Damage self-sensing and strain monitoring of glass-reinforced epoxy composite impregnated with graphene nanoplatelet and multiwalled carbon nanotubes

2.1 Materials

The developed composites comprised glass fibre, epoxy resin, GNPs, and MWCNTs obtained from a local supplier in Malaysia. The reinforcing fibre used was commercial plain-woven E-type glass with a surface density of 50 g/cm². Smooth-On EpoxAmite 100 series resin with 103 slow hardener was used as a matrix. The epoxy resin has good mechanical properties, moderate gel time, and good wettability [16]. GNP and MWCNT nanofillers were incorporated as conductive reinforcement materials, as shown in Table 1.

2.2 Composite fabrication

In this study, direct mixing was used to disperse the GNPs and MWCNTs in epoxy. This method demonstrated a significant improvement on the stability and dispersion of GNPs-MWCNTs in epoxy at a low cost, and the method was environmentally friendly and fast [17]. Different amounts of GNP and MWCNT nanofillers were added to the epoxy with weight contents of 0.5, 1.5, 3, and 5 wt% as reported in Table 2. These composites were fabricated through the high shear mixing and sonication followed by the hand lay-up method, as illustrated in Figure 1. The mixtures were mixed first using a mechanical stirrer at 1,000 rpm for 10 min and placed in a high shear mixer (Thinky Mixer ARE 310) for 12 min mixing and 2 min deforming. This step continued using an ultrasonication probe for 20 min at 30% amplitude and 24 kHz in an ice-water bath to prevent excessive heating during the process [18]. The purpose of mixing process using a high shear mixer with an ultrasonication probe is to break the nanofiller aggregates. Subsequently, the curing agent was added and mixed manually at a constant speed and stirred for 5 min. The hardener was mixed with the nanofiller–epoxy mixture in a weight ratio according to the manufacturer’s specifications. The GNP-epoxy and MWCNT-epoxy mixtures were evenly laminated on 5 layers of woven glass fabrics (200 mm² × 200 mm²) using a hand roller and vacuum-bagging method. The roller helps the mixture penetrate and remove the bubbles inside the glass fabric plies. The glass fabric and epoxy were weighed before each hand lay-up process. In addition, the in-mould pressure was maintained below 3 mbar when the vacuum pump was switched on to force the epoxy mixture to favour the compaction until the end of the curing process. The composite was precured at normal room temperature and pressure for 12 h, followed by post curing for an additional 2 h at 80°C [19].

Figure 1

Preparation process of the GNP-glass and MWCNT-glass composites.

2.3 Electrical test

The GNP-glass and MWCNT-glass composites were cut using a Dremel 4000 tool in accordance with ASTM D257 [20]. Specimens with a diameter of 13 mm and thickness of less than 1 mm were prepared. The electrical test of the two-point probe method was set up using a power DC digital multimeter (6.5 digit multimeter Agilent 34401A) and a voltage supply of 5 V. Calibration was performed prior to the test to ensure that accurate data were obtained. The electrical properties of the specimens were tested for electrical properties containing different GNP and MWCNT nanofiller weight contents. Five replicates of each specimen were tested, and the average values were recorded. A silver conductive paint was used to minimise the contact resistance between the specimen and electrode encountered during the test. The electrical conductivity was calculated using equation (1) [21].

where R is the resistance, r is the resistivity of the material, A is the area of the specimen, and l is the thickness of the specimen.

2.4 Mechanical tests

A rectangular shape was cut according to the ASTM D3039-14 and ASTM D7264-14 standards for the tensile and flexural tests, respectively [22,23]. The tensile test was performed on a Shimadzu universal testing machine with a load cell of 100 kN at a strain rate of 2 mm/min. The composite was placed in a position parallel to the loading direction during tensile testing. The flexural test was conducted using a universal micro tester (INSTRON 5848) with a 2 kN load and crosshead velocity of 2.5 mm/min. Each specimen was connected using copper connectors to a 6.5 digit multimeter (Agilent 34401A) controlled via LabVIEW software, which measures the resistance values of the electromechanical response. During each experiment, the force, stroke, and electrical resistance as a function of time were measured simultaneously. All data and results were used as the mechanical–electrical measurement for damage self-sensing behaviour of the GNP-glass and MWCNT-glass composites.

