Damage Detection in Nanofiller-Modified ... - Purdue Engineering

Proceedings of the ASME 2018 Conference on Smart Materials, Adaptive Structures and Intelligent Systems SMASIS2018 September 10-12, 2018, San Antonio, TX, USA

SMASIS2018-8008 DAMAGE DETECTION IN NANOFILLER-MODIFIED COMPOSITES WITH EXTERNAL CIRCUITRY VIA RESONANT FREQUENCY SHIFTS T. N. Tallman School of Aeronautics and Astronautics, Purdue University West Lafayette, IN, USA ABSTRACT Conductive nanofiller-modified composites have received a lot of attention from the structural health monitoring (SHM) research community in recent years because these materials are piezoresistive (i.e. they have deformation and damagedependent electrical conductivity) and are therefore self-sensing. To date, the vast majority of work in this area has utilized direct current (DC) interrogation to identify and/or localize damage. While this approach has been met with much success, it is also well known that nanofiller-modified composites possess frequency-dependent electrical behavior. This behavior can be roughly modeled as a parallel resistor-capacitor circuit. However, much less work has been done to explore the potential this frequency-dependent behavior for damage detection. To this end, the work herein presented covers some preliminary results which leverage high-frequency electrical interrogation for damage detection. More specifically, carbon nanofiber (CNF)/epoxy specimens are produced and connected to an external inductor in both series and parallel configurations. Because the CNF/epoxy electrically behaves like a resistorcapacitor circuit, the inclusion of an inductor enables electrical resonance to be achieved. Changes in resonant frequency are then used for rudimentary damage detection. These preliminary results indicate that the potential of SHM via the piezoresistive effect in nanofiller-modified composites can be considerably expanded by leveraging alternating current (AC) interrogation and resonant frequency principles.

state which gives rise to an observed conductivity change via the process of piezoresistive inversion [18-20]. Despite the success of these studies, piezoresistive-based damage detection and SHM have well-known limitations. For example, this approach has poor sensitivity to incipient, earlystage damage, conductivity imaging techniques like EIT have poor spatial resolution and require a burdensome number of electrodes, and the low conductivity of nanocomposites requires large power supplies which may not be practical for in-operation or in-field applications with limited access to power. This first limitation is important particularly for fiber-reinforced composites where early and accurate damage detection and tracking is important to life-cycle management. The poor sensitivity of conductivity-based damage detection was explored to some extent by Tallman et al. [14] who looked at the lower limit of through-hole detection in a carbon black (CB)-modified glass fiber/epoxy laminate via EIT. They found that 1.59 mm diameter hole was not distinguishable from background noise in a 95 mm × 95 mm plate using a 16 electrode EIT system. Likewise, Loyola et al. [7] could not clearly locate a 1.59 mm diameter through hole in a 78 mm × 78 mm glass fiber/epoxy laminate plate with an embedded strain-sensitive multi-wall carbon nanotube (MWCNT)-poly(vinylidene fluoride) (PVDF) film using a 32 electrode EIT system. This poor sensitivity is a consequence of the diffusive nature of current flow. That is, current can easily propagate around very small damage without much difficulty. This ultimately manifests as a very small resistance change particularly in plate-like structures with ample alternative paths for current. It is also important to note that the vast majority of work in piezoresistive nanocomposites has levered direct current (DC) or low-frequency alternating current (AC) excitation. This is important because AC principles could potentially be used to overcome the previously outlined limitations of piezoresistivitybased self-sensing. For example, electrical resonance could be used to increase sensitivity to small damage. Furthermore, work by Hassan et al. [21] has shown that conductivity imaging techniques can be markedly improved through the inclusion of

INTRODUCTION Nanocomposites have received tremendous attention from the structural health monitoring (SHM) research community because they are piezoresistive and therefore self-sensing. To date, much success has been had utilizing the piezoresistive effect to monitor damage initiation and accumulation via resistance change methods [1-3], strain identification [4-6], and in-plane damage localization via conductivity imaging techniques such as electrical impedance tomography (EIT) [717]. Efforts are even being made to discern the precise strain

1 Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 12/13/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Copyright © 2018 ASME

FIGURE 1. Left: simple parallel resistor-capacitor equivalent circuit model for the CNF/epoxy. Right: CNF/epoxy specimen with a 0.125” through hole near its center.

