ACF Journal - Asian Concrete Federation€¦ · tive property of strain hardening Ultra High Perfor-mance Fiber Reinforced Cementitious Composite (UHPFRC), so the accurate and representative

ISSN 2465-7964eISSN 2465-7972

Asia

n C

on

crete

Fe

de

ratio

n

AC

F Jo

urn

al

Vol. 4, No. 1, June 2018Vol. 4, No. 2, December 2018

Vol. 4, No. 1, June 2018Proposal of a fatigue life prediction method for RC slabs failed under traveling wheel-type load test 1-11Kyoko TAKEDA, Natsuko HAMADA and Yasuhiko SATOQuality improvement of recycled concrete aggregate by a large-scale tube mill with steel rod 12-21Lapyote Prasittisopin, Chawis Thongyothee, Phattarakamon Chaiyapoom and Chalermwut SnguanyatEffect of pre-loading on chloride diffusion in concrete 22-28Truyen T. Tran, Quyet V. Truong, H. Ranaivomanana and A. KhelidjPerformance evaluation of basalt fiber reinforced mortar under freeze-thaw and chloride-rich environments 29-34Yiming Guo and Hiroshi YokotaComparison of concrete strength from cube and core records by bootstrap 35-46Saha Dauji and Kapilesh BhargavaInvestigation on quality of thin concrete cover using mercury intrusion porosimetry and non-destructive tests 47-66Liyanto Eddy, Koji Matsumoto, Kohei Nagai, Piyaphat Chaemchuen, Michael Henry and Kota Horiuchi

Vol. 4, No. 2, December 2018Tensile behavior of UHPFRC under uniaxial and biaxial stress conditions 67-78Xiujiang Shen and Eugen BrühwilerDeformation mechanism of hardened cement paste under high stress and application of flow law 79-88Yuya SakaiStrength, shrinkage and creep of concrete including CO2 treated recycled coarse aggregate 89-102Gombosuren Chinzorigt, Donguk Choi, Odontuya Enkhbold, Batzaya Baasankhuu, Myung Kwan Lim and Hee Seob LimVisual investigation method and structural performance evaluation for DEF induced damaged Indian Railway PC sleepers 103-115Rajamurugan Sundaram, Koji Matsumoto, Kohei Nagai and Anupam Awasthi

Journal of Asian Concrete Federation

Vol. 4, No. 2, pp. 67-78, December 2018

ISSN 2465-7964 / eISSN 2465-7972

https://doi.org/10.18702/acf.2019.1.4.2.67

Technical Paper

Tensile behavior of UHPFRC under uniaxial and biaxial

stress conditions

Xiujiang Shen* and Eugen Brühwiler

(Received: March 21, 2018; Accepted: December 14, 2018; Published online: January 04, 2019)

Abstract: Representative and accurate characterization of the tensile behavior of strain hardening Ultra High

Performance Fiber Reinforced Cementitious Composite (UHPFRC) remain a challenge. Currently, the uniaxial

methods, like direct tensile test (DTT) and 4-point bending test (4PBT), are commonly applied, although the

biaxial tensile condition has been widely recognized in the UHPFRC applications, e.g. thin UHPFRC layers

as external reinforcement for RC slabs. In this paper, results from ring-on-ring testing of circular slab-like

specimens are presented to determine the equi-biaxial tensile response by means of inverse analysis using 3D

finite element method (FEM). In addition, DTT, using structural specimens cut from large square plates, and

4PBT, using standard specimens cast in mould individually, were carried out. The tensile response from 4PBT

was derived through inverse analysis using 2D FEM. Finally, the corresponding results from the three different

testing methods under either uniaxial or biaxial stress condition were analyzed and compared in terms of tensile

characteristic parameters, tensile material law, fracture process, and energy absorption capacity. While the

three testing methods did not show significant difference in tensile strength, significantly higher strain hard-

ening deformation was identified in the case of biaxial stress conditions.

Keywords: biaxial stress condition, FEM, inverse analysis, ring-on-ring test, UHPFRC, uniaxial stress con-

dition, tensile behavior.

1. Introduction

The tensile response is a fundamental constitu-

tive property of strain hardening Ultra High Perfor-

mance Fiber Reinforced Cementitious Composite

(UHPFRC), so the accurate and representative char-

acterization of this response is necessary for the de-

sign of a given UHPFRC application. In general, this

characterization is achieved by means of uniaxial test

methods, especially direct tensile test (DTT) and 4-

point bending test (4PBT) using small-scale labora-

tory specimens casting in moulds individually. Un-

fortunately, these tests exhibit considerable scatter

and the results are often considered as an upper

bound in case of small-scale laboratory specimens,

hardly reproducing real design situations. Most in-

frastructures, in particular bridge decks and floors,

are principally under biaxial stress condition, far

from uniaxial stress state [1]. In this context, the ac-

tual tensile performance of UHPFRC under biaxial

stress condition should be investigated and com-

pared with that from uniaxial stress condition care-

fully.

The DTT using dumbbell-shaped specimen is

commonly applied to determine directly the uniaxial

tensile properties of UHPFRC for given preparation

conditions (moulds, casting, and curing). For reliable

results, wise design and preparation are required

when conducting DTT. In order to avoid largely the

initial eccentricity with bending effects, the speci-

men was built-in the testing machine by applying the

principle “gluing without adherence”, developed by

Helbling & Brühwiler [2]. This method was also ap-

plied in ref. [3,4]. Proposed by Graybeal et al. [5,6],

the tapered aluminum plates were fixed to both sides

of each specimen end to ensure final fracture occurs

out of the central constant part. Otherwise, one or

two layers of steel wire mesh were used to strengthen

each end of specimen, as applied in ref. [7,8]. Addi-

tionally, boundary conditions also have important in-

fluence on test results, and the fixed conditions is

Corresponding author Xiujiang Shen is a doctoral can-

didate, Maintenance and Safety of Structures (MCS-

ENAC), Ecole Polytechnique Fédérale de Lausanne

(EPFL), GC B2-402, Station 18, CH-1015 Lausanne,

Switzerland (Email: [email protected]).

Eugen Brühwiler is a professor, director of Maintenance

and Safety of Structures (MCS-ENAC), Ecole Polytech-

nique Fédérale de Lau-sanne (EPFL), GC B2-402, Sta-

tion 18, CH-1015 Lau-sanne, Switzerland

(Email: [email protected]).

67

recommend for reliable estimation of tensile re-

sponse of UHPFRC after elastic limit, as confirmed

by Kanakubo [9]. The 4PBT, alternatively, was used

successfully to identify the tensile property of UHP-

FRC indirectly by means of inverse analysis, includ-

ing analytical method and FEM [10–12].

Regarding direct biaxial tests, four actuators

and a big frame are generally necessary, and inher-

ently, many challenges pertaining to uniform load

distribution, frictional effect, accurate boundary con-

dition and load control need to be addressed care-

fully. Therefore, only few studies have been con-

ducted on the biaxial behavior of concrete [13,14],

especially no experimental study on biaxial behavior

of UHPFRC has been recorded through direct biaxial

test. Recently, the ring-on-ring test, as a 3D version

of 4PBT, has been developed to investigate the biax-

ial flexural strength of concrete [15–17]. This

method was extended to UHPFRC by several re-

searchers [1]. The limited test results show that the

ring-on-ring test allowed the actual development of

fiber bridging effects between cracks, thus accu-

rately representing the behavior of UHPFRC mem-

bers subjected to biaxial flexural loading.

In this paper, the ring-on-ring test on circular

slab-like specimens has been developed to determine

the equi-biaxial tensile response. In addition, direct

tensile tests using dumbbell specimens cut from

large square plates and 4PBT using small plates cast

in molds were carried out. The corresponding tensile

response from five DTT, six 4PBT and four ring-on-

ring tests were analyzed and compared (See Fig. 1).

The main objective was to examine the differences

and relationships of the tensile response of UHPFRC

under uniaxial and biaxial stress conditions, and to

propose the most appropriate test method to deter-

mine the tensile property for a given UHPFRC appli-

cation.

2. Experimental Program

2.1 Ring-on-ring test

The ring-on-ring test method was applied for in-

direct characterization of the tensile behavior under

biaxial stress condition, using circular slab-like spec-

imens with a diameter R = 600 mm and a thickness

h = 50 mm. This method has been extensively

adopted and even standardized by ASTM [18] in the

ceramics and glass domain. Recently, this method

was modified and validated to measure the biaxial

flexural strength of concrete and UHPFRC [15–

17,19,20]. The updated ring-on-ring test yielded sta-

ble test results with small scatter, and it is promising

to be a reliable and rational means to investigate bi-

axial flexural behavior of UHPFRC.

Figure 2 shows the full test set-up and devices

applied in this experimental campaign. The slab was

simply supported on a steel support ring with R =

500 mm. Loading was imposed by a hydraulic jack

acting on the center of slab through a steel force

transmitting ring with r = 150 mm. All the slabs were

subjected to three loading–unloading cycles to 20 kN

with an actuator displacement rate of 1.0 mm/min.

Afterwards, monotonic loading with the same dis-

placement rate was applied up to the peak force, fol-

lowed by a rate of 4.0 mm/min until the actuator dis-

placement reached 80 mm. Under loading, the uni-

form stress is introduced on the bottom surface

within the force transmitting ring area, where biaxial

stress condition is assumed.

Fig. 1 – Approach for the comparison of the tensile response of UHPFRC under uniaxial and biaxial stress

conditions

68

Journal of Asian Concrete Federation, Vol. 4, No. 2, December 2018

Fig. 2 – Schematic description of test setup (unit: mm)

The slabs were tested with the casting surface

facing upwards, allowing the observation of tensile

crack propagation on the smooth sheathed surface.

Before testing, the casting surface was polished and

a mortar layer was placed between support ring and

bottom surface to level out both surfaces. Two rub-

ber pads (thickness of 10 mm, E = 500 MPa) were

positioned between the slab surfaces and the two

rings to distribute the force evenly.

As illustrated in Fig. 2, Digital Image Correla-

tion (DIC) technique was applied to observe the de-

flection development, strain field, and micro-crack-

ing during the whole testing process. Two digital

cameras were placed underneath the slab at a dis-

tance of 0.5 m and an angle of 23 degrees to the ver-

tical. The tensile surface of the slab was painted with

matte white paint, and then spayed black speckle pat-

tern with size less than 1 mm. The targeted area,

which was visible to the DIC, was about Ø 500 mm

on the center of the slab. In such case, the DIC meas-

urement accuracy can reach around 5 με. In addition,

several LVDTs were installed on the top surface to

measure the deflection. All deflection measurements

were performed with respect to the strong floor. The

measurement frequency was 5 Hz. Further details

about the ring-on-ring test applied in this study can

be found in [21].

2.2 Direct tensile test (DTT)

The dumbbell shaped specimens, with a con-

stant cross section of 80 mm × 50 mm at the central

part, were adopted for uniaxial DTT. The geometry

of specimen was designed based on the equation of

Neuber’s spline [22,23]. In total five specimens were

extracted from a large square plate (1,100 mm ×

1,100 mm × 50 mm) with the same thickness and

casting procedure as that for the circular slab-like

specimen (See Fig. 3). This allowed to assess the var-

iability of tensile behavior in the plate.

The tensile tests for all specimens were per-

formed on a universal servo-hydraulic testing ma-

chine with a capacity of 1,000 kN, according to SIA

2052 [24]. The Digital Image Correlation (DIC)

technique and three different series of sensors were

adopted to measure the deformation and crack open-

ing of the UHPFRC, as shown in Fig. 4. Further de-

tails about the developed DTT in this study can be

found in ref. [3].

2.3 Four-point bending test (4PBT)

In total six small plate specimens with dimen-

sion of 500 mm × 100 mm × 3 mm were cast indi-

vidually in molds. The 4PBT for all specimens was

performed on a universal servo-hydraulic testing ma-

chine with a capacity of 200 kN, according to SIA

2052 [10,24]. The total span of the four-point bend-

ing test set up was 420 mm (See Fig. 5), and the sup-

ports allowed free displacement of the specimen

along its longitudinal axis. Two transducers placed

on a measuring frame on each side of the specimen

measured the net deflection in the center of the span.

The measurements were taken at a frequency of 5Hz

during the test.

69


Fig. 3 – Extracting of dumbbell specimens and dimension of dumbbell specimen (unit: mm)

Fig. 4 – Tensile test setup and instrumentation for the dumbbell specimens (unit: mm)

Fig. 5 – Four-point bending test setup and instrumentation (unit: mm) [24]

2.4 Fabrication and curing

The chosen UHPFRC is an industrial premix

containing 3.8% by volume of straight steel fibers

with length of 13 mm and diameter of 0.175 mm, and

its water/cement ratio is 0.15. The UHPFRC was

mixed to obtain a batch of 180 liters. The large

square plate and circular slab-like specimens were

cast in one step: the fresh UHPFRC mixture was

(a) two application points of t

he displacements;

(b) reference wafer fixed on th

e upper specimen surface;

(c) metallic frame placed at m

id-height of the specimen

70


poured in the center of the formworks, and let flow

without any pulling or vibration. Regarding the small

plate specimens for 4PBT, the fresh mixture was

poured from one side and let flow. Once the casting

was completed, a plastic sheet was pulled over the

specimens to allow for auto-curing of the material.

The specimens were demolded after 24 hours, then

kept under moist curing conditions (20℃, 100% hu-

mility) for the following seven days; and subse-

quently, stored inside the laboratory until testing.

The test age was more than 60 days, given that more

than 90% of the UHPFRC material properties is at-

tained after 60 days [4,25].

3. Test Results

3.1 Uniaxial tensile response from DTT

In Figure 6, the DTT results of five specimens

are presented in terms of stress-strain (σ-ε) curves, in

which the thick black line represents the average re-

sponse. The stress is defined as the measured force

divided by the constant cross-sectional area of dumb-

bell specimen, while the strain is based on the aver-

age value measured from two short LVDTs with

measuring length of 160 mm. The main characteris-

tic tensile parameters for each specimen are summa-

rized in Table 1, including elastic modulus EU, elas-

tic limit point (stress fUte and corresponding strain

εUte), and ultimate point (fUtu and εUtu). Here, the end

of the linear relationship in σ-ε curve is regarded as

elastic limit point, and the beginning of the tensile

softening response is defined as ultimate point. The

average curve is obtained through averaging 5 nor-

malized curves, where the stress and strain are di-

vided by the corresponding values at peak point (fUtu

and εUtu), respectively. As shown in Fig. 6 and Table

1, a considerable variation of tensile response, strain-

hardening behavior in particular, is observed. This is

attributed to the variability of fiber distribution in

different specimens depending on the distance from

the pouring point [3]. In this case, due to the high

fluidity and workability of the UHPFRC material,

the fresh mixture flowed freely from the center to the

border in radial direction. The flow exerted forces on

the fibers, pushing fibers to align more perpendicu-

larly to the flow direction, as illustrated in Fig. 7. Ad-

ditionally, based on a previous study [3], random fi-

ber distribution and orientation can be assumed for

specimen T2~T4 around the pouring point.

Based on DIC analysis using VIC-3D, the

whole microcracking and fracture process of each

specimen is captured effectively during the loading

process. The initiation and propagation of fine mi-

cro-cracks in the strain-hardening domain, in partic-

ular, can be detected visually in DIC full-field strain

maps. The representative fracture process shown by

T3 is illustrated in Fig. 8. Generally, point A (elastic

limit) refers to the start of strain-hardening response

in UHPFRC, symbolized by the activation of first

fine micro-cracks; while point B (ultimate limit)

stands for the end of this response, characterized by

the formation of one single localized fictitious crack

by grouping of several fine micro-cracks. Afterwards

(beyond point B), the fictitious crack shows signifi-

cant stress transfer through the fibers bridging the

two crack sides and develops with increasing crack

opening until no more stress is transferred when a

crack opening of half the fiber length is reached, i.e.,

in present case, 6.5 mm (= Lf/2). It is important to

note that the microcracks are mostly concentrated in

two local zones with random distribution of several

micro-cracks over a large extent (> Lf).

Fig. 6 – Tensile response from DTT

71


Table 1 – Tensile parameters from DTT

N° EU [GPa] fUte [MPa] fUtu [MPa] fUtu/fUte εUte [‰] εUtu [‰]

T1 45.60 9.70 12.23 1.26 0.22 3.48

T2 49.09 9.41 11.21 1.19 0.20 2.33

T3 47.39 7.74 9.62 1.24 0.17 2.68

T4 49.64 7.25 9.48 1.31 0.17 2.15

T5 48.38 7.20 9.10 1.26 0.16 0.68

Average 48.02 8.26 10.33 1.25 0.18 2.26

Std. dev. 1.59 1.21 1.33 0.04 0.03 1.02

COV 0.03 0.15 0.13 0.03 0.14 0.45

Fig. 7 – Final crack positions of dumbbell specimens and schematic view of fiber distribution in the

plate

Fig. 8 – Representative microcracking and fracture process of UHPFRC specimen under DTT

72


Fig. 9 – Bending response from 4PBT

3.2 Uniaxial tensile response based on inverse

analysis from 4PBT

Figure 9 presents the bending behavior from

4PBT in terms of force-deflection curves, in which

the thick black curve is the average response. All

curves agree well with each other, showing compa-

rable bending behaviour. This results from the simi-

lar fiber distribution and orientation in the small

plate specimens, given that they were fabricated in-

dividually in moulds following the same casting pro-

cedure.

