Feasibility study on real-scale, self-healing concrete slab by developing a [609291]

Feasibility study on real-scale, self-healing concrete slab by developing a
smart capsules network and assessed by a plethora of advanced
monitoring techniques
Eleni Tsangouria,⇑, Jordy Lelona, Pieter Minneboa, Hisafumi Asaueb, Tomoki Shiotanib,
Kim Van Tittelboomc, Nele De Beliec, Dimitrios G. Aggelisa, Danny Van Hemelrijcka
aDept. Mechanics of Materials & Constructions, Faculty of Engineering Sciences, Vrije Universiteit Brussel, 1050 Brussels, Belgium
bLaboratory on Innovative Techniques for Infrastructures ITIL Kyoto, Graduate School of Engineering, Kyoto University, 615-8540 Kyoto, Japan
cMagnel Laboratory for Concrete Research, Dept. Structural Engineering, Faculty of Engineering & Architecture, Ghent University, Technologiepar k-Zwijnaarde 60,
9052 Ghent, Belgium
highlights
/C15Vascular network of macro-tubes that carries healing agent into concrete is designed.
/C15Healing feasibility tested on real-size concrete slab loaded in cycles under bending.
/C15Crack formation, agent filling and re-cracking tracked by plethora of monitoring methods.
/C15Cracks locally repair proven by pulse velocity regain after healing.
/C15Elastic wave tomography and pulse emission on drilled cores verify crack healing.
article info
Article history:
Received 7 May 2019
Received in revised form 9 August 2019Accepted 21 August 2019
Keywords:
Autonomous crack healingConcretePlane capsules networkAcoustic EmissionElastic Wave Tomography
Ultrasound Pulse Velocity
Digital Image CorrelationIntegrated monitoring systemabstract
The study presents the design, manufacturing and testing of the first concrete structural element that car-
ries a plane network of long, brittle macro-capsules into which chemical repair agent circulates to pro-vide autonomous and repeatable healing of formed cracks under service loads. The vascular network
design emerges as the most promising concept based on a decade of research, the milestones of which
are briefly reported. In this experimental attempt, the network of simultaneously occurring and interact-ing cracks formed on a real-scale steel reinforced concrete slab loaded under four-point bending aretracked by an integrated sensing configuration that combines advanced acoustic monitoring techniques
(Acoustic Emission, Elastic Wave Tomography, Ultrasound Pulse Velocity measurements) with Digital
Image Correlation and visual crack inspection. Macro-capsules rupture is depicted by AE monitoring ofhigh energy burst signals. It is shown that effective healing occurs locally, detected by sealing of open
cracks and regain in mechanical properties. Indicatively, pulse velocity regain up to 100% is detected
on healed zones, result also obtained by elastic wave tomography. Limited repair is measured on crackswith openings larger than 0.5 mm. This outcome could only be verified by correlating the experimental
evidence of different monitoring methods, highlighting the need for design optimization and establish-
ment of an advanced tempo-spatial structural health monitoring protocol.
/C2112019 Elsevier Ltd. All rights reserved.
1. Introduction
1.1. An outline of self-healing concrete’s history
The history of self-healing concrete can be separated in three
phases. In the early 1990s the concept is introduced by the
pioneering work of Dry designing a configuration that permits in
time release of repair chemicals from fibers embedded into the
https://doi.org/10.1016/j.conbuildmat.2019.116780
0950-0618/ /C2112019 Elsevier Ltd. All rights reserved.⇑Corresponding author.
E-mail addresses: Eleni.Tsangouri@vub.be (E. Tsangouri), Jordy.Lelon@vub.be
(J. Lelon), Pieter.Minnebo@vub.be (P. Minnebo), Asaue.Hisafumi.7a@kyoto-u.ac.jp
(H. Asaue), Shiotani.Tomoki.2v@kyoto-u.ac.jp (T. Shiotani), Kim.VanTittelboom@u-
gent.be (K. Van Tittelboom), Nele.DeBelie@ugent.be (N. De Belie), Dimitrios.
Aggelis@vub.be (D.G. Aggelis), Danny.Van.Hemelrijck@vub.be (D. Van Hemelrijck).Construction and Building Materials 228 (2019) 116780
Contents lists available at ScienceDirect
Construction and Building Materials
journal homepage: www.elsevier.com/lo cate/conbuildmat

cementitious matrix (Phase 1) [1]. The study ends with a series of
queries that remained unanswered until a decade after. Inspired by
the study of White et al. that launched the first autonomously
healed polymer, research towards the same direction searching
for an autonomously healed concrete system has flourished in
the 2000s (Phase 2) [2]. By 2010, the research community agreed
that the optimal design for macro-capsule based healing consists
of a one-dimensional array of tubular capsules that carry
polymer-based healing agent and are embedded into concrete dur-
ing casting (illustrated in Fig. 1 a)[3]. The capsules are ruptured
due to stresses releasing the agent that fills the crack void achiev-
ing in this way almost instant crack sealing and mechanical
restoration [4]. The capsules nature was selected to be brittle
enough to break at the moment of matrix crack formation, but
strong enough to survive the concrete mixing process and alsothe aggressive concrete environment [5,6] .
The path has been cleared towards autonomous healing con-
crete, though the ideal configuration was still being sought.
Numerous parametric studies have emerged evaluating the cap-
sules geometry [7](thickness, length ( Fig. 1 b)[8], inner/outer
diameter [9]), the capsules nature [10] (from soft elastomers and
polymers [11] to brittle glass, ceramics and cement [9]), the heal-
ing agent chemistry [12,13] , the capsules-concrete interface bond-
ing, the crack-healing agent interaction [14], etc.
