Mechanoelectrical Effects in Natural Fiber-Reinforced Polymers as Structural Health Monitoring

Abstract

Natural fiber-reinforced polymers are gaining popularity as sustainable structural materials. However, their inherent variability can limit their reliability in load-bearing applications. To address this issue, we investigate a novel structural health monitoring method that leverages mechanoelectrical effects in flax fiber-reinforced epoxy composites. In our study, a contactless capacitive coupled measurement setup records electrical polarization during fatigue testing at four load levels. The polarization signals we observed increased with increasing load levels. Additionally, changes in polarization correlate with changes in dynamic modulus, providing early indicators of potential failure. This work lays the foundation for a new type of structural health monitoring in natural fiber-reinforced polymers.

1. Introduction

Natural fiber-reinforced polymers have great potential due to their positive carbon footprint [1] and mechanical properties [2], which can be on the order of magnitude of glass fiber-reinforced polymers [3,4]. The major disadvantage of natural fiber-reinforced polymers is the fluctuation in material properties caused by the natural plant fibers [5,6] that replace synthetic fibers such as glass fiber or carbon fiber in the composite. Therefore, structural health monitoring systems [7] are essential for using natural fiber-reinforced polymers in load-bearing structures.

The authors recently published a paper presenting a dynamic deformation sensor for natural fiber-reinforced polymers [8]. This work extends beyond the earlier published work in [8] by evaluating the mechanoelectric effect as a structural health monitoring system.

Several methods can be used to evaluate the service life of fiber-reinforced polymers in a non-destructive manner [9]. Baron et al., for example, developed a system for natural fiber-reinforced polymers based on an embedded copper conductor. This sensor system uses high-frequency alternating currents to estimate the boundary layer condition between the conductor and the composite. Afterward, the condition and the composite’s remaining service life can then be estimated [10,11], but the correlation between the boundary layer and the remaining service life of the composite requires further research. Additionally, larger conductor diameters result in a more significant effect size, but may negatively impact the composite material properties. Another approach uses electrical conductivity by incorporating conductive filler materials, such as carbon nanotubes (CNTs), into the liquid resin [12,13]. When damage occurs within the composite, the conductivity changes upon the application of voltage. This method allows for the detection of fiber breaks, matrix cracks, and delamination. However, it has a significant drawback: the manufacturing of the composite can be significantly complicated by the amount of filler used [13]. The nanofillers tend to agglomerate, necessitating the use of special mixing methods to achieve a uniform dispersion [13]. Additionally, the viscosity of the matrix increases with the amount of filler, leading to higher porosity and consequently a decrease in compressive strength [14]. To address this issue, a complex degassing procedure is required for matrices with high filler content [14]. Recording the acoustic emissions of composite materials using piezoelectric acoustic sensors is also feasible, which typically involves multiple sensors that can detect acoustic events and ascertain the location of damage based on wave velocity [15,16]. These acoustic sensors must be integrated into the material to achieve an optimal signal-to-noise ratio [17]. However, this integration presents a challenge, as the sensors can affect the mechanical properties of the composite [18]. Additionally, the ceramic structure of piezoelectric sensors makes them susceptible to damage under strain [19]. Techniques that utilize optical measurements include digital image correlation [20], X-ray radiography [21], and shearography [22]. However, these methods frequently face challenges due to high equipment costs and their sensitivity to environmental factors such as lighting, vibrations, and rigid body motion [23,24].

In contrast to the previously discussed methods, our approach leverages mechanoelectric effects within the composite material, which can be measured using capacitive coupling techniques. This method enables a comprehensive characterization of the overall condition of the composite material. Furthermore, our method provides the additional benefit of flexible sensor deployment, that is sensors in the form of conductive plates or foils, that can be integrated into the composite. As a result, it enables quantifying changes in the material’s mechanical properties through electrical signals, offering valuable insights into the composite’s condition in an easy, cheap, and safe way. This work serves as the initial proof of a phenomenon that paves the way for new methods of monitoring the service life of composite materials. Additionally, we wanted to gain a deeper understanding on the mechanoelectrical effect more precisely.

