- Email: [email protected]
Piezoresistive in-situ strain sensing of composite laminate structures
Composites: Part B 69 (2015) 534–541
Contents lists available at ScienceDirect
Composites: Part B journal homepage: www.elsevier.com/locate/compositesb
Piezoresistive in-situ strain sensing of composite laminate structures Michael C. Koecher, John H. Pande, Scott Merkley, Sam Henderson, David T. Fullwood ⇑, Anton E. Bowden Brigham Young University, Provo, UT 84602, USA
a r t i c l e
i n f o
Article history: Received 30 October 2013 Received in revised form 23 September 2014 Accepted 24 September 2014 Available online 12 October 2014 Keywords: A. Carbon fiber A. Smart materials A. Nano-structures B. Electrical properties
a b s t r a c t Various methods have been developed to monitor the health and strain state of carbon fiber reinforced polymers, each with a unique set of pros and cons. This research assesses the use of piezoresistive sensors for in situ strain measurement of carbon fiber and other composite structures in multidirectional laminates. The piezoresistive sensor material and the embedded circuitry are both evaluated. For the piezoresistive sensor, a conductive nickel nanocomposite sensor is compared with the piezoresistivity of the carbon fiber itself. For the circuit, the use of carbon fibers already present in the structure is compared with the use of nickel coated carbon fiber. Successful localized strain sensing is demonstrated for several sensor and circuitry configurations. Numerous engineering applications are possible in the ever-growing field of carbon-composites. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Carbon fiber reinforced polymers (CFRP) possess a unique set of material properties that include high strength-to-weight ratio, low thermal expansion, and good fatigue characteristics; these benefits have led to rapid expansion in numerous areas of engineering [1–3]. Given the various loading conditions and environments that this relatively new material has to endure, there is a pressing need for a wide range of structural health and condition monitoring solutions. A monitoring solution that includes instantaneous strain sensing could provide useful feedback for control, actuation and data logging functions. This paper discusses the development of piezoresistive sensing solutions for carbon fiber structures. Of particular interest is the embedded circuitry that enables the sensing of local strain in a component via connection to an external resistance meter. The two in-situ piezoresistive sensors that are evaluated are 1. embedded nickel nanostrand (NiN) nanocomposites and 2. neat prepreg carbon fiber. For the connecting circuitry, nickel coated carbon fibers and carbon fiber prepreg alone are compared as pseudowires to the piezoresistive sensor; the probing configuration of the external meter is also considered. Envisaged applications include strain and health monitoring of structural members, as well as the sensing of deflection in compliant components such as sports equipment and actuation devices. For control and data logging purposes. Actual engineering components might include ⇑ Corresponding author at: 435 CTB, Brigham Young University, Provo, UT 84602, USA. Tel.: +1 801 422 6316. E-mail address: [email protected] (D.T. Fullwood). http://dx.doi.org/10.1016/j.compositesb.2014.09.029 1359-8368/Ó 2014 Elsevier Ltd. All rights reserved.
bridge decks, composite beams, wind turbine blades, golf clubs, compliant mechanisms, etc. The reported examples of a tensile specimen, a beam under bending, and a carbon-fiber pressure vessel, illustrate uses of the methodology that could be adapted for various engineering applications.
1.1. Current strain sensing methods Various methods exist for measuring strain in carbon fiber structures. Fiber Bragg gratings, using embedded fiber optics measure strain [4,5], and other exotic and novel methods (e.g. [6–8]) are making tremendous progress towards providing a suite of solutions that will no longer be confined to expensive laboratory situations. More traditional methods include strain gauges composed of thin metal films [9]. The advantages of these types of gauges are that they are fairly simple to install, low cost, and have proved to be successful through years of use in industry. However, metal foils also have a very limited strain range, and because these gauges are adhered to the surface they are susceptible to damage. Furthermore, wires must be routed across or though the structure to carry the required signal to the monitoring unit. Carbon fiber is itself piezoresistive by nature, and can thus act as a strain gauge [10–12]. However, transverse plies in a multidirectional laminate largely short circuit the piezoresistive response of the carbon fiber structure [13]. The issue of routing circuitry to the point of interest must also be considered for this approach. Carbon nanotubes [14,15], carbon black [16], and nickel nanostrands [17]. This paper will focus on nickel nanostrand composites due
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
to their extremely large piezoresistive effect. Johnson et al. discovered that directly embedding a nickel nanostrand patch into unidirectional carbon fiber laminates yielded a significant piezoresistive response [18,19]. However, when the same method was applied to multidirectional laminates the signal quickly short-circuited and no piezoresistive effect was found. To measure piezoresistivity in multidirectional CFRP layups, a nickel nanostrand nanocomposite was insulated in fiberglass to resolve the short-circuiting issue, and embedded between the layers of carbon fiber. This method not only proved successful for measuring strain but also for detecting damage to the structure. However, due to the large stress concentrations induced by the embedded insulated patch, the strength of the composites was severely compromised. The current paper will evaluate methods to measure strain in multidirectional carbon fiber laminates without significantly altering the strength of the carbon fiber structure. Piezoresistive sensing material will compose of either the carbon fiber structure itself, or local regions of nano-composite with nickel nanostrands added to the matrix. The signal path to the monitoring system will be generated using only the carbon fiber or using embedded nickel coated carbon fibers (NCCF). These coated fibers are at least three times more conductive than bare carbon fiber [20] thus reducing the amount of noise in the piezoresistive signal. The optimized configuration of probes for detecting the composite strain will be evaluated, and the resultant methodology will be tested in an engineering situation: a cylindrical carbon fiber pressure vessel. 2. Materials and methods 2.1. Materials Nickel nanostrands (NiNs) display a high strand aspect ratio and a bifurcated structure that allows conductivity of a nanocomposite to be obtained at small volume fractions of nanostrands [21]. When combined with a polymer matrix they form a piezo-resistive material that has been used as the basis for strain sensors [22]. The piezoresistivity of such nanocomposites has been most successfully modeled using percolation theory and a quantum tunneling mechanism [23,24]. With a sufficiently large volume fraction of conductive filler material nanojunctions will form between filler particles in which electrons can tunnel through the insulating barrier. Once this occurs an electrical network is created through the nanocomposite and it becomes conductive. The resistivity of a nanojunction can be calculated using Eq. (1):
h
2
q ¼ 2 pffiffiffiffiffiffiffiffiffiffi exp e
2mk
pffiffiffiffiffiffiffiffiffiffi ! 4p 2mk s h
ð1Þ
where h is the Planck constant (J s), e is the electron charge (C), m is the electron mass (kg), k is the tunneling barrier height (J), and s is
535
the distance between particles (m) [17]. Thus the resistivity of a nanojunction is a function of the inter-particle distance, s. As a nanocomposite is strained s will change and a piezoresistive signal is obtained. Another novel material, nickel coated carbon fiber (NCCF), is tested as the basis for the sensor circuitry in this paper. Carbon fiber receives a uniform coating of nickel through a chemical vapor deposition process. These coated fibers are much more conductive than bare carbon fiber and are often used to aid in electrical shielding in carbon fiber structures [25]. Both the nickel nanostrands and the nickel coated carbon fiber were produced and provided by Conductive Composites Company (Heber, Utah). The structure of these materials can be seen in the SEM images depicted in Fig. 1. Along with these novel materials, a traditional unidirectional carbon fiber prepreg is used as the structural material that is to be monitored: ZR6-P35, provided by Zoltek. Components with various pre-preg layups were created to determine the capability and limitations of the embedded conductive materials for in situ strain sensing. For initial sensor evaluation, laminates were formed in configurations suitable for tensile testing. Each carbon fiber laminate was composed of layers of prepreg cut to 250 mm 25 mm. Woven fiber glass tabs were adhered to the ends of the carbon fiber samples to insulate the sample from the metal grips of the tensile tester and ensure that these did not interfere with the signal. Each sample consisted of a sensor patch location and a sensor patch material. The patch location refers to the area that strain is to be measured and the patch material was either prepreg embedded with NiNs or simply the carbon fiber prepreg itself. To limit variation and guarantee uniformity in the samples with embedded nanostrands, 0.02 g of nickel nanostrands were filtered through a 60 mesh screen (250 lm) before being placed in a patch area of 19 mm by 12 mm. For carbon fiber and NiN patches the [0] direction (prepreg oriented parallel to the applied strain) and [90] direction (prepreg oriented transverse to the applied strain) were evaluated as the sample was strained in the direction of its length. As mentioned previously, the use of NCCF was assessed as a means of forming the sensor circuit. The alternative to NCCF was to allow the signal to carry along the carbon fibers in the prepreg itself. It was hypothesized that the NCCF would better control the signal flow and limit scatter and short-circuiting across the cross plies of prepreg. Thus nickel coated carbon fiber was embedded onto the surface of various samples and compared to control samples with no embedded materials. Fig. 2 is an example of a sample with embedded nanostrands and nickel coated carbon fiber. As can be seen in Fig. 2 a small gap was placed between NCCF bundles at the patch location. The gap in the NCCF was required to ensure the signal traveled through the carbon fiber or NiN patch at the patch location. Table 1 shows the various sample configurations tested. Also included in Table 1 are signal-to-noise ratio
Fig. 1. SEM images (from left to right) of nickel nanostrands and nickel coated carbon fiber (NCCF used courtesy of Nathan Hansen, Conductive Composites Company).