The dawn of a new era in self-healing concrete was set around
2015 with a series of feasibility studies [15,16] that investigated
the design of vascular networks carrying healing agent and aiming
to provide repetitive sealing and regain in mechanical properties
in real-size construction elements. The agent can be stored into long
thin capsules [17] that additionally provide beneficial effect on the
concrete toughness and partly into reservoirs connected to the cap-
sules and attached to the concrete element ( Fig. 1 c). Internal pres-
sure controls the agent release in the presence of cracks that
cause capsule breakage. The agent circulates into the capsule net-
work due to applied pressure achieving simultaneous cracks healing
at different locations in the concrete. Repeatable healing is feasible
since additional healing agent can be delivered through the reser-
voirs. This optimized healing system permits two crack scenarios:
– Cracks are sealed and mechanically restored, therefore material
continuity is reinstated, and the element is again homogeneous
and robust. Under service loads, a crack forms at a different
position (not at the effectively healed region). Crack redistribu-
tion permits stress release and enhances the fracture toughness.
– The healed crack reopens under service loads, but crack propa-
gation is significantly delayed due to the presence of polymer-ized agent that interlocks the fractured concrete surfaces.
Fracture toughness increases due to the development of new
crack surfaces.
In this study, the feasibility and performance of a 2D capsule
network ( Fig. 1 d) is experimentally investigated by testing a real-
scale concrete slab under repeated loading. The damage onset,
healing activation and crack repair is assessed using a monitoring
system that combines the most promising inspection techniques
previously evaluated in small-scale concrete samples. In Section 2 ,
the selected experimental methods are presented and their effec-
tiveness on healing assessment is critically discussed. In Section 3 ,
the healing system configuration is described. In Section 4 , the
experimental outcome is presented and the main findings that
contribute to the healing system characterization are given.
It should be highlighted that this is the first time in literature
that repeatable healing is assessed on large scale concrete struc-
tural elements, therefore the study appears a first approach to
establish the healing concept in concrete technology and industry.
The need for design improvements is reported since up to now only
local crack sealing and regain in mechanical properties is achieved.
1.2. Measuring and sensing of autonomous healing by advanced
monitoring techniques
In most of the studies assessing the healing efficiency an estab-
lished test procedure is followed: regardless the capsule design,
bending mode is used for crack creation and the load is applied
in loading/reloading cycles:
– At the first test cycle a unique crack or several cracks form and
propagate, triggering capsule rupture and activating the healing
mechanism. The test is crack controlled and stops when a
macro-crack of a few hundred micrometers is formed. Initial
strength, stiffness and fracture toughness are measured.
– An up to 24 h pause is required to permit the agent to polymer-
ize in the crack void. Later, the bending test is repeated. The
healing efficiency is calculated based on the stiffness and
strength regain. The loading/reloading regime can be repeatedly
applied when capsule networks are used, and agent refilling is
possible through reservoir suppliers ( Fig. 2 c–d).
The use of advanced monitoring methods is the key behind the
fast progress of concrete healing technology [18]. Traditionally,
mechanical test analysis was accompanied by fracture models
and monitoring methods to understand the cracking phenomena
Fig. 1. Design configurations: a) Short tubular capsules; b) Long tubular capsules; c) Prototype pipes with 3D printed agent reservoirs; d) 2D network with no des and agent
suppliers.2 E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780

in concrete. Constitutive damage laws can describe in detail the
way that concrete responds to fracture. However, ‘healing revolu-
tion’ introduces the reverse to the ‘fracture process’: crack closureand mechanical restoration. The autonomous repair procedure
introduces a condition that cannot be assessed based on conven-
tional fracture mechanic tools, standardized material laws or tradi-
tional experimental techniques. The material response after
healing remains a black box and advanced or modified experimen-
tal tools are urgently needed to verify repair and decode its
mechanisms.
Phase 2 and 3 progresses in healing technology go along with
the redesign and establishment of advanced integrated monitoring
systems. It is imperative to configure a link between experimen-
tally measured fracture variables (strength, stiffness, toughness,
fracture energy) and continuous monitoring using numerous
non-destructive and other advanced methods. Acoustic Emission
(AE), Ultrasound Pulse Velocity (UPV), Digital Image Correlation
(DIC), Computed Tomography scanning, Capillary Water Absorp-
tion are invoked. It is shown that the synergy of different methods
provides a comprehensive understanding of autonomous repair in
concrete. Each method contributes at a distinctive moment of test-
ing. For instance, DIC measures the crack reopening after healing
[14]. AE detects the capsule rupture and sets the healing triggering
moment [19]. UPV verifies crack closure and material restoration
[20]. Below each method’s contribution is briefly presented.
1.2.1. Acoustic Emission (AE)
A number of acoustic transducers mounted on the concrete sur-
face can capture the elastic waves emitted during occurrence,
widening or reopening (after healing) of a crack [21]. The wave
arrival time and magnitude are considered to identify the fracture
source localization and mode. AE contributes by detecting in time
and in space the capsule rupture, and therefore the healing trigger-
ing mechanism [19,22] . The latter is important in large scale sam-
ples since the location of newly formed cracks is crucial for the re-
distribution of the healing agent through vascular capsule
networks.
1.2.2. Digital Image Correlation (DIC)
Full-field plane strain and deformation maps are key tools to
detect cracking onset, opening and closure after healing. Studieson small samples using a pair of high-resolution cameras and
post-processing speckle-pattern image analysis software success-
fully detected new cracks formed after healing and healed cracksreopening delayed in time [20]. Additionally, DIC analysis has been
used to detect debonding at the healed agent-concrete cracked sur-
face [14]. The DIC application is nowadays extended to track cracks
propagation on alternative healing concrete solutions (i.e. [23]).
1.2.3. Ultrasound Pulse Velocity (UPV)
Ultrasound velocity can be measured using a pair of emitter-
receiver transducers before and after each loading/reloading cycle
and during the healing agent polymerization process (discretemeasurements during curing). The velocity is associated to the
structural health condition of the sample. The measurement is
straight-forward in the case of small-scale samples carrying a
unique bending crack: velocity along the crack’s height increases
after healing indicating effective crack repair [24,25] . An alterna-
tive application was recently utilized to assess healing efficiency:
the piezoelectric transducers are embedded into the concrete dur-
ing casting [20]. In this way wave reflections are eliminated and
the measurement is focused on local zones where healing is pre-
sent. The studies have proven that embedded sensing measure-
ments are effective in gauge lengths up to 1 m [26]. The analysis
becomes more complex in larger samples carrying several cracks.
