models for crushed
aggregates showed how RCA physical properties such as RCA par-
ticle size, adhered mortar content, and RCA ratio can affect the
5 strength properties.
Modulus of elasticity of RCA systems quantifies the resistance
of elastic deformation under an applied load. The stress-strain rela-
0 tionships in Fig. 7 depict a variety of elastic moduli levels depending
0 0.0002 0.0004 0.0006 0.0008 0.001 0.0012 on the composite stiffness of the RCA systems. Representative
Applied Strain elastic moduli of RCA systems were calculated by computing the
gradient between the point of zero stress and strain, and the point at
Fig. 7. Stress-strain relationships for crushed RCA systems through 40% of the maximum compressive stress [ASTM C469/C469M
finite-element simulations. (ASTM 2014) and BS EN 1992-1-1 (BSI 2004)]. The highest elastic
modulus was obtained in simulation C-25L1-0.50, where a large
portion of the aggregate area was covered by the natural aggregate
phase. In contrast, the lowest elastic modulus was observed in sim-
stiffness of the recycled aggregate was reduced, and it showed com- ulation C-12H2-0.50, which had the highest adhered mortar content
patibility improvements with enhanced mechanical response com- in the RCA system. Based on these results, the RCA elastic modulus
pared with the other crushed RCA systems. was governed by the stiffness of the natural aggregate portion be-
This behavior was characterized through shrinkage testing by cause the material stiffness of this phase was greater than any of the
Adams et al. (2016), and numerically confirmed by Jayasuriya et al. material stiffness in the RCA system.
(2018a), where it was concluded that higher adhered mortar contents Table 4 presents the variation of compressive strength and elas-
led to an improved cracking resistance against drying shrinkage, tic modulus obtained from the numerical analyses where a trend
which allowed the material to further deform under shrinkage was observed that the natural aggregate ratio of the total concrete
stresses. In contrast, the simulation C-12N0-0.50 (i.e., natural ag- area was proportional to the elastic modulus of the RCA system.
gregate system) exhibited a lower strain capacity due to the exist- The ratios of the natural aggregate phase were calculated based
ence of large stiffness heterogeneity of material stiffnesses in the on the number of corresponding blue color elements in the natural
concrete system with no adhered mortar attached to the natural aggregate area over the total number of elements (i.e., 40,000
aggregate. elements) in each RCA model. When the adhered mortar content
When the maximum aggregate size was increased for RCA sys- was increased from moderate to high levels, the elastic modulus
tems with lower amount of adhered mortar content, the compressive was drastically decreased due to the low stiffness of the adhered
strength dropped, as shown in the simulation C-25L1-0.50 of Fig. 7. mortar resulting a lower global stiffness. Similar type of material
As the compressive load was applied on this system, the strain de- behaviors were found based on other numerical studies simulated
formation favored to propagate toward the material phase that had by the authors for stress-strain behavior with RCA cored samples
the highest stiffness value (i.e., natural aggregate). However, the under the same loading condition for varying adhered mortar con-
strain deformation in natural aggregate was extremely low due to tents adopting a constant aggregate geometry (Jayasuriya et al.
the higher material stiffness and material toughness. Therefore, ten- 2018a, b).
sile strains from the lateral deformations were easily developed on Overall, the RCA compressive strength performance improved
the outside of the natural aggregate region and penetrated into the with increasing adhered mortar levels. For example, it was found
other material phases followed by the interfacial boundaries. Among that the compressive strength improved by 14% when increasing
the simulations with a maximum aggregate size of 12 mm and ag- the adhered mortar content from 8.60% to 78.43%. Further, at
gregate ratio of 0.5 (i.e., C-12L1-0.50, C-12M2-0.50, and C-12H2- higher adhered mortar level, the RCA system reached larger strains
0.50), it was observed that, as the adhered mortar content was before softening. On average, the peak simulated strain was 891 μϵ
increased from 8.60% to 78.43%, the peak strain deformation in- with a standard deviation of 20 μϵ for systems containing low to
creased by approximately 28%. These results confirmed that higher moderate adhered mortar levels; however, the strain capacity in-
adhered mortar contents in RCA systems resulted higher defor- creased by 35% when a high adhered mortar level was used in the
mation capacities. This behavior was mainly due to the improved RCA system. The elastic modulus was governed by the area of
stiffness compatibility between the adhered mortar and mortar ma- natural aggregate in the RCA systems and decreased with increas-
trix which resulted in minor stress localization upon loading. The ing amounts of adhered mortar content. For example, the elastic
simulation C-12L1-0.30 showed the lowest compressive strength of modulus decreased by 7.1% when the adhered mortar increased
