Pellet-based fused deposition modeling for the development of soft compliant robotic grippers with integrated sensing elements

As described previously [42, 43], a conductive thermoplastic composite was mixed, (Torque rheometer HAAKE PolyLab OS, Thermofisher, Durlach, Germany) at 195 °C for 60 min, extruded in the form of a filament with a 1.75 mm diameter, using extrusion temperature of 195 °C (Capillary rheometer RH7, Netzsch-Gerätebau, Selb, Germany) cut into pellets with a length between 3 and 5 mm. The capillary rheometer was equipped with a 1.75 mm die and a 500 mbar pressure sensor. In order to ensure that the filler was homogeneously distributed in the matrix, the pressure was recorded with time (figure S1 (available online at stacks.iop.org/FPE/7/025010/mmedia)). After the extrusion process, the filaments were cooled to room temperature and cut into pellets with a length between 3 and 5 mm. The composite consisted of styrene-ethylene-butylene-styrene triblock copolymer from Kraiburg TPE (Waldkraiburg, Germany) and carbon black from TIMCAL (Bodio, Switzerland). The two components were mixed in 1:1 mass ratio. The large carbon black concentration (50%wt) ensured that the resulting composite had a resistivity at the conduction zone of the percolation curve of the composite [44, 45].

For the printing of the TPE structures, 3 mm styrene-ethylene-butylene-styrene triblock copolymer (TPS) pellets, with three different Shore hardness (Sh25A, Sh50A and Sh65A), from Kraiburg TPE (Waldkraiburg, Germany), were used. For the 3D printing, the pellet printer TU-maker Voladora NX+ (International Technology 3D Printers S.L., Valencia, Spain), equipped with a single-screw extruder (figure 1(a)) was used. The temperature used for the printing (250 °C) was higher than the temperature used in section 2.1. The reason for that was that heating zone was significantly shorter compared to the capillary Rheometer and therefore, the material spent shorter time in the heating zone. To compensate for this fact and ensure an adequate adhesion for the printed elements, a higher temperature was necessary. Printing of the substrate and the sensor parts could be obtained using the following parameters: Nozzle size of 0.6 mm, Nozzle temperature of 250 °C, printing bed temperature of 45°, layer height of 0.2 mm and printing speed of 15 mm s⁻¹. An extrusion multiplier of 8 was used for the substrate of Shore hardness 65 A, 12 for the 50 A and 16 for the 25 A. The extrusion multiplier has been defined a parameter for controlling the extrusion rate, expressed by the volume of the melted plastic material that flows through the nozzle per unit time [46].

Zoom In Zoom Out Reset image size

To achieve multi-material printing with a single pellet printing head, we first printed the substrate material and finally the sensor material on top of it. For the tensile testing the resulted structure will be called sensor strip (figure 1(b)).

To investigate the sensor behaviour of the single piezoresistive element for later comparison, fiber with a diameter of 0.6 mm were extruded using the TU-maker Voladora NX+ 3D printer. Pellets of the piezoresistive sensor were fed through a hopper and the threads were extruded with a temperature of 250 °C. The cross-section of the fiber was measured as the average of three samples.

To investigate the piezoresistive behaviour of the fiber sensor and the sensor strips, a tensile test with a simultaneous recording of the electrical signal was performed. For the tensile tests, a Zwick Roell Z005 tensile testing machine with pneumatic clamps, with a pressure of 4 bars and a 200 N load cell (ZwickRoell, Ulm, Germany) was used. A gauge length of 50 mm, for the fibers and the sensor strips, was used.

For the electrical measurements, a Keithley 2450 multi-meter (Keithley Instruments, Solon, USA) was used. The sampling rate of 10 Hz was selected. Tensile tests were performed up to the point of fracture. In addition, dynamic conditions (cyclical testing) and also quasistatic testing involving a dwell time of 10 s were performed. The cross-head speed during the tensile experiments was 400 mm min⁻¹. This speed was selected based on the speed of the servomotor, when performing the robot experiments.

The relative resistance (R_rel) was calculated using the following formula:

$$\begin{equation}{R_{{\text{rel}}}} = \frac{{R - {R_0}}}{{{R_0}}}\end{equation} \tag{ 1 }$$

where R is the value of the resistance and R₀ the value of the resistance when no strain is applied to the sensor.

For defining the sensitivity of the sensor, the GF was used, as defined by the following equation:

$$\begin{equation}{\text{GF}} = \frac{{\Delta {R_{{\text{rel}}}}}}{{\Delta \varepsilon }}.\end{equation} \tag{ 2 }$$

As for the resistivity, the following formula was used

$$\begin{equation}\rho = \frac{{RA}}{l}\end{equation} \tag{ 3 }$$

where A is the cross-section, measured with the optical microscope and l the distance between the electrodes, performing the measurement.