2. Materials and Methods

2.1. Manufacturing of Flax Fiber-Reinforced Polymer Samples

Compression molding was selected as the manufacturing process for the natural fiber-reinforced polymers, as the material properties in terms of stiffness and strength are superior or identical to most other processes such as hand lay-up, resin transfer molding, or vacuum infusion [25,26]. The samples were produced using a heated press with plates measuring 150 × 200 mm², limited by the dimensions of the heating surface. Each laminate consisted of 8 layers of unidirectional flax (BComp ampliTex [27]), each fiber was itself twisted from individual fibers [28,29]), infused with Entropy Resins 305 epoxy [30], and combined with Entropy Resins CPF-fast hardener [31]. The weights of the flax fibers and the epoxy resin were measured using a Kern digital scale [32] (0.01 g tolerance) prior to manufacturing. During the assembly process, epoxy resin was applied to the flax layers after every two layers to ensure uniform distribution [33]. The curing process was conducted at a temperature of 82 °C [30,31] under a pressure of 13.6 bar, which led in our previous tests to the best material properties and a sample thickness of 2.35 ± 0.2 mm. The completed composite panel underwent milling procedures using a milling machine to produce 170 × 17 mm² samples. These dimensions were taken from DIN EN ISO 574-4 [34] and are compatible with other common static and dynamic standards such as ISO 13003 [35], ASTM D3039 [36], and ASTM 3479 [37]. The fiber volume fraction of all samples was 50 ± 2% according to the calculation method presented by Roe et al. [38]. Subsequently, the samples were sealed under vacuum conditions at room temperature until further tests began to mitigate environmental influences, such as moisture absorption [39].

2.2. Dynamic Investigation of the Composite Material

A fatigue test was performed on the Instron 8872 Servohydraulic fatigue testing system using a tension–tension setup to measure the mechanical and mechanoelectrical effects. The fatigue testing system is equipped with hydraulic clamping jaws. The fatigue test follows standard testing procedures as defined in ASTM D3479 [37], applying load levels of 50%, 60%, 70%, and 80% of the Ultimate Tensile Strength (UTS), with an r-ratio of 0.1, consistent with other studies [25,40]. Prior to the fatigue test, tensile tests were conducted according to ASTM 3039 [36] to determine stress levels, resulting in a UTS of 360 MPa. The maximum and minimum stress levels of a sinusoidal load correspond to

$$\begin{array}{c}{\sigma }_{max}=\alpha ·UTS\end{array}$$

(1)

$$\begin{array}{c}{\sigma }_{min}=R·{\sigma }_{max}=R·\alpha ·UTS\end{array}$$

(2)

where $UTS$ refers to the Ultimate Tensile Strength of the material, $\alpha$ is the load level as a fraction of UTS, R is the r-ratio, ${\sigma }_{max}$ is the maximum stress, and ${\sigma }_{min}$ is the minimum stress, which was used during the test. To accelerate the general fatigue test, using the highest possible frequency similar to [40,41] is important. However, because the temperature increase depends on both the load level and the frequency [42,43], it is not feasible to use an arbitrarily high frequency. The frequency at different load levels was chosen so the temperature would not exceed 50 °C [43]. One drawback of this approach is that the elevated temperature can accelerate the aging process even at lower loads [43,44,45]. Consequently, it can be concluded that the number of cycles performed at lower load levels and higher frequencies should be considered conservative. In other words, the samples would withstand at least this number of cycles at the same load level but with a lower frequency. Preliminary tests determined the following frequencies: 5 Hz for 80% and 70% of the UTS and 10 Hz and 20 Hz for 60% and 50% UTS, respectively. For the mechanical evaluation, we selected the normalized dynamic modulus $E_{norm}^{\ast }$, $tan\delta$, and displacement as key figures. The dynamic modulus is commonly referenced in the literature [41,46,47,48] and serves as a preferred method for investigating the viscoelastic behavior and stiffness characteristics of materials under cyclic loading. Furthermore, normalization enables the comparison of different load levels with one another. Our normalized dynamic modulus is defined as

$$E_{norm}^{\ast }={\displaystyle \frac{E^{\ast }}{E_{0.2}^{\ast }}}$$

(3)

where $E^{\ast }$ refers to the dynamic modulus and $E_{0.2}^{\ast }$ refers to the dynamic modulus at the 0.2 normalized cycles. In all tests, a transient phase was observed at the beginning of loading, during which both mechanical and electrical variables exhibited disproportionately high fluctuations. Therefore, instead of normalizing to the first cycle, which could amplify noise and obscure trends, the dynamic modulus was normalized to the value at 0.2 normalized cycles. This point was selected to ensure that the data reflect stabilized material behavior and enable more robust comparisons. This relative normalization enables consistent intra-sample comparisons across both mechanical and electrical domains. In the results, we specifically compared 0.2 and 0.95 normalized cycles across the samples, as these points indicate stable and end-of-life behavior, respectively. Additionally, the absolute values at 0.2 normalized cycles were compared based on various parameters such as load level and frequency. The loss factor $tan\delta$ represents the ratio of the loss modulus to the storage modulus ($tan\delta ={\displaystyle \frac{E^{″}}{E^{\prime }}}$) [37]. Flax fiber-reinforced polymers typically exhibit a higher $tan\delta$ compared to other composites [49,50,51], which suggests a relatively greater loss modulus in relation to the storage modulus. This characteristic leads to increased damping and heat build-up during dynamic loading of natural fiber-reinforced polymers [52,53]. The displacement observed throughout load cycles enables the assessment of time-dependent deformation behavior, allowing for the early detection of damage mechanisms within the material. Notably, abrupt changes in length amplitude serve as indicators of critical conditions, signaling the potential onset of crack formation or structural failure. A total of 12 tests were conducted per load level, resulting in a total of 48 tests. Two datasets from every load level were chosen to illustrate the mechanical and electrical behavior. The total test duration was 3 months for all 48 tests.