The uniaxial tensile response of UHPFRC is

evaluated indirectly by means of inverse analysis of

4PBT results using non-linear Finite Element Anal-

ysis (FEA). A 2D FE model was built using the non-

linear FE analysis software DIANA (smeared crack

model), targeting at simulating the bending behav-

iour in terms of force-deflection response and crack-

ing pattern of small plate specimens under 4PBT.

The best results of FE model fitting with the average

experimental curve is shown in Fig. 9, where a close

fit is achieved, as indicated by the thick red curve.

The corresponding uniaxial tensile parameters are

summarized in Fig. 13 and Table 2.

3.3 Biaxial tensile response based on inverse

analysis from ring-on-ring test

The biaxial flexural responses of four UHPFRC

circular slabs from ring-on-ring tests are presented in

terms of force-deflection curve (F - δ) of the center

point, as shown in Fig. 10. The recorded force value

was adjusted considering a geometry factor that ac-

counts for the precise thickness of each slab. The de-

flection was measured by DIC on the bottom surface,

excluding the deformation of the rubber pad meas-

ured from three LVDTs on the top surface. It is obvi-

ous that all slabs show a consistent flexural response

with little scatter. Up to a force value of about 40kN,

the flexural behavior of the UHPFRC slabs in terms

of F-δ curve is almost linear, and the end of this lin-

earity is herein defined as elastic limit (point A). Af-

terwards, a quasi-linear response (II, A-B) with a

slight decrease of stiffness is noticed, in which the

formation and propagation of multiple microcracks

are expected. And sequentially, significant deflection

hardening behaviour is identified until the peak point

(C) is reached. Afterwards, the slabs exhibit im-

portant ductility with high residual resistance in the

softening phase (IV C-D). Further details about the

test results are described in [21].

Furthermore, the representative microcracking

and fracture process from S1-3, as observed by DIC

on the visualized central portion (400 mm × 400 mm),

is shown in Fig. 11. The selected DIC images in Fig.

11 represent the crack patterns in different character-

istic phases following the F-δ curve, and the white

dash circle marks the position of the force transmit-

ting ring.

Accordingly, several phenomena characterizing

microcracking process of UHPFRC slabs are identi-

fied. Once the elastic limit (A) is reached, the first

microcracks with random distribution initiate within

the force transmitting ring area, where the tensile

stress is uniform and maximal in all directions. Af-

terwards, these microcracks start to propagate irreg-

ularly, and more new microcracks are generated until

point B is reached. In this phase II, the microcrack

pattern changes continuously with increasing deflec-

tion and has a complex distribution, microcracks in-

itiate randomly with irregular propagation paths;

some of them produce multiple branches, and some

73


cross each other or combine together during propa-

gation. At point B, several fine fictitious cracks (w ≥

0.05 mm) are detected by DIC, and afterwards, more

microcracking appears with increasing deflection.

Fine fictitious cracks initiate from some of previous

microcracks in phase II, and largely concentrate

within the force transmitting ring area. Some fine fic-

titious cracks are localized at peak force (point C),

and then propagate radially from the center to the

edge with increasing crack opening in the softening

phase (C-D). No additional cracks form beyond

point C.

Similarly, in order to determine the biaxial ten-

sile response of UHPFRC, the inverse analysis of the

ring-on-ring test results was conducted by means of

a 3D FE analysis using DIANA software. Consider-

ing the random fiber distribution in the slab, the full

scale of the slab element was modeled, and the

smeared crack concept was adopted. The boundary

conditions in simulation of ring-on-ring test have a

strong influence on the stability of the numerical pro-

cedure and on the fracture pattern. During testing, the

specimens were not prevented from sliding and lift-

ing from the supporting ring along fracture growth.

Therefore, the two adjacent ¼ points of the support

ring were only constrained in tangential directions,

and the top surface central point was constrained in

both X and Y directions. Additionally, the interface

elements were built between the specimen and the

rubber pads, avoiding tensile reaction force from the

supporting ring. The base and verification of FE

model are described in ref. [21]. The modeling re-

sults with best fitting of F-δ curve with respect to the

average response is illustrated in Fig. 10 (red line),

where a close fit is achieved. The results from the

inverse analysis are summarized in Fig. 13 and Table

2.

4. Discussion

Finally, all the tensile responses under uniaxial

or biaxial stress condition based on different meth-

ods, are summarized in Fig. 13 and Table 2. Regard-

ing DTT, the average response from specimen

T2~T4 is used for further comparison, since their lo-

cations in the large plate correspond to the area under

biaxial stress condition in ring-on-ring test. The uni-

axial tensile response from 4PBT cannot be applied

directly, due to favorable fiber alignment in the small

plates along the loading direction. This phenomenon

is attributed to the small geometry of the mould and

specific casting process as described in section 2.4.

Thus, fiber distribution and orientation effect should

be considered for better comparison.

Based on a previous study on an UHPFRC layer

with thickness of 50 mm [12], the average fiber ori-

entation factor (μ0) was identified in the range of

0.53~0.60. In this study, the mean value is applied,

namely μ0 = 0.57, for the DTT specimens, given that

random fiber distribution may be assumed in the cen-

tral part of the large plate. For a UHPFRC layer or

small specimen with thickness of 30 mm, μ0 was de-

termined in the range of 0.61~0.70 in ref. [12]. Thus,

the upper limit (μ0 = 0.70) is chosen for 4PBT speci-

mens, respecting preferential fiber alignment in

small plates. The corresponding efficiency factor

(μ1) is obtained for both cases (0.94 and 0.96, respec-

tively) based on Fig. 10. Accordingly, the uniaxial

tensile response from 4PBT is modified to be repre-

sentative for the large plate in the case of random fi-

ber distribution, the results are also summarized in

Fig. 13 and Table 2.

As observed in Fig. 13 and Table 2, in general,

the tensile performance in terms of strength from

both uniaxial and biaxial stress states are quantita-

tively similar. This phenomenon can be attributed to

the fact that all the test methods provide the speci-

mens with a certain area of uniform stress, allowing

for initiation of microcracks and localization of fic-

titious cracks at local weaker zones with respect to

the main stress direction. Thus, the tensile perfor-

mance largely depends on the distribution and size of

local weaker zones. In the case of random fiber dis-

tribution, the local weaker zones can be assumed to

distribute randomly without any considerable prefer-

ence in all directions.

Additionally, it should be noted that a signifi-

cant increase in hardening strain εUtu is found under

biaxial stress state. This effect can be due to the fact

that many more fibers in different directions contrib-

uted to the bridging and debonding effects under bi-

axial stress state, offering considerably higher duc-

tility and toughness including larger deformation,

compared with the results from specimens with uni-

axial stress state, where fibers perpendicular to the

loading direction have no contribution. This differ-

ence can also be explained by the different cracking

patterns under different stress states, as illustrated in

Figs. 8 and 11. The circular slab subjected to biaxial

stress state shows a large amount of microcracks dis-

tributed densely and randomly on the tensile surface

in strain-hardening domain (phase II, AB), and most

microcracks developed along irregular paths are in-

terlocked. Sequentially, several fictitious cracks ap-

peared. In the case of the DTT specimens under uni-

axial stress state, the microcracks concentrated lo-

cally at weaker zones and propagated from one edge

to the other following a relatively linear path, and

only one or two fictitious cracks localized finally.

74


Fig. 10 – Biaxial flexural response from the ring-on-ring tests

Fig. 11 – Representative microcracking and fracture process of UHPFRC slab of a ring-on-ring test

specimen.

Table 2 – Tensile parameters from the three different test methods

Test Method EU [GPa] fUte [MPa] fUtu [MPa] fUtu/fUte εUte [‰] εUtu [‰]

DTT 49 8.13 10.10 1.24 0.18 2.39

4PBT 51 10.00 14.00 1.40 0.20 3.92

4PBT (modified) 51 8.00 11.20 1.40 0.16 3.14

Ring-on-ring test 50 9.60 11.35 1.18 0.19 5.54

75


Fig. 12 – Relationship between orientation factor μ0 and efficiency factor μ1 [12]

Fig. 13 – Tensile responses from the three different test methods

5. Conclusions

This study investigated the tensile behavior of

UHPFRC under uniaxial and biaxial stress condi-

tions by means of DTT, inverse analysis based on

4PBT, and ring-on-ring test results using FEM. The

results suggest that the tensile response of UHPFRC

is not an intrinsic property and depends on several

factors, including the specimen geometry, flow re-

gime of fresh mixture during casting, and the stress

condition imposed on the specimen.

In the case of random fiber distribution, there is

no significant difference of tensile performance in

terms of strength between uniaxial and biaxial stress

conditions. Furthermore, since more fibers are acti-

vated, the multiple microcracking behavior is more

76


pronounced and complex under biaxial stress condi-

tion, resulting in a significantly higher strain harden-

ing deformation εUtu,compared with uniaxial stress

condition. Consequently, when the fibers are distrib-

uted randomly, it is conservative to use uniaxial ten-

sile parameters as obtained from DTT or 4PBT, to

design UHPFRC structural elements or UHPFRC

strengthening layers on concrete substrates subjected

to biaxial stress condition.

References 1. J. Kim; D.J. Kim; S.H. Park; and G. Zi (2015)

"Investigating the flexural resistance of fiber

reinforced cementitious composites under bi-

axial condition," Composite Structures, 122,

pp. 198–208.

2. A. Helbling and E. Brühwiler (1987) "Eine

neue Halterung für Zugversuche mit Beton-

Probekörper," Material Und Technik, 4, pp.

103–107 (in German).

3. X. Shen and E. Brühwiler "Characterization of

Tensile Beahvior in UHPFRC Thin Slab Using

NDT Method and DIC System," Proceedings

PRO 129, Fuzhou, China, November 2018.

4. K. Habel (2004) Structural behaviour of ele-

ments combining ultra-high performance fibre

reinforced concretes (UHPFRC) and rein-

forced concrete, Doctoral Thesis, EPFL, Swiz-

terland.

5. B.A. Graybeal and F. Baby (2013) "Develop-

ment of Direct Tension Test Method for Ultra-

High-Performance Fiber-Reinforced Con-

crete," ACI Materials Journal, 110(2), pp. 177-

186.

6. A. Abrishambaf; M. Pimentel; and S. Nunes

(2017) "Influence of fibre orientation on the

tensile behaviour of ultra-high performance fi-

bre reinforced cementitious composites," Ce-

ment and Concrete Research, 97, pp. 28–40.

7. S. Gwon and M. Shin (2016) "Direct-tensile

and flexural strength and toughness of high-

strength fiber-reinforced cement composites

with different steel fibers," Journal of Asian

Concrete Federation, 2(1), pp. 67–80.

8. M. Shin; S.-W. Gwon; K. Lee; S.W. Han; and

Y.W. Jo (2014) "Effectiveness of high perfor-

mance fiber-reinforced cement composites in

slender coupling beams," Construction and

Building Materials, 68, pp. 476–490.

9. T. Kanakubo (2006) "Tensile Characteristics

Evaluation Method for Ductile Fiber-Rein-

forced Cementitious Composites," ACT, 4, pp.

3–17.

10. E. Denarié; L. Sofia; and E. Brühwiler (2017)

"Characterization of the tensile response of

strain hardening UHPFRC - Chillon Viaducts,"

AFGC-ACI-Fib-RILEM Int. Symposium on

Ultra-High Performance Fibre-Reinforced

Concrete, UHPFRC 2017.

11. SAMARIS D26 (2006) Modelling of UHPFRC

in composite structures, European project 5th

FWP / SAMARIS – Sustainable and Advanced

Materials for Road Infrastructures – WP 14:

HPFRCC (http://samaris.zag.si/).

12. M. Bastien-Masse; E. Denarié; and E.

Brühwiler (2016) "Effect of fiber orientation

on the in-plane tensile response of UHPFRC

reinforcement layers," Cement and Concrete

Composites, 67, pp. 111–125.

13. S.-K. Lee; Y.-C. Song; and S.-H. Han (2004)

"Biaxial behavior of plain concrete of nuclear

containment building," Nuclear Engineering

and Design, 227, pp. 143–153.

14. W. Swanepoel (2011) The behaviour of fibre

reinforced concrete (SHCC) under biaxial

compression and tension, Stellenbosch: Stel-

lenbosch University.

15. G. Zi; H. Oh; and S.-K. Park (2008) "A novel

indirect tensile test method to measure the bi-

axial tensile strength of concretes and other

quasibrittle materials," Cement and Concrete

Research, 38, pp. 751–756.

16. O. Ekincioglu (2010) A discussion of paper “A

novel indirect tensile test method to measure

the biaxial tensile strength of concretes and

other quasibrittle materials” by G. Zi, H. Oh,

S.K. Park, Cement and Concrete Research, 40,

pp. 1769–1770.

17. J. Kim; D.J. Kim; and G. Zi (2013) "Improve-

ment of the biaxial flexure test method for con-

crete," Cement and Concrete Composites, 37,

pp. 154–160.

18. ASTM C1499-05 (2009) Standard test method

for monotonic equibiaxial flexural strength of

advanced ceramics at ambient temperature,

ASTM International, West Conshohocken,

Pennsylvania.

19. D.-Y. Yoo; G. Zi; S.-T. Kang; and Y.-S. Yoon

(2015) "Biaxial flexural behavior of ultra-high-

performance fiber-reinforced concrete with

different fiber lengths and placement meth-

ods," Cement and Concrete Composites, 63, pp.

51–66.

20. D.-Y. Yoo; N. Banthia; G. Zi; and Y.-S. Yoon

(2017) "Comparative Biaxial Flexural Behav-

ior of Ultra-High-Performance Fiber-Rein-

forced Concrete Panels Using Two Different

Test and Placement Methods," Journal of Test-

ing and Evaluation, 45(2), pp.624-641

21. X. Shen and E. Bruehwiler (2017) "Flexural

quasi-static behavior of UHPFRC circular slab

specimens," AFGC-ACI-Fib-RILEM Int.

Symposium on Ultra-High Performance Fibre-

Reinforced Concrete, UHPFRC 2017.

77


22. H. Neuber (1969) "Der zugbeanspruchte

Flachstab mit optimalem Querschnittsüber-

gang," Forschung Im Ingenieurwesen, 35, pp.

29–30 (in German).

23. S.D.P. Benson and B.L. Karihaloo (2005)

"CARDIFRC® -Development and mechanical

properties. Part III: Uniaxial tensile response

and other mechanical properties," Magazine of

Concrete Research, 57, pp. 433–443.

24. Technical Leaflet SIA 2052 UHPFRC – Mate-

rials, design and construction, March 2016 (in

German and French).

25. K. Habel; M. Viviani; E. Denarié; and E.

Brühwiler (2006) "Development of the me-

chanical properties of an Ultra-High Perfor-

mance Fiber Reinforced Concrete (UHP-

FRC)," Cement and Concrete Research, 36, pp.

1362–1370.

78




ISSN 2465-7964 / eISSN 2465-7972

https://doi.org/10.18702/acf.2019.1.4.2.79

Technical Paper

Deformation mechanism of hardened cement paste under

high stress and application of flow law

Yuya Sakai*

(Received: June 06, 2018; Accepted: November 02, 2018; Published online: January 04, 2019)

Abstract: In this study, a creep test was performed on hardened cement paste (HCP) with stepwise load

increase at different confining pressures and saturation degrees. The strain rate–stress relationship, obtained

under a stress of >75% of the maximum strength and plotted on a log–log chart, showed a slope of six. From

previous studies on crystalline materials, such as rock, metal, and ice, it can be inferred that this slope indi-

cates a deformation governed by dislocation creep. If dislocation creep occurs in HCP, the deformation may

be governed by crystalline hydrates other than calcium silicate hydrate (C-S-H) because dislocation creep

is generally not defined for gel materials. Further study and careful discussion are required because a slope

of six is a necessary condition for the dislocation creep. The activation volume was evaluated, and the flow

law was applied to calculate the strain rate of HCP. The obtained activation volume gives a better fit for the

measured results than the previously reported values.

Keywords: cement paste, triaxial test, flow law, dislocation creep.