536
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 2. [90,0,0,90] Laminate with embedded NiNs and NCCF (gray lines). A small piece of carbon fiber is placed over the NiN patch to prevent inadvertent dispersion of the NiNs during curing.
Table 1 Tensile sample configurations that were tested and their accompanying signal-tonoise ratio (SNR) results. Sample
Material
Patch
SNR
Standard deviation
A B C D E F
[0,90]s Carbon fiber [0,90]s Carbon fiber [90,0]s Carbon fiber [90,0]s Carbon fiber Woven fiberglass Woven fiberglass
Carbon fiber [0] NiN Carbon fiber [90] NiN Carbon fiber [90] [90] and NiN
0.86 2.16 17.90 4.32 14.50 80.54
0.28 0.21 1.44 0.05 0.90 1.93
(SNR) results which will be discussed in the results section. Fiberglass samples were also prepared and tested in order to test a piezoresistive patch (of either carbon fiber/NiNs or carbon fiber alone) in non-conductive structures. Patches consisting of nanostrands and carbon fiber were embedded in the fiberglass structure to compare the piezoresistivity of the carbon fiber and the NiN patch. Four samples were made for each configuration and typically five measurements were taken on each sample (as described in Section 2.2) to produce the statistics reported later.
significant as compared to the resistance of the component [26]. In this method a known current is passed through the outer probes and a voltage is read between the inner probes. Since no current is passing through the inner probes there is no contact resistance. A National Instruments NI 9219 multifunctional module was set up for four probe resistance readings for these tests. A constant current of 500 lA was applied to the outer probes while a voltage was read across the inner probes. Using Ohms law an output of resistance was obtained. Three different configurations of probing to measure the piezoresistive change at the patch location were evaluated in this research. In the first method, referred to as the ‘‘collinear’’ probing method, four collinear probes were placed on the carbon fiber sample to eliminate contact resistance. For the second method the probes were placed in a ‘‘box’’ configuration. The box configuration is similar to the collinear method except there are two pseudowires instead of four and each wire has two probes on it. In the final method, referred to as ‘‘longitudinal’’ probe method, the probes were place on each side of the patch. In each method the pseudo-wires consist of either the conductive carbon fiber prepreg or the embedded NCCF. Fig. 3 depicts the difference between the three probing methods. Samples were cyclically loaded in tension in an Instron tensile tester to a prescribed strain of 0.30% at a rate of 0.1 Hz in order to determine the feasibility of each configuration as in situ strain sensing for multidirectional carbon fiber laminates. Three point bending tests were also performed on select samples using supports 12.5 cm apart with a 10 mm vertical displacement of the wedge to determine the piezoresistive changes due to tension and compression. During loading the resistance measurements were sampled at a rate of 100 Hz and a moving average of 10 was used to improve signal smoothness.
3. Results and discussion 3.1. Tensile tests
2.2. Method The four probe resistance measurement method was used to measure the piezoresistive signal at a point remote from the desired strain measurement; this is a common method to eliminate contact resistance readings when contact resistance is
3.1.1. Nickel coated carbon fiber To determine the feasibility of using NCCF as pseudo-wires to the piezoresistive patch it was first necessary to determine if piezoresistive properties exist in NCCF. If the NCCF display a piezoresistive response, this will confuse the signal expected from the
Fig. 3. (a) Transverse collinear probe method, (b) box probe method, and (c) longitudinal probe method. In each method the gray lines represent the circuit path using either the carbon fiber strands from the structure or embedded NCCF. EX+ and EX are the input and output current probes while HI and LO are the voltage reading probes.
537
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 4. Comparison of the piezoresistivity of carbon fiber to the more conductive NCCF.
sensor. Tensile tests were performed on carbon fiber and NCCF and compared in Fig. 4. It can be seen that the NCCF is much more conductive than carbon fiber and does not exhibit piezoresistive properties. This suggested that using NCCF as pseudo-wires to the patch does not contribute to the piezoresistive signal, and one can be confident that the strain being measured is precisely located at the patch location. 3.1.2. Circuit and probing methods A wide variety of tests were performed on samples with various probing configurations (see Fig. 3) with results reported in Table 2. The signal-to-noise ratio (SNR) was calculated and used as a metric in the laboratory to quantitatively compare the piezoresistive signal between samples. A larger SNR correlated to a higher quality piezoresistive signal. In general, it was observed that a SNR value lower than about 0.30 suggested that there is no distinguishable signal that can be correlated to strain; an ideal SNR for non-laboratory situations has not been determined. Comparing Tests 3, 4, and 9 to Tests 5, 6, and 8, respectively, it can be seen that using embedded NCCF as pseudo-wires allow for a stronger piezoresistive signal at greater distances than just using carbon fiber prepreg. It can also be seen in Table 2 that the longitudinal probing method allowed for remote probing of the piezoresistive sensor at greater distances. It is of note that the longitudinal probing method also yielded more consistent repeatable results. Thus, each subsequent sample was tested using the longitudinal probing method. 3.1.3. Sensor tests Each sample in Table 1 was subsequently probed using the longitudinal method. Fig. 5 shows the results obtained from tensile tests performed on samples A and C. Each of these samples consisted of a carbon fiber patch with different carbon fiber orientations. In Fig. 5a the carbon fiber on the surface and NCCF are
Fig. 5. Comparison of piezoresistive signals obtained from (a) sample A and (b) sample C from Table 1.
oriented in the [0] direction. In Fig. 5b the carbon fibers on the surface are oriented in the [90] direction while the NCCF are oriented in the [0] direction. Thus it can be seen that when the carbon fiber is oriented parallel to the NCCF the piezoresistive signal does not have a strong correlation to the strain, but when carbon fibers and NCCF are oriented transverse to one another the piezoresistivity relates well to strain. It is further confirmed in Fig. 6 that when the carbon fiber and NCCF are oriented parallel to one another no correlation between the piezoresistive signal and strain is found.
Table 2 SNR calculations for various samples with different probing methods. Test number
Orientation
NiN patch
NCCF
Probe method
Distance (cm)
SNR
Standard deviation
1 2 3 4 5 6
[0,90,0,90]
No
No
Yes
No
Parallel collinear Transverse collinear Box
NA NA 2.5 4 2.5 7.5
0.0518 0.8767 3.0018 0.1454 6.2408 0.4556
0.043 0.099 1.202 0.022 2.448 0.049
7 8 9
[90,0,0,90]
Longitudinal
4 14 14
0.5099 0.4818 0.2534
0.196 0.068 0.032
Yes Yes
Yes No
538
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 7. Comparison of fiberglass samples (a) E and (b) F.
