Civil, Construction and Environmental EngineeringPublications Civil, Construction and Environmental Engineering

12-2013

Soft Elastomeric Capacitor Network for StrainSensing Over Large SurfacesSimon LaflammeIowa State University, laflamme@iastate.edu

Husaam S. SaleemIowa State University, hsaleem@iastate.edu

Bharath K. VasanIowa State University, bharathk@iastate.edu

See next page for additional authors

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

Part of the Electrical and Computer Engineering Commons, and the Structural EngineeringCommons

The complete bibliographic information for this item can be found at http://lib.dr.iastate.edu/ccee_pubs/16. 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 DigitalRepository. It has been accepted for inclusion in Civil, Construction and Environmental Engineering Publications by an authorized administrator ofIowa State University Digital Repository. For more information, please contact digirep@iastate.edu.

Soft Elastomeric Capacitor Network for Strain Sensing Over LargeSurfaces

AbstractField applications of existing sensing solutions to structural health monitoring (SHM) of civil structures arelimited. This is due to economical and/or technical challenges in deploying existing sensing solutions tomonitor geometrically large systems. To realize the full potential of SHM solutions, it is imperative to developscalable cost-effective sensing strategies. We present a novel sensor network specifically designed for strainsensing over large surfaces. The network consists of soft elastomeric capacitors (SECs) deployed in an arrayform. Each SEC acts as a surface strain gage transducing local strain into changes in capacitance. Results showthat the sensor network can track strain history above levels of 25 με using an inexpensive off-the-shelf dataacquisition system. Tests at large strains show that the sensor's sensitivity is almost linear over strain levels of0-20%. We demonstrate that it is possible to reconstruct deflection shapes for a simply supported beamsubjected to quasi-static loads, with accuracy comparable to resistive strain gages.

KeywordsElectrical and Computer Engineering

DisciplinesElectrical and Computer Engineering | Structural Engineering

CommentsThis is an author's manuscript of an article from IEEE/ASME Transactions on Mechatronics 18 (2013):1647–1654, doi.10.1109/TMECH.2013.2283365. Posted with permission.

AuthorsSimon Laflamme, Husaam S. Saleem, Bharath K. Vasan, Randall L. Geiger, Degang J. Chen, Michael R.Kessler, and Krishna Rajan

This article is available at Iowa State University Digital Repository: http://lib.dr.iastate.edu/ccee_pubs/16

IEEE/ASME TRANS. ON MECHATRONICS 1

Soft Elastomeric Capacitor Network for StrainSensing over Large Surfaces

Simon Laflamme, Hussam S. Saleem, Bharath K. Vasan, Randall L. Geiger, Degang Chen,

Michael R. Kessler, Krishna Rajan

Abstract—Field applications of existing sensing solutionsto structural health monitoring (SHM) of civil structuresare limited. This is due to economical and/or technicalchallenges in deploying existing sensing solutions to moni-tor geometrically large systems. To realize the full poten-tial of SHM solutions, it is imperative to develop scalablecost-effective sensing strategies. We present a novel sensornetwork specifically designed for strain sensing over largesurfaces. The network consists of soft elastomeric capaci-tors (SECs) deployed in an array form. Each SEC acts asa surface strain gage transducing local strain into changesin capacitance. Results show that the sensor network cantrack strain history above levels of 25µε using an inexpensiveoff-the-shelf data acquisition systems. Tests at large strainsshow that the sensor’s sensitivity is almost linear over strainlevels of 0-20%. We demonstrate that it is possible to re-construct deflection shapes for a simply supported beamsubjected to quasi-static loads, with accuracy comparableto resistive strain gages.

Index Terms—Sensor network, condition assessment,prognosis, soft elastomeric capacitor, structural health mon-itoring, distributed strain gages.

I. Introduction

STRUCTURAL health monitoring (SHM) can be de-fined as the automation of damage assessment. When

applied to geometrically large systems, such as bridges,aircraft, and wind turbines, this task becomes com-plex because of the existing economical and/or technicalchallenges in deploying existing sensing solution at themesoscale. For instance, static-based methods, which in-cludes fiber-optics [1], [2], [3], typically lead to direct sig-nal processing, but the technologies can be expensive todeploy over large surfaces. On the other hand, dynamic-based methods, which include accelerometers [4], [5], [6],are typically inexpensive to deploy, but they often overlookthe complexity of damage diagnosis and localization [7].