22.6 MPa due to the reduction of aggregate ratio with low adhered from 8.60% to 78.43%.
Maximum
12 mm 12 mm 25 mm 12 mm 12 mm 12 mm
Aggregate size
Adhered mortar 0/0% 1/8.60% 1/9.97% 1/6.50% 2/31.02% 2/78.43%
location/content
Aggregate 0.50 0.50 0.50 0.30 0.50 0.50
ratio
Downloaded from ascelibrary.org by University of Birmingham on 04/10/20. Copyright ASCE. For personal use only; all rights reserved.
(a)
RCA principal
compressive
strain profiles
Maximum
12 mm 12 mm 25 mm 12 mm 12 mm 12 mm
Aggregate size
Adhered mortar 0/0% 1/7.44% 1/10.97% 1/9.38% 2/46.98% 2/63.03%
location/content
Aggregate 0.50 0.50 0.50 0.30 0.50 0.50
ratio
(b)
Fig. 9. Principal compressive strain profile for the simulated RCA systems: (a) crushed RCA systems; and (b) round RCA systems.
profiles of RCA systems, especially near the peak response, be- (Xiao et al. 2013a, b), which further confirms that the current
cause failure of concrete can be promoted due to large deformation numerical simulation results exhibit a realistic point of view.
under compression or due to the lateral deformation of the material Despite the shape of the recycled aggregates, the principal ten-
caused by the global response. sile strains have initiated near the new ITZ at first and then further
It can be seen from Fig. 9 that under the uniaxial load, the RCA propagated into the old ITZ. However, this sequential order of the
systems have performed quite satisfactorily such that the principal cracking pattern could be reversed depending on the geometric ori-
compressive strains have just reached the concrete crushing strain entation of a recycled aggregate that would pertain an interfacial
capacity of 0.003 at the peak level that can allow further deforma- transition zone (i.e., new or old ITZ) along the loading direction.
tions under compression. Therefore, it can be confirmed that the Alternations in strain transfer mechanism were attributed to strain
strain distributions of all the concrete systems simulated under com- localization near the intersection of the ITZ phases (i.e., location
pression showed similar trends regardless of the induced hetero- where both new and old ITZs). Rapid tensile strains localization
geneity from the material stiffness, and at the peak response, the was observed at locations when the old ITZ was aligned in the
new ITZ was the only material phase that was subjected to higher direction of the loading plane, causing accelerated damage pro-
strain levels. But considering Fig. 10, the corresponding principal gression. Thus, more tensile strain deformations were accumulated
tensile strains have already reached that capacity, and the material on such areas and interfacial cracking failure potentials were
fracturing was initiated even before the ultimate crushing occurred. expedited.
This was identified as the material having undergone a softening The two RCA systems that contained 25-mm maximum aggre-
phase where the material can no longer resist the applied strain de- gate size showed that larger particles tend to induce higher ten-
formation due to the drastic change of the overall material stiffness. sile strain deformations (i.e., cracking potential) as depicted under
During this period, according to the mesoscopic simulations per- simulations C-25L1-0.50 and R-25L1-0.50 in Figs. 10(a and b).