1.2.4. Other monitoring methods
Previous work of Van Tittelboom et al. has shown that capsule
rupture and agent release into the crack void can be visualized
by advanced computed tomography [3]. In post-mortem stage
and in small-size samples, the crack volume that is sealed was
measured and related to the agent rheology [8]. However, the
method cannot be applied in real-size structures since the sample
size can only be up to few hundred millimeters. Capillary Water
Absorption was previously applied to assess the sealing of cracks
after healing in small-scale concrete samples [8,27] . The method
obtains robust and quantitative results in the case that a unique
crack has been formed, but it is less effective in large scale concrete
samples where multiple connected cracks interact. The latter
cracks form several alternative flow paths, therefore the water flow
cannot be accurately mapped. Additionally, a modified test config-
uration was used on large-scale concrete beams and it was proven
Fig. 2. a) Aluminium nut with the embedded rod; b) Centering rod; c) Clay capsules after vitrification.E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780 3

that water penetration affects the healing process since hydration
modified the fracture process zone and interfered with the poly-
merization of the healing agent [8].
2. Case-study of the optimal healing concept
2.1. Capsules design
Long capsules were cast in stoneware clay (Ateliermasse Rot
2505) with grain sizes of 0–0.5 mm. Clay was selected after com-
paring the material’s physical and mechanical characteristics to
other materials commonly used in literature (i.e. borosilicate glass,
cement, ceramic, polymer). Clay is brittle enough to break in the
presence of cracks with a width of 0.1 mm, is cost-effective, can
survive the concrete casting procedure and is chemically inert to
the healing agent. Considering the long-life healing systems design,
clay does not degrade over time when being embedded into the
concrete aggressive environment. Moreover, clay’s plastic nature
makes it easily workable, therefore facilitates the capsules casting.
Vitrification and extrusion procedures are applied to optimize
the capsule design and eliminate clay’s porosity (to avoid the agent
absorption by the clay capsules and agent leakage at defective sec-
tions). The clay is embedded into a metallic container attached to
an aluminum nut with a die ( Fig. 2 a). At ambient conditions and
under hydraulic ram pressure, the clay is pushed towards and
through the die forming hollow tubes with outer diameter of
10.6 mm. A rod is used to set the hole in the middle with a diam-
eter of 2 mm ( Fig. 2 b). The capsules length was set at 270 mm
(Fig. 2 c). The extruded capsules were stored in ambient conditions
for 24 h after casting in the course of which the embedded water
progressively evaporated and in this way early brittle failure due
to rapid expansion was avoided. Afterwards, the capsules were
progressively heated up to 1250 /C176C (vitrification). At these high
temperatures sintering occurs and the clay becomes ceramic and
thus brittle enough to break in the presence of macro-cracks
formed in concrete.
2.2. Building of the vascular network
The vascular network is embedded at the middle zone of a con-
crete slab loaded under four-point bending and its size is limited to
cover the zone where tensile cracks will form ( Fig. 3 a).
The clay capsules are interconnected (to achieve healing agent
circulation through paths) using cylindrical nodes that carry open-
ings into which the capsules are fixed ( Fig. 3 b). Node openings are
set to connect two or four capsules ( Fig. 3 c) and set in different
angles in order to build a 30 /C176angle rhomboid capsule mesh
(Fig. 3 d). Nodes are 20 mm high, have a diameter of 40 mm and
are made of PVC.
PVC was selected since shaping of the node and drilling of open-
ings can be effectively done. In total, 17 nodes and 24 capsules
were set to prepare a healing network of 950 mm wide and
810 mm long.
Four nodes at the edges of the network are connected to flexible
plastic pipes to supply the healing agent into the network (their
position is indicated in Fig. 3 e). In the presence of cracks, the heal-
ing agent is manually poured into the pipes and travels throughout
the capsule network. The reservoir pipes are filled up to a certain
level and the drop of the agent volume is indicative of the amount
of agent distributed into the network.
The healing agent fast flowed through the network with the aid
of pumping. One can verify the agent circulation in the capsules by
tracking the level of agent into the flexible supply pipe. It was
observed that the agent level dropped a few centimeters per min-
ute after filling. The level progressively decreased lower the follow-ing hour. One hour after the healing agent was poured, the pipes
were filled with cleaning agent which is flushed through using
pressurized air and in this way the capsules were emptied and
cleaned again. The healing and cleaning cycles can be repetitive
as long as the pipes remain clean.
Fig. 3. a) Loading set-up and slab dimensions; b) Steel rebars mesh and vascular
network capsules embedded into the slab mold; c) Node and their geometry; d)Inside view of the slab where the vascular network is illustrated; e) Healing agentdeposit tubes; f) AE transducers position; g) Transmitter-receiver illustration of AEtransducers used for ultrasound pulse velocity measurements; h) Hammer tapping
points throughout the bottom of the slab; i) Core drilling positions.4 E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780

Polyurethane-based one-component resin (HA Flex SLV AF) is
selected as healing agent. The agent interacts with water and air,
expands and cures fast in ambient conditions. The agent was
selected based on previous research, is optimally viscous and it
can effectively seal cracks up to 0.5 mm wide [20]. In a recent
study, it was reported that the agent can effectively seal the crack
void and the agent-crack face interfacial bonding appears strong
enough to control and delay the re-opening of the cracks [14].A
flushing liquid solvent (Washing agent ECO) commonly used to
dissolve resin was selected as a cleaning agent.
