3D-Printed Load Cell Using Nanocarbon Composite Strain Sensor

sensors

Communication 3D-Printed Load Cell Using Nanocarbon Composite Strain Sensor

Kwan-Young Joung 1,2, Sung-Yong Kim 3 , Inpil Kang 3 and Sung-Ho Cho 1,*

1 Department of Electronic Engineering, Hanyang University, Seoul 04763, Korea; [email protected] 2 Department of Innovative Smart Manufacturing R&D, Korea Institute of Industrial Technology, Cheonan 31056, Korea 3 Department of Mechanical and Design Engineering, Pukyong National University, Busan 48513, Korea; [email protected] (S.-Y.K.); [email protected] (I.K.) * Correspondence: [email protected]; Tel.: +82-2-2220-0390

Abstract: The development of a 3D-Printed Load Cell (PLC) was studied using a nanocarbon composite strain sensor (NCSS) and a 3D printing process. The miniature load cell was fabricated using a low-cost LCD-based 3D printer with UV resin. The NCSS composed of 0.5 wt% MWCNT/epoxy was used to create the ﬂexure of PLC. PLC performance was evaluated under a rated load range; its output was equal to the common value of 2 mV/V. The performance was also evaluated after a calibration in terms of non-linearity, repeatability, and hysteresis, with ﬁnal results of 2.12%, 1.60%, and 4.42%, respectively. Creep and creep recovery were found to be 1.68 (%FS) and 4.16 (%FS). The relative inferiorities of PLC seem to originate from the inherent hyper-elastic characteristics of polymer sensors. The 3D PLC developed may be a promising solution for the OEM/design-in load cell market and may also result in the development of a novel 3D-printed sensor.

Keywords: load cell; carbon nanotube; strain sensor; 3D printing; piezoresistivity

Citation: Joung, K.-Y.; Kim, S.-Y.; Kang, I.; Cho, S.-H. 3D-Printed Load Cell Using Nanocarbon Composite 1. Introduction Strain Sensor. Sensors 2021, 21, 3675. Load cells convert forces such as pressure or load into electrical signals. They are https://doi.org/10.3390/s21113675 some of the most widely used sensors in our daily lives, with industrial, automotive, and measurement applications. Load cell types include hydraulic, pneumatic, piezoelectric, Academic Editor: Iren E. Kuznetsova capacitance, and strain gauge, with strain gauge being the most common [1]. Within the internal structure of a strain gauge load cell (called a flexure), a structure such as a beam Received: 2 April 2021 or a membrane is deformed under the loading conditions. These load cells commonly Accepted: 20 May 2021 measure the deformation of the flexure with conventional foil or semiconductor strain Published: 25 May 2021 gauges. A typical foil strain gauge uses a piezoresistive mechanism based on electrical resistance fluctuations due to changes in its length and cross-section under the load. The Publisher’s Note: MDPI stays neutral load cell is generally designed to meet the desired specifications of a specific structure, with regard to jurisdictional claims in so that its performance is closely related to the characteristics of the strain gauge. A foil published maps and institutional affil- strain gauge is manufactured by processing the alloy material into a fine pattern using iations. an etching method; its sensitivity, which is represented by the gauge factor, determines the alloy properties [2]. A foil strain gauge has low sensitivity and weak durability. A semiconductor strain gauge has high sensitivity with a larger output, which can be used to develop high-performance load cells [3]. A semiconductor strain gauge is manufactured Copyright: © 2021 by the authors. using a silicon crystal cutting process; however, its brittleness makes it less producible and Licensee MDPI, Basel, Switzerland. less durable, which limits its suitability for many industrial applications [4]. This article is an open access article Newly emerging materials may be used to develop sensors that overcome some of distributed under the terms and these problems [5,6]. Nanocarbon allotrope, such as carbon nanotubes (CNT) and graphene, conditions of the Creative Commons may enable the development of new load cells with higher sensitivity, better mechanical Attribution (CC BY) license (https:// strength, and simpler machining [7–9]. Nanocarbon allotrope consisting of continuous creativecommons.org/licenses/by/ carbon atoms have high mechanical strength and excellent electrical conductivity, making 4.0/).

Sensors 2021, 21, 3675. https://doi.org/10.3390/s21113675 https://www.mdpi.com/journal/sensors Sensors 2021, 21, 3675 2 of 12

