3.1 Mechano-electrical behavior of piezoresistive fibers and 3D printed elements
The two different TPS composite sensor fibers were characterized with tensile testing up to the point of fracture. From the response of the stress (Fig. 4a), it was seen that both fibers showed the characteristic necking that has been already reported for styrene-based tri-block co-polymer (TPS) composites with higher filler content26,28,32. As for the yield point, it appeared at strain 10.3% for the fiber based on TPS 40A and 17.3% for the sensor fiber based on TPS 50A. Both fibers exhibited a strain hardening effect that was more prominent for the sensor fiber based on 50A. It had both a higher elongation at the point of fracture (400%) and larger ultimate strength (14.8 MPa). In comparison, the sensor fiber based on TPS 40A only reached an elongation of 270% (point of fracture) and 3.9 MPa (ultimate strength). These findings are in agreement with the results of a previous study that showed that decreasing the Shore hardness, below 50A, leads to a decrease in the point of fracture for TPS/CB sensor fibers24.
Up to 20% strain, the electrical response is similar for both sensor fibers (Fig. 4b). For both fibers, the sensitivity increased significantly for strains above the yield point. A gauge factor (GF) of 117 and 70 was calculated between 20–100% strain for the sensor fibers based on TPS 40A and 50A, respectively. Similar behavior has been reported earlier for styrene-based tri-block co-polymer (TPS) carbon black composites with higher filler content24.
Similar tensile tests were performed for the printed sensor elements. Because of the later lamination process at 105°C, the elements were tested before and after heating at 105°C. Comparing the elements based on TPS 40A and 50A, it can be seen that the sensor element based on TPS 50A could endure larger elongations (up to 250%) and had a higher ultimate strength of 5.12 MPa (Fig. 5a). The sensor element based on TPS 40A could endure elongations of up to 151% and the tensile strength was 3.1 MPa. Comparing the sensor elements with the sensor fibers, it can be seen that the ultimate strength was lower for the sensor elements. The elongation at the point of fracture also decreased for the sensor elements but the strain hardening effect is also present for the sensing element based on TPS 50A.
After the thermal treatment, the mechanical properties of the sensor element based on TPS 50A did not change significantly. The elongation at the point of fracture was 238% and the ultimate strength was 4.63 MPa. This was not the case for the sensor element based on TPS 40A. The elements broke at 14% strain, after the thermal treatment. Thus, the sensor composite based on TPS 40A is incompatible with self-healing matrix materials that require thermal treatment for healing.
For the relative resistance response, both sensor elements showed a positive piezoresistive response (Fig. 5b). It can be observed that lower Shore hardness resulted in a higher sensitivity (GF = 82), which is in good agreement with Fig. 4b. The GF is lower for the sensor elements, compared to the sensor fibers. The response of the sensor signal (relative resistance) of the sensing element based on TPS 50A looks almost identical before and after heating.
3.2 Mechano-electrical behavior of the fiber composite strips (SFCs)
After the analysis of the sensor fibers, the two fibers were embedded into two different self-healing supramolecular elastomers (UPy1 and UPy2). The mechanical and electrical characterization was repeated for the self-healing sensor fiber composite strips (SFC). From the response to the stress (Fig. 6a), it was seen that the SFC UPy2 50A had a higher ultimate strength (1.1 MPa) than the other SFCs (ca 0.3 MPa). The elongation at the point of fracture was similar for all the SFCs (ca 300%). The two SFCs, with the UPy1 matrix, exhibited a negative slope for strains higher than 50%. The strain hardening behavior that was seen in Fig. 4a does not appear for all the SFCs. A local decrease in the cross section was observed for the SFC UPy1 SFCs. Generally, elastomers have a viscoelastic behavior and the negative slope and decrease in cross-section are good indications that at higher strains, the viscous behavior dominates the mechanical behavior of both SFCs with UPy1 matrix. Additionally, the elongation at the point of fracture increased slightly (20%) for both the SFCs 40A compared to the strain sensor fiber based on TPS 40A. The opposite trend was observed for the fiber based on TPS 50A, as both the SFCs 50A had an elongation at a break 100% smaller than the fiber based on TPS 50A.
