Civil, Construction and Environmental Engineering

Civil, Construction and Environmental Engineering

Publications

7-2013

Robust Flexible Capacitive Surface Sensor for

Structural Health Monitoring Applications
Simon Laflamme
Iowa State University, laflamme@iastate.edu

Matthais Kollosche
Universität Potsdam

Jerome J. Connor
Universität Potsdam

See next page for additional authors

Follow this and additional works at: http://lib.dr.iastate.edu/ccee_pubs

Part of the Structural Engineering Commons
The complete bibliographic information for this item can be found at http://lib.dr.iastate.edu/
ccee_pubs/17. For information on how to cite this item, please visit http://lib.dr.iastate.edu/
howtocite.html.

This Article is brought to you for free and open access by the Civil, Construction and Environmental Engineering at Iowa State University Digital
Repository. It has been accepted for inclusion in Civil, Construction and Environmental Engineering Publications by an authorized administrator of
Iowa State University Digital Repository. For more information, please contact digirep@iastate.edu.
Robust Flexible Capacitive Surface Sensor for Structural Health
Monitoring Applications
Abstract
Early detection of possible defects in civil infrastructure is vital to ensuring timely maintenance and extending
structure life expectancy. The authors recently proposed a novel method for structural health monitoring
based on soft capacitors. The sensor consisted of an off-the-shelf flexible capacitor that could be easily
deployed over large surfaces, the main advantages being cost-effectiveness, easy installation, and allowing
simple signal processing. In this paper, a capacitive sensor with tailored mechanical and electrical properties is
presented, resulting in greatly improved robustness while retaining measurement sensitivity. The sensor is
fabricated from a thermoplastic elastomer mixed with titanium dioxide and sandwiched between conductive
composite electrodes. Experimental verifications conducted on wood and concrete specimens demonstrate
the improved robustness, as well as the ability of the sensing method to diagnose and locate strain.

Keywords
Strain gauge, Structural health monitoring, Strain monitoring, Capacitive sensor, Dielectric polymer,
Stretchable sensor, Flexible membrane, Sensing skin

Disciplines
Structural Engineering

Comments
This is an author's final manuscript of an article from Journal of Engineering Mechanics 139 (2013): 879–885,
doi:10.1061/(ASCE)EM.1943-7889.0000530. Posted with permission.

Authors
Simon Laflamme, Matthais Kollosche, Jerome J. Connor, and Guggi Kofod

This article is available at Iowa State University Digital Repository: http://lib.dr.iastate.edu/ccee_pubs/17

 1 Robust Flexible Capacitive Surface Sensor for Structural Health

2 Monitoring Applications

3 Simon Laflamme 1 , A.M. ASCE

2
Matthias Kollosche
Jerome J. Connor 3 , F. ASCE
4
Guggi Kofod

4 ABSTRACT
5 Early detection of possible defects in civil infrastructure is vital to ensure timely mainte-
6 nance and extend structure life expectancy. A novel method for structural health monitoring
7 based on soft capacitors was recently proposed by the authors. The sensor consisted of an
8 off-the-shelf flexible capacitor that could be easily deployed over large surfaces, with the
9 main advantages of being cost-effective, easy to install, and allowing simple signal process-
10 ing. In this paper, a capacitive sensor with tailored mechanical and electrical properties is
11 presented, resulting in a greatly improved robustness while retaining measurement sensitiv-
12 ity. The sensor is fabricated from a thermoplastic elastomer mixed with titanium dioxide,
13 then covered with conductive composite electrodes. Experimental verifications conducted on
14 wood and concrete specimens demonstrate the improved robustness, as well as the ability of
15 the sensing method to diagnose and locate strain.
16 Keywords: structural health monitoring, strain monitoring, capacitive sensor, dielectric
17 polymer, stretchable sensor, flexible membrane, sensing skin
1
Department of Civil, Construction, and Environmental Engineering, Iowa State University, Ames, IA,
50011
2
Applied Condensed-Matter Physics, Institute of Physics and Astronomy, University of Potsdam, D-14476
Potsdam, Germany
3
Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge,
MA, 02139 USA
4
Applied Condensed-Matter Physics, Institute of Physics and Astronomy, University of Potsdam, D-14476
Potsdam, Germany

1
18 INTRODUCTION
19 Structural health monitoring (SHM), defined as methods for damage diagnosis and local-
20 ization, is expected to play a predominant role in management of large-scale infrastructure.
21 Their long lifespan, often combined with insufficient maintenance and increased utilization,
22 enhances the risks associated with their failure (Karbhari 2009; Harms et al. 2010). In the
23 U.S., the number of bridges classified as in need of repairs, due either to deficiency or obso-
24 lescence, is excessive, and estimated costs of repairs is beyond 2 trillions USD (Mason et al.
25 2009; Simon et al. 2010). Timely maintenance and inspection programs are essential in
26 improving health and ensuring safety of civil infrastructure (Brownjohn 2007), consequently
27 enhancing their sustainability.
28 The complexity of damage diagnosis and localization in large-scale infrastructure arises
29 from their inherent size, which impacts monitoring solutions both technically and economi-
30 cally. Damage localization techniques, such as non-destructive evaluation using pulse-echo,
31 dynamic response, or acoustic emission, are too expensive to deploy over the entire moni-
32 tored infrastructure for complete monitoring. In fact, known complete SHM systems rapidly
33 become economically unfeasible as the scale is increased.
34 The authors recently proposed a novel monitoring technique for civil structures (Laflamme
35 et al. 2012a), the sensing skin, consisting of soft elastomeric capacitors arranged in a matrix
36 form which could be deployed over an entire structure. This low-cost approach is capable
37 of discretely monitoring changes in strain over large areas without compromising structural
38 integrity, a combination of properties hardly found in other SHM technologies. It is anal-
39 ogous to biological skin, where local changes can be monitored over an entire surface, and
40 therefore represents one possible solution to the complexity of applying damage localization
41 methods globally.
42 The skin-type property of bio-inspired sensing methods arises from the use of flexi-
43 ble films, for which elastomeric substrates are common (Lumelsky et al. 2001). For in-
44 stance, Harrey et al. (Harrey et al. 2002) researched a humidity sensor using polymide and