1. Introduction

Concrete is one of the most important construc-

tion materials in the world and has been used for a

long time. However, the mechanism of concrete de-

formation has not yet been fully understood. For ex-

ample, creep deformation was reported more than

100 years ago [1], but its mechanism is still a subject

of discussion. The main mechanisms of creep defor-

mation proposed so far are based on microcrack for-

mation [2], water seepage from hardened cement

paste (HCP) [3], occurrence of slip between the glob-

ules of calcium silicate hydrate (C-S-H) [4], occur-

rence of microprestress [5], etc. Gaining a clear un-

derstanding of the concrete deformation mechanism

is important for the safe and rational design and

maintenance of concrete structures. With regard to

creep deformation, it is known that the stress higher

than the threshold stress causes creep fracture, and

this threshold stress is called the sustained load

strength [6–8]. The sustained load strength usually

ranges from 70% to 80% of the maximum strength

in normal concrete [6,9,10], although higher frac-

tions (e.g. > 85%) have been reported for high-

strength concrete [11,12]. Hsu et al. [13] studied the

development of cracks in concrete and reported that

matrix cracks formed continuous crack patterns un-

der a stress of more than 70% of the maximum

strength. Because this fraction (percentage of the

sustained load strength in relation to the maximum

strength) agrees with the ratio of the applied stress to

the maximum strength, crack development may be

related to the sustained load strength; however, the

origin of the sustained load strength has not yet been

clearly explained. The ratio between the sustained

load strength and maximum strength for various

types of concrete is almost consistent, which indi-

cates that they have a common mechanism; gaining

an understanding of the deformation mechanism un-

der high stress (stress higher than the sustained load

strength) may lead to a better understanding of the

failure mechanism of concrete.

In this study, the deformation mechanism of

HCP was studied by performing triaxial tests. The

results showed that a shear plane was not formed in

HCP that was subjected to a certain confining pres-

sure, even though the HCP deformed upon the appli-

cation of up to 10% strain [14]. Based on this result,

it was inferred that the HCP shows plastic flow when

subjected to a certain confining pressure. The plastic

flow of HCP was clearly observed during the com-

paction of crushed HCP [15]. The mechanism of the

flow was studied by performing a creep test with a

stepwise increase in the load. Furthermore, the strain

rate was calculated assuming that the HCP reached a

static state within 20 min after the load increase [14].

Corresponding author Yuya Sakai is an Assistant Pro-

fessor of Institute of Industrial Science, The University

of Tokyo, Tokyo, Japan.

79

However, this assumption was not appropriate be-

cause HCP requires a longer time to reach a static

state under sustained load [6]. The obtained results

were not consistent and not easy to analyze or inter-

pret. Moreover, a model to describe the deformation

of HCP was not proposed. The strain rate due to plas-

tic deformation is described by the flow law that is

often applied to inorganic materials, such as rocks,

ceramics and metals [16,17]. However, to the best of

our knowledge, the application of the flow law to

HCP has not been studied yet.

Therefore, in this study, another analysis

method was applied to the results of the stepwise

creep test, additional experiments were performed to

apply the flow law, and a quantitative discussion was

included on the deformation mechanism of HCP.

First, a loading test was performed at a constant

strain rate to obtain the maximum strength. The sam-

ples were cut after the test and the cross-sectional

surfaces were observed. Subsequently, a creep test

was performed with a stepwise increase in the load

corresponding to 30–95% of the maximum strength.

The deformation mechanism of the HCP was dis-

cussed based on the obtained strain rate–stress rela-

tionship. The activation volume was evaluated, and

the flow law was applied to describe the deformation

of the HCP under high stress.

2. Methodology

2.1 Sample preparation

In this research, cement paste (water-to-cement

ratio = 0.4) made from ordinary Portland cement was

used. The properties of the cement are presented in

Tables 1 and 2. The mixing procedure was based on

JIS R 5201. The paste was first mixed for 60 seconds

in a mixer operating at a low speed (orbital rotation:

62 ± 5 rpm, planetary rotation: 140 ± 5 rpm). The

mixer was stopped for 30–60 seconds to scrape off

cement paste on the sides of the mixing bowl and

paddle. Then, the paste was mixed for 90 seconds at

a high speed (orbital rotation: 125 ± 5 rpm, planetary

rotation: 285 ± 5 rpm). The mixed paste was cast in

a plastic mould (250 × 150 × 100 mm) and sealed. It

was demoulded 24 hours after casting and then kept

under water for two months. The temperature of the

room and water was 24 °C. After curing, cylinders of

φ10 mm were cored from the HCP mass. Only the

part deeper than 2 cm from the surface was used.

Both ends of the cylinder were ground to achieve flat

and parallel surfaces. The prepared cylinders (φ10 ×

24 mm) were immersed into acetone for 24 hours to

stop the hydration reaction and reduce capillary suc-

tion in the subsequent drying period. After immer-

sion, the cylinders were dried in a desiccator at 24 °C

and 20% RH until the weight change over 24 hours

because of moisture loss was less than 0.1% of the

specimen weight. Saturated samples were prepared

by immersing the cylinders into tap water after dry-

ing at 24 °C and 20% RH until the weight change

over 24 hours was less than 1% of the specimen

weight. The specimen names are composed of the al-

phabets D or W followed by numbers (e.g. D0, W50);

D and W indicate dry and saturated samples, respec-

tively, and the number indicates the confining pres-

sure Pc. Assuming that the saturation degrees of the

saturated and oven-dried (at 105 °C) samples were

unity and 0, respectively, the calculated saturation

degree of a sample dried at 24 °C and 20% RH was

0.31. Considering that HCP was immersed in ace-

tone, this saturation degree (0.31) was a mixture of

water and acetone. The testing age varied in the

range of 5–6 months. The carbonation depth of the

sample, which was kept in the desiccator (24 °C and

20% RH) after all tests, was measured using a 1%

solution of phenolphthalein in ethyl alcohol. The

sample was split horizontally at the middle using a

chisel, and the solution applied on the fracture sur-

face indicated that the sample had a carbonation

depth of 0.5 mm. The porosity of the sample, which

was calculated using the following equation, was

found to be 0.39.

)()(1 watersatwateroven WWWW (1)

where φ is the porosity, Woven is the oven-dried

weight, Wwater is the saturated weight under wa-

ter, and Wsat is the saturated weight in the air.

Table 1 – Chemical composition of cement

Chemical property (%)

Ig. loss Insol. SiO2 Al2O3 Fe2O3 CaO MgO SO3 Na2O K2O

2.84 0.27 19.83 4.56 2.95 61.28 3.78 2.83 0.35 0.57

Table 2 – Physical properties of cement

Density (g/cm3) Specific surface area (cm2/g)

3.12 4,110

80


2.2 Triaxial test

2.2.1 Test apparatus and sample assembly

The Paterson-type [18] triaxial test apparatus

was used with argon gas as the pressure medium to

generate confining pressure on the sample. The sche-

matic view of the apparatus is shown in Fig. 1. A

sample was placed between alumina and zirconia

pistons enclosed in a heat-shrinkable (polyolefin)

tube that sealed the sample from the confining me-

dium. The end face of the HCP sample was exposed

to the laboratory atmosphere through small centre

holes in the spacers and pistons to avoid pore pres-

sure being generated during experiments. To achieve

shrinkage, the tube was heated from the outside us-

ing a heat gun. To check the temperature of the HCP

specimen during the heat treatment on the tube, a 2-

mm-diameter hole was created on the HCP sample

and the temperature in the hole was measured using

a needle probe thermometer [19]. The distance from

the sample surface to the hole was 2 mm. A heat-

shrinkable tube was placed over the HCP sample and

heated until the tube touched the sample completely.

The maximum temperature reached was 46 °C;

therefore, the heat might have some effect near the

surface of the sample during this heating process. A

saturated sample was wrapped with a plastic film be-

fore being covered with a heat-shrinkable tube. Steel

anvils were then attached to the top and bottom of

the zirconia pistons and fixed with steel wires over

the heat-shrinkable tube. Figure 2 shows a photo-

graph of the sample assembly. The prepared sample

assembly was inserted into the pressure vessel of the

test apparatus. Load was applied by moving the

lower piston upward. The stress was calculated by

dividing the load measured using the internal load

cell by the cross-sectional area of the sample. The

apparatus was designed such that the confining pres-

sure did not affect the stress measured by the internal

load cell; therefore, the calculated stress corre-

sponded to the differential stress. The axial strain

was calculated by dividing the displacement meas-

ured using the transducer located below the sample

assembly by the initial length of the sample. The load,

displacement, and confining pressure were measured

at intervals of 1 second. All the tests were performed

at room temperature (24 °C).

2.2.2 Constant strain rate loading test

Triaxial tests with constant strain rate loading

were performed at 1.4 × 10−4 s−1. Two dry samples

were tested at each value of Pc; one of the two sam-

ples was loaded until the strain reached 10%, and the

Fig. 1 – Schematic diagram of triaxial test-ing apparatus

Specimen

Furnace

Internal load cell

Pressure compen-sating piston

Up and down move-ment driven by actua-tor

Vent

Confining pressure

Fig. 2 – Sample assembly

Specimen

Alumina piston

Zirconia piston

Anvil

Alumina piston

Zirconia piston

Anvil

81


other one was loaded until the stress started decreas-

ing after reaching the maximum stress (σmax). Sam-

ples tested at Pc = 0 MPa were loaded until the stress

decrease stagnated.

After the triaxial tests with constant strain rate

loading, the sample covered with a heat-shrinkable

tube was cut out from the assembly and impregnated

into a two-component epoxy resin (room-tempera-

ture curing). The sample was then cut parallel to the

long axis when the epoxy resin hardened. The cutting

surface was polished using alumina powder (average

diameter = 55 and 15 μm) and observed by the naked

eye.

2.2.3 Stepwise creep test

A stepwise creep test was performed to obtain

strain rate–stress relationship. In this test, the applied

stress was increased stepwise from 30% to 95% of

σmax. The stress was increased every 30–60 minutes.

The strain rate–stress relationship is known to show

more internal consistency in the stepwise creep test

than in the conventional creep test [20, 21]. In a pre-

vious study, the strain rate was calculated when the

slope of the strain–time curve was regarded to be in

a static state, but the slope actually kept decreasing

and did not reach a static state [14]. Thus, as in the

study by Zhang and Spiers [22], who studied the

compaction mechanism of calcite powder, the strain

rates were calculated from the slope of the strain in-

crement-time relationship at certain strain values.

The strain rate of D0 was calculated using data ob-

tained 1000–2000 seconds after each value of stress

was attained, and that of W0 was obtained by fitting

the exponential approximate function for the entire

data set because the strain in the samples without Pc

was small. The strain rate at steady-state creep is de-

scribed by the following flow law [23]:

(2)

where 𝜀 is the strain rate (s−1), A is a constant

(mm/Pan), σ is the differential stress (Pa), n is the

stress exponent, d is the grain size (m), m is the

grain size exponent, Q is the activation energy

(J/mol), P is the pressure (Pa), V is the activation

volume (m3/mol), R is the gas constant (= 8.3

J/K/mol), and T is the absolute temperature (K).

The slope of the strain rate-stress relationship

corresponds to n.

2.3 Hydrostatic pressure test

The value of the activation volume is necessary

for applying the flow law to the deformation of HCP.

The equation to calculate the activation volume is

derived from Eq. (2) by taking the natural logarithm

and differentiating partially with respect to P (as-

suming σ to be a constant) as follows:

(3)

(4)

(5)

To calculate V using Eq. (5), it is necessary to de-

termine 𝜀 for different P (Pc) values with σ being

constant. However, as shown later in Section 3.2, the

strain rate–stress relationship follows the flow law

only when σ is more than 75% of σmax. Because σmax

of dry samples increases with an increase in Pc, the

constant σ can be less than 75% of σmax with increas-

ing Pc and V would be calculated in the region where

the data do not follow the flow law. Therefore, V was

calculated by measuring the strain rate of a saturated

sample for different Pc values with a constant σ (92

MPa, 80% of σmax in W20) because σmax of the satu-

rated sample hardly depends on Pc, as shown later in

Section 3.1. For the saturated sample, Pc was first set

to 20 MPa, σ corresponding to 80% of σmax was ap-

plied, and Pc was then increased by 10 MPa up to 50

MPa to obtain the ln 𝜀–Pc relationship.

3. Results

3.1 Constant strain rate loading test

The differential stress–strain curves of the dry

samples are shown in Fig. 3(a). The results for the

samples loaded until ε = 10% and those loaded until

the stress started decreasing are represented by bro-

ken lines and solid lines, respectively. The two sam-

ples with the same Pc showed similar results. D0

showed a sudden stress decrease after σmax. σmax in-

creased with an increase in Pc. The stress kept in-

creasing until ε = 10% for Pc = 100 MPa, but the

stiffness decreased in the low-strain region. Figure

3(b) shows the results for the saturated samples.

Compared to the dry samples tested at the same Pc,

σmax for the saturated samples decreased. W20 and

W50 showed similar curves.

Figure 4 shows the cut and polished surfaces of

the samples after the triaxial tests. Except for D0, the

left image shows the sample loaded up to 10% strain

and the right image shows the sample loaded until

the stress started decreasing. D30 shows only the

sample loaded up to 10% strain. The black lines in

the samples are the epoxy resin that penetrated

through the cracks. Vertical cracks developed in D0,

and shear planes were formed in D10, D30, and D50.

A shear plane was not formed in D100. 3.2 Stepwise creep test

The obtained strain increment–time curves are

shown in Fig. 5. The values in the legends indicate

stress (the values in parentheses indicates the ratio of

82


the applied stress to σmax). The values in parentheses

at the upper right corner are the strain values at which

𝜀 was calculated. Figure 6 shows the relationship be-

tween 𝜀 and σ, with the lines indicating slopes of

three and six. The data for stress values less than 0.75

× σmax are shown as plots represented by short lines,

and the data for stress values equal to or greater than

0.75 × σmax are shown as plots represented by circles.

The plots represented by the short lines are scattered,

whereas the plots represented by the circles are dis-

tributed linearly, except for D0, and have slopes

close to six.

3.3 Hydrostatic pressure test

Figure 7 shows the strain–time relationship and

ln 𝜀–Pc relationship of a saturated sample in the hy-

drostatic pressure test. According to Eq. (5), the

slope of the plots in Fig. 7(b), −3 × 10−8, multiplied

by -RT is equal to V, and the calculated V for the sat-

urated samples was 7.4 × 10−5 m3/mol.

4. Discussion

4.1 Deformation mechanism of HCP under high

stress

The dry samples showed stress decrements, and

vertical cracks or shear planes were formed when Pc

was 0–50 MPa. When Pc was 100 MPa, the stress

increased up to 10% strain and no shear plane was

formed. In rock mechanics, fractures at lower strains

with vertical cracks or shear planes are considered

brittle, whereas those at higher strains without shear

planes are considered ductile [18]. Following this

classification, the HCP deformation changed from

brittle to ductile with an increase in Pc. These results

are consistent with those obtained in a previous study

[14]. The horizontal crack in the sample shown on

the right side of Fig. 4(d) might have been formed

during unloading.

As seen in Fig. 3(a), the stress did not decrease

when Pc was 100 MPa; a similar tendency has been

reported for concrete [24]. However, HCP showed a

decrease in stress in the low-strain region, whereas

concrete did not. A possible reason for this differ-

ence is that the aggregate and concrete were in con-

tact with each other. Consequently, the aggregate

generated resistance that prevented a decrease in

stress. In porous rocks, stress decreases due to mi-

croscopic damage has been reported [25,26], and a

similar phenomenon might have occurred in HCP. In

Figure 6, the strain rate–stress relationship at stress

values greater than 0.75 × σmax is seen to be distrib-

uted linearly and the slope is approximately six. This

slope corresponds to n in Eq. (2).

Fig. 3 – Differential stress–axial strain relationship with constant strain rate loading

Axial strain (%)

Dif

fere

nti

al s

tres

s (M

Pa)

(a) Dry specimens (b) Saturated specimens

D0 D10

D50

D100

D30

W50

W0

W20

Dif

fere

nti

al s

tres

s (M

Pa)

Axial strain (%)

(a) D0 (b) D10 (c) D30 (d) D50 (e) D100

Fig. 4 – Cross section of dry specimens after constant strain rate loading test

83


The creep mechanisms of crystalline materials

are classified according to the value of n: n = 1 is

considered to represent diffusion creep (creep defor-

mation governed by atomic diffusion), and n > 1 is

considered to represent dislocation creep (creep de-

formation governed by dislocation movement)

[26,27]. Bürgmann and Dresen [28] classified n = 3–

6 as creep governed by dislocation climb. Various

crystalline materials such as metal [29] and ice [30]

have similar equations and classifications. Thus, the

slope of six seen in Fig. 6 indicates that the defor-

mation of HCP for stress values larger than 0.75 ×

σmax is governed by dislocation climb. The same

slopes for the dry and saturated samples indicate that

the deformation mechanism does not change accord-

ing to the saturation degree. Concrete generally fails

when it is subjected to a stress larger than a sustained

load strength that is approximately 75% of the max-

imum strength [6,9,10]. This fraction corresponds to

the threshold stress above which data are distributed

with the slope of six in Fig. 6. This correspondence

indicates that the creep failure in concrete may be at-

tributed to deformation due to dislocation creep.

Many materials show plastic deformation, which is

sometimes followed by brittle fracture. In addition,

the interaction between cracks and dislocation plays

an important role in plastic deformation [31]. There-

fore, the propagation of cracks in HCP at stress val-

ues of more than 70% of the maximum strength may

be affected by dislocations. However, it should be

noted that the slope of six is a necessary, but not suf-

ficient, condition for dislocation creep.

