applied

sciences
Article
Meso-Scale Failure of Freezing–Thawing Damage of
Concrete under Uniaxial Compression
Zheng Si *, Xiaoqi Du, Lingzhi Huang and Yanlong Li
State Key Laboratory Base of Eco-Hydraulic Engineering in Arid Area (Xi’an University of Technology), Xi’an
University of Technology, Xi’an 710048, China; 2170420092@stu.xaut.edu.cn (X.D.); hlzsz@126.com (L.H.);
liyanlong@xaut.edu.cn (Y.L.)
* Correspondence: sizheng@xaut.edu.cn




Received: 15 January 2020; Accepted: 10 February 2020; Published: 13 February 2020


Featured Application: Model established by discrete element method can predict damage degree
of freeze-thaw concrete.

Abstract: In this study, discrete element PFC3D software was used to simulate the uniaxial compression
of C30 concrete specimen models after freezing–thawing damage. Based on the thickness of the
freezing–thawing damage layer of the concrete and the uniaxial compression test, the meso-parameters
of cement mortar in concrete and the meso-parameters of weak interface between the cement
mortar and coarse aggregate under different freezing–thawing cycles were calibrated. The axial
compression deformation produced in the experiment was compared with the axial displacement
of particles obtained by numerical simulation to verify the accuracy of the model and the selected
meso-parameters. Based on the numerical model, the effects of freeze–thaw damage on the mechanical
properties of concrete were studied from the perspectives of the contact force between particles, crack
development, and concrete cracking, and the characteristics of meso-scale failure of concrete model
after freezing–thawing damage were analyzed. The results show that the stress–strain curves obtained
by uniaxial compression simulation in the PFC3D-based numerical model of freezing–thawing concrete
are consistent with the test results. During the uniaxial compression process in the freezing–thawing
concrete model, the contact force at different stage points between the particles, crack development,
and the number of cracked concrete fragments quantitatively reflected the degree of meso-scale
concrete failure. With increasing numbers of freeze–thaw cycles, the failure severity of the model
increased with obvious penetrating cracks, which is consistent with the failure pattern obtained by
physical experiments. Therefore, numerical experiments can be used to study the meso-scale failure
modes of concrete.

Keywords: C30 concrete; freezing–thawing damage; uniaxial compression test; particle flow;
meso-scale failure

1. Introduction
Concrete is the dominant component of infrastructure around the world, for which the service
environment has become increasingly complex. Freezing–thawing damage has become one of the main
risk factors affecting concrete health and longevity, especially in cold areas [1,2]. Freezing–thawing
cycles can affect the basic mechanical properties of concrete, such as compressive strength, tensile
strength, and elastic modulus [3], further affecting the performance and load bearing capacity of
concrete structures, which not only influence daily production and work, but also jeopardizes the safe
operation and use of concrete structures.
Since the mid-20th century, freezing–thawing damage of concrete has been studied from the
perspective of materials. Although some problems remain unresolved, a freezing–thawing damage

Appl. Sci. 2020, 10, 1252; doi:10.3390/app10041252 www.mdpi.com/journal/applsci

Appl. Sci. 2020, 10, 1252 2 of 18

theory has been established [4–6], providing theoretical support for past and current research. Most of
the studies on concrete performance during the freeze–thaw cycle are mainly focused on the degradation
of mechanical properties of concrete. Specifically, during the concrete freezing–thawing damage process,
micro-cracks and expansion appear due to the change in the internal pore structure of concrete, resulting
in a change in mechanical properties such as the strength and deformation modulus of concrete [7].
At present, extensive experimental studies are carried out on the mechanical properties of concrete
after freezing–thawing damage, but the test methods used have been quite different. Cao [8] studied the
deterioration process of freezing–thawing concrete by uniaxial compression tests, and established the
freezing–thawing damage equation of concrete. Hasan et al. [9] conducted uniaxial compression tests of
concrete subjected to different freezing–thawing cycles, and studied the degradation law of the number
of freezing–thawing cycles on the mechanical properties of concrete. Bogas et al. [10] performed
uniaxial compression tests on concrete after freezing–thawing cycles, and clarified the relationships
among dynamic elastic modulus, compressive strength, and freezing–thawing cycles. Shang et al. [11]
completed biaxial compression tests on concrete cubes under different freezing–thawing cycles to
study the effect of freezing–thawing cycle on the mechanical properties of concrete, and established
the failure criteria and stress–strain relationship model of freezing–thawing concrete under biaxial
compression. Ma et al. [12] compared the uniaxial compression constitutive model of concrete damaged
by salt freezing with the corresponding uniaxial compression test results of salt-frozen concrete, and
determined the influence of freeze–thaw cycles on the mechanical properties of concrete. Sun et al. [13]
determined the damage parameters of concrete under the coupled conditions of salt solution and
freezing–thawing cycles by putting concrete under freezing–thawing cycles in different salt solutions.
Zhu et al. [14] established a finite element numerical model of uniaxial compression according to the
change in numbers of freeze–thaw cycles, studied the influence of different freeze–thaw cycles on the
mechanical properties of concrete, and compared it with the physical test results. The results of the
numerical simulation and physical test were consistent. Hasan et al. [15] reported the stress–strain
curve model of concrete under freezing–thawing cycles by fitting and analyzing the data obtained from
a large number of freezing–thawing tests. Liu et al. [16] found the strength changing law of concrete
through freezing–thawing tests of concrete, proposing a new freezing–thawing damage parameter
to describe the process of concrete strength change, and introduced the stress damage equation of
concrete. Duan et al. [17] added a new random damage variable for freezing–thawing degradation in
concrete through freezing–thawing tests, and discussed the influence of this variable on the mechanical
properties of freezing–thawing damage of concrete. Most of these experiments and numerical models
focused on the macro-mechanical properties of freezing–thawing concrete, whereas the meso-scale
characteristics of freezing–thawing concrete have not been studied in depth.
In this study, the effect of the number of freeze–thaw cycles on the thickness of the damaged layer
and the uniaxial compressive strength of concrete were obtained using freezing–thawing damage testing.
The particle flow discrete element method was used to establish the meso-scale mechanical model of
freeze–thaw-damaged concrete specimens for uniaxial compression simulation. Based on the numerical
model, the characteristics and modes of meso-scale failure of concrete freezing–thawing damage were
analyzed, providing theoretical support for the control and prevention of freezing–thawing damage of
concrete in cold regions.

2. Experimental Materials and Methods

2.1. Freezing–Thawing Tests of Concrete

In order to study the uniaxial compression failure morphology of concrete under different numbers
of freezing–thawing cycles, a basis for meso-parameter calibration in a numerical simulation has been
provided. The test specimen was C30 prismatic concrete, with dimensions of 100 × 100 × 300 mm.
The coarse aggregate was two graded pebbles with a roughly ellipsoid and oblate shape and density
of 2700 kg/m3 . The fine aggregate was medium sand with fineness modulus of 2.5 and density of
 Appl. Sci. 2020, 10, x 3 of 20

In order to study the uniaxial compression failure morphology of concrete under different
numbers of freezing–thawing cycles, a basis for meso-parameter calibration in a numerical simulation
has Sci.
Appl. been provided.
2020, 10, 1252 The test specimen was C30 prismatic concrete, with dimensions of 100 × 100 ×3 300 of 18
mm. The coarse aggregate was two graded pebbles with a roughly ellipsoid and oblate shape and density
of 2700 kg/m 3. The fine aggregate was medium sand with fineness modulus of 2.5 and density of 2000
3 . The cement was po42.5 ordinary Portland cement with density of 3000 kg/m3 . There were
2000 kg/m
kg/m3. The cement was po42.5 ordinary Portland cement with density of 3000 kg/m3. There were 7
7groups
groups of of prismatic
prismatic concrete
concretespecimens,
specimens,of ofwhich
which66 were
were subjected
subjected to
to freezing–thawing
freezing–thawingcycles
cyclesand
and
the
theother
othergroup
groupincluded thecontrol
included the controlspecimens.
specimens. Each
Each group
group included
included threethree concrete
concrete specimens
specimens that
that were
were used to test the thickness of the concrete damage layer under different freezing–thawing
used to test the thickness of the concrete damage layer under different freezing–thawing cycles and the cycles
and the uniaxial
uniaxial compressioncompression
strengthstrength after freezing–thawing
after freezing–thawing cycles. Thecycles. The mix proportions
mix proportions of concrete of concrete
specimens
specimens
are listed inareTable
listed1.in Table 1.

Table1.1. Mix
Table Mix of
of C30
C30 concrete.
concrete.

