Journal Pre-proof

Appropriate drying shrinkage prediction models for lightweight concrete containing

coarse agro-waste aggregate

Mehdi Maghfouri, Payam Shafigh, Vahid Alimohammadi, Yashar Doroudi,

Muhammad Aslam
PII: S2352-7102(19)30913-1
DOI: https://doi.org/10.1016/j.jobe.2019.101148
Reference: JOBE 101148

To appear in: Journal of Building Engineering

Received Date: 7 June 2019

Revised Date: 19 November 2019
Accepted Date: 23 December 2019

Please cite this article as: M. Maghfouri, P. Shafigh, V. Alimohammadi, Y. Doroudi, M. Aslam,
Appropriate drying shrinkage prediction models for lightweight concrete containing coarse agro-waste
aggregate, Journal of Building Engineering (2020), doi: https://doi.org/10.1016/j.jobe.2019.101148.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier Ltd.

 Appropriate drying Shrinkage Prediction Models for lightweight concrete

containing coarse agro-waste aggregate

Mehdi Maghfouri a∗, Payam Shafighb, , Vahid Alimohammadic, Yashar Doroudid, Muhammad Aslame

a
Department of Civil Engineering, Faculty of Engineering, University of Malaya, 50603 Kuala Lumpur,

Malaysia

b
Department of Building Surveying, Faculty of Built Environment, University of Malaya, 50603 Kuala

Lumpur, Malaysia

c
Young Researchers and Elite Club Azad University, Arak Branch, 3836119131 Arak, Iran

d
School of Civil Engineering, The University of Queensland, St Lucia, QLD 4074, Australia

e
Department of Civil Engineering, School of Engineering & Technology, Institute of Southern Punjab, Multan,

Pakistan

Abstract

Oil palm shell (OPS) agro-waste aggregate was successfully employed to produce

lightweight aggregate concretes. However, the high rate of drying shrinkage for this type of

concrete can impair the intended functionality of the structures or reduce service life. In this

study, drying shrinkage strain behaviour of normal-weight and lightweight aggregate

concretes was compared with prediction models. Twelve different mix proportions were

designed by using of crushed granite (NWA) and OPS aggregates as replacement of the

NWA at increasing intervals of 20%. Actual drying shrinkage at early and long-term ages

(275 days) was compared against theoretical results from drying shrinkage prediction models

such as ACI209R, EN1992, GL2000, B3 and SAK. Comparison between the test results and

predicted results indicated that the EN1992 was the most precise model at early-ages as well

∗
Corresponding author: Tel.: +60-111-4440-584
E-mail address: m.maghfouri@gmail.com (M. Maghfouri)

1
as GL2000 at long-term ages. In contrast, B3 was found the least precise model at both early

and long-term ages.

Keywords: Oil palm shell; Lightweight aggregate concrete; Drying shrinkage; Prediction

model

1. Introduction

Concrete as a composite material made from a mixture of Portland cement, sand, aggregate

and water, simulates the properties of rock. Shrinkage in concrete is defined as time-

dependent reduction in volume or length of the hardened concrete, which is mainly caused by

moisture transfer in concrete [1,2]. There are several types of shrinkage such as plastic,

chemical, autogenous, carbonation, thermal and drying shrinkage. Amongst the different

types of shrinkage mentioned, the drying shrinkage has high magnitude and remarkable

influence on durability of concrete and plays an important role for the structural member as it

would be harmful when it is restrained. Drying shrinkage in long term ages causes cracks,

which is one of the most objectionable type of defects that appear on the surface of concrete

members such as pavements, slabs and floors. However, this phenomenon is not critical when

concrete is utilized for insulation, filling purposes or non-structural panels [3]. The

development of micro-cracks and further crack propagation due to shrinkage and creep,

affects significantly on the durability of concrete structures [4, 5]. The magnitude of the

shrinkage depends on several factors such as quantity and type of binder and aggregates,

relative humidity, water to cement ratio, curing condition, member size and mixing method.

Although, the complexity of such factors together has not been well understood [6]. The

complexity of the long term drying shrinkage process including the pressure of disjunction,

capillary pressure, and the tension on the surface, as well as their impacts on the properties of

cement matrix makes it complicated to generate a prediction model [5]. More complex

2
shrinkage models do not necessarily increase the accuracy of the results comparing to the

simple models [7]. As it is not common to conduct long-term shrinkage test as a routine test

for every single concrete mixture, the structural engineer requires the prediction of time-

dependent strains of hardened concrete in order to assess the risk of concrete cracking as well

as deflection stripping-reshoring. Designers could only rely on models to predict the rate of

drying shrinkage during the design stage. Therefore, selection of the most reliable model to

estimate drying shrinkage for different type of concrete mixture is noteworthy.

Generally drying shrinkage value of the conventional concrete is ranging between 200 to 800

micro-strain [8] while it could be different for other types of concrete . Clarke [9] reported

the normal shrinkage rate for lightweight aggregate concrete comparing to the normal-weight

concrete (NWC) is 1.4 to 2 times higher. Generally, shrinkage of the concrete can be opposed

by the aggregates as it can provide enough restriction to reduce the contraction of cement

paste. Basically, lightweight aggregate concretes (LWAC) are containing less rigid, porous

aggregate and high volume of cement paste. The high cement content is applied mainly due

to the enhancement of mechanical properties of hardened concrete as well as modification of

the workability and stability of the mixtures [10, 11]. Therefore, lightweight concrete would

have greater long-term shrinkage value compared to the NWC.

Oil palm shell (OPS) is one of the agro waste material which can be replaced as coarse

aggregate for concrete mixtures. Previous studies [12, 13] highlighted that the OPS

lightweight aggregate concrete has good mechanical properties and durability performance.

However, its drying shrinkage rate is about five times higher than the NWC due to the lower

modulus of elasticity [14-16]. Shafigh et al. [17] found that high drying shrinkage of OPS

concrete is mainly due to high cement and OPS content as coarse aggregate. Therefore, a

practical solution to control and reduce the drying shrinkage strain is to optimise the volume

of coarse OPS aggregates in the concrete mixture.

3
In the present study, drying shrinkage of concretes containing different volume of OPS as

coarse aggregate (0-100%) was investigated. As no model has been defined for prediction of

drying shrinkage in agro-waste lightweight aggregate concretes, five mathematical drying

shrinkage models extracted from codes of practice and researches, such as ACI209R,

EN1992, GL2000, B3 and SAK were applied in order to comparison between experimental

and predicted results of drying shrinkage strain. The accuracy of the predicted results was

then evaluated by the error percentage and coefficient of variation methods.

Experimental Program

1.1. Materials

Ordinary Portland Cement (OPC) and Fly Ash (FA) were utilised as the blended binders in

this investigation. OPC met the requirements of ASTM C 150 [18] with the specific gravity

of 3.15 and Blaine surface area of 3710 cm2/g was used as well as locally available class F fly

ash (FA), which fulfilled the requirements in ASTM: C618 [19] with the specific gravity and

Blaine surface area of 2.18 and 7290 cm2/g, respectively. The chemical composition of

cementitious materials is shown in Table 1.

For the mixing of raw materials and curing of the concrete specimens, potable water was

used. In order to achieve better dispersion of cement particles, adequate workability and

proper compaction of fresh concrete, Polycarboxylate superplasticizer (SP) was added

together with mixing water.

The crushed granite and OPS with a maximum size of 12.5mm supplied by the local palm oil

plant were utilised as coarse aggregates. After collecting disordered fresh OPS aggregates

from palm oil mill, they were stored in an open area for approximately 7 months. After

drying process, the fibers on the surface of the OPS were removed. The next step was

washing the OPS aggregates using detergent powder to remove the oil and other impurities

4
from the OPS surface. Mining sand, provided from local resources with a maximum grain

size of 4.75 mm, the specific gravity of 2.67 and the fineness modulus of 2.89, was

considered as a fine aggregate. The physical properties of the fine and coarse aggregates are

presented in Table 2.