Based on the dynamic tensile testing, the mechanical and electrical drift between different cycles was calculated (figure 2(a)). The electrical drift was defined as a percentage of the difference in the value of the relative resistance at the same strain during two different cycles of the dynamic testing. Using quasistatic testing, the mechanical and electrical relaxation were analyzed. Based on the mechanical and electrical signal at the beginning and the end of the dwell time the relaxation was calculated. Figure 2(b) shows the investigation of signal relaxation behaviour schematically. For the tensile measurements, the figures represent the testing for one sample.

Zoom In Zoom Out Reset image size

The dimensions of the printed structures were investigated with a light microscope from Carl Zeiss AG, (Jena, Germany). Before the imaging assessment, the samples were immersed in liquid nitrogen and then cut in the middle, so that the cross-section was visible. A polygonal fitting tool was used for calculating the cross-section area. For calculations of the cross-section, the average of three measurements was used. Standard deviation for the height was not calculated, because it was not constant over the width of the cross-section. The standard deviation values were reported for the cross-section area.

The compliant soft robotic gripper structure is based on a stiff base and a soft compliant part. The stiff base was printed with poly-lactic acid (PLA) using a Raise3D Pro 2 3D printer (Raise 3D, Irvine, USA). The soft compliant part with integrated strain sensor element was printed with TU-maker Voladora NX+. In order to be able to detect deformations at the entire length of the compliant gripper, the electrodes were placed at positions A and B, as can be seen in figure 3(b). For the planned application, a maximum deformation of 10% between points A and B was calculated based on the bending angle, however, this value depends on the size of the spool gripped by the gripper. Two Tower Pro MG90S micro servos (Adafruit Industries, New York, USA) were used to move the gripper device. The control of the motors was performed with an Arduino microcontroller.

Zoom In Zoom Out Reset image size

In order to assess the sensor behaviour of the piezoresistive elements, extruded fiber sensors were tested with a tensile test up to the point of fracture (figure 4). The test procedure to verify the sensor performance of fiber piezoresistive materials has been already reported by Georgopoulou et al [47]. The fiber was extruded using the pellet-based 3D printer and a constant diameter of 0.58 ± 0.02 mm diameter was produced.

Zoom In Zoom Out Reset image size

For the stress–strain curve (figure 4(a)), a characteristic necking (yield point) at 13% strain can be observed. This necking is well-known from previous studies about TPS based sensors with high filler concentration [42, 47]. The sensor fiber can endure elongations up to 370% strain. Looking at the response of the electrical resistance (figure 4(b)), it can be seen that the sensor fiber exhibited a positive piezoresistive effect with the increasing strain. At the linear parts of the curve, the GF was 0.5 (strain 0%–24%), 19 (strain 24%–75%) and 38 (strain 200%–370%). The GF was not calculated at the non-linear parts of the curve. At strain below 24%, the resistance increased only very slightly with the increasing strain, but the sensitivity rapidly increased at higher strains (figure 4(c)).

Testing the sensor fiber revealed characteristics of the piezoresistive behaviour, however, a free standing sensor fiber is time consuming to integrate into a structure and according to this, sensor strips, based on TPS substrate with sensor element printed on top of the substrate, were fabricated. To investigate the mechano-electrical response of the strips, a tensile test up to the point of fracture was performed (figure 5).

Zoom In Zoom Out Reset image size

In figure 5, it can be seen that the sensor strips made with higher Shore hardness are stiffer, as expected. The necking at the yield point, observed for the sensor fiber, is not present for the strips. The elongation at the point of fracture decreased with the increasing Shore hardness of the support structure and therefore it can be concluded that the TPE of the substrate, and not the TPE-based sensing element with high filler content, will limit the fracture behaviour for soft robotic applications. The strain at point of fracture for the sensor strip (Sh25A) and the sensor fiber are similar; therefore, strain limitation due to the support structure can be neglected. Based on the analysis of Decroly et al, 400% strain is more than sufficient for the motoring of actuated soft robotic structures [48]. For all 3D printed sensor strips, the electrical signal response during the tensile test to the point of fracture are presented in figure 6.

