Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Mechanics Research Communications 118 (2021) 103791

Contents lists available at ScienceDirect

Mechanics Research Communications


journal homepage: www.elsevier.com/locate/mechrescom

Prediction of the relationship between strength and porosity of pervious


concrete: A micromechanical investigation
Viet-Hung Vu d , Bao-Viet Tran a,b ,∗, Ba-Anh Le a,c , Hoang-Quan Nguyen a,b
a
Research and Application Center for Technology in Civil Engineering (RACE), University of Transport and Communications, 3 Cau Giay, Dong Da, Hanoi, Viet
Nam
b
Faculty of Construction Engineering, University of Transport and Communications, 3 Cau Giay, Dong Da, Hanoi, Viet Nam
c
Faculty of Civil Engineering, University of Transport and Communications, 3 Cau Giay, Dong Da, Hanoi, Viet Nam
d
Campus in Ho Chi Minh City, University of Transport and Communications, 450-451 Le Van Viet, Thu Duc City, Viet Nam

ARTICLE INFO ABSTRACT

Keywords: This paper aims to develop a micromechanical model to evaluate the compressive strength–porosity relation-
Pervious concrete ship of the pervious concrete mesostructure from its compositions. By considering this material within the
Micromechanical approach framework of the quasi elasto-brittle approach, failure is governed by quadratic strain and stress averages.
Compressive strength
Through this approach, a micromechanical uniaxial compressive strength criterion of von Mises-type is
Porosity
constructed from phase-averaged properties. Due to its ability to predict percolation thresholds, the use of
Brittle fracture
the self-consistent scheme is appropriate for determining the homogenized stiffness tensors. For the ease of
broad engineering applications, two weighted self-consistent schemes, in which complex geometric properties
are simplified by several free parameters, are proposed. First, the influences of these parameters on the
macroscopic strength of pervious concrete are investigated. Then, to show the reliability and accuracy of
the proposed approach, the present model’s predicting strength is compared to experimental observations and
existing models in the literature.

1. Introduction bearing surface. In general, its main components are composed of


uniform/single-sized aggregates, which have little or no fine aggre-
In the last few decades, due to rapid industrial development and gates, cement, and water, resulting in a large, open pore structure.
urbanization, the impermeable surface area (i.e., concrete pavements, The permeability of PC is determined by an interconnected macro-pore
buildings or other civil infrastructures, etc.) for water runoff in large structure (ranging from 2 to 8 mm [2]) and a relatively high void
cities has significantly reduced. As a result, there has been severe content/porosity, typically ranging from 15% to 35% by volume, which
flooding, which has seriously impacted the lower underground water corresponds to a possible and effective permeability up to 6 mm/s [2].
level and has created environmental problems (i.e., heat urban island During the design stage, the CS of normal Portland cement con-
effect, traffic noise, etc.) and has weakened driver’s safety (i.e., water crete is always controlled first. Since its main property/function is
splashing, slippery driving surfaces, etc.). In order to solve these prob- water permeability, common PC mixture proportion design methods
lems, several environmentally friendly approaches have been recently
are aimed with a target porosity. Generally, the following steps are
proposed and studied around the world, such as the sustainable urban
required: firstly, a target value of total porosity is chosen [2,3]; sec-
drainage system (SUDS), low-impact development (LID), etc. Contrary
ondly, CS is estimated based on the total porosity. However, in order
to conventional water drainage systems, the SUDS is a solution to
to apply PC in practical structures (i.e. car parking plots, local driveway
increase the permeable capability of surface layers by directly and
pavements, etc.), PC should be strong enough to carry service loading
quickly capturing all rainfall in the base/sub-base layer or facilitating a
part of accumulated rainwater volume that infiltrates underground [1]. and withstand wear and tear. Therefore, the balance between the PC
Portland cement pervious concrete (PC), also known as little or permeability and strength should be considered when defining a PC
no fine aggregate concrete or permeable concrete, plays an important mixture proportion. Additionally, it is difficult to manufacture an ideal
role in the SUDS. It serves dual purposes as it is one of the best PC with concurrently high permeability and CS. For instance, it is
stormwater management systems and also has a low/moderate load widely accepted that the more porous concrete is, the less strength

∗ Corresponding author.
E-mail address: viettb@utc.edu.vn (B.-V. Tran).

https://doi.org/10.1016/j.mechrescom.2021.103791
Received 11 August 2021; Received in revised form 6 October 2021; Accepted 7 October 2021
Available online 15 October 2021
0093-6413/© 2021 Elsevier Ltd. All rights reserved.
V.-H. Vu et al. Mechanics Research Communications 118 (2021) 103791

it holds, thus the greater its permeable capacity, and vice versa. In fractions 𝑣𝛼 (𝛼 = 1, … , 𝑛). The 𝛼-component has elastic stiffness fourth
the complex structure of pervious concrete, the pores can be present order tensor C𝛼 that is characterized by bulk and shear modulus (𝑘𝛼 ,
from the nano-scale (pores within cement paste) to the macro-scale 𝜇𝛼 ), a specific ellipsoidal shape noted by an aspect ratio 𝑟𝛼 . By hypoth-
(interconnected voids between coarse aggregate). It conducts that the esizing that the inclusions are distributed non-preferably (randomly)
porosity of porous concrete could be defined differently. In this study, in the space, the composite product is isotropic. The asymptotically
for clarity, we limit the relationship between the compressive strength exact solutions for the homogenized moduli of a dilute suspension of
of the mesostructure (e.g., a cubic sample 10 × 10 × 10 cm) with well-separated randomly oriented ellipsoids inclusions (called Dilute
the total macro-porosity, while the affection of the micro-porosity is Approximation) are given by [15]
examined in the relationship with the strength of cement paste.