Table 1. Chemical compositions and LOI of cement (OPC) and fly ash (FA)

wt (%)
Oxide compositions
OPC FA
SiO2 19.80 64.60
CaO 63.40 3.67
Fe2O3 3.10 4.00
MgO 2.50 0.66
AL2O3 5.10 21.7
SO3 2.40 0.30
K 2O 1.00 1.20
Na2O 0.19 0.32
Loss on ignition (LOI) 1.80 5.10

Table 2. Physical properties of the aggregates

Coarse aggregates Fine aggregate

Properties
Granite OPS Mining sand

Specific gravity 2.65 1.20 2.67

Compacted bulk density (kg/m3) 1490 610 1660

24-h water absorption (%) <1 20.5 1.2

1.2. Concrete mix proportions and testing methods

Twelve different mix proportions in two groups with the same volume of binder and

aggregate were designed using crushed granite (NWA) and OPS aggregates. In both groups,

the NWC with only granite aggregate was considered as a control mix. Partially replacement

of the normal-weight aggregate for the conventional concrete at an interval of 20% was

applied for both groups. The type of the binder was the only difference between group (1)

and group (2). Class F fly ash with a dosage of 25% by weight of binder was used for group

(2). Generally, evaluating the effect of the Fly ash on the mechanical properties, workability,

5
setting time and durability of concrete can lead to determine the optimum level of the Fly ash.

According to ACI 211.1-91 [20], the recommended substation level of class F fly ash is

ranging from 15 to 25% of total cementitious material. The designed mix proportions, slump,

density and 28-day compressive strength for all concrete mixes are shown in Table 3.

Table 3. Concrete mix proportions

Coarse
Binders
Mix Water SP Sand aggregate Slump Density
(kg)
Groups
ID (litre) (%cement) (kg/m3) (kg/m3) (mm) (kg/m3)
fcu*
OPC FA Granite OPS
C1 898 0 205 2340 74.4
C2 715 81 130 2205 64.2
C3 540 163 90 2110 54.4
Group (1)
C4 480 0 163 1 819 359 244 70 2015 50.1
C5 179 325 65 1990 44.8
C6 0 406 40 1900 40.5
F1 898 0 245 2255 59.1
F2 715 81 110 2180 51.3
F3 540 163 110 2100 48.8
Group (2)
F4 360 120 162 1 819 359 244 105 1990 40.5
F5 179 325 125 1815 35.5
F6 0 406 90 1715 31.2
*fcu is the 28-day compressive strength in MPa

For all concrete mixes in groups (1) and (2), the ordinary Portland cement, class F fly ash,

crushed granite and the pre-soaked saturated surface dried OPS aggregates were blended in

the revolving drum mixer. Dry mixing of the raw materials was continued for 2-mintues and

followed by adding 70% of the mixing water containing SP. The remainder of water was

added to the mixture for another 5-minute mixing. Then, the slump test for fresh concrete was

carried out prior to the concrete sampling. The slump test results presented in Table 3.

For each concrete mix, two 100×100×300mm prisms were prepared and kept in the

laboratory environment for 24 hours. After demoulding and proper curing, the pre-drilled

stainless-steel discs namely DEMEC points (Fig.1) were bonded with adhesive on three sides

of the prisms with the spacing of 200 mm.

6
 Fig. 1. Pre-drilled stainless steel discs (DEMEC pints)

Drying shrinkage strains of the concrete prisms were precisely measured by the digital

version of the DEMEC Mechanical Strain Gauge, which was calibrated to measure based on

micro-strain. As shown in Fig.2, the DEMEC gauge consist of an invar main beam with two

conical locating points, one fixed and the other pivoting on a special accurate knife edge. The

gauge points locate in the DEMEC points.

Fig. 2. DEMEC Mechanical Strain Gauge

1.3. Curing condition

For all drying shrinkage prisms, 7 days’ water curing was applied. After proper curing, the

specimens were placed in the laboratory environment for shrinkage measurement in different

ages. However, in order to measure the compressive strength of concrete mixes, 100mm

cubes were immersed in water for 28 days. The average temperature of the curing water was

7
22 ± 2˚C, whereas the temperature in the lab environment with a relative humidity of about

74 % was 30 ± 2˚C.

3.0 Results and Discussions

3.1 Drying shrinkage prediction models

In the construction industry, the prediction of the time-dependent strain for the hardened

concrete, is an approach for structural engineers to predict the serviceability of the concrete

structure and assess the risk of deflection and cracking. Based on the experimental results of

drying shrinkage and relevant theories, several mathematical models were proposed and

added to the standards and codes of practices for prediction of the drying shrinkage strain.

Generally, they were designed on the basis of two main factors namely, mathematical form of

the model relating to the time, and fitting of the parameters and resulting expressions. In this

study, five prediction models, ACI-209R, EN1992 (EC2), Gardner and Lockman (GL2000),

Bazant and Baweja (B3) and Sakata (SAK) were applied to estimate shrinkage behaviour of

the normal-weight, and low-density concretes. The selected set of values considered for the

prediction of drying shrinkage strain of the different concretes is shown in Table 4.

Furthermore, the summary of different factors and parameters considered by the selected

prediction models are presented in Table 5.

Table 4. Selected parameters used for prediction models in this study

Factors Data
Age of concrete 7 days
Relative humidity (%) 70.0 to 76.0
Compressive strength (28-day, MPa) 30.0 to 75.0
Type of cement Normal OPC
3
Cement content (kg/m ) 400 – 550

8
 Slump (mm) 40 to 245
Dimension of specimen (mm) 100 × 100 × 300

Table 5. Input factors for predicting shrinkage

Prediction models for drying shrinkage

Factors
ACI-209R EN1992 GL2000 B3 SAK
Compressive strength (MPa) X X X X X
Curing condition X - - X -
Relative humidity X X X X X
Volume to surface ratio (v/s) X - X X X
Concrete slump X - - - -
Fine agg. to total aggregate (Af /A) X - - - -
Cement type X X X X X
Water - - - - X
Air content X - - - -
Cross-section - X X X X
Environmental - - - - -
Lightweight concrete - X - - -

3.1.1 ACI-209R shrinkage model

The ACI 209R [21] is an empirical model which has been using by designers since 1971 for

the prediction of drying shrinkage strains for lightweight and normal-weight concretes with

low to moderate strength, under controlled environmental conditions. Furthermore, this

model does not predict the shrinkage phenomena and any specific coefficient to penalize the

long-term shrinkage of lightweight concrete. According to ACI-209R, the shrinkage strain as

S (t,tc) is calculated as shown in Equations 1 to 4. Where, t (days) is the age of concrete at the

time of shrinkage, tc (days) the age of concrete at the beginning of drying, and S∞ is the

ultimate shrinkage.

( )
( , ) = × S∞ ……………………………………………………..………………..
( )

(1)

. ∗
f = 26.0 × ….……………………………….………………………......

(2)

9
S∞ = 780 × 10-6 × (Ƴsh) …………….………………………………………...........................

(3)

Ƴsh = Ƴtc × ƳRH × Ƴvs × Ƴs × Ƴψ × Ƴc × Ƴα.………. ………………………………………….

(4)

An average value for f of 35 under 7 days moist curing condition and 55 for up to 3 day steam

curing is recommended by ACI, while f also can be computed from Eq. (2). The volume-

surface ratio is represented as V/S. The Ƴsh represents the product of seven applicable

correction factors that take into consideration as follow;

Ƴtc = Curing time coefficient

ƳRH = Relative humidity coefficient

Ƴvs = Factor depends on volume-surface ratio

Ƴs = Slump factor (mm)

Ƴψ = Ratio of fine aggregate to the total aggregates

Ƴc = Cement content in kg/m3

Ƴα = Air content (%).

As shown in table 5, the ACI 209R [21] prediction model almost covers all the effective

factors on concrete shrinkage. The ACI model was individually applied to each mix to

estimate the shrinkage strain, although each mix has different compressive strength at 28-day.

The comparison of the experimental development of drying shrinkage strain and the ACI

predicted values for concrete mixes in group (1) are shown in Fig. 3. Also table 6 depicts a

comparison between drying shrinkage strain of experimental and predicted results at early

ages.

10
It was found that all the mixes in group (1) gained sharp shrinkage strain during the first

week of drying. However, the increase rate of shrinkage was lower in 14 days for C1 and C2

mixes. The difference between the predicted and experimental results for conventional

concrete (C1) was found about 66%, 50% and 22% for 3-day, 7-day and 14-day, respectively.

