A Spray-On Carbon Nanotube Artificial Neuron ... - Semantic Scholar

0 downloads 0 Views 5MB Size Report
Jul 26, 2016 - Department of Architectural Engineering, Namseoul University, Cheonan ... Smart Structures and Bio-Nanotechnology Laboratory, University of ...
sensors Article

A Spray-On Carbon Nanotube Artificial Neuron Strain Sensor for Composite Structural Health Monitoring Gyeongrak Choi 1 , Jong Won Lee 2 , Ju Young Cha 3 , Young-Ju Kim 4 , Yeon-Sun Choi 5 , Mark J. Schulz 6 , Chang Kwon Moon 7 , Kwon Tack Lim 7 , Sung Yong Kim 7 and Inpil Kang 7, * 1 2 3 4 5 6 7

*

Manufacturing Automation R & BD Group, Korea Institute of Industrial Technology, Chonan 143-701, Korea; [email protected] Department of Architectural Engineering, Namseoul University, Cheonan 331-707, Korea; [email protected] Department of Medical Device, Korea Institute of Machinery & Materials, Daegu 711-880, Korea; [email protected] Mineral Resources Research Division, Korea Institute of Geoscience and Mineral Resources, Daejeon 305-350, Korea; [email protected] School of Mechanical Engineering, Sungkyunkwan University, Suwon 440-746, Korea; [email protected] Smart Structures and Bio-Nanotechnology Laboratory, University of Cincinnati, Cincinnati, OH 45221, USA; [email protected] Graduate School of Pukyong National University, Busan 608-739, Korea; [email protected] (C.K.M.); [email protected] (K.T.L.); [email protected] (S.Y.K.) Correspondence: [email protected]; Tel.: +82-51-629-6167

Academic Editor: Jandro L. Abot Received: 25 February 2016; Accepted: 22 July 2016; Published: 26 July 2016

Abstract: We present a nanocomposite strain sensor (NCSS) to develop a novel structural health monitoring (SHM) sensor that can be easily installed in a composite structure. An NCSS made of a multi-walled carbon nanotubes (MWCNT)/epoxy composite was installed on a target structure with facile processing. We attempted to evaluate the NCSS sensing characteristics and benchmark compared to those of a conventional foil strain gauge. The response of the NCSS was fairly good and the result was nearly identical to the strain gauge. A neuron, which is a biomimetic long continuous NCSS, was also developed, and its vibration response was investigated for structural damage detection of a composite cantilever. The vibration response for damage detection was measured by tracking the first natural frequency, which demonstrated good result that matched the finite element (FE) analysis. Keywords: carbon nanotube; artificial neuron; strain sensor; composites; structural health monitoring; piezoresistivity; damage detection

1. Introduction Composites are becoming crucial structural materials in high technology industries. Composites offer excellent strength/stiffness-to-weight, resistance to corrosion and chemical attack, good electrical behavior, and typically provide sustainable and low maintenance. Due to such advantages of composites, the Boeing 787 designed carbon fiber reinforced composite airframe and primary structure saves 20% more fuel than conventionally similar airplanes [1]. Despite such great merits, composites consisting of multiple laminar layers are inherently prone to failure owing to delamination of the multiple laminar layers, moisture absorption, and low speed impact. Structural applications of composites are still more challenging and less predictable than metals. More sophisticated structural health monitoring (SHM) and nondestructive inspection (NDI) techniques are required to detect defects of composites than metals. Traditional NDI of composites Sensors 2016, 16, 1171; doi:10.3390/s16081171

www.mdpi.com/journal/sensors

Sensors 2016, 16, 1171

2 of 11

include visual inspection, automated tap tests, ultrasonic inspection, radiography, and thermography. To improve the efficiency of the visual inspection conducted with the naked eye, some techniques have been developed such as impact sensitive coatings, liquid penetrants, and magnetic particles. The vibration analysis in terms of natural frequency change is reported as an effective method compared to ultrasound transmission techniques for filament wound carbon fiber reinforced plastic (CFRP) tubes [2]. Some techniques use a bulky transducing system and need to eliminate individual components for testing, which have a high cost and are labor intensive. The inspection cost is more than one-third of acquiring and operating the structure [3]. Therefore, a more economical and effective way is required. A promising sensor should be light, small, and easily installed in a structure with high sensitivity to cover a large target area for an SHM. A sensor made of carbon nanotubes (CNTs) might meet such demands due to their unique properties [4,5]. CNTs can be used to form smart materials and possibly produce various kinds of sensors for composites. CNTs have been studied to develop novel sensors due to their promising structural and electrical characteristics [6–10]. For applying CNTs as strain sensors for engineering purposes, CNTs can be fabricated as a smart nanocomposite through a composite process [11–15]. In addition, a liquid type CNT sensor can also be obtained from a composite process and may be useful as a novel smart paint sensor for SHM. Smart paint has the advantage of being able to produce complex shapes of a target structure with a light sensor equipment load [16–18]. However, the CNT strain sensor is still in early development for SHM. The CNT strain sensing performance by means of voltage responses need to be systematically assessed for structural deterioration monitoring under various loading conditions. Therefore, we report the multi-walled carbon nanotube (MWCNT) composite strain sensing properties under static and dynamic loading conditions and experimentally studied the composite to develop a neuron as a continuous strain sensor. The long continuous strain sensor was named a neuron [19]. Large structures, such as bridges and buildings, need to be monitored over long spans, and the multi-dimensional strain in the structure must be measured. In large structural applications, it is difficult to install a sufficient number of conventional strain gauges because they require a large amount of wiring and electronics. However, a neuron can cover a large area and detect the strain along it due to its continuity. A CNT neuron can measure strains occurring along its length and it can be easily installed on large structures using a spray-on technique [12]. For engineering applications, it is necessary to develop more economic materials for the sensors and to perform systematic studies of the sensing characteristics. Since a displacement transducer has the advantage of monitoring a structure under large bending deflections with low frequency vibration [20], we characterized the strain sensor using the frequency responses. Structural vibration based health monitoring has been used to detect damage in a structure through altered dynamic characteristics and the change is characterized by changes in the natural frequency [21,22]. The vibration response of the neuron was investigated with experimental and finite element (FE) analysis for structural damage detection of a composite cantilever by tracking the first natural frequency. 2. Fabrication of the Neuron The artificial neuron was fabricated with MWCNT and epoxy. Epoxy (Kukdo-chemical Co., Seoul, Korea, YD-128) was used as a matrix for the nanocomposite because it shows excellent mechanical strength and is easily fabricated as a composite. Commercially-obtained MWCNTs (Hanwha Chemical Co., Seoul, Korea, CM-250) were used for the nanocomposite filler. Though single-walled carbon nanotubes (SWCNTs) have excellent electromechanical properties, because of the high cost and difficult incorporation into polymers at high loading, this study used an MWCNT composite to develop a continuous strain sensor. The piezoresistive properties of MWCNT have been greatly improved due to the successful development of the synthesis process. The base MWCNT solution was prepared using a similar process to the one described in our previous study, and the piezoresistive characteristics of both PMMA and epoxy composites are nearly similar [6].