2.3. Concrete composition and slab casting
Normal strength concrete was prepared and cast into rectangu-
lar wooden molds. The sample is demolded two days after casting
and cured in ambient conditions for 28 days before testing. The
material composition is given in Table 1 . The slab length was
4 m, the width is 1 m and the thickness 0.2 m. A reinforcement
mesh is made by ribbed steel rebars with diameter of 8 mm. The
grid opening is 150 mm ( Fig. 3 b). The steel reinforcement mesh
stands 35 mm from the slab’s bottom and just above the capsules
network.
2.4. Loading set-up
The slab was simply supported at both edges and loaded under
four-point bending. Two loading bars were used to apply dis-
tributed forces at the top side of the slab and the middle bending
span was set at 0.7 m ( Fig. 3 a). The test was displacement con-
trolled with a rate of 0.5 mm/min and the loading stopped as soon
as one of the forming cracks in the bending span got 0.2 mm wide.Then, the healing and cleaning agents were released and unloading
followed with a displacement rate of 2 mm/min. A second loading
cycle followed after 24 h, a time frame that permits agent curing
and effective crack healing. The cyclic loading continued as
described above until global failure.
2.5. Monitoring methods
2.5.1. Acoustic Emission
Eight AE transducers with resonance frequency response at
150 kHz (R15, Mistras Group) were mounted using magnetic hold-
ers on the slab’s bottom and side surfaces. The sensor position,
indicated in Fig. 3 f and given in Table 2 , is selected to cover the
middle bending span zone where cracks are expected to form.
The signals were pre-amplified by 40 dB and the signal acquisition
threshold amplitude was set at 35 dB. The AE wave features
(amplitude, duration, rise time, frequency, counts, etc.) and the
waveform shape were collected and digitally stored.
The acoustic source was localized using triangulation 3D local-
ization approach considering a constant wave propagation velocity
at 3500 m/s. AE hit- and event-based post-processing analysis was
performed.
2.5.2. Digital Image Correlation
A limited (due to loading setup configuration) zone of 500 mm
wide at the middle of one of the sides was painted with a randomlydistributed black-white speckle pattern, as illustrated in Fig. 3 a. A
pair of high-resolution CCD cameras were set in parallel, with a
stereo-angle of 30 /C176and facing the speckle pattern-covered side at
600 mm further. Lenses with focal length equal to 23 mm were
used. The cameras developed a stereovision optical system that
simultaneously captured images each 5 sec during testing using
Vic-Snap software. Vic-3D software is used to post-process the col-
lected images and obtain deformation and strain full-field patterns.
The horizontal strain
exx(in %) (along the x axis as defined in
Fig. 3 a) was used to visualize the fracture zone built up surround-
ing the formed cracks. The respective horizontal displacement U
(in mm) was used to measure the crack opening over time and
to stop the test as soon as one of the cracks reached an opening
of 0.2 mm.
2.5.3. Ultrasound Pulse Velocity
Each AE transducer can be used as transmitter of a burst high-
amplitude elastic signal. The transmitted signal was captured by
the AE transducers mounted on the sample surface permitting
the wave velocity calculation along several internal paths. Wave
velocities were measured at the end of each loading/unloading
cycle ( Fig. 3 g). In this way, analysis of the health condition of the
middle zone was achieved.
2.5.4. Acoustic Emission and Elastic Wave Tomography
AE tomography (AET) was applied to estimate the internal
velocity distribution during testing. The wave propagation velocity
through the excitation-AE receiver path is calculated considering
the path length and the propagation time obtained [28]. The
inverse of velocity, referred to as slowness is considered as compu-
tational input for discrete elements on the structure. Based on it,
the theoretical propagation time is obtained by a finite element
model. The element mesh size is considered equal to
0.15 m /C20.25 m. Each element slowness is revised to reduce time
difference and the iteration leads to accurate computation of slow-ness, and therefore to the velocity. AE tomography allows to use
random excitations, without a priori knowing the exact position
or excitation time, taking therefore advantage of the actual AE
sources. Considering different wave propagation paths over the
structure, a tomogram of the elastic wave velocity over the target
area is formed [29,30] .
AE tomography was performed with input of the AE events col-
lected under loading and unloading testing regimes. Events were
obtained from the bending span zone (0.7 m wide and 1 m thick).
Based on AE analysis, the events were localized at the tensile bot-
tom zone of the sample, therefore the analysis is limited only at the
bottom part of the slab with a thickness equal to 0.1 m. Elastic
wave tomography was re-applied after one year by hammering
excitations on the healed slab stored in ambient conditions. The
AE transducers were re-mounted on the previous positions and
were activated to collect the signals due to hammer tapping per-
formed throughout the bottom side of the sample. In detail, ham-
mer tapping was performed at discrete points every 200 mmTable 1
Concrete composition.
Weight percentage (%) Weight (kg/m3)
Sand 0/2 28 1620
Gravel 4/7 20 1150
Gravel 7/14 31 1790
CEM II 52.5 N 14 810Water 7 405Table 2
AE transducers position according to coordinate system set in Fig. 3 a.
# Sensor X (mm) Y (mm) Z (mm)
1 2200 1000 95
2 2200 0 953 2050 250 04 2050 250 05 1750 750 06 1750 750 07 1600 1000 958 1600 0 95E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780 5

(respectively, mesh size considered for the finite element model)
along the X and Y direction as illustrated in Fig. 3 h, covering this
way the slab bottom.
2.5.5. Core drilling and post-mortem ultrasound velocity
measurements
After hammer tapping and a year after the mechanical testing,
cylindrical cores were drilled at different locations at the middle
zone of the slab. The drilled samples have a diameter of 114 mm
and a height of 200 mm (slab’s height). The core drilling positions
are selected in an attempt to detect cracking zones where effective
healing has occurred. Therefore, the cracking pattern at the bottom
of the sample was used to set the drilling position. In total 12 cylin-ders were drilled. Ultrasound pulse velocity measurements were
obtained at each cylinder at 5 height levels (from the slab bottom
reaching the top side: 20/60/100/40/180 mm) using PULSONIC
portable system. The 50 kHz pulser emits a burst signal with a
magnitude of 2500 V and the 50 kHz receiver captures signals with
2 MHz sampling rate and 12-bit resolution.
3. Results
3.1. Mechanical response to damage and repair
DIC strain maps can be used to detect the crack onset and prop-
agation at loading/unloading cycles. Horizontal (
exxin %) strain
maps are presented at early (as cracks form) and severe (at the
end of each loading cycle) damage state in Fig. 4 . Three macro-
cracks are developed at the zone under DIC investigation. Theopening of each crack is measured at the bottom zone of the slab
and the total crack opening is plotted as well. Two more cracks
form further at the middle bending span but cannot be monitored
by DIC.