For the relative resistance (Fig. 6b), in this case, the SFCs with the fiber based on TPS 40A, showed significantly higher sensitivity. This observation correlated with the analysis for the extruded sensor fibers shown in Fig. 4. The response was only linear for strains 20–140% and for strains above 250%. Looking at low strains, it can be seen that the SFCs with the fiber based on TPS 40A showed a plateau (no change in the slope) of the curve. This was not the case for the fiber based on TPS 50A which showed a positive slope at low strains. Comparing the GF in the range 0-100%, it can be seen that the value is identical (94) for the two 40A SFCs and 7 and 13 for the SFCs UPy1 50A and UPy2 50A, respectively. From these values, it is not evident if the matrix materials affect the electrical response, but it seems that in contrast to the response of the stress, for the electrical properties the sensor fiber plays the dominant role. The value is smaller than the GF of the sensor fibers, before the integration in the self-healing matrix. Looking at low strains (below 20%), it was not possible to calculate the GF, because the SFCs didn't exhibit a linear response, but it is evident that the SFCs 50A responded with higher sensitivity (ΔRrel), compared to the SFCs 40A with both self-healing matrix materials. The SFCs 40A shows a sensitive response for strains above 20%.
In order to investigate the reproducibility of the sensor response, dynamic cyclic tensile tests between 0–50% strains were performed for the four different SFCs. The dynamic cycling tests were repeated after the introduction of manual cuts from both edges (e.g. damage stage) and after healing. Figure 6 shows the results of the dynamic cycling tests. The optical microscope analyses for each stage (pristine, damaged, healed) are presented in (Figs. 7a, 7b, 7c).
For the stress curve in the damage stage (Fig. 7d), negative stress values can be observed after unloading (e.g. 0% strain), especially after the damage occurred. However, after healing, the values resembled the stress curve of the pristine one. The electrical response of the sensor signal was reproducible between the 2nd and 10th cycle for all three cases (Fig. 7e). The initial resistance (R0) increased significantly when the damage occurred by manual cutting (Fig. 7f). After healing a similar sensitivity (ΔRrel) can be observed however, the sensor signal was shifted to higher values due to the higher initial resistance and a re-calibration is needed. It is worthwhile to mention that the drift of the sensor signal increased significantly after self-healing process.
In Fig. 8a-f, the results of the SFC based on UPy1 and the sensor fiber based on TPS 50A are summarized. The mechanical behavior of this composite is similar to the one before (Fig. 8d). It can be concluded that the different Shore hardness of the two sensor fibers does not significantly affect the stiffness of the composite strips and the mechanical behavior (recovery of stress after the healing) is dominated by the self-healing UPy1 matrix material. As expected by the previous results, the sensitivity (ΔR) decreased with the higher Shore hardness value of the sensor fiber matrix. Unlike the SFC UPy1 40A, the SFC UPy1 50A doesn't need a recalibration after self-healing. Similar to the previous UPy1 matrix composite the relative resistance and the sensitivity decrease with the damage. After self-healing, the sensor revealed the same electrical signal behavior as the pristine one. However, the drift significantly increased after the heating and this is a disadvantage for this SFC (Fig. 8e and f).
For both SFCs based on UPy1 matrix, a change in electric signal behavior was observed. However, without additional AI (artificial intelligence) methods to cope with the drift behavior, it will be not possible to detect the damage and self-healing of the composite strip easily.
The dynamic tensile testing results for the UPy2 matrix composites with the two different sensor fibers are presented in Fig. 9a-f and Fig. 10a-f. Unfortunately, the SFCs with the TPS sensor fiber based on Shore hardness 40A were too brittle after the self-healing step to obtain any data from the dynamic testing (Fig. 9). The R0 increased significantly but sensitivity (ΔRrel) did not show a significant change after damage, which is in good agreement with previous results of the UPy1 SFCs. It is worthwhile to mention that the mechanical behavior of pristine and damaged samples are very similar. However, it can be concluded that the combination of UPy2 matrix and sensor fiber based on TPS 40A did not result in a composite material with intrigued damage detection.
From the response of the mechanical stress (Fig. 9d), it was seen that the profile of the stress was the same before and after the damage occurred. This is different from what was observed for the fiber SFCs with the UPy1 matrix, confirming again that the profile of the stress is dominated by the type of matrix material. As for the electrical response (Fig. 9e), the same response for the R0 and ΔRrel was seen, as for the other SFCs. The R0 increased and ΔRrel decreased after the damage occurred. No significant difference in the drift was observed after the healing.
Only the stress-strain behavior changed after damage and self-healing (Fig. 10). The negative stress in the pristine samples can be explained by the viscoelastic behavior of the UPy2, already reported in previous studies12,17. The sensor signal did not show significant changes between pristine, damaged and healed stages (Fig. 10e). Therefore, only the value of the R0 can be used for detecting the presence of damage. Similar to the SFC UPy1 50A, a re-calibration is not necessary after the healing. The drift remained low after the healing and the quality between the different samples is very consistent (low standard deviation).