Sensors 2016, 16, 1171 Sensors2016, 2016,16, 16,1171 1171 Sensors

3 of 11 of10 10 33 of

mixedMWCNT MWCNTsolution solution was sprayed with an airbrush airbrush on aa patterned patterned surface of aa The mixed was sprayed withwith an airbrush on a patterned surface of a structure The mixed MWCNT solution was sprayed an on surface of structure as shown in Figure 1a. The sprayed-on neuron was cured at room temperature until as shown in Figure 1a. The sprayed-on neuron was cured at room temperature until completely dry. structure as shown in Figure 1a. The sprayed-on neuron was cured at room temperature until completely dry. The The spray process was times repeated several times to tothin obtain uniform thin layer. After The spray process was repeated several to obtain a uniform layer. After the curing process, completely dry. spray process was repeated several times obtain aa uniform thin layer. After curing process, the electrical wires were connected using silver conducting epoxy (Cleantech, the electrical wires were connected using silver conducting epoxy (Cleantech, Ansan, Korea, Elcoat the curing process, the electrical wires were connected using silver conducting epoxy (Cleantech, Ansan,to Korea, Elcoat P-100) resistance. to reduce reduce the the contactlayer resistance. An epoxy epoxy layer or another another polymer P-100) reduce the contact An contact epoxy or another polymer layer can be optionally Ansan, Korea, Elcoat P-100) to resistance. An layer or polymer layer can can be optionally coated on the thethe neuron layer tocontamination protect the the neuron neuron from contamination contamination and coated on be theoptionally neuron layer to protect neuron fromto and external damage. Figure 1b layer coated on neuron layer protect from and external damage. Figure 1b shows the neuron on a plastic beam. Patterning was required to install shows the neuron on a plastic beam. Patterning was required to install the neuron on a structure with external damage. Figure 1b shows the neuron on a plastic beam. Patterning was required to install the neuron neuronsensor on aa structure structure with with controlled sensor dimensions. The length length and thickness affect the controlled dimensions. Thecontrolled length andsensor thickness affect the resistance ofand the neuron thataffect is related the on dimensions. The thickness the resistance of the neuron that is related to the sensitivity neuron. If the sensor electrode of the to the sensitivity of the neuron. If the sensor electrode of the nanocomposite strain sensor (NCSS) resistance of the neuron that is related to the sensitivity of the neuron. If the sensor electrode of the nanocomposite strain sensor (NCSS) is wider, the strain response of the sensor might be mixed with is wider, the strain response of the sensor might be mixed with axial and transverse deformation. nanocomposite strain sensor (NCSS) is wider, the strain response of the sensor might be mixed with axial andto transverse deformation. In order order to to read read the proper proper strain information, the pattering of the the In order read the proper strain information, the pattering ofstrain the sensor electrodethe should be studied. axial and transverse deformation. In the information, pattering of sensor electrode should be studied. The major advantage of the spray-on neuron as a strain sensor The major advantage of the spray-on neuron as a strain sensor was ease of use on complex surface sensor electrode should be studied. The major advantage of the spray-on neuron as a strain sensor was ease easeincluding of use use on onwelded complexparts, surface shapes, including welded can parts, where stress concentrations can shapes, where stress concentrations occur andstress are difficult to measure was of complex surface shapes, including welded parts, where concentrations can occur and and are difficult difficult to measure measure using conventional conventional strain strain gauges. gauges. using conventional strain gauges. using occur are to

(a) (a)

(b) (b)

Figure 1. Fabrication of of the multi-walled multi-walled carbon carbon nanotube nanotube (MWCNT)/epoxy (MWCNT)/epoxy neuron neuron with with spray: spray: Figure Figure 1.1. Fabrication Fabrication the of the multi-walled carbon nanotube (MWCNT)/epoxy neuron with (a) spray on a patterned bar; (b) fabricated neuron on a plastic ruler (3 wt. % MWCNT, (a) spray a patterned bar; bar; (b) fabricated neuron on on a plastic spray: (a) on spray on a patterned (b) fabricated neuron a plasticruler ruler(3(3wt. wt. %% MWCNT, MWCNT, 220 mm ×× 10 10 mm ×× 0.33 0.33 mm, RR == 3.07 3.07 kΩ). kΩ). 220 220 mm mm ˆ 10mm mm ˆ 0.33mm, mm, R = 3.07 kΩ).