additional, material-derived constraints. This work was presented in the context of inducing known global deformations within a piezoresistive material, but in the context of AC excitation, this could come in the form of phase-shift information. And lastly, nanocomposite impedance decreases considerably at high-frequencies. This frequency dependence could be leveraged to alleviate power requirements. The limited work that does exist in understanding the complex, frequencydependent electrical properties of nanocomposites has shown that the net electrical input-output response of these materials can be approximated as an equivalent circuit such as the simple version shown in Figure 1 or more sophisticated variations thereof [13, 22-24]. Further, Loh et al. [22] studied the effect of deformation by conducting electrical impedance spectroscopy (EIS) measurements during deformation, fitting the EIS data to an equivalent circuit model via simulated annealing, and determining how strain affects various components of the equivalent circuit. In light of the preceding discussion, this manuscript conducts a preliminary investigation into how sensitivity to damage in nanofiller-modified polymer composites can be improved by leveraging AC principles. More specifically, the capacitive nature of a carbon nanofiber (CNF)/epoxy nanocomposite will be exploited by connecting the material in parallel and series with an external inductor. This gives rise to electrical resonance within the nanocomposite-inductor system, and shifts in resonant frequency will be used for rudimentary damage detection. The remainder of this manuscript is organized as follows. First, the process by which two CNF/epoxy specimens were produced is described. Second, the experimental setup for frequency response measurements is detailed. Third, small, medium, and large through-hole damages are induced in the specimens. EIS, impedance magnitude versus interrogation frequency, and phase lag versus interrogation magnitude plots are then generated using an impedance analyzer for each damage state and compared the undamaged data. And lastly, a brief section on summary and conclusions of this preliminary study is presented.

FIGURE 2. Top: series and parallel inductor arrangements used in this study. Bottom: the red box shows the non-ideal equivalent behavior of the inductor. This non-ideal behavior becomes important to the frequency-dependent response of the CNF/epoxy-inductor circuit at high interrogation frequencies.

weight (i.e. the total weight of the combined resin and curing agent). To help facilitate dispersion, a surfactant (Triton X-100) was also added to the CNF/resin mixture at a surfactant-tonanofiller weight ratio of 0.76:1. This combination of CNFs, resin, and surfactant was then mixed in a planetary centrifuge (THINKY AR-100) for 5 minutes. Following this mixing process, the CNF/resin/surfactant mixture was sonicated for 5 minutes (sonication took a total of 10 minutes with the sonicator operating in pulse mode, 1 second on/off pulses) with a horn sonicator (Qsonica Q125) operating at 125 W and 20 kHz. After sonication, the CNF/resin/surfactant mixture was allowed to cool to room temperature. An appropriate amount of curing agent was then added (curing agent-to-resin weight ratio of 27:100). BYK A-501 air-release agent was also added at 0.3% of the final CNF/epoxy weight. The CNF/resin/surfactant/curing agent/airrelease agent was then mixed again in the planetary centrifuge for another 5 minutes, degassed at room temperature for 30 minutes, poured into an open silicone mold, and cured for 5 hours at 85 °C. Following the elevated-temperature cure, the specimen was allowed to sit overnight at room temperature before being removed from the mold. The next day, the specimen was cut into two pieces. Each piece was trimmed to 1” × 1.3” using a water-cooled tile saw with an as-molded thickness of 0.125”. The ends of the specimens were then cleaned with acetone and covered with two coatings of colloidal silver paste (Ted Pella). After allowing the silver paste to dry for several hours, copper tape was applied over the silver paste with additional copper tape securing the specimens on an acrylic substrate as also shown in Figure 1. EXPERIMENTAL PROCEDURE Electrical measurements were collected from both specimens in a damage-free state. Measurements were collected again after drilling a 0.125” diameter hole through the specimen and after boring the initial hole out to 0.25” and to 0.375”. All holes were drilled near the center of the specimen as shown in Figure 1. Resistance measurements were collected with a handheld digital multi-meter (DMM). Impedance and phase angle versus interrogation frequency data was collected using a

SPECIMEN MANUFACTURING To produce the CNF/epoxy specimens, CNFs (Pyrograf III PR-24-XT-HHT) were added to epoxy resin (System 2000 resin and curing agent from Fibre Glast) at 1% of the final epoxy

2 Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 12/13/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Copyright © 2018 ASME