2.3. Electrical Measurement Setup

The electrical response of the specimen was measured by utilizing a plate capacitor setup. The polarization resulting from mechanical deformation in the material leads to charge displacement inside the material, which can be observed as current flowing from one electrode to the other, which is then measured. When demonstrating the sensor principle in [8], an embedded copper conductor represented one electrode and the surrounding test setup the other. However, this approach presents several significant drawbacks, primarily the risk of conductor detachment from the composite, which cannot be ignored. The resulting microphonic effects [54] from this setup may interfere with other phenomena, such as piezoelectric or flexoelectric effects, complicating the accurate identification of the specific effect being measured. Furthermore, should the integrated conductor become detached from the composite material during dynamic mechanical analysis, it could introduce additional unforeseen effects. As a result, this setup might be more indicative of the boundary layer between the conductor and the composite (cf. [11]), rather than the composite itself. Consequently, utilizing an integrated conductor is not advisable at this stage of the research.

Instead, we implemented a contactless measurement method where no microphonic effect or triboelectric effect can occur due to an integrated conductor (cf. Figure 1a,b). This method offers the significant advantage of not impacting the composite material, thereby avoiding the introduction of additional mechanoelectric effects. Two electrodes were placed on opposite sides of the sample, creating a capacitor with three stacked dielectrics: air, NFRP, air. One electrode was connected to the central conductor of a triaxial cable, while the other was connected to the inner shielding of the same cable. To minimize noise, it was crucial to employ additional shielding around the test setup in the form of a Faraday cage, which was also connected to the inner shielding of the triaxial cable. The inner and outer shields were short-circuited by a triaxial to biaxal connector. This triaxial cable was connected to a high-resolution ammeter (Keysight B2987B, Keysight Technologies, Santa Rosa, CA, USA). The following current range was chosen: 200 pA, with a resolution of 100 aA.

The lower clamping jaws of the fatigue testing machine were fixed, while the upper jaw was connected to the actuator of the fatigue testing machine. The actuator’s position was transmitted as a voltage signal to a data acquisition card (DAQ, NI-9234) with a sampling rate of 8533.33 Sa/s. The ammeter sent a trigger signal to the measurement card, synchronizing the position data and electrical signals at the beginning of each measurement. The ammeter can store a maximum of 100,000 measurement points, making the measurement of the complete test at once impossible. With a sampling rate of 1500 Sa/s, one measurement period equals 1 min. To reduce the evaluation workload, at the beginning of the study, it was chosen to increase the pause duration after each measurement (no pause at the test beginning, 2 min after 30 measurements, and 15 min after 40 measurements). Since significant changes in the signaldid not occur at the beginning of each test, as expected, the sampling rate was decreased to 750 Sa/s, and no pause between measurements for the load levels of 50% and 60% UTS, resulting in only a few seconds interruption between measurements due to data transmission and resetting of the ammeter. As a result, sudden events may not be recorded for specimen reaching a higher end-of-life of the load levels of 70% and 80% UTS. After each measurement, the position and electrical data are transmitted to the computer. The mechanical parameters recorded by the Instron Servohydraulic testing system can be synchronized with the measurement data of the electrical measurement setup using the position data.

2.4. Derivation of Composite Material Polarization

To enhance the comparability of future studies, the material’s polarization was calculated, which is common when working with mechanoelectrical effects [55,56,57], rather than relying on the measured current, as demonstrated in our previous work [8]. Therefore, the theoretically measured voltage between the two electrodes is calculated first, which can be derived from the line integral of the electric field [58]:

$$\begin{array}{c}{V}_{meas}={\int }_a^b\mathbf{E}·\mathbf{dl}\end{array}$$

(4)

where E is the electric field vector and $dl$ is the infinitesimal line element along the path from point a to point b. Only the component of the electric field that acts in the direction of the surface is taken into account. The total voltage can be expressed as (cf. Figure 1c)

$$\begin{array}{c}{V}_{meas}=E_{air}{l}_{air}+E_{nfrp}{l}_{nfrp}+E_{air}{l}_{air}\end{array}$$