𝑛 ∑
𝑛
Although, within research, the question remains how porosity and 𝑘ℎ = 𝑘1 + 𝑣𝛼 (𝑘𝛼 − 𝑘1 )𝐴𝑠𝛼 , 𝜇ℎ = 𝜇1 + 𝑣𝛼 (𝜇𝛼 − 𝜇1 )𝐴𝑑𝛼 (1)
other properties affect the strength of PC. Due to the complexity of 𝛼=2 𝛼=2
the PC structure, several models are widely derived from experimental
where 𝐴𝑠𝛼 , 𝐴𝑑𝛼 are the shape functions of the inclusion phase 𝛼 and
studies on this topic [4,5]. According to these models, the strength–
coincide respectively with the spherical and deviatoric components of
porosity relationship depends on the concrete matrix strength (whether
the strain localization fourth order tensor A𝛼 calculated based on the
macro porosity is equal to zero) and particular free parameters. There
Eshelby solution [15] for the geometric property of inclusions. It is
are three main types of models, which include linear, power, and
necessary to note that in (1), the 1-component plays the role of the
exponential (see in [4]). Based on basic equations, modified models
infinitive matrix medium.
were proposed to account for specific properties of the material, such
The differential approximation was constructed from the differential
as aggregate size [5], pore size [6,7], water–cement ratio [8], and
scheme process and the dilute solution result. At each step of the
elastic modulus of aggregate [4,9]. However, these quasi-experimental
procedure, we added proportionally infinitesimal volume amounts 𝑣𝛼 𝛥𝑡
type models have some disadvantages. For examples, while the linear
(𝛥𝑡 ≪ 1) of randomly oriented inclusions into the already constructed
regression formula cannot describe the non-linear relationship between
composite of the previous step, which contained volume fractions 𝑣𝛼 𝑡
the CS and porosity, the exponential and power models cannot pre-
dict thresholds where the material collapses. Therefore, it was merely of the inclusion phases (0 < 𝑡 < 1). Differential approximation for the
suitable for specific experimental cases. homogenized elastic moduli of the 𝑛-component matrix composite is
From a novel approach, the micromechanical model appears suit- the solution 𝑘ℎ = 𝑘(1), 𝜇 ℎ = 𝜇(1) of coupled differential equations

1 ∑ 1 ∑
able for this research topic, because it utilizes and accounts for the 𝑛 𝑛
𝑑𝑘 𝑑𝜇
available information about the concrete composition, microstructure = 𝑣𝛼 (𝑘𝛼 − 𝑘)𝐴𝑠𝛼 , = 𝑣 (𝜇 − 𝜇)𝐴𝑑𝛼 , (2)
𝑑𝑡 1 − 𝑣𝑡 𝛼=2 𝑑𝑡 1 − 𝑣𝑡 𝛼=2 𝛼 𝛼
of composites (e.g., orientations, shapes, volume fractions, spatial dis-
tribution of inclusions, etc.), and the behavior of each constituent. A with
large number of basic micromechanical models have been developed ∑
𝑛

within the framework of homogenization techniques for porous media, 𝑘(0) = 𝑘1 , 𝜇(0) = 𝜇1 , 0 ≤ 𝑡 ≤ 1, 𝑣= 𝑣𝛼 . (3)
𝛼=2
which can be found in the mechanical handbook of Dormieux et al. [10]
and several reviews [11,12]. Several theoretical formulas based on the If, in Eq. (2), the matrix phase is eliminated, then 𝑣 = 1 and at
micromechanical framework are presented in the literature [13,14] for the end of the process, the multiplier 1∕(1 − 𝑣𝑡) → ∞, hence the sums
∑𝑛 𝑠 ∑𝑛 𝑑
the strength–porosity relationship of PC. In [13], porous concrete is 𝛼=1 𝑣𝛼 (𝑘𝛼 − 𝑘)𝐴𝛼 → 0 and 𝛼=1 𝑣𝛼 (𝜇𝛼 − 𝜇)𝐴𝛼 → 0 as 𝑘, 𝜇 should be
modeled as two-phase material (pore and concrete matrix). The average finite. Thus, we obtain what is called the Self-consistent Approxima-
stress in the concrete domain was constructed in hydrostatic and simple tion (SC) for homogenized elastic moduli (𝑘ℎ , 𝜇ℎ ) of the 𝑛-component
shear conditions. Then, a peak average stress criterion was applied to mixture of particulates, which is the solution of the self-consistent
calculate the effective strength of the material. Although, due to the equations
hydrostatic stress/strain condition, this model is not a good fit for the ∑
𝑛 ∑
𝑛
current experimental data. Also, Li et al. [14] recently proposed a local 𝑣𝛼 (𝑘𝛼 − 𝑘ℎ )𝐴𝑠𝛼 = 0, 𝑣𝛼 (𝜇𝛼 − 𝜇 ℎ )𝐴𝑑𝛼 = 0. (4)
peak stress criterion at poles of a spherical pore to define the uniaxial 𝛼=1 𝛼=1
compressive strength of a porous medium. However, this simplified
central pore model is not enough to account for the complexity of 2.2. Homogenization of uniaxial compressive strength for elasto-brittle ma-
concrete structures. terial
As stated above, it appears that a micromechanics-based approach
to investigate the strength properties of PC is promising. Therefore, In line with earlier research on strength of cementitious composite
beginning in Section 2, fundamentals of a homogenized technique are materials [16], herein pervious concrete material is considered within
developed to establish a uniaxial strength formula for multi-component the frame work of the quasi elasto-brittle approach. The failure of the
composite materials. A short review connecting experimental obser- entire RVE depends on the microscopic failure of particulate phases.
vations with existing models is presented in Section 3. To balance According to the von Mises criterion, failure is governed by phase-
the practical effectiveness of the micromechanical model, an adaptive averaged quadratic strain and stress rather than by phase-averaged
strategy is proposed in Section 4, in which two self-consistent schemes strain/stress. Based on the successfully developed mathematical for-
with weighted parameters are investigated. Furthermore, a comparison mulation of Kreher [17], Dormieux et al. [18], Pichler et al. [16] for
is made between the present model and the collected experimental calculating quadratic strain averages in terms of the solid phase elastic
data, alongside other existing models to show findings. energy, the following equations are deduced:
( 𝜕(Cℎ ) )1∕2
1
2. Homogenized elasticity and strength of multi-component brit- 𝜀𝑑𝛼 = 𝐄∶ ∶𝐄 ,
4𝑣𝛼 𝜕(𝜇𝛼 )
tle composite material