them suitable for use in new smart nanocomposite materials with ultra-light, high-strength conductivity [10–14]. A nanocarbon-based composite may be an ideal sensory material because of its electrical and mechanical properties [15,16]. When used as fillers in the composite, nanocarbon allotropes improve the mechanical strength and create conductive paths in the composite [17]. When a nanocarbon-based composite is deformed, the composite’s electrical properties change due to internal conductive conditions in the process of piezoresistivity. Therefore, we developed nanocarbon composite strain sensors (NCSS) using a nanocarbon piezoresistive composite (NCPC), which showed similar performance to commercial foil strain gauges [18,19]. Under compression loading, NCPC internal nanofillers move closer to each other and thus increase their electrical conductivity, thereby reducing the electrical resistance of the NCPC. Conversely, NCPC electrical resistance is increased under tensile load due to its broken internal conductive networks. NCPC can be used to develop new types of strain sensors and load cells because of its promising material characteristics, which are different from conventional metal or ceramic materials. Specifically, we focus on the beneficial machining properties of NCPC, which allow various load cell structures to be easily manufactured with a three-dimensional (3D) printing process. A 3D printing process using NCPC paint or filaments may be used to fabricate novel load cells. With 3D printing technology, a device can be easily designed and manufactured using polymeric materials [20–23]. 3D printing combined with NCPC has been used in the development of sensors. Kim et al. fabricated 3D-printed multiaxial force sensors using CNT/TPU filaments. In this study, the electrical characteristics of the 3D-printed sensor were demonstrated through various experiments, such as the three-point bending and a repeated bending cycle test [24]. Christ et al. fabricated TPU/MWCNT strain sensors using 3D printing and evaluated their electrical properties [25]. Al-Rubaiai et al. fabricated stretchable strain sensors using 3D printing for wind sensing [26]. Stano et al. fabricated load cell structures and sensor parts with an FFF method using TPU and CNT filaments [27]. 3D-printed strain-gauge microforce sensors were recently developed. Qu et al. demonstrated the effectiveness of 3D-printed force sensors in a proof-of-concept demonstration. In their study, 3D-printed microforce sensors with micronewton sensing resolutions were fabricated by a fused deposition method (FDM) or stereo lithography (SLB) method. Only the flexure was fabricated with a 3D printing process; commercial strain gauges were used to develop the force sensors [28]. A load cell using the NCSS for its sensory part may be a promising solution in the development of a novel 3D-printed sensor. Despite the fact that many studies have investigated 3D printing technology for sensors, problems arising from the use of polymer materials still exist. Polymer-based 3D printing sensors have some unstable electrical characteristics such as hysteresis, recovery, and response time due to the viscoelastic properties of polymer. To overcome these problems, it is necessary to study the characteristics of 3D printing polymer sensor systematically. Since the performance of the sensor directly relates to its material property and structural forming, the aim of our contribution is to study the piezoresistive behavior of NCPC when it forms a sensing structure such as a flexure of load cell fabricated by a polymer-based 3D printing process. A miniaturized load cell with a simple beam flexure was developed to be easily fabricated using a 3D printing process. A finite element method (FEM) was conducted to design the load cell flexure, and piezoresistive paint made of MWCNT/epoxy composite was used to fabricate the sensor part of the flexure. The load cell was fabricated using a low-cost LCD-type 3D printer. The sensing performance of the printed load cell (PLC) was studied in terms of loading voltage output characteristics.

2. 3D Printed Load Cell Design and Fabrication 2.1. 3D Printed Load Cell Design A load cell generally requires various ﬂexure structure types to optimize or maximize its sensor performance. Although the fabrication of some ﬂexures was traditionally limited Sensors 2021, 21, x FOR PEER REVIEW 3 of 13