We present a novel sensor network designed for cost-effective SHM of large surfaces, applicable to civil struc-tures and other geometrically large systems. The sen-sor in this network consists of a distributed array of softelastomeric capacitors (SECs) constituting surface strain

S. Laflamme is with the Department of Civil, Construction, andEnvironmental Engineering, and the Department of Electrical andComputer Engineering, Iowa State University, Ames, IA 50011 (cor-responding author: S. Laflamme, e-mail: laflamme@iastate.edu),

H. Saleem is with the Department of Civil, Construction, andEnvironmental Engineering, Iowa State University, Ames, IA 50011,

B. Vasan, R. Geiger and D. Chen are with the Department ofElectrical and Computer Engineering, Iowa State University, Ames,IA 50011,

M. Kessler is with the School of Mechanical and Materials Engi-neering at Washington State University, Pullman, WA 99164,

K. Rajan is with the Department of Materials Sciences and Engi-neering, Iowa State University, Ames, IA 50011.

gages. Each SEC is fabricated from of a thermoplasticelastomer matrix of poly-styrene-co-ethylene-co-butylene-co-styrene (SEBS) doped with titanium dioxide (TiO2),sandwiched between electrodes composed of SEBS mixedwith carbon black (CB).

Soft elastomeric sensors have previously been proposedfor SHM of civil structures [8], [9], [10], [11]. Popularapplications include carbon nanotubes within the elas-tomeric substrate to create resistive strain sensors [12],[13]. Capacitance-based film-type sensors have also beenproposed, which include applications to humidity [14], [15],pressure [16], strain [17], [18], and tri-axial force [19] mea-surements. The proposed SEC differs from literature bycombining both a large physical size and high initial ca-pacitance, resulting in a larger surface coverage and highersensitivity. The SEC constitutes a promising candidate forstrain sensing over large surfaces.

Compared with other existing sensing solutions, the pro-posed sensor is an alternative to fiber optics. With bothfiber optic sensors and the SEC technology, strain data canbe measured over large systems. The SEC network offersthe advantages of being 1) cost-effective; 2) operable atlow frequencies; 3) mechanically robust; 4) low-powered;5) easy to install onto surfaces; and 6) customizable inshapes and sizes. The proof-of-concept of the SEC tech-nology has been demonstrated by the authors with an off-the-shelf flexible capacitor [20], and with the nanoparticlemix described above [21].

In this paper, we characterize the sensor performance forstrain sensing as a single sensor and in a network config-uration. This characterization is conducted by comparingthe performance of the SEC against off-the-shelf resistivestrain gages (RSGs). We also demonstrate the applicationof a SEC network for reconstructing deflection shapes of asimply supported beam.

The paper is organized as follows. Section II describesthe sensor fabrication, sensing principle, transducer model,and shape reconstruction algorithm. Section III verifies asingle SEC at tracking time history of strain, and validatesthe performance of four SECs organized in a network tocapture deflection shapes. Section IV discusses the resultsand concludes the paper.

II. Background

A. Sensor Fabrication

The sensing hardware developed for strain sensing overlarge surfaces is an array of SECs acting as large-scale sur-face strain gages. Each SEC is composed of a nanocom-posite mix of SEBS (Dryflex 500120) doped with rutile

2 IEEE/ASME TRANS. ON MECHATRONICS

TiO2 (Sachtleben R 320 D) serving as a dielectric for thecapacitor. The TiO2 is selected to increase the materi-als permittivity and robustness with respect to mechanicaltempering. This dielectric is sandwiched between compli-ant electrodes fabricated with the same SEBS mixed withCB particles (Printex XE 2-B) [21]. Fig. 1 shows the pic-ture of a single SEC. Fig. 2 illustrates the fabrication pro-cess of a SEC. The process is initiated by the fabricationof a SEBS/toluene solution. Part of this solution is usedto create the nanoparticle mix, in which TiO2 particles areadded and dispersed using an ultrasonic tip. The resultingmix is drop-casted on a glass slide, and dried over 5 daysto allow complete evaporation of the solvent. Meanwhile,the remaining SEBS/toluene solution is used to create thecompliant electrodes. Here, CB particles are added insteadof TiO2 to create a conductive mix. Finally, the CB mix issprayed or painted on both surfaces of the dried polymer.

Fig. 1. A single SEC (70 × 70 mm2)

Fig. 2. Sensor fabrication

B. Sensing Principle

The capacitance C of a SEC is written

C = e0erA

h(1)

where e0 = 8.854 pF/m is the vacuum permittivity, er thedimensionless polymer relative permittivity, A = w · l thesensor area with width w and length l, and h the height of

the dielectric. Given a unidirectional strain in the length lof the sensor (∆w = 0, due to the epoxy), a small changein C can be obtained from Eq. (1) by expressing the dif-ferential ∆C as

∆C =

(∆l

l− ∆h

h

)C (2)