2.1. Self-consistent approximation of elasticity ( 𝜇2 𝜕(Cℎ ) )1∕2


𝛼
and 𝜎𝛼𝑑 = 2𝜇𝛼 𝜀𝑑𝛼 = 𝐄∶ ∶𝐄 (5)
𝑣𝛼 𝜕(𝜇𝛼 )
In this paper, we examine a Representative Volume Element (RVE)
of a multi-component pervious concrete material that occupies region where 𝜀𝑑𝛼 , 𝜎𝛼𝑑 respectively denotes quadratic strain and stress averages
𝛺 consisting of 𝑛 components occupying regions 𝛺𝛼 ⊂ 𝛺 of volume of 𝛼-phase, 𝐂ℎ is the homogenized elastic stiffness tensor, and 𝐄 is

2
V.-H. Vu et al. Mechanics Research Communications 118 (2021) 103791

the uniform macrostrain imposed on the RVE. Based on the inverse


elasticity law between macrostress Σ and macrostrain
( 𝜇2 𝜕(1∕Cℎ ) )1∕2
𝐄 = (1∕Cℎ ) ∶ Σ ⇒ 𝜎𝛼𝑑 = − 𝛼 Σ ∶ ∶Σ . (6)
𝑣𝛼 𝜕(𝜇𝛼 )
For simplicity, the isotropic linear elastic stiffness tensor can be ex-
pressed as follows:

Cℎ = 3𝑘ℎ J + 2𝜇 ℎ K, J = 1∕3𝟏 ⊗ 𝟏, K = I − J, (7)

with I, 𝟏 are the fourth and second order (symmetric) identity tensors
of components in an orthonormal frame. For instance, we are interested
in the uniaxial compressive strength problem (along axe 1) where the
macrostress at RVE has form:

Σ = −𝛴𝑒1 ⊗ 𝑒1 with 𝛴 > 0. (8)

Substitution of Eqs. (7) and (8) into Eq. (6), the formulation of quadratic
strain average for phase 𝛼 can be drawn:
Fig. 1. Comparisons between existing models and experimental observations.
( 𝜇2 )
1 𝜕(1∕𝑘ℎ ) 1 𝜕(1∕𝜇 ℎ ) 1∕2
𝜎𝛼𝑑 = 𝛴 − 𝛼 ( + ) . (9)
𝑣𝛼 9 𝜕(𝜇𝛼 ) 3 𝜕(𝜇𝛼 )
Table 1
Specification of (9) for one homogeneous 𝛼-component case reads as: Existing models for porous material’s strength prediction.

( 𝜇2 ) Ref. Equation Relationship 𝑅2


1 𝜕(1∕𝑘𝛼 ) 1 𝜕(1∕𝜇𝛼 ) 1∕2 𝛴
𝜎𝛼𝑑 = 𝛴 − 𝛼 ( + ) = √ . (10) [4] 𝑆 ℎ = 𝑆𝑐 (1 − 𝑚𝑣𝑝 ) Linear 0.72
𝑣𝛼 9 𝜕(𝜇𝛼 ) 3 𝜕(𝜇𝛼 ) 3 [5] 𝑆 ℎ = 𝑆𝑐 (1 − 𝑚𝑣𝑝 )( 𝑑𝑑 )𝑛 Linear -
0

According to the elasto-brittle theory, 𝛼-component is characterized [4] 𝑆 ℎ = 𝑆𝑐 (1 − 𝑣𝑝 )𝑛 Power 0.84


ln(𝑑 )
[6] 𝑆ℎ = 𝛼0 + 𝛼1 [ 𝑀𝐹 𝑆 ]+ Power -
by a critical value 𝑑
𝜎𝛼,𝑐𝑟𝑖𝑡 .
The behavior of 𝛼-component maintains 𝑣
ln(𝑑𝑛 )
+𝛼2 ( 𝑆𝑝 )−1 + 𝛼3 ln(𝛤3𝐷 )
elasticity when the quadratic strain average is below this yield stress 𝑝

[4] 𝑆 ℎ = 𝑆𝑐 exp(−𝑚𝑣𝑝 ) Exponential 0.83


value. The condition corresponds to a von Mises-type elastic limit [9] 𝑆ℎ = 𝑆𝑐 exp[−(𝑚 − 𝑛𝐹 𝑀)𝑣𝑝 ] Exponential -
criterion, where the equal sign presents the elastic limit. If 𝑆𝛼 denotes [8] 𝑆 ℎ = (𝑎 + 𝑏𝑥 + 𝑐𝑥2 ) ln 𝑣𝑝 + 𝑒 Logarithm 0.83

the uniaxial strength of homogeneous 𝛼-component, corresponding to [4] 𝑆 ℎ = 𝑆𝑐 (1 − 𝑣𝑝 )𝑚 exp (−𝑛𝑣𝑝 ) Mixed 0.84
maximum macrostress whether the failure criterion is valid, we have (1−𝑣𝑝 )2
the criterion for one homogeneous phase: [14] 𝑆ℎ = 𝑆𝑐 5𝑣𝑝 +1
Micromechanical 0.65