2. 3D Printed Load Cell Design and Fabrication 2.1. 3D Printed Load Cell Design Sensors 2021, 21, 3675 3 of 12 A load cell generally requires various flexure structure types to optimize or maximize its sensor performance. Although the fabrication of some flexures was traditionally limited to conventional MEMS processes if they had complicated structures, 3D printing pro- cessesto conventional using NCPC MEMS are free processes from such if theyproblems. had complicated If a 3D printer structures, filament 3D is printingdeveloped processes with NCPC,using the NCPC complex are free flexure from can such be problems. manufactur If aed 3D by printer 3D-printing filament additive is developed processes. with NCPC,Fur- thermore,the complex it may flexure be able can to be fabricate manufactured an integral by 3D-printing load cell with additive a flexure processes. that includes Furthermore, the sensingit may parts. be able In toa previous fabricate study, an integral we reported load cell on with a miniaturized a flexure that pressure includes sensor the sensing built byparts. a 3D FDM In a previousprinter with study, an ABS we reported filament, on and a miniaturizedits sensing part pressure was made sensor of NCPC built bypaint a 3D [29].FDM The printer cantilever with sensing an ABS part filament, of the andpressure its sensing sensor partshowed was a made performance of NCPC improve- paint [29]. mentThe of cantilever approximately sensing 200% part ofwith the linear pressure output sensor voltage showed compared a performance to a bulk-sensing improvement part of approximately 200% with linear output voltage compared to a bulk-sensing part without without specific structural features. Therefore, we planned to develop a 3D-printed load specific structural features. Therefore, we planned to develop a 3D-printed load cell based cell based on the previous works and to hire epoxy resin to build a more tough body than on the previous works and to hire epoxy resin to build a more tough body than the previous the previous one. We benchmarked the specifications of a commercial load cell to develop one. We benchmarked the specifications of a commercial load cell to develop a miniaturized a miniaturized 3D-printed load cell. The scheme of the miniaturized load cell involved a 3D-printed load cell. The scheme of the miniaturized load cell involved a flexure that had flexure that had a beam with two fixed ends, which could be manufactured using a 3D a beam with two fixed ends, which could be manufactured using a 3D printer to allow for printer to allow for simple fabrication. simple fabrication. The flexure dimensions were based on general resin strength (E = 3.8GPa), maximum The flexure dimensions were based on general resin strength (E = 3.8GPa), maximum load, and PLC size. The PLC diameter was 30 mm. A maximum load (P) of 2 kg was ten- load, and PLC size. The PLC diameter was 30 mm. A maximum load (P) of 2 kg was tatively selected based on the fragility of the resin. Considering the interior space of the tentatively selected based on the fragility of the resin. Considering the interior space PLC,of thethe PLC,final beam the final dimensions beam dimensions were 17.3 mm were × 2.6 17.3 mm mm × 1× mm2.6 (length mm × ×1 width mm (length× thick-× ness).width × thickness). TheThe flexure flexure structural structural behavior behavior was was simula simulatedted by by an an FEM FEM analysis analysis using using ANSYS ANSYS to to determinedetermine the the maximum maximum beam beam deflection deflection and and load, load, with with the the results results depicted depicted in inFigureFigure 1 1 andand Table Table 1. 1After. After applying applying 2 kg 2 to kg the to theflexure, flexure, ANSYS ANSYS showed showed a maximum a maximum stress stressof of 5555 MPa MPa around around the thespoke, spoke, which which just exc justeeded exceeded the resin the resinepoxy epoxy tensile tensile yield strength yield strength (54.6(54.6 MPa). MPa). 3D 3D printed printed parts parts are are not not 100% 100% solid solid and and their their internal internal structure structure significantly significantly affectsaffects their their mechanical mechanical properties. properties. Therefore, Therefore, to to design design the the flexure, flexure, we we hired hired a a safety safety fac- factor tor(1.5) (1.5) considering considering the the structural structural uncertainties. uncertainties. The The rated rated capacity capacity of of the the PLC PLC was was less less than than1.3 1.3 kg consideringkg considering the safetythe safety factor factor and theand experimental the experimental load criterionload criterion was determined was deter- to minedbe less to thanbe less 1 kg than by a1 sensorkg by a characterization sensor characterization test. We concludedtest. We concluded that the beam that the design beam was designacceptable was acceptable based on thebased strength on the of strength the 3D-printed of the 3D-printed resin and the resin maximum and the load. maximum We plan load.to conduct We plan a to future conduct study a future on the study optimized on the flexure optimized shape flexure using computationalshape using computa- structural tionalanalysis structural to improve analysis the to performance improve the of performance the load cell. of the load cell.

(a) (b) (c)

Figure 1. Finite element method (FEM) analysis using ANSYS: (a) deformation analysis; (b) strain Figure 1. Finite element method (FEM) analysis using ANSYS: (a) deformation analysis; (b) analysis; and (c) stress analysis.

TableTable 1. 1.FiniteFinite element element method method (FEM) (FEM) analysis analysis result. result.

Max.Max. Load Load (kg) (kg) Max.Max. Deflection Deﬂection (mm) (mm) Max. Strain Strain (%) (%) Max. Max. Stress Stress (MPa) (MPa) 2 0.29 2 55 2 0.29 2 55

2.2. Load Cell Fabrication Using 3D Printing Process To fabricate the sensing part on the ﬂexure of the load cell, we created a piezoresistive

paint using the NCPC process depicted in Figure2. The load cell body was built using a popular and low-cost 3D printer (Creality 3D LD-002R) with 4thWave Resin, as shown in Sensors 2021, 21, x FOR PEER REVIEW 4 of 13

Sensors 2021, 21, x FOR PEER REVIEW 4 of 13

2.2. Load Cell Fabrication Using 3D Printing Process To fabricate2.2. theLoad sensing Cell Fabrication part on the Using flexure 3D ofPrinting the load Process cell, we created a piezoresistive Sensors 2021, 21, 3675 4 of 12 paint using the NCPCTo fabricate process thedepicted sensing in part Figure on the2. The flexure load of cell the body load cell,was webuilt created using a a piezoresistive popular and low-costpaint using 3D printerthe NCPC (Creality process 3D depicted LD-002R) in with Figure 4thWave 2. The loadResin, cell as bodyshown was in built using a Figure 3; 4th Wavepopular Resin and is low-cost a UV tough 3D printerresin and (Creality its mechanical 3D LD-002R) properties with 4thWaveare better Resin,than as shown in ABS; thus, itFigure wasFigure mainly3; 4th3; 4th Wave used Wave Resinin Resinthis is experiment. ais UV a UV tough tough It resin has resin mechanical and and its its mechanical mechanical properties properties properties of 85 MPa, are are better better than than tensile strengthABS;ABS; 41.3 thus, thus, MPa, it it was flexuralwas mainly mainly strength used used in 47in this MPa,this experiment. experiment. elastic modulus It It has hasmechanical 570.8, mechanical elongation properties properties 9.2, of of 85 85 MPa, MPa, and shrinkagetensiletensile 6.9% strength [30].strength The 41.3 3D41.3 MPa,printer MPa, flexural flexural used an strength strength LCD technique 47 47 MPa, MPa, elasticwith elastic a modulus0.4 modulus mm nozzle 570.8, 570.8, for elongation elongation 9.2, 9.2, UV resin, whichandand shrinkageprovided shrinkage an 6.9% 6.9% in-plane [ 30[30].]. TheprintingThe 3D3D printer printerresolution used of an0.075 LCD mm technique technique and a vertical with with aprint- a 0.4 0.4 mm mm nozzle nozzle for ing resolutionforUV of UV 0.05resin, resin, mm. which which The provided sensor provided structure an anin-plane in-plane was printing built printing by resolution 100% resolution infill of by 0.075 of the 0.075 mm3D printer mm and anda vertical a vertical print- and the flexureprintinging was resolution resolutionbuilt by of slender 0.05 of 0.05 mm. and mm. The thin Thesensor four-way sensor structure structuresupport was beams wasbuilt builtby to 100% withstand by 100% infill infillbythe the by 3D the printer 3D target load andprinterand to theobtain and flexure the a flexuremaximum was wasbuilt bending built by slender by effect. slender and andthin thin four-way four-way support support beams beams to towithstand withstand the thetarget target load load and and to to obtain obtain a maximum a maximum bending bending effect. effect.

