Development of fiber-based piezoelectric sensors for the load monitoring of dynamically stressed fiber-reinforced composites

Continuous load monitoring of fiber-reinforced composites represents a complex challenge for the composite sector. In this paper, the development and characterization of piezoelectric sensors for structural health monitoring applications based on polyvinylidene fluoride (PVDF) are presented. The basic sensor structures are melt spun bicomponent filaments in a core-sheath configuration. The core is a volumetric mixture of polypropylene (PP) and Pre-Elec PP, a PP-compound modified with carbon black. The sheath is made of PVDF. Three variants with increasing volumetric Pre-Elec ratio in the core (50 vol.%, 60 vol.% and 70 vol.%) are manufactured and analyzed by conducting optical and resistance measurements as well as tensile tests. In the subsequent process step, the bicomponent filaments are braided with copper fine wires. The PP core and copper braiding of the coaxial tricomponent yarn are used as inner and outer electrode, respectively. By generating an electrostatic high voltage field between these electrodes, the PVDF interlayer is subjected to contact polarization to initialize the piezoelectric effect. The piezoelectric characteristics and thereby the sensory potential of the produced sensor yarns are quantified and analyzed on the fiber and composite scale by investigating their piezoelectric behavior during cyclic tensile tests. The results of the different sensor yarn variants are compared with each other by evaluating the measured voltage signal as a function of the induced cyclic stresses.