𝑆
𝑑
𝜎𝛼𝑑 ≤ 𝜎𝛼,𝑐𝑟𝑖𝑡 = √𝛼 . (11)
3
3. Data on the strength–porosity relationship of pervious concrete
Returning to RVE situation, linear elastic behavior prevails as long
as the quadratic deviatoric stress averages over each phase remain
It is widely recognized that the compressive strength (CS) of pervi-
𝑑
below the critical strength value 𝜎𝛼,𝑐𝑟𝑖𝑡 . Together with Eqs. (9) and ous concrete (PC) is primarily dependent on its porosity, which in turn
(11), the uniaxial macroscopic elastic limit criterion of the isotropic highly depends on aggregate gradation, cementitious material content,
multi-component material becomes: water-to-cement ratio, and compaction effort during placement. Based
( 𝜇2 ) on our open literature review, 164 PC samples with different mixture
1 𝜕(1∕𝑘ℎ ) 1 𝜕(1∕𝜇 ℎ ) 1∕2 𝑆
𝛴 − 𝛼( + ) 𝑑
≤ 𝜎𝛼,𝑐𝑟𝑖𝑡 = √𝛼 , ∀𝛼 (12) proportions and experimental processing were compiled in order to
𝑣𝛼 9 𝜕(𝜇𝛼 ) 3 𝜕(𝜇𝛼 ) 3
( ) observe the relationship between porosity and strength in PC (see in
( 𝜇2 𝜕(1∕𝑘ℎ ) 𝜕(1∕𝜇 ℎ ) )−1∕2
1 Fig. 1). The total porosity of PC is the sum of two components: open
⇒ 𝛴 ≤ 𝑆𝛼 − 𝛼 ( + ) , ∀𝛼. (13)
𝑣𝛼 3 𝜕(𝜇𝛼 ) 𝜕(𝜇𝛼 ) (or effective) porosity and isolated porosity. While the effective pores
Finally, the algorithm for evaluating the homogenized uniaxial com- are considered to be a crucial factor of permeability/infiltration, the
pressive strength of the multi-component linear elastic brittle material strength of PC is greatly affected by the total porosity. As mentioned in
is deduced from the criterion (13). Section 1, several researchers have aimed to observe its effect on the
( ) macroscopic strength of material. Typically, the three aforementioned
( 𝜇2 ℎ 𝜕(1∕𝜇ℎ ) )−1∕2
ℎ 𝛼 1 𝜕(1∕𝑘 ) general model types have been developed for different porous materials
𝑆 = min 𝑆𝛼 − ( + ) . (14)
𝛼=1..𝑛 𝑣𝛼 3 𝜕(𝜇𝛼 ) 𝜕(𝜇𝛼 ) in terms of semi-empirical equations, as presented in Table 1, such
It is necessary to stress that, the homogeneous strength by phase as linear, power and exponential equations, which were originally
𝑆𝛼 plays an important role in (14). For a normal pervious concrete for glass, metal, and ceramic, respectively. Even though PC has been
structure where damage zone propagates in the cement paste, 𝑆 has to studied and developed for a long time, it can be seen that few strength
be determined for cement paste medium and transition zones. On the prediction models have been proposed exclusively for relatively high
one hand, the strength can be measured by the compressive test. On porosity (typically in the ranges of 15 to 40%) and environmentally
the other hand, 𝑆 can be predicted by applying (14) in the multi-scale friendly material. In order to illustrate the reliability and accuracy of
framework [19]. Even if we can predict the average strength of cement the existing models in predicting the CS from the total porosity of
paste/ITZ based on the upscaling procedure and/or empirical works, PC, the best fitting curves are derived and plotted through comparing
the geometric parameters in (14) still have to be calibrated individually collected experimental databases with current models, as shown in
caused by the complexity of pervious concrete structures. Therefore, Fig. 1. The corresponding determination coefficients 𝑅2 for each model
instead of constructing an upscaling procedure with a number of un- are also calculated in Table 1. It is necessary to note that there is not
certain parameters, we limit the model at the mesostructure scale in a much to physically observe from the fitting curves from 14 different
simple adaptive homogenized procedure with a minimum number of sources. However, qualitative comparisons and interesting observations
free parameters (Section 4). can be made from this study.

3
V.-H. Vu et al. Mechanics Research Communications 118 (2021) 103791

𝑣𝑐 (𝜇𝑐 − 𝜇 ℎ )𝐴𝑑𝑐 + 𝑣𝑝 (0 − 𝜇 ℎ )𝐴𝑑𝑝 = 0. (16)

𝑘𝑐 , 𝜇𝑐 are bulk and shear moduli of concrete matrix (at macro porosity
being equal to zero) and 𝐴𝑠𝛼 , 𝐴𝑑𝛼 (𝛼 = 𝑐, 𝑝) are respectively spherical
and deviatoric shape functions that depended on elastic properties of
homogenized medium (𝑘ℎ , 𝜇ℎ ), inclusions (𝑘𝑐 , 𝜇𝑐 , zero for pore), and
shape of inclusions (characterized by ratio 𝑟𝑐 and 𝑟𝑝 in this situation).
The formulation of 𝐴𝑠𝛼 , 𝐴𝑑𝛼 (𝛼 = 𝑐, 𝑝) can be evaluated analytically
based on the solution of the localization problem of isotropically dis-
tributed ellipsoidal inclusions in infinitive homogeneous domain [15],
and is presented in Appendix A. Substituting (𝑘ℎ , 𝜇ℎ ) calculated from
Eqs. (15), (16) into Eq. (14), we get
( 𝜇𝑐2 )
1 𝜕(1∕𝑘ℎ ) 𝜕(1∕𝜇 ℎ ) −1∕2
𝑆 ℎ = 𝑆𝑐 − ( + ) , (17)
(1 − 𝑣𝑝 ) 3 𝜕(𝜇𝑐 ) 𝜕(𝜇𝑐 )
Fig. 2. Self-consistent schemes: (a) - two-phase model; (b) - three-phase model. where 𝑆𝑐 denotes the uniaxial CS of concrete matrix.
Once the elastic modulus of the concrete matrix is available, the
strength–porosity relationship is estimated from Eq. (17) depending
on strength of concrete and shape of material phases. For simplicity,
we consider the Poisson ratio 𝜈𝑐 = 0.2 (from Miled et al. [20]). This
implies that the normalized effective Young’s modulus, bulk modulus,
and shear modulus have the quasi same forms for all the considered
homogenization schemes. Furthermore, the empirical relationship be-
tween strength and Young’s modulus of concrete matrix proposed in
ACI-318 could be accepted

𝐸𝑐 = 3650 𝑆𝑐 (MPa). (18)