Figure 2. Process to fabricate the printed load cell (PLC) structure and the nanocarbon piezoresistive composite flexure. FigureFigure 2. 2.Process Process to to fabricate fabricate the the printed printed load load cell cell (PLC) (PLC) structure structure and and the the nanocarbon nanocarbon piezoresistive piezoresistive composite composite ﬂexure. flexure.

Nano-Carbon Piezoresistive Composite (NCPC) Nano-Carbon Piezoresistive Composite (NC)

Electrode Electrode Figure 3. The 3D PLC using nanocarbon composite strain (diameter: 30 mm, height: 10 mm, re- Figure 3. The 3D PLC using nanocarbon composite strain (diameter: 30 mm, height: 10 mm, sistance: 920 kΩFigure). 3. The 3D PLC using nanocarbon composite strain (diameter: 30 mm, height: 10 mm, re- resistance: 920 kΩ). sistance: 920 kΩ). The sensory part of the load cell was fabricated by NCSS made of NCPC. We blended 0.5 wt% MWCNT (Hanhwa Chemical Co., Seoul, Korea, CM-280) and 99.5 wt% epoxy (Kukdo Chemical Co., Seoul, Korea, YD-128) to make the NCPC paint. The NCPC paint was prepared using a similar process to the one described in our previous study. The

Sensors 2021, 21, x FOR PEER REVIEW 5 of 13 Sensors 2021, 21, x FOR PEER REVIEW 5 of 13

The sensory part of the load cell was fabricated by NCSS made of NCPC. We blended Sensors 2021, 21, 3675 0.5 Thewt% sensory MWCNT part (Hanhwa of the load Chemical cell was Co.,fabricated Seoul, by Korea, NCSS CM-280) made of andNCPC. 99.5 We wt% blended epoxy5 of 12 0.5(Kukdo wt% MWCNT Chemical (Hanhwa Co., Seoul, Chemical Korea, YD-128) Co., Seoul, to make Korea, the CM-280) NCPC paint. and 99.5 The wt%NCPC epoxy paint (Kukdowas prepared Chemical using Co., a Seoul,similar Korea, process YD-128) to the one to makedescribed the NCPCin our previouspaint. The study. NCPC The paint ratio was prepared using a similar process to the one described in our previous study. The ratio ofratio the ofMWCNT the MWCNT and polymer and polymer was selected was selected according according to the percolation to the percolation threshold threshold experi- of the MWCNT and polymer was selected according to the percolation threshold experi- mentexperiment from our from previous our previous works, works, which whichhad a gauge had a gaugefactor factorof approximately of approximately 2 [18]. 2 [18]. ment from our previous works, which had a gauge factor of approximately 2 [18]. The NCPC paint was injected to groove on the flexure ﬂexure and cured to form the sensing The NCPC paint was injected to groove on the flexure and cured to form the sensing part of thethe PLC.PLC. AfterAfter the the curing curing process process for fo ther the sensing sensing part, part, two two electrical electrical wires wires were were con- part of the PLC. After the curing process for the sensing part, two electrical wires were connectednected at the at the end end point point of the of the NCPC NCPC layer layer using using silver silver conducting conducting epoxy epoxy (ElcoatCleantech, (ElcoatClean- connected at the end point of the NCPC layer using silver conducting epoxy (ElcoatClean- tech,Ansan, Ansan, Korea, Korea, Elcoat Elcoat P-100) P-100) to pick to uppick a piezoresistiveup a piezoresistive signal. signal. tech, Ansan, Korea, Elcoat P-100) to pick up a piezoresistive signal. 3. Experimental Experimental 3D-Printed 3D-Printed Load Cell Sensing Characteristics 3. Experimental 3D-Printed Load Cell Sensing Characteristics 3.1. Experimental Experimental Setup Setup 3.1. Experimental Setup To evaluate the PLC output characteristics for load changes, its voltage outputs were comparedTo evaluate toto aa commercial thecommercial PLC output industrial industrial characteristics miniature miniature for load load cell changes,cell (SETech, (SETech, its Seoul, voltage Seoul, Korea, outputs Korea, YC33-50K). wereYC33- compared50K).A test A jig test wasto ajig fabricatedcommercial was fabricated with industrial the with 3D printerthe miniature 3D toprinter add load weight to celladd on(SETech, weight the two on Seoul, overlapped the two Korea, overlapped load YC33- cells 50K).loadat the Acells sametest at jig time,the was same as fabricated illustrated time, as withillustrated in Figure the 3D4 .in printer The Figure 3D to PLC4. addThe voltage weight3D PLC responses on voltage the two wereresponses overlapped compared were loadcomparedto the cells reference at tothe the same loadreference time, cell; as theload illustrated responses cell; the in responses were Figure measured 4. were The 3Dmeasured by PLC general voltage by signal general responses processing, signal were pro- as comparedcessing,depicted as to in depicted Figurethe reference5. in Figure load 5. cell; the responses were measured by general signal processing, as depicted in Figure 5. (a) (b)