D50 (0.08, 0.12, 0.16, 0.20)

W0

W20 (0.04, 0.08, 0.12)

D0

44 MPa (30%) 73 MPa (50%) 88 MPa (60%)

103 MPa (70%) 110 MPa (75%) 117 MPa (80%)

Ax

ial

stra

in i

ncr

emen

t (%

)

57 MPa (50%) 74 MPa (65%) 80 MPa (70%) 86 MPa (75%) 92 MPa (80%) 97 MPa (85%)

103 MPa (90%) 109 MPa (95%)

109 MPa (50%) 131 MPa (60%) 153 MPa (70%) 164 MPa (75%) 175 MPa (80%) 186 MPa (85%) 197 MPa (90%) 208 MPa (95%)

18 MPa (30%) 31 MPa (45%) 37 MPa (55%) 43 MPa (65%) 49 MPa (70%) 52 MPa (75%) 55 MPa (80%) 58 MPa (85%) 61 MPa (90%) 64 MPa (95%)

Fig. 5 – Strain increment–time relationship in stepwise creep test

W50 (0.04, 0.08, 0.12, 0.16)

34 MPa (30%) 58 MPa (50%) 69 MPa (60%) 81 MPa (70%) 87 MPa (75%) 93 MPa (80%) 98 MPa (85%)

104 MPa (90%)

Time after reaching certain stress (s) Time after reaching certain stress (s)

Time after reaching certain stress (s)

Time after reaching certain stress (s) Time after reaching certain stress (s)

Ax

ial

stra

in i

ncr

emen

t (%

)

Ax

ial

stra

in i

ncr

emen

t (%

)

Ax

ial

stra

in i

ncr

emen

t (%

)

Ax

ial

stra

in i

ncr

emen

t (%

)

84


4.2 Application of flow law

The flow law (Eq. (2)) was applied to the ob-

tained results. Figure 6 shows the results for n = 6

and m = 0 for the dislocation creep. The value of Q

was set as 35,000 J/mol [39,32] for the dry sample

and 17,500 J/mol for the saturated sample [31–33]

based on the results for rocks. The obtained V for the

saturated sample was 7.4 × 10−5 m3/mol. Sammis et

al. [36] reported the Q and V values for various ma-

terials; the values obtained in this study were close

to those of the ionic crystals. The value of V for the

dry sample was also assumed to be 7.4 × 10−5 m3/mol

because a similar value was previously reported for

dry and wet rocks [33–35]. The value of V reported

by Klug and Wittmann [37] was 1 × 10−20 cm3 (6 ×

10−3 m3/mol), which was approximately 100 times

larger than the one obtained in this study. For fitting

the calculation results to the measured data, A was

set to 1 × 10−23 for the dry samples and 5 × 10−15 for

the saturated samples. The calculated curves using

these parameters are shown in Fig. 8(a) along with

the measured data for stress values larger than 0.75

× σmax; the calculated and measured data are in good

agreement. Figure 8(b) shows the calculated curves

obtained using the value of V reported by Klug and

Wittmann [37]. The value of A was adjusted to ob-

tain the best possible fit for the data (A = 1 × 1015 for

dry and saturated samples) but the calculated curves

varied significantly depending on Pc and deviated

from the measured data. In Equation (2), V expresses

the dependency of 𝜀 on Pc; therefore, the value of V

reported by Klug and Wittmann [37] is likely to be

too large to describe the deformation of HCP under

high stress based on the flow law.

4.3 Contribution of C-S-H to deformation of

HCP

The strain rate–stress relationship of HCP in Fig.

6 indicates that the deformation of HCP was gov-

erned by dislocation similar to the case of crystalline

materials such as rock [27,28], metal [29], and ice

[30]. In concrete, dislocation was reported to occur

in calcium hydroxide because of the stress caused by

drying shrinkage [38]. However, 50–60% of the

HCP volume is composed of C-S-H, which is gener-

ally regarded as a gel. Klug [39] and Klug and Witt-

mann [37] discussed the creep deformation of HCP,

assuming the deformation of the amorphous solid

skeleton. However, the plastic deformation mecha-

nism of an amorphous material, such as metallic

glass, is explained by the shear band formation or lo-

cal atomic jump [41,42]. In both these mechanisms,

the strain rate is a hyperbolic function of stress, and

the value of n changes from 1 to a very large number.

The consistent slope of six obtained in this study in-

dicates that the deformation mechanism of HCP is

closer to that of a crystalline material than an amor-

phous material. The time-dependent response of C-

S-H has been studied using various approaches, and

recently, atomistic simulation has appeared as a

powerful tool for its investigation. Morshedifard et

al. [40] carried out a molecular dynamics simulation

to study the time-dependent response of C-S-H, and

they reported a behavior often seen in glassy systems.

In addition, these authors reported that the amount of

interlayer water changes the time-dependent re-

sponse of C-S-H. If the response of C-S-H under

high stress is similar to metal glass and if dislocation

creep occurs in HCP, crystalline hydrates such as

calcium hydrates or ettringite may govern the defor-

mation under high stress. As these crystalline hy-

drates are not dominant in volume, they likely form

a skeleton to resist the applied load, and the defor-

mation of this structure may be governed by the dis-

location creep [43]. The calcium hydroxide crystal is

large, and its aspect ratio is high. Generally, to form

a skeleton, the element volume fraction needs to be

more than about 16% [44,45]. However, when the

aspect ratio is high, the required volume fraction de-

creases significantly [46,47], and a skeleton is

formed more easily. The volume fraction of the hy-

drates other than C-S-H is around 40%, which is suf-

ficiently large for the hydrates to form a structure.

Further study is necessary to conclude the defor-

mation mechanism.

5. Conclusion

In this study, a stepwise creep test was per-

formed to understand the deformation mechanism of

hardened cement paste. Subsequently, the flow law

was applied to the obtained test results. The follow-

ing conclusions were derived:

(1) The maximum strength of the saturated samples

was lower than that of the dry samples under

any given confining pressure.

(2) The strain rate and differential stress relation-

ship obtained in the creep test with a stepwise

increase in the load showed a slope of six on a

log–log chart. This slope indicates that the de-

formation is governed by dislocation creep.

(3) Dislocation creep might have possibly caused

the fracture of concrete when the stress was

larger than the sustained load strength.

(4) The calculated strain rate based on the flow law

showed that the obtained activation volume was

reasonable whereas the one reported by a previ-

ous study was too large.

85


(5) Because 50–60% of the cement paste volume is

composed of the gel hydrate C-S-H, dislocation

cannot be defined in a gel. Other crystalline hy-

drates could likely govern the deformation, simi-

lar to the deformation governed by dislocation

creep.

The results of this study indicate that dislocation

might play an important role in the deformation of

hardened cement pastes under high stress. If this is

true, dislocation may govern various failure patterns

of concrete, such as fatigue. Further study and care-

ful discussion in this regard are necessary because

the slope of six in the strain rate–stress relationship

is a necessary but not a sufficient condition for dis-

location creep.

Str

ain

rat

e (s

-1)

Differential stress (MPa)

1

3

1

6

W50

D0 W0

D50

W20

Fig. 6 – Strain rate–differential stress relationship in stepwise creep test

Str

ain

(%

)

Time (s)

Fig. 7 – Strain–time and strain rate–confining pressure relationships in hydrostatic pressure test

Confining pressure (MPa)

ln (

Str

ain

rat

e) (

s-1)

Pc = 50 MPa

Pc = 20 MPa

Pc = 40 MPa

Pc = 30 MPa

(a) Strain–time relationship (b) Strain rate–confining pressure relationship

Str

ain

rat

e (s

-1)


Fig. 8 – Strain rate–differential stress relationship in stepwise creep test along with calculated strain rate based on flow law

W50

D0

W0

D50

W20

Str

ain

rat

e (s

-1)


W50

D0

W0

D50

W20

(a) V = 7.4 × 10−5 m3/mol (b) V = 6 × 10−3

m3/mol

86


Acknowledgements

This study was supported by the Foundation for

the Promotion of Industrial Science, Tokyo, Japan.

The triaxial tests in this study was performed using

the apparatus owned by the Department of Earth, En-

vironmental, and Planetary Science, Brown Univer-

sity, Rhode Island, USA.

References

1. Hatt, W.K. (1907) “Notes on the effect of time

element in loading reinforced concrete beams,”

Proceedings ASTM, 7, pp. 421~433.

2. Pickett, G. (1942) “The effect of change in

moisture content on the creep of concrete under

a sustained load,” Journal of American Con-

crete Institute, 38, pp. 333~355.

3. Powers, T.C. (1968) “The thermodynamics of

volume change and creep,” Matériaux et Con-

struction, 1(6), pp. 487~507.

4. Ali, I. and Kesler, C. (1964) “Mechanisms of

creep in concrete,” Symposium on Creep of

Concrete, Detroit, ACI Publication SP-9, pp.

35~63.

5. Bazant, Z.P.; Hauggaard, A.B.; Baweja, S.; and

Ulm, F.J. (1997) “Microprestress-solidification

theory for concrete creep. I: Aging and drying

effects,” Journal of Engineering Mechanics,

123(11), pp. 1188~1194.

6. Rüsch, H. (1960) “Researches toward a general

flexural theory for structural concrete,” Journal

of American Concrete Institute, 57(1), pp. 1~28.

7. Carpinteri, A.; Valente, S.; Zhou, F.P.; Ferrara,

G.; and Melchiorri, G. (1997) “Tensile and

flexural creep rupture tests on partially dam-

aged concrete specimens,” Materials and Struc-

tures, 30(5), pp. 269~276.

8. Ranaivomanana, N.; Multon, S.; and Tu-

ratsinze, A. (2013) “Tensile, compressive and

flexural basic creep of concrete at different

stress levels,” Cement and Concrete Research,

52, pp. 1~10.

9. Ngab, A.S.; Nilson, A.H.; and Slate, F.O. (1981)

“Shrinkage and creep of high strength concrete,”

Journal of American Concrete Institute, 78(4),

pp. 255~261.

10. Smadi, M.M.; Slate, F.O.; and Nilson, A.H.

(1982) “Time-dependent behavior of high-

strength concrete under high sustained com-

pressive stresses,” Research Report No. 82 –16,

Department of Structural Engineering, Cornell

University, Ithaca, NY.

11. Smadi, M.M.; Slate, F.O.; and Nilson, A.H.

(1985) “High-, medium-, and low-strength

concretes subject to sustained overloads-strains,

strengths, and failure mechanisms,” Journal of

American Concrete Institute, 82(5), pp.

657~664.

12. Iravani, S. and MacGregor, J.G. (1994) “High

performance concrete under high sustained

compressive stresses,” Structural Engineering

Report No. 200, Department of Civil Engineer-

ing, University of Alberta.

13. Hsu, T.T.C; Slate, F.O.; Sturman, G.M.; and

Winter, G. (1963) “Microcracking of plain con-

crete and the shape of the stress-strain curve,”

Journal of American Concrete Institute, 60(2),

pp. 209~224.

14. Sakai, Y and Kishi, T. (2017) “Deformation

mechanism of hardened cement paste and ef-

fect of water,” Journal of Advanced Concrete

Technology, 15(1), pp. 19~28.

15. Sakai, Y.; Tarekegne, B.T.; and Kishi, T. (2016)

“Recycling of hardened cementitious material

by pressure and control of volumetric change,”

Journal of Advanced Concrete Technology,

14(2), pp. 47~54.

16. Goetze, C. (1978) "The mechanisms of creep in

olivine," Philosophical Transactions of the

Royal Society of London. Series A, Mathemat-

ical and Physical, 288(1350), pp.99-119.

17. Frost, H.J. and Ashby, M.F. (1982) "Defor-

mation-mechanism Maps: The Plasticity and

Creep of Metals and Ceramics," Pergamon

Press.

18. Paterson, M. and Wong, T. (2005) “Experi-

mental Rock Deformation - The Brittle Field,”

Springer Science & Business Media, Berlin.

19. Sakai, Y.; Nakatani, M.; Takeuchi, A.; Omorai,

Y.; and Kishi, T. (2016) “Mechanical behavior

of cement paste and alterations of hydrates un-

der high-pressure triaxial testing,” Journal of

Advanced Concrete Technology, 14(1), pp.

1~12.

20. Heap, M.J.; Baud, P.; Meredith, P.G.; Bell,

A.F.; and Main, I.G. (2009) “Time-dependent

brittle creep in Darley Dale sandstone,” Journal

of Geophysical Research: Solid Earth, 114,

B07203.

21. Brantut, N.; Heap, M.J.; Meredith, P.G.; and

Baud, P. (2013) “Time-dependent cracking and

brittle creep in crustal rocks: A review,” Jour-

nal of Structural Geology, 52, pp. 17~43.

22. Zhang, X. and Spiers, C.J. (2005) “Compaction

of granular calcite by pressure solution at room

temperature and effects of pore fluid chemistry,”

International Journal of Rock Mechanics and

Mining Sciences, 42, pp. 950~960.

23. Jackson, I. (2000) “The Earth's Mantle: Com-

position, Structure, and Evolution,” Cambridge

University Press, Cambridge.

24. Sfer, D.; Carol, I.; Gettu, R.; and Etse, G. (2002)

“Study of the behavior of concrete under triax-

ial compression,” Journal of Engineering Me-

chanics, 128(2), pp. 156~163.

87


25. Lubarda, V.A.; Mastilovic, S.; and Knap, J.

(1996) “Brittle-ductile transition in porous

rocks by cap model,” Journal of Engineering

Mechanics, 122(7), pp. 633~642.

26. Zheng, Z.; McLennan, J.D.; and Martin, J.W.

(1991) “Compressive stress-induced mi-

crocracks and effective elastic properties of

limestone and concrete,” Final Technical Re-

port to USAF, Topic No. AF90-186, Terra Tek

Report TR91-107.

27. Mei, S. and Kohlstedt, D.L. (2000) “Influence

of water on plastic deformation of olivine ag-

gregates: 2. Dislocation creep regime,” Journal

of Geophysical Research, 105(B9), pp.

21471~21481.

28. Bürgmann, R. and Dresen, G. (2008) “Rheol-

ogy of the lower crust and upper mantle: Evi-

dence from rock mechanics, geodesy, and field

observations,” Annual Review of Earth and

Planetary Sciences, 36, pp. 531~567.

29. Sato, E.; Yamada, T.; Tanaka H.; and Jimbo, I.

(2005) “Categorization of ambient temperature

creep behavior of metals and alloys on their

crystallographic structures,” Journal of Japan

Institute of Light Metals, 55(11), pp. 604~609.

30. Goldsby, D.L. and Kohlstedt, D.L. (1995) “The

transition from dislocation to diffusion creep in

ice,” Abstracts of the Lunar and Planetary Sci-

ence Conference, 26, pp. 473~474.

31. Ohr, S.M. (1987) “Dislocation-crack interac-

tion,” Journal of Physics and Chemistry of Sol-

ids, 48(11), pp. 1007~1014.

32. Riding, K.A. and Siddiqui, S. (2012) “Effect of

calculation methods on cement paste and mor-

tar apparent activation energy,” Advances in

Civil Engineering Materials, 1(1), pp. 1~19.

33. Karato, S.I. and Wu, P. (1993) “Rheology of

the upper mantle: A synthesis,” Science,

260(5109), pp. 771~778.

34. Karato, S.I. and Jung, H. (2003) “Effects of

pressure on high-temperature dislocation creep

in olivine,” Philosophical Magazine, 83(3), pp.

401~414.

35. Rybacki, E.; Gottschalk, M.; Wirth, R.; and

Dresen, G. (2006) “Influence of water fugacity

and activation volume on the flow properties of

fine-grained anorthite aggregates,” Journal of

Geophysical Research: Solid Earth, 111(B3),

B03203.

36. Sammis, C.G.; Smith, J.C.; and Schubert, G.

(1981) “A critical assessment of estimation

methods for activation volume,” Journal of Ge-

ophysical Research: Solid Earth, 86(B11), pp.

10707~10718.

37. Klug, P. and Wittmann, F. (1974) “Activation

energy and activation volume of creep of hard-

ened cement paste,” Materials Science and En-

gineering, 15(1), pp. 63~66.

38. Harutyunyan, V.S.; Abovyan, E.S.; and Mon-

teiro, P.J.M. (2003) “X-ray diffraction investi-

gations of deformations and dislocation con-

Figureuration in calcium hydroxide crystallites

of concrete,” Physica Status Solidi, 200(2), pp.

307~325.

39. Klug, P. (1969) “Activation energy of creep of

hardened cement paste,” Matériaux et Con-

struction, 2(1), pp. 11~16.

40. Morshedifard, A.; Masoumi, S.; and Abdolhos-

seini Qomi, M.J. (2018) “Nanoscale origins of

creep in calcium silicate hydrates,” Nature

Communications, 9, 1785.

41. Schuh, C.A.; Hufnagel, T.C.; and Ramamurty,

U. (2007) “Mechanical behavior of amorphous

alloys,” Acta Material, 55(12), pp. 4067~4109.

42. Kassner, M.E.; Smith, K.; and Eliasson, V.

(2015) “Creep in amorphous metals,” Journal

of Materials Research and Technology, 4(1),

pp. 100~107.

43. Sakai, Y. and Uehara, Y. (2018) “Deformation

mechanism of cement paste and hydrates under

high stress,” Journal of Advanced Concrete

Technology, 16(6), pp. 262~271.