Keysight E4990A Impedance Analyzer sweeping from 20 Hz to 10 MHz. For measurements of the CNF/epoxy specimens in series or parallel with an external inductor, an inductor of 22,000 μH and a DC resistance of 55 Ω was used. A schematic of the inductor arrangements can be seen in Figure 2. EXPERIMENTAL RESULTS AND ANALYSIS Initial resistance measurements of the specimens before damage were made with 𝑅𝑅1 = 499.3 kΩ and 𝑅𝑅2 = 530.1 kΩ where 𝑅𝑅1 and 𝑅𝑅2 designate the resistance of specimen 1 and specimen 2, respectively. EIS plots were generated for both specimens before damage and after damage for all three damage states. These plots can be generated from impedance and phase angle versus interrogation frequency data as shown below in equation (1) where 𝑍𝑍 and 𝜃𝜃 are the impedance and phase angle measured by the impedance analyzer, respectively. The EIS results are presented in Figure 3. For the equivalent circuit model of the CNF/epoxy shown in Figure 1, the expressions for the real and imaginary components of 𝑍𝑍 can be rewritten as shown below in the right-most equality of equation (1) where 𝜔𝜔 is the angular frequency at which the material is being interrogated, 𝑅𝑅𝑠𝑠 is the series resistor, 𝑅𝑅𝑝𝑝 is the parallel resistor, and 𝐶𝐶 is the capacitor in Figure 1. From Figure 3 and leveraging the insights of Loh et al. [22], the introduction of through-hole damage seemingly increases 𝑅𝑅𝑝𝑝 . Re(𝑍𝑍) = |𝑍𝑍| cos(𝜃𝜃) = 𝑅𝑅𝑠𝑠 + Im(𝑍𝑍) = |𝑍𝑍| sin(𝜃𝜃) =

𝑅𝑅𝑝𝑝

2 𝐶𝐶 2 1+𝜔𝜔2 𝑅𝑅𝑝𝑝

2 𝐶𝐶 𝜔𝜔𝑅𝑅𝑝𝑝

2 𝐶𝐶 2 1+𝜔𝜔2 𝑅𝑅𝑝𝑝

(1a)

(1b)

Next, consider the effect of damage on the impedance and phase angle as a function of interrogation frequency. Figure 4 shows the results of this for both specimens in series with an external inductor. The frequency response of the inductor by itself is plotted for reference. Note that the inductor has a selfresonance due to its non-ideal qualities as schematically illustrated in Figure 2. From Figure 4, it can be seen that there is little activity of interest over the majority of the frequency range. Therefore, Figure 5 shows a close up of the resonant behavior between 150 and 450 kHz for both specimens in series with the external inductor. From Figures 4 and 5, it can be seen that damage is clearly impacting the resonant properties of the series CNF/epoxyinductor circuit. However, no clear trend with damage size is evident (i.e. resonant frequency does not consistently increase or decrease with increasing damage size). In the absence of any trend for the series configuration, consider next the parallel CNF/epoxy-inductor setup as shown in Figure 2. As before, the net impedance and phase angle of the CNF/epoxy-inductor circuit is plotted against interrogation frequency. This data is collected for the undamaged state and after each new throughhole damage for both specimens. The frequency range again ranges from 20 Hz to 10 MHz. The results of this analysis are shown in Figures 6 and 7.

FIGURE 3. EIS plots showing the real and imaginary components of the impedance of the CNF/epoxy specimens as a function interrogation frequency. Note these EIS plots do not include the external inductor.

Observing Figures 6 and 7, a clear trend can be seen between damage size and resonant behavior. More specifically, as the damage size increases, the impedance peaks shift increasingly to the right. To investigate the potential of these resonant frequency shifts as a damage detection metric, the percent change in resonant frequency was plotted against through-hole diameter. For comparison, the resistance change ratio is also shown as a function of through-hole diameter. This plot is shown in Figure 8. Due to the limited data points near the peak impedance values in Figure 7, resonant frequencies were approximated by determining where the phase angle is equal to zero for each damage state. This was done by finding the data points immediately greater than and less than zero for each damage state. These two points were then used to fit a linear curve such that linear interpolation could be used between these points to approximate the frequency at which the phase angle of the CNF/epoxy-inductor system is zero, 𝜃𝜃 = 0.

3 Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 12/13/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Copyright © 2018 ASME

FIGURE 4. Impedance and phase angle of CNF/epoxy in series with an external inductor. Top plots are for specimen 1 and bottom plots are for specimen 2. The frequency response of the inductor by itself is included for reference. Note these plots show the entire frequency range from 20 Hz to 10 MHz.

FIGURE 6. Impedance and phase angle of CNF/epoxy in parallel with inductor. Top plots are for specimen 1 and bottom plots are for specimen 2. The frequency response of the inductor by itself is included for reference. Note these plots show the entire frequency range from 20 Hz to 10 MHz.

FIGURE 5. Close up of impedance and phase angle of CNF/epoxy in series with an inductor. Here, it can be seen that damage is clearly impacting the resonant behavior of the circuit. However, no clear correlation between damage and frequency-shift is evident.