Figure 4. TheThe 3D 3D printed printed jig jig for for loading loading test: test: (a (a) )CAD CAD draft; draft; and and (b (b) )the the fabricated fabricated top top and and bottom bottom Figure 4. The 3D printed jig for loading test: (a) CAD draft; and (b) the fabricated top and bottom part of the Jig. part of the Jig.

FigureFigure 5. 5. 5.Schematic SchematicSchematic illustration illustration illustration of of the the of 3D 3D the printing printing 3D printing load load cell cell load sensin sensin cellg g sensingcharacteristics characteristics characteristics measurement measurement measure- setup. setup.ment setup.

As the ﬂexure was bent due to an external load, the deﬂection (δ) induced a resistance change (∆R) in the NCPC sensing part on the PLC, which was converted to micro voltage output through a Wheatstone quarter-bridge. The Wheatstone bridge can compensate

for temperature effects; a full bridge is commonly used to cancel the signals caused by Sensors 2021, 21, x FOR PEER REVIEW 6 of 13

Sensors 2021, 21, 3675 As the flexure was bent due to an external load, the deflection (δ) induced a resistance6 of 12 change (ΔR) in the NCPC sensing part on the PLC, which was converted to micro voltage output through a Wheatstone quarter-bridge. The Wheatstone bridge can compensate for extraneoustemperature forces. effects; In a thisfull study,bridge weis commonly used a quarter-bridge used to cancel to the monitor signals the caused basic by responses extra- ofneous the PLCforces. without In this anystudy, compensation. we used a quarte Ther-bridge voltage to output monitor from thethe basic Wheatstone responses of bridge the wasPLC amplifiedwithout any and compensation. filtered via a The data voltage acquisition output system from the (QuantumX-MX-840, Wheatstone bridge HBMwas am- Co., Marlborough,plified and filtered MA, via USA) a data with acquisition a 50 Hz system sampling (QuantumX-MX-8 frequency and40, 5HBM Hz low-passCo., Marlbor- filter. Figureough, 6MA, shows USA) the with experimental a 50 Hz sampling apparatus freq withuency the jig and and 5 dataHz low-pass acquisition filter. system. Figure 6 shows the experimental apparatus with the jig and data acquisition system.

FigureFigure 6. Experimental apparatus with with a a 3D 3D prin printedted jig jig and and a a data data acquisition acquisition system. system. 3.2. 3D PLC Sensing Characteristics 3.2. 3D PLC Sensing Characteristics The PLC sensing properties were characterized for indoor conditions based on the The PLC sensing properties were characterized for indoor conditions based on the standard calibration process of load cells while considering experimental constraints in an standard calibration process of load cells while considering experimental constraints in early laboratory-level development stage. High-precision standard weights were applied an early laboratory-level development stage. High-precision standard weights were ap- to the load cells by the 3D-printed jig and voltage outputs were measured. However, due plied to the load cells by the 3D-printed jig and voltage outputs were measured. However, to the limitations of the 3D-printed jig precision, a reference load cell was also used to due to the limitations of the 3D-printed jig precision, a reference load cell was also used evaluate the loading conditions. The PLC outputs were measured five times after reaching to evaluate the loading conditions. The PLC outputs were measured five times after reach- steady state and were averaged after eliminating the maximum and minimum data points. ing steady state and were averaged after eliminating the maximum and minimum data For the sensor characterization, the PLC response stability was first evaluated by points. measuring the voltage outputs under no loading and under a static load (0.2 kg), as For the sensor characterization, the PLC response stability was first evaluated by shown in Figure7. The PLC voltage output showed a stable response with some deviation measuring the voltage outputs under no loading and under a static load (0.2 kg), as shown (±0.03 mV/V, ±20 g). A noisy response has been reported for NCPC due to a resistance drift in Figure 7. The PLC voltage output showed a stable response with some deviation of the composite resistance originating from the inherent electrical changing properties of (±0.03 mV/V, ±20 g). A noisy response has been reported for NCPC due to a resistance CNTs based on the surrounding environment [31]. However, the electrical signal stability of drift of the composite resistance originating from the inherent electrical changing proper- PLC with NCSS is sufficient for engineering applications, as we reported previously [18,19]. ties of CNTs based on the surrounding environment [31]. However, the electrical signal The response times of the two load cells, presented in Figure7b, were nearly identical. stability of PLC with NCSS is sufficient for engineering applications, as we reported pre- A series of weights was successively added to the PLC to qualitatively determine its viously [18,19]. The response times of the two load cells, presented in Figure 7b, were characteristics; the responses are shown in Figure8. Reproducible responses were exhibited nearly identical. by the PLC, which showed a drift problem similar to other polymer-based strain sensors. A series of weights was successively added to the PLC to qualitatively determine its The PLC response was initially well matched with the reference load cell; however, after characteristics; the responses are shown in Figure 8. Reproducible responses were exhib- some time, gaps were found due to the drift. The drift seemed to be caused by hysteresis, ited by the PLC, which showed a drift problem similar to other polymer-based strain sen- because it mostly happened during unloading and later. Other studies have reported that sors. The PLC response was initially well matched with the reference load cell; however, the drift response is affected by the inherent characteristics of polymers [32,33].