44. Scher, H. and Zallen, R. (1970) “Critical Den-

sity in Percolation Processes,” The Journal of

Chemical Physics, 53(9), pp. 3759~3761.

45. Bentz, D.P. and Garboczi, E.J. (1991) “Perco-

lation of phases in a three-dimensional cement

paste microstructural model,” Cement and

Concrete Research, 21(2), pp. 325~344.

46. Garboczi, E.J.; Snyder, K.A.; Douglas, J.F.;

and Thorpe, M.F. (1995) “Geometrical perco-

lation threshold of overlapping ellipsoids,”

Physical Review E, 52(1), pp. 819~828.

47. Saar, M.O. and Manga, M. (2002) “Continuum

percolation for randomly oriented soft-core

prisms,” Physical Review E, 65(5), 056131.

88




ISSN 2465-7964 / eISSN 2465-7972

https://doi.org/10.18702/acf.2019.1.4.2.89

Technical Paper

Strength, shrinkage and creep of concrete including CO2

treated recycled coarse aggregate

Gombosuren CHINZORIGT, Donguk CHOI*, Odontuya ENKHBOLD, Batzaya BAASANKHUU,

Myung Kwan LIM and Hee Seob LIM

(Received: August 15, 2018; Accepted: December 22, 2018; Published online: January 04, 2019)

Abstract: Strength development and long-term behaviour of recycled aggregate concrete (RAC) were inves-

tigated. Recycled coarse aggregate (RA) replaced natural coarse aggregate (NA) by 30%, 50%, and 100% by

vol. in this study while natural fine aggregate was always used. For the investigation of strength development,

3, 7, 28, 56, 90, and 180 day compressive strengths and 28 day split tensile strength and flexural strength have

been determined. Shrinkage and creep behaviours of RAC with 0%, 50%, and 100% RA replacing NA have

been studied up to 6 months. The compressive strength of RAC was not affected by 30% replacement, but it

was reduced by 50% and 100% replacement. Split tensile strength was not affected significantly while flexural

strength was reduced with increasing amount of replacement. Shrinkage strain of RAC with 50% RA was

similar to that of natural aggregate concrete, but shrinkage increased with 100% replacement. Specific creep

of RAC with 100% RA increased by 38% over that of natural aggregate concrete. The strength of concrete

with CO2 treated RA was lower than that of concrete with RA without CO2 treatment. Shrinkage of RAC with

and without CO2 treatment was similar, while the creep of RAC with CO2 treated RA was smaller than that of

RAC including RA without CO2 treatment.

Keywords: recycled coarse aggregate, recycled aggregate concrete, strength, creep, shrinkage, carbonation.

1. Introduction

In South Korea, an economic boom has started

in late 1960s about 50 years ago and many building

and civil engineering infrastructures have been built

starting from the late 1960s. The amount of construc-

tion and demolition waste (C&DW) generation is

huge and takes about 50% of national waste genera-

tion, primarily due to demolition of old structures

constructed during the economic boom period. An-

nual C&DW generation was 68 million tonnes in

2011, where the waste concrete took about 65% fol-

lowed by waste asphalt concrete (19%), mixed waste

(10%), and others [1].

The effective reutilization of waste concrete

typically includes use as road subbase material,

secondary product such as bricks and blocks as well

as recycled aggregates. Although the environmental

impact for the reutilization varies depending on the

end product, the most effective way to reutilize waste

concrete is in the form of recycled aggregate espe-

cially from the view point of resource conservation.

It should be noted that the use of structural quality

recycled aggregate is not wide spread practice yet:

Only 1-2% of waste concrete is used as structural

grade recycled aggregate in South Korea for example.

There are many different standards that regulate

the quality of recycled aggregate in different coun-

tries. For example, in Europe, only recycled coarse

aggregate is accepted [2]. In South Korea and Japan,

both recycled fine aggregate and recycled coarse ag-

gregate are used [3-6]. Despite complicated modern

day production technology of recycled aggregate in-

cluding multiple-stage crushing, it is not possible to

completely remove old mortar adhered to original

natural aggregates. The adhered mortar makes the

mechanical properties of the recycled aggregate in-

ferior to those of natural aggregate, especially den-

sity and water absorption [7].

Corresponding author Donguk CHOI is a professor of Dept. of Architectural Engineering, Hankyong National University

(HKNU), Anseong, Korea.

Gombosuren CHINZORIGT is a M.S. student of Dept. of Ar-

chitectural Engineering, HKNU, Anseong, Korea. Batzaya BAASANKHUU is a M.S. student of Dept. of Archi-

tectural Engineering, HKNU, Anseong, Korea..

Odontuya ENKHBOLD is a Ph.D. student of Dept. of Civil and

Environmental Engineering, Mongolian University of Science and Technology, Ulaanbaatar, Mongolia.

Myung Kwan LIM is a professor of Dept. of Architectural En-

gineering, Songwon University, Gwangju, Korea.

Hee Seob LIM is a research professor of Dept. of Civil Engi-neering, Hannam University, Daejeon, Korea.

89

Two different approaches can be employed to

improve the mechanical properties of recycled ag-

gregate: i.e. strengthen the adhered mortar or remove

the adhered mortar. One of the strengthening ap-

proach of the adhered mortar is accelerated carbona-

tion [8-12].

There have been many studies on the long-term

behaviour of recycled aggregate concrete, but very

few on the shrinkage and creep behaviour of recycled

aggregate concrete using carbonated recycled aggre-

gate [13-20].

This study aims to investigate the following:

(1) Improvement of mechanical properties of recy-

cled coarse aggregate (RA) by accelerated car-

bonation to lower absorption and to increase

density;

(2) Monitor strength development of recycled ag-

gregate concrete (RAC) with RA replacing NA

by 30%, 50%, and 100% by vol. compared to

that of natural aggregate concrete (NAC); and

(3) Investigate the long-term properties such as

shrinkage and creep of RAC with RA replacing

NA by 50% and 100% by vol. compared to

those of NAC.

2. Materials properties and preparation

for long-term test

2.1 Aggregates and accelerated carbonation of

recycled coarse aggregate Crushed natural coarse aggregate (NA) and

recycled coarse aggregate (RA) of 25-mm nominal

size were used. Crushed natural fine aggregate (FA)

was always used. RA was supplied by a local

commercial waste concrete treatment company.

Figure 1 shows sieve analysis results of NA, RA, and

FA [21]. RA did not satisfy the density and

absorption requirements by KS F 2573, which is

density of 2,500 kg/m3 or greater (O.D.) and water

absorption of 3% or smaller. Since RA did not satisfy

the requirements in terms of density and water

absorption, it was attempted to improve the quality

of RA by accelerated carbonation following KS F

2584 [22] as the carbonation of concrete would result

in increased strength and reduced permeability [23].

The carbonation of adhered mortar can be expressed

by Eq. (1).

Ca(OH)2 + CO2 = CaCO3 + H2O (1)

RA was carbonated in a carbonation chamber

for three days (72 hours). The rate of carbonation

depends on the moisture content of the adhered

mortar and the relative humidity of the ambient

medium [23,24]. To achieve the moisture content

ideal for carbonation, the following procedure was

adopted in this study.

RA was soaked in water for 10 minutes. Dry

cloth was then used to clean up surface

moisture of the aggregates. In the next step, RA

was exposed to room environmental condition

for five hours where the temperature was 21°C

typ. and R.H. was 40%-45% typ. Moisture

content of RA was measured every one hour

during the five-hour period. As a result, the

moisture content at entry to the carbonation

chamber ranged between 63% and 67%.

During the three-day-long accelerated carbona-

tion, the temperature inside the chamber was main-

tained at 20 ± 2 °C, R.H. was 60 ± 5%, and carbon

dioxide (CO2) concentration was 5 ± 0.2%, respec-

tively, while the pressure inside the chamber was the

same as the atmospheric pressure. The carbonated

recycled coarse aggregates thus produced are called

CRA in this study.

2.2 Adhered mortar amount of RA

RA mechanical properties are dependent on

amount of adhered mortar. Pre-soaking in acid was

the method adopted in this study to determine ad-

hered mortar amount [27]. After RA was oven dried

for 24 hours, RA was soaked in 20% hydrochloric

acid (HCL) at 20°C for 24 hours and then soaked in

distilled water. Difference in weights before and af-

ter soaking was used to determine adhered mortar

amount as shown in Eq. (2):

Adhered mortar amount =

(W1 – W2)/W1 x 100, % (2)

where W1 is bulk weight of aggregate before

soaking (O.D.) and W2 = bulk weight of ag-

gregate after soaking (O.D.).

Table 1 summarizes the mechanical properties

of all aggregates determined in this study in terms of

water absorption, density, adhered mortar amount,

crushing value, and fineness modulus (F.M.). Figure

2 shows RA with adhered mortar (before soaking)

and RA without adhered mortar (after soaking).

2.3 Mix design

Volumetric concrete mix design of 100%

natural aggregate concrete (NAC) was for a target

strength of 30 MPa. Table 2 shows the mix design

and properties of fresh concrete. All aggregates were

prepared and mixed in SSD condition. Recycled

aggregate concrete (RAC) was produced by

substituting 30%, 50%, and 100% of NA with RA by

volume. Multiple test cylinders (Φ100 x 200 mm)

were made for compressive strength test and split

tensile strength test, while prismatic specimens (100

x 100 x 400 mm) were used for flexural strength test.

90


All cylindrical and prismatic specimens were

demolded one day after casting and then cured under

water until the test age. Compressive strength was

tested at 3, 7, 28, 56, 90, and 180 days. Elastic

modulus of concrete and Poisson’s Ratio were also

determined at 28 days during compressive strength

test while three replicate specimens were tested.

Split tensile strength and flexural strength were

determined only at 28 days while two replicate

specimens were tested.

2.4 Preparation for shrinkage measurement

Shrinkage tests were conducted using a pris-

matic specimen (100 x 100 x 400 mm) for five dif-

ferent concretes: NAC, RAC-50, RAC-100, CRAC-

50, and CRAC-100. One day after casting, beam

mold was removed and the specimens were brought

to inside an environmental chamber where tempera-

ture was set at 20°C and R.H. was set at 60%. A ther-

mocouple was used to measure temperature inside

the chamber while a portable hygrometer was used

to measure R.H. As the specimens were exposed to

air, shrinkage measurement started immediately us-

ing an embedded strain gauge (60-mm length) and

additional two strain gauges (60-mm length)

mounted on two side faces of the beam. Teflon sheet

was used at bottom surface of the beam to eliminate

friction between the steel base plate and the concrete

beam. Shrinkage data were taken using a data logger

connected to a computer at every 1 hour. The shrink-

age measurement continued for 180 days.

2.5 Preparation for creep test

Creep test started 35 days after casting using a

150 x 300 mm cylinder for five different concretes:

NAC, RAC-50, RAC-100, CRAC-50, and CRAC-

100. The creep test specimens were cured under wa-

ter for 35 days after which they were placed in the

same environmental chamber for the shrinkage

measurement until the end of the creep test which

continued for 150 days. Two cylinder specimens

were placed in a loading frame typ. on a 50,000-lb-

(220-kN) capacity load cell, while the applied sus-

tained load was about 30% of 28-day compressive

strength. From a companion cylinder stored right

next to the creep test specimen, shrinkage data were

also retrieved. The creep measurement was made by

means of three strain gauges per cylinder: one em-

bedded strain gauge and two strain gauges mounted

on two side faces symmetrically (same as shrinkage

measurement). The second data logger connected to

a computer was used while the data acquisition rate

was one data set at every ten minutes.

Table 1 Mechanical properties of aggregates

Aggregate type Water

absorption

(%)

Density,

SSD

(kg/m3)

Crushing

value

(%)

Adhered mor-

tar amount

(%)

F.M.

Coarse aggregate NA 0.48 2,690 17.4 -- 7.3

RA 3.84 2,430 21.2 24.2 7.4

CRA 3.14 2,490 -- -- --

Fine aggregate (FA) 0.78 2,590 -- -- 2.56

Table 2 Mix design for 1 m3 concrete and slump and air content of fresh mix

Index W/C S/A C

(kg)

W

(kg)

Sand

(kg)

NA

(kg)

RA

(kg)

CRA

(kg)

Ad.

(kg)

Slump

(mm)

Air

content

(%)

NAC

0.5

0.48

364

182

806

909 -- --

2.73

155 5.0

RAC-30 640 248 -- -- 4.4

RAC-50 457 413 -- 160 4.9

RAC-100 -- 821 -- 165 5.1

CRAC-30 640 -- 254 150 4.3

CRAC-50 457 -- 423 155 5.7

CRAC-100 -- -- 847 150 5.6

NOTE: W/C was 0.5 by wt.; S/A was 0.48 by vol.; super plasticizer was used at 2.73 kg; RAC-30 recycled aggregate

concrete with 30% replacement of NA by RA; CRAC-50 carbonated recycled aggregate concrete with 50% replacement

of NA by RA; natural fine aggregate was used for all mixes.

91


Fig. 1 – Gradation of NA, RA and FA

(a) RA before soaking (b) RA soaked in 20% HCL (c) RA after soaking

Fig. 2 – Method used to determine adhered mortar amount

(a) Shrinkage (b) Creep

Fig. 3 – Shrinkage measurement and creep test in progress

3. Test Results

3.1 Mechanical properties of NA, RA and CRA

In Table 1, the mechanical properties of NA,

RA, and CRA are summarized as well of those of

fine aggregates. It is seen that the water absorption

of RA (3.84%) is significantly larger than that of NA

(0.48%), SSD density of RA (2,430 kg/m3) is 90% of

NA (2,690 kg/m3) while the crushing value of RA

(21.2%) is larger than that of NA (17.4%). The ad-

hered mortar amount which plays a major role for the

high absorption and low density is 24.2% for RA in

0

10

20

30

40

50

60

70

80

90

100

pan 0.15 0.3 0.6 1.25 2.5 5 10 20 25

Pas

sing (

%)

Sieve size (mm)

FA

NA

RA

92


Table 1. It is also shown in Table 1 that, after car-

bonation treatment, the density of RA increases from

2,430 kg/m3 (SSD) to 2,490 kg/m3 (SSD) and the wa-

ter absorption decreases from 3.84% to 3.14%.

Therefore the three-day accelerated carbonation

scheme adopted in this study was effective and it can

be safely assumed that all or part of the adhered mor-

tar got carbonated which resulted in decreased ab-

sorption and increased density.

3.2 Strength, elastic modulus and Poisson’s

Ratio of hardened concrete

Table 3 and Figures 4 and 5 show the compres-

sive strength development. Compressive strengths of

NAC can be compared with those of RAC-30, RAC-

50, and RAC-100 in Table 3 and Figs. 4 and 5. Fig-

ure 5(a) shows that the compressive strength is sim-

ilar for NAC and RAC-30, which indicates that the

compressive strength is not affected by 30% replace-

ment of NA with RA. Compressive strengths of

RAC-50 and RAC-100 are lower than that of NAC.

The results indicate that the strength is not influ-

enced with 30% replacement of NA with RA but it

is influenced with increasing replacement ratio of

RA by 50% and higher. The strength reduction is not

large even with 50% and 100% replacement. For ex-

ample, at 28 days, the compressive strength of NAC

is 34.4 MPa while it is 33.1 MPa (96% of NA) and

34.4 MPa (100%), respectively, for RAC-50 and

RAC-100. After 180 days, the compressive strength

is 39.4 MPa for NAC while it is 37.5 MPa for RAC-

50 (95%), and 36.6 MPa for RAC-100 (93%).

Table 3 and Figure 5(a) also show that the

strength development after 28 days is similar be-

tween NAC and RAC: i.e. the strength increases

slowly and steadily even after 28 days up to 180 days

as shown. KS F 2573 currently allows maximum 30%

RA replacement [3]. Test results confirm that current

30% limit is valid. It is noted that the strengths of

RAC-50 and RAC-100 are not much lower than that

of NAC. There is a possibility that the current maxi-

mum replacement limit of 30% can be raised in case

of good quality RA that meets the requirements of

KS F 2573.

Figure 5(b) compares the strength development

of NAC and CRAC-30, CRAC-50, and CRAC-100.

Again the compressive strength of CRAC-30 with 30%

replacement of NA with CRA is similar to that of

NAC. However, the compressive strengths of

CRAC-50 and CRAC-100 are significantly lower

than that of NAC at all test ages. Test results suggest

that, although the compressive strength is not af-

fected by 30% replacement of CRA, it is reduced by

replacement of CRA by 50% or higher. It must be

noted in Table 3 that the compressive strengths of

CRAC-50 and CRAC-100 are lower than those of

RAC-50 and RAC-100, respectively, at all test ages.

This unexpected results need explanation because,

after accelerated carbonation, the mechanical prop-

erties of CRA such as density and absorption im-

proved over that of RA. At present authors do not

have a clear explanation to this phenomenon. A hy-

pothesis is suggested that, with three-day accelerated

carbonation scheme adopted in this study, only the

part of adhered mortar gets carbonated, which may

result in poor bond with new cement paste at surface

of carbonated adhered mortar.