2
45 polyethersulphone (PES) substrates. Zhang et al. (Zhang et al. 2003) and Hurlebaus & Gaul
46 (Hurlebaus and Gaul 2004) developed sensor skins using polyvinylidene (PVDF) substrates.
47 Carlson et al. (Carlson et al. 2006) and Tata et al. (Tata et al. 2009) opted for a poly-
48 imide substrate. Lipomi et al. (Lipomi et al. 2011) developed a stretchable capacitor made
49 from conducting spray-deposited films of single-walled carbon nanotubes on polydimethyl-
50 siloxane (PDMS). Carbon nanotubes have also been used as resistance-based strain sensors
51 (Gao et al. 2010), and for damage localization by measuring changes in electrical impedance
52 tomographical conductivity (Loh et al. 2009). Here, carbon nanotubes (carbon black) are
53 added to the SEBS substrate to form the sensor electrodes.
54 Other bio-inspired solutions have been proposed using stiff ceramic piezoelectric (PZT)
55 sensors. Often used in civil engineering for dynamic strain measurements (Zhang et al.
56 2003; Lin et al. 2009; Giurgiutiu 2009), PZT patches can be arranged in an array to form
57 piezoelectric wafer active sensors (PWAS), which allows damage localization (Zhang 2005;
58 Yu and Giurgiutiu 2009; Zagrai et al. 2010; Yu et al. 2011). Fiber optic-based health
59 monitoring techniques are also commonly used in the field. Their concept is analogous to a
60 nervous system, where fiber optic sensors can be deployed linearly for damage localization.
61 Large-scale deployments of fiber optic systems have been studied in Ref. (Chen et al. 2005;
62 Zou et al. 2006; Lin et al. 2007).
63 Previously, the sensing skin was constructed using a commercially available silicone
64 elastomer-type material (Danfoss PolyPowerTM ) with pre-defined smart metallic electrodes.
65 Results showed that the sensing solution was promising at monitoring large-scale surfaces,
66 capable of both detecting and localizing cracks on concrete specimens (Laflamme et al. 2010;
67 Laflamme et al. 2012a). Despite these promising results, practical issues arose due to the
68 concurrence of relatively low stiffness and thickness (80 µm) of the sensing material. In the
69 experiments presented by the authors in Ref. (Laflamme et al. 2012a), practical challenges
70 arose in the electrical contact on the electrodes, and small surface features on the tested
71 concrete specimens sometime led to tears in the sensing patches. Also, a few months after

3
72 the experiments were conducted and the results published, the authors noticed degradation
73 of the silver electrodes due to environmental condition. These issues inspired the authors’
74 efforts to develop a robust capacitive-based sensor using the same monitoring principle, with
75 the objective to improve the applicability, robustness, and life cycle of the sensing technology
76 for broad implementation onto civil infrastructure.
77 As it will be demonstrated in this paper, the replacement of materials results in a greatly
78 enhanced robustness, a fundamental property for SHM solutions applied to civil infras-
79 tructure. Here, both the dielectric elastomer and the sensing electrodes are replaced with
80 nanocomposites based on a thermoplastic elastomer and the sensor itself has a higher thick-
81 ness. It follows that this paper also demonstrates the generality of the sensing principle, and
82 the practical opportunities for improving the technology.
83 The paper is organized as follows. The next section describes the robust capacitive-based
84 sensor. The sensing principle is discussed, the sensor fabrication process is briefly described,
85 and the proposed sensor is verified. The subsequent section verifies the measurement method
86 in a laboratory setup. Here, the sensor robustness is demonstrated, along with its capacity
87 to diagnose and locate damages on wood and concrete specimens. The last section concludes
88 the paper.

89 ROBUST CAPACITIVE-BASED SENSOR

90 Sensing Principle

91 The sensing skin consists of several individual capacitive sensors, termed sensing patches,
92 which can have individually shaped areas. A single sensing patch is constructed using a soft
93 incompressible polymer membrane covered by highly compliant and stretchable electrodes
94 to form a stretchable capacitor.
95 The capacitance C of such a sensing patch depends on the surface area A, the thickness
96 d, the vacuum permittivity 0 , and the permittivity of the polymer M :

C = 0 M A/d (1)

4
 97 The change in the geometry of the capacitor due to deformation is directly related to its
98 capacitance, allowing measurement of high precision strain responses.
99 Fig. 1 illustrates a sensor patch deployed over a monitored surface. In the figure, a crack
100 causes a change ∆A in the sensor area, which can be measured as a change in capacitance
101 (Laflamme et al. 2010; Kollosche et al. 2011). A differential measurement mode can also be
102 employed in which two sensors are compared directly (Laflamme et al. 2010).