Introduction
Recently, the demand of smart textiles and composites is increasing rapidly [1][2][3][4][5][6][7]. Functional integration like structural health monitoring (SHM) systems, consisting of various sensing and electrical components, enable the continuous load monitoring of complex stress scenarios of fiber-reinforced composites (FRC), which offers a promising potential in the age of digitalization, automation and Internet of Things [8]. The acquired sensor signals can be used to obtain and derive information about the condition of the structure, such as critical damaged areas or residual load-bearing capacity. This SHM approach ensures the functionality and integrity of FRC structures by continuously monitoring the structural state and enables the identification of potential damage at an early stage and the initiation of appropriate countermeasures.
At present, commonly used types of sensors for SHM applications are fiber Bragg grating (FBG) sensors, resistive strain sensors, and piezoelectric sensors [9]. FBG sensors are widely deployed in the construction sector, especially for the SHM of bridge structures. Li and Siwowski et al developed an SHM system for the smart monitoring of road bridges and derived that a distributed fiber optic sensor network proved its effectiveness for SHM of civil infrastructures [10,11]. The analysis of the measured data showed that the strains caused by normal vehicle loads are low, indicating a reliable and functional state of the bridge. However, the handling and installation of FBG sensors are complicated, manually labor-intensive and time consuming, which greatly increases the resulting costs.
Another commonly used sensor type are resistive strain gauges (SG) for the direct measurement of the strain response of FRC structures and components. Cui et al used multiple strain sensors on a Cantilever beam and showed that it is possible to accurately detect and localize the induced structural damage [12]. For strain measurements on and in FRC panels, Ruzek et al used FBG and SG sensors simultaneously and compared them to each other. In general, a good agreement between the FBG and SG data was achieved. The authors discussed, that the positioning of the SG significantly influenced the sensor response [13]. Furthermore, ambient conditions, such as temperature changes, humidity, chemicals and mechanical loads demand additional compensation and protection approaches. The installation is labor-intensive and critical for the reliable function of the SG. Another major disadvantage is the lacking structural compatibility and therefore in-situ strain measurement capability of SG. On the one hand, the SG can be applied to the structural surface with the limitation of only measuring surface strains. On the other hand, SG can be integrated into FRC structures, e.g. by gluing them to the textile reinforcement. This approach offers more insight information of the composite behavior but also influences the structure itself by an additional potential discontinuity [7].
Piezoelectric sensors based on polyvinylidene fluoride (PVDF) are widely deployed and have been extensively used in sensing applications especially as weight, pressure and strain sensor [9,[14][15][16] to capture dynamic events. PVDF is a polymorphic thermoplastic polymer, meaning that it can crystallize in α-, β-, γ-and/or δ-phases, respectively. In general, the non-polar and most stable α-phase is naturally formed, when the PVDF is cooled from a melt. The αphase can be transformed to polar β-phase by cold drawing e.g. within the melt spinning process. Thereby, the molecule chain is oriented along the stretching direction with a perpendicular but random orientation of the dipoles. Hence, the resulting dipole moment is zero. An additional polarization of the β-phase PVDF in a high electro-static field is required to ensure the piezoelectric material property. The β-phase shows the highest polarity and is essential for the piezoelectricity of PVDF [17,18]. In general, either corona-or contact polarization is used, whereby the most crucial poling parameters are time, temperature and electric field strength. The contact method requires two electrodes that are connected to a high voltage supply for polarizing the PVDF interlayer. Due to electrical breakdowns and flashovers, this poling process can be error-prone and difficult to handle. A constant PVDF thickness and optimal poling parameters are mandatory [19,20]. The non-contact corona-poling is realized by applying a high potential corona discharge near the specimen. The generated ions charge the surface of the specimen. By applying an electro-static field between the charged surface and its grounded counterpart, a polarization or an orientation of the dipoles respectively, is conducted. This method allows the poling process with only one electrode as well as higher electric field strengths before an electrical breakdown, compared to the contact poling [21].
Meng and Yideveloped a PVDF-based stress gauge to determine the stress-strain curve of concrete under impact [15]. The stress gauge showed a high sensitivity and signal quality. Piezoelectric sensors are generally predestined and only suitable for dynamic applications. However, most of the piezoelectric PVDF is used as a thin film [22] and is therefore not suitable for the processing in textile machines and only have a low structural compatibility in smart textiles and FRC. Lund, Nilsson and Walter et al presented first results of melt spun bicomponent PVDF fibers with a conductive core and characterize their manufacturing and sensing potential. However, the polarization and piezoelectric characterization of those fibers was realized by external electrodes e.g. between two metallic electrodes, by applying silver paint or copper films [17,18,21,23].
In contrast, the objective of this research work is the development of a new approach to manufacture fully fiberbased piezoelectric sensors for SHM applications. As mentioned, common monitoring methods in the field of SHM are often externally applied, less structural compatible and cost-intensive, such as SG, optical measurements, ultrasonic or x-ray-based systems [7]. However, textile-technologically processible and fully fiber-based sensors enable in-situ SHM, because of their high structural compatibility, for the identification of local structural stresses within the FRC, particular for dynamic loading scenarios and are also suitable for tactile sensing in smart textiles. The sensor manufacturing and textile-technological integration is highly time-and costefficient due to the high degree of automation and production speeds of textile machines. Based on melt spun bicomponent filaments consisting of an inner conductive modified polypropylene (PP) electrode and a PVDF sheath that represents the piezoelectric material, the outer electrode is applied by a braiding process. The result is a coaxial tricomponent structure with both a conductive inner and outer electrode separated by the PVDF interlayer. On the one hand, the sensor design can be advantageously used for contact polarization of the PVDF in an electro-static field, and on the other hand, it allows measuring the sensor signal as an electric potential shift (piezoelectric output voltage) generated by a dynamic stress. The piezoelectric characteristics of the sensor are investigated and analyzed in cyclic tensile tests on the yarn and composite scale.

Process chain for development of fiber-based piezoelectric sensors
The process chain for the realization of piezoelectric sensor yarns for SHM applications in FRC based on PVDF is schematically shown in figure 1. The presented approach is discussed in detail in the following sections.

Materials
The used granulate for the core is a volume mixture of Pre-Elec PP 1353 that is modified with carbon black (Premix GmbH, Finland) and conventional PP Moplen HR561R (Basell Polyolefine GmbH, Germany). The PVDF used for the sheath was Solef 1008 (Solvay Solexis, Milan Italy). The essential parameters of the materials are listed in table 1 and are according to the suppliers.
Copper fine wires with a diameter of 0.05 mm (Dahmen GmbH & Co. KG, Germany) and low electrical resistivity of 0.002 mΩ cm were used to realize the outer electrode of the bicomponent filaments in a textile braiding process. In a further step, the tricomponent sensor yarns were braided again with PA66 to ensure an electrical insulation to the textile reinforcement within the carbon fiber reinforced plastic (CFRP) specimen. A carbon fiber (CF) biaxial non-crimp fabrics with 1600 tex CF rovings (24 K) in warp and 800 tex CF rovings (12 K) in weft direction, with an area mass of 680 g m −2 according to the manufacturer (Gustav Gerster GmbH & Co. KG, Germany), was chosen. The CFRP specimen were manufactured by using a commerically available two component epoxy resin RIMR135/RIMH137 (Hexion GmbH, Germany).