Furthermore, it was observed that by increasing the ratio of OPS in NWC, the drying

shrinkage strain sharply raised. The sharpest increment in shrinkage strain and the remarkable

difference between the predicted and experimental test result was recorded for C6 mix.

At early ages, the model showed under-estimated shrinkage value by increasing the drying

period. At later-ages, the ACI model presented a similar trend to the C1 and C2 mixes. It

over-estimated (about 17%) for C1 mix, and under-estimated (about 5%) for C2 mix.

However, for all the remaining mixes, the ACI model predicted lower results. After 9 months

of drying, for the C3, C4, C5 and C6 mixes, the model predicted lower values of shrinkage by

about 26%, 39%, 54%, and 60%, respectively. This shows that incorporation of OPS in

conventional concrete significantly increases the rate of shrinkage development at early-ages,

which significantly resulted in higher drying shrinkage strain at long-term ages. Therefore, it

could be concluded that this type of aggregate cannot be utilised as total coarse aggregate in

concrete mixtures due to the fast increment of drying shrinkage in concrete containing high

volume of OPS. Maghfouri et al. [22] reported the OPS content as coarse aggregate in

concrete based on mechanical and durability properties. It was recommended to be limited to

60% of total volume of the coarse aggregate in the concrete mixture.

Table 6. Early-age measured and ACI predicted drying shrinkage strains

Mix Experiment ACI predicted results *

Groups
code 3-day 7-day 14-day 3-day 7-day 14-day
C1 60 89 102 21 (-66%) 45 (-50%) 80 (-22%)
C2 64 96 115 21 (-68%) 45 (-53%) 80 (-30%)
C3 85 106 166 21 (-76%) 45 (-58%) 80 (-52%)
Group (1)
C4 108 130 160 21 (-81%) 45 (-66%) 80 (-50%)
C5 125 220 286 21 (-84%) 45 (-80%) 80 (-72%)
C6 141 301 335 21 (-85%) 45 (-85%) 80 (-76%)
Group (2) F1 55 98 135 27 (-51%) 58 (-41%) 102 (-25%)

11
 F2 76 115 158 27 (-64%) 58 (-50%) 102 (-36%)
F3 93 119 184 27 (-71%) 58 (-51%) 102 (-45%)
F4 102 144 199 27 (-73%) 58 (-60%) 102 (-49%)
F5 120 151 222 27 (-77%) 58 (-62%) 102 (-54%)
F6 161 249 386 27 (-83%) 58 (-77%) 102 (-74%)
* Value in parenthesis presents ratio of the predicted results to the experimental values.
(-) The model under-estimated the drying shrinkage result
(+) The model over-estimated the drying shrinkage result
900

800

700
Drying Shrinkage (microstrain)

600

500

400

300

200 C1-7D C2-7D

C3-7D C4-7D
100 C5-7D C6-7D
Avg. ACI
0
0 50 100 150 200 250 300
Age (days)

Fig. 3. Development of the drying shrinkage for group (1) mixes with ACI-209R [21] model

Similar to the group (1), the comparison of the drying shrinkage strain development of the

experimental result and the ACI predicted values for group (2) is presented in Fig. 4. The

ACI model was individually applied to each mix to estimate the shrinkage strain. Despite the

different 28-day compressive strength for each mix, almost similar results were predicted by

ACI for each mix with the average difference between 3-5%. Therefore, the average values

for the ACI model were considered and indicated in Fig. 4. The comparison between

experimental and predicted results at early-ages is shown in Table 6. It was observed that, the

increase rate in shrinkage for group (2) mixes (containing fly ash) was higher than group (1)

mixes.

For the conventional concrete (C1), drying shrinkage strain was recorded about 9% lower

compared to the same mix containing fly ash (F1) at 7-day age. However, for both mixes, the

ACI model predicted different results because of different compressive strengths. In general,

12
it was found that as the substitution of OPS increased, sharper drying shrinkage strain was

observed in group (2) concretes. F3 mix at 14 days’ age showed about 45% higher drying

shrinkage strain compared to the predicted values. Mix F6 showed a sharp increase in

shrinkage strain, which is almost similar to the C6 mix with about 75% average difference

between the experimental and predicted results.

In general, at early-ages it was observed that with increasing of drying, the predicted result

was lower than actual shrinkage result for all mixes in group (2) as shown in Fig. 4. In later-

ages of about 9 months of drying, predicted values of the ACI model were approximately

12%, 25%, 30%, 42%, 55%, and 62% smaller than the experimental values for F1, F2, F3,

F4, F5, and F6 mixes, respectively. In addition, the group (2) mixes showed higher

experimental drying shrinkage strains compared to group (1) mixes. The test result of the

mixes in group (2) show that the addition of OPS in fly ash concrete has a significant effect

on the rate of shrinkage development in early ages for all mixes, which also resulted in higher

drying shrinkage strain at long-term ages.

1000

900

800
Drying Shrinkage (microstrain)

700

600

500

400

300
F1-7D F2-7D
200
F3-7D F4-7D
100 F5-7D F6-7D
Avg. ACI
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 4. Development of the drying shrinkage for group (2) mixes with ACI-209R [21] model

3.1.2 Eurocode (EC2) Drying Shrinkage Model

13
It has been specified by European Standards [23] that the drying shrinkage of the low-density

concretes can be estimated by some expressions defined for NWC. In this model, the final

drying shrinkage of lightweight concrete is modified by an observational factor of 1.2. As

represented in Eqs. 5 to 9 total shrinkage considered by EN1992 prediction model is

composed of two components, the drying shrinkage strain and the autogenous shrinkage

strain[23]. Autogenous shrinkage in lightweight concrete with pre-soaked aggregates was

assumed considerably smaller than NWC.

ABC = ABD + ABF …………………………………………………………………………….

(5)

Where, ABC is the total shrinkage, ABD and ABF show the autogenous shrinkage and the drying

shrinkage, respectively. The estimated model of drying shrinkage is also measured as follow:

ABD = GDC (H, HC ). IJ . ABD, ….…….……………………………………………………….….(6)

RST .
ABD, = 0.85. GMN . (220 + 110. PDC ). exp ( U
) ………….…………………………......(7)

( T)
GDC (H, HC ) = ...…………………………………………………………......(8)
( T) . VJWX

ABC (H, HC ) = YZ ∗ ABD ……….………………………………………………………………. (9)

Where,

GDC (H, HC ) = The age of experimental samples,

IJ = Coefficient depending on the notional size (h0)

h0 = Notional dimension of the concrete member (mm)

T = Age of concrete (days),

ts = Age of concrete at the beginning of drying (days),

For prediction of drying shrinkage by using Eurocode model, the selected parameters are

presented in Table 5.

14
The development of the experimental drying shrinkage strain and the EC2 prediction model

results for mixes in group (1) are shown in Fig. 5. Similar to ACI model, the EC2 model was

applied individually to each mix to estimate the shrinkage strain. Although each mix had a

different 28-day compressive strength, almost similar results were predicted by EC2. The

average values for the EC2 model were indicated in Fig. 5. All the experimental drying

shrinkage results were observed and studied at early and long-term ages. Sharp increase in

drying shrinkage was recorded for all the mixes at early-ages as shown in Table 7. The EC2

showed a similar trend to the experimental result for C1 mix at 3 days. It was also observed

that by increasing the drying age of the concrete specimens from 3 to 14 days, the difference

between the actual and the predicted results rised. For instance, the predicted results for C2

mix at the ages of 3 and 7 days showed the average difference of about 7% and 18%,

respectively. While at the age of 14 days, the EC2 significantly overestimated the values with

the average difference of about 48%. In general, it was found that the EC2 model is not

giving appropriate results for C1 and C2 mixes at 14 days. Nonetheless, as the substitution

ratio with OPS increased, the reliable results were achieved for C3 and C4 mixes. The

average difference between the actual and predicted results for C3 mix at 7 and 14 days were

about 7% and 2%, respectively. Similarly, for C4 mix the difference was found about 13%

and 6% at 7, and 14 days, respectively. Furthermore, the C5 and C6 mixes experienced the

sharpest increase in shrinkage strain at early-ages.