The poisson ratio of pure SEBS materials has been re-ported to be 0.49 [22]. The elastomer can be assumed tobe incompressible, where the nominal volume V = w · l · hwill be preserved after the SEC geometric deformations ∆land ∆h:

w · l ·h= (l+ ∆l)(h+ ∆h)w

∆l

l≈−∆h

h

(3)

In now follows from Eqs. (1) to (3) that

∆C

∆l= 2

e0erw

hor

∆C

εs= 2C (4)

where εs is the sensor strain. Eq. (4) represents the sen-sitivity of the sensor. This sensitivity can be increased bydecreasing the SEC thickness, increasing the width, or in-creasing the dielectric permittivity, which is attained byaltering the nanocomposite mix [23]. It is therefore possi-ble to customize the sensitivity for a given geometry. Forthe SEC shown in Fig. 1 (C ≈ 595 pf, w = l = 70 mm,h= 0.3 mm), the resulting sensitivity is

∆C

∆l= 17.0 pF/mm or

∆C

εs= 1190 pF/εm (5)

Note that the sensitivity of each SEC may vary by ±20%due to the manual fabrication process.

The sensing principle for the sensor consists of directlymeasuring the relative changes in capacitance of a SECover time. Fig. 3 illustrates the measurement principle.The sensor is adhered onto the monitored surface usingan epoxy. A strain on the monitored surface provokes achange in the sensor geometry ∆l, which is measured bythe data acquisition system (DAQ) as a change in the ca-pacitance ∆C, or the relative change ∆C/C, where C isthe nominal capacitance.

Fig. 3. Measurement Principle (layers not scaled)

: TITLE 3

C. SEC Signal-Strain Model

Several models have been studied to describe the stress-strain relationship for thermoplastic elastomers, which isdominated by a nonlinear rate-dependant response. Thisnonlinear response is further complicated by additionalnonlinearities from the hysteresis and Mullins’ effect [24].A typical constitutive model used to represent the mechan-ical behavior is a three-parameters rheological model, inwhich an elastic spring and a visco-plastic dashpot in seriesare installed in parallel with a hyper-elastic rubbery spring[25]. The three-parameter model has been used in civil en-gineering, for system identification of rubber bearings usedfor base-isolation [26], [27]. In the proposed application, weassume that the materials on which the polymer is bondedsignificantly stiffer than the SEC. For a two-dimensionalbending beam subjected to low frequency excitations, onecan write:

εs = εm = −c δ2y

δx2(6)

where εm is the strain of the monitored beam surface, c isthe distance from the surface to the centroid of the beam,y is the deflection (downwards in Fig. 4(c)), and x is thelongitudinal Cartesian coordinate (leftwards in Fig. 4(c)).Using Eqs. (4) and (6), one obtains a gage factor indepen-dent of the nanoparticle mix: :

∆CC

εm= 2 (7)

D. Shape Reconstruction

The problem of real-time reconstruction of deflectionshapes from position and curvature measurements fromsensor networks has been widely studied, with applica-tions to condition assessment, SHM, and shape control[28], [29], [30], [31]. Here, we select the polynomial inter-polation method for reconstructing the deflection shapesfrom a network of curvature data, a technique also usedto smoothen data. The algorithm consists of fitting thecurvature data using a polynomial function. In the case ofa two-dimensional beam equipped with four sensors, thefitting function is taken as a third degree polynomial toavoid possible over-fitting [28], [31].

ε̂m,j = a0 + a1xj + a2x2j + a3x

3j (8)

where the hat denotes an estimation for the jth sensor.Minimizing the error J for n sensors:

J =

n∑j

(εm,j − ε̂m,j)2 (9)

leads to the expression:

A = (XTX)−1XT Ξm (10)

with:

A =

a0

a1

a2

a3

Ξm =

εm,1

εm,2

· · ·εm,n

X =

1 x1 x2

1 x31

1 x2 x22 x3

2...

......

...1 xn x2

n x3n

(11)

Once the parameters A are determined, the deflectionshape is obtained by integrating the curvature twice:

y(x) =

∫ L

0

∫ L

0

δ2y

δx2dx2

=

∫ L

0

∫ L

0

−1

c(a0 + a1xj + a2x

2j + a3x

3j )dx2

=−1

c

(a0x2

2+ a1

x3

6+ a2

x4

12+ a3

x5

20

)+ b1x+ b2

(12)

where L is the length of the beam. Enforcing the boundaryconditions y(0) = y(L) = 0 for a simply-supported beam,one obtains:

b1 =1

c

(a0L

2+ a1

L2

6+ a2

L3

12+ a3

L4

20

)b2 = 0

(13)

III. Laboratory Verifications

A. Experimental Setup

In this section, the proposed sensor network is validatedusing 1) a single SEC; and 2) four SECs organized in anetwork, both measuring surface strain of a bending beam.An additional test is conducted on a free-standing sensorto study its behavior under large levels of strain (0-20%).We use the same SEC sensor size as shown in Fig. 1 tocharacterize the performance of a full-scale sensor. Bend-ing beam tests are conducted in a three-point load setup ona simply supported aluminum beam of support-to-supportdimensions 406.4× 101.6× 6.35 mm3 (16× 4× 0.25 in3).A typical experimental setup is shown in Fig. 4. SECsand RSGs are installed following a similar procedure. Themonitored surface is sanded, painted with a primer, anda thin layer of an off-the-shelf epoxy (JB Kwik) is appliedon which the sensors are adhered.