Thus, we have Eq. (17) as the weighted two-phase self-consistent


approximation (WSC2) to predict the relationship between strength
and porosity of such PC. From an empirical point of view, in this
simple situation, the model (17) has one material parameter (strength
of concrete matrix) and two free parameters (ratios 𝑟𝑝 and 𝑟𝑐 ) as the
empirical models for porous concrete cited above.
To evaluate this model, the strength–porosity relationship in the
Fig. 3. Weighted self-consistent estimate of dimensionless homogenized uniaxial com-
case of spherical inclusions is first established, then references this for
pressive strength of a two-phase porous medium (WSC2) as a function of 𝑣𝑝 for case other situations. From Eqs. (15), (16), we have
of 𝑟𝑝 = 𝑟𝑐 < 1.
𝜇ℎ √
𝑘ℎ
= = 1 − 2𝑣𝑝 , then 𝑆 ℎ = 𝑆𝑐 (1 − 𝑣𝑝 )(1 − 2𝑣𝑝 ). (19)
𝑘𝑐 𝜇𝑐
In Fig. 3, we fixed 𝑟𝑐 = 𝑟𝑝 and changed its value from 0 to 1. A
4. Weighted self-consistent models
few interesting curves can be obtained when the ratios tend to zero
(𝑟𝑐 = 𝑟𝑝 < 0.1). This shows that, for some cases it could be interesting
Following Section 2, one mathematical elastic limit criterion of
to use only one free parameter to create simulated curves that agree
multi-component composite material was constructed in Eq. (13),
well with the database.
which allowed us to determine the uniaxial CS of composite material
that depends on mechanical and geometrical properties of the com-
4.2. Three-phase model
ponent phase material (Eq. (14)). This section explores whether this
criterion has the ability to model the relationship between uniaxial CS
Recent research on cement paste’s influence on strength, water-to-
and the porosity of PC material.
cement ratio, aggregate mixture to the strength of pervious concrete [3,
5,8] suggests that a more reliable model can be created by accounting
4.1. Two-phase model
for detailed properties of the material. From the image processing
results and procedures for characterizing the skeleton structures of
Firstly, a particularly simple situation, where the two-component
pervious concretes [3], it was considered that all coarse aggregates
composite is a suspension of random same shape ellipsoidal pores of
were coated by thin layers of cement paste and bonded by another
porosity 𝑣2 = 𝑣𝑝 in a continuous concrete matrix of volume fraction 𝑣1 =
part of cement paste. Pores were then formed from inter-space between
𝑣𝑐 = 1 − 𝑣𝑝 , is considered. This configuration is based on the empirical
coated mortar-aggregate inclusions. So, we propose a three-pattern self-
models cited below, in which the uniaxial CS of PC depends only on
consistent scheme for pervious concrete consisting of: (b1) a suspension
the strength of the concrete matrix and total porosity. The main idea
of coated spheres in an infinitive homogenized domain; (b2) cement
of the self-consistent scheme is that there are two types of inclusions
paste phase of ellipsoidal shape embedded in an infinitive homogenized
embedded in an effective equivalent infinite medium corresponding to
domain; (b3) porous phase of ellipsoidal shape embedded in an infini-
two localization problems in Fig. 2(a): (a1) concrete matrix inclusions
tive homogenized domain (Fig. 2(b)). The three phases examined in
of ellipsoidal form characterized by an aspect ratio 𝑟𝑐 ; (a2) pores of
this model are cement paste (characterized by elastic moduli 𝑘𝑐𝑚 , 𝜇𝑐𝑚 ,
ellipsoidal form characterized by an aspect ratio 𝑟𝑝 . Mathematically,
uniaxial compressive strength 𝑆𝑐𝑚 , volume fraction 𝑣𝑐𝑚 ), aggregate
from Eqs. (4), the couple self-consistent equations for the two-phase
(characterized respectively by 𝑘𝑎 , 𝜇𝑎 , 𝑆𝑎 , 𝑣𝑎 ), and pore.
model can be deduced respectively as follows
For the localization problem 1 (Fig. 2 (b1)), to balance between
𝑣𝑐 (𝑘𝑐 − 𝑘ℎ )𝐴𝑠𝑐 + 𝑣𝑝 (0 − 𝑘ℎ )𝐴𝑠𝑝 = 0, (15) effectiveness and applicability of the model, we considered here coated

4
V.-H. Vu et al. Mechanics Research Communications 118 (2021) 103791

Fig. 4. Weighted self-consistent estimate of dimensionless homogenized uniaxial com-