Fig. 6. Comparison of piezoresistive signals obtained from (a) sample B and (b) sample D.
From Fig. 5 it has been shown that the piezoresistivity of the neat carbon fiber prepreg can be exploited to determine the percent strain in a multidirectional laminate carbon fiber structure. Also, NCCF can be used as pseudo-wires to a patch location thus allowing strain measurement at a specific location while remotely probing from the patch location. Sample B and sample D are similar to sample A and sample C except a NiN patch is included rather than using the neat carbon fiber as a patch. These samples were tested to determine how a NiN patch affects the piezoresistive signal (which has resulted in good results for other authors for long-range strain gauges) [27]. The results from Fig. 6 yield similar results to those in Fig. 5. This suggests that using a NiN patch does nothing to improve upon the piezoresistive signal. Thus the relative orientation of carbon fiber in the laminate and the use of NCCF has more of an effect than using a NiN patch. To further investigate the effect of the NiNs on the piezoresistive signal samples E and F were tested and the results can be seen in Fig. 7. Each sample has a piezoresistive signal that strongly correlates to the strain which again shows that a NiN patch does little to affect the piezoresistive signal. Table 1 shows a calculation for the signal-to-noise (SNR) of each sample. It is of interest to note that the signals obtained in Fig. 7 have less noise than those in Figs. 5 and 6. This can be attributed to the fact that even though the NCCF is much more conductive than carbon fiber, the signal will slightly scatter across the carbon fiber thus adding noise to
the overall signal in the carbon-fiber laminate samples. In samples E and F the fiberglass is completely insulating and there is minimal scatter in the signal and the amount of noise is dramatically decreased. 3.2. Bending tests A [90,0,0,90]s carbon fiber laminate with NCCF embedded on the top and bottom surfaces was placed in bending to evaluate the variation of piezoresistivity due to compressive and tensile strain across a single component. From Fig. 8a, increasing the displacement of the wedge caused a compressive strain on the top surface and a decrease in resistance was obtained. The opposite is true on the bottom surface; the tensile strain increased with displacement and an increase in resistance was obtained. This test demonstrates the ability to measure varying strain at different points in a sample using the NCCF circuits to pin-point the measurement site. The three point bending test was also conducted on fiberglass Sample E from Table 1. Similar results were obtained and can be seen in Fig. 9. The fiberglass sample once again yielded a higher signal-to-noise ratio of 24.76 as compared to the 6.79 signal-tonoise ratio of the carbon fiber, which further suggests that there is some signal loss due to the conductive nature of the carbon fiber. 3.3. Failure tests The results so far demonstrate that it is feasible to use NCCF to sense strain at a remote distance from a probing location in a sample in either tension or compression. For this to be a useful method
539
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541 Table 3 Failure tests results.
Failure stress (MPa) Standard deviation Strain to failure (%) Standard deviation
Embedded NiNs/NCCF
No NiNs/NCCF
Total
9578 364 1.75 0.42
9445 1179 2.01 0.59
9511 875 1.88 0.53
ASTM D3039 standard. These results were compared with samples of the same layer orientation without embedded NiNs or NCCF. Table 3 shows stress to failure for various samples with and without embedded NiNs and NCCF. Embedding NiNs and NCCF in the middle or top layer of a carbon fiber laminate showed no significant effect on the failure stress. Thus it appears that the embedded NCCF and NiNs do not create significant stress concentrations in the bulk structure, and these materials can be used without compromising the integrity of the carbon fiber structure. 3.4. Application
Fig. 8. Three point bending tests on a [90,0,0,90]s carbon fiber laminate with NCCF embedded on the (a) top surface and (b) bottom surface.
As a proof of concept a cylindrical pressure vessel was fabricated to show the feasibility and advantages of using nickel coated carbon fiber as pseudo-wires to allow remote sensing of piezoresistive carbon fiber patches. Three layers of roll-wrapped carbon fiber prepreg were laid around a cylindrical 15 cm diameter mandrel. The two interior layers were wrapped with the fibers aligned in the longitudinal (axis) direction while the outer layer was aligned in the hoop direction. Bundles of nickel coated carbon fiber were placed in the longitudinal direction on the outer surface and in the hoop direction on the interior of the pressure (see Fig. 10). A second set of longitudinal NCCF bundles were embedded on the opposite side for redundancy in the sensor setup. This arrangement ensured that at each sensor patch location the carbon fibers and NCCF were oriented transverse one relative to another to enable the best possible piezoresistive signal to be obtained as found in Figs. 5 and 6. Once cured, caps were adhered to both ends to seal the pressure vessel. The pressure vessel was pressurized using distilled water to prevent electrical short circuiting and alleviate safety concerns with pressurized air. The vessel was pressurized using an Ametek M&G T-65 twin seal pressure pump while resistance measurements were simultaneously being recorded. Results for longitudinal strain on the outside surface can be seen in Fig. 11. As the pressure increased the longitudinal strain increased and an increase in the resistance was obtained. The pressure vessel was rapidly depressurized and accordingly the resistance dropped. Thus, as expected, with a tensile strain in the longitudinal direction the piezoresistive signal is positive.
Fig. 9. Three point bending test results from Sample E.
to measure strain in a carbon fiber structure it is important to know how embedding this NCCF material affects the strength of the carbon fiber structure. Samples were embedded with NCCF and NiNs in various positions, and loaded to failure according the
Fig. 10. Pressure vessel design (gray lines represent NCCF bundles). The gap between each NCCF bundle is transverse to the carbon fiber thus allowing for piezoresistive readings.
540
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 11. Longitudinal strain measurement in pressure vessel. The vessel was pressurized to 552 kPa and then rapidly depressurized starting at approximately 12 s.
Fig. 14. Extreme piezoresistive changes when failure occurs (gray region) suggest that failure can be detected in a carbon fiber structure.
longitudinal strain creates a positive piezoresistive signal while measuring hoop strain yields negative piezoresistivity. Further loading the pressure vessel above 827 kPa caused the vessel to start failing. Resistance measurements were taken on both longitudinal bundles and the hoop bundle as the pressure vessel began to fail. In Fig. 14 it can be seen as the fibers began to delaminate and break large resistance changes were detected. It is evident that this method of strain measurement could be used to detect failures within a carbon fiber component. The different amplitudes of resistance changes between the three locations could suggest the location of where failure began. 4. Conclusions Fig. 12. Strain reading on inside surface of pressure vessel. Pressure reached a max of 827 kPa at approximately 17 s.
The results measuring the hoop strain on the interior surface can be seen in Fig. 12. It was assumed that as the pressure increased there would be a tensile strain in the hoop direction and thus a positive piezoresistive signal would be obtained, yet the results show a negative piezoresistive signal with increasing stress. These results are being attributed to the compression of the longitudinal carbon fibers in the radial direction. As the pressure increased the carbon fibers were compressed into intimate contact with one another creating a lower resistance (see Fig. 13). Unfortunately, the data recording was interrupted before complete depressurization and subsequently the vessel was damaged so a return of the resistance reading to its unpressurized state was not seen. Yet, the results still indicate a negative piezoresistive signal in the hoop direction. From the resistance measurements obtained it is evident that using NCCF as pseudo-wires to a patch location can facilitate remote strain measurements in a carbon fiber structure. Measuring
Fig. 13. Pressure causes the fibers to compress into one another thus lowering the resistance in the hoop direction.