(5)

$$\begin{array}{c}{V}_{meas}=2E_{air}{l}_{air}+E_{nfrp}{l}_{nfrp}\end{array}$$

(6)

where $E_{air}$ and $E_{nfrp}$ are the electric fields in air and the composite material, respectively, while $l_{air}$ and $l_{nfrp}$ represent the air gap between the electrode and the composite surface and the thickness of the composite, respectively. The electric field in air, $E_{air}$, is determined using the electric displacement field [58]:

$$\begin{array}{c}{E}_{air}={\displaystyle \frac{D_{air}}{{\epsilon }_{air}{\epsilon }_0}}\end{array}$$

(7)

where $D_{air}$ is the displacement field in air, ${\epsilon }_{air}$ is the air permittivity, and ${\epsilon }_0$ is the vacuum permittivity. The relative permittivity of air is assumed to be one. For the composite material, the electric field $E_{nfrp}$ is defined in terms of both the displacement field and the material’s polarization [58]:

$$\begin{array}{c}{E}_{nfrp}={\displaystyle \frac{D_{nfrp}-P_{nfrp}}{{\epsilon }_0{\epsilon }_r}}\end{array}$$

(8)

where $D_{nfrp}$ is the displacement field in the composite, $P_{nfrp}$ represents its polarization, and ${\epsilon }_r$ is its relative permittivity. Since no free surface charges are present as a decharging procedure is performed upfront (see next section), Gauss’s law states that

$$\begin{array}{c}{D}_{nfrp}=D_{air}=D\end{array}$$

(9)

Substituting this into the voltage equation gives

$$\begin{array}{c}{V}_{meas}=2{\displaystyle \frac{D}{{\epsilon }_0}}{l}_{air}+{\displaystyle \frac{D-P_{nfrp}}{{\epsilon }_0{\epsilon }_r}}{l}_{nfrp}\end{array}$$

(10)

Under short-circuit conditions, where $V_{meas}=0$, we solve for the polarization $P_{nfrp}$:

$$\begin{array}{c}{P}_{nfrp}=D\left(2{\epsilon }_r{\displaystyle \frac{l_{air}}{l_{nfrp}}}+1\right)\end{array}$$

(11)

Since the system is modeled as a parallel-plate capacitor (neglecting edge effects), the displacement field D is given by

$$\begin{array}{c}D={\displaystyle \frac{Q_{meas}}{A}}={\displaystyle \frac{{\int }_0^t{I}_{meas}dt}{A}}\end{array}$$

(12)

where $Q_{meas}$ is the measured charge, $I_{meas}$ is the measured current, and A is the electrode surface area. Substituting this into the polarization equation yields

$$\begin{array}{c}{P}_{nfrp}={\displaystyle \frac{{\int }_{t_0}^t{I}_{meas}dt}{A}}\left(2{\epsilon }_r{\displaystyle \frac{l_{air}}{l_{nfrp}}}+1\right)\end{array}$$

(13)

For the measurements, the area of the electrodes A, the distance of the electrode to the material $l_{air}$, and the thickness of the material $l_{nfrp}$ is given by geometry. In the following, it is assumed that the variable $l_{nfrp}$ remains constant. With a Poisson’s ratio ${\nu }_{12}$ of approximately 0.4 [59,60] and a maximum mean displacement of approximately 2 mm, the change in thickness corresponds to about 11.05 $\mu m$ based on our sample geometry. All that remains is the measured current $I_{meas}$ and the relative permittivity ${\epsilon }_r$. Several papers have studied the permittivity of flax fiber-reinforced polymers with epoxy resin [61,62,63,64]. Our experiments require frequency ranges of 5–20 Hz and temperatures between 20 and 40 °C. However, none of these studies provide the specific combination of temperature range, manufacturing process, and frequency range required for our tests. As a result, it is necessary to conduct additional experiments, which are detailed in Appendix A.

2.5. Procedure Before the Experiment

Before every test, the specimen and surrounding test equipment (as seen in Figure 1b) were discharged using an ionizer benchtop Panasonic ER-Q to ensure no static charges on the specimen or the test setup are responsible for the measured current signal instead of the internal effects of the specimen. After discharge, the actuator was moved with only the top clamp closed, so no mechanical force acts on the specimen, to see whether any remaining static charges or disturbances from the test equipment would lead to false measurements. An example of the current signal before and after ionization can be seen in Figure 2a,b. Before ionization (a), static charges on the specimen or actuator cause a current signal according to the movement of the specimen and actuator. After ionization (b), this is reduced to a negligible minimum.