FIGURE 7. Close up of impedance and phase angle of CNF/epoxy in parallel with an inductor. Here, it can be seen that damage is again impacting the resonant behavior of the circuit. However, unlike the series configuration, a clear trend between phase angle and damage size is observable for the parallel CNF/epoxyinductor configuration.

From Figure 8, it can be seen that the percent change in resonant frequency is greater than the percent change in resistance for any given damage state in both specimens. This increase is relatively modest for the case of a 0.125” hole in specimen 1 and quite appreciable for the 0.375” hole in specimen 2. Despite this variation, there is nonetheless a consistent improvement. Therefore, these preliminary results seem to indicate that the sensitivity to small damage of nanofillermodified composites can be improved by leveraging resonant principles and high-frequency AC interrogation.

resonant frequencies were used to detect damage in a CNFmodified epoxy. To achieve resonance, the CNF/epoxy nanocomposite was connected in series and in parallel with an external inductor. This was done in light of the known capacitive behavior of nanocomposites. The series CNF/epoxy-inductor configuration did not show any discernable correlation between damage size and resonant frequency shift. However, the parallel CNF/epoxy-inductor circuit did show a clear trend between damage size and resonant frequency shift with larger through-hole damage causing larger frequency shifts. Further, the percent change in resonant frequency was appreciably larger than the percent change in DC resistance for any given damage state in both specimens. In light of these preliminary findings, it appears as though AC principles such as resonance do indeed have potential to address the well-known limitations of piezoresistive

SUMMARY AND CONCLUSIONS This manuscript has presented a very preliminary investigation into how AC principles can be used to bolster the potential of piezoresistive nanocomposite-based damage detection and SHM. More specifically, shifts in electrical

4 Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 12/13/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Copyright © 2018 ASME

[4] I. Kang, M. J. Schulz, J. H. Kim, V. Shanov and D. Shi, "A carbon nanotube strain sensor for structural health monitoring," Smart Materials and Structures, vol. 15, pp. 737-748, 2006. [5] K. J. Loh, J. P. Lynch and N. Kotov, "Inductively coupled nanocomposite wireless strain and pH sensors," Smart Structures and Systems, vol. 4, pp. 531-548, 2008. [6] K. J. Loh, J. Kim, J. P. Lynch, N. W. S. Kam and N. A. Kotov, "Multifunctional layer-by-layer carbon nanotubepolyelectrolyte thin films for strain and corrosion sensing," Smart Materials and Structures, vol. 16, pp. 429-438, 2007. [7] B. R. Loyola, T. M. Briggs, L. Arronche, K. J. Loh, V. La Saponara, G. O'Bryan and J. L. Skinner, "Detection of spatially distributed damage in fiber-reinforced polymer composites," Structural Health Monitoring, vol. 12, pp. 225-239, 2013. [8] B. R. Loyola, V. La Saponara, K. J. Loh, T. M. Briggs, G. O'Bryan and J. L. Skinner, "Spatial sensing using electrical impedance tomography," IEEE Sensors, vol. 13, pp. 23572367, 2013. [9] H. Dai, G. J. Gallo, T. Schumacher and E. T. Thostenson, "A novel methodology for spatial damage detection and imaging using a distributed carbon nanotube-based composite sensor combined with electrical impedance tomography," Journal of Nondestructive Evaluation, vol. 35, 2016. [10] G. J. Gallo and E. T. Thostenson, "Spatial damage detection in electrically anisotropic fiber-reinforced composites using carbon nanotube networks," Composite Structures, 2015. [11] T. N. Tallman and J. A. Hernandez, "The effect of error and regularization norms on strain and damage identification via electrical impedance tomography in piezoresistive nanocomposites," Nondestructive Testing & Evaluation International, vol. 91, pp. 156-163, 2017. [12] T. N. Tallman and K. W. Wang, "Damage and strain identification in multifunctional materials via electrical impedance tomography with constrained sine wave solutions," Structural Health Monitoring, vol. 15, pp. 235244, 2016. [13] T. N. Tallman, Gungor, W. K. W. S and C. E. Bakis, "Damage detection and conductivity evolution in carbon nanofiber epoxy via electrical impedance tomography," Smart Materials and Structures, vol. 23, p. 045034, 2014. [14] T. N. Tallman, S. Gungor, K. W. Wang and C. E. Bakis, "Damage detection via electrical impedance tomography in glass fiber/epoxy laminates with carbon black filler," Structural Health Monitoring, vol. 14, pp. 100-109, 2014. [15] T. N. Tallman, S. Gungor, K. W. Wang and C. E. Bakis, "Tactile imaging and distributed strain sensing in highly flexible carbon nanofiber/polyurethane nanocomposites," Carbon, vol. 95, pp. 485-493, 2015.