Bicomponent melt spinning
The basis for the sensor yarns were melt spun bicomponent multifilaments in a core-sheath-configuration manufactured on a melt spinning plant (DIENES Apparatebau GmbH, Germany). Figure 2 shows a schematic representation of the bicomponent melt spinning plant, equipped with one twin screw extruder for the core, and one single screw extruder for the sheath. The high-viscosity polymer melt, especially caused by the high Pre-Elec PP 1353 volume in the bicomponent core,  required a high process pressure of up to p ext = 10 MPa. A relatively high extrusion temperature of the final zone, compared to the melting temperatures, of T ext = 245 • C was chosen for both extruders to reduce the melt viscosity and therefore p ext . The spinneret in figure 2(a) contains 60 holes with a diameter of 0.6 mm each and an aspect ratio of L/D = 3. The dosage rate of the polymers to the spinneret determines the relative amount of core and sheath material in the multifilaments [21] and was  set to approximate an equal volumetric ratio (core/sheath) of 50:50. An external winder was used since the built-in winder had a minimal take-up speed of 300 m min −1 , which was too high for the presented application. The optimal winder speed for a constant melt spinning process was determined empirically with focus on a high draw ratio (DR) [18,24] but also minimal fiber breakage and manufacturing interruptions, respectively. In [22,25] a DR of 3.0 up to 5.0 was reported. It was found that the transformation of α-into β-phases within the PVDF is very high at this range, resulting into β-phase amounts of 60% up to 100%. The winder velocities were iteratively increased during the melt spinning process until a stable production with no fiber breakage was ensured. With the empirical determined winder velocities and the constant extrusion speed of v 0 = 2.9 m min −1 , an approximate DR can be calculated using equation (1). Table 2 summarizes the relevant melt spinning process parameters. For the conducted investigations and manufacturing steps presented in this paper, the bicomponent multifilaments were separated into individual filaments and will be further referred to as bicomponent fibers for better readability. In this research, three different volumetric mixtures were

Determination of the electrical resistivity
In order to determine the dependence of the electrical conductivity of the bicomponent core on the volume fraction of Pre-Elec PP 1353 and the draw ratio, the electrical resistance was measured by using an electrometer (6517a/DM 2000, IPF Dresden) as illustrated in figure 3. The PVDF sheath of the bicomponent fiber (A) was carefully removed using scalpel tools and then mechanical fixed (B) into the electrometer. The defined measurement length (C) was 60 mm. The electrical resistivity ρ was then calculated with equation (2).
where R is the electrical resistance, A is the cross-section, d core is the core diameter of the bicomponent fiber and l is the measurement length.

Optical measurements
Additionally, a microscopic analysis was conducted to measure the diameter of the fiber core and to evaluate the PVDF layer thickness and quality by using a light microscope Axio Imager M1m with a digital color camera AxioCam MRc5 (Carl Zeiss AG, Germany) and the image processing software ImageJ. The determined value for the total diameter of the bicomponent fiber was then compared to a theoretically calculated diameter according equation (3). The required density was approximated using equation (4) (see table 3).
where ρ Bico ,ρ Core ,ρ Sheath ,ρ Pre-Elec and ρ PP are the density of the bicomponent fiber, core, sheath, Pre-Elec PP and PP, Tt f is the fineness of the bicomponent fiber and φ Pre-Elec and φ PP are the volumetric fraction of Pre-Elec PP and PP in the core.