Fig. 5 shows the development trend of actual drying shrinkage and predicted results for all

mixes. At later-ages up to about 2 months of drying, the EC2 model presented a similar trend

to the C3 and C4 mixes. The same trend also was observed for C2 mix after 3 months. At 9

months of drying, the model showed about 17% higher shrinkage strain for C1 mix, whereas

at same age of drying, predicted values for C3, C4, C5 and C6 mixes were approximately

25%, 38%, 54%, and 60% lower than the experimental values, respectively. It clearly shows

15
that the incorporation of OPS in conventional concrete significantly increases the rate of

shrinkage development at early and long-term ages.

Table 7. Early-age difference between experimental and EC2 predicted drying shrinkage strains

Experimental drying
Mix EC2 predicted results *
Groups shrinkage strain
code
3-day 7-day 14-day 3-day 7-day 14-day
C1 60 89 102 60 (00%) 113 (+27%) 170 (+67%)
C2 64 96 115 60 (-07%) 113 (+18%) 170 (+48%)
C3 85 106 166 60 (-30%) 113 (+07%) 170 (+02%)
Group (1)
C4 108 130 160 60 (-45%) 113 (-13%) 170 (+06%)
C5 125 220 286 60 (-52%) 113 (-49%) 170 (-41%)
C6 141 301 335 60 (-58%) 113 (-62%) 170 (-49%)
F1 55 98 135 67 (-22%) 127 (+29%) 190 (+41%)
F2 76 115 158 67 (-12%) 127 (+10%) 190 (+20%)
F3 93 119 184 67 (-28%) 127 (+06%) 190 (+03%)
Group (2)
F4 102 144 199 67 (-34%) 127 (-12%) 190 (+04%)
F5 120 151 222 67 (-44%) 127 (-16%) 190 (+14%)
F6 161 249 386 67 (-58%) 127 (-49%) 190 (-51%)
* Value in parenthesis presents ratio of the predicted results to the experimental values.
(-) The model under-estimated the drying shrinkage result
(+) The model over-estimated the drying shrinkage result
900

800

700
Drying Shrinkage (microstrain)

600

500

400

300

200
C1-7D C2-7D
C3-7D C4-7D
100
C5-7D C6-7D
Avg. EC2
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 5. Development of the drying shrinkage for group (1) mixes with EC2 [24] model

The development of experimental drying shrinkage strain and the average EC2 predicted

values for mixes in group (2) is presented in Fig. 6. Compared to the group (1), all mixes

containing fly ash in group (2), gained sharp shrinkage strain at the first week of drying.

Meanwhile, as can be seen in Table 7, the rate of increase in shrinkage for group (2) mixes

16
was higher compared to mixes in group (1). Consistent increase in shrinkage was also

observed for group (2) mixes compared to group (1) at early-ages. The F1 mix showed about

24% higher drying shrinkage at 14 days compared to C1 mix. However, for the same mix,

EC2 predicted significantly higher results of about 41%, which was mainly because of sharp

development of shrinkage strain in conventional concrete due to the addition of fly ash.

Furthermore, consistent increase in drying shrinkage was observed for F2 to F5 mixes as the

contribution level of OPS aggregates increased from 20 to 80% with the same amount of fly

ash content. The predicted results for F6 mix were found significantly lower due to its higher

shrinkage strain at early- and long-term ages.

The comparison of experimental and the EC2 predicted results for group (2) mixes at long-

term ages is shown in Fig. 6. At later-ages of about 9 months of drying, the EC2 model

underestimated the shrinkage values compared to the experimental results with the

percentages of about 11%, 24%, 29%, 41%, 54%, and 62% for F1, F2, F3, F4, F5, and F6

mixes, respectively. This might be due to a sharp increase in drying shrinkage at early-ages

for all mixes due to the addition of OPS and fly ash in NWC, which significantly showed

higher long-term shrinkage strains.

1000

900

800
Drying Shrinkage (microstrain)

700

600

500

400

300
F1-7D F2-7D
200
F3-7D F4-7D
100 F5-7D F6-7D
Avg. EC2
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 6. Development of the drying shrinkage for group (2) mixes with EC2 [24] model

17
3.1.3 Gardner and Lockman (GL-2000) model

In 2001, Gardner and Lockman [25] proposed a model for prediction of drying shrinkage

which was based on the modification of Gardner and Zhao (GZ) model. The model is called

GL2000 which is suitable to apply for conventional concrete having water/cement (W/C) in

the range of 0.4 to 0.6 with 28 days compressive strength up to 82 MPa. The effect of

additives, chemical and mineral admixtures, curing method and casting temperature are not

taken into consideration for this model.

The detailed expressions of GL2000 model are presented below in equations 10 and 11.
/
( W)
ACJ (H, H ) = ACJ[ ∗ (1 − 1.18 ∗ ℎ ) ∗ ^( c …………………………………….…. (10)
W) . _∗(`ba)

f
Z
ACJ[ = 1000 ∗ d ∗ e ∗ 10 g
……………………………………………….…………....…. (11)

Where,

ACJ = The shrinkage strain

ACJ[ = Ultimate shrinkage strain

t = Age of concrete (days),

t0 = Age of concrete at the beginning of drying (days),

h = Relative humidity

V = Volume of the concrete specimen (mm3)

S = Concrete specimen surface (mm2)

K = The cement type factor

hBi = Compressive strength at the age of 28-day (MPa)

For prediction of drying shrinkage by using GL2000 model, the selected parameters are

presented in table 5.

Fig. 7 presents the comparative trends for the experimental and the average GL2000

predicted results of drying shrinkage strain for group (1) mixes. At early-ages, it was found

18
that the GL2000 model predicted significantly higher results for C1 and C2 mixes as can be

seen in Table 8. This was mainly due to a sharp prediction from the GL model. Furthermore,

the over-estimated prediction results for C3 and C4 mixes were found much closer to the

experimental results at 14 days’ ages with the average difference of about 15%. For C5 and

C6 mixes, the GL model under-estimated the results at all ages 3, 7 and 14 days with the

increment percentages of about 30%, 46%, and 39%, respectively. For all mixes in group (1)

it was observed that prediction results for minimum substitution of OPS (C1 and C2 mixes)

were significantly higher than the experimental results. Whereas, the mixes containing

medium range of OPS substitution showed closer results. The prediction results were found

significantly lower for mixes containing a high volume of OPS aggregates.

Fig. 7 shows the development trend of drying shrinkage experimental and the predicted

results for all mixes. Up to 2 months drying, the trend of GL2000 model was found much

close to C3 and C4 mixes. However, at the same age, the experimental results for C1 and C2

mixes were found significantly lower. At later-ages, the GL model trend followed the trend of

C3 mix. In general, at about 9 months of drying, the GL2000 model showed about 34% and

16% higher shrinkage strain for C1 and C2, respectively. While at the same age of drying, the

predicted values were approximately 6%, 23%, 42%, and 49% lower than the experimental

values for the C3, C4, C5 and C6 mixes, respectively.

Table 8. Early-age difference between experimental and GL2000 predicted drying shrinkage strains

Experimental drying
Mix GL2000 predictions *
Groups shrinkage strain
code
3-day 7-day 14-day 3-day 7-day 14-day
C1 60 89 102 93 (+55%) 139 (+56%) 187 (+84%)
C2 64 96 115 93 (+46%) 139 (+44%) 187 (+63%)
C3 85 106 166 93 (+10%) 139 (+31%) 187 (+13%)
Group (1)
C4 108 130 160 93 (-14%) 139 (+07%) 187 (+17%)
C5 125 220 286 93 (-25%) 139 (-37%) 187 (-34%)
C6 141 301 335 93 (-34%) 139 (-54%) 187 (-44%)
F1 55 98 135 118 (+114%) 175 (+79%) 237 (+75%)
Group (2)
F2 76 115 158 118 (+55%) 175 (+52%) 237 (+50%)

19
 F3 93 119 184 118 (+27%) 175 (+47%) 237 (+29%)
F4 102 144 199 118 (+15%) 175 (+21%) 237 (+19%)
F5 120 151 222 118 (-02%) 175 (+16%) 237 (+07%)
F6 161 249 386 118 (-27%) 175 (-30%) 237 (-39%)
* Value in parenthesis presents ratio of the predicted results to the experimental values
(-) The model under-estimated the drying shrinkage result
(+) The model over-estimated the drying shrinkage result

900

800

700
Drying Shrinkage (microstrain)

600

500

400

300

200 C1-7D C2-7D

C3-7D C4-7D
100 C5-7D C6-7D
Avg. GL2000
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 7. Development of the drying shrinkage for group (1) mixes with GL2000 [25] model

Table 8 shows the development of shrinkage strain at early-ages for mixes in group (1) and

(2). It was found that the GL2000 model predicted significantly higher results for C1 and C2

mixes, similar to F1 and F2 mixes. Furthermore, the over-estimated prediction results for F5

mix was found much closer to the experimental results at 14 days’ ages with the difference of

about 7%. For C6 and F6 mixes, the GL model under-estimated the results at all ages 3, 7 and

14 days.