Based on the load-crack opening graph it is shown that at the
first loading cycle, the concrete elastically deforms until a load of
20 kN is reached. Beyond this point, cracks progressively form
and open wider and the steel rebar plastic deformation state is
reached approximately at a load of 30 kN. At the second loading
cycle, lower initial stiffness is measured than at the early loading
stage, proving that the existing cracks re-open. The latter can beindicative of insufficient healing or of less stiff elastic response of
the released agent [14]. Once again, loading stops soon after the
plateau of steel plastic deformation is reached. At the beginning
of the third loading cycle, plastic deformation of concrete cracks
is observed, as at early loading stage, there is already extended
crack opening. As the same cracks re-open, the plateau of steel
plastic deformation is exceeded and load drops as the total crack
opening is beyond 0.65 mm, an indication that global failure is
soon expected. Based on load-deflection curves, the stiffness recov-
ery is measured equal to 53% and 50% at the second and third load-
ing cycle relative to the initial loading cycle.
The fact that all the cracks reopened after the first healing cycle
and the stiffness at early loading stage did not restore its initial
value, can be indicative of limited crack sealing (meaning preven-
tion of liquids ingress) and ineffective healing, however this con-clusion cannot be reached considering only the surface crack
response (volume response of the crack is crucial) and since the
global stiffness of the sample is partially controlled by steel rebars
reinforcement. Further investigation using other monitoring meth-
ods may provide robust conclusions on healing efficiency.
An interesting observation concerns the cracks profile along the
sample height: the cracks propagate forming two branches and do
not follow a straight line; on the other hand, the crack bends. The
observation that the cracks bend lead to the conclusion that the
capsule network has an influence on the crack evolution, confirm-
ing what was found in small scale studies [9].
3.2. Healing activation by capsule breakage
The capsule rupture in the presence of cracks is the first proof
that the healing mechanism is activated and can effectively repair
the open cracks. Acoustic Emission hits magnitude analysis was
previously used to detect glass capsules rupture phenomena on
both small- and large-scale concrete beams [19,20] . Indicatively,
it was shown that glass rupture emits a burst and strong signal
with energy a scale greater than concrete cracking hits. Compared
to previous studies, in this study capsules are made of ceramic,
more brittle than glass and of nature similar to concrete, therefore
their rupture may emit hits lower in magnitude that are non-
detectable and masked by concrete cracking hits. Also, the testing
Fig. 4. DIC strain maps derived at early/severe damage state and load-crack opening graphs as measured considering part of the cracks at the middle bending sp an at the a)
first; b) second; c) third loading cycle.6 E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780

sample is enlarged, therefore wave attenuation effect may intro-
duce complexity on hits analysis.
The AE hits collected at each loading cycle were tracked and it
was found that in the data timeline, series of hits that are captured
by almost all transducers and localized as events, carry energy a
scale greater than the rest of hits simultaneously emitted. These
are classified and associated to ceramic capsules rupture. The rest
of AE hits are attributed to crack formation and propagation. AE
hits clustering was effectively achieved in this large-scale test
and the two groups of AE hits are illustrated in Fig. 5 . The number
of hits and events and the average values of their AE energy, dura-tion and amplitude are listed in Table 3 . Both duration and ampli-
tude appear to follow the same tendency to energy.
It is shown that the number of hits emitted due to capsule
breakage is limited compared to the hits population attributed to
concrete cracking. Respectively, few events, most of them at the
first loading cycle are localized at positions where capsules break.
Regarding the AE activity attributed to concrete failure, one
may notice that the emitted hits are low in magnitude throughout
testing. The scenario changes at the end of the third loading cycle,
as global failure is reached. At this stage, the hits number accumu-
lates, and their energy reaches peak values.
Hits derived by capsule breakage
Hits derived by concrete cracking Events derived by capsule breakage
Events derived by concrete cracking a)
b)
c) LOAD1 LOAD2 LOAD3
Fig. 5. AE hits energy distribution and AE events projected on X-Y plane at the a) first; b) second and c) third loading cycle. The AE activity is clustered and col ored differently
based on emission source: capsules breakage in red and concrete cracking and failure activity in black respectively.E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780 7

It should be highlighted that capsule breakage events are
detected even at the third loading cycle, as the sample reaches ulti-
mate failure. Based on Fig. 5 , the first capsule breakage occurs
when a load of 20 kN is reached and according to the DIC crack
opening measurements (see Fig. 4 a), at this moment the cracks
form and propagate. It is proven that cracks onset is the trigger
of capsule rupture. Beyond this point, the AE hits attributed to cap-
sule rupture are well spread in time reaching the end of each load-
ing cycle.
At the following loading cycles, capsule rupture is detected at
higher load levels, phenomena attributed to potential new crack
formation or further propagation of open cracks.
Additionally, the AE events are projected on an X-Y plane to
illustrate their spatial distribution. The crack path cannot be
detected since the AE events are widely spread covering therespective fracture zone built up surrounding these cracks. The
AE events attributed to capsule breakage are similarly spread in
space.
It should be noted that the number of hits and events at the sec-
ond loading cycle is limited compared to the initial and third load-
ing cycle. This can be related to the absence of new cracks after
healing and potential elastic reopening of existing cracks.