Specimen Water/Binder Sand 1 m3 Material Content

1 m3 Material of Concrete/kg
Content of Concrete/kg
Specimen Strength Water/Binder Ratio Sand Ratio
Strength Ratio Ratio Water Cement Sand Sand Stone
Water Cement Stone FlyFly
AshAshAdmixture
Admixture
C30 0.53 0.4 170 256 765 1150 64 9.6
C30 0.53 0.4 170 256 765 1150 64 9.6

Four days before the freezing–thawing test, the specimens were removed from the maintenance
Four days before the freezing–thawing test, the specimens were removed from the maintenance
site for appearance inspection, and then immersed in water at 18 ◦ C for four days before the
site for appearance inspection, and then immersed in water at 18 °C for four days before the freezing–
freezing–thawing test. Each freezing–thawing cycle lasted 4 h, in which freezing required 2.5 h and
thawing test. Each freezing–thawing cycle lasted 4 h, in which freezing required 2.5 h and thawing
thawing required 1.5 h. The 6 different groups of specimens were frozen and thawed 25, 50, 75, 100,
required 1.5 h. The 6 different groups of specimens were frozen and thawed 25, 50, 75, 100, 125, or
125, or 150 times, while the control group specimens were maintained in the standard climate chamber.
150 times, while the control group specimens were maintained in the standard climate chamber. The
The freezing–thawing temperature of concrete was set to –22 ◦ C. Once the freezing–thawing cycle was
freezing–thawing temperature of concrete was set to –22 °C. Once the freezing–thawing cycle was
completed, uniaxial compression failure tests were carried out on all specimens at the same time.
completed, uniaxial compression failure tests were carried out on all specimens at the same time.
2.2. Damaged Layer of Frozen-Thawed Concrete
2.2. Damaged Layer of Frozen-Thawed Concrete
Concrete specimens subjected to the various numbers of freeze–thaw cycles are shown in Figure 1.
Concrete specimens subjected to the various numbers of freeze–thaw cycles are shown in Figure 1.
As concrete is categorized as a multiphase heterogeneous material, different degrees of micro-cracks
As concrete is categorized as a multiphase heterogeneous material, different degrees of micro-cracks
appeared. As the freezing–thawing test proceeded, the number of micro-cracks on the surface of
appeared. As the freezing–thawing test proceeded, the number of micro-cracks on the surface of concrete
concrete increased after 25 freezing–thawing cycles, and the surface of the specimens began to be rough,
increased after 25 freezing–thawing cycles, and the surface of the specimens began to be rough, with
with a small amount of cement mortar falling off. After 50 freezing–thawing cycles, cracks were visible
a small amount of cement mortar falling off. After 50 freezing–thawing cycles, cracks were visible on
on the surface of the specimens, which had become obviously rough; after 75 cycles, the micro-cracks
the surface of the specimens, which had become obviously rough; after 75 cycles, the micro-cracks on
on the surface of concrete gradually expanded, part of the surface began to peel off, and some aggregate
the surface of concrete gradually expanded, part of the surface began to peel off, and some aggregate
was exposed. After 100 freezing–thawing cycles, the concrete surface seriously peeled off with some
was exposed. After 100 freezing–thawing cycles, the concrete surface seriously peeled off with some
coming off in scales or layers, and aggregate was more exposed; after 125 freezing–thawing cycles, more
coming off in scales or layers, and aggregate was more exposed; after 125 freezing–thawing cycles,
concrete peeled off in scales or layers with some corners even peeling off. After 150 freezing–thawing
more concrete peeled off in scales or layers with some corners even peeling off. After 150 freezing–
cycles, aggregate was exposed in large areas and more areas in the corner peeled off.
thawing cycles, aggregate was exposed in large areas and more areas in the corner peeled off.

Figure1.1.Concrete
Figure Concrete specimens
specimens with
with the different number of freezing–thawing
freezing–thawing cycles.
cycles.

Concrete
Concretefreezing–thawing
freezing–thawing isis aa gradual
gradual process. The outermost
process. The outermost layer
layer of
of concrete
concrete isis seriously
seriously
damaged,
damaged,butbutthe
thedeeper
deeperin
inthe
theconcrete,
concrete,the
theless
less the
the damage. Therefore, concrete
damage. Therefore, concreteisisconsidered
consideredininthree
three
parts:
parts:freezing–thawing
freezing–thawing damage
damage layer,
layer, freezing–thawing damage transition
freezing–thawing damage transition layer,
layer, and
and undamaged
undamaged
layer [18]. The freezing–thawing damage transition layer is a dense and thin area. To simplify the
analysis of such behavior, this part of the concrete is considered the undamaged layer of concrete.
The degradation of mechanical properties of freezing–thawing concrete is mainly caused by the
degradation of mechanical properties of the damaged layer of concrete [19]. A simplified sketch of a
shown in Figure 2b.

Figure 2b shows that the thickness of the damaged layer of concrete increased with increasing
number of freezing–thawing cycles, the growth rate of concrete damage layer thickness from 25 to 125
freeze–thaw cycles was basically unchanged, and the growth rate of the layer of damaged concrete
Appl. Sci. 2020, 10, 1252 4 of 18
thickness with more than 125 freeze–thaw cycles obviously accelerated. In the process of repeated
freezing–thawing of free capillary water in concrete, the damage stress caused by volume expansion
concrete
increasedsection is shown
gradually, in Figure
causing 2a. This paper
new micro-cracks to assumed
continue that the damaged
developing layer oftoconcrete
and damage was
accumulate
uniformly spread around the section.
continuously. When the damage value exceeds a certain range, the damage will increase sharply.

(a) (b)

Figure 2. (a) Damaged layer distribution of freezing–thawing concrete section and (b) the curve of
Figure 2. (a) Damaged layer distribution of freezing–thawing concrete section and (b) the curve of
relationship between the number of freezing–thawing cycles and damage thickness.
relationship between the number of freezing–thawing cycles and damage thickness.
The thicknesses of the damaged layer of the six groups of concrete specimens under the
2.3. Uniaxial Compression
different numbers Test of Concretecycles
of freezing–thawing Damaged
werebymeasured
Freezing–Thawing
using ultrasonic nondestructive testing
technology [20]. HC-U81 concrete strength defect detector
With the goal of determining the uniaxial compression failure was used modeto measure the thickness
and stress–strain curve of
of
concrete
concrete freeze–thaw
damaged bydamage layer, and the
freezing–thawing, measurement
7 groups accuracy
of specimens werewas 0.01under
tested mm. uniaxial
A concrete specimen
compression
was
on adivided into three regions, inelectro-hydraulic
microprocessor-controlled which the thicknesses
servoofpressure
the damaged
tester.layer of concrete
According to ThewasTest
detected
Code
by ultrasonic nondestructive testing, and the average value of the results was
for Hydraulic Concrete [21], the average value of peak stress of three specimens was taken as theused. Using Origin
software (OriginLab,
compressive strengthUSA), theconcrete
of the polynomial fitting ofDuring
specimens. the thickness d ofthe
the test, freezing–thawing
concrete specimen damage
was
layer with the number of freezing–thawing cycles N was conducted, and the fitting
preloaded at 2 KN, and then the lifting speed of the testing machine was set to 0.5 mm/s until the curve is shown in
Figure 2b.
strain reached 3.0 × 10 , then the test stopped. The failure modes and stress–strain curves of freezing–
–3

Figure 2b shows
thawing-damaged that the
concrete thickness
under of the
uniaxial damaged layer
compression of concrete
were obtained as increased
shown in with increasing
Figures 3 and 4.
number of freezing–thawing cycles, the growth
The peak stress–strain curves are shown in Figure 5. rate of concrete damage layer thickness from 25 to 125
freeze–thaw cycles was basically unchanged, and the growth rate of the layer of damaged concrete
thickness with more than 125 freeze–thaw cycles obviously accelerated. In the process of repeated
freezing–thawing of free capillary water in concrete, the damage stress caused by volume expansion
increased gradually, causing new micro-cracks to continue developing and damage to accumulate
continuously. When the damage value exceeds a certain range, the damage will increase sharply.