pressive strength of a three-phase porous medium (WSC3) as a function of 𝑣𝑝 for case
of 𝑓 = 0.1, 𝑆𝑎 = 𝑆𝑐𝑚 , 𝑓𝑎 = 𝑣𝑎 ∕(𝑣𝑎 + 𝑣𝑐𝑚 ) = 0.6, 𝑟 < 1.
Fig. 5. Comparison of WSC2 (𝑟𝑐 = 0.02, 𝑟𝑝 = 0.015, 𝑆𝑐 = 120) and WSC3 (𝑟 = 0.12, 𝑓 =
0.1, 𝑆𝑐𝑚 = 80, 𝑆𝑎 ∕𝑆𝑐𝑚 = 1.2, 𝑣𝑎 ∕(𝑣𝑎 + 𝑣𝑐𝑚 ) = 0.5) with the experimental data and the
existing models.
inclusions having spherical form. This can be solved through the mi-
cromechanical framework for ellipsoidal coated inclusions, however,
this scheme directs to more complex calculations. It is noted that If the database in Fig. 1 is considered as a reference, the calculations
𝐴𝑠𝑎 , 𝐴𝑑𝑎 , 𝐴𝑠𝑐𝑚1 , 𝐴𝑑𝑐𝑚1 are respectively spherical and deviatoric shape func- show a strong correlation between the results of the present approx-
tions of spherical core aggregate and shell layer cement paste. The imation and the data in the literature when 𝑟, 𝑓 take values close to
strategy to find the tensors is based on the solution of displacement 0.1. To evaluate the effectiveness of the model, a similar illustration
fields in polar coordinates [21]. Explicit solutions of four functions of Fig. 1 is reported in Fig. 5, in which the two approximations WSC2
are constructed in [22], and will be represented in Appendix B. Fur- and WSC3 are fitted with the collected experimental data, while the
thermore, determining the volume fraction of layer cement paste is linear, micromechanical, and mixed models are reported as references.
a challenge. By using the result from the image processing technique With 𝑟𝑐 = 0.02, 𝑟𝑝 = 0.015, 𝑆𝑐 = 120, the WSC2 yields 𝑅2 = 0.68,
(in [3]), the thickness of this layer can be measured. However, this while 𝑅2 of the WSC3 is equal to 0.84 for 𝑟 = 0.12, 𝑓 = 0.1, 𝑆𝑐𝑚 =
calculation is sensitive and particular. If we note 𝑓 , factor between 80, 𝑆𝑎 ∕𝑆𝑐𝑚 = 1.2, 𝑣𝑎 ∕(𝑣𝑎 +𝑣𝑐𝑚 ) = 0.5. The two approximations predict the
volume fraction of layer cement paste to the total volume fraction of pore percolation threshold close to 0.4. The WSC2 fits well for porosity
cement paste, 𝑓 could play a role in the weighted parameter of the from 0.2 to 0.3. A good correlation between the WSC3 and the data
model. was observed over all the ranges of porosity. This result shows that,
The localization problems (b2), (b3) in Fig. 2 are equivalent to the WSC3 can be applied for any particular mixtures of PC.
localization problems (a1), (a2) of the two-phase model. Subscript 𝑐 is
replaced by 𝑐𝑚2. In accepting the result from Fig. 3, for simplicity, it is 5. Conclusions
considered that the aspect ratio of pore and cement paste has the same
value, denoted by 𝑟. So, 𝑓 , 𝑟 present two weighted parameters of this The uniaxial compressive strength of pervious concrete material
three-pattern self-consistent model to predict the relationship between can be well simulated by a micromechanical model in the framework
uniaxial strength and total porosity of pervious concrete. of the homogenized technique and quasi-brittle failure governed by
The couple self-consistent equations have form a von Mises-type criterion. Once all of the mechanical and geomet-
rical properties of component phases at the mesostructure scale are
𝑣𝑎 (𝑘𝑎 − 𝑘ℎ )𝐴𝑠𝑎 + 𝑓 𝑣𝑐𝑚 (𝑘𝑐𝑚 − 𝑘ℎ )𝐴𝑠𝑐𝑚1 + available, the strength of pervious concrete can be estimated through
Eq. (14). The consistent results demonstrate that the present approach
+(1 − 𝑓 )𝑣𝑐𝑚 (𝑘𝑐𝑚 − 𝑘ℎ )𝐴𝑠𝑐𝑚2 + 𝑣𝑝 (0 − 𝑘ℎ )𝐴𝑠𝑝 = 0, (20)
can sufficiently predict the CS of PC based on the material property.
𝑣𝑎 (𝜇𝑎 − 𝜇ℎ )𝐴𝑑𝑎 + 𝑓 𝑣𝑐𝑚 (𝜇𝑐𝑚 − 𝜇ℎ )𝐴𝑑𝑐𝑚1 + Because of the complexity of pervious concrete structures, it is not
easy to apply the approximations to direct engineering applications.
+(1 − 𝑓 )𝑣𝑐𝑚 (𝜇𝑐𝑚 − 𝜇 ℎ )𝐴𝑑𝑐𝑚2 + 𝑣𝑝 (0 − 𝜇ℎ )𝐴𝑑𝑝 = 0, (21) Therefore, an adaptive strategy to connect theoretical and experimental
results is proposed. Firstly, two simple schemes are proposed: one is
and formula (14) can be used in the form a classical self-consistent scheme of pore and concrete matrix phase;
( )
( 𝜇2 ) the other is a combination of classical self-consistent schemes with
1 𝜕(1∕𝑘ℎ ) 𝜕(1∕𝜇 ℎ ) −1∕2
𝑆 ℎ = min 𝑆𝛼 − 𝛼 ( + ) . (22) the generalized self-consistent scheme (Hashin composite sphere as-
𝛼=𝑎,𝑐𝑚 𝑣𝛼 3 𝜕(𝜇𝛼 ) 𝜕(𝜇𝛼 )
semblage) to account for the interaction of three phases: aggregate,
In Eq. (22), the uniaxial strength of pervious concrete can be pre- cement paste, and pore. The weighted parameters (𝑟, 𝑓 ), characterizing
dicted depending on mechanical properties of solid skeleton structures geometric properties of materials, hence might be flexible enough to
and two geometric weighted parameters 𝑓 , 𝑟. In the following, we approximate practical mixtures. Elastic properties of the solid skeleton
derivate the Eq. (22) for particular situations to clarify some influencing could be estimated by engineering formulation, in case of limitation.
factors. Clearly, the Poisson ratio is always equal to 0.2 and the Secondly, instead of tensorial calculations, a series of explicit algo-
relationship between compressive strength and Young’s modulus of the rithms resumed by Eqs. ((15)–(17), (23) for the two-phase model
solid skeleton is expressed as in Eq. (18). (WSC2) and by Eqs. (20)–(22), (23)–(26) for the three-phase model
To evaluate the influence of free parameters 𝑟, 𝑓 , we fix 𝑓 = 0.1, (WSC3) are presented. This result can reduce the number of exper-
change 𝑟 from 1 to 0 (Fig. 4) with 𝑆𝑐𝑚 = 𝑆𝑎 and 𝑓𝑎 = 𝑣𝑎 ∕(𝑣𝑎 +𝑣𝑐𝑚 ) = 0.6. iments, whereas we need only two or three experimental values to

5
V.-H. Vu et al. Mechanics Research Communications 118 (2021) 103791

determine the free parameters of the model such as 𝑆, 𝑟, 𝑓 . In this


regard, the proposed approach appears more as a refined procedure ⎧
⎪ 𝑦 = (𝑓 𝑣𝑐𝑚 ∕(𝑓 𝑣𝑐𝑚 + 𝑣𝑎 ))1∕3
( ( ) )( ( )) ( )2
based on an analytical homogenization strategy rather than a multi- ⎪ 𝐴 = −4 𝜂3 − 2 𝛽 4 − 5 𝜈𝑐𝑚 𝑦3 𝜂1 𝑦7 − 𝜂2 7 − 10 𝜈𝑐𝑚 − 126 𝛽 𝜂2 𝑦3 1 − 𝑦2
⎪ ( ( ) 3) ( ( )) ( )
scale analytical approach itself. Therefore, the approximations appear 2 2
⎪ 𝐶 = − 𝜂3 + 𝛽 7 − 5 𝜈𝑐𝑚 𝑦 7 3
4 𝜂1 𝑦 + 𝜂2 7 + 5 𝜈𝑚 − 126 𝛽 𝜂2 𝑦 1 − 𝑦