2.6. Data Processing

The evaluation was conducted using Matlab (2023b). For mechanical analysis, evaluation was performed automatically for each cycle using Instron WaveMatrix 2. Regarding mechanoelectrical analysis, polarization was calculated using Equation 13. Subsequently, the norm of the resulting complex polarization was calculated. Finally, the root mean square (RMS) polarization was determined for each cycle. The cycles were normalized using the respective end-of-life cycle. Both the dynamic modulus and the polarization were normalized based on the respective values at 0.2 normalized cycles. A summary of all values used in the normalization process can be found in Appendix C.

3. Results and Discussion

3.1. Mechanical Analysis

In the initial step, the mechanical properties of two selected datasets per load level were analyzed and compared, including $E^{\ast }$, $tan\delta$, and material displacement, with those documented in the literature. This comparison aims to assess the degree to which our findings align with previously reported results by other authors. Figure 3a–d illustrates the normalized dynamic modulus $E^{\ast }$ at various load levels for two tested samples per load level. During the first approximately 1000 cycles, all samples exhibit strain hardening, which increases the dynamic modulus by about 0.05. After this initial phase, the dynamic modulus shows a slow increase for the 70% and 80% load levels. This hardening effect is often attributed to the reorientation of cellulose fibers in the composite [42,47,65]. After the initial rise, 50% and 60% load levels experience a nearly monotonic decrease for most of the service life. Ueki et al. describe a combination of the actual stiffening effect and the softening that can be caused by damage to the fiber, which may occur during fatigue tests, explaining our observed behavior for loads of 50% and 60% [65]. Notably, the number of cycles to failure increases as the load level decreases. Detailed information on the exact number of cycles can be found in Appendix C. Just before reaching maximum service life, around 0.95, the dynamic modulus sharply declines across all load levels. Other authors have observed a significant increase in acoustic emissions in this region [41,66,67]. In some instances, the samples exhibit a step-like progression in the dynamic modulus. This phenomenon has also been noted in multiple other studies [41,47,48,68]. This behavior may be due to significant damage, such as delamination or fiber breaks [42,48,69,70].

Figure 3e–h illustrates the behavior of $tan\delta$ at different load levels. In our measurements, the value of $tan\delta$ starts at approximately 0.08 and decreases approximately linearly with the number of cycles in logarithmic scale, regardless of the applied load levels. The consistent starting values for $tan\delta$ suggest that the samples were at similar temperatures at the experiment’s beginning. Alternatively, the samples could have different temperatures, but $tan\delta$ remains insensitive within this temperature range, which could also be the case because the typical sensitive temperature range of flax composites is at higher temperatures [71,72]. At lower load levels and the increasing number of maximum cycles, $tan\delta$ can fall to lower values before the break. For example, $tan\delta$ drops to about 0.05 at 80% UTS, 0.04 at 70% UTS and 0.03 for 60% and 50%. However, towards the end of the testing, $tan\delta$ sharply increases, which is attributed to material breakage.

Figure 3i–l illustrate the displacement of the tested samples. Each load level initiates at a different displacement due to the initial ramp, which elevates the sample to the appropriate mean load level. At high load levels of 80% and 70% UTS, the displacement increases rapidly after the substantial initial increase. For the 80% load level, the displacement at break reaches approximately 2000 $\mathsf{\mu }$m, where a notable increase is observed. In contrast, for the lower load levels (60% and 50%), the displacement increases steadily after the initial significant change. Towards the end of the service life, there is a sharp increase in displacement, indicating the break of the composite. In some cases (e.g., Figure 3 in blue), the drop in the dynamic modulus results in a stronger sudden increase in displacement. This observation is not consistent with all conducted measurements (e.g., Figure 3 in red). In general, a characteristic S-curve can be recognized due to the strong change in displacement at the beginning and end of the service life [40,48]. In summary, it is notable that the observed mechanical behavior is consistent with already published results in [40,41,47,48,68].