2.3. Uniaxial Compression Test of Concrete Damaged by Freezing–Thawing

With the goal of determining the uniaxial compression failure mode and stress–strain curve of
concrete damaged by freezing–thawing, 7 groups of specimens were tested under uniaxial compression
on a microprocessor-controlled electro-hydraulic servo pressure tester. According to The Test Code
for Hydraulic Concrete [21], the average value of peak stress of three specimens was taken as
the compressive strength of the concrete specimens. During the test, the concrete specimen was
preloaded at 2 KN, and then the lifting speed of the testing machine was set to 0.5 mm/s until
the strain reached 3.0 × 10–3 , then the test stopped. The failure modes and stress–strain curves of
freezing–thawing-damaged concrete under uniaxial compression were obtained as shown in Figures 3
and 4. The peak stress–strain curves are shown in Figure 5.
Figures 3 and 4 show that the severity of the uniaxial compression damage to concrete increased
with the increase in the number of freezing–thawing cycles. When the concrete specimens without
freezing–thawing were damaged, diagonal cracks mainly occurred. However, with the increase in
Appl. Sci. 2020, 10, 1252 5 of 18

freezing–thawing cycles, concrete became more brittle and the risk of bursting became less serious.
With 25 freezing–thawing cycles, the diagonal main cracks deepened and large breaks in flakes or
bulk along the direction of the main cracks occurred. The situation became more obvious with the
increase in the number of freezing–thawing cycles. So when the specimen was damaged, many
vertical main cracks usually occurred. With increasing freezing–thawing cycles, the micro-cracks
inside the concrete gradually increased [5], which further deformed the specimen transversely under
compression, resulting in the vertical cracking of concrete for its transverse tension. With increasing the
compressive load, the change law of stress–strain curve of freezing–thawing-damaged concrete under
uniaxial compression was similar to the one shown above (Figure 4), with a linear elastic stage, pre-peak
plastic stage, peak point, and post-peak decline stage. In the linear elastic stage, the elastic modulus
decreased with the increase in the number of freezing–thawing cycles. The peak stress decreased
with the increase in the number of freezing–thawing cycles while the peak strain increased with the
increase in the number of freezing–thawing cycles. This occurred because the internal micro-structure
of concrete specimens was damaged after freezing–thawing cycles. With the increase in the number of
freezing–thawing cycles, the internal damage of concrete developed continuously, leading to a change
in the concrete constitutive behavior.
Appl. Sci. 2020, 10, x 5 of 20
Appl. Sci. 2020, 10, x 5 of 20

Figure
Figure 3.
3.Failure
3. Failure modes
modes
Failure of
of concrete
modes concrete under uniaxial
underunder
of concrete uniaxial compression
compression
uniaxial with
with different
compression different numbers
numbers of
with different of freezing–
freezing–
numbers of
freezing–thawing cycles. thawing
thawingcycles.
cycles.

Figure
Figure 4.
4.Stress–strain
Stress–straincurve
Stress–strain curveof
of concrete
concrete under
under different
different numbers
numbers of
of freezing–thawing
freezing–thawing cycles.
freezing–thawing cycles.

To study the change in the axial dimensions of concrete specimens before and after uniaxial
compression under increasing number of freezing–thawing cycles, the axial height of specimens before
and after compression was measured using Vernier calipers. Each specimen was measured three
times in the axial direction, and the average value was used as the axial compression deformation of
specimens. In each group, we measured three different specimens to determine the size change of
Appl. Sci. 2020, 10, 1252 6 of 18

concrete specimens after uniaxial compression under different freezing–thawing cycles, as shown in
Table 2. Figure 4. Stress–strain curve of concrete under different numbers of freezing–thawing cycles.

(a) (b)
Figure 5. (a)5.Peak (a)
Figure stress and stress
Peak (b) peak strain
and (b)ofpeak
concrete under
strain of different
concretenumbers
under of freezing–thawing
different numbers cycles.
of
freezing–thawing cycles.
Figures 3 and 4 show that the severity of the uniaxial compression damage to concrete increased
with the increase in the number of freezing–thawing cycles. When the concrete specimens without
Table 2. Axial deformation of concrete specimens under different freeze–thaw cycles.
freezing–thawing were damaged, diagonal cracks mainly occurred. However, with the increase in
freezing–thawing cycles, concrete Cycles
Number of Freezing-Thawing became more 0 brittle
25 and the
50 risk of
75 bursting
100 became
125 less150
serious.
With 25 freezing–thawing cycles, the diagonal main cracks deepened and large breaks in flakes or
Average Value of Test Axial
bulk along theDeformation the main cracks 0.897
direction of(mm) 0.894
occurred. 0.894
The situation 0.891
became 0.890
more 0.890
obvious0.889
with the
increase in the number of freezing–thawing cycles. So when the specimen was damaged, many
vertical main Model
3. Numerical cracks of
usually occurred.
Concrete Damaged Withbyincreasing freezing–thawing cycles, the micro-cracks
Freezing–Thawing
inside the concrete gradually increased [5], which further deformed the specimen transversely under
compression, resulting in the vertical cracking of concrete for its transverse tension. With increasing
3.1. Model Setup
the compressive load, the change law of stress–strain curve of freezing–thawing-damaged concrete
The basic principle of the discrete element method [22–24] is to treat the whole medium as a series
of discrete and independent moving elements that have certain geometric (shape, size, arrangement,
etc.), physical, and chemical characteristics. These elements are microscopic and only interact with
adjacent elements, whose motion is controlled by the classical equations of motion. The deformation
and evolution of the whole medium are described by the motion and mutual position of each element.
Particle Flow Code (PFC) [25] is software based on the discrete element code, which allows discrete
particles to undergo displacement and rotation. With the automatic identification of new contacts in
the calculation process, particles can be combined to simulate deformed polyhedral particles [26,27].
In PFC software, the linear parallel bond model is usually used to simulate the contact between
aggregate particles. The linear parallel bond model is regarded as many parallel springs with certain
stiffness distributed on the contact surface of particles, which can transfer the force and moment
between particles. When the linear parallel bond model is adopted, the total contact force Fi . and total
contact moment Mi between particles need to be calculated, as follows:
n s
Fi = Fi + Fi
n s (1)
Mi = Mi + Mi

n n
where Fi is the normal contact force between particles, Mi is the normal contact moment between
s s
particles. Fi is the tangential contact force between particles, and Mi is the tangential contact moment
between particles. The principle of the linear parallel bond model between particles is shown in
Figure 6.
Generation of concrete numerical model specimens involves first generating the boundary wall in
the calculation range, then generating the small spherical particles randomly in the boundary wall
Appl. Sci. 2020, 10, 1252 7 of 18
Appl. Sci. 2020, 10, x 7 of 20

range to
pebble simulate
coarse the cement
aggregate mortarshapes,
of different to fill thein specimen,
which the and usingparticles
spherical clump technology
and the coarse to generate
aggregate the
Appl. Sci. 2020, 10, x 7 of 20
pebble coarse aggregate of different shapes, in which the spherical
are subject to a Gaussian distribution. A test piece of pebble concrete with good contact and particles and the coarse aggregate are
subject to a Gaussian
compactness wasaggregatedistribution.
generated; the A test piece
coarse of pebble concrete
aggregate withincluded
good contact and compactness
pebble coarse of different shapes, in whichof thethe pebble
spherical particles and two kinds
the coarse of coarse
aggregate
was generated;
aggregates the
withtooblatecoarse aggregate of
sphere distribution. the pebble
and ellipsoidAshapes, included
as shown two kinds
in Figure of coarse
7a,b. aggregates
Thegood model with
size ofoblate
the
are subject a Gaussian test piece of pebble concrete with contact and
compactness was generated; the coarse aggregate of the pebble included two kinds of coarse ×
sphere
specimensand ellipsoid
was 100 × shapes,
100 × 300 as shown
mm. in Figure
Considering 7a,b.
the The
influencemodel of size
the of
mix the specimens
ratio of concrete was 100
(Table × 100
1) and
300 size
the mm.ofConsidering
aggregates the with
cement the
oblate influence
mortar
sphere andof
with the the mixshapes,
running
ellipsoid ratio of
speedasconcrete in(Table
of computer,
shown Figure 1)7a,b.
we and the model
assumed
The size ofsize
that the cement
the cement
of the
mortar
mortar with
specimens the
was composed running
was 100 ×of speed
100particlesof
× 300 mm.computer,
with we
radii of 2–5
Considering assumed
the mm, that
andof
influence the
the cement
thedensity
mix ratio mortar
was was
set to 2000
of concrete composed
(Tablekg/m of
3; the
1) and
particles with 3
size of the
the size ofradii
coarse of 2–5 mm,
theaggregate
cement was
mortarandthe the
with samedensity
the as inwas
running set
thespeed to 2000
physical kg/m
test,
of computer, ; the
which size
wewas of thethat
composed
assumed coarseof aggregate
the 5–20
cement mm
was mortar
clump theparticles,
same wasascomposed
in
andthethephysical
density test,
of particles
waswhich
with was
radii
set to 2700composed
of 2–5
kg/m mm, ofand
3. The 5–20themm
cement clump
density
mortarwasparticles,
set to 2000
particles and the density
kg/m
accounted 3; the
for
was set to 2700 kg/m 3 . The cement mortar particles accounted for 51% of the total volume of particles
size of the coarse aggregate was the same as in the physical test, which
51% of the total volume of particles and the coarse aggregate accounted for 49%. The total number of was composed of 5–20 mm
and clump
particles wasparticles,
the coarse 6384. andmodel
aggregate
The the density
accounted was
was built set tothe
for 49%.
with 2700
The kg/m
Ztotal 3. The cement mortar particles accounted for
axisnumber
as of particles
the vertical was and
direction 6384.theThe model
X and was
Y axes
built51%
withof the
the total
Z axisvolume
as the of particles
vertical and the
direction coarse
and aggregate
the X and accounted
Y axes
as the horizontal directions, as shown in Figure 7c. According to the thickness of freezing–thawing as for
the 49%. The
horizontal total number
directions, ofas
shownparticles
in was
Figure 6384.
7c. The
Accordingmodel was
to the built with
thickness theofZ axis as the
freezing–thawing vertical
damage layers under different numbers of freezing–thawing cycles obtained in the previous test, the direction
damage and
layers the X and
under Y axes
different
as theof
numbers horizontal directions,cycles
freezing–thawing as shown in Figure the7c. According tothe
thedamaged
thicknessand of freezing–thawing
damaged and undamaged areas of obtained
concrete in were previous
established test,
using PFC3D software, undamaged
as shown areas
in
of damage
concrete layers
were under
established different
using numbers
PFC3D of freezing–thawing
software, as shown cycles
in Figureobtained
7d,e, inin the
which previous
the greentest,area
the is
Figure 7d,e, in which the green area is the damaged area of frozen-thawed concrete, and the blue area
damaged and undamaged areas of concrete were established using PFC3D software, as shown in
the
is damaged
the undamaged area concrete.
of frozen-thawed concrete, and the blue area is the undamaged concrete.
Figure 7d,e, in which the green area is the damaged area of frozen-thawed concrete, and the blue area
is the undamaged concrete.