103 Sensor Fabrication

104 The dielectric material for the capacitive sensor is a nanocomposite with soft dielectric
105 elastomer and ceramic insulating nanoparticles. The elastomer matrix is poly-styrene-co-
106 ethylene-co-butylene-co-styrene, (SEBS - Dryflex 500120, ρ = 930 kg/m3 ), provided by VTC
107 Elastoteknik AB (Sweden). The incompressible thermoplastic SEBS is a soluble tri-block
108 copolymer, which forms a reversible physically crosslinked elastomer network (Laurer et al.
109 1996). The initial stiffness of the SEBS is Y = 312 kPa (Kollosche et al. 2009) and the
110 permittivity of the pure material is M = 2.2 (McCarthy et al. 2009). In this study the
111 handling and robustness of the capacitive sensors is improved by increasing the sensor film
112 thickness in the range of 200 µm to 400 µm. The increase in thickness leads to a reduction in
113 basic capacitance (Eq. (1)), which can be adjusted by increasing the permittivity M through
114 addition of nanoparticles. The authors have demonstrated a significant sensing enhancement,
115 approximately 46 times, aimed towards sensing skin for robots; this was achieved by grafting
116 a small amount of highly polarizable polyaniline (PANI) on the matrix elastomer backbone
117 to form a molecular composite (Stoyanov et al. 2010; Kollosche et al. 2011).
118 Here, an elastic nanocomposite of SEBS and ceramic nanoparticles of rutile titanium
119 dioxide TiO2 (R 320 D, ρ = 4200 kg/m3 , Sachtleben Chemie GmbH, Germany) is proposed
120 to form a soft capacitive-based sensor for direct application to health monitoring of civil
121 infrastructure. The nanoparticles have an average diameter of 300 nm, a static dielectric
122 constant of  = 128, and are coated with a silicon oil. This coating facilitates the dispersion
123 of the nanoparticles in the SEBS matrix, which improves the mechanical and dielectric prop-

5
124 erties of the composite (Stoyanov et al. 2011). The fabrication of the dielectric membrane
125 is as follows:

126 1. Two SEBS/toluene solutions are prepared by dissolving the polymer at concentrations
127 of 200 g/L and 300 g/L.
128 2. A mass of TiO2 particles is weighed to obtain a concentration of 15%vol of the pre-
129 pared elastomer solutions. The TiO2 is added to the dissolved elastomer and stored
130 in an ultra sonic bath for 1 hr to homogenize the mixture. The nanoparticles are
131 dispersed in the polymer solution using an ultra-sonic tip (high intensity ultrasonic
132 processor Vibracell 75041, Sonics & Materials Inc.,USA). The dispersion of the TiO2
133 particles within the polymer matrix is investigated for selected samples using scanning
134 electron microscopy (SEM) (Zeiss Ultra 55+, cryo-fracture with 5 nm sputtered Au,
135 2.0 kV). Fig. 2 shows the dispersion of the nanoparticles within the polymer matrix.
136 3. The polymer concentrations are drop-casted from solution on glass slides and covered
137 to control the evaporation rate. The samples are left to dry on a heat plate at 50◦ C
138 for 24 hr and then stored in an oven at 50◦ C for 4 more days to allow evaporation of
139 the remaining solvent.
140 4. All membranes are visual inspected and the thicknesses are measured using a mi-
141 crometer screw at randomly selected locations. Membranes with inhomogeneities
142 or varying thicknesses are discarded. The selected polymer concentrations result in
143 thicknesses of approximately 200 µm or 400 µm.
144 5. The membrane are peeled of from the glass slides and equipped with plastic masks on
145 both sides to define an area of dimension 70 × 70 mm for application of the electrodes
146 (Fig. 3(a)).

147 Subsequent to the fabrication of the elastomeric membrane, stretchable electrodes are
148 applied on both surfaces of the doped polymer. These elastic electrodes are made of carbon
149 black (CB) (Printex XE 2-B) particles dispersed in SEBS solution. The fabrication of the

6
150 electrodes is as follows:

151 1. A batch of SEBS/toluene solution is prepared by dissolving the polymer at a concen-

152 tration of 400 g/L, followed by the incorporation of the CB for a final concentration
153 of 15%vol.
154 2. The CB solution is stored in an ultra sonic bath for 2 hrs to disperse the CB. The
155 solution is sonicated with a ultra-sonic tip for 10 min and stored a second time in the
156 ultra sonic bath for 2 hrs to homogenize.
157 3. A mask is applied to the membrane to define an area of approximately 70 × 70 mm.
158 (Fig. 3(a)). The conductive elastomer solution is brushed onto the membrane, and let
159 drying for 20 min . The process is repeated onto the other side. Because the solvents
160 used to fabricate the membrane and the electrodes are identical, the CB solution is
161 coalesced with the dielectric membrane, forming a uniform conductive layer.

162 The completed process gives a highly stretchable capacitor of known dimensions, shown
163 in Fig. 3(b). The prepared sample is attached to plastic frames equipped with aluminum
164 strips to form electrical connections. The resistance of the electrodes is verified using a
165 circuit analyzer. Typical values range between 2-10 kΩ. Finally, the relative permittivity of
166 elastic nanocomposite is backtracked from the capacitive measurement performed with an
167 LCR meter (HP 4284A, sampling rate of 100 s−1 ) and the dimension of the capacitor. The
168 permittivity of the materials ranges between M = 6.5−7, in agreement with results obtained
169 for well dispersed particles at identical volume concentration (Stoyanov et al. 2011).