Manufacturing of the outer electrical conductive and insulating layer
For the contact polarization of PVDF, but also for the measurement of the generated piezoelectric voltage, two electrodes are required. One approach was shown in [21,26], where the authors applied a silver paint matrix on the bicomponent fibers to realize the outer electrode. Another approach was conducted by [27], where the bicomponent fibers were embedded into a conductive thermoplastic matrix to form the outer electrode.
In this research work a copper fine-wire mesh, acting as electrical conductive outer electrode, was applied onto the bicomponent fibers using a circular braiding machine RU 2/12-80 (Herzog Maschinenfabrik GmbH, Germany), depicted in figure 4(a) to provide a fully textile-based tricomponent piezoelectric sensor yarn. The density of the braided structure was determined by the take-up speed of the bicomponent fibers. A braid density of 11.5 braids per cm was chosen and realized by using six braiding bobbins actively, to ensure an open braided structure. This ensured a compromise between sufficient outer electrode material for contact polarization and high residual flexibility for further textile processing. In addition, an open-mesh electrode improves the fiber-matrixadhesion of the piezoelectric PVDF sheath to the epoxy matrix when used in FRC components. Therefore, an excellent mechanical coupling of the sensory piezoelectric PVDF layer and the surrounding matrix of the FRC was ensured. This interaction is essential for transferring the structural strain of the monitored FRC component to the PVDF layer and should improve the generated piezoelectric signals. Such a completely textile-based approach is predestined for further lowdamage and function-retaining textile processing on standard textile machinery. Figure 4(b) shows a detailed view of  the braiding point where the bicomponent fibers were braided by copper fine wires. For a stable production process, the rotation speed of the braiding machine was set to 450 min −1 to reach production speeds of up to 0.2 m min −1 . Figure 4(c) schematically shows the realized structure of the tricomponent sensor yarn with the inner and outer electrode as well as the PVDF interlayer. An exemplary microscopic image is visualized in figure 10. Before the tricomponent sensors were integrated into the textile reinforcement, they were braided again to provide electrical insulation to the conductive base material as seen in figure 4(d).

Activation of the piezoelectric interlayer by contact polarization
A description of the required process steps for the applied contact polarization is provided in figure 5. Beforehand, the conductive core and the braiding structure were manually exposed with precision tools. The prepared tricomponent yarn was then wound on a bobbin and positioned into an oven. The contact polarization of the PVDF layer was performed based on the schematic setup visualized in figure 6. Using a bypass within the wall of the drying oven, the poling wires of the high voltage source HEK 100 (Maag Flockmaschinen GmbH, Germany) were inserted into the chamber. The inner (figure 6(a)) and outer electrodes ( figure 6(b)) of the tricomponent yarn were electrically contacted with micro clamps. Two samples were poled simultaneously by connecting them electrically in parallel to obtain the same poling parameters for both tricomponent yarns. Subsequent, the yarn was heat up to the specific poling temperature.
Several researchers [28][29][30] have investigated and varied the poling parameters for the polarization of PVDF. In general, a typical applied field strength and poling temperature is in the range of 50-300 MV m −1 and 50 • C-130 • C, respectively. The poling time varied in the range of seconds to minutes [21,22,25,31]. The values were mostly determined iteratively, depending on the PVDF thickness and appearance of electrical breakdowns. The poling parameters used in this research work are described in detail in 3.2.

Manufacturing of the CFRP specimen
The tetracomponent sensor yarns (figure 4(d)) were textile technologically integrated on CF non-crimp fabrics by using a tailored fiber placement machine as shown in figure 7(a) (SGY 0200-6500, ZSK Stickmaschinen GmbH, Germany). A total of two layers was required to realize a specimen thickness of at least 2 mm, according to the norm DIN EN ISO 14125. The sensor position was therefore approximately in the middle of the specimen thickness. The CF non-crimp fabrics with integrated piezoelectric sensors were impregnated with epoxy resin in a hand lay-up, tempered in an oven according to the datasheet of the epoxy resin and cut to a size of 250 mm in length and 25 mm in width. The specimen had an average thickness of 2.8 ± 0.2 mm. An exemplary CFRP specimen is shown in figure 7(b). One linear sensor per specimen was realized and the sensor over length was used to measure the generated piezoelectric voltage.

Piezoelectric characterization of the tricomponent yarn on fiber and composite scale
For the investigation of the piezoelectric properties of the tricomponent yarns on fiber and composite scale, cyclic tensile tests were conducted on a tensile tester (ZmartPro, ZwickRoell GmbH & Co. KG, Germany).
2.9.1. Fiber scale. Therefore, the yarns were fixed into electrical insulated clamps and were electrically contacted with micro clamps outside of the testing length to avoid mechanical influences or interruptions at the contacting points. The specimen length were l 0 = 500 mm for PrEl50 and PrEl60 and 200 mm for variant PrEl70 due to insufficient functional fiber lengths. A displacement-controlled rectangular wave with a maximum testing speed of v = 1000 mm min −1 was chosen for the testing procedure, that corresponds to a testing frequency of approximately 0.8 Hz. A preload force of 0.5 N was applied and a total number of 60 cycles were performed in each test. An appropriate testing regime was derived by conducting preliminary tensile tests of the bicomponent fibers to ensure minimal plastic deformations during the cyclic tests. The generated voltage changes induced by the piezoelectric PVDF due to stretching and relaxation were measured simultaneously with a multimeter 8864a (Fluke GmbH, Germany). A schematic representation of the testing setup is shown in figure 8(a).