In the single SEC bending tests, the SEC and strain gageare located under the beam at x = 0.50L. For the fourSECs tests, the SECs and strain gages are located underthe beam at x= {0.20,0.40,0.60,0.80}L, as shown in Figs.4(b) and 4(c). The load is applied using a hand operatedhydraulic test system (Enerpac) for the static tests, and aservo-hydraulic fatigue testing machine (MTS) for quasi-static tests. All tests are repeated three times. In the free-standing tests, the sensor is pre-stretched at approximately

4 IEEE/ASME TRANS. ON MECHATRONICS

Fig. 4. Laboratory setup for four SECs. (a) General setup withfour SECs installed on the bottom of the beam; (b) setup schematic- from side; and (c) setup schematic - from under the beam.

1.5% strain and subjected to a uniaxial tensile strain usingan Instron universal testing machine (model 5569).

Data from the SECs are acquired using an inexpen-sive off-the-shelf data acquisition system (ACAM PCap01)sampled at 95.4 Hz for the single SEC setup (includingthe free-standing setup), and 48.0 Hz for the four SECssetup. The SEC readings are compared against RSG withresolution of 1 µε (Vishay Micro-Measurements, CEA-06-500UW-120). Strain gage data are acquired using aHewlett-Packard 3852 data acquisition system, and datasampled at 1.7 Hz when using the hand hydraulic, and 55Hz when using the MTS.

Data are filtered using a low-pass filter, and zeroed usingthe average capacitance while the beam is unloaded.

B. Validation of the SEC

A single SEC is first validated using a step load. Fig.5 is a plot of the results, in which the capacitance sig-nal has been transformed into strain using Eq. (7). TheSEC signal raises above noise beyond 25 µε, but for strainlevels above 55 µε, there is a constant difference of ap-proximately 8 µε. Fig. 6 shows the results from a typicalquasi-static load test. The excitation history consists ofa displacement-based triangular wave loads with increas-ing frequencies from 0.0167 to 0.40. Results show thatthe SEC is capable of tracking a quasi-static strain historywithin a given level of resolution. Fig. 7 is a plot of theSEC readings in function of the strain measured by theRSG, validating the linearity of the sensor. Its measuredsensitivity is in agreement with Eq. (5).

Given that the dielectric permittivity does not changesignificantly in the low frequency range (<100 Hz) [32], alarge portion of the measurement errors in both Figs. 5and 6 can be attributed to the electronics. Firstly, there isparasitic capacitance in the cables connecting to the sen-sors, which cause variations in the measured capacitance.Because SECs require very small power, this noise can beminimized by digitizing the signal at the source, enablinglong distance transmissions, either over a wired or wire-less link, with essentially no signal degradation. Secondly,the DAQ itself may not have the sufficient resolution formeasuring small changes in capacitance. For instance, us-ing Eq. (5), the measurement of 1 µε over the SEC cor-responds to measuring a change of 0.00140 pF. Thirdly,

Fig. 5. Strain history of SEC versus RSG (step load).

Fig. 6. Strain history of SEC versus RSG (triangular load) andactuator displacement.

there is a small linear drift in the signal. This drift canbe seen in both Figs. 5 and 6, where the SEC signal doesnot return to the zero strain line. While this drift couldbe filtered out, future developments in a dedicated DAQsystems for SEC measurements may minimize this noise.Other sources of error may come from imperfections in thesensor fabrication and geometry, and/or a slight angle inthe application of the sensor which would change the geo-metric relationship described in Eq. (2).

A net advantage of the SEC is its high elasticity com-pared to conventional strain transducers, enabling mea-surements of large strain levels. To study the behavior ofthe SEC at large strains, the sensor is subjected to a trian-gular strain ramping from 0 to 20% in approximately 2%increments. Fig. 8 shows a plot of the measured capac-itance versus applied strain after pre-stretch, along withthe applied strain history. Results show that the sensorexhibits a slight nonlinearity over the range 0-20%. Notethat, in a free-standing setup, the sensitivity cannot be ob-tained using Eq. (5), because εy 6= 0. Also, the differentialform of Eq. (1) given by Eq. (2) does not apply for large

: TITLE 5

Fig. 7. Sensitivity of the SEC.

strain, which explains the loss in linearity compared toFig. 7. Nevertheless, it is possible to model the nonlinearsensitivity of the sensor for large strain measurements.