170 Experimental verification of a sensing patch

171 The strain sensing properties of a sensing patch were determined in a tensile tester
172 (Zwick Roell Z005) combined with a capacitance meter (LCR meter, HP 4284A, sampling
173 rate of 100 s−1 ). These experiments were performed using a direct capacitance measurement
174 mode, where the capacitance is measured directly for a freestanding sample, in opposition
175 to a differential measurement mode, where the differential capacitance between two sensors

7
176 is measured directly. A sample was clamped into the tensile tester on the frame pieces,
177 and metallic leads were connected to the capacitance meter using shielded cable to record
178 the change in capacitance under variation in strain. The sample was stored in an oven to
179 avoid variation in ambient conditions, and pre-strained to approximately 5% of its initial
180 length while the electrodes retained their conductivity. The capacitive measurements were
181 performed far below of the cut off frequency to avoid any influence of changes in resistance
182 or capacitance on the sensing electronics (RC behavior). Remark: local strain is transduced
183 by a variation in capacitance of the dielectric membrane, not by a variation in resistance of
184 the electrodes.
185 An experiment was also conducted in which parts of the sensing materials have been
186 removed using a regular hole punch (see Fig. 4(b)). The sample was placed on a stack of
187 paper, then punched through with the hole punch. The diameter of the hole was 14 mm.
188 Afterwards, the sample was characterized as described. Remark that the sample capaci-
189 tance was 692.7 pF before the hole was punched, and 667.8 pF after. The capacitance of
190 the removed section was 23.8 pF, for a total of 691.6 pF, which is almost identical to the
191 initial value. This is an indication that the film can survive this disruptive procedure while
192 conserving its sensing properties almost intact.
193 The sensor capacitance with respect to strain is shown in Fig 5 for a time history of
194 increasing strain targets (unfiltered data). Measurements are presented for a freestanding
195 sensor, and for the same sensor after the 14 mm hole was punched through it. Both curves
196 clearly show a capacitive response to the variation in strain. Already at about 0.00001%
197 (0.1 ppm) strain, jumps in the capacitive signal can be discerned, and above 0.2 ppm the
198 signal raises well above the noise. Here, 0.1 ppm corresponds to approximately 0.750 µm.
199 The hole punched through the sample does not affect the sensing properties, the signal
200 appears to be as responsive as before the intrusive procedure. This experiment demonstrates
201 the high level of the sensor robustness with respect to controlled mechanical tempering.
202 An obvious feature of both capacitive curves is the baseline drift, which is as large as the

8
203 signal. The authors suggest that these drifts could be caused by the impact of relaxation
204 behavior due to pre-stretching of the free-standing sensor. A successful use of these sensors
205 must therefore take these drifts into account. In what follows, the differential capacitance
206 measurement mode is used, with the benefits to cancel out measurement drifts and envi-
207 ronmental impacts (e.g. changes in temperature and humidity), similar to bridge circuits in
208 resistance-based strain gauges (Hannah and Reed 1992).

209 LABORATORY VERIFICATION OF THE SENSING METHOD

210 The sensing method is tested on wood and concrete specimens using the differential
211 capacitance measurement mode, where the change in capacitance between two adjacent
212 patches is measured directly. This way, strain (e.g. cracks) can be localized among several
213 patches.
214 The objective of the tests discussed in this section is to demonstrate that the novel flexible
215 membrane is robust, and capable of detecting and localizing damages. First, a robustness
216 test is conducted by damaging the sensor. Then, experiments employing one differential
217 patch set on wood specimens are presented. Finally, four patches are installed on a concrete
218 beam to demonstrate two independent patch sets, enabling a sensing solution capable of
219 localizing cracks.

220 Experimental Setup

221 The newly developed sensing patches are shown in Fig. 6 glued to a concrete beam. Each
222 individual patch was connected at two corners using electrical connectors, which were glued
223 using the previously described electrode solution. Each single patch covers an area of 80 ×
224 80 mm (3.15 × 3.15 in) with an active sensing area of 70 × 70 mm (2.75 × 2.75 in). The
225 wood and concrete specimens were sanded, then a primer was brushed on the surface and
226 allowed to dry. The sensor patches were glued to this area and smoothed by hand (which
227 may cause a slight prestretch). The glue was a two-component epoxy. Electrical connectivity
228 could easily be established using wires and clips.

9
229 The wood specimens were mounted on an Instron servo-hydraulic testing machine, while
230 experiments on concrete beams were performed using a Baldwin universal compression/tension
231 machine with an Admit computer control servo hydraulic system. Each differential sensor
232 pair was attached to an ACAM PSA21-CAP signal comparison amplifier, which was con-
233 nected directly to the measurement computer. The sampling rate was set to 10 s−1 . Remarks:
234 the ACAM PSA21-CAP outputs the differential capacitance between two capacitors. Thus,
235 it is not possible to obtain plots of individual sensors. Also, while the ACAM PSA21-CAP
236 is an inexpensive data acquisition system, the use of this equipment involves wiring of the
237 sensors, which decreases applicability of the sensing method. However, development of im-
238 proved electronics will result in fewer wires, and ultimately allows wireless data acquisition
239 of several patches using a single data acquisition system (Laflamme et al. 2012b).