Composite scale.
A similar testing setup was used for the three-point bending tests of the CFRP specimen on the composite scale ( figure 8(b)). The sensor functionality was investigated in the linear elastic regime. Preliminary tests showed that a flexural deformation of 1 mm is appropriate. The testing configuration was chosen according to the norm DIN EN ISO 14125. The preload force, testing speed and total number of cycles were 1 N, 1000 mm min −1 and 30 respectively. The distance between the supports was 80 mm. An exemplary specimen is seen in figure 8(c). An additional piece of paper was placed between the specimen and loading pin to ensure an electrical insulation.

Characterization of the melt spun bicomponent fibers
3.1.1. Tensile test. Uniaxial tensile tests figure 9(a) were performed to investigate the mechanical properties of the three of bicomponent fiber variants (table 3). It can be seen, that the bicomponent fibers PrEl60 with the highest DR have the highest average tensile strength, but lowest strain capacity with 44.9 ± 7.5 MPa and 3.2 ± 1.0%, respectively. Due to the higher winder velocity during the melt spinning, the resulting fibers have inhomogeneous properties, which can also be derived by the highest standard deviation and highest deviation to the theoretical calculated stress using the calculated diameter d Bico,c as seen table 4. However, the bicomponent fiber PrEl50 show the lowest tensile strength, but highest strain capacity with 29.8 ± 3.7 MPa and 4.2 ± 0.7%, respectively. During the melt spinning, variant PrEl50 was the most stable configuration allowing a low error production, which can also be derived by the overall lowest standard deviation. The tensile strength (32.9 ± 3.9 MPa) and strain capacity (3.7 ± 0.8%) of the bicomponent fibers PrEl70 are located between variant PrEl50 and PrEl60. Both theoretically calculated stress curves of PrEl50 and PrEl70 show high agreement with the practically determined curves. A maximum linear elongation of 1.5% was derived from the stress-strain curves and was used as maximum boundary value for the cyclic tensile tests to determine the piezoelectric properties of the sensor yarn within a linear elastic deformation regime.  shows the results of the electrical resistivity measurement. By analyzing the trend, it can be determined, that an increasing volumetric Pre-Elec PP content in the core of the bicomponent fiber directly leads to an exponential decrease of the electrical  resistivity. However, higher volume fractions complicate both the melt spinning process due to increasing melt viscosities and further textile processing due to the increasing brittleness of the fibers. It can be assumed that a PP/Pre-Elec PP volume ratio of 50%-60% in the core represents an optimum between electrical conductivity and technical processability. Due to the high melt viscosity of the core material, the dosage rate was increased to ensure a more stable and continuous extrusion of the core, resulting in the higher volumetric core and therefore lower sheath fraction. However, it can be observed that there is a concentricity of both core and sheath components as well as an enclosed PVDF layer except of minor imperfections in variant PrEl70. On the one hand, the high volumetric Pre-Elec PP content of variant PrEl70 caused an increase of the melt viscosity and therefore an increase of the process pressure as well. This resulted into a slight eccentric position of the core, which led to locally non-closed PVDF layers. On the other hand, the melt spun fibers showed a more brittle behavior than variant PrEl50 as well as PrEl60, observed by an increased occurrence of fiber ruptures during production. Consequently, the winder velocity for variant PrEl70 was set lower compared to the other configurations (table 3). The optical measurements and determination of the fiber fineness also showed the expected lowest geometric dimensions of fiber variant PrEl60, due to the highest realized draw ratio. By using the fineness, the theoretical total diameter of the bicomponent fibers was calculated with equation (3).
Comparing the optical measured and calculated total diameter of the bicomponent fibers it becomes clear, that only configuration PrEl60 deviates from the theoretical value (18% relative deviation). It is assumed that this was caused by the high winder velocity and slightly changing production conditions during melt spinning. However, the theoretically and practically determined values of variant PrEl50 and PrEl70 agree very well.