Embedded nickel nanostrands and nickel coated carbon fiber were evaluated as tools for in-situ strain sensing in a multidirectional carbon fiber laminate. Tests were conducted via altering of the strain sensor patch and the connecting circuitry. Previous research has used the piezoresistivity of directly embedded nickel nanostrand nanocomposites to measure strain in unidirectional layup. Testing a variety of strain sensor configurations has shown that directly embedding nickel nanostrands into a multidirectional laminate is ineffective for strain measurement and does little to improve upon the piezoresistivity of the carbon fiber. Carbon fiber’s piezoresistivity is a good measure of strain, and thus a valid strain sensor, but if the carbon fiber is also used as the current carrier to a probe location, the probing location must be extremely close to the area where strain is to be measured to allow a discernible electrical signal to be obtained. The more conductive nickel coated carbon fiber does not have the same piezoresistive properties as carbon fiber and hence is a better choice for supplying a circuit from the sensor area to the probe location. Using NCCF as the connecting circuitry provides pseudo-wires to remotely measure the strain at a desired location without piezoresistivity in the circuit affecting the signal. It has been shown that the best results are obtained when the NCCF circuit is oriented perpendicular to the carbon fibers that are in the sensor region. Furthermore the optimal probe geometry involves the four-probe longitudinal method. Future work will be performed to determine if embedding the NCCF circuit between layers of a laminate can accurately determine strain from arbitrary positions within a sample. Samples embedded with nickel coated carbon fiber and nickel nanostrands were loaded to failure in tension to determine the adverse effects these materials have on the strength of the carbon fiber. It has been shown that the embedded materials do not
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
significantly alter the strength of the structure. Thus, using embedded NCCF is a safe, easy way to remotely measure strain within a carbon fiber structure without compromising structural integrity. In addition to tensile and bending tests of laminates, a pressure vessel was built as a proof of concept application of remote strain sensing using embedded nickel coated carbon fiber for the connecting circuit. When pressurized, longitudinal and hoop strain were detected. Longitudinal strain yielded positive piezoresistive results while hoop strain yielded negative piezoresistive results. It was also found that when the pressure vessel began to fail, the failure was detectable through an extreme change in the resistance reading. The tests illustrate the potential for applying such an approach in a wide range of engineering applications for health monitoring, strain sensing and control of composite components. The distance between the location of measured strain and the probe location is modest in the tests (up to 14 cm); the ability to place the probes at much greater distances from the location of interest is desirable, and may be increased with further optimization of the method. However, for many applications a surface probe at similar distances from the strained location may be acceptable. Acknowledgement This work was supported, in part, by a National Science Foundation Grant, Contract Number: CMMI-1235365. References [1] Gibson RF. Principles of composite materials mechanics. 2 ed. Boca Raton: CRC Press; 2007. [2] Strong AB. Fundamentals of composites manufacturing: materials, methods, and applications. 2 ed. Dearborn: Society of Manufacturing Engineers; 2008. [3] Roberts T. The Carbon Fibre Industry Worldwide 2008–2014. MTP; 2009. [4] Kim H-S, Yoo S-H, Chang S-H. In situ monitoring of the strain evolution and curing reaction of composite laminates to reduce the thermal residual stress using FBG sensor and dielectrometry. Compos B Eng 2013;44(1):446–52. [5] Ling H-Y, Lau K-T, Su Z, Wong ET-T. Monitoring mode II fracture behaviour of composite laminates using embedded fiber-optic sensors. Compos B Eng 2007;38(4):488–97. [6] Fu T, Liu Y, Lau K-T, Leng J. Impact source identification in a carbon fiber reinforced polymer plate by using embedded fiber optic acoustic emission sensors. Compos B Eng 2014;66:420–9. [7] Habib F, Martinez M, Artemev A, Brothers M. Structural health monitoring of bonded composite repairs – a critical comparison between ultrasonic Lamb wave approach and surface mounted crack sensor approach. Compos B Eng 2013;47:26–34.
541
[8] Caminero MA, Lopez-Pedrosa M, Pinna C, Soutis C. Damage monitoring and analysis of composite laminates with an open hole and adhesively bonded repairs using digital image correlation. Compos B Eng 2013;53:76–91. [9] Tuttle M, Brinson H. Resistance-foil strain-gage technology as applied to composite materials. Exp Mech 1984;24(1):54–65. [10] Wang X, Fu X, Chung DDL. Strain sensing using carbon fiber. J Mater Res 1999;14(3):12. [11] Xiaojun W, Chung DDL. Continuous carbon fibre epoxy-matrix composite as a sensor of its own strain. Smart Mater Struct 1996;5(6):796. [12] Wang X, Chung DDL. Self-monitoring of fatigue damage and dynamic strain in carbon fiber polymer-matrix composite. Compos B Eng 1998;29(1):63–73. [13] Angelidis N, Wei CY, Irving PE. The electrical resistance response of continuous carbon fibre composite laminates to mechanical strain. Compos A Appl Sci Manuf 2004;35(10):1135–47. [14] Park J-M, Kim P-G, Jang J-H, Wang Z, Kim J-W, Lee W-I, et al. Self-sensing and dispersive evaluation of single carbon fiber/carbon nanotube (CNT)–epoxy composites using electro-micromechanical technique and nondestructive acoustic emission. Compos B Eng 2008;39(7–8):1170–82. [15] Pham GT, Park Y-B, Liang Z, Zhang C, Wang B. Processing and modeling of conductive thermoplastic/carbon nanotube films for strain sensing. Compos B Eng 2008;39(1):209–16. [16] Stübler N, Fritzsche J, Klüppel M. Mechanical and electrical analysis of carbon black networking in elastomers under strain. Polym Eng Sci 2011;51(6):1206–17. [17] Johnson O, Gardner C, Seegmiller D, Mara N, Dattelbaum A, Rae P, et al. Multiscale model for the extreme piezoresistivity in silicone/nickel nanostrand nanocomposites. Metall Mater Trans A 2011;42(13):3898–906. [18] Johnson OK, Gardner CJ, Fullwood DT, Adams BL, Hansen G, Hansen N. The colossal piezoresistive effect in nickel nanostrand polymer composites and a quantum tunneling model. Comput, Mater Continua 2010;15(2):24. [19] Johnson TM, Fullwood DT, Hansen G. Strain monitoring of carbon fiber composite via embedded nickel nano-particles. Compos B Eng 2013;43: 1155–63. [20] Hansen G. The roles of nanostrands and nickel coated fibers in electrically conductive composites design. In: Proceedings of 37th ISTC. Seattle, WA, Conference October 31–November 3, Conference; 2005. [21] Hansen N, Adams DO, DeVries KL, Goff A, Hansen G. Investigation of electrically conductive structural adhesives using nickel nanostrands. J Adhes Sci Technol 2011;25(19):2659–70. [22] Hyatt T, Fullwood D, Bradshaw R, Bowden A, Johnson O. In: Proulx T, editor. Nano-composite sensors for wide range measurement of ligament strain experimental and applied mechanics, vol. 6. New York: Springer; 2011. p. 359–64. [23] Zhou JF, Song YH, Zheng Q, Wu Q, Zhang MQ. Percolation transition and hydrostatic piezoresistance for carbon black filled poly(methylvinylsilioxane) vulcanizates. Carbon 2008;46(4):679–91. [24] Meilikhov EZ. Giant piezoresistivity of nanocomposites with the tunnel conductivity. Sens Actuators, A 2009;153(2):187–90. [25] Composites C. Nickel Chemical Vapor Deposition (CVD) Coated Fibers; 2011. [26] Yoshimoto S, Murata Y, Kubo K, Tomita K, Motoyoshi K, Kimura T, et al. Fourpoint probe resistance measurements using PtIr-coated carbon nanotube tips. Nano Lett 2007;7(4):956–9. [27] Johnson TM, Fullwood DT, Hansen G. Strain monitoring of carbon fiber composite via embedded nickel nano-particles. Compos: Part B 2012;43:1155–63.