The development of the drying shrinkage strain and the theoretical results for group (2) mixes

is shown in Fig. 8. At first 60 days of drying, for F1 to F4 mixes almost similar results of

predicted and experimental shrinkage values were observed. At later-ages, F4 mix gained a

sharp increase in shrinkage, while F1 mix experienced lower shrinkage strain compared to the

predicted model. The difference between the experimental and predicted results for F1 mix

was found about 21%. At 9 months of drying, the GL2000 model showed similar results for

20
F2 and F3 mixes. Whereas, at the same age of drying, the predicted values were

approximately 16%, 35%, and 46% lower than the experimental values for the F4, F5 and F6

mixes, respectively.

1000

900

800
Drying Shrinkage (microstrain)

700

600

500

400

300

200 F1-7D F2-7D

F3-7D F4-7D
100 F5-7D F6-7D
Avg. GL2000
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 8. Development of the drying shrinkage for group (2) mixes with GL2000 [25] model

3.1.4 Bazant and Baweja (B3) shrinkage model

B3 model which is based on a mathematical description of several physical phenomena

related to time-dependent creep and shrinkage, was developed by Bazant and Baweja [26].

This model was calibrated by a computerized data bank comprising practically all the

relevant test data obtained in various laboratories throughout the world. It was stated by

Bazant and Baweja [26] that B3 model has a very low coefficient of variations compared to

the ACI209R and EN1992 model. The model applies the compliance function which reduces

the risk of errors due to inaccurate values related to the modulus of elasticity. The proposed

prediction equations for the drying shrinkage strain are as follow:

ACJ (H, H ) = −ACJj ∗ IJ ∗ tan ℎ V k W

……………………………………………………(12)
Tl

sWt
( )f/
ACJj = −P ∗ P ∗ m0.00856 ∗ n .
∗ (145 ∗ hBi ) . p
+ 270r ∗ uvW.wx∗sWt
yW vzTl ……...(13)
( )f/
uvW.wx∗(yW vzTl )

21
where,

ACJ = Drying shrinkage strain,

ACJj = Ultimate shrinkage strain,

t = Age of concrete (days),

t0 = Age of concrete at the beginning of drying (days),

α1 = Cement type factor,

α2 = Curing factor,

w = Water content (kg/m3),

hBi = 28-day compressive strength (MPa),

IJ = Humidity dependent factor,

{CJ = Size and shape dependent factor.

The selected parameters for prediction of drying shrinkage using B3 model are presented in

Table 5.

The comparison between experimental and B3 analytical model results at early-ages is shown

in Table 9. This model significantly under-estimated the results for all the mixes in both

groups at early-ages. The development of shrinkage strain comparisons at long-term ages are

shown in Figs. 9 and 10. For both groups, not a suitable correlation between the experimental

and predicted results was observed. This might be due to the factors which have been covered

by this model. Therefore, this model cannot be considered as a designed model for the mixes

prepared in this study.

22
 Table 9. Early-age difference between experimental and B3 predicted drying shrinkage strains

Experimental drying
Mix B3 predicted results *
Groups shrinkage strain
code
3-day 7-day 14-day 3-day 7-day 14-day
C1 60 89 102 32 (-47%) 48 (-46%) 66 (-35%)
C2 64 96 115 32 (-50%) 48 (-50%) 66 (-43%)
C3 85 106 166 32 (-63%) 48 (-55%) 66 (-60%)
Group (1)
C4 108 130 160 32 (-71%) 48 (-63%) 66 (-59%)
C5 125 220 286 32 (-75%) 48 (-78%) 66 (-77%)
C6 141 301 335 32 (-77%) 48 (-84%) 66 (-80%)
F1 55 98 135 30 (-45%) 46 (-53%) 63 (-53%)
F2 76 115 158 30 (-60%) 46 (-60%) 63 (-60%)
F3 93 119 184 30 (-67%) 46 (-62%) 63 (-66%)
Group (2)
F4 102 144 199 30 (-70%) 46 (-68%) 63 (-68%)
F5 120 151 222 30 (-75%) 46 (-70%) 63 (-71%)
F6 161 249 386 30 (-81%) 46 (-82%) 63 (-84%)
* Value in parenthesis presents ratio of the predicted results to the experimental values
(-) The model under-estimated the drying shrinkage result
(+) The model over-estimated the drying shrinkage result

900
C1-7D
800 C2-7D
C3-7D
700
Drying Shrinkage (microstrain)

C4-7D
600 C5-7D

500 C6-7D
Avg. B3
400

300

200

100

0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 9. Development of the drying shrinkage for group (1) mixes with B3 [26] model

23
 1000
F1-7D
900 F2-7D
800 F3-7D
F4-7D
Drying Shrinkage (microstrain)

700
F5-7D
600 F6-7D
Avg. B3
500

400

300

200

100

0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 10. Development of the drying shrinkage for group (2) mixes with B3 [26] model

3.1.5 Sakata (SAK) shrinkage prediction model

Since the early 1980s Sakata has been working on a research project to develop a prediction

model based on a statistical method from many experimental data. He proposed new

prediction equations of creep and shrinkage of concrete which was published in 1996 by

Japan society of civil engineers in the standard specification for design and construction of

the concrete structure. The 1996 revision of the JSCE specification was noteworthy as for the

first time it was presenting an original Japanese shrinkage model. For this model, the creep

and shrinkage tests were carried out corresponding to the actual and controlled ambient

conditions [27, 28]. The effect of temperature during casting of concrete also was taken into

consideration. The detailed expressions of the model are presented as follow:

ɛTl} ∗( W)
ACJ (H, H ) = ~ (
..…………………………………………………………………..(14)
W)

ɛTl•
ACJj = ..…………….……………………………………………………………..(15)
€∗ W

R ( J)‚
ACJ• = xWW
…….…………………………………………………………….(16)
_ ƒ„… † ˆ
‡e ( w)

where,

24
ACJ = Final value of shrinkage strain,

ACJj = Ultimate shrinkage strain,

t = Age of concrete (days),

t0 = Age of concrete at the beginning of drying (days),

α = Cement type factor,

h = Relative humidity

W = Water content (kg/m3),

hBi = 28-day compressive strength (MPa),

The selected parameters for prediction of the drying shrinkage using SAK model are

summarised in Table 5.

The early-age differences between the results are highlighted in Table 10. For conventional

concrete (C1), the SAK model predicted closer results at 7 day age while the difference was

higher for 14 days. However, the model showed reliable results for mixes containing 20 to

60% OPS (C2 to C4), with the average difference in the range of 15 to 21%. Furthermore,

due to a sharp gain of shrinkage for C5 and C6 mixes, the SAK model predicted significantly

lower results with the average difference of about 58%.

Fig. 11 shows the development trend of drying shrinkage experimental and SAK predicted

results for all mixes in group (1). At later-ages, the SAK model showed a similar trend to the

experimental values of C2, C3 and C4 mixes up to 30 days, whilst the C4 mix due to sharper

gain in shrinkage left the trend. At about 9 months of drying, the SAK model showed about

28% and 10% higher shrinkage strain for C1 and C2 mixes, respectively, whilst, at same age

of drying, the predicted values for C3, C4, C5 and C6 mixes were approximately 13%, 28%,

47%, and 53%, respectively, lower than the experimental values.