FIGURE 8. Percent change in DC resistance and resonant frequency plotted against through-hole diameter for both specimens in parallel with the external inductor.

nanocomposite-based SHM. At the very least, these results indicate that the poor sensitivity of conductivity changes to small damage can be improved. However, there are several puzzling aspects of these preliminary results. First, it is not clear why there was no apparent trend between damage size and frequency shift for the CNF/epoxy in series with the external inductor. And second, specimen 2 exhibited a pronouncedly greater frequency shift than specimen 1 despite both specimens coming from the same material and enduring the same damage. While this could be brusquely attributed to the non-uniformity of the nanofiller dispersion, even that supposition belies a more fundamental question: that is, it is not clear how exactly the complex, frequency-dependent electrical properties of nanocomposites depend on micro-scale nanofiller networking. Therefore, future work should seek to elucidate a more fundamental understanding of nanocomposite impedivity and piezoimpedivity than is available through equivalent circuit models. REFERENCES [1] E. T. Thostenson and T. W. Chou, "Carbon nanotube networks: sensing of distributed strain and damage for life prediction and self healing," Advanced Materials, vol. 18, pp. 2837-2841, 2006. [2] E. T. Thostenson and T. W. Chou, "Real-time in situ sensing of damage evolution in advanced fiber composites using carbon nanotube networks," Nanotechnology, vol. 19, p. 215713, 2008. [3] J. J. Ku-Herrera, V. L. La Saponara and F. Aviles, "Selective damage sensing in multiscale hierarchical composites by tailoring the location of carbon nanotubes," Journal of Intelligent Material Systems and Structures, 2017.

5 Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 12/13/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Copyright © 2018 ASME

[16] T. N. Tallman, F. Semperlotti and K. W. Wang, "Enhanced delamination detection in multifunctional composites through nanofiller tailoring," Journal of Intelligent Material Systems and Structures, vol. 26, p. 2565–2576, 2015. [17] Y. Zhao, S. Gschossmann and M. Schager, "Observing the fracture behavior of a center crack via electrical impedance tomography using inkjet-printed carbon nanotube thin films," in International Workshop on Structural Health Monitoring, Stanford, 2017. [18] T. N. Tallman and K. W. Wang, "An Inverse Methodology for Calculating Strains from Conductivity Changes in Piezoresistive Nanocomposites," Smart Materials and Structures, vol. 25, p. 115046, 2016. [19] T. N. Tallman, S. Gungor, G. M. Koo and C. E. Bakis, "On the inverse determination of displacements, strains, and stresses in a carbon nanofiber/polyurethane nanocomposite from conductivity data obtained via electrical impedance tomography," Journal of Intelligent Material Systems and Structures, pp. 1-13, 2017. [20] H. Hassan and T. N. Tallman, "Predicting failure from conductivity changes in piezoresistive nanocomposites," in Proc. SPIE 10598, Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems, Denver, 2018. [21] H. Hassan, F. Semperlotti, K. W. Wang and T. N. Tallman, "Enhanced imaging of piezoresistive nanocomposites through the incorporation of nonlocal conductivity changes in electrical impedance tomography," Journal of Intelligent Material Systems and Structures, 2018. [22] K. J. Loh, J. P. Lynch, B. S. Shim and N. A. Kotov, "Tailoring Piezoresistive Sensitivity of Multilayer Carbon Nanotube Composite Strain Sensors," Journal of Intelligent Material Sytems and Structures, vol. 19, pp. 747-764, 2008. [23] C. Zhao, W. Yuan, H. Liu, B. Gu, N. Hu, A. Y. Ning and F. Jia, "Equivalent circuit model for the strain sensing characteristics of multi-walled carbon nanotube/polyvinylidene fluoride films in alternating current circuit," Carbon, vol. 129, pp. 585-591, 2018. [24] A. A. Sahreai, M. Ayati, M. Baniassadi, D. Rodrigue, M. Baghani and Y. Abdi, "AC and DC electrical behavior of MWCNT/epoxy nanocomposite near percolation threshold: Equivalent circuits and percolation limits," Journal of Applied Physics, vol. 123, p. 105109, 2018.

6 Downloaded From: https://proceedings.asmedigitalcollection.asme.org on 12/13/2018 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Copyright © 2018 ASME