240 Robustness Test

241 Here, the sensor robustness with respect to physical damages (e.g. caused during the
242 installation or vandalism) is verified experimentally. During the experiment, five long cuts
243 were made into one patch (a short cut followed by four cuts running over approximately
244 50% of the patch length, see Fig. 7(a)). The change in differential capacitance after each cut
245 stabilized immediately (Fig. 7(b)), demonstrating that the sensor was still operational.

246 Strain and Crack Detection on a Wood Specimens

247 A set of experiments have been conducted on wood specimens of dimension 185 × 185 ×
248 38 mm (7.25 × 7.25 × 1.5 in). Each test consisted of inducing a flexural crack in the specimen
249 using a three-point load setup, under constant displacement at a rate of 0.500 mm/min until
250 failure. For these tests, the reference sensing patch of identical geometry and comparable
251 capacitance was not glued to the beam, but placed nearby as shown in Fig. 8(d). The
252 measurement setup is shown in Fig. 8(b), showing the sensing patch glued to the bottom-
253 center of the wood specimen. Tests have been conducted on a total of five specimens.
254 Fig. 8(a) shows a typical result (sensor data are filtered). During the experiment, the sen-
255 sor was capable of detecting the crack before it caused failure of the specimen, as illustrated

10
256 by the two jumps in the differential capacitance measurements. These two jumps indicate
257 the initial deformation of the sensor resulting from the formation of the crack and secondary
258 impulsive expansion of the crack at failure. The decrease in the sensor capacitance after fail-
259 ure corresponds well with a recovery of the sensor after the specimen was unloaded. After
260 unloading, the measured differential signal change reported by the PSA21-CAP was 28.3%.
261 Remark that the direction of the capacitance signal, in this case positive with increasing
262 strain, depends on the pole connections with respect to the reference sensor. Fig. 8(c) shows
263 the cracked specimen. This crack detection test has been successfully repeated on the four
264 additional specimens.
265 The sensor capability to measure a quasi-static strain history can be determined by
266 computing the correlation capacitance-load, as the load-curvature relationship can be ap-
267 proximated as linear in the case of the 3-point load setup for small deformations. Table ??
268 lists the regression study for the load-capacitance relationship before the formation of the
269 first crack. Specimens 1-2 used thin sensing patches while specimens 3-5 used thick sensing
270 patches. All samples depicted a good fit in the linear regression (R2 ), showing that the
271 sensor accurately captured the change in curvature of the beam. Thin samples exhibited
272 a larger variance, which can be explained by the higher sensitivity associated with smaller
273 thicknesses d.

274 Strain and Crack Detection on a Concrete Beam

275 A second set of experiments have been conducted on a medium-scale concrete beam of
276 dimension 1370 × 90 × 150 mm (54 × 3.5 × 6 in). The objective was to test the capacity of
277 the nanocomposite sensor to both detect and localize cracks. Four sensing patches have been
278 installed in a matrix arrangement to form a sensing skin, using a vertical pattern. The sensing
279 patches were organized in pairs in order to measure changes in differential capacitance, one
280 pair at the extremities (sensor 1), and one pair in the middle (sensor 2). Fig. 6 illustrates
281 the monitored surface of the bottom flange of the beam, along with the location of the
282 comparative sensing patches. The sensor patches were applied off-center to avoid regions

11
283 of constant differential curvature. The localization of two comparative patches with respect
284 to each other in the differential capacitance measurement mode does not affect the sensing
285 accuracy. For enhanced precision on strain localization, the differential capacitance can
286 be measured between all sensors. Such additional precision was out of the scope of this
287 experimental campaign.
288 Flexural cracks were induced using a four-point load setup, with a constant load rate of
289 6000 N/min, until failure of the beam. Remark: the failure mode was in shear, thus not
290 reflected in the measurements. This is a consequence of the underlying pedagogic purpose of
291 the beam failure experiment, which was performed in the context of an undergraduate class.
292 Fig. 9 (a) plots the differential capacitance measurements for both sets of sensors against
293 time, and Fig. 9 (b) shows the cracked specimen at the end of the test, where the sensing
294 patches have been removed while the beam was loaded to verify the final locations of the
295 cracks. A first substantial jump in the measurements is observed with sensor 1, at the
296 beginning of the experiment (less than 100 sec). That jump indicates the formation of a
297 crack (crack 1) at the extremity of the sensor patches. The crack was visually observable at
298 that time. A second jump in the measurement curve can also be seen, which coincides with
299 the formation of two additional cracks (cracks 2a and 2b) which could have been formed
300 simultaneously, along with a symmetrical crack (crack 2c) monitored by the second set of
301 sensors. The small jumps in sensor 1 measurements and the slope in sensor 2 measurements
302 probably result from a combination of the slow expansion of the cracks and the increasing
303 beam curvature. At the end of the sensor 2 curve (after 900 sec), there is a significant jump
304 in the data, which coincided with the formation of a third crack (crack 3) that was visually
305 verifiable.
306 Additional features in Fig. 9(a) data include a jump in sensor 1 measurements after 1000
307 sec, a fluctuation in sensor 2 measurements after 900 sec (at crack 3), and a higher level in
308 the change in capacitance compared with results obtained in Fig. 8 (experiments on wood
309 specimens). The upward jump in the sensor 1 is a result of an inversion of the electrode poles