Current test results agree well with that in the

existing literature [8-20]. Andal et al. [18] tested

strength and shrinkage of concrete using 20-mm re-

cycled coarse aggregates of preserved quality and

commercial quality (density = 2,310-2,320 kg/m3,

absorption = 4.88-5.32%, w/c = 0.45). They have

suggested that the use of 100% RA resulted in some

reduction in the strength, but the use of 30% RA as

partial replacement of coarse aggregate produced

compressive strength similar to that of concrete with

100% virgin coarse aggregate. Doming et al. [16]

used 20-mm nominal size RA which replaced NA at

20%, 50%, and 100% (density = 2,460 kg/m3, ab-

sorption = 5.19-6.08%, w/c = 0.5). They have ob-

served that when the effective w/c was maintained

constant, the compressive strength was the same so

the substitution of NA by RA did not have a signifi-

cant effect. Geng et al. [14] used 25-mm nominal

size RA which replaced NA by 100% (density =

2,713 kg/m3, absorption = 5.07%, w/c = 0.45). They

have observed some reduction of compressive

strength from 100% RA concrete than NA concrete,

but the reduction was less than 10%.

Figure 6 shows the tensile strength test results

in terms of both split tensile strength (black bar) and

flexural strength (white bar) at 28 days. The split ten-

sile strength (fsp) test data show that the split tensile

strength is not much affected by replacement of NA

by RA (or CRA) with exception of RAC-100. On the

other hand, the flexural strength (fr) tends to decrease

with increasing replacement ratio of NA with RA (or

CRA). However, the flexural strengths are above the

flexural strength predicted by the KCI Structural

Concrete Design Code [28]. It can be concluded that

the split tensile strength is not significantly affected

by replacing NA with RA up to 100%. The flexural

strength is negatively affected with increasing re-

placement ratio of NA with RA, but it still satisfies

the code required flexural strength.

Table 3 and Figure 7 show elastic modulus

measured at 28 days. The 30% replacement of NA

by RA (or CRA) does not influence the elastic mod-

ulus. For 50% and 100% RA replacement, the elastic

modulus reduces a little but are at least 90% that of

NAC and are within 10% margin from the predicted

value by KCI Structural Concrete Design Code [28].

It needs to be noted that due to relatively soft adhered

93


mortar, the elastic modulus of RA is typically lower

than that of NA which in turn negatively affect the

elastic modulus of RAC. Such negative effect is not

clearly shown in Table 3 and Fig. 7. Table 3 also

shows Poisson’s Ratio for NAC, RAC, and CRAC.

Despite some scatter in test data, the Poisson’s Ra-

tios range between 0.18 and 0.22 and there is no clear

indication that the Poisson’s Ratio is influenced by

replacement of NA by RA even up to 100% replace-

ment.

3.3 Shrinkage of recycled aggregate concrete

Figure 8 shows temperature and R.H. in the en-

vironmental chamber during the test period. It can be

seen that the temperature is relatively constant at

19.5°C and varies between 19°C and 21°C. R.H.

ranges between 48% and 85% in the chamber reflect-

ing seasonal variation of outside weather with a

mean of 59.7%.

Figure 9 shows measured shrinkage strain ver-

sus time. Shrinkage strains show steady and fast in-

crease during the first two weeks followed by steady

increase during the next three months. After about 3-

1/2 months, the shrinkage strain development some-

what levels out (or it still keeps increasing but at a

much lower rate). Therefore, in Table 4, the shrink-

age strain values are summarized at each time mark,

i.e. after 2 weeks, 3-12/ months, and 6 months (180

days).

Table 3 – Summary of strength, elastic modulus and Poisson’s Ratio

Index Compressive strength, fcu

(MPa)

Tensile strength

(MPa)

Elastic

Modulus

(GPa)

Poisson’s

Ratio

f3 f7 f28 f56 f90 f180 f sp fr

NAC

mean

min.

max.

22.4

21.9

22.8

25.6

25.6

25.7

34.4

33.2

36.9

35.6

35.2

36.3

37.9

37.6

38.8

39.4

39.1

39.5

2.56

2.55

2.57

6.10

5.81

6.39

26.1

25.9

26.2

0.19

0.18

0.20

RAC-30

mean

min.

max.

25.2

25.2

25.3

28.8

27.5

29.8

37.2

36.3

37.8

37.9

36.9

38.8

37.5

34.7

39.9

39.6

38.0

42.6

2.62

2.49

2.75

6.00

5.82

6.18

28.1

26.2

29.9

0.20

0.18

0.22

RAC-50

mean

min.

max.

21.0

20.9

21.2

25.1

23.4

26.2

33.1

31.9

34.3

34.2

33.9

34.4

35.0

33.0

36.0

37.5

37.1

37.8

2.67

2.66

2.68

5.67

5.42

5.92

23.6

21.5

25.9

0.21

0.18

0.24

RAC-100

mean

min.

max.

22.1

21.7

22.7

24.5

22.8

26.0

34.4

33.8

34.7

33.2

32.5

34.1

35.0

34.0

37.2

36.6

33.1

40.3

2.24

1.85

2.63

5.25

5.12

5.38

25.2

24.5

26.2

0.20

0.17

0.22

CRAC-30

mean

min.

max.

23.1

22.6

23.7

27.5

26.0

28.5

33.6

32.6

34.3

36.1

35.2

36.9

37.8

36.1

39.9

41.0

40.7

41.3

2.71

2.20

3.22

5.59

5.50

5.69

25.7

24.6

27.1

0.22

0.18

0.27

CRAC-50

mean

min.

max.

15.5

15.2

15.7

19.5

18.3

21.3

27.0

24.6

29.5

28.0

26.8

29.6

28.5

26.42

30.34

32.2

31.0

33.3

2.48

2.43

2.54

5.38

5.18

5.57

25.4

23.7

26.5

0.20

0.19

0.23

CRAC-100

mean

min.

max.

17.6

16.2

19.5

19.0

18.0

20.7

27.5

24.4

30.6

30.3

29.6

31.0

27.4

24.4

32.8

30.3

29.2

32.8

2.73

2.64

2.82

5.16

5.14

5.18

23.8

23.4

24.0

0.18

0.17

0.20

NOTE: Compressive strength is average of three test; tensile strength is average of two tests; elastic modulus and Pois-

son’s Ratio are taken at 28 days.

94


Fig. 4 – Compressive strength test results by age

(a) NAC vs. RAC (b) NAC vs. CRAC

Fig. 5 – Compressive strength development of NAC, RAC and CRAC

Fig. 6 – Tensile strength at 28 days

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

45.0

f3 f7 f28 f56 f90 f180

Com

pre

ssiv

e s

trength

(M

Pa)

Ages (day)

NAC RAC30 RAC50 RAC100 CRAC30 CRAC50 CRAC100

0

10

20

30

40

50

0 50 100 150 200

NAC

RAC-30

RAC-50

RAC-100Com

pre

ssiv

e s

trength

(M

Pa)

Time (days)

0

10

20

30

40

50

0 50 100 150 200

NAC

CRAC-30

CRAC-50

CRAC-100

Time (days)

Com

pre

ssiv

e s

trength

(M

Pa)

0

2

4

6

8

10

NAC RAC-30 RAC-50 RAC-100 CRAC-30 CRAC-50 CRAC-100

Tensile

strength

(M

Pa)

split tensile flexural strength

Code predicted flexural strength: 0.63√𝑓𝑐′

95


Fig. 7 – Elastic modulus at 28 days

In Table 4, in case of NAC, the shrinkage strain

is 206 μm/m after 2 weeks, 690 μm/m after 3-1/2

months, and 712 μm/m after 6 months. It is seen that

about 29% of the maximum shrinkage developed

during the first two weeks and about 97% of the max-

imum shrinkage (recorded in 6-months period) de-

veloped during the first 3-1/2 months.

Table 4 also shows the shrinkage strains of RAC

and CRAC normalized by that of NAC. In Table 4

and Figure 9, the shrinkage of NAC, RAC-50, and

CRAC-50 is almost the same (84%-103% of NAC).

The shrinkage of RAC-100 and CRAC-100 is larger

than that of NAC at all ages (103%-114% of NAC).

It can be suggested that the shrinkage of 50% recy-

cled coarse aggregate concrete is the same as that of

NAC and the shrinkage of 100% recycled coarse ag-

gregate concrete is larger than that of natural aggre-

gate concrete. In addition, there are no differences in

the shrinkage behavior between RAC and CRAC for

all replacement ratios tested.

Manzi et al. [19] reported shrinkage strain of

about 860 μm/m at 180 days from RAC with 63.5%

RA (density 2,250-2,430 kg/m3, w/c = 0.48, absorp-

tion = not disclosed) and 650-680 μm/m at 180 days

from RAC with 36.5% RA. The shrinkage strains re-

ported by Manzi et al. are comparable to those meas-

ured in this study. Other researchers reported larger

shrinkage strains. Andal et al. [18] reported 50%

more shrinkage from 100% replacement of RA (den-

sity = 2,310-2,320 kg/m3, absorption = 4.88-5.32%,

w/c = 0.45) after 180 days. Domingo et al. [16] re-

ported about 20% increased shrinkage from RAC

with 50% RA replacement and about 70% more

shrinkage with 100% RA replacement after 180 days

(density = 2,460 kg/m3, absorption = 5.19-6.08%,

w/c = 0.5). The reason for the relatively small

amount of shrinkage strains compared to that of

NAC determined from this study can be relatively

good quality RA with low amount of water absorp-

tion (absorption = 3.84%, 4.88-5.32%, and 5.19-6.08%

for RA used in this study, by Andal et al., by Do-

mingo et al., respectively). Although the effective

w/c can be maintained during batching, concrete us-

ing RA with higher absorption capacity will lead to

more porous concrete, which can be more suscepti-

ble for moisture loss inducing increased drying

shrinkage.

Figure 10 shows the shrinkage strains for NAC,

RAC-50, and RAC-100 predicted by ACI 209 Tech-

nical Committee report [29]. Shrinkage strain-time

curves of NAC and RAC-50 predicted by ACI 209

are in good agreement with the current test data

especially after 180 days. Measured shrinkage of

RAC-100 in this study is higher than the ACI 209

predicted value, while the difference is about 15%.

3.4 Creep of recycled aggregate concrete

Creep test began 35 days after casting and con-

tinued for 150 days. Creep test specimens were cured

under water right after demolding until the test day (t

= 35 days). Sustained load that corresponds to about

30% of the 28-day compressive strength (See Table

3) was applied at 35 days and the same sustained load

was maintained for the duration of the creep test. A

total of five different concretes was tested: NAC,

RAC-50, RAC-100, CRAC-50, and CRAC-100.

Since two creep specimens were tested using one

loading frame, RAC-50 and RAC-100 were tested

using the same loading frame and hence were sub-

jected to the same sustained load. CRAC-50 and

CRAC-100 were also tested using the same loading

frame. NAC was tested while the overall test proce-

dure followed ASTM C512 recommendations [30].

0

5

10

15

20

25

30

35

40

NAC RAC-30 RAC-50 RAC-100 CRAC-30 CRAC-50 CRAC-100

Ela

stic

modulu

s (G

Pa)

Code predicted elastic modulus: 8,500√𝑓𝑐′3

96


Table 4 – Shrinkage strain vs. time (unit: μm/m)

Index Time after shrinkage measurement Shrinkage normalized by NAC at time of

2

weeks

3-1/2

months

6

months

2

weeks

3-1/2

months

6

months

NAC 206 690 712 1.0 1.0 1.0

RAC-50 175 697 725 0.85 1.01 1.02

RAC-100 220 747 803 1.07 1.08 1.13

CRAC-50 173 706 731 0.84 1.02 1.03

CRAC-100 213 746 809 1.03 1.08 1.14

Fig. 8 – Temperature and R.H. in environmental chamber

Fig. 9 – Shrinkage vs. time

Table 5 summarizes the creep test results.

Figure 11 shows the total strain developed in all

creep test specimens that include short-term strain

(elastic strain) at t = 35 days and long-term strain that

consists of shrinkge strain and creep strain. Both

elastic strain and srinkage strain were deducted from

the total strain to determine the net creep strains, and

the results are shown in Fig. 12. It needs to be noted

that the sustained load level each creep test specimen

is subjected to is different, and therefore the creep

strains shown in Fig. 12 need to be normalized in

terms of unit sustained stress (i.e. 1 MPa). The results

are shown as specific creep in Fig. 13. In Table 5 and

Figure 13, it is seen that NAC experiences the

smallest specific creep of 81 μm/m/MPa after 150

days. The largest specific creep is determined from

RAC-100 that is 112 μm/m/MPa (138% of NAC)

followed by RAC-50 with 101 μm/m/MPa (125% of

NAC). The creep test results show that the creep of

RAC is significantly larger than that of NAC.

0.0

20.0

40.0

60.0

80.0

100.0

0 50 100 150 200

Tem

pera

ture

(°C

) and rela

tive

hum

idity

(%)

Time (days)

0

200

400

600

800

1000

0 50 100 150 200

Shrinkage s

train

(μm

/m)

Time (days)

NAC

RAC-50

RAC-100

CRAC-50

CRAC-100

97


Fig. 10 – Shrinkage prediction model by ACI 209 [29]

Current test data match well with those

available in the literature. Manzi et al. [19] showed

that specific creep of RAC with 63.5% RA (density

2,250-2,430 kg/m3, w/c = 0.48) substitution was 90

μm/m and 90-95 μm/m for RAC with 36.5% RA

substitution after 150 days. Domingo et al. [16]

reported that the specific creep of RAC with 50% RA

(density = 2,460 kg/m3, absorption = 5.19-6.08%,

w/c = 0.5, sustained loading about 30% of compres-

sive strength started at 28 days) substitution

increased over that of NAC by 29% and increased

for RAC with 100% RA substitution by 32% over

that of NAC. In addition, Rye et al. [15] after

systematic analysis of existing data matrix available

in literature has concluded that the creep of concrete

increases with a decreasing rate with increasing RA,

giving an average increase of 32% at 100% RA

content. The specific creep of CRAC is smaller than

that of RAC: i.e. it is 96 μm/m/MPa for CRAC-50

(119% of NAC). Also specific creep of 91

μm/m/MPa is determined from CRAC-100 that is

112% of NAC. It may needs to be noted that the

specific creep of CRAC-100 is smaller than CRAC-

50 after 150 days, which is an unexpected test result.

In this study the duration of creep test is only 150

days. In the longer term, the difference between the

two creep test data may become smaller in Fig. 13.

Creep coefficient (Φ) is shown in Table 5 and

Fig. 14 for all creep test specimens which ranges

between 1.97 for NAC and 2.76 for RAC-100.

Therefore the creep coefficient of RAC-100 is as

large as 140% of NAC. In case of RAC-50, the creep

coefficient is 136% of NAC. The creep coefficient of

CRAC-50 and CRAC-100 is 115% and 120% that of

NAC, respectively. The Φ value is smallest for NAC

as expected followed by CRAC-50, CRAC-100,

RAC-50, and RAC-100. Again the current test data

match well with those published. Geng et al. [14]

reported that the creep coefficient ratio (Φ/ΦNAC) is

about 1.3 for RAC with w/c = 0.5 including 100%

RA (RA density = 2,713 kg/m3, absorption = 5.07%).

Figure 15 shows the creep coefficient predicted by

ACI 209 Technical Committee report for NAC,

RAC-50, and RAC-100 [29]. Creep coefficient-vs.-

time curves predicted by ACI 209 are lower than the

current test data shown in Fig. 14. The difference is

about 30% for NAC and it is much larger for RAC-

50 and RAC-100. The prediction of creep strains de-

pends on many influencing factors such as time of

loading, temperature, R.H., strength as well as some

fresh concrete properties. Especially the large differ-

ences between the ACI predicted values and the cur-

rent test data means that the ACI 209 formula devel-

oped exclusively for NAC is not applicable for RAC.

Table 5 – Creep test results 150 days after sustained load application

Index NAC RAC-50 RAC-100 CRAC-50 CRAC-100

Total strain (μm/m) 1,575 1,601 1,754 1,358 1,347

Shrinkage (μm/m) 321 262 292 303 351

Elastic strain (μm/m) 423 364 389 322 297

Creep strain (μm/m) 831 976 1,074 733 699

Specific creep (μm/m/MPa) 81 101 112 96 91

Creep coefficient 1.97 2.68 2.76 2.27 2.36

Creep coefficient Φ/ΦNAC 1.0 1.36 1.40 1.15 1.20

NOTE: Applied stress is 10.3 MPa for NAC; 9.63 MPa for RAC-50 and RAC-100; and 7.66 MPa for CRAC-50 and

CRAC-100.