25
 Table 10. Early-age difference between experimental and SAK model drying shrinkage strains

Experimental drying
Mix SAK predicted results *
Groups shrinkage strain
code
3-day 7-day 14-day 3-day 7-day 14-day
C1 60 89 102 38 (-37%) 78 (+12%) 132 (+29%)
C2 64 96 115 38 (-41%) 78 (+18%) 132 (+15%)
C3 85 106 166 38 (-56%) 78 (-26%) 132 (-21%)
Group (1)
C4 108 130 160 38 (-65%) 78 (-40%) 132 (-18%)
C5 125 220 286 38 (-70%) 78 (-64%) 132 (-54%)
C6 141 301 335 38 (-73%) 78 (-74%) 132 (-61%)
F1 55 98 135 39 (-29%) 82 (-17%) 138 (+02%)
F2 76 115 158 39 (-49%) 82 (-29%) 138 (-13%)
F3 93 119 184 39 (-58%) 82 (-31%) 138 (-25%)
Group (2)
F4 102 144 199 39 (-62%) 82 (-43%) 138 (-31%)
F5 120 151 222 39 (-67%) 82 (-46%) 138 (-38%)
F6 161 249 386 39 (-76%) 82 (-67%) 138 (-64%)
* Value in parenthesis presents ratio of the predicted results to the experimental values
(-) The model under-estimated the drying shrinkage result
(+) The model over-estimated the drying shrinkage result

At early ages for F1 mix, the predicted results were found almost similar to the experimental

result at 14 days’ age, while the 3 and 7 days results were under-estimated by the model.

Similar behaviour was also found for F2 mix, the average difference between the results was

about 13% at 14 days’ age. Furthermore, as the substitution of OPS increased beyond 20% in

the mixes, the average difference between the experimental and predicted values was also

increased. Due to the sharp gain of shrinkage for F6 mix, the SAK model predicted

remarkably lower results at 14 days’ age with the difference of about 64%.

Fig. 12 shows the development trend of the experimental drying shrinkage and SAK

predicted results for all mixes in group (2). At later-ages, the SAK model presented a similar

trend to the experimental values of F1 to F3 mixes up to 30 days. At later ages, the trend of

the model was found closer for F1 mix. At about 9 months of drying, the model predicted

lower results than the experimental values with the differences of about 18%, 23%, 36%,

51%, and 59% for F2, F3, F4, F5, and F6 mixes respectively, lower than the experimental

values.

26
 900

800

700
Drying Shrinkage (microstrain)

600

500

400

300

200
C1-7D C2-7D
C3-7D C4-7D
100
C5-7D C6-7D
Avg. SAK
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 11. Development of the drying shrinkage for group (1) mixes with SAK [27] model

1000

900

800
Drying Shrinkage (microstrain)

700

600

500

400

300

200 F1-7D F2-7D

F3-7D F4-7D
100 F5-7D F6-7D
Avg. SAK
0
0 30 60 90 120 150 180 210 240 270 300
Age (days)

Fig. 12. Development of the drying shrinkage for group (2) mixes with SAK [27] model

3.2.The accuracy of the prediction models

The drying shrinkage developments for the conventional concrete containing OPS as coarse

aggregate (group (1)) and the same mixes with the replacement of cement with fly ash (group

(2)) were compared with five prediction models, ACI209R, EN1992, GL2000, B3 and SAK

as shown in Figs. 3 to 12. Generally, it was observed that most of the prediction models

showed much lower results compared to the experimental values, which was mainly due to a

sharp gain of drying shrinkage strain for the mixes containing OPS aggregates.

27
Two analysis namely error percentage (EP) and coefficient of variation (CV) were carried out

to evaluate the accuracy of the prediction models that applied for this study. For EP analysis,

the residual values were taken into the calculation at the early age (14-day) and the long-term

age (275-day) by considering the respective equation (Eq. 17).

Residual value = Predicted shrinkage value - Experimental shrinkage value………………

(17)

From the calculations, the positive or negative residual values show that the model

overestimates or underestimates the data respectively, compared to the experimental data

[29]. A model with a minimum average of error percentage is the best predictor in the EP

analysis while for the CV is the model with the lowest value.

The error percentage analysis and the coefficient of variation were calculated as follows:

EP = (Residual value*100)/Experimental shrinkage value …...……….….……………….

(18)

CV = Standard deviation / Mean ...……………………………………………………...….

(19)

The summarised results of the EP and the CV for the mixes in both groups at the age of 14-

day are indicated in Table 11 and Table 12, respectively. In group (1) mixes, for C1 concrete

mixture, according to EP and CV, the ACI209R model was the best predictor for the drying

shrinkage, although, it underestimates the results with the average error percentage of about

22%. Rank ascending to the remaining models for the conventional concrete (C1) was found

as SAK, B3, EC2 and GL2000. Furthermore, the SAK model was found as better predictor

for C2 concrete mixture, followed by ACI209, B3, EC2 and GL2000 models. For C3 and C4

mixes, the EC2 model was found as the best predictor at early-ages for both mixes with the

28
average difference in the range of about 2-6%, followed by the GL2000, SAK, ACI and B3

models. The higher substitution of OPS from 80 to 100% (C5 and C6) in conventional

concrete showed a sharp increase in the drying shrinkage, therefore, no any suitable model

was found which can predict better results for such concretes. Although, all the models

selected in this study showed a significant difference between the experimental the predicted

results with the average difference in the range of about 35 to 80%.

Similarly, in group (2) mixes, for F1 concrete mixture, according to EP and CV, the best

shrinkage prediction model at early-ages was the SAK model as it predicted almost 100%

similar results. The ascending rank of the other models for the conventional concrete

containing fly ash (F1) was found as ACI, EC2, B3 and GL2000. Furthermore, the SAK

model was also found as a better prediction model for F2 mix, followed by EC2, ACI209,

GL2000 and B3 models. Similar to C3 and C4 mixes, the EC2 model was also found as a best

predictor at early-ages for both F3 and F4 mixes with the average difference in the range of

about 3-4%, followed by the GL2000, SAK, ACI and B3 models. The GL2000 and the EC2

models were found as good prediction model for the F5 mix, the average difference between

results was found in the range of 7-14%, whereas, the other models predicted significantly

lower results. The OPS concrete (F6) showed sharp increase in the drying shrinkage,

therefore, no any suitable model was found which can predict better results for such concrete.

Table 11. Error percentage analyses for the mixes at early-ages (14 days)

Mix Prediction models

Groups
code ACI209 EC2 GL2000 B3 SAK
C1 -21.56 +66.66 +83.33 -35.29 +29.41
C2 -30.43 +47.82 +62.60 -42.60 +14.78
C3 -51.80 +02.40 +12.65 -60.24 -20.48
Group (1)
C4 -50.00 +06.25 +16.87 -58.75 -17.50
C5 -72.02 -40.55 -34.61 -76.92 -53.84
C6 -76.11 -49.25 -44.17 -80.29 -60.60
F1 -24.44 +40.74 +75.55 -53.33 +02.22
F2 -35.44 +20.25 +50.00 -60.12 -12.65
Group (2)
F3 -44.56 +03.26 +28.80 -65.76 -25.01
F4 -48.74 -04.52 +19.09 -68.34 -30.65

29
 F5 -54.05 -14.41 +06.75 -71.62 -37.83
F6 -73.58 -50.77 -38.60 -83.67 -64.24

Table 12. Coefficient of variation analyses for the mixes at early-ages (14 days)

Mix Prediction models

Groups
code ACI209 EC2 GL2000 B3 SAK
C1 0.17 0.35 0.41 0.30 0.18
C2 0.25 0.27 0.34 0.38 0.09
C3 0.49 0.01 0.08 0.60 0.16
Group (1)
C4 0.47 0.04 0.11 0.58 0.13
C5 0.79 0.36 0.30 0.88 0.52
C6 0.86 0.46 0.40 0.94 0.61
F1 0.19 0.23 0.38 0.51 0.01
F2 0.30 0.13 0.28 0.61 0.10
F3 0.40 0.02 0.17 0.69 0.20
Group (2)
F4 0.45 0.03 0.12 0.73 0.26
F5 0.52 0.10 0.04 0.79 0.32
F6 0.82 0.48 0.33 1.01 0.66

Tables 13 and 14 present the summary of the EP and CV for both groups mixes at long-term

ages (275 days). For conventional concrete (C1), the best prediction models were ACI and

EC2. Although, they overestimated the results with the average difference of about 20%. This

might be due to high strength of the conventional concrete. Rank ascending to other models

for the conventional concrete (C1) were found as SAK, B3 and GL2000. However, the same

models (ACI and EC2) were also found as the best predictors for the C2 mix with a high

accuracy of about 96%-100%. Meanwhile, shrinkage results for C2 concrete mixture with the

error percentage of 11% and 20% was predicted by the SAK and GL2000 models,

respectively. At long-term ages, the GL2000 model was the best predictor for mix C3 as it

showed the accuracy of about 94%, followed by SAK which could predict the results with the

accuracy in the range of 87-90%. The ascending rank of other models for C3 mix was EC2,

ACI and B3. Furthermore, the mixes containing high volume OPS aggregates (C4 to C6)

showed significantly higher experimental results. Therefore, no accurate prediction model

can be suggested for those mixes.