Fig. 8. Sensitivity of the SEC under large strain levels.

C. Shape detection using Sensor Network

This subsection demonstrates a SHM application using anetwork of SECs. Strain measurements from four SECs areused to reconstruct the deflection shapes using the repre-sentation from Eq. (12). Note that the double integrationof strain data to reconstruct the deflection shape may causean accumulation of errors. This effect is minimized by theutilization of the polynomial fit that inherently filters thestrain data, and by the enforcement of the boundary con-ditions (y(0) = y(L) = 0). In implementations, it wouldbe also possible to further reduce the error, for instance,by averaging the resulting shapes within small periods oftime, depending on the sampling rate.

Fig. 9 shows the strain time history of the four SECssubjected to a static load. The comparison of results be-tween Figs. 9(a) and 9(b) show that the results from

the SECs agree with the RSGs at low levels of strain(<100 µε), but the discrepancy increases with the level ofstrain, reaching up to 100 µε. Note that the small differ-ence between symmetric strain gages (RSG1 versus RSG4

and RSG2 versus RSG3) can be explained by a slight off-centered installation of sensors and/or application of theload.

Fig. 9. Strain history for four SECs. (a) SECs; and (b) RSGs.

Using the surface strain data for Fig. 9, the deflectionshapes are estimated and shown in Fig. 10 for a typicalresult (at time t = 25 s). The shape obtained from theSECs is compared against the shape obtained from theRSGs using the same procedure as described in SubsectionII.D. Results are compared against the analytical solutionusing Euler-Bernoulli beam theory:

y(x) = − Px

48EI(3L2 − 4x2) for 0 ≤ x ≤ L

2(14)

where y(x) is symmetry about L/2, P = 120 N is the ap-plied load at time t = 25 s, and E and I are the Young’smodulus and moment of inertia of the beam, respectively.Fig. 10(a) compares the non-normalized results. The un-derestimation of strain for the SECs clearly results in anunderestimation of the magnitude of the shape. In ap-plications to SHM, the normalized deflection shape mightbe more informative to detect damages or changes in thestructural behaviors. Fig. 10(b) shows the shapes nor-malized to a maximum deflection of -1. Results from theSECs compare well against RSGs and the analytical so-lution. The average root-mean-square (RMS) error of thenormalized deflection shapes for the duration of the loadP = 118 N (for 20.6 s ≤ t ≤ 28.8 s) is 0.208 mm/mm forthe SECs and 0.376 mm/mm for the RSGs, showing a com-parable, yet improved, normalized shape. Taken over theentire length of the test, the average RMS errors augmentto 1.32 mm/mm and 3.50 mm/mm for the SECs and RSGs,respectively. This increase is due to the larger errors at theunloaded conditions, during which the polynomial recon-struction gives inaccurate results.

Lastly, we subject the specimen to the triangular load(displacement-based) with increasing rate described in Sec-tion III.B. The strain time histories for the SECs andRSGs, and the actuator displacement history are shownin Fig. 11. Results show that the SECs are capable oftracking the strain history, but that the SECs increasingly

6 IEEE/ASME TRANS. ON MECHATRONICS

Fig. 10. Deflection shapes. (a) Non-normalized; and (b) normalized.

underestimate strain with increasing strain magnitude, ex-cept for SEC3 which is overestimating strain. Another fea-ture is that SEC2 is showing an important difference withrespect to SEC3 in the measurements, a difference mini-mized between RSG2 and RSG3. Note that this differenceshould theoretically be zero if the sensors are placed sym-metrically. This difference was not observed in the trian-gular load test (Fig. 9). As discussed above, a small offsetin the installation can explain the difference between RSG2

and RSG3. Fig. 12 compares the root mean square (RMS)error of the normalized deflection shapes with respect tothe analytical solution. The SEC network obtains a moreaccurate shape than the RSG network beyond an initiallevel of loading. Here as well, when the beam is unloaded,the noise in the sensors signals results in inaccurate de-flection shapes. The significant difference in performancebetween both sensors can be attributed to the SECs av-eraging strain over a large area, while the RSGs measurea localized strain. SECs are less sensitive to placementerrors.

Fig. 11. Strain history for four SECs. (a) SECs; and (b) RSGs andactuator displacement.

IV. Discussion and Conclusion

We have presented a sensor network developed for strainsensing over large surfaces. The network consists of SECsarranged in an array form, transducing strain into changesin capacitance. These elastomeric sensors are fabricated inlaboratory using solutions of SEBS+TiO2 for the dielectricand SEBS+CB for the compliant electrodes.