formed, it was believed to have microcracks initiated at the new This was due to the higher effective curvature of the interface crack-
interfacial transition zones, and eventually failed due to interfacial ing path associated with the larger aggregate size. This resulted a
cracking while the material response continued to proceed beyond poor performance of the RCA system providing a low compressive
the peak level as shown in Fig. 10. Similar material behavior strength. Further, it can be seen from the simulation R-25L1-0.50
was observed in other experimental findings in the literature in Fig. 10(b) that when the RCA system reached its peak stress,
Maximum
12 mm 12 mm 25 mm 12 mm 12 mm 12 mm
Aggregate size
Adhered mortar 0/0% 1/8.60% 1/9.97% 1/6.50% 2/31.02% 2/78.43%
location/content
Aggregate 0.50 0.50 0.50 0.30 0.50 0.50
ratio
Downloaded from ascelibrary.org by University of Birmingham on 04/10/20. Copyright ASCE. For personal use only; all rights reserved.
(a)
RCA principal
tensile strain
profiles
Maximum
12 mm 12 mm 25 mm 12 mm 12 mm 12 mm
Aggregate size
Adhered mortar 0/0% 1/7.44% 1/10.97% 1/9.38% 2/46.98% 2/63.03%
location/content
Aggregate 0.50 0.50 0.50 0.30 0.50 0.50
ratio
(b)
Fig. 10. Principal tensile strain profile for the simulated RCA systems: (a) crushed RCA systems; and (b) round RCA systems.
the principal tensile strains created a clear cracking surface on the procedures with cube and cylinder specimens to quantify the com-
material due to the smooth formation and interconnectedness of pressive strength, where the elastic modulus was evaluated using
the tensile strain contours around the round aggregates. This ex- cylindrical specimens.
plains that the mechanical performance of RCA systems with larger In total, 282 mechanical properties were found during the liter-
round aggregate particles and low adhered mortar content level is ature investigation, where 102 samples were found for systems with
highly vulnerable of getting damaged under applied compressive 0% replacement levels, and the remaining samples were obtained
loading. from 100% replacement levels. The histogram distribution of the
compressive strengths and elastic moduli for those replacement lev-
els are depicted in Figs. 11(a–d). The histograms of mechanical
Comparison of Numerical Analysis Results with properties exhibited a slight skewness to the left because the data
Experimental Data through a Literature were fitted to a Weibull distribution. It has been found out that the
Investigation Weibull distribution is more applicable to fit brittle material proper-
The numerical study was carried out based on randomly generated ties and provide reliable justifications on the variability of such ma-
aggregate distributions for RCA systems containing 0% and 100% terials in comparison with normal-distribution functions (Colman
replacement levels of recycled aggregates. In order to verify the et al. 2014). Compressive responses from the numerical simulations
mechanical responses of simulated RCA systems, a number exper- for 0% and 100% RCA replacement levels are indicated by light
imental studies were analyzed and compared with the numerical gray circular markers in Figs. 11(a–d), in which the numerical
analysis results. The Appendix lists the literature studies that were results were within two standard deviations of the overall mean dis-
incorporated to study the variability of compressive strength and tribution in the corresponding data set.
elastic modulus of RCA systems compared with the numerical sim- The compressive strength of those concrete systems with 0%
ulations obtained from the current study. The literature investigation and 100% replacement levels in Figs. 11(a and c) deviated by 6.5
included normal-weight concrete systems consisting of recycled and 6.7 MPa, respectively, compared with the mean of the exper-
aggregates and natural aggregates, where no system was higher imental data set. The deviation for the elastic modulus was 6.3 and
than a compressive strength of 40 MPa so that potential outliers 5.2 GPa for 0% and 100% RCA replacement levels in the concrete
were eliminated. The experimental studies used standardized testing systems, respectively, as shown in Figs. 11(b and d). The numerical
Fig. 11. Histogram of the material properties of two systems: (a) compressive strength for 0% replacement level; (b) elastic modulus for 0%
replacement level; (c) compressive strength for 100% replacement level; and (d) elastic modulus for 100% replacement level.
results on elastic moduli had showed less variability compared with morphologies and geometries, 12 numerical simulations were per-
the data in the literature, which varied within one standard devia- formed for crushed and round RCA systems subjected to a uniaxial
tion of the mean distribution. Due to the fact that the compressive compressive load to evaluate the capabilities of the methodology.