2.5 Morphological analysis

The microstructures of the GNP-glass and MWCNT-glass composites were observed using field emission scanning electron microscopy (FESEM, Nova NanoSEM 450). This method uses a focused beam of high-energy electrons to generate a variety of signals at the specimen surface, which shows the morphological observation, chemical composition, and crystalline structure of the specimens. FESEM was used to investigate the dispersion state of the GNP and MWCNT nanofillers in the matrix and the failure mechanism. Images of the fractured surface specimens were observed using an accelerating voltage in the range of 3–15 kV [19].

3.1 Electrical conductivity

The electrical responses of the GNP-glass and MWCNT-glass composites were studied by measuring the conductivity with respect to the nanofiller weight content, as shown in Figure 2. The conductivity of the composites increased as the GNP and MWCNT nanofillers changed from 0 to 5 wt%. GNP-glass and MWCNT-glass of 0 wt% had the lowest conductivity value of approximately 10⁻⁹ 1/Ω cm owing to the nonavailability of conducting channels and insulating glass fibre impediment in the composite. Thus, the conducting network of the nanofiller was the main factor that enhanced the conductivity of the composites [24,25]. In this study, the conductivity of the GNP and MWCNT nanofiller systems was considerably influenced by the dispersion state. The close dispersion of the nanofiller formed a conductive pathway through which electrons moved at a faster rate. The nanofillers then realigned and contacted each other, resulting in more effective conductive pathways that enabled the electrons to travel [26]. The entanglement and agglomeration of the nanofiller contribute to the enhancement of the conducting path. Theoretically, the parameter that influences the electrical conductivity is the percolation threshold (p _c), as stated in equation (2) [24]. It is often used to characterise the transition of a conductive polymeric material from an insulator to a conductor. The variation in p _c was attributed to many factors, such as surface modification, nanofiller loading, and polymer viscosity, of the polymer composite that contains the conductive nanofiller [27,28]. The conductivity increases as the nanofiller weight content approaches the critical concentration.

where σ is the electrical conductivity of the specimen, k is a constant, p is the weight content of the nanofiller, p _c is the critical weight content or percolation threshold of the nanofiller, and t is the critical exponential that indicates the dimensionalities of the conductive network in the composite.

Figure 2

Conductivity of GNP-glass and MWCNT-glass composites.

There was a significant increase in conductivity in the range of 1.5–3 wt% for the GNP-glass and MWCNT-glass composites. As shown in Figure 2, there was a possible p _c zone in the range where the conductivity increased by four orders of magnitude. A previous study determined that the principle of p _c can be regarded as the concentrations approaching the critical value at which the electrical conductivity of nanofiller loading increases dramatically [6,24,29,30,31]. This p _c shows the combination of the minimum number of GNP and MWCNT nanofiller contact points and the minimum average distance between the nanofillers to create a conductive network. The best fits of the experimental conductivity data to the log–log plot of p _c give values of GNP-glass and MWCNT-glass of 2.55 and 2.50 wt%, respectively. After the addition of 3 and 5 wt% GNP and MWCNT nanofillers, the conductivity value almost levelled off, and a stable electrical signal was obtained. This indicates that an optimal formation of a conductive network existed between electron and electronic charge transportation. The additional charge carriers offer more direct contact between the GNP and the MWCNT nanofillers, which facilitate rapid rate transfer of electrons [26]. Based on quantum mechanics materials, electrons can jump between neighbouring nanofillers separated by insulating polymer chains. Hashemi and Weng [32] found that electron hopping among nanofiller particles led to an increase in conductivity. The combination of direct contact between the nanofiller, intrinsic conduction of GNP and MWCNT nanofiller, and electron tunnelling caused a sharp increase in the electrical conductivity. These results support the idea of an in situ polymerisation process that enhances the interfacial interaction between the nanofiller and epoxy, thereby improving the composite conductivity [33].