0

150

300

450

600

750

0 40 80 120 160 200

Shrinkage (μm/m

)

Time (days)

NAC

RAC-50

RAC-100

98


Fig. 11 – Total strain vs. time

Fig. 12 – Creep vs. time after loading

Fig. 13 – Specific creep vs. time after loading

0

200

400

600

800

1000

1200

1400

1600

1800

0 50 100 150 200

Tota

l cr

eep s

train

(μm

/m)

Time (days)

NAC

RAC-50

RAC-100

CRAC-50

CRAC-100

t=35

0

200

400

600

800

1000

1200

0 40 80 120 160

Cre

ep s

train

(μm

/m)

Time (days)

NAC

RAC-50

RAC-100

CRAC-50

CRAC-100

0

20

40

60

80

100

120

0 40 80 120 160

Speci

fic

creep (μm

/m/M

Pa)

Time (days)

NAC

RAC-50

RAC-100

CRAC-50

CRAC-100

99


Fig. 14 – Creep coefficients vs. time after loading

Fig. 15 – Creep coefficient prediction model by ACI 209 [29]

4. Conclusions

Strength development, shrinkage, and creep of

RAC was investigated. RA replaced NA by 30%,

50%, and 100% by vol. For the investigation of

strength development, 3, 7, 28, 56, 90, and 180 day

compressive strength and 28 day split tensile

strength and flexural strength have been determined.

Shrinkage and creep behavior of RAC with 0%, 50%,

and 100% RA replacing NA has been studied up to

180 days and 150 days, respectively. RAC including

RA treated with 3-day accelerate carbonation (CRA)

was also used. The following conclusions are drawn

from this study.

(1) The physical properties of RAC, such as density

and water absorption, improve by accelerated

carbonation, but the improvement of mechani-

cal properties is not directly related to the

strength improvement of RAC incorporating the

CO2 treated RA.

(2) Compressive strength of RAC with 30% RA or

CRA is similar to that of NAC and the compres-

sive strength of RAC with 50% or 100% re-

placement is reduced from that of NAC, while

the strength reduction is smaller than 10%.

(3) Split tensile strength is not significantly af-

fected by RA replacement up to 100%. Flexural

strength decreases with increasing amount of

replacement, but the flexural strengths are

above the value required by structural concrete

design code.

(4) Elastic modulus of RAC tends to decrease with

increasing RA replacement, but it is not signifi-

cantly reduced (less than 10%).

(5) Shrinkage of RAC with 50% RA is similar to

that of NAC. Shrinkage of RAC with 100% RA

increases over that of NAC up to 13% after 180

days.

(6) Specific creep of RAC with 50% and 100% RA

increases over that of NAC by 25% and 38%,

0.0

0.8

1.6

2.4

3.2

0 40 80 120 160

Cre

ep c

oeffic

ient

Time (days)

NAC

RAC-50

RAC-100

CRAC-50

CRAC-100

0

0.4

0.8

1.2

1.6

0 40 80 120 160

Cre

ep c

oeffic

ient

Time (days)

NAC

RAC-50

RAC-100

100


respectively, after 150 days of sustained loading.

(7) RAC with CO2 treated RA has similar shrinkage

behavior to RAC with RA not treated by CO2

while creep of RAC with CO2 treated RA is

smaller than that of RAC with RA not treated

with CO2.

Acknowledgement

This work was supported by the Korea Technol-

ogy and Information Promotion Agency for SMEs

(TIPA) grant funded by the Korea government (Proj.

No.: C0531527). Authors also gratefully

acknowledge the support from Dongbu-ENT Corp.

in the form of recycled coarse aggregate supply

throughout this study.

References

1. Choi Donguk et al. (2016) Concrete and Envi-

ronment, 2nd ed., Kimoondang, 505 pp.

2. RILEM Recommendation (1994) Specifica-

tions for concrete with recycled aggregates, Ma-

terials and Structures, 27(9), pp. 557-559.

3. KS F 2573 (2011) Concrete Recycled Aggre-

gate, Korea Standard Association.

4. JIS A 5021 (2005) Recycled aggregate for con-

crete-class H, Japan Standards Association.


crete-class M, Japan Standards Association.


crete-class L, Japan Standards Association.

7. Kim J.-M. et al. (2017) Recycled Aggregate

Concrete, KCI Manual of Practice, Korea Con-

crete Institute, 2017 (in Korean).

8. Shi, C.; Li Y., Zhang, J.; Li, W.; Chong, L.; and

Xie, Z. (2016) “Performance enhancement of

recycled concrete aggregate - A review,” Jour-

nal of Cleaner Production, 112, pp. 466–472.

9. Kou, S.C.; Zhan, B.J.; and Poon, C.S. (2014)

“Use of a CO2 curing step to improve the prop-

erties of concrete prepared with recycled aggre-

gates,” Cement & Concrete Composites, 45, pp.

22–28.

10. Silva, R.V.; Neves, R.; de Brito, J.; and Dhir,

R.K. (2015) “Carbonation behavior of recycled

aggregate concrete,” Cement & Concrete Com-

posites, 62, pp. 22-32.

11. Tam, V.W.Y.; Butera, A.; and Le, K.N. (2016)

“Carbon-conditioned recycled aggregate in

concrete production,” Journal of Cleaner Pro-

duction, 133, pp. 672-680.

12. Zhan, B.; Poon, C.S.; Qiong, L.; Kou, S.; and

Shi, C. (2014) “Experimental study on CO2 cur-

ing for enhancement of recycled aggregate

properties,” Construction and Building Materi-

als, 67, pp. 3–7.

13. Silva, R.V.; de Brito, J.; and Dhir, R.K. (2015)

“Comparative analysis of existing prediction

models on the creep behavior of recycled aggre-

gate concrete,” Engineering Structures, 100, pp.

31-42.

14. Geng, Y.; Wang, Y.; and Chen J. (2016) “Creep

behavior of concrete using recycled coarse ag-

gregates obtained from source concrete with

different strengths,” Construction and Building

Materials, 128, pp. 199-213.

15. Lye, C.-Q.; Dhir, R.K.; Ghataora, G.S.; and Li,

H. (2016) “Creep strain of recycled aggregate

concrete,” Construction and Building Materials,

102, pp. 244-259.

16. Domingo, A.; Lazaro, C.; and Gayarre, F.L.

(2010) “Long term deformations by creep and

shrinkage in recycled aggregate concrete,” Ma-

terials and Structures, 43, pp. 1147-1160.

17. Tam, V.W.Y.; Kotrayothar, D.; and Xiao, J.

(2015) “Long-term deformation behavior of re-

cycled aggregate concrete,” Construction and

Building Materials, 100, pp. 262-272.

18. Andal, J.; Shehata, M.; and Zacarias, P. (2016)

“Properties of concrete containing recycled

concrete aggregate of preserved quality,” Con-

struction and Building Materials, 125, pp. 842-

855.

19. Manzi, S.; Mazzotti, C.; and Bignozzi, M.C.

(2013) “Short and long-term behavior of struc-

tural concrete with recycled concrete aggregate,”

Cement & Concrete Composites, 37, pp. 312-

318.

20. Sindy, S.-P.; Belen, G.-F.; Fernando, M.-A.;

and Iris, G.-T. (2016) “Time-dependent behav-

ior of structural concrete made with recycled

coarse aggregates. Creep and shrinkage,” Con-

struction and Building Materials, 122, pp. 95-

109.

21. KS F 2502 (2014) Standard test method for

sieve analysis of fine and coarse aggregates,

Korea Standard Association.

22. KS F 2584 (2010) Standard test method for ac-

celerated carbonation of concrete, Korea Stand-

ard Association.

23. Neville A. M. (2011) Properties of Concrete, 5 th

ed., Pearson.

24. Mindess S., Young J. F., Darwin, D. (2003)

Concrete, 2nd ed., Prentice Hall.

25. KS F 2503 (2014) Standard test method for den-

sity and absorption of coarse aggregate, Korea

Standard Association.

26. KS F 2541 (2017) Standard test method for de-

termination of Aggregate-Crushing-Value

(ACV), Korea Standard Association.

27. de Brito, J. and Saikia, N. (2013) Recycled Ag-

gregate in Concrete, Springer.

28. KCI Structural Concrete Design Code (2012)

Korea Concrete Institute.

101


29. ACI 209R-92 (1992) Prediction of Creep,

Shrinkage, and Temperature Effects in Con-

crete Structures, American Concrete Institute.

30. ASTM C512-02 (2002) Standard test method

for creep of concrete in compression, American

Society for Testing and Materials.

102




ISSN 2465-7964 / eISSN 2645-7972

https://doi.org/10.18702/acf.2019.1.4.2.103

Technical Paper

Visual investigation method and structural performance

evaluation for DEF induced damaged Indian Railway PC

sleepers

Rajamurugan Sundaram*, Koji Matsumoto, Kohei Nagai and Anupam Awasthi

(Received: June 24, 2018; Accepted: November 21, 2018; Published online: January 04, 2019)

Abstract: Indian Railways uses pre-stressed concrete (PSC) sleepers for its tracks. In the recent years in north

and central part of railways, premature cracks were observed in sleepers. Cracks were observed for sleepers 6

to 9 years after their manufacture. In this research, structural performance of damaged sleepers was evaluated

in both material and structural level. For material level investigation, core samples were taken from the dam-

aged sleepers. From the cores, reduction in static elastic modulus and compressive strength were observed for

damaged sleepers when compared to undamaged sleepers. For structural level investigation, based on the level

of the damage, sleepers were categorized into damaged and undamaged sleepers, and flexural test and shear

test for sleepers were conducted. From the bending and shear test results, relationship between the crack pattern

and capacity reduction in damaged sleepers was studied. In the flexural test, it was found out that more layers

of longitudinal cracks in the central parts reduced its capacity up to 35% when compared to undamaged sleep-

ers. Shear test results showed that it was very close to the minimum requirement for capacity of the sleepers.

Keywords: delayed ettringite formation (DEF), alkali silica reaction (ASR), sleeper, concrete.

1. Introduction

The main objective of railways is to provide

safe transport for passengers. Indian Railways uses

pre-cast pre-stressed concrete sleepers from factories

for its track. Concrete sleepers are important part of

railway track to hold the rail in the position and trans-

fer the load to the supporting structure below. In

northern and central part of the railways, these sleep-

ers start cracking 6 to 9 years after their manufacture.

Even unloaded sleepers, which are not used in the

tracks, also get cracked. As per the recent survey

conducted in the central part of India, in a 100-km

railway track and out of 128,000 sleepers, 45,000

sleepers were damaged with premature cracking.

This means, in this particular railway track of 100

km, almost 35% of the sleepers are damaged. The

damaged sleepers are replaced annually: when fail-

ure cracks at insert location is observed sleepers are

replaced immediately. No clear investigation has

been carried to find the current capacity of these

damaged sleepers. This research aims to find out the

structural performance of damaged sleepers based on

the existing crack patterns in the sleepers. For this

purpose of investigation, research was carried out to

find out sleepers with different levels of damage.

This problem is a potential threat to the Indian Rail-

ways and it affects the safety of the passengers trav-

elling in trains.

In Indian Railways, high temperature more than

70○C has been recorded during steam curing in the

manufacturing process of sleepers. This may cause

Delayed Ettringite Formation (DEF) and be the main

cause of these premature cracking. Research was

conducted in the past study in Indian Railways to

find out the cause of expansive damages in concrete

sleepers, where it was found out the cause of these

cracks were due to both DEF and Alkali Silica Reac-

tion (ASR) [1]. The ASR is caused by the chemical

reaction between cement alkali and reactive aggre-

gate that generates expansive ASR gel. In DEF, dis-

solved ettringite under high temperature curing is re-

formed after the concrete hardening to cause expan-

sive stress. Experimentation with German high early

Corresponding author Sundaram Rajamurugan is an

Engineer at Sumitomo Mitsui Construction Co. Ltd.

Koji Matsumoto is a Project Assistant Professor at In-

stitute of Industrial Science, The University of Tokyo,

Tokyo, Japan.

Kohei Nagai is an Associate Professor at Institute of

Industrial Science, The University of Tokyo, Tokyo,

Japan.

Anupam Awasthi is a Deputy Chief Engineer at Con-

struction Organization in Indian Railways, India.

103

strength cement concluded that delayed expansion

occurs when specimens were cured above 80○C [3].

Boundary temperature conditions for occurrence of

DEF were between 60○C and 70○C [4]. To under-

stand the DEF a holistic approach for late sulphate

release, micro-cracking, and exposure to water was

proposed [5]. Cement composition (alkalis, C3S,

C3A, SO3, and MgO) and fineness also influence the

effect of DEF [2]. Typical crack patterns observed in

the sleepers are shown in Fig. 1. Map cracks are ob-

served at the ends and longitudinal cracks are ob-

served at the midspan. In the past research of prem-

ature cracking, similar problems were observed in

other parts of the world. In 2004, prestressed mon-

obloc concrete sleepers placed in Portugal had

shown premature cracking [6]. Sleepers manufac-

tured in the years between 1992 and 1996 in Sweden

have started to deteriorate and cracks were observed

in sleepers [7]. As per a report in Finland 20,000

sleepers are replaced every year [8]. Distress in pre-

stressed concrete sleepers was observed in eastern

coast of United States [9]. It was predicted that

causes of these cracks could be either DEF or ASR.

From all these research, it is observed that damaged

sleepers are huge in number and it is not possible

economically to replace all the sleepers immediately.

For rational and efficient maintenance, it is essential

to set priorities for replacement considering the

structural performances of sleepers. However, the

past research including the authors study do not suf-

ficiently clarify structural performances [1]. For this

purpose of investigation, sleepers with various de-

grees of damages were selected to study the crack

patterns. Material level testing and structural level

testing were conducted in the sleepers to find out the

current damage conditions of the sleeper. For mate-

rial testing, core samples were taken from the dam-

aged and undamaged sleepers, where compressive

strength and static elastic modulus were found from

the compressive test. For structural level investiga-

tion, crack patterns in the damaged sleepers were

studied where map cracks were observed at the end

of the sleepers and longitudinal cracks at the central

part of the sleepers. Side view and top view of the

damaged sleepers are shown in Fig. 1. Cracks are lo-

cated at the insert location of the sleeper, which is a

failure crack in the sleepers. Based on the level of

damage, cores and sleepers were collected from the

damaged and undamaged sleepers. Structural level

testing was also conducted on both undamaged and

damaged sleepers, where flexural test and shear test

were conducted on the damaged and undamaged

sleepers to find out the current capacity of the sleep-

ers. Existing crack patterns from the damaged sleep-

ers were studied and failure pattern was analyzed.

The relationship between existing crack patterns and

capacity reduction in damaged concrete sleepers was

studied.

Sleeper are placed on ballasted track bed as

shown in Fig. 2. This ballast track bed acts as an

elastic bed and transfer the load coming from sleeper

to wider area of formwork. This formation is

earthern embankment and its not bonded to concrete

structure.

2. Proposal of visual inspection method

Visual inspection was conducted. Based on the

damages observed in the sleepers, sleepers were

classified into two categories such as Damaged and

Undamaged sleepers (See Table 1). Within damaged

and undamaged sleepers, sleepers were further clas-

sified into five categories. Undamaged sleeper’s two

categories were Undamaged new and Undamaged

old. Sleepers in these categories did not have any vis-

ual cracks. Damaged sleepers were categorized into

three categories such as Mild, Moderate and Severe

damage. In Mild damaged sleepers edge cracks were

observed in the sleepers, whereas in Moderate dam-

aged sleepers one longitudinal crack was observed in

the center of the sleepers. In Severe damaged sleep-

ers, more than 2 to 3 longitudinal cracks were ob-

served in the center.

Description of these cracks with crack widths is

shown in Table 1. Typical view of these crack pat-

terns is shown in Fig. 11. Cracks typically occur

around 6 to 9 years after their manufacture. The bal-

last profile and crack sequence are shown in Figs. 2

and 3, respectively. When sleepers are in service, i.e.

installed condition on track, first cracks are usually

seen on the side face of the sleeper. The side face is

covered with ballast and cracks are visible only after

opening the ballast. In later stages cracks are also

seen on the top surface of the sleeper. Ultimately fail-

ure occurs near the inserts [1].

Indian Railways uses M55 grade sleepers for its

tracks (See Table 2). 3 ply of 3 mm high tensile

strength strands are used in sleepers (See Fig. 4). Ac-

cording to the specification, each reinforcing strand

is to be tensioned with initial force of 27 kN. Typical

tendon profile and cross section of sleepers are

shown in Fig. 4.

SEM analysis was conducted in the past study

from concrete samples collected from Indian Rail-

ways, where presence of DEF from the concrete

samples was observed (See Fig. 5). In order to find

the effect of ASR in concrete, the authors also con-

ducted the chemical analysis. It was found from the

chemical analysis that alkali content in cement sam-

ples was greater than 5 kg/m3 (Table 3). Whereas as

per Japanese standards, alkali amount should not ex-

ceed more than 3 kg/m3. This high alkali content in

samples also promotes ASR.