Sensors 2021, 21, x FOR PEER REVIEW 7 of 13

after some time, gaps were found due to the drift. The drift seemed to be caused by hys- Sensors 2021, 21, 3675 teresis,after some because time, it gaps mostly were happened found due during to th eunloading drift. The anddrift later.seemed Other to be studies caused have by hys- re-7 of 12 portedteresis, thatbecause the driftit mostly response happened is affected during by unloading the inherent and characteristicslater. Other studies of polymers. have re- [32,33].ported that the drift response is affected by the inherent characteristics of polymers. [32,33].

FigureFigure 7. 7.Response Response proﬁlesprofiles of the 3D 3D PLC PLC and and the the commercial commercial load load cell: cell: (a) voltage (a) voltage outputs outputs under under no loading; no loading; and (b) and voltageFigure outputs7. Response under profiles a static of load the 0.23D kgPLC (step and response). the commercial load cell: (a) voltage outputs under no loading; and (b) (b) voltage outputs under a static load 0.2 kg (step response). voltage outputs under a static load 0.2 kg (step response).

Figure 8. A successive loading test of the PLC compared with YC-33. Figure 8. A successive loading test of the PLC compared with YC-33. Figure 8. A successive loading test of the PLC compared with YC-33. The resistance of NCPC changes with temperature fluctuations [34]. These changes in resistanceTheThe resistance resistance may affect ofof NCPC NCPC the output changeschanges voltage withwith temperature temperaturefrom the temperature ﬂuctuationsfluctuations of [[34].34the]. PLC.These Figure changes 9 in resistanceshowsin resistance the maychange may affect ratio affect the of outputthethe outputresistan voltage voltagece in from relation from the temperaturetothe the temperature temperature. of the of PLC.the PLC. Figure Figure9 shows 9 theshows change the ratiochange of ratio the resistance of the resistan in relationce in relation to the to temperature. the temperature. The effect of temperature on the output characteristics of a traditional load cell is the biggest factor impairing accuracy. Thus, the PLC output voltage also needs to be adjusted relative to temperature changes. The temperature dependency of PLC resistance tends to change linearly to 50 ◦C, after which it remarkably changes. The temperature sensitivity was found to be −0.4 × 10−3 K−1, which was comparable to other 0.5 wt% MWCNT/epoxy cases that had temperature sensitivities of −0.6 × 10−3 K−1 [35]. Considering the polymeric characteristics of PLC, its operating temperature can be restricted beyond the allowable temperature range (~70 ◦C) of industrial load cells. The output voltages were measured to evaluate the linearity and sensitivity of PLC. To begin the characterization test of the PLC, we applied loads up to the theoretical maximum (2 kg) to investigate its behavior, as depicted in Figure 10. Sensors 2021, 21, x FOR PEER REVIEW 8 of 13

Sensors 2021, 21, 3675 8 of 12

The PLC held up to the analytical maximum load of 2 kg. However, it showed nonlinearly for a loaded response above the 1 kg range and also exhibited hysteresis. From this test, the rated load span was conservatively determined to be 0.8 kg to protect the Sensors 2021, 21, x FOR PEER REVIEWﬂexure and to obtain a linear response. The PLC characteristics were tested again within8 of 13 the rated load range, as shown in Figure 11.

Figure 9. Temperature dependence of the PLC.