3.3. Detection of local healing phenomena using UPV matrix generator
Subsequently, the cracks effective sealing and repair was inves-
tigated. Ultrasound pulse velocity measurements were performed
at the end of the loading and unloading cycles as each AE trans-
ducer transmits a strong burst elastic wave that was received by
the other sensors. Considering 27 pairs of transducers, 27 local
paths are built. Indicatively in Fig. 6 , the velocity evolution is pre-sented as measured along two lines standing at the middle span of
the slab.
The velocity along the black colored line appears low at the end
of the first loading cycle indicating the presence of multiple cracks
in this zone. After agent release (unLOAD1 cycle) and further agent
polymerization (LOAD2), the velocity seems not to be regained and
beyond this stage, the velocity progressively drops further. It is
concluded that the black colored zone was damaged, but not effec-
tively healed. On the contrary, the red colored line stands along a
less damaged zone (initial velocity is higher compared to the one
above) and the velocities measured at the second and third loading
cycle appear restored and in unLOAD2 case equally high to initial
stage. The latter proves effective healing of cracks standing along
this line. This is a promising result, however, the zones where
velocity repair was obtained were few (9 out of 27 paths), thereforezonal healing may occur that cannot provide global mechanical
restoration.
3.4. Planar velocity measurements using AE tomography to detect
damage and local healing
AET developed at ITIL Laboratory (Kyoto Univ.) was used to
extend the above observation from line velocity measurements
to planar view of velocity distribution at different loading stages.
InFig. 7 a, the X-Y maps present the wave velocity distribution at
a horizontal plane standing at the bottom tensile zone of the sam-
ple. The AE events captured during testing (see Fig. 5 ) and localized
at the bottom zone of the slab are used as input to the AET compu-
tational analysis. The color maps are based on AE events collected
at each loading cycle comprising early and severe damage state
and at unloading stage.Table 3
AE hits, events population and average values of AE features for each loading cycle considering the AE hits clusters of capsule breakage and concrete c racking.
Capsule breakage Concrete cracking
Load 1 Load 2 Load 3 Load 1 Load 2 Load 3
Hits population 203 60 75 214,614 23,574 190,305
Events population 45 12 16 1235 60 271Energy (
lVs) 293.3 324.6 257.653 1.6 2.1 2.5
Duration ( ls) 2574.3 7005.1 2405.6 182.2 239.5 273.6
Amplitude (V) 0.4981 0.1764 0.2235 0.0122 0.0119 0.0123
Fig. 6. AE transducers set-up with red and black line illustrating the transmitter-receiver pair under consideration; Wave velocity as measured along the b ) black and c) red
line.8 E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780

Based on these velocity maps, one can qualitatively assess the
health condition of the sample at different loading stages. It is
shown that the cracks formed at the bending span lead to zonal
velocity drops at the bottom plane even at the 1st cycle and at
early damage state. The damaged zone at the red colored edges
appears to be expanded at the second loading cycle and damage
appears dominant only at the plane edges at the third loading
cycle. The velocity distribution is indicatively presented in Fig. 7 b
for a zone at the left side of the plane. It is shown that the velocitysignificantly drops as severe damage occurs at the end of the first
loading cycle, but velocity restores its value at the unloading stage
as crack closure occurs.
Effective healing is proven at this local zone since the velocity
does not drop further beyond this point and maintains relatively
high until failure of the slab. In another zone colored in black
(Fig. 7 c), velocity regain appears only after agent polymerization
into the cracks void, therefore velocity regain is obtained only at
early loading stage of the second and third loading cycle.
Fig. 7. a) Velocity maps at middle span at different loading stages (A-G); Wave velocity as measured at b) red; c) black colored zone.E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780 9

It should be highlighted that the UPV measurements described
in Section 4.3 consider waves that are emitted and captured by
sensors attached at the slab bottom plane. Therefore, the UPV mea-
surements are obtained at the crack level with the greatest open-
ing. In contrast to that, AE tomography mapping represents the
velocity distribution in volume considering the AE events dis-tributed throughout the bottom zone of the slab. The latter is the
reason for a relative overestimation of ultrasound pulse velocity.
A year after testing, elastic wave velocity computational analy-
sis was performed using hammer tapping as excitation at numer-
ous points throughout the slab bottom. The outcome is presented
at the map in Fig. 8 e. The velocity is significantly restored at the
Fig. 8. a) X-Y total AE events map; b) X-Y wave velocity map at the third loading cycle; c) cracks pattern after testing; d) cracks drawing based on previous figur e; e) wave
velocity map a year after testing.
a) b) c)
d) e)
Polymerized agent out of the capsule
Fig. 9. a) Drilled core with crack; b) Ultrasound velocity for three cylindrical samples; c) Sealed crack at the sample bottom; d) Crack with opening wider tha n 0.5 mm;
e) Yellow colored zone at the crack plane where agent is released and detail of polymerized agent standing at the surrounding of a capsule.10 E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780

bending span since no more red colored zones are depicted and the
lowest velocity measured is equal to 2887 m/s (at location high-
lighted with a (*) symbol).
For reasons of completeness, the total AE events map ( Fig. 8 a),
the wave velocity map at the end of the third loading cycle
(Fig. 8 b) and a drawing of cracks ( Fig. 8 d) obtained based on visual
inspection ( Fig. 8 c) after testing are presented. Assembling the
findings derived from the analyses above indicates the complexity
of the fracture process that can only be effectively assessed by
monitoring damage from different perspectives.
3.5. Core drilling in cracked zones to detect effective healing
After hammer tapping, cores were drilled throughout the slab
height (see Fig. 9 a) at different positions (mapped in Fig. 3 i) and
the cylinders are afterwards examined by ultrasound pulse veloc-
ity. The transmitter-receiver pair of sensors was mounted perpen-
dicular to the crack open faces. Three representative cases are
discussed in this section: two cores carrying cracks with a width
at the bottom smaller than 0.5 mm and a third carrying a crack
wider than 0.5 mm. The wave velocity was measured at five differ-
ent heights (20/60/100/140/180 mm) along the cracks and the
result is presented in Fig. 9 b.