3.2 Evaluation of tensile strength using fractional change in resistance response (FCRR)

The GNP-glass and MWCNT-glass composites were subjected to tensile loading using the two-point probe method to obtain the electrical response under strain. The result was plotted in a graph of the typical response for FCRR and stress to strain under tensile load. Four variations in GNP and MWCNT weight content, namely, 0.5, 1.5, 3, and 5 wt%, in the composites were analysed. Figure 3 shows that the tensile strength increased when 0.5–3 wt% nanofiller was incorporated and plateaued after further addition of 5 wt% due to homogeneity. The tensile strength for 3 wt% was the highest among all composites, and this value improved by 16.7% from 198 to 237 MPa for GNP-glass and 24.4% from 164 to 217 MPa for MWCNT-glass. This high strength was attributed to the stiffness and reinforcing effects of the nanofiller in the epoxy. The good interfacial bonding between the nanofiller and epoxy made the load transfer possible, which enhanced the mechanical properties of the composite. In general, the tensile strength of the composites was improved by adding GNP and MWCNT nanofillers up to the optimal content [7,34]. However, with an increase in the nanofiller content up to 5 wt%, the inhomogeneous dispersion of the nanofiller can lead to the appearance of agglomeration. The tensile strength of the specimen could not be further improved owing to the larger amount and increasing difficulty in dispersing the nanofiller. The strong π–π interaction in the GNPs and van der Waals forces in the MWCNTs caused the nanofillers to reagglomerate. This phenomenon led to a weaker interfacial interaction between the nanofiller and the epoxy in the composite. Thus, the agglomeration of a higher amount of nanofiller contributes to the defect and loss of tensile strength. This can be seen in Figure 3(d) with a decrease in tensile strength by 10.4 and 20.1% for 5 wt% of nanofiller in the GNP-glass and MWCNT-glass composites, respectively, compared with 3 wt% for the specimen. Seretis et al. [35] also found that the GNPs addition used up to 5 wt% until 30 wt% had weakened the tensile performance of the laminated nanocomposite.

Figure 3

Tensile stress and FCRR as a function of strain in GNP-glass and MWCNT-glass composites: (a) 0.5, (b) 1.5, (c) 3, and (d) 5 wt%.

The incorporation of GNPs and MWCNTs into the glass composite makes the surface of the composite electrically conductive because of the formation of continuous pathways from the nanofiller. The composite structure becomes sensitive to external stimuli, which facilitates the fabrication of a smart sensor material. The FCRR was related to the environmental and loading changes, which were calculated by the ratio of changed resistance (ΔR) during tensile loading to the unstrained resistance (R _o) of the composite [28]. As shown in Figure 3, the FCRR curve exhibits a nonlinear behaviour with an exponential increase with the applied strain until the final fracture. Specimens with 3 and 5 wt% content of nanofiller had values larger than p _c and established an abundant conductive network that led to different types of FCRR curve trend. Compared with the specimens with 0.5 and 1.5 wt% nanofiller loading, the slope of the FCRR curve for 3 and 5 wt% exhibits a smoother shape. A softening electrical resistance response was achieved close to p _c owing to the creation of a higher contact number of conductive paths along the matrix when a large number of nanofillers are present in the composite structure [36]. The increased conductive path leads to the consideration of the influence of the strain sensing mechanism because of the loss of electrical contact between the nanofiller particles. Furthermore, a minimum content of GNP and MWCNT nanofillers was added to ensure that the composite was electrically conductive. This conductivity can be achieved through the tunnelling effect mechanism. There was a lack of sufficient path to create contact paths, or their concentration was negligible. This can be used to explain the contribution of electrical contact resistance; thus, it can be neglected, and the signal can be fitted to make the tunnelling conductivity effect the dominant mechanism. Thus, the tunnelling effect leads to a sharp increase in the composite resistance in the application of a high tensile load.