Results from the experimental validation demonstratedthat the sensor compares well against off-the-shelf resistivestrain gages. Given the relative size of a single SEC com-

Fig. 12. RMS error of the normalized deflection shapes with respectto the analytical solution.

pares with a RSG, it follows that the sensor can be usedas a surface strain gage capable of covering large areas. Itwas also shown that the sensor can measure large strains,in the levels of 0-20%. The installation procedure used forthe laboratory experiments consisted of a hand-installationusing an off-the-shelf epoxy. This demonstrates the easyapplicability of the sensing solution.

Results also shown the sensor has a tendency to un-derestimate strain. The underestimation of strain can bedue to parasitic capacitance from the wires, complexity inmeasuring low changes in capacitance using off-the-shelfDAQ, drift in the signal, impurities in the sensor fabri-cation, and/or inconsistencies during the manual installa-tion. We envision the development of a dedicated dataacquisition systems for measuring differential capacitance,which would increase the resolution of the sensor at levelscomparable to conventional strain gages. This underesti-mation of strain limits the capability of extracting accuratephysics-based features associated with non-normalized dis-placement values.

The SEC sensor network has been demonstrated in afour SECs setup to detect deflection shapes of a simplysupported beam. The deflection shapes have been recon-structed using a simple polynomial algorithm. Resultshave shown that the SECs were underestimating the realshape, consistent with results discussed above, but wereestimating the normalized deflection shapes accurately.Comparisons with RSGs showed that the SECs were es-timating the normalized shapes with less error. This goodperformance can be explained partly by the capacity of theSECs to average strain over a large area, unlike RSGs thatmeasure localized strain. This particularity represents astrong advantage of SECs over RSGs.

The proposed sensor network is a promising tool forconducting SHM at the mesoscale. In a network setup,the network could be used to collect two-dimension straindata, from which physics-based features can be extracted,including over-stresses, torsion, utilization history, etc.

: TITLE 7

These features can subsequently be utilized as input toforecast models to conduct structural diagnostic and prog-nostic. Possible applications other than extraction of de-flection shapes include crack detection and localization onconcrete structures, detection of permanent deformationson steel members, and weigh-in-motion sensing.

Acknowledgments

This work is supported in part by grant 1001062565 fromthe Iowa Alliance for Wind Innovation and Novel Develop-ment (IAWIND), and by grants ECCS 0903530 and ECCS1102213 from the National Science Foundation (NSF). Thissupport is gratefully acknowledged, along with the tech-nical support of Douglas L. Wood from the Departmentof Civil, Construction, and Environmental Engineering atIowa State University.

References

[1] B. Lin, V. Giurgiutiu, Z. Yuan, J. Liu, C. Chen, J. Jiang, A.S.Bhalla, and R. Guo, “Ferroelectric thin-film active sensors forstructural health monitoring,” in Proc. of SPIE, 2007, vol. 6529,p. 65290I.

[2] V. Sotoudeh, R.J. Black, B. Moslehi, and M.A. Plavsic,“Fiber-optic defect and damage locator system for wind tur-bine blades,” Tech. Rep., Intelligent Fiber Optic SystemsCorp.(IFOS), 2010.

[3] A. Turner and T.W. Graver, “Structural monitoring of windturbine blades using fiber optic bragg grating strain sensors,”Experimental Mechanics on Emerging Energy Systems andMaterials, Volume 5, pp. 149–154, 2011.

[4] T. Harms, S. Sedigh, and F. Bastianini, “Structuralhealth monitoring of bridges using wireless sensor networks,”Instrumentation & Measurement Magazine, IEEE, vol. 13, no.6, pp. 14–18, 2010.

[5] D. Adams, J. White, M. Rumsey, and C. Farrar, “Structuralhealth monitoring of wind turbines: method and application toa hawt,” Wind Energy, 2011.

[6] S.J. Dyas, J. Scheidler, S.G. Taylor, K. Farinholt, and G. Park,“Structural health monitoring of wind turbine blades under fa-tigue loads,” Topics in Experimental Dynamics Substructuringand Wind Turbine Dynamics, Volume 2, pp. 227–245, 2012.

[7] JM Ko and YQ Ni, “Technology developments in struc-tural health monitoring of large-scale bridges,” Engineeringstructures, vol. 27, no. 12, pp. 1715–1725, 2005.

[8] Y. Zhang, GM Lloyd, and ML Wang, “Random VibrationResponse Testing of PVDF Gages for Long-span Bridge Mon-itoring,” in Proc. 4th Int. Workshop on Structural HealthMonitoring, Stanford, 2003, p. 115.