Figure 4 illustrates the results of all 48 tests conducted. The mechanical key figures presented were all recorded at 0.2 normalized cycles. In Figure 4a, the dynamic modulus is plotted in relation to the load levels. A one-way ANOVA ($\alpha$ = 0.05) indicated that there is no significant change in the dynamic modulus at load levels of 50% and 60% of the UTS. However, a significant decline was observed at 70% and 80% of the UTS relative to 50% and 60%. As verified in Figure 4b, the different excitation frequencies at load levels of 50% (20 Hz) and 60% (10 Hz) UTS do not significantly influence the dynamic modulus. The higher load levels (70% and 80% UTS) were tested at a consistent frequency of 5 Hz, which shows a significant change in dynamic modulus, relative to 50% and 60%. In Figure 4c, $tan\delta$ is plotted as a function of the applied load level. The ANOVA indicated a statistically significant effect of the load level on $tan\delta$. While $tan\delta$ remains relatively stable between 50% and 60% UTS, a clear and significant increase is observed at 70% and 80% UTS. Figure 4d illustrates the effect of excitation frequency on $tan\delta$. A significant decrease in $tan\delta$ is observed with increasing frequency. The values measured at 5Hz are significantly higher than those at 10 Hz and 20 Hz. In Figure 4e, displacement is shown as a function of load level. Displacement increases progressively and significantly with increasing UTS, with the highest value observed at 80% UTS. Figure 4f shows displacement in relation to excitation frequency. A significant reduction in displacement is evident with increasing frequency, with the largest displacement recorded at 5 Hz and the smallest at 20 Hz. In summary, the mechanical characterization reveals that increasing load levels beyond 60% UTS result in a significant decline in dynamic modulus and a corresponding rise in both $tan\delta$ and displacement. In contrast, excitation frequency primarily affects $tan\delta$ and displacement, with higher frequencies leading to significant reductions in both while having no notable impact on dynamic modulus at lower load levels.

When examining the change in ratios from 0.2 to 0.95 (cf. Figure 5), it is evident that the change in the dynamic modulus decreases as the load level increases, eventually stabilizing at the highest load levels of 70% and 80%. In contrast, the change in dynamic modulus increases with rising frequency (cf. Figure 5a,b). Additionally, the change in $tan\delta$ increases with higher load levels but decreases with increasing frequency (cf. Figure 5c,d). Notably, most changes in displacement occur at the maximum load level of 80%, with a corresponding decrease as the UTS declines.

3.2. Mechanoelectrial Analysis

This section compares mechanical and electrical measurements at all load levels. First, eight samples, two per load level, are compared over their entire service life. Figure 6 displays the relationship between normalized dynamic modulus and RMS polarization over normalized cycles for specimen tested at 80% (a and b), 70% (c and d), 60% (e and f), and 50% (g and h) UTS. For each sample, the polarization and displacement data are provided for the two points of interest (0.2 and 0.95 normalized service life). Across all load levels, the dynamic modulus typically shows a sharp increase at the onset of testing, followed by a period of gradual increase or decrease. This trend often culminates in a significant drop in dynamic modulus. In contrast, the RMS polarization exhibits a more varied behavior: it may display a slight monotonic decline, plateau, or even an initial decrease followed by a rise or fall in later stages. Nonetheless, substantial changes in RMS polarization consistently correspond to abrupt shifts in the dynamic modulus. Notably, the sample depicted in Figure 6h exhibits minimal changes in both dynamic modulus and RMS polarization, as it did not break within the test’s stopping criterion of 2 million cycles. Although the waveform shape and the phase relationship between polarization and displacement remain relatively constant throughout the service life, there is a pronounced change in amplitude (see details at Markings I and II). This suggests that monitoring the signal strength alone is sufficient to detect variations in the composite material. Overall, the consistent correlation in timing and nature of the changes, particularly the synchronized signal alterations between dynamic modulus and RMS polarization, underscores the effectiveness of the electrical signal as an indicator of changes in material.

In the following sections, the behavior of all selected tested samples is examined in more detail. It is necessary to investigate whether the RMS polarization is influenced by the frequency and load level used. Figure 7a illustrates the RMS polarization in relation to varying load levels. Overall, there is a slight increase in RMS polarization as the load level rises. At the higher load levels (70% and 80%), measured at an excitation frequency of 5 Hz (as shown in Figure 7b), the median RMS polarization increases with load level; however, none of these changes are statistically significant according to one-way ANOVA ($\alpha$ = 0.05). Meanwhile, Figure 7b presents the RMS polarization in relation to the excitation frequency, revealing a decrease in the median as frequency increases, which is also not significant according to one-way ANOVA. Consequently, no significant influences of frequency or load level on the RMS polarization can be observed. Since the RMS polarization behaves similarly to the dynamic modulus (see Figure 6), we compare these two key figures in the following. To achieve this, the mechanical and electrical key figures at 0.2 and 0.95 normalized cycles were taken (dynamic modulus and RMS polarization) and the absolute percentage change for each was calculated. The advantage of using absolute percentage change is that it does not depend on whether the polarization is increasing or decreasing, and the varying absolute values do not affect the resulting key figure. From a sensory perspective, the ratio of the absolute change in polarization to the absolute change in dynamic modulus reflects the sensitivity of the method, as shown in Figure 7c. A trend is apparent: greater changes in stiffness are associated with larger changes in polarization, showing a statistically significant positive relationship (p < 0.001, $R^2$ = 0.434). The samples exhibiting the least sensitivity, meaning they showed minimal changes in polarization despite significant changes in stiffness, were subjected to a load level of 50%. These samples reached the maximum cycle count of 2 million, and testing was terminated before breakage occurred. Overall, changes in stiffness always result in a greater change in polarization. This study also reveals a high level of variance, which is also evident in the polarization results (see Figure 7a,b). Nevertheless, a correlation between the two examined key figures can be observed. The calculated sensitivity corresponds to 4.7 (% polarization per % dynamic modulus).