Figure
Figure 6.
6. Schematic
Schematicdiagram
diagram of
of linear
linear parallel
parallel bond
bond model
model between
between particles.
particles.
Figure 6. Schematic diagram of linear parallel bond model between particles.

(a) (a) (b)

(b)

(c) (d) (e)

(c) (d) (e)
Figure 7. Model
Model regional
regional groupingdiagram
grouping diagramand
andnumerical
numerical model
model ofof freezing–thawing cycle. (a)(a)
Figure
Figure 7. Model
7. regional grouping diagram and numerical model of freezing–thawing
freezing–thawing cycle.
cycle. (a)
Oblate sphere
Oblate sphere coarse
sphere coarse aggregate;
coarse aggregate;
aggregate; (b)(b) Ellipsoid
(b) Ellipsoid coarse
Ellipsoidcoarse aggregate;
coarseaggregate; (c)
aggregate;(c) Freezing–thawing
(c)Freezing–thawing concrete
Freezing–thawingconcrete model;
concretemodel;
model;
Oblate
(d) Regional grouping diagram
(d) Regional grouping diagram ofofspecimens
specimens model;
model; €€Grouping
Groupingdiagram
diagram of contact attribute area of
of contact attribute area
(d) Regional
specimen.grouping diagram of specimens model; € Grouping diagram of contact attribute area of
of specimen.
specimen.
Appl. Sci. 2020, 10, 1252 8 of 18

3.2. Calibration of Meso-Parameters

When the particle discrete element method was used for this concrete research, the biggest
problem was the calibration of meso-parameters of the materials. Because the meso-parameters are
needed for calculation in PFC, they cannot be obtained directly through experiments. Therefore, it was
necessary to determine the meso-parameters suitable for macro-physical parameters through multiple
numerical simulation tests after continuous selection and trial calculation to establish the relationship
between them. Based on the various sensitivities of different meso-parameters to the macro-mechanical
properties of concrete [28,29], in this study, the meso-parameters of the damaged area were calibrated
by an inversion trial-and-error method according to the thickness of the freezing–thawing damage
layer and stress–strain curve under uniaxial compression. The PFC numerical model generally
only considers the meso-parameters between cement and mortar [30,31], but does not consider the
meso-parameters of the weak interface between cement mortar and coarse aggregate. In this study, the
meso-parameters were divided into two parts: the meso-parameters between the cement and mortar
and the meso-parameters of the weak interface between cement mortar and coarse aggregate, as shown
in Tables 3 and 4.

Table 3. Meso-parameters of cement mortar bonding of the damaged layer.

Number of
Freezing–Thawing emod/Gpa kratio fric pb_emod/Gpa pb_kratio pb_nstrength/Mpa pb_strength/Mpa pb_radius
Cycles
0 40 0.1 0.6 40 0.42 68 68 0.5
25 32 0.1 0.6 32 0.42 64 64 0.5
50 30 0.1 0.6 30 0.42 62.25 62.25 0.5
75 28 0.1 0.6 28 0.42 61.6 61.6 0.5
100 26 0.1 0.6 26 0.42 63 63 0.5
125 24.5 0.1 0.6 24.5 0.42 61 61 0.5
150 26 0.1 0.6 26 0.42 62 62 0.5

Table 4. Meso-parameters of cement aggregate bonding of the damaged layer.

Number of
pb_nstrength pb_strength
Freezing–Thawing emod/Gpa kratio fric pb_emod/Gpa pb_kratio pb_radius
/Mpa /Mpa
Cycles
0 40 0.1 0.6 40 0.42 34 34 0.5
25 32 0.1 0.6 32 0.42 32 32 0.5
50 30 0.1 0.6 30 0.42 31.25 31.25 0.5
75 28 0.1 0.6 28 0.42 30.8 30.8 0.5
100 26 0.1 0.6 26 0.42 31.5 31.5 0.5
125 24.5 0.1 0.6 24.5 0.42 30.5 30.5 0.5
150 26 0.1 0.6 26 0.42 31 31 0.5

Note: emod represents the effective modulus of the linear part in the meso-model of concrete, kratio represents
the normal to tangential stiffness ratio of the linear part, fric represents the friction coefficient between particles,
pb_emod represents the effective modulus of the parallel bonding part, pb_kratio represents the normal to tangential
stiffness ratio of the parallel bonding part, pb_nstrength represents the normal strength of the parallel bonding part,
pb_strength represents the tangential strength of the parallel bonding part, and pb_radius represents the parallel
bond radius.

3.3. Model Validation

With the discrete element software PFC3D, the uniaxial compression of freezing–thawing concrete
was simulated using the meso-parameters of freezing–thawing concrete in Tables 3 and 4, keeping the
upper wall still, and applying an upward velocity of 0.5 mm/s to the lower wall to simulate the upward
lifting of the testing machine to obtain the peak stress and peak strain of the concrete specimen model
under different freezing–thawing cycles, as shown in Figure 5. The stress–strain curve of concrete
obtained by numerical simulation is shown in Figure 8.
 With the discrete element software PFC3D, the uniaxial compression of freezing–thawing concrete
was simulated using the meso-parameters of freezing–thawing concrete in Tables 3 and 4, keeping the
upper wall still, and applying an upward velocity of 0.5 mm/s to the lower wall to simulate the upward
lifting of the testing machine to obtain the peak stress and peak strain of the concrete specimen model
under
Appl. Sci.different freezing–thawing cycles, as shown in Figure 5. The stress–strain curve of concrete
2020, 10, 1252 9 of 18
obtained by numerical simulation is shown in Figure 8.

Figure 8.
Figure 8. Stress–strain
Stress–straincurve of of
curve concrete under
concrete freezing–thawing
under cycles
freezing–thawing by Particle
cycles Flow Code
by Particle Flow (PFC)
Code
(PFC) stimulation. stimulation.