The effect of temperature on the output characteristics of a traditional load cell is the biggest factor impairing accuracy. Thus, the PLC output voltage also needs to be adjusted relative to temperature changes. The temperature dependency of PLC resistance tends to change linearly to 50 °C, after which it remarkably changes. The temperature sensitivity was found to be −0.4 × 10−3 K−1, which was comparable to other 0.5 wt% MWCNT/epoxy cases that had temperature sensitivities of –0.6 × 10−3 K−1 [35]. Considering the polymeric characteristics of PLC, its operating temperature can be restricted beyond the allowable temperature range (~70 °C) of industrial load cells. The output voltages were measured to evaluate the linearity and sensitivity of PLC. To begin the characterization test of the PLC, we applied loads up to the theoretical maximum (2 kg) to investigate its behavior, as depicted in Figure 10. The PLC held up to the analytical maximum load of 2 kg. However, it showed nonlinearly for a loaded response above the 1 kg range and also exhibited hysteresis. From this test, the rated load span was conservatively determined to be 0.8 kg to protect the flexure and to obtain a linear response. The PLC characteristics were tested again within the rated load range, as shown in Figure 11. Figure 9. Temperature dependencedependence ofof thethe PLC.PLC.

The effect of temperature on the output characteristics of a traditional load cell is the (a) biggest factor impairing accuracy.(b) Thus, the PLC output voltage also needs to be adjusted relative to temperature changes. The temperature dependency of PLC resistance tends to change linearly to 50 °C, after which it remarkably changes. The temperature sensitivity was found to be −0.4 × 10−3 K−1, which was comparable to other 0.5 wt% MWCNT/epoxy cases that had temperature sensitivities of –0.6 × 10−3 K−1 [35]. Considering the polymeric characteristics of PLC, its operating temperature can be restricted beyond the allowable temperature range (~70 °C) of industrial load cells. The output voltages were measured to evaluate the linearity and sensitivity of PLC. To begin the characterization test of the PLC, we applied loads up to the theoretical maximum (2 kg) to investigate its behavior, as depicted in Figure 10. The PLC held up to the analytical maximum load of 2 kg. However, it showed non- Figure 10. Loading/unloading test of the PLC to the maximum analytical load, 2 kg: (a) hysteresis characteristic; and (b) 5 Sensors 2021Figure, 21, 10.x FORLoading/unloading PEER REVIEW linearly test offor the a loaded PLC to theresponse maximum above analytical the 1 kg load, range 2 kg: and (a) hysteresisalso exhibited characteristic; hysteresis.9 and of 13From time measurements. (b) 5 time measurements. this test, the rated load span was conservatively determined to be 0.8 kg to protect the flexure and to obtain a linear response. The PLC characteristics were tested again within the rated load range, as shown in Figure 11. (a) (b)

Figure 11. Loading/unloading test of the PLC to the rated load, 0.8 kg: (a) 5 time measurements; and (b) hysteresis char- Figure 11. Loading/unloading test of the PLC to the rated load, 0.8 kg: (a) 5 time measurements; and (b) hysteresis acteristic. characteristic. Figure 10. Loading/unloading test of the PLC to the maximum analytical load, 2 kg: (a) hysteresis characteristic; and (b) 5 At the rated load range, the PLC showed a linear response, including hysteresis. The time measurements. PLC output was evaluated to the most common value of 2 mV/V. The PLC performance parameters were derived based on this test, as depicted in Figure 12. The first preliminary evaluation (Vout1) is presented in Figure 11. It was fitted to a second-order polynomial, shown in Equation (1), in order to compensate for the PLC output. The compensated PLC output is presented in Figure 13.

Figure 12. PLC performance non-compensated characterization.

Figure 13. The calibrated PLC performance characterization.

Sensors 2021, 21, x FOR PEER REVIEW 9 of 13

Sensors 2021, 21, 3675 9 of 12 Figure 11. Loading/unloading test of the PLC to the rated load, 0.8 kg: (a) 5 time measurements; and (b) hysteresis characteristic.

AtAt the the rated rated load load range, range, the the PLC PLC showed showed a a linear linear response, response, including including hysteresis. hysteresis. The The PLCPLC output output was was evaluated evaluated to to the the most most common common value value of of 2 2 mV/V. mV/V. The The PLC PLC performance performance parametersparameters were were derived derived based based on on this this test, test, as as depicted depicted in in Figure Figure 12 12.. V out1 TheThe ﬁrstfirst preliminary preliminary evaluationevaluation ((Voutout1) is presented in Figure 11.11. It It was was fitted ﬁtted to to a asecond-order second-order polynomial, polynomial, shown shown in in Equation Equation (1), (1), in inorder order to tocompensate compensate for forthe thePLC PLC out- output.put. The The compensated compensated PLC PLC output output is presented is presented in inFigure Figure 13. 13 .

FigureFigure 12. 12.PLC PLC performance performance non-compensated non-compensated characterization. characterization.