strength is governed by aggregate size, aggregate shape, and relative An image analysis procedure was adopted to generate the aggregate
stiffness, it may show higher variability in performance in com- geometry and the finite-element mesh consisting of eight-noded
parison with the elastic modulus, which is mainly controlled by the quadratic quadrilaterals. Corresponding mechanical responses and
amount of aggregate area of the concrete system and thus showed principal strain distribution patterns were studied for each RCA
lower variability. Hence, the statistical results conveyed a realistic system, and eventually, the numerical results were compared with
response of the computational results from randomly generated existing experimental data through a database that was developed
aggregates. by the authors from the literature. The following conclusions
are drawn from the numerical and statistical analyses performed
through the aggregate generation for RCA concrete systems:
Conclusion • Due to the combined effects of higher geometric angularity and
aggregate interlock, crushed RCA systems showed higher peak
A two-dimensional modeling methodology to generate random strengths compared with round RCA systems.
aggregate structures for RCA systems was proposed in this paper. • Higher adhered mortar levels contributed to relieving strain
The aggregate generation was formulated with a convex hull al- concentrations during loading, which allowed RCA systems to
gorithm, where morphological aggregate properties such as RCA reach higher strain capacities before softening.
maximum size, shape, aggregate ratio, gradation parameters, ad- • The general failure path due to lateral deformation of the con-
hered mortar content, and locations can be manipulated for a de- crete systems initiated from new ITZ to old ITZ and branched
sired mesolevel RCA system. Considering those random aggregate into other material phases.
response because it was governed by many factors such as in mechanical properties of RCA systems.
Appendix. RCA Mechanical Properties Obtained from Experimental Results through a Literature Investigation
Data Availability Statement aggregates.” Constr. Build. Mater. 102 (Part 1): 802–810. https://doi
.org/10.1016/j.conbuildmat.2015.11.022.
Some or all data, models, or code generated or used during the Ajdukiewicz, A., and A. Kliszczewicz. 2007. “Comparative tests of beams
study are available in a repository or online in accordance with and columns made of recycled aggregate concrete and natural aggregate
funder data retention policies developed by Jayasuriya et al. (2019). concrete.” J. Adv. Concr. Technol. 5 (2): 259–273. https://doi.org/10
.3151/jact.5.259.
Andreasen, A. H. M. 1930. “Ueber die beziehung zwischen kornabstufung
Acknowledgments und zwischenraum in produkten aus losen körnern (mit einigen exper-
imenten).” Kolloid-Zeitschrift 50 (3): 217–228. https://doi.org/10.1007
The authors gratefully acknowledge the support of the American /BF01422986.
Concrete Institute (ACI) Foundation for their generous funding and ASTM. 2014. Standard test method for static modulus of elasticity and
continuous support provided under the Project ACI CRC 18.571. Poisson’s ratio of concrete in compression. ASTM C469/C469M.
Special thanks are also acknowledged to Emily S. Shibata and Tola West Conshohocken, PA: ASTM.
Chen at New Jersey Institute of Technology in John A. Reif, Jr., Barber, C. B., D. P. Dobkin, and H. Huhdanpaa. 1996. “The quickhull
Department of Civil and Environmental Engineering for their con- algorithm for convex hulls.” ACM Trans. Math. Software 22 (4):
tribution in data collection for the statistical database analysis. 469–483. https://doi.org/10.1145/235815.235821.
Bažant, Z., M. Tabbara, M. Kazemi, and G. Pijaudier-Cabot. 1990.
“Random particle model for fracture of aggregate or fiber composites.”
References J. Eng. Mech. 116 (8): 1686–1705. https://doi.org/10.1061/(ASCE)
0733-9399(1990)116:8(1686).
Adams, M., T. Fu, A. G. Cabrera, M. Morales, J. Ideker, and O. Isgor. 2016. Beddow, J., and T. Meloy. 1980. Testing and characterization of powders
“Cracking susceptibility of concrete made with coarse recycled concrete and fine particles. London: Heyden.