3.3 Evaluation of flexural strength using FCRR

Figure 4 shows the FCRR curves obtained for composites with different GNP and MWCNT nanofiller weight contents from 0.5 to 5 wt%. A similar trend was observed in the tensile test, whereas the flexural strength increased as the nanofiller weight content increased from 0.5 to 3 wt% due to homogeneity. The 3 wt% of GNP-glass and MWCNT-glass resulted in the highest flexural strength values of 294 and 293 MPa, respectively, because of the stiffness and reinforcing effect of the nanofiller in the epoxy. The good interfacial bonding between the incorporated nanofiller and epoxy made load transfer possible. Thus, incorporation of the GNP and MWCNT nanofillers up to the optimal content improved the flexural properties of the composites. However, the addition of nanofiller concentration beyond the optimal value increases the matrix viscosity, which converts the exfoliated structure within the epoxy into a partially exfoliated structure because of the nanofiller agglomerated structure [37]. The strong π–π interactions in the GNPs and van der Waals forces in the MWCNTs caused the nanofiller to re-agglomerate with each other and weakened the interfacial bonding in the epoxy. The nanofiller dispersed in the epoxy tended to move randomly owing to collisions between particles and matrix; hence, it was possible that the nanofiller particles could collide. As a result, the GNP and MWCNT nanofillers started to re-agglomerate, forming a large particle cluster because of the presence of instantaneous dipoles in each nanofiller under van der Waals forces. Thus, the agglomeration of a higher amount of nanofiller resulted in poor flexural strength. This explains the flexural strength in Figure 4(d) for 5 wt% nanofiller loading of the GNP-glass and MWCNT-glass composites with decreases of 19.9 and 27.3% compared with the 3 wt% specimen, respectively.

Figure 4

Flexural stress and FCRR as a function of strain in GNP-glass and MWCNT-glass composites: (a) 0.5, (b) 1.5, (c) 3, and (d) 5 wt%.

GNPs and MWCNTs have higher moduli; therefore, they apply an even higher modulus to the composite [38,39]. The dissimilarity between the flexural and tensile moduli can be attributed to the deformation during the test, geometry, and dimensions of the specimen. The difference between the flexural strengths of GNP-glass and MWCNT-glass was caused by the stiffness between the layered structure of the GNPs and entanglement of the MWCNTs, respectively. It was possible to monitor the strain related to the bending configuration because the arrangement of the adjacent nanofiller varied with the tensile stress configuration. The FCRR increased exponentially with the increase in applied strain until breakage of the specimen, as shown in Figure 4. The GNP and MWCNT nanofillers further accelerated the breaking of the conductive network, causing an increase in the resistivity of the composites. Tunnelling conductivity is an important mechanism for obtaining a good electrically conductive and highly sensitive strain sensor for the specimen [2]. The curve for FCRR started to increase linearly and became smooth for 3 and 5 wt% after the nanofiller content was increased, although the exponential curve trend was similar for all specimens, as shown in Figure 4(c) and (d). The smoother electrical response curves were attributed to the creation of a higher contact of conductive paths along the epoxy because of the higher number of nanofiller particles present in the composite. One study reported that smoother curves occurred after the addition of the nanofiller, and good sensitivity was achieved when the nanofiller content was close to p _c [36]. The minimum GNP and MWCNT nanofiller content added results in a good conductive composite.

3.4 Damage sensing behaviour using FCRR method

The stress applied to the specimen in the tensile and flexural tests resulted in a strain that caused an incremental increase in electrical resistance. The failure of testing corresponds to the breakage and damage detected as an increase in electrical resistance that comprises two stages, as shown in Figures 3 and 4. This damage behaviour depends on the disruption of the conductive network pathway. Macroscopic damage generally includes matrix cracking, fibre–matrix interfacial debonding, transverse cracking, delamination, fibre breakage, and fibre pull-out [14,40]. The damage evolution and FESEM images of the GNP-glass and MWCNT-glass composites are presented in Figure 5. There was an increase in the electrical resistance before the crack started to propagate. The breakage of the electrical network resulted from the formation of internal damages; even at high strain, the tunnelling effect plays a role in piezoresistivity. This is because the GNP and MWCNT particles start to separate from each other and induce new tunnelling links that result in a further increase in the sensitivity [6]. The FCRR curves in stage 1 (also called the “elastic region”) tend to have a more stable dependence on the applied strain up to 0.025 and 0.02 for the tensile and flexural test, respectively. This curve presents a linear relationship corresponding to the initial matrix microcracking in the composite, as shown in Figure 5(a) and (b), thus providing an early warning for the destruction of the specimen. The FCRR variation increases slightly and can be exemplified by the relationship between the electrical resistance and geometric properties of the specimen equation (2). When the load was applied to the specimen, the length started to increase, and the cross-sectional area started to diminish, leading to an increase in the electrical resistance.