Zoom In Zoom Out Reset image size

From the response of the electrical signal, it can be seen that in general, the shape of the resistance strain-curve was similar for all the sensor strips with the three different Shore hardness. The relative resistance, first sharply increased (positive piezoresistive effect), then decreased (negative piezoresistive effect) and then increased (positive piezoresistive effect) again until fracture occurred. The strain ranges of the minimum in the relative resistance curve (saddle point), appeared at lower strains with lower Shore hardness. Unexpectedly, the electrical response of the sensor strips is significantly different from the sensor fiber, where only a positive piezoresistive effect could be observed (figure 4). It is worthwhile to mention that for the printing of the senor strips and the compliant soft gripper, first the soft TPS structure was printed and later the piezoresistive sensor material was printed on top of it. We assume that printing of the piezoresistive sensor on a soft substrate resulted in different sensor performance because the soft substrate deformed during printing process by the normal pressure during printing. Normal pressure during printing occurs, because the extruded thread of 0.6 mm will be elliptically deformed to a smaller thickness (figure 7). After the printing, the soft substrate will move back (spring back effect) and a stress exerted on the printed sensor element will occur (figure 7(b)). Based on the total height of the printed sensor element (3 × 0.2 mm), it can be seen the element on the lower Shore hardness (figure 7(c)) is integrated deeper inside the substrate resulting in a larger cross-section area. In the 65 A sample (figure 7(e)); the sensor element was not deep inside the substrate. Therefore, height of the sensor element decreased with the increasing Shore hardness and the width increased (figures 7(c)–(e)). The cross-section area of sensor elements analysed by the optical microscope is shown in the table in figure 7. Even though that the soft substrate deformed during printing of the sensor elements, the cross-section area of all sensors is quite similar, resulting in a similar extrusion volume for all the strips. This was not the case in an older study, for TPU filament-based strips, where the cross-section showed a dependency to the Shore hardness of the substrate [18]. However, in the earlier study, the printing parameters were the same for all strips with different Shore hardness and therefore the differences could be attributed to a die swell effect. In the current study, however, the printing parameters were altered to counter for the die swell effect and thus, it was possible to obtain a similar cross-section for the three different Shore hardness strips. Therefore, we had to optimize the extrusion multiplier for each Shore hardness, as mentioned in the experimental section, the optimal multiplier was 8, 12 and 16 for the Shore hardness 25 A, 50 A and 65 A, respectively.

Zoom In Zoom Out Reset image size

It is worthwhile to mention that printing the sensor elements on softest substrate (Shore hardness 25 A) resulted in a smearing problem. Therefore, the top layer of the substrate is contaminated with carbon over a wide area. Unfortunately, it was not possible to print the TPS structure on top of the piezoresistive sensor. In that case, the sensor would smear over the whole printing area of the TPS structure, potentially causing clogging to the 3D printing nozzle.

For comparison, the GF was evaluated between 200% and 230% strains, because in that range all the strips and the fiber show a monotonic response. The GF was calculated 70 for the Sh25A, 30 for the Sh50A and 35 for the Sh65A. From the values of the GF it can be seen that for both the 50 A and 65 A, the sensitivity is similar compared to the pure sensor fiber (GF = 38). However, the value is almost double for the Sh25A strip. This significant difference shows one more time the effect of the substrate stiffness on the sensor response. However, in comparison to TPU sensor strips, the trend of the GF is reversed [17, 18]. This fact might be explainable by the different polymer materials SEBS versus TPU. Georgopoulou et al has demonstrated, that polymer matrix material can significantly change the sensor behaviour of the piezoresistive material [49].

According to Decroly et al, soft actuators in robotic application often have a deformation up to 100%, therefore, the first dynamic tensile testing was first performed at a range of strains 0%–100% (figure 8).

Zoom In Zoom Out Reset image size

From the dynamic cyclical response, it was seen that the sensor fiber (figure 8(a)) and the two TPS sensor strips 25 A and 65 A show similar relative resistance behaviour (figures 8(b) and (d)). As expected from the previous tensile test experiments, the sensor signal for the Sh50A strip is significantly smaller (figure 8(c)). Interestingly, the sensor strips printed with Sh50A coiled up after removing from the printing bed (figure S2). This was not the case for the other strips that remained straight after removing from the printing bed. This contrast could originate in the differences in the compositions to achieve the different Shore hardness of the TPS substrate material. In the case of the 50 A, the coiling up of the strips after the printing, could lead to the development of internal stresses in the sample, which could impact the sensor response.