In summary, the change in RMS polarization serves as an indicator of changes in the dynamic modulus, which in turn reflect changes in stiffness. Both increases and decreases in polarization can be detected. Additionally, a minimum change of approximately 2% in dynamic modulus is necessary for this to affect the polarization significantly.

3.3. Cause of the Electrical Effect

When the authors first discovered the mechanoelectrical effect, several potential origins were considered [8], including the triboelectric effect, electrostriction, piezoelectricity, and the flexoelectric effect. The following discussion assesses whether these mechanisms could account for the measured signals.

In the authors’ previous investigation, a copper conductor was integrated into the composite [8]. In contrast, in this case, a contactless setup was chosen, which by design cannot facilitate macroscopic contact–separation events, as required for effective charge seperation in Triboelectric Nanogenerators [73,74] or as observed in Cable Microphonics [54,75]. Although the absence of a directly attached electrode eliminates conventional charge separation mechanisms, polarization effects may still arise from internal friction, such as interactions between flax fibers and the surrounding epoxy matrix. However, relative motion between the fibers and the matrix is highly restricted, as the flax fibers become embedded in and absorb the resin [76,77]. Analogous to triboelectric nanogenerators, any charge transfer would require repeated contact (contact mode [78]) rather than a single sliding motion [78]. Sliding-induced effects are transient and non-repetitive, whereas contact mode can lead to a repeatable electrical response [79], suggesting the presence of a load-dependent capacitance at the interface between the fiber and the matrix, similar to cable microphonics [54,75]. In this unidirectional composite, which is loaded along the fiber axis, the material undergoes thinning under load, compressing both the fibers and the matrix. Unlike triboelectric contact mode, however, this compression does not involve separation and recontact, making traditional frictional charge transfer unlikely.

Electrostriction occurs in all insulators [80], but inverse electrostriction (the conversion of mechanical energy to electrical energy) requires an electrical bias [81]. Since this bias is absent in our case, this effect was ruled out.

The flexoelectric effect can occur in any crystal symmetry [82], but it requires a strain gradient that disrupts local centrosymmetry [83]. This phenomenon typically manifests during bending and is measured using a bending setup [84,85]. Although flexoelectricity can also be observed under axial loading, it necessitates a gradient in mechanical properties, such as stiffness or thickness, along the specimen (cf. [86] and Appendix B). In natural fibers like flax, inherent structural heterogeneities often cause variations in both stiffness [87,88] and thickness [89], making them promising candidates for observing axial flexoelectric effects. Additionally, manufacturing imperfections such as skewing can further alter local stiffness, thus enhancing the flexoelectric response. Flexoelectricity has also been demonstrated in cellulose [84], the fundamental building block of plant fibers [90] and therefore our composite, particularly in thin films, but its magnitude decreases significantly as the thickness increases [91]. Furthermore, the flexoelectric coefficient is several orders of magnitude smaller than the piezoelectric coefficient [92]. Thus, while the flexoelectric effect cannot be dismissed entirely, it is unlikely to be the primary factor behind our observations.

The piezoelectric effect, in general described as linear electromechanical interaction between mechanical and electrical states [93], requires a non-centrosymmetric crystal structure [82,94], has been observed in cellulose. In cellulose, hydroxyl groups are arranged non-centrosymmetrically through hydrogen bonding, leading to the formation of a dipole moment [95,96,97,98]. Both the direct [99] and converse [100] piezoelectric effects have been demonstrated in bulk materials (solid wood). The strength of the piezoelectric effects can vary significantly depending on the fiber direction [100] and pre-treatment [95,101], which may explain the fluctuating material properties observed. Another indication of polarizability is the decrease in ${\epsilon }^{\prime }$ with increasing frequency (as seen in [64] and in Appendix A), a phenomenon characteristic of orientation polarization, as seen in dipolar molecular chains [84,102]. Furthermore, Knuffel et al. found a correlation between the piezoelectric effect and stiffness in solid wood, a relationship also reflected in our results [103]. Furthermore, the piezoelectric effect depends on the mechanical amplitude [104], which is also the case in our experiments. Therefore, the piezoelectric effect can be identified as a potential cause of our observations.