.011.
aggregates and pores.” Constr. Build. Mater. 75 (Jan): 35–45. https://
Otsuki, N., S. I. Miyazato, and W. Yodsudjai. 2003. “Influence of recycled
doi.org/10.1016/j.conbuildmat.2014.09.069.
aggregate on interfacial transition zone, strength, chloride penetration
Wang, Z., A. Kwan, and H. Chan. 1999. “Mesoscopic study of concrete I:
and carbonation of concrete.” J. Mater. Civ. Eng. 15 (5): 443–451.
Generation of random aggregate structure and finite element mesh.”
https://doi.org/10.1061/(ASCE)0899-1561(2003)15:5(443).
Comput. Struct. 70 (5): 533–544. https://doi.org/10.1016/S0045-7949
Padmini, A. K., K. Ramamurthy, and M. S. Mathews. 2009. “Influence
(98)00177-1.
of parent concrete on the properties of recycled aggregate concrete.”
Constr. Build. Mater. 23 (2): 829–836. https://doi.org/10.1016/j Wardeh, G., E. Ghorbel, and H. Gomart. 2015. “Mix design and properties
.conbuildmat.2008.03.006. of recycled aggregate concretes: Applicability of Eurocode 2.” Int. J.
Pedro, D., J. De Brito, and L. Evangelista. 2014. “Performance of concrete Concr. Struct. Mater. 9 (1): 1–20. https://doi.org/10.1007/s40069
made with aggregates recycled from precasting industry waste: Influ- -014-0087-y.
ence of the crushing process.” Mater. Struct. 48 (12): 3965–3978. Wei, X. 2006. “Experimental study on influence of recycled coarse aggre-
https://doi.org/10.1617/s11527-014-0456-7. gates contents on properties of recycled aggregate concrete.” Concrete
Peng, Y., Y. Liu, J. Pu, and L. Zhang. 2013. “Application of base force 10: 45–47.
element method to mesomechanics analysis for recycled aggregate Winkler, E. 1974. “Stone: Properties, durability in man’s environment.” In
concrete.” Math. Prob. Eng. 2013 (1): 1–8. https://dx.doi.org/10 Vol. 2 of Environmental geomorphology and landscape conservation,
.1155/2013/292801. 230. New York: Springer.
Pepe, M., R. D. T. Filho, E. A. B. Koenders, and E. Martinelli. 2014. Wittmann, F., P. Roelfstra, and H. Sadouki. 1985. “Simulation and analysis
“Alternative processing procedures for recycled aggregates in structural of composite structures.” Mater. Sci. Eng. 68 (2): 239–248. https://doi
concrete.” Constr. Build. Mater. 69 (Oct): 124–132. https://doi.org/10 .org/10.1016/0025-5416(85)90413-6.
.1016/j.conbuildmat.2014.06.084. Xiao, J., J. Li, and C. Zhang. 2005. “Mechanical properties of recycled
Pereira, P., L. Evangelista, and J. de Brito. 2012. “The effect of superplas- aggregate concrete under uniaxial loading.” Cem. Concr. Res. 35 (6):
ticizers on the mechanical performance of concrete made with fine re- 1187–1194. https://doi.org/10.1016/j.cemconres.2004.09.020.
cycled concrete aggregates.” Cem. Concr. Compos. 34 (9): 1044–1052. Xiao, J., W. Li, D. Corr, and S. Shah. 2013a. “Effects of interfacial tran-
https://doi.org/10.1016/j.cemconcomp.2012.06.009. sition zones on the stress–strain behavior of modeled recycled aggregate
Rahal, K. 2007. “Mechanical properties of concrete with recycled coarse concrete.” Cem. Concr. Res. 52 (Oct): 82–99. https://doi.org/10.1016/j
aggregate.” Build. Environ. 42 (1): 407–415. https://doi.org/10.1016/j .cemconres.2013.05.004.