It is shown that effective healing was obtained only in the case
of the core colored in red ( Fig. 3 i,Fig. 9 c), where wave velocity has
a relatively high value and remains almost constant along the sam-
ple height. On the other hand, the blue colored core has low veloc-
ity at the bottom zone and velocity gets higher only at higher levels
for zones out of the vascular network where healing cannot be
obtained. The black colored line plots the velocity of a cylinder that
carries a crack with opening wider than 0.5 mm at the slab bottom
(Fig. 9 d), therefore velocity values are relatively low at the bottom
and middle zone of the sample. Based on the above observation, it
is proven that sealing is effective only for small size cracks. As
macro-cracks widen, the crack void cannot be fully sealed by the
released agent, therefore limited and zonal healing can potentially
occur.
A cylinder was later split in two pieces and the crack plane was
visually inspected to detect agent release and absorption at the
crack plane. Agent release was proven as presented in Fig. 9 e and
furthermore polymerized agent was detected at the surrounding
of a capsule.
4. Conclusions
4.1. Healing system feasibility study
The plethora of results obtained by combining different moni-
toring methods are puzzled together towards the characterization
of the healing system. It is concluded that effective healing on large
scale concrete structures using this network of vascular capsules
can only be local. The mechanical features (load-displacement
graphs, stiffness or strength regain after healing) cannot be consid-
ered for the healing assessment due to structural and loading com-
plexity. The combination of optical and acoustic inspection
provides an insight view of damage and repair that can effectively
track crack sealing and material continuity restoration.
Based on different ultrasound pulse velocity measurements, it is
concluded that crack filling with healing agent is evident, therefore
the healing agent circulation and deposit is successfully obtained.
The crack repair is invoked since the agent effectively polymerizes
into the crack void. However, it is noted that cracks with a larger
volume cannot be treated since only limited sealing and almost
no mechanical restoration is measured. Therefore, the vascular
network should be tuned to repeatably and on time provide suffi-cient agent as cracks form and are restricted to less than 0.5 mm
opening.
4.2. Sensing tool for future industrial application
The brief overview of advanced monitoring methods has shown
that the transition from small- to large-scale concrete elements
introduces challenges for the monitoring process [31]. The pro-
posed sophisticated monitoring system combines different tech-
niques that have been previously tested in small-scale healing
studies. Upscaling to real size structures, one can verify that ultra-
sound pulse velocity measurements and acoustic emission are two
approaches that can detect healing triggering and efficiently char-
acterize crack repair. Especially the use of AE tomography appears
promising since the method provides both qualitative and quanti-
tative full-field information of crack evolution.
However, other techniques that provided essential information
on small-scale tests are used as supplementary methods that verify
the findings of other techniques in voluminous concrete tests.
Indicatively, Digital Image Correlation contributes by partially
tracking the crack formation, but the method is restricted only
on the measurement of surface movements. Visual inspection
and core drilling are interesting post-mortem tools, but their appli-
cation in the field appears limited as well for practical reasons. In
any case, experience obtained at both scales provides very interest-
ing feedback for the design of a healing system applied on real con-
crete elements.
Future studies should focus on the performance of concrete ele-
ments carrying vascular networks and tested under repetitive
incremental and/or dynamic loading. This way, the feasibility of
autonomous healing will be evaluated in realistic loading configu-
rations, reaching progressively the construction materials market
[32].
Declaration of Competing Interest
None.
Acknowledgments
Financial support of the Research Foundation Flanders (FWO-
Vlaanderen, Project No 28976) and SIM-SECEMIN research project
under the program SHE for this study is gratefully acknowledged.
References
[1]C.M. Dry, Three designs for the internal release of sealants, adhesives, and
waterproofing chemicals into concrete to reduce permeability, Cem. Concr.
Res. 30 (12) (2000) 1969–1977 .
[2]S.R. White, N.R. Sottos, P.H. Geubelle, J.S. Moore, M. Kessler, S.R. Siriam, E.N.
Brwon, S. Viswanathan, Autonomic healing of polymer composites, Nature 409(6822) (2001) 794
.
[3]K. Van Tittelboom, N. De Belie, D. Van Loo, P. Jacobs, Self-healing efficiency of
cementitious materials containing tubular capsules filled with healing agent,
Cem. Concr. Compos. 33 (4) (2011) 497–50533 .
[4]M. Wu, B. Johanneson, M. Geiker, A review: self-healing in cementitious
materials and engineered cementitious composite as a self-healing material,
Constr. Build. Mater. 28 (1) (2012) 571–583 .
[5]K. Van Tittelboom, N. De Belie, Self-healing in cementitious materials-A
review, Materials 6 (6) (2013) 2182–2217 .
[6]A. Al-Tabbaa, C. Litina, P. Giannaros, A. Kanellopoulos, L. Souza, First UK field
application and performance of microcapsule-based self-healing concrete,
Constr. Build. Mater. 208 (2019) 669–685 .
[7]A. Kanellopoulos, P. Giannaros, D. Palmer, A. Kerr, A. Al-Tabbaa, Polymeric
microcapsules with switchable mechanical properties for self-healing
concrete: synthesis, characterisation and proof of concept, Smart Mater.
Struct. 26 (2017) .
[8]K. Van Tittelboom, E. Tsangouri, D. Van Hemelrijck, N. De Belie, The efficiencyof self-healing concrete using alternative manufacturing procedures and more
realistic crack patterns, Cem. Concr. Compos. 57 (2015) 142–152
.
[9] E. Tsangouri, F.A. Gilabert, D. Aggelis, N. De Belie, D. Van Hemelrijck, Concrete
Fracture Energy Increase by Embedding Capsules with Healing Ability: TheE. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780 11

Effect of Capsules Nature, in 2nd International RILEM/COST Conference on
Early Age Cracking and Serviceability in Cement-based Materials and
Structures – EAC2, Brussels, pp. 175-181, 2017.