Poling parameters for the contact polarization of the PVDF interlayer
Based on section 2.6 an average electrical field strength of approximately E = 100 MV m −1 was chosen and iteratively varied by increasing the poling voltage without the occurrence of electrical breakdowns. To estimate the required poling voltage range, the structure of the tricomponent yarn was abstracted as a plate capacitor and equation (5) was used for simplification.
where U Pol is the polarization voltage, t PVDF is the thickness of the PVDF interlayer and E is the electrical field strength.
The PVDF thickness was determined in the microscopy analysis of the bicomponent fiber cross-section as described in 3.1.
The applied poling voltage during the contact polarization was set to a range of U Pol.min ⩽ U Pol ⩽ U Pol,max starting with U Pol.min = 1.0 kV and iteratively increasing it to a maximum of U Pol,max = 3.0 kV. The PVDF interlayer was then polarized for t Pol = 5 min at a temperature of T Pol = 80 • C to orientate the randomly distributed dipole moments of the PVDF β-phase along the electric field lines. Afterwards, the tricomponent yarns were cooled down to room temperature with a remaining electro-static field to ensure a permanent polarization.

Piezoelectric behavior of the tricomponent sensor yarns
Considering the results of the preliminary tensile test, cyclic tensile tests were performed using strain states of 0.5%, 1% and 1.5% and corresponding strain rates of 0.033 s −1 for variant PrEl50 and PrEl60 (testing length 500 mm) or 0.083 s −1 for variant PrEl70 (testing length 200 mm), respectively.
Exemplary results of the piezoelectric behavior of the tricomponent sensor yarns for an applied strain of 1% are shown in figures 11(a)-(c) for each variant. The diagrams illustrate a clear dependence of applied cyclic strain and piezoelectric output voltage for all variants. In figure 11(d) an enlarged view of one strain cycle of a tricomponent yarn variant PrEl70 is visualized. In phase I, an increasing strain from 0% to 1% was applied to the sensor yarn, causing the generation of a positive output voltage. As soon as the strain change remained constant, seen at the reversal points (phase II), the voltage dropped to approximately zero again. In the relaxation phase III, the load was removed from the sensor yarns, resulting into negative generated signals.  This well-known dependency illustrates the measurement capability of piezoelectric sensors exclusively for dynamic applications, where a piezoelectric output voltage is only measured when a change in strain occurs. A clearer illustration can be seen in figures 11(e)-(g), where the first derivative of strain over the time (strain rate) is used to illustrate the change in applied strain. In the enlarged view seen in figure 11(h), the dependence on the strain rate and the generated output voltage is more evident than in figure 11(d). Both curves show a very high degree of similarity in their course, characterized by the increase of the signals in phase I, the constant part in phase II and the decrease of the signals in phase III.
However, it was observed that the output voltage linearly increased with increasing number of cycles. This observation is confirmed by conducting a peak analysis of output voltage as seen in figure 12(a). It is assumed, that due to the cyclic stretching an improvement of the piezoelectric response was achieved by increasing the β-phase within the PVDF. Due to the external winder that could not be precisely adjusted, a lower DR than expected might be present. Therefore, there was still a potential for an increase in the β-phase caused by the stretching procedure. Furthermore, it was observed, that the voltage maxima were within a similar range from ± 100 mV to ± 200 mV indicating similar piezoelectric properties across all three variants. The maximum output voltages generated by each sensor yarn in the different strain states are depicted in figure 12(b) and summarized in table 5. In general, it can be seen that an increasing applied strain on the sensor yarns is causing higher output voltages. Here, variant PrEl50 shows the lowest average increase of the output voltage with increasing strain. Variant PrEl60 and PrEl70 have similar quantitative piezoelectric behavior. However, variant PrEl60 showed the lowest voltages with the highest standard deviation caused by the lowest as well as fluctuating PVDF thickness. The highest average output voltages were generated by variant PrEl50 and PrEl70 at a strain of 1.5% with a positive signal amplitude of 214 mV and a negative signal amplitude of −231 mV during tension and relaxation, respectively.
By comparing the output voltages of the three different tricomponent sensor yarns, it can be concluded, that an increase of the conductive Pre-Elec PP content in the core does not influence the piezoelectric behavior of the PVDF significantly. However, consistent fiber properties and a homogeneous PVDF thickness and thus a stable melt spinning process are more crucial. A volumetric Pre-Elec PP content of 50% is therefore sufficient and a compromise between the electrical conductivity of the core needed for contact polarization and the melt spinning producibility. A lower pre-elec PP content in the core, was not further considered in this work, but could also ensure sufficient electrical conductivity and functionality as long as the percolation threshold is not undercut and could even improve the manufacturing process (lower melt viscosity) and resulting fiber quality. A higher pre-elec PP content in the core was technologically not possible with our melt spinning plant.