12
310 for ’sensor 1 left’. This inversion allowed an additional level of damage localization with two
311 comparative patches, provided that the direction of strain was known to be increasing and
312 constant for both patches. Thus, the crack formed under the patch ’sensor 1 left’ (crack 4)
313 was measured as a positive change in capacitance. The fluctuation in the measurements at
314 crack 3 could be explained by the patch oscillating between two states of strain due to a local
315 plastic deformation of the material. This fluctuation tends to converge to the higher strain
316 state. Finally, the higher level of differential capacitance compared with the experiments
317 on wood specimens are partly attributed to the use of thicker patches (approximately 10%
318 thicker).
319 The good correspondence between visually observable results and the sensor pair curves
320 demonstrated that the sensing skin was capable of both detecting and localizing cracks in
321 the concrete beam.

322 CONCLUSION
323 In this paper, the concept of the sensing skin was substantially improved by developing a
324 capacitive sensing patch with tailored mechanical and electrical properties, de facto demon-
325 strating the generality of the sensing method. The sensing patch consists of a thermoplastic
326 elastomer doped with titanium dioxide and sandwiched by stretchable CB composite elec-
327 trodes. The proposed sensing method comprises several practical advantages, making the
328 technology directly applicable for health monitoring of civil infrastructure:

329 • Inexpensive material: SEBS-based materials are inexpensive and easy to process.
330 • Can cover large areas: the flexible nature of the membrane allows full-scale imple-
331 mentation, a powerful advantage compared against conventional mechanical strain
332 gauges.
333 • Low voltage (1-2.5 volts) requirement: capacitance-based measurements only require
334 a fraction of the electrical energy compared to resistance-based strain sensors.
335 • Easy to install: the sensor can be applied easily on irregular surfaces using off-the-shelf

13
336 epoxy and necessitates minor surface preparation.
337 • Customizable surface monitoring: shapes and sizes of the electrodes sandwiching the
338 polymer can be user-defined.
339 • Robust with respect to physical damages: damaged sensing patches can still be used
340 for measuring strain.
341 • Damage localization: individual sensor pairs can localize damage, and the polarity of
342 a signal change may give further precision within a sensor pair.

343 Tests on wood have shown that the new sensor is robust towards mechanical tampering
344 or accidents, and is capable of detecting cracks. Test on a medium-size concrete specimen
345 demonstrated the promising capability of the sensing skin to both detect and locate cracks,
346 an advantage over most existing SHM technologies. Potential impacts of the sensing skin
347 include:

348 1. Enhanced life-span of civil infrastructure via timely maintenance and visual inspection
349 programs, improving sustainability of civil infrastructure.
350 2. Cost savings arising from timely repairs and reduced number of visual inspections.
351 3. Improved safety by diagnosing and localizing problems before they endanger struc-
352 tural integrity.

353 ACKNOWLEDGMENTS
354 The authors thank John T. Germaine for his priceless help and guidance during the
355 testing process, as well as Andreas Pucher and Stephen W. Rudolph for their valuable
356 support with the measurement setups. The authors also acknowledge support from the
357 WING initiative of the German Ministry of Education and Research through its NanoFutur
358 program (Grant No. NMP/03X5511).

359 REFERENCES

360 Brownjohn, J. (2007). “Structural health monitoring of civil infrastructure.” Philosophical

14
361 Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences,
362 365(1851), 589–622.
363 Carlson, J., English, J., and Coe, D. (2006). “A flexible, self-healing sensor skin.” Smart
364 materials and structures, 15, N129.
365 Chen, G., Sun, S., Pommerenke, D., Drewniak, J., Greene, G., McDaniel, R., Belarbi, A.,
366 and Mu, H. (2005). “Crack detection of a full-scale reinforced concrete girder with a
367 distributed cable sensor.” Smart Materials and Structures, 14(3), S88.
368 Gao, L., Thostenson, E., Zhang, Z., Byun, J., and Chou, T. (2010). “Damage monitoring
369 in fiber-reinforced composites under fatigue loading using carbon nanotube networks.”
370 Philosophical Magazine, 90(31-32), 4085–4099.
371 Giurgiutiu, V. (2009). Encyclopedia of structural health monitoring. Wiley, Chapter Piezo-
372 electricity Principles and Materials, 981–991.
373 Hannah, R. and Reed, S. (1992). Strain Gage Users’ Handbook. Springer.
374 Harms, T., Sedigh, S., and Bastianini, F. (2010). “Structural Health Monitoring of Bridges
375 Using Wireless Sensor Networks.” Instrumentation & Measurement Magazine, IEEE,
376 13(6), 14–18.
377 Harrey, P., Ramsey, B., Evans, P., and Harrison, D. (2002). “Capacitive-type humidity
378 sensors fabricated using the offset lithographic printing process.” Sensors and Actuators
379 B: Chemical, 87(2), 226–232.
380 Hurlebaus, S. and Gaul, L. (2004). “Smart layer for damage diagnostics.” Journal of intel-
381 ligent material systems and structures, 15(9-10), 729–736.
382 Karbhari, V. M. (2009). Encyclopedia of structural health monitoring. Wiley, Chapter Design
383 Principles for Civil Structures, 1467–1476.
384 Kollosche, M., Melzer, M., Becker, A., Stoyanov, H., McCarthy, D. N., Ragusch, H., and
385 Kofod, G. (2009). “The influence of mechanical properties in the electrical breakdown in
386 poly-styrene-ethylene-butadiene-styrene thermoplastic elastomer.” Proc. SPIE, Vol. 7287,
387 728729.