Figure 22 shows shows the the FE-SEM FE-SEM images images of of the the fractured fractured side side surface surface of of an an MWCNT/epoxy MWCNT/epoxy Figure Figure 2 shows the FE-SEM images of the fractured side surface of an MWCNT/epoxy composite composite fabricated fabricated using using the the same same dispersion dispersion process process that that was was used used to to form form the the neuron. neuron. Good composite Good fabricated using the same dispersion process that was used to form the neuron. Good dispersion of dispersion of MWCNTs in the epoxy composite was achieved using this fabrication process and the dispersion of MWCNTs in the epoxy composite was achieved using this fabrication process and the MWCNTs the epoxypercolation composite level was achieved using this 0.1 fabrication process and thecompared electrically electricallyin conducting was approximately approximately wt. %, %, which which is normal normal to electrically conducting percolation level was 0.1 wt. is compared to conducting percolation level was approximately 0.1 wt. %, which is normal compared to the literature [22]. the literature literature [22]. [22]. the

(a) (a)

(b) (b)

Figure2.2. 2.FE-SEM FE-SEMimages imagesof ofsprayed sprayedMWCNT/epoxy MWCNT/epoxy (3 (3 wt. wt. %) %) samples: samples: (a) (a) 5000× 5000× and and (b) 25,000×. Figure FE-SEM images of sprayed MWCNT/epoxy Figure 5000ˆ and(b) (b)25,000×. 25,000ˆ.

Sensors 2016, 16, 1171

4 of 11

Sensors 2016, 16, 1171

4 of 10

3. Characteristics of the MWCNT/Epoxy Strain Sensor 3. Characteristics of the MWCNT/Epoxy Strain Sensor 3.1. Piezoresistive Characteristics of the MWCNT/Epoxy Composites 3.1. Piezoresistive Characteristics of the MWCNT/Epoxy Composites The piezoresistive characteristics of the nanocomposites were investigated to determine the The piezoresistive characteristics of the nanocomposites were investigated to determine the variations of electrical properties under structural deformations. The film type specimens were variations of electrical properties under structural deformations. The film type specimens were fabricated to investigate strain sensing properties under static and dynamic loading conditions and fabricated to investigate strain sensing properties under static and dynamic loading conditions and to calibrate neurons. The short length specimens were fabricated in the same way as the nano-spray to calibrate neurons. The short length specimens were fabricated in the same way as the nano-spray process, with the change in CNT wt. % fraction set as 0.7, 2, 3, and 10. Specimens were installed on process, with the change in CNT wt. % fraction set as 0.7, 2, 3, and 10. Specimens were installed on a a cantilever and connected to wires with silver conductive epoxy on the surface to build the electrodes cantilever and connected to wires with silver conductive epoxy on the surface to build the electrodes of the NCSS. TheThe resistance withaamulti-meter multi-meter of the NCSS. resistancechanges changesofofthe theNCSS NCSSelectrodes electrodes were were measured measured with (Yokogawa, Tokyo, Japan, 732) and the cantilever beam deflection was measured using laser (Yokogawa, Tokyo, Japan, 732) and the cantilever beam deflection was measured using a alaser displacement sensor (Keyence, Osaka, displacement sensor (Keyence, Osaka,Japan, Japan,il-300) il-300)as asshown shownin in Figure Figure 3. 3.

Figure 3. Schematic illustration of the strain measurement. Figure 3. Schematic illustration of the strain measurement.

The measured displacements were converted to strains by cantilever beam theory, and the The measured were to strains by cantilever beam theory, and the resistance changesdisplacements were normalized (RN)converted with Equation (1): resistance changes were normalized (RN ) with Equation (1):

Rs  R0 Rs ´RR0 0 RN “ RN 

(1)

(1) R0 R0 is the resistance without any displacement and Rs is the measured sensor resistance when the R0 isisthe resistance without anythe displacement and Rs ismodel the measured sensor resistance when the beam strained. Figure 4 shows experimental strain of the MWCNT/epoxy NCSS from beam strained. Figure shows the experimental strain 4a, model of the MWCNT/epoxy NCSS from theisresults of the above4conversions. As shown in Figure the MWCNT/epoxy composites showed the excellent results oflinearity the above conversions. As shown in Figure 4a, the showed under compression and expansion loads. TheMWCNT/epoxy slope represents composites the sensitivity (Sg) of the fabricated NCSScompression and corresponds the gaugeloads. factor The of theslope sensor determined Equation(S (2): excellent linearity under andto expansion represents theby sensitivity g ) of the fabricated NCSS and corresponds to the gauge factor of the sensor determined by Equation (2): RN Sg  (2) Na ∆R Sg “ (2) ∆ε a Here, εa is the axial strain due to the beam deflection. To investigate the linearity and gauge factor MWCNT piezoresistivity, the relationship the content ratio ofand MWCNT and Here,ofεathe is the axial strain due to the beam deflection. between To investigate the linearity gauge factor the gauge factor of the strain sensors was investigated as shown in Figure 4b. of the MWCNT piezoresistivity, the relationship between the content ratio of MWCNT and the gauge shows fairly linear strain factor ofThe the piezoresistivity strain sensors was investigated as bidirectional shown in Figure 4b.response over various contents of NCSS. The piezoresistive mechanism of CNT composites strongly relate electrical contactof The piezoresistivity shows fairly linear bidirectional strain response overtovarious contents resistance change of the fillers in the matrix and tunneling effects of the fillers [6,23]. The resistance NCSS. The piezoresistive mechanism of CNT composites strongly relate to electrical contact resistance change in case of lower MWCNT contents is much larger than higher contents of MWCNT change of the fillers in the matrix and tunneling effects of the fillers [6,23]. The resistance change in