sufficiently general, simple, flexible, and are expected to be useful for ⎪ 𝑋0 = 𝜇ℎ ∕𝜇𝑐𝑚
⎪ ( ) ( )
broad engineering applications. ⎪
𝜂1 = 49 − 50 𝜈𝑐𝑚 𝜈𝑎 𝛽 + 35 (1 + 𝛽) 𝜈𝑎 − 2 𝜈𝑐𝑚 + 70 𝜈𝑎 − 35 𝜈𝑐𝑚
( )
⎪ 𝜂2 = 7 + 5 𝜈𝑎 (1 + 𝛽) + 28 − 40 𝜈𝑎
⎨ ( )
⎪ 𝜂3 = 2 (1 + 𝛽) 4 − 5 𝜈𝑐𝑚 + 7 − 5 𝜈𝑐𝑚
Declaration of competing interest ⎪ 𝛽 = 𝜇𝑎 ∕𝜇𝑐𝑚 − 1
⎪ ( ( ) ( ) )( )
⎪ 𝛥 = 2 4 − 5 𝜈 ℎ 𝐶 + 7 − 5 𝜈 ℎ 𝐴𝑋0 𝑋0 − 1 + ⋯
⎪ ( )( ( ) 3 ( ) )
⎪ ...525 𝜂2 1 − 𝜈𝑐𝑚 2 𝛽 𝜈𝑐𝑚 − 𝜈 𝑦 + 1 − 𝜈 ℎ 𝜂3 𝑋0

The authors declare that they have no known competing finan- ⎪ 𝜈𝑎 = (3𝑘𝑎 − 2𝜇𝑎 )∕(6𝑘𝑎 + 2𝜇𝑎 ), 𝜈𝑐𝑚 = (3𝑘𝑐𝑚 − 2𝜇𝑐𝑚 )∕(6𝑘𝑐𝑚 + 2𝜇𝑐𝑚 )

cial interests or personal relationships that could have appeared to ⎪ 𝜈 ℎ = (3𝑘ℎ − 2𝜇 ℎ )∕(6𝑘ℎ + 2𝜇 ℎ )

influence the work reported in this paper.
(26)

Acknowledgment
References

This research is supported by Vietnam Ministry of Education and [1] B. Ferguson, Porous Pavements, CRC Press, 2005.
Training under the grant number B2020-GHA-07. [2] American Concrete Institute (ACI) Pervious Concrete, 522-R10: ACI 522
Committee Report, ACI, Farmington Hills, MI, 2010.
[3] X. Xie, T. Zhang, C. Wang, Y. Yang, A. Bogush, E. Khayrulina, Z. Huang, J. Wei,
Q. Yu, Mixture proportion design of pervious concrete based on the relationships
Appendix A. Localization tensor of isotropically distributed ellip-
between fundamental properties and skeleton structures, Cem. Concr. Compos.
soidal inclusions in infinitive homogeneous domain 113 (2020) 103693.
[4] C. Lian, Y. Zhuge, S. Beecham, The relationship between porosity and strength
for porous concrete, Constr. Build. Mater. 25 (11) (2011) 4294–4298.
[5] R. Zhong, K. Wille, Compression response of normal and high strength pervious
⎧ 𝐴𝑠𝛼 = 𝐹 1∕𝐹 2
⎪ ( ) concrete, Constr. Build. Mater. 109 (2016) 177–187.
⎪ 𝐴𝑑𝛼 = 1∕5 2 𝐹 3−1 + 𝐹 4−1 + (𝐹 4𝐹 5 + 𝐹 6𝐹 7 − 𝐹 8𝐹 9)(𝐹 2𝐹 4)−1 [6] O. Deo, N. Neithalath, Compressive behavior of pervious concretes and a
⎪ ( )
⎪ 𝐹 1 = 1 + 𝐴𝐴 3∕2 𝑓𝑥 + 3∕2 𝜃 − 𝑅 (3∕2 𝑓 𝑓 + 5∕2 𝜃 − 4∕3) quantification of the influence of random pore structure features, Mater. Sci.
( ( ))
⎪ 𝐹 2 = 1 + 𝐴𝐴 1 + 3∕2 𝑓𝑥 + 3∕2 𝜃 − 1∕2 𝑅 3 𝑓𝑥 + 5 𝜃 + ⋯ Eng. A 528 (1) (2010) 402–412.
⎪ ( ( )) [7] R. Zhong, K. Wille, Linking pore system characteristics to the compressive
⎪ ... + 𝐵 (3 − 4 𝑅) + 1∕2 𝐴𝐴 (𝐴𝐴 + 3 𝐵) (3 − 4 𝑅) 𝑓𝑥 + 𝜃 − 𝑅 2 𝜃 2 + 𝑓𝑥 − 𝜃
⎪ ( ( )) behavior of pervious concrete, Cem. Concr. Compos. 70 (2016) 130–138.
⎪ 𝐹 3 = 1 + 𝐴𝐴 1 − 𝑓𝑥 − 3∕2 𝜃 + 𝑅 𝑓𝑥 + 𝜃
( ( )) [8] L.G. Li, J.-J. Feng, J. Zhu, S.-H. Chu, A.K.H. Kwan, Pervious concrete: effects of
⎪ 𝐹 4 = 1 + 1∕4 𝐴𝐴 𝑓𝑥 + 3 𝜃 − 𝑅 𝑓𝑥 − 𝜃
⎪ ( ( ) ) porosity on permeability and strength, Mag. Concr. Res. 73 (2) (2021) 69–79.
⎪ 𝐹 5 = 𝐴𝐴 𝑅 𝑓𝑥 + 𝜃 − 4∕3 − 𝑓𝑥 + 𝐵𝜃 (3 − 4 𝑅) [9] P. Chindaprasirt, S. Hatanaka, N. Mishima, Y. Yuasa, T. Chareerat, Effects of
⎪ ( ( ))
⎨ 𝐹 6 = 1 + 𝐴𝐴 1 + 𝑓𝑥 − 𝑅 𝑓𝑥 + 𝜃 + 𝐵 (1 − 𝜃) (3 − 4 𝑅) binder strength and aggregate size on the compressive strength and void ratio
⎪ ( ( )) of porous concrete, Int. J. Miner. Metall. Mater. 16 (6) (2009) 714–719.
𝐹 7 = 2 + 1∕4 𝐴𝐴 3 𝑓𝑥 + 9 𝜃 − 𝑅 3 𝑓𝑥 + 5 𝜃 + 𝐵𝜃 (3 − 4 𝑅)
⎪ ( ) [10] L. Dormieux, D. Kondo, F.-J. Ulm, Microporomechanics, John Wiley & Sons,
⎪ 𝐹 8 = 𝐴𝐴 1 − 2 𝑅 + 1∕2 𝑓𝑥 (𝑅 − 1) + 1∕2 𝜃 (5 𝑅 − 3) + 𝐵 (1 − 𝜃) (3 − 4 𝑅)
⎪ ( ) 2006.
⎪ 𝐹 9 = 𝐴𝐴 𝑓𝑥 (𝑅 − 1) − 𝑅𝜃 + 𝐵𝜃 (3 − 4 𝑅) [11] R. Penta, J. Merodio, Homogenized modeling for vascularized poroelastic
⎪ 𝐴𝐴 = 𝜇𝛼 ∕𝜇𝑖𝑛 − 1
⎪ materials, Meccanica 52 (14) (2017) 3321–3343.
⎪ 𝑅 = (1 − 2𝜈 𝑖𝑛 )∕(2 − 2𝜈 𝑖𝑛 ) [12] H. Dehghani, I. Noll, R. Penta, A. Menzel, J. Merodio, The role of microscale solid