Figure
Figures 4, Figure
4, 5, and 85,showandthat
Figure 8 show that the damage
the freezing–thawing freezing–thawing damage
curve of concrete curve ofbyconcrete
simulated PFC3D
simulated by PFC3D was basically consistent with the law obtained by experiment
was basically consistent with the law obtained by experiment in the rising section and stress–strainin the rising section
and
peakstress–strain peak value.
value. However, in theHowever,
post-peakindecline
the post-peak decline
stage after thestage
peakafter thewas
stress peakreached,
stress was thereached,
bonds
the bonds between particles compressed and fractured, the restriction and binding
between particles compressed and fractured, the restriction and binding force of particles decreased, force of particles
decreased, the shearof
the shear resistance resistance
materialsofdecreased,
materials and
decreased, and the deformation
the deformation of the model ofchanged
the model to changed
the state to
of
the state
shear of shearBecause
expansion. expansion. Becauseshape
the particle the particle shape in the
in the numerical numerical
model model
is single is single
and the and is
bite force the bite
weak,
force is weak, softening
the post-peak the post-peakof thesoftening
numericalof the
modelnumerical modelWhen
is enhanced. is enhanced.
there wereWhenno there
or fewwere no or
freezing–
few freezing–thawing
thawing cycles, the curvecycles, the curve
simulated bysimulated by PFCfaster;
PFC decreased decreased faster; the higher
the higher number theofnumber of
freezing–
freezing–thawing
thawing cycles, the cycles, the slower
slower the curve
the curve decreased.
decreased. When When
the the stress
stress exceeded
exceeded thethepeak
peakstress,
stress, the
softening effect of concrete slowed with increasing numbers of freezing–thawing cycles. In general,
the stress–strain curves of concrete obtained by PFC3D simulation were in good agreement with the
experimental data, which shows that the model established in this paper can be used to simulate the
mechanical properties of concrete
concrete after
after freezing–thawing
freezing–thawing damagedamage under
under uniaxial
uniaxial compression.
compression.
The axial displacement
displacement of concrete particles can
concrete model particles can be used to reflect
reflect the
the compression
compression
deformation
deformation of of model specimens. Therefore, the accuracy
specimens. Therefore, the accuracy of of the concrete numerical
numerical model and the
meso-parameters was verified by comparing the axial displacement of the concrete numerical model at
the strain of 3 × –3 with the axial compression deformation of Table 5 shows
× 10–3 with the axial compression deformation of the concrete specimen. Table
the comparison between
between the the test
testaxial
axialcompression
compressiondeformation
deformationand andthetheparticle
particleaxial
axialdisplacement.
displacement.

Table
Table 5. Comparison of
5. Comparison uniaxial compression
of uniaxial compression deformation
deformation and particle axial
and particle axial displacement
displacement of
of concrete
concrete
under different numbers of freezing–thawing cycles.
under different numbers of freezing–thawing cycles.

Number of Average Value of DifferenceDifference Value Relative

ErrorError of
Number of Average Value of Axial Axial Value between Relative of Numerical
Freezing– Test Axial between Numerical Numerical Model
Freezing–ThawingTest Axial Displacement of
Displacement Numerical
of Model and Test Model and Test in Axial
Thawing Cycles Deformation Particles (mm) Model and Test(mm)
in Axial and Test in Axial
Cycles Deformation (mm) in Axial
Particles (mm) Deformation Deformation (%)
(mm) Deformation (mm) Deformation (%)
0 0 0.897 0.897 0.899 0.899 0.0020.002 0.223
0.223
25 0.894 0.895 0.001 0.112
25 50 0.894 0.894 0.895 0.898 0.0010.004 0.112
0.447
75 0.891 0.894 0.003 0.337
50 100 0.894 0.890 0.898 0.896 0.0040.006 0.447
0.674
125 0.890 0.897 0.007 0.787
150 0.889 0.890 0.001 0.112

Table 5 shows that with the increase in freezing–thawing cycles, the axial deformation obtained by
the test gradually reduced, and the axial displacement of particles was not much different. This occurred
because with the increase in freezing–thawing cycles, the damage degree of concrete increased, and
the height of concrete specimen decreased gradually. When reaching the same strain of 3 × 10–3 , the
 150 0.889 0.890 0.001 0.112

Table 5 shows that with the increase in freezing–thawing cycles, the axial deformation obtained
by the test gradually reduced, and the axial displacement of particles was not much different. This
occurred
Appl. because
Sci. 2020, 10, 1252 with the increase in freezing–thawing cycles, the damage degree of concrete 10 of 18
increased, and the height of concrete specimen decreased gradually. When reaching the same strain
of 3 × 10–3, the compression deformation decreased, while the height of the numerical simulation
compression
specimen diddeformation decreased,
not change before while
uniaxial the height ofsothe
compression, thenumerical simulation
axial displacement ofspecimen did not
particles differed
change before uniaxial compression, so the axial displacement of particles differed
little. The axial deformation obtained from the test was very close to the axial displacement little. The axialof
deformation
particles obtainedobtainedfromfrom the test was
the numerical very close
simulation, andtothe
therelative
axial displacement
error was notof particles
more obtained
than 1%, which
from
showed the that
numerical simulation,
the concrete and the
numerical relative
model anderror
the was not more
calibrated than 1%, whichwere
meso-parameters showed that The
correct. the
concrete numerical model and the calibrated meso-parameters were correct. The axial
axial displacements of particles in the 75 and 150 cycles of freezing–thawing concrete numerical displacements of
particles
model are in shown
the 75 and 150 cycles
in Figure 9. of freezing–thawing concrete numerical model are shown in Figure 9.

(a) (b)

Figure 9. Axial displacement of concrete particles under (a) 75 and (b) 150 freezing–thawing cycles.
Figure 9. Axial displacement of concrete particles under (a) 75 and (b) 150 freezing–thawing cycles.
Figure 9 shows that the axial displacement of particles after 75 cycles of freezing and thawing
Figure 9 shows that the axial displacement of particles after 75 cycles of freezing and thawing
was 0.894 mm, and that of particles after 150 cycles of freezing and thawing was 0.890 mm. In the
was 0.894 mm, and that of particles after 150 cycles of freezing and thawing was 0.890 mm. In the
numerical simulation, the bottom wall was lifted upward to simulate the upward lifting process of
numerical simulation, the bottom wall was lifted upward to simulate the upward lifting process of
pressure testing machine in the macro physical test, so the displacement of the upper particles of
pressure testing machine in the macro physical test, so the displacement of the upper particles of
concrete was 0, and the axial displacement was the displacement of the lower particles of concrete.
concrete was 0, and the axial displacement was the displacement of the lower particles of concrete.
4. Numerical Simulation of Meso-Scale Failure of Freezing–Thawing Damage Concrete under
4. Numerical
Uniaxial Simulation of Meso-Scale Failure of Freezing–Thawing Damage Concrete under
Compression
Uniaxial Compression
Figures 4 and 8 show that the uniaxial compression process of concrete could be divided into
three Figures 4 andthe
stage points: 8 show that point
first stage the uniaxial
was when compression process
the stress of of concrete
the specimens in thecould berising
curve divided into
section
three stage
reached points:the
0.5 times thepeak
firststress,
stage point was when
the second stagethe stress
point wasofwhen
the specimens
the stress ofin the concrete
curve rising section
specimens
reachedthe
reached 0.5peak
times the peak
stress, and the stress,
third the
stagesecond stagewhen
point was pointthewas when
strain of thethe stress ofreached
specimens the concrete
3.0 ×
10 –3 . Because
specimens reached the peak
the concrete stress,damaged
became and the after
thirdthe
stage pointstage
second was point
whenin thethestrain of the
uniaxial specimens
compression
reached the
process, 3.0 second
× 10–3. Because
and third the concrete
stage pointsbecame damaged
were selected forafter the second
analysis stage point
and research in thisinpaper.
the uniaxial
Based
compression
on the PFC3Dprocess,
numericalthemodel,
secondthe and third of
effects stage points were selected
freezing–thawing damage foron
analysis and research
the mechanical in this
properties
paper.
of Based
concrete were onstudied
the PFC3D
from the numerical
aspect ofmodel,
contactthe effects
force of freezing–thawing
between damage on the
particles and the development of
mechanical
cracks. The properties
cracking of ofspecimens
concrete were andstudied from the failure
the meso-scale aspect of contact
mode force between particles
of freezing–thawing and
concrete
were analyzed.

4.1. Analysis of Contact Force between Concrete Particles

In the PFC model, the contact force between particles is the force produced by extrusion and shear
between particles. The contact force between particles can quantitatively reflect the damage degree
of concrete under uniaxial compression. Therefore, we studied the maximum value of the contact
force between particles in the two different stage points of freezing–thawing-damaged concrete, and
analyzed the change in contact force between particles during uniaxial compression of the concrete
model under different numbers of freezing–thawing cycles. The maximum contact force at the two
different stage points is shown in Figure 10.
shear between particles. The contact force between particles can quantitatively reflect the damage
degree of concrete under uniaxial compression. Therefore, we studied the maximum value of the
contact force between particles in the two different stage points of freezing–thawing-damaged
concrete, and analyzed the change in contact force between particles during uniaxial compression of
the concrete
Appl. model
Sci. 2020, 10, 1252 under different numbers of freezing–thawing cycles. The maximum contact11force
of 18
at the two different stage points is shown in Figure 10.

Figure 10. Maximum contact force of freezing–thawing-damage concrete.