Contents lists available at ScienceDirect
Composites: Part B journal homepage: www.elsevier.com/locate/compositesb
Piezoresistive in-situ strain sensing of composite laminate structures Michael C. Koecher, John H. Pande, Scott Merkley, Sam Henderson, David T. Fullwood ⇑, Anton E. Bowden Brigham Young University, Provo, UT 84602, USA
a r t i c l e
i n f o
Article history: Received 30 October 2013 Received in revised form 23 September 2014 Accepted 24 September 2014 Available online 12 October 2014 Keywords: A. Carbon fiber A. Smart materials A. Nano-structures B. Electrical properties
a b s t r a c t Various methods have been developed to monitor the health and strain state of carbon fiber reinforced polymers, each with a unique set of pros and cons. This research assesses the use of piezoresistive sensors for in situ strain measurement of carbon fiber and other composite structures in multidirectional laminates. The piezoresistive sensor material and the embedded circuitry are both evaluated. For the piezoresistive sensor, a conductive nickel nanocomposite sensor is compared with the piezoresistivity of the carbon fiber itself. For the circuit, the use of carbon fibers already present in the structure is compared with the use of nickel coated carbon fiber. Successful localized strain sensing is demonstrated for several sensor and circuitry configurations. Numerous engineering applications are possible in the ever-growing field of carbon-composites. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Carbon fiber reinforced polymers (CFRP) possess a unique set of material properties that include high strength-to-weight ratio, low thermal expansion, and good fatigue characteristics; these benefits have led to rapid expansion in numerous areas of engineering [1–3]. Given the various loading conditions and environments that this relatively new material has to endure, there is a pressing need for a wide range of structural health and condition monitoring solutions. A monitoring solution that includes instantaneous strain sensing could provide useful feedback for control, actuation and data logging functions. This paper discusses the development of piezoresistive sensing solutions for carbon fiber structures. Of particular interest is the embedded circuitry that enables the sensing of local strain in a component via connection to an external resistance meter. The two in-situ piezoresistive sensors that are evaluated are 1. embedded nickel nanostrand (NiN) nanocomposites and 2. neat prepreg carbon fiber. For the connecting circuitry, nickel coated carbon fibers and carbon fiber prepreg alone are compared as pseudowires to the piezoresistive sensor; the probing configuration of the external meter is also considered. Envisaged applications include strain and health monitoring of structural members, as well as the sensing of deflection in compliant components such as sports equipment and actuation devices. For control and data logging purposes. Actual engineering components might include ⇑ Corresponding author at: 435 CTB, Brigham Young University, Provo, UT 84602, USA. Tel.: +1 801 422 6316. E-mail address: [email protected] (D.T. Fullwood). http://dx.doi.org/10.1016/j.compositesb.2014.09.029 1359-8368/Ó 2014 Elsevier Ltd. All rights reserved.
bridge decks, composite beams, wind turbine blades, golf clubs, compliant mechanisms, etc. The reported examples of a tensile specimen, a beam under bending, and a carbon-fiber pressure vessel, illustrate uses of the methodology that could be adapted for various engineering applications.
1.1. Current strain sensing methods Various methods exist for measuring strain in carbon fiber structures. Fiber Bragg gratings, using embedded fiber optics measure strain [4,5], and other exotic and novel methods (e.g. [6–8]) are making tremendous progress towards providing a suite of solutions that will no longer be confined to expensive laboratory situations. More traditional methods include strain gauges composed of thin metal films [9]. The advantages of these types of gauges are that they are fairly simple to install, low cost, and have proved to be successful through years of use in industry. However, metal foils also have a very limited strain range, and because these gauges are adhered to the surface they are susceptible to damage. Furthermore, wires must be routed across or though the structure to carry the required signal to the monitoring unit. Carbon fiber is itself piezoresistive by nature, and can thus act as a strain gauge [10–12]. However, transverse plies in a multidirectional laminate largely short circuit the piezoresistive response of the carbon fiber structure [13]. The issue of routing circuitry to the point of interest must also be considered for this approach. Carbon nanotubes [14,15], carbon black [16], and nickel nanostrands [17]. This paper will focus on nickel nanostrand composites due
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
to their extremely large piezoresistive effect. Johnson et al. discovered that directly embedding a nickel nanostrand patch into unidirectional carbon fiber laminates yielded a significant piezoresistive response [18,19]. However, when the same method was applied to multidirectional laminates the signal quickly short-circuited and no piezoresistive effect was found. To measure piezoresistivity in multidirectional CFRP layups, a nickel nanostrand nanocomposite was insulated in fiberglass to resolve the short-circuiting issue, and embedded between the layers of carbon fiber. This method not only proved successful for measuring strain but also for detecting damage to the structure. However, due to the large stress concentrations induced by the embedded insulated patch, the strength of the composites was severely compromised. The current paper will evaluate methods to measure strain in multidirectional carbon fiber laminates without significantly altering the strength of the carbon fiber structure. Piezoresistive sensing material will compose of either the carbon fiber structure itself, or local regions of nano-composite with nickel nanostrands added to the matrix. The signal path to the monitoring system will be generated using only the carbon fiber or using embedded nickel coated carbon fibers (NCCF). These coated fibers are at least three times more conductive than bare carbon fiber [20] thus reducing the amount of noise in the piezoresistive signal. The optimized configuration of probes for detecting the composite strain will be evaluated, and the resultant methodology will be tested in an engineering situation: a cylindrical carbon fiber pressure vessel. 2. Materials and methods 2.1. Materials Nickel nanostrands (NiNs) display a high strand aspect ratio and a bifurcated structure that allows conductivity of a nanocomposite to be obtained at small volume fractions of nanostrands [21]. When combined with a polymer matrix they form a piezo-resistive material that has been used as the basis for strain sensors [22]. The piezoresistivity of such nanocomposites has been most successfully modeled using percolation theory and a quantum tunneling mechanism [23,24]. With a sufficiently large volume fraction of conductive filler material nanojunctions will form between filler particles in which electrons can tunnel through the insulating barrier. Once this occurs an electrical network is created through the nanocomposite and it becomes conductive. The resistivity of a nanojunction can be calculated using Eq. (1):
h
2
q ¼ 2 pffiffiffiffiffiffiffiffiffiffi exp e
2mk
pffiffiffiffiffiffiffiffiffiffi ! 4p 2mk s h
ð1Þ
where h is the Planck constant (J s), e is the electron charge (C), m is the electron mass (kg), k is the tunneling barrier height (J), and s is
535
the distance between particles (m) [17]. Thus the resistivity of a nanojunction is a function of the inter-particle distance, s. As a nanocomposite is strained s will change and a piezoresistive signal is obtained. Another novel material, nickel coated carbon fiber (NCCF), is tested as the basis for the sensor circuitry in this paper. Carbon fiber receives a uniform coating of nickel through a chemical vapor deposition process. These coated fibers are much more conductive than bare carbon fiber and are often used to aid in electrical shielding in carbon fiber structures [25]. Both the nickel nanostrands and the nickel coated carbon fiber were produced and provided by Conductive Composites Company (Heber, Utah). The structure of these materials can be seen in the SEM images depicted in Fig. 1. Along with these novel materials, a traditional unidirectional carbon fiber prepreg is used as the structural material that is to be monitored: ZR6-P35, provided by Zoltek. Components with various pre-preg layups were created to determine the capability and limitations of the embedded conductive materials for in situ strain sensing. For initial sensor evaluation, laminates were formed in configurations suitable for tensile testing. Each carbon fiber laminate was composed of layers of prepreg cut to 250 mm 25 mm. Woven fiber glass tabs were adhered to the ends of the carbon fiber samples to insulate the sample from the metal grips of the tensile tester and ensure that these did not interfere with the signal. Each sample consisted of a sensor patch location and a sensor patch material. The patch location refers to the area that strain is to be measured and the patch material was either prepreg embedded with NiNs or simply the carbon fiber prepreg itself. To limit variation and guarantee uniformity in the samples with embedded nanostrands, 0.02 g of nickel nanostrands were filtered through a 60 mesh screen (250 lm) before being placed in a patch area of 19 mm by 12 mm. For carbon fiber and NiN patches the [0] direction (prepreg oriented parallel to the applied strain) and [90] direction (prepreg oriented transverse to the applied strain) were evaluated as the sample was strained in the direction of its length. As mentioned previously, the use of NCCF was assessed as a means of forming the sensor circuit. The alternative to NCCF was to allow the signal to carry along the carbon fibers in the prepreg itself. It was hypothesized that the NCCF would better control the signal flow and limit scatter and short-circuiting across the cross plies of prepreg. Thus nickel coated carbon fiber was embedded onto the surface of various samples and compared to control samples with no embedded materials. Fig. 2 is an example of a sample with embedded nanostrands and nickel coated carbon fiber. As can be seen in Fig. 2 a small gap was placed between NCCF bundles at the patch location. The gap in the NCCF was required to ensure the signal traveled through the carbon fiber or NiN patch at the patch location. Table 1 shows the various sample configurations tested. Also included in Table 1 are signal-to-noise ratio
Fig. 1. SEM images (from left to right) of nickel nanostrands and nickel coated carbon fiber (NCCF used courtesy of Nathan Hansen, Conductive Composites Company).