[9] S. Hurlebaus and L. Gaul, “Smart layer for damage diagnostics,”Journal of intelligent material systems and structures, vol. 15,no. 9-10, pp. 729–736, 2004.

[10] JA Carlson, JM English, and DJ Coe, “A flexible, self-healingsensor skin,” Smart materials and structures, vol. 15, pp. N129,2006.

[11] U. Tata, S. Deshmukh, JC Chiao, R. Carter, and H. Huang,“Bio-inspired sensor skins for structural health monitoring,”Smart Materials and Structures, vol. 18, pp. 104026, 2009.

[12] K.J. Loh, T.C. Hou, J.P. Lynch, and N.A. Kotov, “Carbonnanotube sensing skins for spatial strain and impact damageidentification,” Journal of Nondestructive Evaluation, vol. 28,no. 1, pp. 9–25, 2009.

[13] L. Gao, E.T. Thostenson, Z. Zhang, J.H. Byun, and T.W. Chou,“Damage monitoring in fiber-reinforced composites under fa-tigue loading using carbon nanotube networks,” PhilosophicalMagazine, vol. 90, no. 31-32, pp. 4085–4099, 2010.

[14] PM Harrey, BJ Ramsey, PSA Evans, and DJ Harrison,“Capacitive-type humidity sensors fabricated using the off-set lithographic printing process,” Sensors and Actuators B:Chemical, vol. 87, no. 2, pp. 226–232, 2002.

[15] Hyun Pyo Hong, Kyung Hoon Jung, Nam Ki Min, Yong HoonRhee, and Chan Won Park, “A highly fast capacitive-typehumidity sensor using percolating carbon nanotube films as a

porous electrode material,” in Sensors, 2012 IEEE. IEEE, 2012,pp. 1–4.

[16] Darren J Lipomi, Michael Vosgueritchian, Benjamin CK Tee,Sondra L Hellstrom, Jennifer A Lee, Courtney H Fox, andZhenan Bao, “Skin-like pressure and strain sensors basedon transparent elastic films of carbon nanotubes,” Naturenanotechnology, vol. 6, no. 12, pp. 788–792, 2011.

[17] KI Arshak, D McDonagh, and MA Durcan, “Development ofnew capacitive strain sensors based on thick film polymer andcermet technologies,” Sensors and Actuators A: Physical, vol.79, no. 2, pp. 102–114, 2000.

[18] Michael Suster, Jun Guo, Nattapon Chaimanonart, Wen HKo, and Darrin J Young, “A high-performance mems capac-itive strain sensing system,” Microelectromechanical Systems,Journal of, vol. 15, no. 5, pp. 1069–1077, 2006.

[19] Jagoda Anna Dobrzynska and MAM Gijs, “Polymer-basedflexible capacitive sensor for three-axial force measurements,”Journal of Micromechanics and Microengineering, vol. 23, no. 1,pp. 015009, 2013.

[20] S. Laflamme, M. Kollosche, JJ Connor, and G. Kofod, “Softcapacitive sensor for structural health monitoring of large-scalesystems,” Structural Control and Health Monitoring, 2011.

[21] S. Laflamme, M. Kollosche, JJ Connor, and G. Kofod, “Robustflexible capacitive surface sensor for structural health monitor-ing applications,” ASCE J. Eng. Mech., in press.

[22] AN Wilkinson, ML Clemens, and VM Harding, “The effects ofsebs-¡ i¿ g¡/i¿-maleic anhydride reaction on the morphology andproperties of polypropylene/pa6/sebs ternary blends,” Polymer,vol. 45, no. 15, pp. 5239–5249, 2004.

[23] M. Kollosche, H. Stoyanov, S. Laflamme, and G. Kofod,“Strongly enhanced sensitivity in elastic capacitive strain sen-sors,” J. Mater. Chem., 2011.

[24] A. Amin, MS Alam, and Y. Okui, “An improved hyperelasticityrelation in modeling viscoelasticity response of natural and highdamping rubbers in compression: experiments, parameter iden-tification and numerical verification,” Mechanics of materials,vol. 34, no. 2, pp. 75–95, 2002.

[25] HJ Qi and MC Boyce, “Stress–strain behavior of thermoplasticpolyurethanes,” Mechanics of Materials, vol. 37, no. 8, pp. 817–839, 2005.

[26] AR Bhuiyan, Y. Okui, H. Mitamura, and T. Imai, “A rheol-ogy model of high damping rubber bearings for seismic analy-sis: Identification of nonlinear viscosity,” International Journalof Solids and Structures, vol. 46, no. 7, pp. 1778–1792, 2009.