3.4. Future Challenges

In this study, a mechanoelectric effect was observed in 48 samples, showing its potential for use in structural health monitoring under laboratory conditions. However, several challenges need to be addressed to make this suited for real-life conditions, which is briefly discussed in this section.

First, the measurement setup presents a challenge. Conductive structures, such as foils and conductors, must be brought into proximity to the composite material, as contactless measurement is not always feasible. The introduction of these conductive structures may also influence the observed effect. Another challenge lies in the measurements themselves. The observed effects were relatively small and could be disturbed or obscured by electromagnetic coupling (EMC noise). A well-defined approach is necessary to ensure electromagnetic compatibility during measurements. Although a correlation between the measured signals and the stiffness was found, a problem can occur if the stiffness changes only slightly. This limitation could impact the effectiveness of our concept. Scaling the method to larger structures presents new challenges. In particular, very large structures (e.g., wind turbine blades) typically exhibit natural frequencies in the very low range. At such low frequencies, it is essential to ensure that the signal strength remains sufficient for reliable detection. As observed in Figure 7 and discussed in [8], the current amplitude decreases with decreasing frequency, which may limit the detectability of the mechanoelectric effect in large-scale applications. Furthermore, the duration of experiments is significantly affected by frequency. Lower frequencies result in longer cycle times, meaning it takes considerably more time to reach the same relative service life as in smaller-scale tests with higher frequencies. Therefore, it may be necessary to test the long-term stability of the signal and evaluate whether accelerated testing methods can be applied without altering the observed behavior. For larger samples, there is likely a trade-off between spatial resolution and measurement complexity. Higher spatial resolution requires a greater number of sensors or measurement points, which increases system complexity and installation effort. On the other hand, using fewer sensors reduces complexity but may miss localized damage. Moreover, in large structures, minor defects may not only remain undetected but can also be overshadowed by the overall mechanical behavior of the structure, making them harder to distinguish in the measured signals. To determine the optimal balance, future investigations should explore the use of multiple sensors on larger structures, examining signal quality, sensor placement strategies, and the system’s sensitivity to local damage. Finally, the variability of the actual measurement signals should be addressed. However, it is possible that this measurement method only allows for relative monitoring due to fluctuations in signal strength. Further research is required to improve our understanding of the effect and enhance the overall consistency of the signals obtained. One possible explanation for the observed phenomena is the direction in which the fibers grow. A controlled orientation of these fibers could potentially lead to more consistent results.

Future studies should incorporate systematic control experiments and quantitative validation to enhance the mechanistic understanding of the observed effects, which include investigating control groups with and without metallic conductors to eliminate potential triboelectric influences, as well as applying or shielding external electric fields. Additionally, key sample parameters such as fiber type, fiber thickness, fiber volume fraction, matrix material, and layup structure should be varied. Moreover, changes in loading modes (for example, bending versus tension) should be examined to differentiate between possible mechanisms like piezoelectricity, flexoelectricity, and electrostriction. Quantifying the contributions of each effect is essential for refining interpretations and improving reproducibility. Such efforts can also help determine the extent to which these effects are influenced by extrinsic versus intrinsic material properties, thereby establishing clear guidelines for material design and application. Overall, the method described here offers a significant advantage and represents an initial step toward developing a new type of structural health monitoring system for natural fiber-reinforced polymers. Since this approach does not require the integration of sensors, it does not alter the mechanical properties of the composite material. Furthermore, a correlation exists between changes in stiffness and changes in polarization, establishing a direct connection between mechanical and electrical parameters. This feature provides a noteworthy advantage over other structural health monitoring systems and could provide an even greater advantage when used in combination with other structural health monitoring systems.

4. Conclusions

In this study, a mechanoelectric effect, likely of piezoelectric nature, was investigated in natural fiber-reinforced polymer for the first time across four different load levels using contactless measurements during dynamic mechanical analysis. The results show that the measured polarization is not statistically significant in relation to load levels and correlates with the dynamic modulus. Ultimately, monitoring changes in the electrical signal allows for predicting upcoming failures of nearly all the samples. Overall, the method described here offers a significant advantage and marks an important step toward the development of a new class of structural health monitoring systems for natural fiber-reinforced polymers. Unlike conventional approaches, it does not require embedded sensors, thereby preserving the composite material’s mechanical properties. A clear correlation between changes in stiffness and polarization establishes a direct link between mechanical and electrical behavior. This feature provides a distinct advantage over existing systems and holds even greater potential when integrated with other structural health monitoring methods.