536
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 2. [90,0,0,90] Laminate with embedded NiNs and NCCF (gray lines). A small piece of carbon fiber is placed over the NiN patch to prevent inadvertent dispersion of the NiNs during curing.
Table 1 Tensile sample configurations that were tested and their accompanying signal-tonoise ratio (SNR) results. Sample
Material
Patch
SNR
Standard deviation
A B C D E F
[0,90]s Carbon fiber [0,90]s Carbon fiber [90,0]s Carbon fiber [90,0]s Carbon fiber Woven fiberglass Woven fiberglass
Carbon fiber [0] NiN Carbon fiber [90] NiN Carbon fiber [90] [90] and NiN
0.86 2.16 17.90 4.32 14.50 80.54
0.28 0.21 1.44 0.05 0.90 1.93
(SNR) results which will be discussed in the results section. Fiberglass samples were also prepared and tested in order to test a piezoresistive patch (of either carbon fiber/NiNs or carbon fiber alone) in non-conductive structures. Patches consisting of nanostrands and carbon fiber were embedded in the fiberglass structure to compare the piezoresistivity of the carbon fiber and the NiN patch. Four samples were made for each configuration and typically five measurements were taken on each sample (as described in Section 2.2) to produce the statistics reported later.
significant as compared to the resistance of the component [26]. In this method a known current is passed through the outer probes and a voltage is read between the inner probes. Since no current is passing through the inner probes there is no contact resistance. A National Instruments NI 9219 multifunctional module was set up for four probe resistance readings for these tests. A constant current of 500 lA was applied to the outer probes while a voltage was read across the inner probes. Using Ohms law an output of resistance was obtained. Three different configurations of probing to measure the piezoresistive change at the patch location were evaluated in this research. In the first method, referred to as the ‘‘collinear’’ probing method, four collinear probes were placed on the carbon fiber sample to eliminate contact resistance. For the second method the probes were placed in a ‘‘box’’ configuration. The box configuration is similar to the collinear method except there are two pseudowires instead of four and each wire has two probes on it. In the final method, referred to as ‘‘longitudinal’’ probe method, the probes were place on each side of the patch. In each method the pseudo-wires consist of either the conductive carbon fiber prepreg or the embedded NCCF. Fig. 3 depicts the difference between the three probing methods. Samples were cyclically loaded in tension in an Instron tensile tester to a prescribed strain of 0.30% at a rate of 0.1 Hz in order to determine the feasibility of each configuration as in situ strain sensing for multidirectional carbon fiber laminates. Three point bending tests were also performed on select samples using supports 12.5 cm apart with a 10 mm vertical displacement of the wedge to determine the piezoresistive changes due to tension and compression. During loading the resistance measurements were sampled at a rate of 100 Hz and a moving average of 10 was used to improve signal smoothness.
3. Results and discussion 3.1. Tensile tests
2.2. Method The four probe resistance measurement method was used to measure the piezoresistive signal at a point remote from the desired strain measurement; this is a common method to eliminate contact resistance readings when contact resistance is
3.1.1. Nickel coated carbon fiber To determine the feasibility of using NCCF as pseudo-wires to the piezoresistive patch it was first necessary to determine if piezoresistive properties exist in NCCF. If the NCCF display a piezoresistive response, this will confuse the signal expected from the
Fig. 3. (a) Transverse collinear probe method, (b) box probe method, and (c) longitudinal probe method. In each method the gray lines represent the circuit path using either the carbon fiber strands from the structure or embedded NCCF. EX+ and EX are the input and output current probes while HI and LO are the voltage reading probes.
537
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 4. Comparison of the piezoresistivity of carbon fiber to the more conductive NCCF.
sensor. Tensile tests were performed on carbon fiber and NCCF and compared in Fig. 4. It can be seen that the NCCF is much more conductive than carbon fiber and does not exhibit piezoresistive properties. This suggested that using NCCF as pseudo-wires to the patch does not contribute to the piezoresistive signal, and one can be confident that the strain being measured is precisely located at the patch location. 3.1.2. Circuit and probing methods A wide variety of tests were performed on samples with various probing configurations (see Fig. 3) with results reported in Table 2. The signal-to-noise ratio (SNR) was calculated and used as a metric in the laboratory to quantitatively compare the piezoresistive signal between samples. A larger SNR correlated to a higher quality piezoresistive signal. In general, it was observed that a SNR value lower than about 0.30 suggested that there is no distinguishable signal that can be correlated to strain; an ideal SNR for non-laboratory situations has not been determined. Comparing Tests 3, 4, and 9 to Tests 5, 6, and 8, respectively, it can be seen that using embedded NCCF as pseudo-wires allow for a stronger piezoresistive signal at greater distances than just using carbon fiber prepreg. It can also be seen in Table 2 that the longitudinal probing method allowed for remote probing of the piezoresistive sensor at greater distances. It is of note that the longitudinal probing method also yielded more consistent repeatable results. Thus, each subsequent sample was tested using the longitudinal probing method. 3.1.3. Sensor tests Each sample in Table 1 was subsequently probed using the longitudinal method. Fig. 5 shows the results obtained from tensile tests performed on samples A and C. Each of these samples consisted of a carbon fiber patch with different carbon fiber orientations. In Fig. 5a the carbon fiber on the surface and NCCF are
Fig. 5. Comparison of piezoresistive signals obtained from (a) sample A and (b) sample C from Table 1.
oriented in the [0] direction. In Fig. 5b the carbon fibers on the surface are oriented in the [90] direction while the NCCF are oriented in the [0] direction. Thus it can be seen that when the carbon fiber is oriented parallel to the NCCF the piezoresistive signal does not have a strong correlation to the strain, but when carbon fibers and NCCF are oriented transverse to one another the piezoresistivity relates well to strain. It is further confirmed in Fig. 6 that when the carbon fiber and NCCF are oriented parallel to one another no correlation between the piezoresistive signal and strain is found.
Table 2 SNR calculations for various samples with different probing methods. Test number
Orientation
NiN patch
NCCF
Probe method
Distance (cm)
SNR
Standard deviation
1 2 3 4 5 6
[0,90,0,90]
No
No
Yes
No
Parallel collinear Transverse collinear Box
NA NA 2.5 4 2.5 7.5
0.0518 0.8767 3.0018 0.1454 6.2408 0.4556
0.043 0.099 1.202 0.022 2.448 0.049
7 8 9
[90,0,0,90]
Longitudinal
4 14 14
0.5099 0.4818 0.2534
0.196 0.068 0.032
Yes Yes
Yes No
538
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 7. Comparison of fiberglass samples (a) E and (b) F.