Piezoelectric behavior of the tricomponent sensor yarns in CFRP
The results of the cyclic three-point bending tests of the CFRPspecimen and the generated output voltage of the piezoelectric sensors as a function of the flexural deformation (a)-(c) and bending velocity (first derivative of the flexural deformation, (d)-(f)) are shown in figure 13. The flexural deformation was set to ± 1 mm to ensure a complete unloading of the specimen.
In general, the piezoelectric behavior of the sensors in CFRP are very similar to the results on the fiber scale as described in 3.3. A very high correlation of the generated voltage signals and the induced bending load can be observed, considering the bending velocity curves. When the bending velocity is zero, no piezoelectric voltages are generated. A positive bending velocity (loading) causes a positive, a negative bending velocity (unloading) a negative voltage signal. The quantities of the sensor output voltages are in a similar range of ± 100 mV to ± 200 mV as on the fiber scale.
No significant differences were found between the three different sensor variants. However, fluctuating values of the generated peak voltages were observed for all configurations, indicating a systematic error like vibration of the testing machine or minor movements of the micro clamps. This phenomenon can be reduced in future SHM applications by using a more permanent electrical contacting method like soldering. Another reason could be the chosen cyclic testing procedure, that was includes a complete unloading of the test specimen. With the relatively high testing speeds and machine vibrations, it is possible that the loading point of the loading pin varies that causes different output voltages. An unloading to the preload could reduce this phenomenon.
The maximum strain (ε max ) of the piezoelectric sensors can be approximated using the equation (6) where s max is maximum flexural deformation, h is the specimen thickness and L is the distance between the supports. Considering, that the approximated maximum strain of the piezoelectric sensors was less than <0.3%, it can be derived, that the developed sensors have a very high sensitivity, not only on fiber, but also on composite scale.

Conclusion
In this paper, a process chain for the manufacturing and evaluation of fiber-based piezoelectric tetracomponent sensor yarns for SHM applications in CFRP was presented. Further usage of the developed sensors in smart textiles and textile reinforced concrete is conceivable. PVDF was used as piezoelectric interphase layer between two electrical conductive layers acting as electrodes. The electrical conductive core made of PP/Pre-Elec PP was covered by a PVDF layer in a bicomponent melt spinning process. Subsequently, the developed bicomponent fibers were covered by copper wire braided mesh acting as outer electrode. By means of this specific sensor structure, a contact polarization enabled the poling of the PVDF layer. Moreover, it was demonstrated that the chosen parameter set for the contact polarization process (poling voltage U Pol ≈ 1-3 kV, poling temperature T Pol = 80 • C, poling time t Pol = 5 min) was suitable for the polarization of the PVDF. The piezoelectric output voltage was measured between the two electrodes as a function of the applied cyclic loading on the fiber and composite scale. In general, a very good correlation between the applied cyclic strain and strain rate, respectively and piezoelectric output voltage was observed. Peak voltages in the range of ±200 mV ± 30 mV were generated. Furthermore, no significant differences were determined between the three present variants (increasing ratio of the conductive Pre-Elec PP in the core) regarding the piezoelectric behavior. A similar behavior of the developed piezoelectric sensors on fiber and composite scale was observed and the sensory principle for the determination of dynamic mechanical loads was proven. However, the optical analysis, tensile tests and piezoelectric evaluation emphasized that consistent fiber properties and a homogeneous PVDF thickness and thus a stable melt spinning process are crucial.
Further research activities in this area will focus on the quantifiable and more precisely adjustable drawing ratio of the PVDF, during the melt spinning process of the bicomponent fibers. The piezoelectric response of the sensor yarns should be further investigated on the textile, composite and structural level at higher strain rates and also characterized regarding durability and long term reliability especially for the application in smart textiles or textile reinforced concrete.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.