15
388 Kollosche, M., Stoyanov, H., Laflamme, S., and Kofod, G. (2011). “Strongly enhanced sensi-
389 tivity in elastic capacitive strain sensors.” Journal of Materials Chemistry, 21, 8292–8294.
390 Laflamme, S., Kollosche, M., Connor, J. J., and Kofod, G. (2010). “Large-scale capacitance
391 sensor for health monitoring of civil structures.” Proc. of 5WCSCM.
392 Laflamme, S., Kollosche, M., Connor, J. J., and Kofod, G. (2012a). “Soft capacitive sensor
393 for structural health monitoring of large-scale systems.” Structural Control and Health
394 Monitoring, 19(1), 70–81.
395 Laflamme, S., Saleem, H., Vasan, B., Geiger, R., Chaudhary, S., Kollosche, M., Kofod, G.,
396 and Rajan, K. (2012b). “Novel Flexible Membrane for Structural Health Monitoring of
397 Wind Turbine Blades - Feasibility Study.” Proceedings of Materials Challenges in Alter-
398 native & Renewable Energy, Clearwater, FL.
399 Laurer, J. H., Bukovnik, R., and Spontak, R. J. (1996). “Morphological characteristics of
400 sebs thermoplastic elastomer gels.” Macromolecules, 29(17), 5760–5762.
401 Lin, B., Giurgiutiu, V., Bhalla, A., Chen, C., Guo, R., and Jiang, J. (2009). “Thin-film active
402 nano-PWAS for structural health monitoring.” Proc. of SPIE, Vol. 7292, 72921M–1.
403 Lin, B., Giurgiutiu, V., Yuan, Z., Liu, J., Chen, C., Jiang, J., Bhalla, A., and Guo, R.
404 (2007). “Ferroelectric thin-film active sensors for structural health monitoring.” Proc. of
405 SPIE, Vol. 6529, 65290I.
406 Lipomi, D., Vosgueritchian, M., Tee, B., Hellstrom, S., Lee, J., Fox, C., and Bao, Z. (2011).
407 “Skin-like pressure and strain sensors based on transparent elastic films of carbon nan-
408 otubes.” Nature Nanotechnology, 6(12), 788–792.
409 Loh, K., Hou, T., Lynch, J., and Kotov, N. (2009). “Carbon nanotube sensing skins for
410 spatial strain and impact damage identification.” Journal of Nondestructive Evaluation,
411 28(1), 9–25.
412 Lumelsky, V., Shur, M., and Wagner, S. (2001). “Sensitive skin.” Sensors Journal, IEEE,
413 1(1), 41–51.
414 Mason, R., Gintert, L., Hock, V., Sweeney, S., Lampo, R., and Chandler, K. (2009). “A novel

16
415 integrated monitoring system for structural health management of military infrastructure.”
416 DoD Corrosion Conference.
417 McCarthy, D. N., Risse, S., Katekomol, P., and Kofod, G. (2009). “The effect of dispersion
418 on the increased relative permittivity of tio 2 / sebs composites.” Journal of Physics D:
419 Applied Physics, 42(14), 145406.
420 Simon, J., Bracci, J., Gardoni, P., et al. (2010). “Seismic Response and Fragility of Deterio-
421 rated Reinforced Concrete Bridges.” Journal of Structural Engineering, 136, 1273–1281.
422 Stoyanov, H., Kollosche, M., McCarthy, D. N., and Kofod, G. (2010). “Molecular composites
423 with enhanced energy density for electroactive polymers.” Journal of Material Chemistry,
424 20, 7558–7564.
425 Stoyanov, H., Kollosche, M., Risse, S., McCarthy, D. N., and Kofod, G. (2011). “Elastic block
426 copolymer nanocomposites with controlled interfacial interactions for artificial muscles
427 with direct voltage control.” Soft Matter, 7, 194–202.
428 Tata, U., Deshmukh, S., Chiao, J., Carter, R., and Huang, H. (2009). “Bio-inspired sensor
429 skins for structural health monitoring.” Smart Materials and Structures, 18, 104026.
430 Yu, L., Cheng, L., and Su, Z. (2011). “Correlative sensor array and its applications to iden-
431 tification of damage in plate-like structures.” Structural Control and Health Monitoring.
432 Yu, L. and Giurgiutiu, V. (2009). Encyclopedia of structural health monitoring. Wiley, Chap-
433 ter Piezoelectric Wafer Active Sensors, 1013–1027.
434 Zagrai, A., Doyle, D., Gigineishvili, V., Brown, J., Gardenier, H., and Arritt, B. (2010).
435 “Piezoelectric wafer active sensor structural health monitoring of space structures.” Jour-
436 nal of Intelligent Material Systems and Structures, 21(9), 921.
437 Zhang, Y. (2005). “Piezoelectric paint sensor for real-time structural health monitoring.”
438 Proc. of SPIE, Vol. 5765, 1095.
439 Zhang, Y., Lloyd, G., and Wang, M. (2003). “Random Vibration Response Testing of PVDF
440 Gages for Long-span Bridge Monitoring.” Proc. 4th Int. Workshop on Structural Health
441 Monitoring, Stanford, 115.