[10] A. Kanellopoulos, A. Litina, P. Giannaros, A. Al-Tabbaa, Effect of Different Types
of Polymeric Microcapsules on the Self-Healing Efficiency of Cement Based
Composites, 9th International Conference on Fracture Mechanics of Concreteand Concrete Structures (FramCos-9), 2016
.
[11] B. Hilloulin, K. Van Tittelboom, E. Gruyaert, N. De Belie, A. Loukili, Design of
polymeric capsules for self-healing concrete, Cem. Concr. Compos. 55 (2015)
298–307 .
[12] K. Van Tittelboom, K. Adesanya, P. Dubruel, P. Van Puyvelde, N. De Belie,Methyl methacrylate as a healing agent for self-healing cementitious
materials, Smart Mater. Struct. 20 (12) (2011) 125016
.
[13] A. Kanelllopoulos, T. Qureshi, A. Al-Tabbaa, Glass encapsulated minerals for
self-healing in cement-based composites, Constr. Build. Mater. 98 (2015) 780–
791.
[14] J. Feiteira, E. Tsangouri, E. Gruyaert, C. Lors, G. Louis, N. De Belie, Monitoring
crack movement in polymer-based self-healing concrete through digital imagecorrelation, acoustic emission analysis and SEM in-situ loading, Mater. Des.
115 (5) (2017) 238–246
.
[15] H. Huang, G. Ye, Z. Shui, Feasibility of self-healing in cementitious materials–
By using capsules or a vascular system?, Constr Build. Mater. 63 (2014) 108–
118.
[16] P. Minnebo, G. Thierens, G. De Valck, K. Van Tittelboom, N. De Belie, D. Van
Hemelrijck, E. Tsangouri, Novel design of autonomously healed concrete:
towards vascular healing network, Materials 10 (1) (2017) 49 .
[17] S. Sangadji, E. Schlangen, Self-healing of concrete structures-novel approach
using porous network concrete, J. Adv. Concr. Technol. 10 (5) (2012) 185–194 .
[18] D. Snoeck, F. Malm, V. Cnudde, C.U. Grosse, K. Van Tittelboom, Validation of
self-healing properties of construction materials through nondestructive and
minimal invasive testing, Adv. Mater. Interfaces 5 (17) (2018) 1800179 .
[19] E. Tsangouri, D.G. Aggelis, K. Van Tittelboom, N. De Belie, D. Van Hemelrijck,
Detecting the activation of a self-healing mechanism in concrete by acoustic
emission and digital image correlation, Sci. World J. 2013 (2013) .
[20] E. Tsangouri, G. Karaiskos, D.G. Aggelis, A. Deraemaeker, D. Van Hemelrijck,Crack sealing and damage recovery monitoring of a concrete healing system
using embedded piezoelectric transducers, Struct. Health Monit. 14 (5) (2015)
462–474
.
[21] C. Grosse, F. Malm, Combination of inspection and monitoring techniques for
the detection of fractures in concrete with self-healing properties, 9thInternational Conference on Fracture Mechanics of Concrete and ConcreteStructures (FraMCoS-9), 2016
.
[22] K. Van Tittelboom, N. De Belie, F. Lehmann, C. Grosse, Acoustic emission
analysis for the quantification of autonomous crack healing in concrete,
Constr. Build. Mater. 28 (1) (2012) 333–341 .
[23] N. Hasheminejad, C. Vuye, A. Margaritis, B. Ribbens, B. Jacobs, J. Blom, J. Dirckx,S. Vanlanduit, Investigation of crack propagation and healing of asphalt
concrete using digital image correlation, Appl. Sci. 9 (12) (2019) 2459
.
[24] E. Gruyaert, J. Feiteira, N. De Belie, M. Malm, C.U. Grosse, E. Tziviloglou, E.
Schlangen and E. Tsangouri, Non-destructive testing techniques to evaluate
the healing efficiency of self-healing concrete at lab-scale, in 6th InternationalConference on Emerging Technologies in Non-Destructive Testing, Brussels,
2015.
[25]
N.P. Kaur, J.K. Shah, S. Majhi, A. Mukherjee, Healing and simultaneous
ultrasonic monitoring of cracks in concrete, Mater. Today Commun. 18
(2019) 87–99 .
[26] G. Karaiskos, E. Tsangouri, D.G. Aggelis, A. Deraemaeker, D. Van Hemelrijck,
Damage detection monitoring applications in self-healing concrete structuresusing embedded piezoelectric transducers and recovery, J. Phys. Conf. Ser. 628
(1) (2015) 012110
.
[27] W. Li, Z. Jiang, Z. Yang, N. Zhao, W. Yuan, Self-healing efficiency of
cementitious materials containing microcapsules filled with healing
adhesive: mechanical restoration and healing process monitored by waterabsorption, PLoS ONE 8 (11) (2013) e81616
.
[28] Y. Kobayashi, T. Shiotani and H. Shiojiri, Damage identification using seismic
travel time tomography on the basis of evolutional wave velocity distribution
model, in Structural Faults and Repair, 2006.
[29] T. Shiotani, N. Okude, S. Momoki, Y. Kobayashi, Proposal of assessment methodfor infrastructures by AE tomography (in Japanese), National Conference on
Acoustic Emission, 2011
.
[30] Y. Kobayashi, T. Shiotani, Seismic tomography with estimation of source
location for concrete structure, in Structural Faults and Repair, Edinburgh,
2012.
[31] R. Davies, O. Teall, M. Pilegis, A. Kanellopoulos, T. Sharma, A. Jefferson, D.
Gardner, A. Al-Tabbaa, K. Paine, R. Lark, Large scale application of self-healingconcrete: design, Constr., Test., Front. Mater. 5 (2018) 12
.
[32] K. Van Breugel, Is there a market for self-healing cement-based materials?, 1st
International Conference on Self-healing Materials, 200712 E. Tsangouri et al. / Construction and Building Materials 228 (2019) 116780