Fig. 6. Comparison of piezoresistive signals obtained from (a) sample B and (b) sample D.
From Fig. 5 it has been shown that the piezoresistivity of the neat carbon fiber prepreg can be exploited to determine the percent strain in a multidirectional laminate carbon fiber structure. Also, NCCF can be used as pseudo-wires to a patch location thus allowing strain measurement at a specific location while remotely probing from the patch location. Sample B and sample D are similar to sample A and sample C except a NiN patch is included rather than using the neat carbon fiber as a patch. These samples were tested to determine how a NiN patch affects the piezoresistive signal (which has resulted in good results for other authors for long-range strain gauges) [27]. The results from Fig. 6 yield similar results to those in Fig. 5. This suggests that using a NiN patch does nothing to improve upon the piezoresistive signal. Thus the relative orientation of carbon fiber in the laminate and the use of NCCF has more of an effect than using a NiN patch. To further investigate the effect of the NiNs on the piezoresistive signal samples E and F were tested and the results can be seen in Fig. 7. Each sample has a piezoresistive signal that strongly correlates to the strain which again shows that a NiN patch does little to affect the piezoresistive signal. Table 1 shows a calculation for the signal-to-noise (SNR) of each sample. It is of interest to note that the signals obtained in Fig. 7 have less noise than those in Figs. 5 and 6. This can be attributed to the fact that even though the NCCF is much more conductive than carbon fiber, the signal will slightly scatter across the carbon fiber thus adding noise to
the overall signal in the carbon-fiber laminate samples. In samples E and F the fiberglass is completely insulating and there is minimal scatter in the signal and the amount of noise is dramatically decreased. 3.2. Bending tests A [90,0,0,90]s carbon fiber laminate with NCCF embedded on the top and bottom surfaces was placed in bending to evaluate the variation of piezoresistivity due to compressive and tensile strain across a single component. From Fig. 8a, increasing the displacement of the wedge caused a compressive strain on the top surface and a decrease in resistance was obtained. The opposite is true on the bottom surface; the tensile strain increased with displacement and an increase in resistance was obtained. This test demonstrates the ability to measure varying strain at different points in a sample using the NCCF circuits to pin-point the measurement site. The three point bending test was also conducted on fiberglass Sample E from Table 1. Similar results were obtained and can be seen in Fig. 9. The fiberglass sample once again yielded a higher signal-to-noise ratio of 24.76 as compared to the 6.79 signal-tonoise ratio of the carbon fiber, which further suggests that there is some signal loss due to the conductive nature of the carbon fiber. 3.3. Failure tests The results so far demonstrate that it is feasible to use NCCF to sense strain at a remote distance from a probing location in a sample in either tension or compression. For this to be a useful method
539
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541 Table 3 Failure tests results.
Failure stress (MPa) Standard deviation Strain to failure (%) Standard deviation
Embedded NiNs/NCCF
No NiNs/NCCF
Total
9578 364 1.75 0.42
9445 1179 2.01 0.59
9511 875 1.88 0.53
ASTM D3039 standard. These results were compared with samples of the same layer orientation without embedded NiNs or NCCF. Table 3 shows stress to failure for various samples with and without embedded NiNs and NCCF. Embedding NiNs and NCCF in the middle or top layer of a carbon fiber laminate showed no significant effect on the failure stress. Thus it appears that the embedded NCCF and NiNs do not create significant stress concentrations in the bulk structure, and these materials can be used without compromising the integrity of the carbon fiber structure. 3.4. Application
Fig. 8. Three point bending tests on a [90,0,0,90]s carbon fiber laminate with NCCF embedded on the (a) top surface and (b) bottom surface.
As a proof of concept a cylindrical pressure vessel was fabricated to show the feasibility and advantages of using nickel coated carbon fiber as pseudo-wires to allow remote sensing of piezoresistive carbon fiber patches. Three layers of roll-wrapped carbon fiber prepreg were laid around a cylindrical 15 cm diameter mandrel. The two interior layers were wrapped with the fibers aligned in the longitudinal (axis) direction while the outer layer was aligned in the hoop direction. Bundles of nickel coated carbon fiber were placed in the longitudinal direction on the outer surface and in the hoop direction on the interior of the pressure (see Fig. 10). A second set of longitudinal NCCF bundles were embedded on the opposite side for redundancy in the sensor setup. This arrangement ensured that at each sensor patch location the carbon fibers and NCCF were oriented transverse one relative to another to enable the best possible piezoresistive signal to be obtained as found in Figs. 5 and 6. Once cured, caps were adhered to both ends to seal the pressure vessel. The pressure vessel was pressurized using distilled water to prevent electrical short circuiting and alleviate safety concerns with pressurized air. The vessel was pressurized using an Ametek M&G T-65 twin seal pressure pump while resistance measurements were simultaneously being recorded. Results for longitudinal strain on the outside surface can be seen in Fig. 11. As the pressure increased the longitudinal strain increased and an increase in the resistance was obtained. The pressure vessel was rapidly depressurized and accordingly the resistance dropped. Thus, as expected, with a tensile strain in the longitudinal direction the piezoresistive signal is positive.
Fig. 9. Three point bending test results from Sample E.
to measure strain in a carbon fiber structure it is important to know how embedding this NCCF material affects the strength of the carbon fiber structure. Samples were embedded with NCCF and NiNs in various positions, and loaded to failure according the
Fig. 10. Pressure vessel design (gray lines represent NCCF bundles). The gap between each NCCF bundle is transverse to the carbon fiber thus allowing for piezoresistive readings.
540
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
Fig. 11. Longitudinal strain measurement in pressure vessel. The vessel was pressurized to 552 kPa and then rapidly depressurized starting at approximately 12 s.
Fig. 14. Extreme piezoresistive changes when failure occurs (gray region) suggest that failure can be detected in a carbon fiber structure.
longitudinal strain creates a positive piezoresistive signal while measuring hoop strain yields negative piezoresistivity. Further loading the pressure vessel above 827 kPa caused the vessel to start failing. Resistance measurements were taken on both longitudinal bundles and the hoop bundle as the pressure vessel began to fail. In Fig. 14 it can be seen as the fibers began to delaminate and break large resistance changes were detected. It is evident that this method of strain measurement could be used to detect failures within a carbon fiber component. The different amplitudes of resistance changes between the three locations could suggest the location of where failure began. 4. Conclusions Fig. 12. Strain reading on inside surface of pressure vessel. Pressure reached a max of 827 kPa at approximately 17 s.
The results measuring the hoop strain on the interior surface can be seen in Fig. 12. It was assumed that as the pressure increased there would be a tensile strain in the hoop direction and thus a positive piezoresistive signal would be obtained, yet the results show a negative piezoresistive signal with increasing stress. These results are being attributed to the compression of the longitudinal carbon fibers in the radial direction. As the pressure increased the carbon fibers were compressed into intimate contact with one another creating a lower resistance (see Fig. 13). Unfortunately, the data recording was interrupted before complete depressurization and subsequently the vessel was damaged so a return of the resistance reading to its unpressurized state was not seen. Yet, the results still indicate a negative piezoresistive signal in the hoop direction. From the resistance measurements obtained it is evident that using NCCF as pseudo-wires to a patch location can facilitate remote strain measurements in a carbon fiber structure. Measuring
Fig. 13. Pressure causes the fibers to compress into one another thus lowering the resistance in the hoop direction.