Figure 5

Damage illustration and FESEM images for GNP-glass and MWCNT-glass composites: (a) and (b) matrix microcrack, (c) and (d) interfacial bonding and transverse crack, and (e) and (f) delamination, fibre breakage, and longitudinal crack.

The FCRR displays a more irregular shape as the strain increases during stage 2, which can be called the damage-sensing response (damage area). The microcrack growth, transverse crack, and fibre-matrix interfacial debonding are shown in Figure 5(c) and (d). The transverse crack and interfacial debonding probably induce more damage to the conductive network in the composite, as reported by Harizi et al. [41]. At a strain between 0.04 and 0.05, the FCRR curve rose rapidly at the end of stage 2. This failure behaviour of the specimen was caused by the continuous breakage of electrical pathways because of the rupture of the glass fibre, which restricts the electrical current flow.

Figure 5(e) and (f) illustrates the delamination, fibre breakage, and longitudinal cracking or splitting occurring after stage 2. This further breakup of the conduction path led to an increase in the resistivity of the composite. This phenomenon can be explained by the proposed deformation mechanism in a schematic form referred to in the modelling of the tunnelling effect and contact conduction mechanism. A previous study found that the increase in electrical resistance was linked to the failure behaviour of the specimen [14,42,43,44]. This failure may occur at the stages of transverse cracking, interfacial debonding, delamination, and glass fibre breakage. Under mechanical loading, the GNP and MWCNT nanofillers in the composite is oriented, cramped, and folded, resulting in electrical resistance changes [6,40]. The contact resistance between the conductive nanofiller resulting from the tunnelling effect was an important deformation factor of the resistance changes in the composite. A further increase in the strain caused the tunnelling conductive paths for the GNP and MWCNT nanofillers to be broken.

The FCRR exhibits an exponential variable at the beginning with the increase in strain, and it rises abruptly when the specimen fails, as illustrated in Figures 3 and 4. Specifically, the FCRRs of GNP-glass and MWCNT-glass with different nanofiller contents show nonlinear behaviour resulting from the tunnelling effect, which plays a dominant role in the electrical characterisation. As reported in previous literature, the piezoresistive mechanism of conductive composite-based strain sensors is described by the tunnelling effect [6,45,46]. The variation in the electrical resistance changes can be mainly attributed to different conductive mechanisms, such as tunnelling conduction and contact conduction theory. The possible tunnelling path and random modelling distribution were observed using FESEM, as shown in Figures 6 and 7. The tunnelling theory is derived based on Simmon’s formula presented in equation (3) to explore the strain-sensing behaviour of the conductive composite [6].

where L is the number of particles forming a single conductive path, R _t is the tunnelling resistance, N is the number of conductive pathways, h is Planck’s constant, d is the tunnel width, a is the effective cross section, e is the electron charge, m is the mass of the electron, and φ is the height of the potential barrier.

Figure 6

Tunnelling effect: (a) FESEM image of possible tunnelling effect of GNP and (b) modelling distribution of GNP in epoxy.

Figure 7

Contacting conduction: (a) FESEM image of possible contacting conduction among GNPs and (b) modelling distribution of GNP in epoxy.

According to tunnelling theory, electrons can tunnel through insulating material under certain conditions, resulting in transmission between the electrons of the conductive nanofiller [40]. The electrical resistance is altered owing to the separation of the conductive nanofiller and changes in the interparticle distance. In equation (3), the exponent indicates that an increase in the tunnel width can lead to an increase in the tunnelling resistance. This might explain the irregular shape with the changing tunnelling gaps resulting in varying strain, such as in stage 2. With further increase in strain, some tunnelling conductive paths break, increasing the tunnelling resistance. With a constant strain, the distance between adjacent nanofillers exceeds the tunnelling gap because of the deformation of epoxy; thus, the tunnelling conduction gradually disappears. This also suggests that the variation in the resistivity was contributed by the contact conduction. Figure 7 shows an FESEM image and the modelling distribution of the contacting conduction path between adjacent GNPs. For the GNP nanofiller with two contact points at i and ii, the resistance can be expressed as equation (4) [40].

where L _i ii is the distance between the two contacting points, σ is the electrical conductivity, and S is the cross-sectional area of the GNP. Based on equation (4), one can conclude that the resistance increases with increasing distance of the two contacting conduction points during loading. This may be the main reason why the FCRR exhibits an increasing change under a larger strain.