Sensors 2016, 16, 1171

5 of 11

Sensors 2016, 16, 1171 10 that case of lower MWCNT contents is much larger than higher contents of MWCNT composites5 of and makes higher sensitivity of piezoresistivity. The abundant electrical conductive networks in higher composites and that makes higher sensitivity of piezoresistivity. The abundant electrical conductive content composites may hamper the piezoresistive changes. As shown in Figure 4b, the piezoresistive networks in higher content composites may hamper the piezoresistive changes. As shown in sensitivity was affected by the filler concentration and the composites near the percolation threshold Figure 4b, the piezoresistive sensitivity was affected by the filler concentration and the composites displayed higher sensitivity than displayed composites at higher concentrations. To design theconcentrations. sensor sensitivity, near the percolation threshold higher sensitivity than composites at higher the proper content point at the curve kneecontent pointpoint can at bethe selected andpoint the can optimized sensitivity To design the sensor sensitivity, the proper curve knee be selected and considering the weight of MWCNT canthe be weight achieved at 2–3 wt. %.be achieved at 2–3 wt. %. the optimized sensitivity considering of MWCNT can

(a)

(b)

Figure 4. Piezoresistive characteristics of MWCNT/epoxy composites for strain sensors:

Figure 4. Piezoresistive characteristics of MWCNT/epoxy composites for strain sensors: (a) experimental strain model and (b) gauge factor w.r.t MWCNT wt. %. (a) experimental strain model and (b) gauge factor w.r.t MWCNT wt. %.

3.2. Strain Sensing Characteristics of the MWCNT/Epoxy NCSS

3.2. Strain Sensing Characteristics of the MWCNT/Epoxy NCSS

To benchmark to a conventional strain gauge, a foil strain gauge (CAS, Yangju, Korea, AP-11 S30N-120-EC) with thestrain NCSSgauge, on the same as illustrated in Figure 3. The To benchmarkwas to ainstalled conventional a foil cantilever, strain gauge (CAS, Yangju, Korea, AP-11 NCSS should be characterized by means of its voltage responses to compare its performances withNCSS S30N-120-EC) was installed with the NCSS on the same cantilever, as illustrated in Figure 3. The thebe commercial sensor,by and we utilized signal processing for the sensors. should characterized means of itstraditional voltage responses to compare its performances with the Table 1 shows the specification of the foil strain gauge and the NCSS. The experiment was commercial sensor, and we utilized traditional signal processing for the sensors. processed using 3 wt. % NCSS, which shows a similar gauge factor region to the foil strain gauge.

Table 1 shows the specification of the foil strain gauge and the NCSS. The experiment was processed using 3 wt. % NCSS, which shows similar gauge factor region to the foil strain gauge. Table 1.aSensor specification.

Foil Strain Gauge Nano Composite Strain Sensor (NCSS) Table 1. Sensor specification. Resistance [Ω] 120 2.7 k Gauge factor 2.1 2.8 Sensor (NCSS) Foil Strain Gauge Nano Composite Strain Size [mm] w × l × t 6 × 40 × 0.07 9.15 × 51.76 × 2.85 Resistance [Ω] 120 2.7 k Gauge factor 2.1 2.8 To evaluate the sensing stability of the sensors, we simultaneously measured the no loading Size [mm] w ˆ l ˆ t 6 ˆ 40 ˆ 0.07 9.15 ˆ 51.76 ˆ 2.85 sensor outputs and step responses of NCSS and the strain gauge under static loads to determine the response performance of the sensors, as shown in Figure 5. The NCSS output was as as that we of the foil strain gauge. Under a the no loading To evaluate thevoltage sensing stability of stable the sensors, simultaneously measured no loading condition, the maximum absolute deviation of the NCSS voltage output (0.022 mV) was 22% greater sensor outputs and step responses of NCSS and the strain gauge under static loads to determine the than performance that of the commercial strainas gauge (0.018 mV). It5.was confirmed that very stable voltage response of the sensors, shown in Figure outputs were acquired from NCSS through improvements in both the electrical properties of the The NCSS voltage output was as stable as that of the foil strain gauge. Under a no loading nanocomposites and signal processing. Generally, it is difficult to obtain such a stable voltage output condition, the maximum absolute deviation of the NCSS voltage output (0.022 mV) was 22% greater from the CNT composite sensors due to resistance drift of the composite resistance originating from than the thatinherent of the electrical commercial strain gauge (0.018 mV). It was confirmed that very stable voltage changing property of CNTs according to surroundings [24]. However, we outputs were acquired from NCSS through improvements in both the electrical properties of the succeeded in controlling the resistance stability and obtained stable sensing characteristics from the nanocomposites and signal processing. results Generally, it reported is difficult to obtain NCSS. The process and experimental will be in detail later.such a stable voltage output from the CNT composite sensors due to resistance of the composite resistance To assess the linearity of the NCSS, the outputdrift voltage of the sensor was measuredoriginating according tofrom the deformation of the cantilever. Figure 6 shows the linear voltage response of the NCSS the inherent electrical changing property of CNTs according to surroundings [24]. compared However, we to the strain gauge after signal processing. succeeded in controlling the resistance stability and obtained stable sensing characteristics from the

NCSS. The process and experimental results will be reported in detail later.