Figure S3 shows the mechanical stress–strain behaviour of the sensor filament and the 3D printed strips. As expected, Shore hardness of the TPS affects the mechanical stress of the strips. As higher the Shore hardness of the TPS, the higher the mechanical stress. A stress plateau or negative stress value, at low strain indicates bucking behaviour for the sensor fiber and the strips (Table 1). The buckling effect, observed in figure S3, has been already reported for other 3D printed TPEs and it can be resolved by pre-straining the sensor [50–53]. In previous studies, the dynamic sensor behaviour of the piezoresistive element often showed uncertainty of the sensor signal at low strain because of the viscoelastic behaviour of the matrix material that was confirmed by the presence of buckling [49, 54, 55]. In these studies, the sensor was integrated into the soft matrix with casting methods or injection molding and uncertainty with the form of nonlinear response appeared for the sensor signal at low strains. The nonlinearity was then explained by the insufficient load transfer between the sensor fiber and the matrix. In the present work, despite the presence of buckling, the sensor signal did not show an uncertainty (plateau) at low strain. This shows that integrating sensors in soft elastomers, using FDM is a promising response for producing sensors with linear response, even for low Shore hardness materials (25 A). This observation was confirmed by another study, where a similar TPE was used, but in that case, the strips were strained only between 0% and 50% and a direct comparison with the current study, is not feasible [56].

In addition to the dynamic tensile tests up to 100% strain, dynamic testing up to 10% strain were performed, based on the calculated deformation of gripper for the large spool object with a diameter of 90 mm. Figure 9 shows the dynamic tensile cycle test up to 10% strain.

Zoom In Zoom Out Reset image size

Based on the results, no secondary peak can be observed in the sensor signal (figure 9). As expected from the previous analysis (figure 6), for the Sh25A and Sh65A strips a negative piezoresistive response can be observed. For the Sh50A strip, in figure 6, the slope of the relative resistance-strain curve was positive and therefore the positive piezoresistive response of the figure 9(b) correlates nicely to the previous results. Table 2 shows the summary of the mechanical and electrical sensor signal of the 3D printed strips with different Shore hardness.

It can be seen that the value of the resistivity decreased with the increasing Shore hardness of the substrate material. Similar results have been observed previously for injection molded silicone samples with integrated piezoresistive sensor fibers [54]. Based on this, it can be assumed, that the Shore hardness of the structure where sensor elements are integrated has an important effect on the sensing behaviour of the developed piezoresistive SEBS sensor. In addition to this, the initial resistivity results show, that the 3D printing also affects the electrical conductivity of the sensor elements. The reason for this phenomenon has not been investigated and should be part of further research activities. However, based on the optical microscope pictures of the cross section analysis (figure 7) it can be assumed that the softness of the substrate will affect the printing behaviour of the sensor material.

For the mechanical buckling behaviour of the three strips, it can be observed that the results do not correlate with the Shore hardness. It is worthwhile to mention that the SEBS of Kraiburg is a composite based on SEBS, PP (polypropylene), fillers and stabilizers, and therefore the Shore hardness and the mechanical behaviour can be adjusted in a wide range, by changing the material composition. The drift of the sensor signal, for the Sh50A strip is significantly higher and we assume that this is also related to the different material composition of the SEBS based composite. However, the strip based on the commercial TPE with the lowest Shore hardness (Sh25A) showed the smallest drift behaviour of the sensor signal at 10% strain.

Quasistatic tensile experiments were performed between 0%–5% and 0%–10% strain, based on the calculated deformation of gripper for the small and the large spool object, with a diameter of 60 and 90 mm, respectively. Figure 10 shows the results for the strain up to 10%, a dwell time of 10 s was used at maximum and minimum strain. The results of the quasistatic test up to 5% strain are in the supplementary part (figure S5).

Zoom In Zoom Out Reset image size

The stress at 10% strain increase using TPS with higher Shore hardness for the TPE substrate. This can be expected because of the relationship between Shore hardness and Young's modulus of the thermoplastic support material (figures 10(a)–(c)). For the electrical signal (figures 10(d)–(f)), it can be seen that the strips Sh25A and Sh65A exhibited reverse piezoresistivity and the Sh50A strips show positive piezoresistive response, similar to previous analysis (figure 9). The values of the relaxation of the stress and the resistance can be found in table 2. Looking at the mechanical relaxation, no significant differences can be seen between the strips of different Shore hardness. However, for the electrical relaxation at 5% strain, the 25 A and 65 A strips show the lowest relaxation. At 10% strain, a significantly higher electrical relaxation of the 25 A strip can be observed. Looking at figure 6, it can be observed that at around 10% strain the 25 A strip change from negative to positive piezoresistive behaviour. Based on these results, we assume, that it is essential to look at the full stain analysis up to the point of fracture to identify in which strain area the soft sensor material can be efficiently used as a sensor material.