.buildenv.2005.07.033. Xiao, J., W. Li, D. Corr, and S. Shah. 2013b. “Simulation study on the stress
Ramesh, G., E. Sotelino, and W. Chen. 1996. “Effect of transition zone distribution in modeled recycled aggregate concrete under uniaxial
on elastic moduli of concrete materials.” Cem. Concr. Res. 26 (4): compression.” J. Mater. Civ. Eng. 25 (Apr): 504–518. https://doi.org/10
611–622. https://doi.org/10.1016/0008-8846(96)00016-6. .1061/(ASCE)MT.1943-5533.0000598.
Ren, W. Y., Z. J. Yang, and P. Withers. 2013. “Meso-scale fracture Xiao, J., W. Li, Z. Sun, and S. P. Shah. 2013c. “Crack propagation in
modelling of concrete based on X-ray computed tomography images.” recycled aggregate concrete under uniaxial compressive loading.” ACI
In Proc., 5th Asia-Pacific Congress on Computational Mechanics Mater. J. 109 (4): 451–462.
(APCOM), 1–10. Taipei, Taiwan: Taipei International Convention
Xiao, J., Q. Liu, and Y. C. Wu. 2012. “Numerical and experimental studies
Center.
on fracture process of recycled concrete.” Fatigue Fract. Eng. Mater.
Sato, R., I. Maruyama, T. Sogabe, and M. Sogo. 2007. “Flexural behavior
Struct. 35 (8): 801–808. https://doi.org/10.1111/j.1460-2695.2012
of reinforced recycled concrete beams.” J. Adv. Concr. Technol. 5 (1):
.01673.x.
43–61. https://doi.org/10.3151/jact.5.43.
Yang, K. H., H. S. Chung, and A. F. Ashour. 2008. “Influence of type
Schlangen, E., and J. G. M. Van Mier. 1992. “Simple lattice model for
and replacement level of recycled aggregates on concrete properties.”
numerical simulation of fracture of concrete materials and structures.”
Mater. Struct. 25 (9): 534–542. https://doi.org/10.1007/BF02472449. ACI Mater. J. 105 (3): 289–296.
Schorn, H., and U. Rode. 1991. “Numerical simulation of crack propaga- Yang, Z., X. Su, J. Chen, and G. Liu. 2009. “Monte Carlo simulation
tion of microcracking to fracture.” Cem. Concr. Compos. 13 (2): 87–94. of complex cohesive fracture in random heterogeneous quasi-brittle
https://doi.org/10.1016/0958-9465(91)90003-Z. materials.” Int. J. Solids Struct. 46 (17): 3222–3234. https://doi.org/10
Sheen, Y.-N., H.-Y. Wang, Y.-P. Juang, and D.-H. Le. 2013. “Assessment .1016/j.ijsolstr.2009.04.013.
on the engineering properties of ready-mixed concrete using recycled Ye, H. 2009. “Experimental study on mechanical properties of concrete
aggregates.” Constr. Build. Mater. 45 (Aug): 298–305. https://doi.org made with high quality recycled aggregates.” Sichuan Build. Sci. 35 (5):
/10.1016/j.conbuildmat.2013.03.072. 195–199.
Thirumalaiselvi, A., N. Anandavalli, and J. Rajasankar. 2019. “Mesoscale Yin, A., X. Yang, and Z. Yang. 2013. “2D and 3D fracture modeling of
studies on the effect of aggregate shape idealisation in concrete.” Mag. asphalt mixture with randomly distributed aggregates and embedded
Concr. Res. 71 (5): 244–259. https://doi.org/10.1680/jmacr.17.00184. cohesive cracks.” Procedia IUTAM 6: 114–122. https://doi.org/10
Thomas, C., J. Setién, J. Polanco, P. Alaejos, and M. Sánchez de Juan. .1016/j.piutam.2013.01.013.
2013. “Durability of recycled aggregate concrete.” Constr. Build. Zaitsev, Y., and F. Wittmann. 1977. “Crack propagation in a two-phase
Mater. 40: 1054–1065. https://doi.org/10.1016/j.conbuildmat.2012.11 material such as concrete.” Fracture 3: 1197–1203. https://doi.org/10
.106. .1016/B978-0-08-022144-1.50100-0.