Embedded nickel nanostrands and nickel coated carbon fiber were evaluated as tools for in-situ strain sensing in a multidirectional carbon fiber laminate. Tests were conducted via altering of the strain sensor patch and the connecting circuitry. Previous research has used the piezoresistivity of directly embedded nickel nanostrand nanocomposites to measure strain in unidirectional layup. Testing a variety of strain sensor configurations has shown that directly embedding nickel nanostrands into a multidirectional laminate is ineffective for strain measurement and does little to improve upon the piezoresistivity of the carbon fiber. Carbon fiber’s piezoresistivity is a good measure of strain, and thus a valid strain sensor, but if the carbon fiber is also used as the current carrier to a probe location, the probing location must be extremely close to the area where strain is to be measured to allow a discernible electrical signal to be obtained. The more conductive nickel coated carbon fiber does not have the same piezoresistive properties as carbon fiber and hence is a better choice for supplying a circuit from the sensor area to the probe location. Using NCCF as the connecting circuitry provides pseudo-wires to remotely measure the strain at a desired location without piezoresistivity in the circuit affecting the signal. It has been shown that the best results are obtained when the NCCF circuit is oriented perpendicular to the carbon fibers that are in the sensor region. Furthermore the optimal probe geometry involves the four-probe longitudinal method. Future work will be performed to determine if embedding the NCCF circuit between layers of a laminate can accurately determine strain from arbitrary positions within a sample. Samples embedded with nickel coated carbon fiber and nickel nanostrands were loaded to failure in tension to determine the adverse effects these materials have on the strength of the carbon fiber. It has been shown that the embedded materials do not
M.C. Koecher et al. / Composites: Part B 69 (2015) 534–541
significantly alter the strength of the structure. Thus, using embedded NCCF is a safe, easy way to remotely measure strain within a carbon fiber structure without compromising structural integrity. In addition to tensile and bending tests of laminates, a pressure vessel was built as a proof of concept application of remote strain sensing using embedded nickel coated carbon fiber for the connecting circuit. When pressurized, longitudinal and hoop strain were detected. Longitudinal strain yielded positive piezoresistive results while hoop strain yielded negative piezoresistive results. It was also found that when the pressure vessel began to fail, the failure was detectable through an extreme change in the resistance reading. The tests illustrate the potential for applying such an approach in a wide range of engineering applications for health monitoring, strain sensing and control of composite components. The distance between the location of measured strain and the probe location is modest in the tests (up to 14 cm); the ability to place the probes at much greater distances from the location of interest is desirable, and may be increased with further optimization of the method. However, for many applications a surface probe at similar distances from the strained location may be acceptable. Acknowledgement This work was supported, in part, by a National Science Foundation Grant, Contract Number: CMMI-1235365. References [1] Gibson RF. Principles of composite materials mechanics. 2 ed. Boca Raton: CRC Press; 2007. [2] Strong AB. Fundamentals of composites manufacturing: materials, methods, and applications. 2 ed. Dearborn: Society of Manufacturing Engineers; 2008. [3] Roberts T. The Carbon Fibre Industry Worldwide 2008–2014. MTP; 2009. [4] Kim H-S, Yoo S-H, Chang S-H. In situ monitoring of the strain evolution and curing reaction of composite laminates to reduce the thermal residual stress using FBG sensor and dielectrometry. Compos B Eng 2013;44(1):446–52. [5] Ling H-Y, Lau K-T, Su Z, Wong ET-T. Monitoring mode II fracture behaviour of composite laminates using embedded fiber-optic sensors. Compos B Eng 2007;38(4):488–97. [6] Fu T, Liu Y, Lau K-T, Leng J. Impact source identification in a carbon fiber reinforced polymer plate by using embedded fiber optic acoustic emission sensors. Compos B Eng 2014;66:420–9. [7] Habib F, Martinez M, Artemev A, Brothers M. Structural health monitoring of bonded composite repairs – a critical comparison between ultrasonic Lamb wave approach and surface mounted crack sensor approach. Compos B Eng 2013;47:26–34.
541
[8] Caminero MA, Lopez-Pedrosa M, Pinna C, Soutis C. Damage monitoring and analysis of composite laminates with an open hole and adhesively bonded repairs using digital image correlation. Compos B Eng 2013;53:76–91. [9] Tuttle M, Brinson H. Resistance-foil strain-gage technology as applied to composite materials. Exp Mech 1984;24(1):54–65. [10] Wang X, Fu X, Chung DDL. Strain sensing using carbon fiber. J Mater Res 1999;14(3):12. [11] Xiaojun W, Chung DDL. Continuous carbon fibre epoxy-matrix composite as a sensor of its own strain. Smart Mater Struct 1996;5(6):796. [12] Wang X, Chung DDL. Self-monitoring of fatigue damage and dynamic strain in carbon fiber polymer-matrix composite. Compos B Eng 1998;29(1):63–73. [13] Angelidis N, Wei CY, Irving PE. The electrical resistance response of continuous carbon fibre composite laminates to mechanical strain. Compos A Appl Sci Manuf 2004;35(10):1135–47. [14] Park J-M, Kim P-G, Jang J-H, Wang Z, Kim J-W, Lee W-I, et al. Self-sensing and dispersive evaluation of single carbon fiber/carbon nanotube (CNT)–epoxy composites using electro-micromechanical technique and nondestructive acoustic emission. Compos B Eng 2008;39(7–8):1170–82. [15] Pham GT, Park Y-B, Liang Z, Zhang C, Wang B. Processing and modeling of conductive thermoplastic/carbon nanotube films for strain sensing. Compos B Eng 2008;39(1):209–16. [16] Stübler N, Fritzsche J, Klüppel M. Mechanical and electrical analysis of carbon black networking in elastomers under strain. Polym Eng Sci 2011;51(6):1206–17. [17] Johnson O, Gardner C, Seegmiller D, Mara N, Dattelbaum A, Rae P, et al. Multiscale model for the extreme piezoresistivity in silicone/nickel nanostrand nanocomposites. Metall Mater Trans A 2011;42(13):3898–906. [18] Johnson OK, Gardner CJ, Fullwood DT, Adams BL, Hansen G, Hansen N. The colossal piezoresistive effect in nickel nanostrand polymer composites and a quantum tunneling model. Comput, Mater Continua 2010;15(2):24. [19] Johnson TM, Fullwood DT, Hansen G. Strain monitoring of carbon fiber composite via embedded nickel nano-particles. Compos B Eng 2013;43: 1155–63. [20] Hansen G. The roles of nanostrands and nickel coated fibers in electrically conductive composites design. In: Proceedings of 37th ISTC. Seattle, WA, Conference October 31–November 3, Conference; 2005. [21] Hansen N, Adams DO, DeVries KL, Goff A, Hansen G. Investigation of electrically conductive structural adhesives using nickel nanostrands. J Adhes Sci Technol 2011;25(19):2659–70. [22] Hyatt T, Fullwood D, Bradshaw R, Bowden A, Johnson O. In: Proulx T, editor. Nano-composite sensors for wide range measurement of ligament strain experimental and applied mechanics, vol. 6. New York: Springer; 2011. p. 359–64. [23] Zhou JF, Song YH, Zheng Q, Wu Q, Zhang MQ. Percolation transition and hydrostatic piezoresistance for carbon black filled poly(methylvinylsilioxane) vulcanizates. Carbon 2008;46(4):679–91. [24] Meilikhov EZ. Giant piezoresistivity of nanocomposites with the tunnel conductivity. Sens Actuators, A 2009;153(2):187–90. [25] Composites C. Nickel Chemical Vapor Deposition (CVD) Coated Fibers; 2011. [26] Yoshimoto S, Murata Y, Kubo K, Tomita K, Motoyoshi K, Kimura T, et al. Fourpoint probe resistance measurements using PtIr-coated carbon nanotube tips. Nano Lett 2007;7(4):956–9. [27] Johnson TM, Fullwood DT, Hansen G. Strain monitoring of carbon fiber composite via embedded nickel nano-particles. Compos: Part B 2012;43:1155–63.