Comparing the cycling experiments, for all SFCs, the R0 can be used to monitor damage and healing steps directly. Using sensitivity (ΔRrel) might be possible, but recalibration might be necessary, depending on the self-healing matrix. Overall, it is evident that the SH composites with the sensor fiber based on TPS 40A are not suitable, because after self-healing step they got too brittle to be further used in cycling tests. This fact was expected based on the brittleness of the elements based on TPS 40A, seen in Fig. 5 after heating at the 105°C. Based on this observation, the sensor fibers and elements based on TPS 40A will not be used for further analysis.
3.3 Mechano-electrical behavior of the 3D printed sensing element composites (SECs)
In addition to the SFCs previously discussed, 3D printed sensor elements were integrated into the supramolecular matrix materials, namely UPy1 and UPy2. Even though the sensor elements based on TPS 40A resulted in higher sensitivity, sensor elements with TPS 50A were only integrated into the self-healing matrix materials, because of the brittle behavior after self-healing step. The results of the tensile testing tests are shown in Fig. 11.
From the response of the mechanical stress (Fig. 11a), it was seen that the slope of the stress-strain curve became negative for strains higher than 50% strain for the SEC UPy1 50A. The SEC UPy2 50A showed a strain hardening behavior that was also seen in the case of the sensor element based on TPS 50A. The same trend was seen in the case of the SFCs (Fig. 6). The elongation at the point of fracture at 370% strain was higher than the sensor element based on TPS 50A (240%), showing that the integration of the sensor element in the self-healing matrix has a positive effect on this property. The value is in the same range as the SFCs, showing for one more time that the self-healing matrix is the dominant influence in the mechanical behavior of the composite strip
As for the electrical behavior (Fig. 11b), the SECs showed a positive piezoresistive response for the entire range of the sensor. The UPy2 50A strip showed better sensitivity at low strains (below 20%). The values of the GF are slightly lower than the value reported for the sensor element (14), but this behavior was also seen in the case of the SFCs. The values of the GF are in the same range, as the SFCs with the same self-healing elastomer matrix. Overall, despite the differences seen between sensor fibers and elements, the composite strips have similar mechanical and electrical behavior. The SFCs showed a slightly increased sensitivity and the SECs a higher elongation at break, but in both cases, the differences are small.
Dynamic tensile testing results of the SECs are summarized in (Fig. 12a-f and Fig. 13a-f). The results are similar to those presented for the SFCs (Fig. 8 and Fig. 10). A calibration after the healing is needed for both the UPy1 50A and UPy2 50A SECs, since the value of the ΔRrel does not recover after the healing. The UPy1 50A SEC showed a significant increase in the drift after the healing, a behavior also seen in the UPy1 50A SFC. The sensitivity (ΔRrel) is higher for UPy2 50A SEC, but this behavior matches with the results seen in Fig. 11b for low strains. The values are in a similar range as the two SFCs.
It can be concluded that composite piezoresistive sensors by extrusion (fibers) or 3D printing (elements) can be used for monitoring self-healing materials. Since the value of the ΔRrel does not recover after the healing (in all strips, but the UPy1 50A SFC), it is evident that recalibration after the healing step has to be performed for real applications, like closed-loop control systems for soft robots33,34.
3.4 SECs as composite strips for monitoring a soft robotic actuator module
In soft robotics, tendon or pneumatic bending actuators are often used 11,24,35,36. Due to the limited amount of self-healing material, the performance of the SECs was investigated under bending conditions on a 3D printed TPU hinge (Fig. 14a-b). An internal setup was developed for the dynamic testing on tendon-driven hinges. The SECs were glued on the top of the hinge to investigate sensor signals in pristine, damaged and healed states (Fig. 14c-d).
As expected from the cycling tensile experiments, the initial resistance R0 can be used to monitor the different states (e.g. damage and healing) for the self-healing composite with both matrix materials. Even though the values were different for the two composites, the trend in the values was the same. For both the composites, the value of the R0 increased when the damage occurred and returned to the initial values of the pristine after the healing (Figs. 14e and 14f). Remarkably for bending experiments, the drift of the electrical signal for both self-healing composites was significantly below 1% (e.g. 0.1%), including the first cycle. This is an improvement from the tensile experiments, where the values of the drift were, as high as 18%.