Sensors 2016, 16, 1171

6 of 11

To assess the linearity of the NCSS, the output voltage of the sensor was measured according to the deformation of the cantilever. Figure 6 shows the linear voltage response of the NCSS compared to the strain gauge Sensors 2016, 16,after 1171 of 10 Sensors 2016, 16, 1171signal processing. 6 of6 10

Sensors 2016, 16, 1171

6 of 10

(a)

(a)

(b)

(b)

Figure 5. Response profiles of the MWCNT/epoxy nanocomposite strain sensor (NCSS) and strain Figure 5. Response profiles of of thethe MWCNT/epoxy nanocomposite strain sensor (NCSS) and Figure 5. Response MWCNT/epoxy strain sensor (NCSS) and load strainstrain gauge: (a) NCSS profiles voltage output without a load; nanocomposite (b) NCSS voltage output under a static gauge: (a) NCSS voltage output without a load; (b) NCSS voltage output under a static gauge: (a) NCSS voltage output without a load; (b) NCSS voltage output under a static load load (a) (b) (step response).

(step (step response). response).

Figure 5. Response profiles of the MWCNT/epoxy nanocomposite strain sensor (NCSS) and strain gauge: (a) NCSS voltage output without a load; (b) NCSS voltage output under a static load (step response).

Figure 6. Strain sensing linearity of the MWCNT/epoxy NCSS.

6. Strain sensing linearity ofscale the MWCNT/epoxy NCSS. In this test, Figure the Figure end6. point linearity was 3.8% full (FS) non-linearity, which was fairly linear Strain sensing linearity of the MWCNT/epoxy NCSS. for a strain sensor, and the response performances of the two sensors were nearly identical. Figure 6. Strain sensing linearity of the MWCNT/epoxy NCSS. In this the end point linearity full (FS) non-linearity, which was fairly linear The test, dynamic characteristics werewas also 3.8% tested, as scale shown in Figure 7.

In test,sensor, the end point linearityperformances was 3.8% full scale (FS) non-linearity, which was fairly linear forthis a strain and thepoint response of scale the two were nearly In this test, the end linearity was 3.8% full (FS)sensors non-linearity, which identical. was fairly linear for a strain sensor, and the response performances of the two sensors were nearly identical. The dynamic characteristics were also tested, as shown in Figure 7. for a strain sensor, and the response performances of the two sensors were nearly identical. The dynamic characteristics were also shownininFigure Figure The dynamic characteristics were alsotested, tested, as as shown 7. 7.

(a)

(b)

Figure 7. Dynamic characteristics MWCNT/epoxy NCSS; (a) voltage output response under 28 Hz excitation and (b) NCSS power spectra.

(a)

(b)

(a) (b) Figure 7. Dynamic characteristics MWCNT/epoxy NCSS; (a) voltage output response under 28 Hz Figure 7. Dynamic characteristics MWCNT/epoxy NCSS; (a) voltage output response under 28 Hz Figure 7. Dynamic characteristics MWCNT/epoxy NCSS; (a) voltage output response under 28 Hz excitation and (b) NCSS excitation and (b) NCSSpower powerspectra. spectra.

excitation and (b) NCSS power spectra.

Sensors 2016, 16, 1171

7 of 11

Sensors 2016, 16, 1171

7 of 10

The NCSS was confirmed for vibration monitoring of the structures. The phase discrepancy between NCSS and the strain gauge on Figure 7a should be structures. related to The a delay ordiscrepancy filtering in the The NCSS wasfoil confirmed for vibration monitoring of the phase data acquisition. Figure 7bfoil shows power spectra NCSS supports theorvibration between NCSS and the strainthe gauge on Figure 7a of should be that related to a delay filtering inanalysis the applications in this study. data acquisition. Figure 7b shows the power spectra of NCSS that supports the vibration analysis applications in this study. to measure the reliable static and dynamic strain information under The NCSS was validated The NCSS was validated to measure the reliable static andofdynamic strain information under loading and deformed structural conditions. The NCSS made the MWCNT/epoxy shows linear loading and deformed structural conditions. The NCSS made of the MWCNT/epoxy shows strain responses compared to those of the foil strain gauge. Thus, it is expected that linear it can be strain responses compared to those of the foil strain gauge. Thus, it is expected that it can be used as used as a new strain sensor. It is necessary to investigate the thermal and other characteristics a new strain sensor. It is necessary to investigate the thermal and other characteristics of the NCSS. of the NCSS. The MWCNT/epoxy based strain sensor can be fabricated as various types and applied The MWCNT/epoxy based strain sensor can be fabricated as various types and applied to to structural SHM. structural SHM. 4. Cross-Sectional Damage Detection Cantilever 4. Cross-Sectional Damage DetectionininaaVibrating Vibrating Cantilever A strain gaugegauge conventionally measures the point strain,strain, but the can measure the averaged A strain conventionally measures the point butneuron the neuron can measure the strainaveraged along the length of the structure where it was installed. Damages such as corrosion strain along the length of the structure where it was installed. Damages such as corrosion and and deterioration can be represented as a change the cross section a beam segment.Therefore, Therefore, this deterioration can be represented as a change in theincross section of aofbeam segment. study assumes theshown beam shown in Figure 8, which a cross sectional damagedzone zone wherein wherein the studythis assumes the beam in Figure 8, which has ahas cross sectional damaged the bending rigidity is reduced due to the change in cross sectional area. bending rigidity is reduced due to the change in cross sectional area.