Figure 10 shows that with increasing freezing–thawing cycles, the maximum contact force between
particles in the second
second stage
stage point
point gradually
gradually decreased,
decreased, and the maximum
maximum contact force between between
particles in the third stage point gradually increased; the maximum contact force at the second stage
point was significantly higher than the maximum contact force between particles in the third stage
point because with the increase in freezing–thawing cycles, at the second stage point, the bite force
between particles gradually decreased, and the peak stress gradually decreased, so the maximum maximum
contact force
forcebetween
betweenparticles
particlesgradually
graduallydecreased.
decreased. With
Withthethe
increase
increasein freezing–thawing
in freezing–thawing cycles, the
cycles,
residual
the residualstress of the
stress model
of the gradually
model graduallyincreased, so the
increased, maximum
so the maximum contact
contactforce between
force betweenparticles in
particles
the third
in the stage
third point
stage gradually
point increased,
gradually but but
increased, the residual stress
the residual of the
stress ofmodel was obviously
the model less than
was obviously less
the
thanpeak
the stress, so the so
peak stress, maximum contactcontact
the maximum force inforce
the second stage point
in the second stagewas
pointobviously higher than
was obviously the
higher
maximum contact force
than the maximum in the
contact third
force stage
in the point.
third stageContact force in force
point. Contact the second
in the stage
secondpoint
stagewith
point75with
and
150 cycles
75 and 150ofcycles
freezing–thawing is shownisinshown
of freezing–thawing Figurein11; the third
Figure stage
11; the point
third contact
stage force
point is shown
contact force in
is
Figure
shown 12.
Appl. Sci.in Figure
2020, 10, x 12. 13 of 20

(a) (b)
Figure 11. Contact force distribution of concrete during the second stage point (unit: N) with (a) 75
Figure 11. Contact force distribution of concrete during the second stage point (unit: N) with (a) 75
and (b) 150 cycles of freeze–thaw.
and (b) 150 cycles of freeze–thaw.
 (a) (b)

Appl. Sci.
Figure 10,Contact
2020,11. 1252 force distribution of concrete during the second stage point (unit: N) with (a) 75
12 of 18
and (b) 150 cycles of freeze–thaw.

(a) (b)
Figure 12. Contact force distribution
distribution of
of concrete
concrete at the third stage point (unit: N)
N) with
with (a)
(a) 75
75 and
and (b)
(b)
150 cycles of freeze–thaw. 150 cycles of freeze–thaw.

Figures 11 and 12 show that the the maximum

maximum contact
contact force between
between particles with 75 freeze–thaw
cycles was 193.24 KN, and that between particles particles with 150 cycles was 185.12 KN at the second stage
point. The
The maximum
maximum contact
contact forces
forces between
between particles
particles with
with 75
75 and 150 cycles were 142.35 and 154.42
KN at the third stage point, respectively. In the second stage point, the maximum contact force between
particles
particles with 150 cycles of freezing–thawing
freezing–thawing was was 4%
4% lower
lower than that of 75 cycles, which is basically
consistent with the peak stress reduction rate of concrete under under the
the same
same freezing–thawing
freezing–thawing cycles.
cycles. In
the third stage point, the maximummaximum contact force between particles with 150 cycles of freeze–thaw
freeze–thaw
increasedby
increased by7.8%
7.8%compared
compared with
with 75 cycles,
75 cycles, whichwhich is basically
is basically consistent
consistent with
with the the residual
residual stress
stress growth
growth
rate rateconcrete
of the of the concrete modelthe
model under under
samethefreezing–thawing
same freezing–thawing cycles under
cycles under uniaxial
uniaxial compression
compression strain
strain –3
× 10 shows
of 3.0 ×of103.0. This
–3 . Thisthat
shows
the that the modeled
modeled contact
contact force of force of freezing–thawing
freezing–thawing concreteconcrete
particlesparticles is
is correct.
correct.
By By controlling
controlling the termination
the termination conditioncondition of uniaxial
of uniaxial compression
compression of concrete
of concrete numerical
numerical model,model,
the
the contact
contact forceforce between
between particles
particles at any
at any point
point onon thestress–strain
the stress–straincurve
curveisisobtained
obtainedto
to estimate
estimate the
damage degree of the concrete structure structure atat different
different points.
points.

4.2. Crack
4.2. Crack Development
Development Law
Law of
of Concrete
Concrete Model
Model
In the concrete model, the contact between particles is a linear parallel bond (Pb). In the Pb model,
when the stress value of the Pb exceeds the value of the Pb strength, the Pb frame breaks. At this time,
a micro-crack occurs; when the shear bond strength value is exceeded, the generated crack is a shear
crack; when the normal bond strength value is exceeded, the generated crack is a tension crack. For
the Pb model, once the bond between particles breaks, the internal stress of concrete is redistributed,
and the bond between adjacent particles also breaks. Then, through the connection and expansion of
micro-cracks, the concrete material shows macro damage. Therefore, we studied the number of cracks
after uniaxial compression at two different stage points of freezing–thawing-damaged concrete, and
analyzed the micro-crack state of two different stage points of the freezing–thawing-damaged concrete
model. The number of cracks in the two different stage points of concrete uniaxial compression is
shown in Figure 13.
softening stage. At this time, the concrete lost bearing capacity completely, so the number of micro-
cracks increased significantly. Due to the large shear force between particles in the process of uniaxial
compression, the number of shear cracks at the same stage point was significantly higher than the
number of tension cracks. The number of concrete cracks in the second stage point of 75 and 150
freezing–thawing
Appl. Sci. 2020, 10, 1252cycles is shown in Figure 14; the number of cracks in the third stage point is shown
13 of 18
in Figure 15.

Figure
Figure 13.
13. The
The number
number of
of cracks
cracks in
in two
two stage
stage points
points of
of freezing–thawing-damaged
freezing–thawing-damaged concrete.
concrete.

Figure 13 shows that with the increase in freezing–thawing cycles, the number of cracks in
the concrete model gradually increased. The number of cracks in the second stage point was
significantly lower than in the third stage point, and the number of shear cracks in the same stage
point was significantly higher the number of tension cracks. This occurred because with the increase
in freezing–thawing cycles, the internal damage of concrete was more serious and the number of
micro-cracks increased. The second stage point was the peak stress point of concrete. At this time,
the concrete specimen began to lose load capacity. At the third stage point, the concrete was in the
post-peak softening stage. At this time, the concrete lost bearing capacity completely, so the number of
micro-cracks increased significantly. Due to the large shear force between particles in the process of
uniaxial compression, the number of shear cracks at the same stage point was significantly higher than
the number of tension cracks. The number of concrete cracks in the second stage point of 75 and 150
freezing–thawing cycles is shown in Figure 14; the number of cracks in the third stage point is shown
in Figure 15.
Appl. Sci. 2020, 10, x 15 of 20

(a) (b)
Figure 14.14.
Figure Number ofof
Number concrete
concretecracks
cracksatatthe
the second pointwith
second stage point with(a)
(a)7575and
and(b)(b) 150
150 freeze–thaw
freeze–thaw cycles.
cycles.
Appl. Sci. 2020, 10, 1252
(a) (b) 14 of 18
Figure 14. Number of concrete cracks at the second stage point with (a) 75 and (b) 150 freeze–thaw cycles.

(a) (b)
Figure
Figure15.
15.Number
Numberofofconcrete
concretecracks
cracksat
atthe
the third
third stage point with
with (a)
(a)75
75and
and(b)
(b)150
150freeze–thaw
freeze–thawcycles.
cycles.

Figures 14 and 15 show that at the second second stage

stage point,
point, the number
number of cracks
cracks between
between particles
with 75 freezing–thawing cycles was 7941, and the number of cracks between particles with 150
freezing–thawing cycleswas
freezing–thawing cycles was8488.
8488.InInthe
thethird
thirdstage
stagepoint,
point,thethe number
number ofof cracks
cracks between
between particles
particles 75
75 freezing–thawing
freezing–thawing cycles
cycles was 21,120,
was 21,120, and the and the number
number of cracks of between
cracks between
particles particles
with 150 with 150
freezing–
freezing–thawing
thawing cycles wascycles
23,710.was 23,710.
In the secondIn stage
the second
point, stage point,
the crack the crack
growth rate forgrowth ratewas
75 cycles for 75
6.8%cycles
that
was
of 1506.8% that of 150 freezing–thawing
freezing–thawing cycles;
cycles; in the third inpoint,
stage the third stage growth
the crack point, theratecrack
for 75growth
cycles rate
was for 75
12.2%
cycles
that ofwas
150 12.2%
cycles that of 150 cycles of freezing–thawing.
of freezing–thawing. Therefore, the crack Therefore,
growth the
ratecrack
in thegrowth rate inpoint
third stage the third
was
stage point was
significantly significantly
higher than in thehigher
secondthan
stageinpoint,
the second
and thestage
crackspoint, and the
generated cracks
in the generated
uniaxial in the
compression
uniaxial compression concrete
of freezing-thawing of freezing-thawing concrete
mainly occurred inmainly occurreddecline
the post-peak in the post-peak
stage. Bydecline stage. the
controlling By
controlling
termination the termination
condition condition
of uniaxial of uniaxial
compression compression
of the of the concrete
concrete numerical model,numerical
the number model, the
of cracks
number of cracks
at any point on theatstress–strain
any point on the stress–strain
curve is obtained tocurve
predictis the
obtained
damage to degree
predictof the damage
concrete degree of
structure at
concrete
different structure
points. at different points.