⎪ 𝑓𝑥 = 𝑟2 (2 − 3𝜃)(𝑟2 − 1)−1 matrix compressibility on the mechanical behaviour of poroelastic materials, Eur.
⎪ ( ) J. Mech. A Solids 83 (2020) 103996.
⎪ 𝐵 = 1∕3 𝑘𝛼 ∕𝑘𝑖𝑛 − 𝜇𝛼 ∕𝜇 𝑖𝑛
( √ )( )−3∕2 [13] X. Du, L. Jin, G. Ma, Macroscopic effective mechanical properties of porous dry
⎪ 𝜃 = 𝑟 arccos (𝑟) − 𝑟 −𝑟2 + 1 −𝑟2 + 1 , 0 ≤ 𝑟 ≤ 1.
⎩ concrete, Cem. Concr. Res. 44 (2013) 87–96.
[14] D. Li, Z. Li, C. Lv, G. Zhang, Y. Yin, A predictive model of the effective tensile
(23)
and compressive strengths of concrete considering porosity and pore size, Constr.
Build. Mater. 170 (2018) 520–526.
It is necessary to note that, in order to use (23) in Section 4, we have
[15] S. Torquato, Random Heterogeneous Materials. Interdisciplinary Applied
to replace superscript 𝑖𝑛 by ℎ , while 𝛼 =(𝑐,𝑝) for the two-phase model, Mathematics, Vol. 10, No. 1007, Springer-Verlag, New York, 2002.
and 𝛼 =(𝑐𝑚2,𝑝) for the three-phase model. [16] B. Pichler, C. Hellmich, J. Eberhardsteiner, Spherical and acicular representation
of hydrates in a micromechanical model for cement paste: prediction of early-age
elasticity and strength, Acta Mech. 203 (3) (2008) 137.
[17] W. Kreher, Residual stresses and stored elastic energy of composites and
Appendix B. Strain localization tensors of coated spherical inclu-
polycrystals, J. Mech. Phys. Solids 38 (1) (1990) 115–128.
sions in infinitive homogeneous domain [18] L. Dormieux, A. Molinari, D. Kondo, Micromechanical approach to the behavior
of poroelastic materials, J. Mech. Phys. Solids 50 (10) (2002) 2203–2231.
[19] M. Königsberger, M. Hlobil, B. Delsaute, S. Staquet, C. Hellmich, B. Pichler,
( )
⎧ 𝐴𝑠 = 3 𝑘ℎ +4 𝜇ℎ (3 𝑘𝑐𝑚 +4 𝜇𝑐𝑚 ) Hydrate failure in ITZ governs concrete strength: A micro-to-macro validated
⎪ 𝑎 (3 𝑘𝑐𝑚 +4 𝜇ℎ )(3 𝑘𝑎 +4 𝜇𝑐𝑚 )+12 𝑦3 (𝜇ℎ −𝜇𝑐𝑚 )(𝑘𝑐𝑚 −𝑘𝑎 ) engineering mechanics model, Cem. Concr. Res. 103 (2018) 77–94.
⎨ ( ℎ ) (24) [20] K. Miled, K. Sab, R. Le Roy, Effective elastic properties of porous materials:
3 𝑘 +4 𝜇ℎ (3 𝑘𝑎 +4 𝜇𝑐𝑚 )
⎪ 𝐴𝑠 = Homogenization schemes vs experimental data, Mech. Res. Commun. 38 (2)
⎩ 𝑐𝑚1 (3 𝑘𝑐𝑚 +4 𝜇 )(3 𝑘𝑎 +4 𝜇𝑐𝑚 )+12 𝑦3 (𝜇ℎ −𝜇𝑐𝑚 )(𝑘𝑐𝑚 −𝑘𝑎 )

(2011) 131–135.
[21] B.V. Tran, D.C. Pham, T.H.G. Nguyen, Equivalent-inclusion approach and
⎧ 𝑑 effective medium approximations for elastic moduli of compound-inclusion
(1−𝜈 ℎ )(1−𝜈𝑐𝑚 )𝑋0 (−4 (𝑋0 −1)(𝜂1 𝑦7 −𝜂2 (7−10 𝜈𝑐𝑚 ))+35 𝜂2 (1−𝜈𝑐𝑚 ))
⎪ 𝐴𝑎 = 225 𝛥 composites, Arch. Appl. Mech. 85 (12) (2015) 1983–1995.
⎨ 𝑑 (1−𝜈 ℎ )𝑋0 ((𝑋0 −1)(𝐴+60 𝑦3 (1−𝜈𝑐𝑚 )(𝜂1 𝑦7 −𝜂2 (7−10 𝜈𝑐𝑚 )))+35 (1−𝜈𝑐𝑚 )𝜂2 𝜂3 (1−𝑦3 )) [22] V. Marcadon, E. Herve, A. Zaoui, Micromechanical modeling of packing and
⎪ 𝐴𝑐𝑚1 = 15 (1−𝑦3 )𝛥 size effects in particulate composites, Int. J. Solids Struct. 44 (25-26) (2007)

8213–8228.
(25)

You might also like