30
In group (2) mixes, for F1 concrete mixture, according to EP and CV, the SAK model was

the best predictor for the long-term drying shrinkage as it predicted almost 97% similar

results. The ascending rank of the other models for the conventional concrete containing fly

ash (F1) was found as EC2, ACI, GL2000 and B3. Furthermore, the GL2000 model showed a

better correlation between experimental and predicted values for F2 mix, although, the

difference was found about 8%, followed by SAK, EC2, ACI209 and B3 models. However,

the same model GL2000 showed almost similar results for mix F3 and under-estimated with

the difference of about 16% for F4 concrete mixture, whereas, the predicted values from

other models were not found in a suitable range for both mixes. The concretes containing a

high volume of OPS (F5 and F6) showed a sharp increase in the drying shrinkage, therefore,

no suitable model was found.

Table 13. Error percentage analyses for the mixes at long-term ages (275 days)

Mix Prediction models

Groups
code ACI209 EC2 GL2000 B3 SAK
C1 +19.56 +19.92 +50.55 -40.22 +39.48
C2 -04.98 -04.62 +19.64 -52.49 +10.85
C3 -25.51 -25.28 -06.20 -62.75 -13.10
Group (1)
C4 -38.63 -38.44 -22.72 -69.31 -28.40
C5 -54.30 -54.16 -42.45 -77.15 -46.68
C6 -59.80 -59.67 -49.37 -79.90 -53.10
F1 -11.79 -10.81 26.53 -59.95 -03.19
F2 -24.90 -24.05 07.74 -65.89 -17.57
F3 -30.15 -29.37 00.19 -68.28 -23.34
Group (2)
F4 -41.53 -40.87 -16.12 -73.45 -35.83
F5 -54.95 -54.45 -35.38 -79.54 -50.56
F6 -62.41 -61.98 -46.07 -82.93 -58.74

Table 14. Coefficient of variation analyses for the mixes at long-term ages (275 days)

Mix Prediction models

Groups
code ACI209 EC2 GL2000 B3 SAK
C1 0.12 0.12 0.28 0.35 0.23
C2 0.03 0.03 0.12 0.50 0.07
C3 0.20 0.20 0.04 0.64 0.09
Group (1)
C4 0.33 0.33 0.18 0.75 0.23
C5 0.52 0.52 0.38 0.89 0.43
C6 0.60 0.60 0.46 0.94 0.51
F1 0.08 0.08 0.16 0.60 0.02
Group (2)
F2 0.20 0.19 0.05 0.69 0.13

31
 F3 0.25 0.24 0.00 0.73 0.18
F4 0.37 0.36 0.12 0.82 0.30
F5 0.53 0.52 0.30 0.93 0.47
F6 0.64 0.63 0.42 1.00 0.59

4. Conclusion

The development of the drying shrinkage strains for the conventional concrete containing oil

palm shell (OPS) as coarse aggregate (C1 to C6 mixes) and the same mixes containing 25%

fly ash by mass of the total cementitious material (F1 to F6 mixes) were compared by

applying five different prediction models such as ACI209R, EN1992, GL2000, B3 and SAK

which were selected from the researches and standards.

1. Substitution of OPS in conventional concrete has a significant effect on the rate of

shrinkage development at early-ages for all mixes. It also resulted in higher drying

shrinkage strain at long-term ages. In group (2) mixes, higher experimental drying

shrinkage strains were observed compared to the concrete mixtures in group (1) due to

the addition of fly ash as a cementitious material. The sharpest increment in shrinkage

strain was observed for mixes C6 and F6, which is about 75% greater than the predicted

results.

2. Most prediction models in this study showed much lower results of drying shrinkage

strain compared to the experimental values at long-term ages (275 days). It was mainly

due to sharp gaining of drying shrinkage strain for the mixes containing OPS aggregates.

3. Based on error percentage and coefficient of variation, Eurocode (EC2) model was the

best prediction model at early-ages as well as GL2000 at long-term ages. In contrast, for

mixes in both groups, B3 was considered as a poor drying shrinkage prediction model at

early and long-term ages.

4. In group (1) and at early-ages (14 days), the best models for prediction of drying

shrinkage were ACI209R, SAK for C1 and C2 mixes, respectively. EC2 model was

32
 found as a practical predictor for mixes C3 and C4 with the average difference in the

range of about 2-6%. However, no accurate prediction model was found for the C5 and

C6 mixes with the higher substitution of OPS from 80 to 100% in conventional concrete.

5. In group (2), for F1 and F2 mixes, the SAK model was considered as the best predictor

with 100% and 90% similarity in results, respectively. The EC2 was the most accurate

model among others for both mixes F3 and F4 with the average difference in the range of

about 3-4%. For mix F5, GL2000 and EC2 models were selected with the average

difference in the range of 7-14%. The F6 mix with 100% OPS substation of aggregate

showed the highest rate of drying shrinkage, as a result, none of the models in this study

could predict the drying shrinkage value for such concrete.

6. In group (1) and at long-term ages (275 days), both ACI and EC2 models were found as

reliable prediction models with the accuracy of about 80% and 96% for C1 and C2

mixes, respectively. The GL2000 model was selected for mix C3 with the accuracy of

about 94% while, no suitable prediction model was found to predict the drying shrinkage

of C4 to C6 mixes.

7. In group (2), the SAK model predicted the drying shrinkage strain with the accuracy of

about 97% for F1 mix. A good correlation between experimental and predicted values

with the difference of about 8% was observed by the GL2000 model for F2 mix. The

same model showed almost similar results for F3 mix, and under-estimated the drying

shrinkage with the difference of about 16% for F4 concrete mix. However, due to the

sharp increase in drying shrinkage for F5 and F6 mixes, no suitable model can be

proposed.