17
442 Zou, L., Bao, X., Ravet, F., Chen, L., Zhou, J., and Zimmerman, T. (2006). “Prediction
443 of pipeline buckling using distributed fiber Brillouin strain sensor.” Proc. 2nd Int. Conf.
444 Structural Health Monitoring of Intelligent Infrastructure, Shenzhen, 393.

18
445 List of Tables

19
446 List of Figures
447 1 Sketch of a capacitive unit, showing the sandwich structure of compliant elec-
448 trodes surrounding an electrically insulating elastomer layer, deployed over a
449 monitored surface using a bounding agent. Layer thicknesses are not scaled. 21
450 2 SEM picture of a cryo-fractured surface of the polymer nanocomposite material 22
451 3 (a) Schematic of the plastic mask used to define the electrode area, (b) Soft
452 stretchable capacitor with conductive electrodes . . . . . . . . . . . . . . . . 23
453 4 (a) The capacitive sensor with an electrode area of 70 mm × 70 mm and
454 thickness of 386 µm (b) The sensor after a hole of diameter 14 mm was
455 punched through it. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
456 5 Sensitivity of a sensing patch with respect to strain (regular sensor above,
457 sensor with hole punched below). The bottom graph shows the strain history. 25
458 6 (a) Four sensing patches deployed on the bottom flange of the concrete beam,
459 forming a sensing skin. (b) Locations of sensing patches on the concrete beam
460 used for the experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
461 7 (a) Sensor after five cuts. (b) Relative capacitance against time. . . . . . . . 27
462 8 (a) Relative capacitance and applied load against time - specimen 4. (b)
463 Typical three-point load setup for the wood specimens. (c) Cracked specimen
464 (bottom flange view). (d) Locations of sensing patches. . . . . . . . . . . . . 28
465 9 (a) Differential capacitance measurements for both set of sensors against time.
466 (b) Cracked specimen (bottom flange view) . . . . . . . . . . . . . . . . . . . 29

20
 stretchable electrodes

A+DA
polymer layer
bonding agent
crack monitored structure

FIG. 1. Sketch of a capacitive unit, showing the sandwich structure of compliant elec-
trodes surrounding an electrically insulating elastomer layer, deployed over a monitored
surface using a bounding agent. Layer thicknesses are not scaled.

21
 1 µm Mag = 10 KX

FIG. 2. SEM picture of a cryo-fractured surface of the polymer nanocomposite material

22
 mask
~70mm
nanocomposite
glass substrate 80mm
80mm

(a) (b)

FIG. 3. (a) Schematic of the plastic mask used to define the electrode area, (b) Soft
stretchable capacitor with conductive electrodes

23
 C0=692.7 pF C=667.8 pF

Ccut=23.8 pF

14 mm

(a) (b)

FIG. 4. (a) The capacitive sensor with an electrode area of 70 mm × 70 mm and

thickness of 386 µm (b) The sensor after a hole of diameter 14 mm was punched
through it.

24
 762.0

761.8

761.6
Capacitance [pF]

SEBS 500120
761.4 15 vol% TiO2

734.4

734.2

734.0
SEBS 500120
733.8 „WITH HOLE“
15 vol% TiO2

0.00012
Strain [%]

0.00008
0.00004
0
0 1000 2000 3000 4000 5000
Time [s]

FIG. 5. Sensitivity of a sensing patch with respect to strain (regular sensor above,
sensor with hole punched below). The bottom graph shows the strain history.

25
 point beam point
beam centerline load centerline load

sensor 1 sensor 2 sensor 2 sensor 1 90 mm

left left right right (a)

sensor 1 sensor 2 sensor 2 sensor 1

left left right right 4 patches @ 80mm = 320 mm

(a) (b)

FIG. 6. (a) Four sensing patches deployed on the bottom flange of the concrete beam,
forming a sensing skin. (b) Locations of sensing patches on the concrete beam used
for the experiment.

26
 60
differential capacitance [ppm]
50
1. Cut. 4. 40 5.
3.
30 4.
3.
5. 20
2. 2.
10 1. Cut

0
-10
0 10 20 30 40 50 60 70
(a) (b) Time [s]

FIG. 7. (a) Sensor after five cuts. (b) Relative capacitance against time.

27
 80 3.5
Unloading
differential capacitance [ppm]

70 Load
3.0
Sensor
60 Crack-
expansion 2.5
(b) (c) crack
50
2.0 Load [kN]
185mm
40
Crack 1.5
30 Sensor
1.0 Ref.
20
185mm

0.5 Sensor
10
0 0
0 100 200 300 400 500
Time [s] beam centerline &
(a) (d) load point

FIG. 8. (a) Relative capacitance and applied load against time - specimen 4. (b)
Typical three-point load setup for the wood specimens. (c) Cracked specimen (bottom
flange view). (d) Locations of sensing patches.

28
 0
relative change in capacitance [ppm]

Sensor 1
-50 Sensor 2
Crack 2c sensor 1
-100 beam centerline
left sensor 2
-150 left sensor 2
Crack 1 right sensor 1
-200 Crack 3 right
crack 4
-250 crack 3
crack 2c
-300 Crack 2a & crack 1
2b
-350 crack 2a&2b
Crack 4
-400
0 200 400 600 800 1000 1200
Time [s]
(a) (b)

FIG. 9. (a) Differential capacitance measurements for both set of sensors against time.
(b) Cracked specimen (bottom flange view)

29