The gauge factor (GF) is a strain sensitivity coefficient that can be referred to as the ratio of the relative change between the fractional electrical resistance changes and the strain given by equation (5) [47]. The GF values of the GNP-glass and MWCNT-glass composites were calculated from the slope of the FCCR versus the strain graph, as shown in Table 2. The GF value was affected by the nanofiller content and strain applied during mechanical loading. The specimen with a lower weight content (0.5 wt%) exhibited a higher GF under the applied strain of tensile and flexural loading. The maximum value of GF achieved at 0.5 wt% was reduced with the further addition of GNP and MWCNT nanofillers. This larger GF was caused by the smaller number of conductive networks inside the composite; thus, the FCRR was larger. Moreover, the GF was contributed by the types and content of the conductive nanofiller, as well as the type of matrix reinforced with the fibre [40,48]. Thus, the values of GF of the GNP-glass and MWCNT-glass composites were sufficient to determine the sensitivity for strain monitoring (Table 3).

From recent literature, Table 4 represent the GF and strain range of carbon-based self-sensing composite (GNP and MWCNT). Xiang et al. [6] found a high gauge factor and stretchability, but the FCRR-strain response (Figures 3 and 4) in current work can identify the various failure modes of GNP-glass and MWCNT-glass composites. The current study’s GF and linearity are comparable to Xiang et al. [6] and Esmaeili et al. [11], but they has better stretchability. Moriche et al. [14], depending on the GNP content, demonstrated higher gauge factor and stretchability than the current result. Their electrical resistance response is exponential and linear, but difficult to determine the various modes of failure.

3.5 Surface morphologies of fractured specimens

After mechanical testing, the fractured specimens were characterised by FESEM. This test reveals the image state of the dispersion content and random distribution of GNP and MWCNT nanofillers. Figure 8 shows the surface morphology of the specimen for the glass, GNP-glass, and MWCNT-glass composites to understand the interfacial interaction between the nanofiller and epoxy. The smooth surface and homogeneity of the glass composite can be recognised as a freeze–fracture surface from the rich epoxy area in the absence of nanofiller, as shown in Figure 8(a). The presence of nanofiller that penetrated the glass fibre formed conductive networks throughout the composite, as indicated by the red circle in Figure 8(b). The image with nanofiller in Figure 8(c)–(f), for the GNP-glass and MWCNT-glass composites, shows an uneven fracture surface that made the surface rougher. The rougher surface was tougher. This rough surface indicates a longer crack growth path that requires high energy for crack propagation and hence, enhances the mechanical properties of the GNP-glass and MWCNT-glass compared with the glass composite. The higher strength needed could result from the strong interfacial interaction between the GNP and MWCNT nanofillers within the epoxy [33,49]. When the nanofiller content is increased, a good distribution is difficult to obtain owing to the appearance of agglomeration. This results from the difficulty of stirring a bundle of all cluster nanofillers from the π–π interaction in the GNPs and van der Waals forces in the MWCNTs. Uniform dispersions of GNPs and MWCNTs were observed within the epoxy, as shown in Figure 8(d) and (f). Thus, increasing the weight percentage loading of GNP and MWCNT nanofiller to the optimal amount yields a nearly uniform distribution. This implies a certain level of adhesion, which indicates the effective stress transfer from the epoxy to the nanofiller. This leads to an improvement in the tensile and flexural properties. Although a high degree of agglomeration must be avoided during the preparation of the composite, agglomeration zones with diameters larger than 10 µm are needed to achieve sufficient electrical conductivity [50]. Furthermore, the clusters of GNP and MWCNT nanofillers can contribute to a higher electrical conductivity and lower p _c. The good dispersion of the GNPs was attributed to the two-dimensional planar geometry, which resembled the layered clay structure in the epoxy [39]. This simplified process was suitable for the fabrication of GNP-glass and MWCNT-glass composites for proper sensor functionality, as reported by Sam-Daliri et al. [36].

Figure 8

FESEM images: (a) fracture surface of glass, (b)–(d) GNP-glass, and (e)–(f) MWCNT-glass composites.