References

33
1. Maghfouri, M., Shafigh, P., Ibrahim, Z. B., & Alimohammadi, V., "Quality control of lightweight
aggregate concrete based on initial and final water absorption tests." IOP Conference Series:
Materials Science and Engineering. 2017, Vol. 210. No. 1. IOP Publishing.
2. ACI Committee 209, "Prediction of Creep, Shrinkage, and Temperature Effects in Concrete
Structures", in ACI Manual of Concrete Practice. American Concrete Institute: Farmington Hills, MI,
1997, p. 209R.1-209R.47.
3. Kosmatka, S., W. Panarese, and B. Kerkhoff, Portland cement association. Design and control of
concrete mixtures, 2002. 5420: p. 60077-61083.
4. Bažant, Z.P., J.-K. Kim, and L. Panula, Improved prediction model for time-dependent
deformations of concrete: Part 1-Shrinkage. Materials and Structures, 1991. 24(5): p. 327-345.
5. Boucherit, D., Kenai, S., Kadri, E., & Khatib, J. M., A simplified model for the prediction of long
term concrete drying shrinkage. KSCE Journal of Civil Engineering, 2014. 18(7): p. 2196-2208.
6. Bazant, Z., Prediction of concrete creep and shrinkage: past, present and future. Nuclear
Engineering and Design, 2001. 203(1): p. 27-38.
7. McDonald, D. and H. Roper. Prediction of drying shrinkage of concrete from internal humidities
and finite element techniques. in RILEM PROCEEDINGS. 1993. CHAPMAN & HALL.
8. Zia, P., A. Shuaib, and M. Leming, High-performance Concretes, a State-of-art Report (1989-
1994). 1997: Federal Highway Administration, FHWA-RD-97-030: McLean (VA).
9. Clarke, J., Structural lightweight aggregate concrete, Taylor & Francis e-Library, CRC Press,
London, United Kingdom, 2002.
10. Holm, T. and T. Bremner, State of the art report on high strength, high durability structural
low-density concrete for applications in severe marine environments, in US Army Corps of
Engineers, Engineering Research and Development Center, ERDC/SLTR-00-3. 2000.
11. Aslam, M., P. Shafigh, and M.Z. Jumaat, Drying shrinkage behaviour of structural lightweight
aggregate concrete containing blended oil palm bio-products. Journal of Cleaner Production, 2016.
127: p. 183-194. 30
12. Mannan, M. A., Alexander, J., Ganapathy, C., & Teo, D. C. L., Quality improvement of oil palm
shell (OPS) as coarse aggregate in lightweight concrete. Building Environment, 2006. 4(9): p. 1239-
1242.
13. Teo, D. C. L., Mannan, M. A., Kurian, V. J., & Ganapathy, C., Lightweight concrete made from
oil palm shell (OPS): structural bond and durability properties. Building Environment, 2007. 42(7): p.
2614-2621.
14. Aslam, M., P. Shafigh, and M.Z. Jumaat, Oil-palm by-products as lightweight aggregate in
concrete mixture: A review. Journal of Cleaner Production, 2016. 126: p. 56-73.
15. Aslam, M., Shafigh, P., Jumaat, M. Z., & Lachemi, M., Benefits of using blended waste coarse
lightweight aggregates in structural lightweight aggregate concrete. Journal of Cleaner Production,
2016. 119: p. 108-117.

34
16. Shafigh, P., Mahmud, H. B., Jumaat, M. Z., & Zargar, M., Agricultural wastes as aggregate in
concrete mixtures-A review. Construction and Building Materials, 2014. 53: p. 110-117.
17. Shafigh, P., Alengaram, U. J., Mahmud, H. B., & Jumaat, M. Z., Engineering properties of oil
palm shell lightweight concrete containing fly ash. Materials Design, 2013. 49: p. 613-621.
18. American Society for Testing and Materials. Specification forPortland Cement, ASTM C 150,
1994.
19. ASTM, C., Standard specification for coal fly ash and raw or calcined natural pozzolan for use in
concrete. 2005, ASTM C685.
20. Dixon, D.E., et al., Standard Practice for Selecting Proportions for Normal, Heavyweight, and
Mass Concrete (ACI 211.1-91). 1991.
21. 209.2R-08., A., Guide for modeling and calculating shrinkage and creep in hardened concrete.
2008, American Concrete Institute.
22. Maghfouri, M., P. Shafigh, and M. Aslam, Optimum Oil Palm Shell Content as Coarse Aggregate
in Concrete Based on Mechanical and Durability Properties. Advances in Materials Science and
Engineering, 2018.
23. Narayanan, R. and A. Beeby, Designers' Guide to EN 1992-1-1 and EN 1992-1-2. Eurocode 2:
Design of Concrete Structures: General Rules and Rules for Buildings and Structural Fire Design.
Thomas Telford London, UK, 2005.
24. 1992-1-1., E., Eurocode 2: Design of concrete structures – Part 1–1: General rules and rules for
buildings. 2010, European Committee for standardization CEN.
25. Gardner, N. and M. Lockman, Design Provisions for Drying Shrinkage and Creep of
NormalStrength Concrete. . ACI Materials Journal, 2001. 98, p. 159-167.
26. Bazant, Z. and S. Baweja, Creep and shrinkage prediction model for analysis and design of
concrete structures: Model B3. ACI Special Publications, 2000. 194: p. 1-84.
27. Sakata, K., Tsubaki, T., Inoue, S., and Ayano, T. Prediction equations of creep and drying
shrinkage for wide-ranged strength concrete. in Creep, Shrinkage and Durability of Concrete and
other quasi-brittle Materials, Proceedings of the Sixth International Conference, Cambridge (MA).
2001.
28. Sakata, K., Prediction of creep and shrinkage of concrete. JapanSociety of Civil Engineer, 1996.
29. Khanzadeh, M., Accuracy of Shrinkage Prediction Models in High Performance Concretes
Containing Slag and Silica Fume, Paper presented at the In Second International
Conference on Sustainable Construction Materials and Technologies: Ancona, Italy, 2010.

35
Research highlights:

Addition of fly ash to the concrete mixtures generated higher drying shrinkage
No suitable prediction model can be proposed for concretes containing high volume
of oil palm shell lightweight aggregates.
EN1992 was the best model for prediction of drying shrinkage at early-ages as well as
GL2000 at long-term ages.
B3 model was considered as a poor drying shrinkage prediction model for all mixes in
both groups at early and long-term ages.
 AUTHORSHIP STATEMENT

Manuscript title: Appropriate Drying Shrinkage Prediction Models for Lightweight

Concrete Containing Coarse Agro-Waste Aggregate

All persons who meet authorship criteria are listed as authors, and all authors certify
that they have participated sufficiently in the work to take public responsibility for
the content, including participation in the concept, design, analysis, writing, or
revision of the manuscript. Furthermore, each author certifies that this material or
similar material has not been and will not be submitted to or published in any other
publication before its appearance in the Journal of Building Engineering.

Authorship contributions

Conception and design of study:

Mehdi Maghfouri, Muhammad Aslam & Payam Shafigh
Acquisition of data:
Mehdi Maghfouri, Vahid Alimohammadi, Muhammad Aslam & Yashar Doroudi
Analysis and interpretation of data:
Mehdi Maghfouri, Muhammad Aslam, Vahid Alimohammadi, Payam Shafigh
& Yashar Doroudi
Drafting the manuscript:
Mehdi Maghfouri, Muhammad Aslam, Vahid Alimohammadi, & Payam Shafigh
Revising the manuscript critically for important intellectual content:

Mehdi Maghfouri, Vahid Alimohammadi, Payam Shafigh & Yashar Doroudi

Approval of the version of the manuscript to be published:

Mehdi Maghfouri, Payam Shafigh & Vahid Alimohammadi

Author’s name

Affiliation

Mehdi Maghfouri

Department of Civil Engineering, Faculty of Engineering, University of Malaya 50603 Kuala Lumpur, Malaysia

Payam Shafigh

Department of Building Surveying, Faculty of Built Environment, University of Malaya, 50603 Kuala Lumpur,
Malaysia

Vahid Alimohammadi

Young Researchers and Elite Club Azad University, Arak Branch, 3836119131 Arak, Iran

Yashar Doroudi

School of Civil Engineering, The University of Queensland, St Lucia, QLD 4074, Australia

Muhammad Aslam

Civil Engineering, School of Engineering & Technology, Institute of Southern Punjab, Multan, Pakistan
 Conflicts of Interest Statement

Manuscript title: Appropriate Drying Shrinkage Prediction Models for Lightweight

Concrete Containing Coarse Agro-Waste Aggregate

The authors whose names are listed below certify that they have participated for
analysis, interpretation of the data, drafting the article and revising for important
intellectual content. The authors also certify that the present manuscript has not been
submitted to, nor is under review at, another journal or other publishing venue.

In this manuscript, the authors have no affiliation with any organization with a direct
or indirect financial interest in the subject matter discussed in the manuscript

Author’s name
Affiliation

Mehdi Maghfouri
Department of Civil Engineering, Faculty of Engineering, University of Malaya 50603 Kuala Lumpur, Malaysia

Payam Shafigh
Department of Building Surveying, Faculty of Built Environment, University of Malaya, 50603 Kuala Lumpur,
Malaysia

Vahid Alimohammadi
Young Researchers and Elite Club Azad University, Arak Branch, 3836119131 Arak, Iran

Yashar Doroudi
School of Civil Engineering, The University of Queensland, St Lucia, QLD 4074, Australia

Muhammad Aslam
Civil Engineering, School of Engineering & Technology, Institute of Southern Punjab, Multan, Pakistan