Figure 8. Damaged cantilever beam with a strain neuron.

Figure 8. Damaged cantilever beam with a strain neuron.

A FE model with ANSYS (Canonsburg, PA, USA) was constructed for the glass fiber cantilever to study the ANSYS dynamic(Canonsburg, response capacity the neuron and establish baseline model for Abeam FE model with PA,ofUSA) was constructed fora the glassFE fiber cantilever damage assessment based on the vibration response ofestablish the neuron. Table 2 outlines the for beam subsequent to study the dynamic response capacity of the neuron and a baseline FE model baselinedamage parameters of the based beam used the analysis. The area, inertia of the the subsequent assessment on theinvibration response of themoment neuron.of Table 2 outlines cross-sectional area, and density in Table 2 were determined from measurements of the beam.

baseline parameters of the beam used in the analysis. The area, moment of inertia of the cross-sectional area, and density in Table 2 were determined from measurements of the beam. Table 2. Estimated geometric parameters of the beam. Thickness t (m) 0.00346

BeamTable width2. Estimated Estimated E Density 1 Length 2 geometric parameters ofLength the beam. B (m) E (Gpa) D (kg/m3) L1 (m) L2 (m) 0.024 25.1 1806 0.035 0.035

Length 3 L3 (m) 0.28

Thickness Beam width Estimated E Density Length 1 Length 2 Length 3 t (m) B (m) E (Gpa) D (kg/m3 ) L1 (m) L2 (m) L3 (m) Damage detection of the cantilever beam was performed using the vibration data from the 0.00346 0.024 such as a25.1 1806or other defect, 0.035 reduced 0.035 0.28 neuron. The local damage, crack, corrosion, the local stiffness of the structure that affects the global dynamic characteristics of the structure [21,22]. Therefore, health monitoring may beof possible by measuring in natural frequencies of the structure. In this Damage detection the cantilever beam the waschange performed using the vibration data from the neuron. study, only the first natural frequency was used because of the canceling effects of some modes (e.g., The local damage, such as a crack, corrosion, or other defect, reduced the local stiffness of the structure 3rd mode of the cantilever beam) when measuring the strain along the entire length of the beam, and that affects the global dynamic characteristics of the structure [21,22]. Therefore, health monitoring the first natural frequency is an important parameter to estimate the behavior of the large may be possible by measuring naturalthe frequencies of the structure. In frequencies this study, only infrastructure. In addition, the it is change easier toinmeasure first frequency than the higher due the first natural frequency wasofused because of the canceling effects of some modes (e.g., 3rd mode of the to the large amplitude vibration. cantilever Element beam) when the strain along entire lengthelement of the beam, therelatively first natural 2 (ΔL2measuring ) shown in Figure 8 was used the as the damaged becauseand of the large strain induced. The loss of bending rigidity of Element 2 for each damage case was frequency is an important parameter to estimate the behavior of the large infrastructure. In5.78%, addition, 11.08%, 20.96%, 25.26%, and 30.05%. cases of damage were artificially applied filing it is easier to17.40%, measure the first frequency thanSix the higher frequencies due to the largebyamplitude

of vibration.

Sensors 2016, 16, 1171

8 of 11

Element 2 (∆L2 ) shown in Figure 8 was used as the damaged element because of the relatively largeSensors strain2016, induced. 11.08%, 16, 1171 The loss of bending rigidity of Element 2 for each damage case was 5.78%, 8 of 10 17.40%, 20.96%, 25.26%, and 30.05%. Six cases of damage were artificially applied by filing both sides sides to of reduce an element reduce the beam led a reduction of the beam The inertia. of anboth element the to width of the thewidth beamof which led awhich reduction of the beam inertia. damage The damage ratios were calculated by cross sectional area reductions of the Element 2. ratios were calculated by cross sectional area reductions of the Element 2. The first natural frequencies of the beam were analyzed with ANSYS and the results shown in The first natural frequencies of the beam were analyzed with ANSYS and the results shown in Figure 9. Figure 9.

(a)

(b) Figure 9. ANSYS analysis examples for the first natural frequencies of the beam; (a) healthy beam

Figure 9. ANSYS analysis examples for the first natural frequencies of the beam; (a) healthy beam case case and (b) 30.5% damaged case. and (b) 30.5% damaged case.

Free vibration tests with the initial displacement method were experimentally performed for a healthy case and tests the six damage cases. The response of method the neuron wasexperimentally sampled at 50 kHz for 5 s. for Free vibration with the initial displacement were performed The first natural frequencies for the healthy and damaged beams based on measurements from a healthy case and the six damage cases. The response of the neuron was sampled at 50 the kHz for vibration tests were compared for to those obtained from the ANSYS in Table andmeasurements Figure 10. The from 5 s. The first natural frequencies the healthy and damaged beams based3 on measured vibration response under a structural deterioration closely matched the ANSYS analysis.

the vibration tests were compared to those obtained from the ANSYS in Table 3 and Figure 10. The measured vibration response under a structural deterioration closely matched the ANSYS analysis.

Sensors 2016, 16, 1171 Sensors 2016, 16, 1171

9 of 11 9 of 10

Table 3. First natural frequencies for various damage cases of Element 2 of the cantilever beam. Table 3. First natural frequencies for various damage cases of Element 2 of the cantilever beam.