104


Table 2 – Concrete mix proportions in sleepers

Coarse aggregate, CA1 (20 mm) 981.49 kg (50.50%)

Coarse aggregate, CA2 (20 mm) 420.64 kg (21.50%)

Fine aggregate, FA 546.28 kg (28.00%)

Cement (53-S) 445.50 kg/m3

Admixture 2.227 kg (0.5%) of cementitious material

Water 142.56 Liters

W/C ratio 0.32

A/C ratio 4.373

Table 1 – Description of damage categories

Category No. of

sleepers

Cracks

Left side Central part Right side

Undamaged new 3 No cracks No cracks No cracks

Undamaged old 3 No cracks No cracks No cracks

Mild damage 2 1 mm cracks 1 mm cracks No cracks

Moderate damage 3 1 mm cracks 1 layer of 1-2 mm cracks 1-2 mm cracks

Severe damage 4 1-3 mm cracks 2 to 3 layers of 2-3 mm cracks 1-2 mm cracks

Ends - Map cracks

Center - Longitudinal

Failure - Insert cracks

Failure cracks due to combined effect of dam-

age and lateral loads Map cracks

at ends

Map cracks

at ends

Longitudinal cracks due to the combined

effect of expansion and pre-stress

Fig. 1 – Typical cracking pattern observed in the sleeper of Indian Railways

105


Fig. 2 – Ballast profile track slab

(a) First Step (Cracks at side face) (b) Second Step (Cracks visible on top surface)

(c) Third Step (Cracks near insert location)

Fig. 3 – Crack sequences

(c) Side view of sleepers

PC strand

PC strand

Fig. 4 – PS tendon profile

(a) Cross section of sleepers (unit: mm) (b) 3 x 3mm wire strands

106


SEM analysis was conducted in the past study

from concrete samples collected from Indian Rail-

ways, where presence of DEF from the concrete

samples was observed (See Fig. 5). In order to find

the effect of ASR in concrete, the authors also con-

ducted the chemical analysis. It was found from the

chemical analysis that alkali content in cement sam-

ples was greater than 5 kg/m3 (Table 3). Whereas as

per Japanese standards, alkali amount should not ex-

ceed more than 3 kg/m3. This high alkali content in

samples also promotes ASR.

Fig. 5 – Presence of DEF in concrete samples [1]

Table 3 – Chemical analysis results in cement samples

Factories Component analysis result of cement (mass %) Alkaline amount

(kg/m3) Na2O K2O Na2O eq.*

1st Factory 0.27 1.28 1.11 5.03

2nd Factory 0.27 1.30 1.13 5.11

3rd Factory 0.27 1.28 1.11 5.03

Average 0.27 1.29 1.12 5.07

3rd

cut 1st cut 2

nd cut

(b) Mild damage (c) Moderate damage (d) Severe damage

(a) Typical side view of Severe damaged sleeper

Fig. 6 – Cut cross section in sleepers

107


3. Internal crack pattern and tests of core

samples

3.1 Cross section cut in sleepers

Sleepers were cut to find whether cracks have

infiltrated inside the sleepers. Sleepers were col-

lected from both damaged and undamaged sleepers

(See Fig. 6), where manufacturing year of sleeper is

also mentioned (See Table 4).

Except “Severe damaged sleeper” where three

sleepers were taken for cross section cut, in rest of

the categories only a sleeper was cut. Typical side

view of cross section cut is shown in Fig. 6(a), where

three cuts were made on each sleeper. Critical cross

sections based on the category with more cracks are

shown in this Fig. 6. For both undamaged new and

undamaged sleepers no cracks were observed inside

the sleepers, crack infiltration up to 20 mm was only

observed in the sleepers and cracks did not infiltrate

fully inside the sleepers. It is observed from the study

that DEF expansion occurs ununiformly and only the

inner portion is expanded, outer side is put in tension

in the circumferential direction as we observed in the

ring tension behavior, resulting in large cracks only

in the outer side of the sleepers.

Table 4 – List of all the specimens used for cross section cut

Category Manufacture

year No. of

sleepers

Crack width Left side Central part Right side

Undamaged new 2015 1 No cracks No cracks No cracks

Undamaged old 1986 1 No cracks No cracks No cracks

Mild damage 2006 1 1 mm cracks 1 mm cracks No cracks

Moderate damage 2006 1 1 mm 1 layer of 1-2 mm 1-2 mm

Severe damage-I 2002 1 1-2 mm 3 layers of 2-3 mm 1-2 mm

Severe damage-II 2006 1 1-2 mm 3 layers of 2-3 mm 1-2 mm

Severe damage-III 2002 1 1-2 mm 3 layers of 2-3 mm 1-2 mm

Typical core sample

Existing cracks- Severe damaged

Existing cracks- Mild damaged

Undamaged

sleeper

New-2012

Old-2002

Mild-2002

Severe-

2002

Damaged

sleeper

Fig. 7 – Side view of damaged and undamaged sleepers for cores samples

Table 5 – Cores samples for compressive test

Category Damaged sleeper Manufacturing year Number of cores

Undamaged New 2015 2

Old 2002 2

Damaged Mild 2002 2

Severe 2002 2

108


3.2 Compressive test of core sample

Cores were collected from both damaged and

undamaged sleepers for compressive test (See Fig.

7), where manufacturing year of sleeper is also in-

cluded (See Table 5). One typical view of core sam-

ple is shown in Fig. 7. Compressive strength and

static elastic modulus were obtained from the com-

pressive test of core samples. From the compressive

test results, peak load of core samples was obtained

and compressive strength of core samples were cal-

culated. It was observed that average compressive

strength for several damaged specimens was 13.2

MPa and whereas average compressive strength of

53 MPa was observed for Mild damaged core. Figure

8 shows the compressive strength results of the core

samples. Static elastic modulus was calculated and

the results are compared in Fig. 9. Minimum static

elastic modulus of 5 GPa was recorded in the Severe

Undam-

aged New

Undam-

aged Old Mild

damage

Severe

damage

Undam-

aged New

Undam-

aged Old Mild

damage

Severe

damage

Fig. 8 – Compressive strength of core samples Fig. 9 – Static elastic modulus of core samples

Table 6 – Number of tested sleepers

based on manufacturing year

Manufacturing

year Number of

sleepers

2015 3

1986 3

2006 3

2006 3

2002 3

Table 7 – Number of tested sleepers based on level of damage

Category Number of sleepers and

manufacturing year

Undamaged new 3 Nos. in 2015

Undamaged old 3 Nos. in 1986

Mild damage 1 Nos. in 2006 1 Nos. in 2002

Moderate damage 3 Nos. in 2006

Severe damage 2 Nos. in 2006 2 Nos. in 2002

242

180 18

22

1195 280

23

5

497.5

22

Transducer

73

Sleeper

Frame

Support 280 497.5

Fig. 10 – Schematic view of flexural test arrangement (Dimensions in mm)

109


damaged core, whereas maximum static elastic mod-

ulus is recorded in undamaged old core of about 37

GPa (Fig. 8). Even though no visual cracks were ob-

served in the cores, reduction in material strength up

to 75% was observed in Severe damaged sleepers

when compared to Mild damaged sleepers. It was

found from the study that outer crack is an indication

to show the capacity reduction in sleepers, but this is

not the main reason for capacity reduction in cores.

Main reason is concrete is affected in material level

due to expansion in concrete due to DEF and ASR.

(a) Undamaged sleeper (b) Mild damaged sleeper

(c) Moderate damage sleeper (d) Severe damage sleeper

Fig. 11 – View of crack patterns of sleepers before loading

Fig. 12 – Result of flexural test

Side view - one side

Side view - other side

Top view

No cracks

Side view- other side


Top view

1 mm thick cracks


Side view - other side

Top View

b) Mild damage Sleeper

Side View- One Si

de

Two layers LongitudinalSide view- other side

Side view- one side

Top View

110


Fig. 13 – Typical view of sleeper after failure (Severe damaged)

Fig. 14 – Side view – comparison of capacity in damaged sleeper

4. Loading test for sleepers

4.1 Flexural test

Fifteen sleepers with different levels of damage

were collected from Indian Railways (See Table 6).

Within these 15 sleepers, sleepers were further cate-

gorized into five categories such as “Undamaged

new”, “Undamaged old”, “Mild damage”, “Moder-

ate damage” and “Severe damage” based on the level

of damage (See Table 7). Flexure test was conducted

to find out the flexural capacity of the sleepers (See

Fig. 10). Transducers were fixed at the loading

points and at the center part of the sleeper to find

midspan deflection. Sleepers were loaded in flexure

and allowed to be loaded until the failure is observed

from the sleeper [1].

1-mm-thick cracks

2-mm-thick cracks

3-mm-thick cracks

111


Fig. 15 – Schematic shear test arrangement

280 280 1195 280 280

Transducer

180 73

37 12

242

Sleeper

Frame

Support

280 280 1195 280 280

Transducer

Sleeper

Frame

Support

a) Typical shear test arrangement on one side

b) Typical shear test arrangement on the other side

a) Undamaged New

b) Severe Damaged Sleeper 3 mm cracks

Table 8 – Number of tested sleepers

based on level of damage for shear test

Category Manufacturing

year Number of

sleepers

Undamaged

new 2015 1

Severe dam-

age 2002 3

Fig. 16 – Typical view of crack patterns of sleepers

112


Fig. 19 – Existing crack patterns and shear test results

Fig. 17 – Result of shear test

Existing cracks Shear cracks Bending cracks Shear cracks Bending cracks

Existing cracks

a) View after failure in one side b) View after failure in other side

Fig. 18 – Typical view of sleeper after shear failure

2 to 3 mm thick, two to

three layers of crack

No cracks

End cracks

113


4.1.1 Results of flexure test

From the flexure test results, peak loads were

obtained for all categories of sleepers (See Fig. 12).

It is clearly seen that not only peak load but also stiff-

ness is significantly reduced in the case of severe

damaged sleepers. One of the typical views of

sleeper after the failure is shown in Fig. 13. Mode of

failure is supposed to be flexural compression failure

which occurs before yielding of PC strands.

4.1.2 Relationships between flexural capacities

and visible crack pattern

From the flexural test results, it shows that Se-

vere damaged sleepers show capacity reduction (See

Fig. 14). Average peak load of Severe damaged

sleeper is 56% capacity reduction. Average peak

load of Moderate damaged sleeper is same as Un-

damaged old sleepers, but in one of the cases Mod-

erate damaged sleeper shows 8% capacity reduction.

Average peak load of Mild damaged sleeper is 7%

less than Undamaged old sleepers, but in the cases

Mild damaged sleeper shows 12% capacity reduc-

tion. By comparing the results of both Undamaged

old and new sleepers less than 2% difference was ob-

served. If 2 to 3 layers of longitudinal cracks were

observed in the central part of sleeper these cracks

reduce the capacity of the sleeper.

Minimum peak load as per Indian standards

for flexural test is 60 kN. Even though severe dam-

aged sleeper showed reduction in peak load, still the

minimum peak load observed in these sleepers was

127 kN. Therefore, sleepers are safe in flexure. The

evaluation method in which damage level is classi-

fied are appropriate. The evaluation method in which

damage level is classified are appropriate, where

damage sleepers show lesser capacity compared to

undamaged sleepers.

4.2 Shear test Shear test was also conducted for sleepers. Il-

lustrated view of shear test arrangement is shown in

Fig. 15. Transducers were fixed near the loading

points and at the central part of the sleeper to find

midspan deflection. Shear test was conducted on one

side first and after completion the test was repeated

over the other side on the same sleeper. Sleepers

were categorized into “Undamaged new” and “Se-

vere damage” (See Table 8). Typical view of these

categories of sleepers is shown in Fig. 16.

4.2.1 Results of shear test

From the shear test results, peak loads were

obtained from Undamaged new and Severe damaged

sleeper (See Fig. 17). View of the sleeper after fail-

ure is shown in Fig. 18. Shear capacity and member

stiffness in severe damaged sleepers significantly re-

duced as observed in the flexure tests.

4.2.2 Relationship between shear capacities and

visible crack patterns

Figure 19 shows the existing crack patterns

available in sleepers. Average peak load of Severe

damaged sleeper is 36% less than the Undamaged

old sleeper, but in one of the cases Severe damaged

sleeper shows 42% capacity reduction. Minimum re-

quirement for the peak load as per Indian standards

is 230 kN. In one of the cases peak load is 250 kN

and this is close to the minimum requirement for the

peak load. Cracks were observed in severe damage

sleeper, where more layers of end cracks and more

layers of longitudinal cracks were observed in dam-

aged sleepers. It indicates that shear capacities are

also correlated.

Future deteriorations such as corrosion in the

mild and moderate damage sleepers seem not to be a

significant issue since the surrounding environment

of the sleepers is rather mild (no chlorides, no sul-

fate, etc.), therefore verifications of durability per-

formances are not an emergent problem. However,

considering that environmental conditions in the

other regions are different from those in this study,

investigation on the durability performance of dam-

aged sleepers is also an important issue in the future.

5. Conclusions

In this study, the evaluation method for damage

level by visual inspection was proposed for damaged

PC sleepers of Indian Railways. The proposed eval-

uation method was validated by comparing results on

the material properties and structural performances.

As a result, following conclusions are obtained.

(1) The evaluation method in which the damage

level is classified into “no damage”, “mild dam-

age”, “moderate damage” and “severe damage”

based on the width and number of cracks ob-

served on the outer surface of OC sleepers was

proposed.

(2) The PC sleepers were cut and the inner crack

patterns were investigated. As a result, the

cracks do not infiltrate to the deeper region even

in the case of “severe damage”. It indicates that

the expansion caused by DEF and ASR is not

uniform but it occurred only in the inner region.

Thus, crack patterns observed on the outer sur-

face are affected by not only amount but also

spatial distribution of the expansion strain.

(3) Compressive tests of core samples taken from

the damaged PC sleepers were conducted. As a

result, compressive strength and elastic modu-

lus of the concrete are significantly reduced es-

pecially in the cases of “severe damage”. DEF

114


and ASR affect the mechanical properties in the

material level.

(4) Loading tests of PC sleepers were conducted.

As a result, both flexural and shear capacities

were significantly decreased in the cases of “se-

vere damage”. Especially, shear capacities of

the “severe damage” cases closed to the mini-

mum requirement level of Indian Railways, in-

dicating that replacement of those sleepers is an

urgent issue. In addition, it was confirmed that

there is a good correlation between the struc-

tural performances and damage levels, indicat-

ing that the evaluation method for damage level

proposed in this study is appropriate to deter-

mine the priority for the replacement.

(5) Mild and Moderate damaged sleepers can be

continued to use by Indian Railways. When

compared material strengths and capacity of the

Mild and Moderate damaged sleepers to those

of the undamaged new core samples, significant

reductions were not noticed. However, consid-

ering the further damage progress in the future,

it is recommended that damaged sleepers are re-

placed from severer ones.

(6) Due to the presence of high alkali content (5 %)

in cement samples. Recommendation has been

made to Indian Railways to maintain alkali con-

tent to less than 3 %.

References

1. Awasthi, A.; Matsumoto, K.; Nagai, K;

Asamoto, S.; and Goto, S. (2017) “Investigation

on Possible Causes of Expansion Damages in

Concrete – a Case Study of Sleepers in Indian

Railways,” Journal of Asian Concrete Federa-

tion, 3(1), pp. 49-66.

2. Sahu, S. and Thaulow, N. (2004) “Delayed

Ettringite Formation in Swedish Concrete Rail-

road Ties,” Cement and Concrete Research, 34,

pp. 1675-1681.

3. Ghorab, H. Y.; Heinz, D.; Ludwig, U.; Mesken-

dahl, T.; and Welter, A. (1980) “On the stability

of calcium aluminate sulphate hydrates in pure

systems and in cements”, 7th Intnl. Congr. Chem.

Cem., Vol IV, pp. 496-503.

4. Hanehara, S. and Oyamada, T. (2010) “Repro-

duction of delayed ettringite formation (DEF) in

concrete and relationship between DEF and al-

kali silica reaction,” Monitoring and Retrofit-

ting of Concrete Structures- B. H. Oh, et al.

(eds), Korea Concrete Institute, Seoul, ISBN

978-89-5708-181-5.

5. Collepardi, M. (2003) “A state-of-the-art review

on delayed ettringite attack on concrete,” Ce-

ment and Concrete Composites, 25, pp. 401-407.

6. Santos, S. (2004) “Diagnosis and Prognosis of

Portuguese Concrete Railway Sleepers Degra-

dation – a Combination of ASR and DEF.” 13th

International conference on AAR, Norway.

7. Hakan T. (2006) “Assessment of Fatigue Re-

sistance and Strength in Existing Concrete

Structures,” Doctoral thesis, ISBN 978-91-

85685-03-5, pp. 133-148.

8. Tepponen, P.; and Eriksson, B. (1987) “Dam-

ages in Concrete Sleepers in Finland,” Nordic

Concrete Research, 6, pp. 199-209.

9. Mielenz, R. C.; Marusin, S. L.; Hime, W G.; and

Jugovic, Z. T. (1995) “Investigation of pre-

stressed concrete railway tie distress,” Concrete

International, 17(12), pp. 62-8.

115


ACF Journal


ACF China ChapterAssociation of Structural Concrete (Russia)Concrete Institute of Australia Indian Concrete InstituteIndonesia Concrete InstituteJapan Concrete InstituteKorea Concrete InstituteMongolian Concrete AssociationSingapore Concrete InstituteTaiwan Concrete InstituteThailand Concrete InstituteVietnam Concrete Institute

is supported by following member institutes ofAsian Concrete Federation

Documents

ACF Journal - Asian Concrete Federation€¦ · tive property of strain hardening Ultra High Perfor-mance Fiber Reinforced Cementitious Composite (UHPFRC), so the accurate and representative