4.3. Cracking Law

4.3. Cracking Law of
of Concrete
Concrete Model
Model Specimens
Specimens
When
When the
the micro-cracks
micro-cracks in in the
the concrete
concrete model
model connected
connected and
and expanded,
expanded, the
the concrete
concrete model
model
showed
showed aa macro failure state,
macro failure state, which
which isis represented
represented by
by the
the number
number of of fragments
fragments in
in PFC3D
PFC3D software.
software.
The
The fragment
fragment was
was generated
generated by
by the penetration of
the penetration many micro-cracks
of many micro-cracks around
around the
the particles. The larger
particles. The larger
the
the number
numberofoffragments,
fragments,thethe
more micro-cracks
more around
micro-cracks the particles.
around When When
the particles. the concrete stress reached
the concrete stress
the peak stress, the crack of the specimen was just beginning, the crack was in the initiation
reached the peak stress, the crack of the specimen was just beginning, the crack was in the initiationstage, and
the number of fragments was 0. Therefore, we only displayed the crack situation at the third stage
point, as shown in Figure 16.
Figure 16 shows that the growth rate of the fragment in the third stage point was basically
the same as that of the damaged layer because with the increase in freezing–thawing cycles, the
thickness of the damaged layer gradually increased. The more serious the internal damage, the more
the fragments are produced by uniaxial compression, so the growth rate was basically the same.
Simultaneously, with the increase in freezing–thawing cycles, the number of concrete fragments under
uniaxial compression gradually increased and the damage worsened. This occurred because with
the increase in freezing–thawing cycles, the internal damage of the concrete continued to develop,
and the number of micro-cracks produced by uniaxial compression gradually increased, which led to
the connection and expansion of micro-cracks and the formation of more fragments, increasing the
seriousness of damage, which is basically consistent with the compression failure pattern obtained
from the test. The particle fragment at the third stage point is shown in Figure 17.
Appl. Sci. 2020, 10, x 16 of 20

stage,Sci.and
Appl. 2020,the
10, number
1252of fragments was 0. Therefore, we only displayed the crack situation 15
at ofthe
18
third stage point, as shown in Figure 16.

Figure 16. Number of fragments at the third stage point of freezing–thawing concrete.

Figure 16 shows that the growth rate of the fragment in the third stage point was basically the same
as that of the damaged layer because with the increase in freezing–thawing cycles, the thickness of the
damaged layer gradually increased. The more serious the internal damage, the more the fragments are
produced by uniaxial compression, so the growth rate was basically the same. Simultaneously, with the
increase in freezing–thawing cycles, the number of concrete fragments under uniaxial compression
gradually increased and the damage worsened. This occurred because with the increase in freezing–
thawing cycles, the internal damage of the concrete continued to develop, and the number of micro-
cracks produced by uniaxial compression gradually increased, which led to the connection and
expansion of micro-cracks and the formation of more fragments, increasing the seriousness of
damage, which is basically consistent with the compression failure pattern obtained from the test.
Figure 16. Number
Number of
of fragments
fragments at the third stage point of freezing–thawing concrete.
The particle fragment at the third stage point is shown in Figure 17.
Figure
Figure16 17 shows
17shows that
showsthat the
thatthe maximum
thegrowth
maximum rate number
of of
offragments
the fragment
number in the damaged
fragments third stageby
damaged concrete
point
by particles
was basically
concrete that
particlesthe did
same
that did
not
not undergo freeze–thaw was 188; the maximum numbers of particle fragments damaged by 25, 75,
as undergo
that of the freeze–thaw
damaged was
layer 188; the
because maximum
with the numbers
increase in of particle fragments
freezing–thawing damaged
cycles, the by 25,
thickness 50,
of the
50,
100, 125, 125,
damaged
75, 100, and
layer150 cycles
gradually
and ofincreased.
150 cycles freezing–thawing
The morewere
of freezing–thawing 194,
serious
were 212,
the 264,
212,292,
internal
194, 264,324,
292,and
damage, the 480,
andrespectively.
324,more the
480,fragments With
are
respectively.
the increase
produced
With the by in freeze–thaw
uniaxial
increase cycles, the
in compression,
freeze–thaw so failure
the
cycles, the mode
growth rateofwas
failure concrete
mode wasthe
basically
of concreteincreasingly
same. severe. The
was Simultaneously,
increasingly number
with
severe. the
The
of concrete
increase in fragments and
freezing–thawing the final failure
cycles, the morphology
number of were
concrete used to
fragments predict
under the
number of concrete fragments and the final failure morphology were used to predict the failure mode failure
uniaxial mode and
compressionthe
damage
and the degree
gradually ofdegree
increased
damage concrete
and of indamage
the macroscopic
theconcrete inworsened. physical test.
This occurred
the macroscopic physical because
test. with the increase in freezing–
thawing cycles, the internal damage of the concrete continued to develop, and the number of micro-
cracks produced by uniaxial compression gradually increased, which led to the connection and
expansion of micro-cracks and the formation of more fragments, increasing the seriousness of
damage, which is basically consistent with the compression failure pattern obtained from the test.
The particle fragment at the third stage point is shown in Figure 17.

Figure 17 shows that the maximum number of fragments damaged by concrete particles that did
not undergo freeze–thaw was 188; the maximum numbers of particle fragments damaged by 25, 50,
75, 100, 125, and 150 cycles of freezing–thawing were 194, 212, 264, 292, 324, and 480, respectively.
With the increase in freeze–thaw cycles, the failure mode of concrete was increasingly severe. The
number of concrete fragments and the final failure morphology were used to predict the failure mode
and the damage degree of concrete in the macroscopic physical test.

(a) (b) (c)

Figure 17. Cont.

(a) (b) (c)

Appl. Sci. 2020, 10, 1252 16 of 18
Appl. Sci. 2020, 10, x 17 of 20

(d) (e)

(f) (g)

Figure 17.
Figure 17. Failure
Failurestates
statesofofconcrete
concreteatatthe
thethird
third stage
stage point
point under
under different
different freezing–thawing
freezing–thawing cycles.
cycles. (a)
(a) Failure
Failure statestate concrete
concrete without
without freeze–thaw;
freeze–thaw; (b) Failure
(b) Failure state ofstate of 25 freezing–thawing
25 freezing–thawing concrete;
concrete; (c)
(c) Failure
Failure
state state
of 50 of 50 freezing–thawing
freezing–thawing concrete; (d) Failure(d)
concrete; stateFailure state of 75 freezing–thawing
of 75 freezing–thawing concrete; (e) concrete; (e)
Failure state
Failure
of state of 100 freezing–thawing
100 freezing–thawing concrete;
concrete; (f) Failure state(f)ofFailure state of 125 freezing–thawing
125 freezing–thawing concrete;
concrete; (g) Failure state (g)
of
Failure
150 state of 150 freezing–thawing
freezing–thawing concrete. concrete.

5.
5. Conclusions
Conclusions
Based
Basedon onthe
theuniaxial compression
uniaxial compressiontesttest
of freezing–thawing-damaged
of freezing–thawing-damaged concrete, a numerical
concrete, model
a numerical
was established using particle flow discrete element theory to study the influence of freezing–thawing
model was established using particle flow discrete element theory to study the influence of freezing–
damage
thawingon the meso-scale
damage failure mode
on the meso-scale of concrete.
failure The following
mode of concrete. conclusions
The following are drawn.
conclusions are drawn.
As the difference between the uniaxial compression axial deformation of the freezing–thawing
As the
concrete modeldifference
and thatbetween the uniaxial
of the numerical compression
simulation axial
was less deformation
than of the
1%, the model freezing–thawing
of freezing–thawing
concrete model and that of the numerical simulation
concrete and the selected parameters is accurate. was less than 1%, the model of freezing–thawing
concrete and the
Through selected
analysis of parameters
the contactisforce
accurate.
between particles in two different stage points of the
numerical model, we found that the softening effect of concrete after the peak force is considerably
Through analysis of the contact force between particles in two different stage points of the
weakened by the freezing–thawing effect on concrete. From 75 to 150 cycles of freezing–thawing,
numerical model, we found that the softening effect of concrete after the peak force is considerably
the softening effect after peak decreased by 39.7%. At the same time, by controlling the termination
weakened by the freezing–thawing effect on concrete. From 75 to 150 cycles of freezing–thawing, the
condition of uniaxial compression of the concrete numerical model, the contact force between particles
softening effect after peak decreased by 39.7%. At the same time, by controlling the termination
at any point on the stress–strain curve can be used to estimate the damage degree of the concrete
condition of uniaxial compression of the concrete numerical model, the contact force between particles
structure at different points.
at any point on the stress–strain curve can be used to estimate the damage degree of the concrete
structure at different points.