Healthy Case 1 Case 2 Case 3 Case 4 Case 5 Case 6

Damage to Element 2 (%) Experimental (Hz) Damage to Element 2 (%) Experimental (Hz) 0 15.80 Healthy 0 15.80 5.78 15.60 Case 1 5.78 15.60 11.08 15.45 Case 2 11.08 15.45 17.40 15.25 Case 3 17.40 15.25 20.96 20.96 15.19 Case 4 15.19 Case 5 15.04 25.26 25.26 15.04 Case 6 14.85 30.05 30.05 14.85

ANSYS (Hz) 15.77 15.77 15.64 15.64 15.49 15.49 15.32 15.32 15.21 15.21 15.05 15.05 14.87 14.87

ANSYS (Hz)

Figure various damage damage cases. cases. Figure 10. 10. First First natural natural frequencies frequencies for for various

5. Conclusions 5. Conclusions A nanocomposite nanocomposite strain strainsensor sensor(NCSS) (NCSS)was wasstudied studiedtotodevelop develop a novel SHM sensor that could A a novel SHM sensor that could be be easily installed in a composite structure. The NCSS was fabricated with epoxy and MWCNT as a easily installed in a composite structure. The NCSS was fabricated with epoxy and MWCNT as a matrix matrix andfiller, nano filler, respectively. The NCSS was patterned on a using structure using atechnique, spray-on and nano respectively. The NCSS was patterned on a structure a spray-on technique, andit this it easy to apply it to existing structures for SHM applications.very We and this made easy made to apply it to existing structures for SHM applications. We accomplished accomplished very stable voltage outputs from the NCSS through improvements of both electrical stable voltage outputs from the NCSS through improvements of both electrical properties of the properties of theand nanocomposite and signal processing. This the study evaluated NCSS sensing nanocomposite signal processing. This study evaluated NCSS sensingthe characteristics by characteristics by benchmarking to those of a conventional foil strain gauge. Under a no loading benchmarking to those of a conventional foil strain gauge. Under a no loading condition, even though even though a maximum deviation of the NCSS voltage greater acondition, maximum absolute deviation of theabsolute NCSS voltage output was greater than output the foil was strain gauge,than the the foil strain gauge, the NCSS voltage output was as stable as that of a commercial sensor. For the NCSS voltage output was as stable as that of a commercial sensor. For the linearity test of the NCSS, linearity test oflinearity the NCSS, endFSpoint linearity was 3.8% FS non-linearity, whichsensor. is quiteThe linear for the end point wasthe 3.8% non-linearity, which is quite linear for a strain NCSS a strain sensor. The NCSS also showed good performance for vibration monitoring of the structures. also showed good performance for vibration monitoring of the structures. The dynamic characteristics Thethe dynamic characteristics the NCSS voltage output response nearly identical the NCSS 28 Hz of NCSS voltage output of response were nearly identical at thewere 28 Hz excitation test.atThe excitation test. The NCSS power spectra proved that it can be available for vibration based power spectra proved that it can be available for vibration based NDI techniques. NDI The techniques. NCSS neuron was developed as a continuous strain sensor. The vibration response of the The NCSS neuron was a continuous strain The cantilever. vibration response of the neuron was investigated fordeveloped structural as damage detection of a sensor. composite The measured neuron was investigated for structural damage detection of a composite cantilever. The measured vibration response and damage detection by tracking the first natural frequency showed good results vibration response damage detection by tracking the first strain naturalsensor frequency showed to good that matched the FEand analysis. The spray type, long continuous is anticipated be results that matched the FE analysis. The spray type, long continuous strain sensor is anticipated to a versatile sensor in mechanical and civil engineering because of its static and dynamic strain be a versatile sensor in mechanical and civil engineering because of its static and dynamic strain measurement with a longer range and less required data acquisition. measurement with a longer range and less required data acquisition. This study confirmed that NCSS might be able to measure reliable information under loading and deformed structural conditions. NCSS can provide assurance of the operational health of a

Sensors 2016, 16, 1171

10 of 11

This study confirmed that NCSS might be able to measure reliable information under loading and deformed structural conditions. NCSS can provide assurance of the operational health of a composite structure without the need for actuators or complex wave propagation analysis. The NCSS artificial neuron may effectively save the cost of SHM by reducing the number of channels of data acquisition compared to point sensors covering the same area. Thus, it may be applied to the state diagnosing system for composite structures. It is necessary to perform studies on the investigation of thermal and other characteristics of the NCSS. We plan to present more details on the strain sensing characteristics of NCSS with sophisticated experimental results including environmental conditions. Acknowledgments: This work was supported by a Research Grant of Pukyong National University (2015). Author Contributions: Gyeongrak Choi, Mark J. Schulz and Inpil Kang conceived and designed the experiments; Ju Young Cha and Sung Yong Kim performed the experiments; Jong Won Lee analyzed the data; Young-Ju Kim, Yeon-Sun Choi, Chang Kwon Moon and Kwon Tack Lim contributed materials and analysis tools; Inpil Kang wrote the paper.” Authorship must be limited to those who have contributed substantially to the work reported. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11.

12.

13.

14. 15.