[27] MK Razzaq, AR Bhuiyan, Y. Okui, H. Mitamura, and T. Imai,“Effect of rubber bearings modeling on seismic response of baseisolated highway bridge,” 7th CUEE and 5th ICEE Joint Con-ference, 7.

[28] RT Jones, DG Bellemore, TA Berkoff, JS Sirkis, MA Davis,MA Putnam, EJ Friebele, and AD Kersey, “Determination ofcantilever plate shapes using wavelength division multiplexedfiber bragg grating sensors and a least-squares strain-fitting al-gorithm,” Smart materials and structures, vol. 7, no. 2, pp. 178,1999.

[29] A. Tessler and J.L. Spangler, “A least-squares variationalmethod for full-field reconstruction of elastic deformations inshear-deformable plates and shells,” Computer methods inapplied mechanics and engineering, vol. 194, no. 2, pp. 327–339,2005.

[30] M. Nishio, T. Mizutani, and N. Takeda, “Structural shape re-construction with consideration of the reliability of distributedstrain data from a brillouin-scattering-based optical fiber sen-sor,” Smart Materials and Structures, vol. 19, no. 3, pp. 035011,2010.

[31] R. Glaser, V. Caccese, and M. Shahinpoor, “Shape monitor-ing of a beam structure from measured strain or curvature,”Experimental mechanics, vol. 52, no. 6, pp. 591–606, 2012.

[32] Hristiyan Stoyanov, Matthias Kollosche, Sebastian Risse, De-nis N McCarthy, and Guggi Kofod, “Elastic block copolymernanocomposites with controlled interfacial interactions for arti-ficial muscles with direct voltage control,” Soft Matter, vol. 7,no. 1, pp. 194–202, 2011.

8 IEEE/ASME TRANS. ON MECHATRONICS

Simon Laflamme received a B.Com (2003)and B.Eng (2006) degrees from McGill Univer-sity, Montreal, Canada. He received a M.Engdegree in High Performance Structures (2007)and a Ph.D degree in Structures and Materials(2011), both from MIT. He is an Assistant Pro-fessor at Iowa State University in the Depart-ment of Civil, Construction, and Environmen-tal Engineering, holds a courtesy appointmentin the Department of Electrical and ComputerEngineering, and is affiliated with the Center for

Nondestructive Evaluation. His research interests include structuralhealth monitoring, structural control, and smart systems.

Hussam S. Saleem received his B.Sc. in CivilEngineering from the University of Jordan, Am-man, Jordan and MS in Civil engineering fromWashington State University, Pullman, WA. Heis a Ph.D candidate in the Department of Civil,Construction, and Environmental Engineeringat Iowa State University.

Bharath K. Vasan received his Ph.D degree inElectrical and Computer Engineering from IowaState University in 2013.

Randall L. Geiger is a Professor in the De-partment of Electrical and Computer Engineer-ing at Iowa State University and holds the Tuncand Lala Professorship. His research is focusedon the design and testing of analog and mixed-signal integrated circuits. Dr. Geiger is a Fel-low of the IEEE, a past president of the IEEECircuits and Systems Society, a past member ofthe IEEE Publications Board, a past chair ofthe Transactions Committee of the IEEE Peri-odicals Council, and currently serves as chair of

the IEEE Kirchoff Award Committee.

Degang Chen is a Professor of Electrical andComputer Engineering at Iowa State University,with PhD from UCSB in 1992 and summer po-sitions at Boeing in 1999, Maxim in 2001 andTexas Instruments in 2011 and 2012. His re-search is on analog and mixed-signal integratedcircuit design and test, specializing in test timereduction and precision test without precisioninstruments. He has published over 200 pa-pers, 12 of which received best paper awards andother honors, including the International Test

Conference Ned Kornfield Best Paper Award in 2013.

Michael Kessler is Professor and Director ofthe School of Mechanical and Materials Engi-neering at Washington State University (WSU).Prior to his appointment at WSU in August2013, he was the Wilkinson Professor of In-terdisciplinary Engineering in the Departmentof Materials Science and Engineering at IowaState University. He obtained his PhD from theUniversity of Illinois at Urbana-Champaign in2002. His research focuses on the mechanics,processing, and characterization of polymer ma-

trix composites and nanocomposites.

Krishna Rajan is a professor in the Materi-als Science and Engineering department at IowaState University and holds the Stanley Chair ofInterdisciplinary Engineering. He is also direc-tor of the international combinatorial sciencesand materials informatics collaboratory, and di-rector of the Institute for Combinatorial Dis-covery. Dr. Rajan received a B.A.Sc. in metal-lurgy and materials science from the Universityof Toronto in 1974 and an Sc.D. in materials sci-ence from the Massachusetts Institute of Tech-

nology in 1978.