With the increase in freezing–thawing cycles, the number of cracks increased significantly. In the
second stage point, the crack growth rate from 75 to 150 cycles of freezing–thawing was 6.8%; in the third
Appl. Sci. 2020, 10, 1252 17 of 18

With the increase in freezing–thawing cycles, the number of cracks increased significantly. In the
second stage point, the crack growth rate from 75 to 150 cycles of freezing–thawing was 6.8%; in
the third stage point, the crack growth rate from 75 to 150 cycles of freezing–thawing was 12.2%.
In the third stage point, the crack growth rate was significantly higher than that in the second stage
point, which indicated that the failure of concrete mainly occurred in the post-peak softening stage.
By controlling the termination condition of uniaxial compression of the numerical model of concrete,
the number of cracks at any point on the stress–strain curve is obtained to predict the degree of
concrete damage.
With the increase in freezing–thawing cycles, the number of concrete fragments increased
gradually, and the damage worsened. From 125 to 150 cycles of freezing–thawing, the growth rate
of concrete fragments was the highest, with an increase of 48.1; the fragmentation after 150 cycles of
freezing–thawing was the most serious. Therefore, testing the mechanical properties of concrete is
necessary when concrete has undergone more than 125 cycles of freezing and thawing. At the same
time, the number of concrete fragments and the final failure morphology are used to predict the failure
mode and the damage degree of concrete in macroscopic physical test.

Author Contributions: Conceptualization, Z.S.; Data curation, X.D.; Formal analysis, Z.S. and X.D.; Investigation,
X.D. and L.H.; Project administration, L.H. and Y.L.; Resources, Z.S., L.H. and Y.L.; Validation, Y.L. All authors
have read and agreed to the published version of the manuscript.
Funding: This study was financially supported by the National Natural Science Foundation of China (Nos.
51879217, 51409207, and 51609200) and the National Science Foundation for Post-Doctoral Scientists of China (No.
2015M582765XB).
Acknowledgments: The author would like to thank the reviewers for their comments.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Hanjari, K.Z.; Utgenannt, P.; Lundgren, K. Experimental study of the material and bond properties of
frost-damaged concrete. Cem. Concr. Res. 2011, 41, 244–254. [CrossRef]
2. Almahmoud, F.; Mechling, J.M. Bond strength of different strengthening systems—Concrete elements under
freeze-thaw cycles and salt water immersion exposure. Constr. Build. Mater. 2014, 70, 399–409. [CrossRef]
3. Li, B.; Mao, J.; Nawa, T. Mesoscopic damage model of concrete subjected to freeze-thaw cycles using mercury
intrusion porosimetry and differential scanning calorimetry (MIP-DSC). Constr. Build. Mater. 2017, 147,
79–90. [CrossRef]
4. Powers, T.C. A working hypothesis for further studies of frost resistance of concrete. Am. Concr. Inst. 1950,
16, 245–272.
5. Powers, T.C. Freezing effect in concrete, Durability of Concrete. Am. Concr. Inst. 1975, 47, 1–12.
6. Powers, T.C.; Helmuth, R.A. Theory of volume changes in hardened Portland cement paste during freezing.
Proc. Highw. Res. Board 1953, 32, 285–297.
7. Hanjari, K.Z.; Kettil, P.; Lundgren, K. Modelling the structural behaviour of frost-damaged reinforced
concrete structures. Struct. Infrastruct. Eng. 2013, 9, 416–431. [CrossRef]
8. Cao, D.F. Experimental study on tensile properties of concrete after freeze-thaw cycles. Constr. Build. Mater.
2012, 4, 42–52.
9. Hasan, M.; Okuyama, H.; Sato, Y. Stress-strain model of concrete damaged by freezing and thawing cycles. J.
Adv. Concr. Technol. 2004, 2, 8999. [CrossRef]
10. Bogas, J.A.; Brito, J.D.; Ramos, D. Freeze–thaw resistance of concrete produced with fine recycled concrete
aggregates. J. Clean. Prod. 2016, 115, 294–306. [CrossRef]
11. Shang, H.S.; Song, Y.P. Experimental study of strength and deformation of plain concrete under biaxial
compression after freezing and thawing cycles. Cem. Concr. Res. 2006, 36, 1857–1864. [CrossRef]
12. Ma, B.; Ye, Y.H.; Sun, Y. Mechanical performance analysis of salt-frozen or freeze-thaw concrete based on
damage theory. J. Build. Struct. 2009, 30, 298–302.
13. Sun, W.; Mu, R.; Luo, X. Effect of chloride salt, freeze-thaw cycling and externally applied load on the
performance of the concrete. Cem. Concr. Res. 2002, 32, 1859–1864. [CrossRef]
Appl. Sci. 2020, 10, 1252 18 of 18

14. Zhu, J.P.; Li, S.C.; Ramos, D. Mechanical property deterioration model for concrete in environment with
freezing-thawing. J. Archit. Civ. Eng. 2009, 26, 62–67.
15. Hasan, M.; Ueda, T.; Sato, Y. Stress-Strain Relationship of Frost-Damaged Concrete Subjected to Fatigue
Loading. J. Mater. Civ. Eng. 2008, 20, 37–45. [CrossRef]
16. Liu, H.K.; Li, J. Constitutive law of attacked concrete subject to freeze-thaw cycles. Constr. Build. Mater. 2011,
14, 736–741.
17. Duan, A.; Tian, Y.; Dai, J.G. A stochastic damage model for evaluating the internal deterioration of concrete
due to freeze–thaw action. Mater. Struct. 2014, 47, 1025–1039. [CrossRef]
18. Shen, K.X. Study on Bearing Capacity of Reinforced Concrete Beams under Freeze-Thaw Cycle. Master’s
Thesis, XI’AN University of Architecture and Technology, Xi’an, China, 2014.
19. Yu, H.; Ma, H.; Yan, K. An equation for determining freeze-thaw fatigue damage in concrete and a model for
predicting the service life. Constr. Build. Mater. 2017, 137, 104–116. [CrossRef]
20. Guan, X. The Analysisi of Damage and the Research of Bearing Capacity of Reinforced Concrete Flexural
Members under the Environment of Freezing and Thawing. Ph.D. Thesis, XI’AN University of Architecture
and Technology, Xi’an, China, 2015.
21. Chen, G.X. Test Code for Hydraulic Concrete; Ministry of Water Resources of the PRC: Beijing, China, 2006.
22. Wang, Z.X. Uniaxial Test Simulation of Concrete Freeze-Thaw Damage Based on Particle Flow Discrete
Element Method. Master’s Thesis, Xi’an University of Technology, Xi’an, China, 2018.
23. Liu, Z.L.; Sun, W.W.; Wang, X.M. Research on numerical simulation method of projectile penetration into
micro concrete based on particle dispersion model. J. Vib. Shock 2016, 35, 162–169.
24. Li, Y.J.; Cui, W.; Qi, L. Numerical simulation of self compacting concrete performance based on discrete
element method. Eng. J. Wuhan Univ. 2019, 52, 33–39.
25. Itasca Consulting Group Inc. PFC3D User’s Manual; Version 5.0; Itasca Consulting Group: Minneapolis, MN,
USA, 2017.
26. Elias, J. Simulation of railway ballast using crushable polyhedral particles. Powder Technol. 2014, 264, 458–465.
[CrossRef]
27. Lichter, J.; Lim, K.; Potapov, A. New developments in cone crusher performance optimization. Miner. Eng.
2009, 22, 613–617. [CrossRef]
28. Geng, J.Y.; Bu, J.Q.; Gao, Y. Numerical Analysis of Concrete Triaxial Compression Test Based on Discrete
Element Method. J. Shijiazhuang Railw. Inst. 2015, 28, 88–91.
29. Xiao, H. Simulation Of Comcrete Failure Process Using Discrete Element Method Based On 2-D Particle
Flow Code. Master’s Thesis, Xiangtan University, Xiangtan, China, 2011.
30. Kang, Z.; Tang, X.W.; Qin, C. End effect of concrete by meso-scale discrete element modeling. J. Harbin Inst.
Technol. 2013, 45, 94–98.
31. Xu, S.J.; Yin, X.T.; Ma, S.K. Numerical Test Study of Concrete Material Based on Particle Flow. J. Exp. Mech.
2009, 24, 251–258.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).