Hale, J. Boeing 787 from the Ground Up. Available online: www.boeing.com (accessed on 16 February 2016). Cawley, P.; Woolfrey, A.M.; Adams, R.D. Natural frequency measurements for production quality control of fibre composites. Composites 1985, 16, 23–27. [CrossRef] Diamanti, K.; Soutis, C. Structural health monitoring techniques for aircraft composite structures. Prog. Aerosp. Sci. 2010, 46, 342–352. [CrossRef] Krishnan, A.; Dujardin, E.; Ebbesen, T.W.; Yianilos, P.N.; Treacy, M.M.J. Young’s modulus of single-walled nanotubes. Phys. Rev. B 1998, 58, 14013–14019. [CrossRef] Baughman, R.H.; Sakhidov, A.A.; de Heer, W.A. Carbon Nanotubes—the Route toward Applications. Science 2002, 297, 787–792. [CrossRef] [PubMed] Kang, I.; Schulz, M.J.; Kim, J.H.; Shanov, V.; Shi, D. A Carbon Nanotube Strain Sensor for Structural Health Monitoring. Smart Mater. Struct. 2006, 15, 737–748. [CrossRef] Knite, M.; Tupureina, V.; Fuith, A.; Zavickis, J.; Teteris, V. Polyisoprene-Multi-wall carbon nanotube composites for sensing strain. Mater. Sci. Eng. 2006, 27, 1125–1128. [CrossRef] Loh, K.J.; Kim, J.; Lynch, J.P.; Kam, N.W.S.; Kotov, N.A. Multifunctional layer-by-layer carbon nanotube-polyelectrolyte thin films for strain and corrosion sensing. Smart Mater. Struct. 2007, 16, 429–438. [CrossRef] Loyola, B.R.; Zhao, Y.; Loh, K.J.; Saponara, V.L. The electrical response of carbon nanotube-based thin film sensors subjected to mechanical and environmental effects. Smart Mater. Struct. 2013, 22, 1–11. [CrossRef] Chang, W.S.; Song, S.A.; Kim, J.H.; Han, C.S. Fabrication of Carbon Nanotube Strain Sensors. KSME 2009, 33, 773–777. [CrossRef] Kang, I.; Yoon, Y.H.; Kim, J.H.; Lee, J.W.; Gollapudi, R.; Subramaniam, S.; Narsimhadevara, S.; Hurd, D.; Kirikera, G.R.; Shanov, V.; et al. Introduction to carbon nanotube and nanofiber smart materials. Compos. Part B Eng. 2005, 37, 382–394. [CrossRef] Kang, I.; Joung, K.Y.; Choi, G.R.; Schulz, M.J.; Choi, Y.S.; Hwang, S.H.; Ko, H.S. The Bulk Piezoresistive Characteristics of Carbon Nanotube Composites for Strain Sensing of Structures. J. Nanosci. Nanotechnol. 2007, 7, 3736–3739. [CrossRef] [PubMed] Zhang, S.; Zhang, H.; Yao, G.; Liao, F.; Gao, M.; Huang, Z.; Li, K.; Lin, Y. Highly stretchable, sensitive, and flexible strain sensors based on silver nanoparticles/carbon nanotubes composites. J. Alloys Compds. 2015, 652, 48–54. [CrossRef] Sebastian, J.; Schehl, N.; Bouchard, M.; Boehle, M. Health monitoring of structural composites with embedded carbon nanotube coated glass fiber sensors. Carbon 2014, 66, 191–200. [CrossRef] Georgousis, G.; Pandis, C.; Kalamiotis, A.; Georgiopoulos, P.; Kyritsis, A.; Kontou, E.; Pissis, P.; Micusik, M.; Czanikova, K.; Kulicek, J.; et al. Strain sensing in polymer/carbon nanotube composites by electrical resistance measurement. Compos. Part B Eng. 2015, 68, 162–169. [CrossRef]

Sensors 2016, 16, 1171

16. 17. 18. 19.

20. 21. 22. 23. 24.

11 of 11

Egusay, S.; Iwasawaz, N. Piezoelectric paints as one approach to smart structural materials with health-monitoring capabilities. Smart Mater. Struct. 1998, 7, 438–445. [CrossRef] Al-Saffar, Y.; Aldraihem, O.; Baz, A. Smart paint sensor for monitoring structural vibrations. Smart Mater. Struct. 2012, 21, 045004. [CrossRef] Capsal, J.F.; David, C.; Dantras, E.; Lacabanne, C. Piezoelectric sensing coating for real time impact detection and location on aircraft structures. Smart Mater. Struct. 2012, 21, 055021. [CrossRef] Goutham, R.K.; Shinde, V.; Schulz, M.J.; Ghoshal, A.; Sundaresan, M.; Allemang, R. Damage localisation in composite and metallic structures using a structural neural system and simulated acoustic emissions. Mech. Syst. Signal Process. 2007, 21, 280–297. Randall, R.B. Vibration-Based Condition Monitoring; Wiley: West Sussex, UK, 2011; pp. 16–17. Pandey, A.K.; Biswas, M.; Samman, M.M. Damage detection from changes in curvature mode shapes. J. Sound Vib. 1991, 145, 321–332. [CrossRef] Doebling, S.W.; Farrar, C.R.; Prime, M.B. A Review of Damage Identification Methods that Examine Changes in Dynamic Properties. Shock Vib. Dig. 1998, 30, 91–105. [CrossRef] Bauhofer, W.; Kovacs, J.Z. A review and analysis of electrical percolation in carbon nanotube polymer composites. Compos. Sci. Technol. 2009, 69, 1486–1498. [CrossRef] Kordas, K.; Mustonen, T.; Toth, G.; Jantunen, H.; Lajunen, M.; Soldano, C.; Talapatra, S.; Kar, S.; Vajtai, R.; Ajayan, P.M. Inkjet printing of electrically conductive patterns of carbon nanotubes. Small 2006, 2, 1021–1025. [CrossRef] [PubMed] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).