FigureFigure 13. 13.The The calibrated calibrated PLC PLC performance performance characterization. characterization. 2 Vout2 = 0.09V out1 + 0.87Vout1 − 0.02 (1) After the compensation, the linearity improved. The end point linearity was 2.12% for full-scale non-linearity, which was less than that of a conventional load cell. To test the creep, we only considered time and tested the load cell output voltage under the maximum rated load of 0.8 kg at room temperature without considering other environmental conditions. Figure 14 shows the creep test. Because the ﬂexure was made of epoxy resin, creep was expected to occur when a load was applied for a period of time. Creep may be more considerable for PLC than for a commercial a load cell made of metal. Sensors 2021, 21, x FOR PEER REVIEW 10 of 13

2 VV=+−0.09 0.87 V 0.02 (1) out211 out out After the compensation, the linearity improved. The end point linearity was 2.12% for full-scale non-linearity, which was less than that of a conventional load cell. To test the creep, we only considered time and tested the load cell output voltage under the maximum rated load of 0.8 kg at room temperature without considering other Sensors 2021, 21, 3675 environmental conditions. Figure 14 shows the creep test. Because the flexure was10 made of 12 of epoxy resin, creep was expected to occur when a load was applied for a period of time. Creep may be more considerable for PLC than for a commercial a load cell made of metal.

Figure 14.14. PLC creep test under maximum ratedrated loadload 0.80.8 kg.kg.

An international standard, OIML R 60, recommends that creepcreep shouldshould bebe evaluatedevaluated using data measured for more than 30 minmin [[36].36]. We alsoalso currentlycurrently suffersuffer difﬁcultiesdifficulties toto obtain long term signal stability due to the viscoelastic properties of polymer. Therefore, only limited steady-state data were availabl availablee to test the creep of PLC. We diddid aa simplesimple creep test to to remove remove the the load load applied applied to tothe the PLC PLC after after applying applying a load a load of 0.8 of kg 0.8 to kg the to PLC the PLCin the in no-load the no-load state. state.The creep The creepand creep and creeprecove recoveryry values values can be can obtained be obtained as “(peak as“(peak value- value-stablestable value)/rated value)/rated output”. output”. Creep Creepand creep and creeprecovery recovery were werefound found to be to1.68 be (%FS) 1.68 (%FS) and and4.16 4.16(%FS). (%FS). The PLC The PLCcreep creep behavior behavior was wasnot comparable not comparable to that to thatof a commercial of a commercial load loadcell, cell,but it but can it be can improved be improved by compensation. by compensation. Based onon thethe parametersparameters conceptually conceptually evaluated evaluated in in this this preliminary preliminary laboratory laboratory inves- in- tigationvestigation (Table (Table2), the 2), performancethe performance of the of PLCthe PLC does does not yetnot meet yet meet industrial industrial market market criteria, cri- whichteria, which require require less than less 1%than [1 1%]. The [1]. inferiorThe inferior results results found found for the for PLC the PLC may may be due be due to the to inherentthe inherent hyper hyper elastic elastic characteristics characteristics of of polymer polymer sensors. sensors. A A future future study study is is plannedplanned toto investigate approaches to improve PLCPLC performance,performance, including compensation for drift, hysteresis, and temperature based on characteristic models. hysteresis, and temperature based on characteristic models. Table 2. PLC performance. Table 2. PLC performance.

ParameterParameter Non-Linearity Non-Linearity [%] [%]Repeatability Repeatability [%] [%]Hysteresis Hysteresis [%] [%] EvaluationEvaluation (1st) (1st) 6.48 6.481.58 1.584.27 4.27 EvaluationEvaluation (compensated) (compensated) 2.12 2.121.60 1.604.42 4.42 MarketMarket requirement requirement [1] [ 1 ]0.25~1 0.25~1 0.1~0.5 0.1~0.5 0.25~1 0.25~1

4. Conclusions The development of a 3D PLC was studied using a nanocarbon composite strain sensor (NCSS) and a 3D printing process. A miniaturized load cell with a simple beam flexure could be easily fabricated using a 3D printing process. To design the sensory part of the load cell, the structural behaviors of the flexure were analyzed using a finite element method. Its loading capacity was conservatively determined to be less than 1 kg based on the strength of the resin epoxy. The miniature load cell was fabricated using a low-cost LCD-based 3D printer with UV resin. An NCSS composed of NCPC with 0.5 wt% MWCNT/epoxy was used to create Sensors 2021, 21, 3675 11 of 12

the flexure. The sensing performance of the PLC was studied in terms of its voltage output characteristics under a 0.8 kg load, which was the linear loading response range. A zero- balance voltage output was stable with some deviation (±0.01 mV/V, ±20 g) due to the inherent noisy response of NCPC. The response time was nearly identical to that of a commercial load cell. The PLC showed drift problems in a successive loading/unloading test, as well as temperature dependence (−0.4 × 10−3 K−1). PLC performance was evaluated under a rated load range; its output was equal to the common value of 2 mV/V. The performance was also evaluated after calibration in terms of non-linearity, repeatability, and hysteresis, with final results of 2.12%, 1.60%, and 4.42%, respectively; these results do not yet meet industrial market criteria. Creep and creep recovery were found to be 1.68 (%FS) and 4.16 (%FS). PLC creep behavior was not comparable to that of a commercial load cell, but it can be improved by compensation. The relative inferiorities of PLC seem to originate from the inherent hyper elastic characteristics of polymer sensors. Thus, a future study is planned to investigate compensation for drift, hysteresis, and temperature based on characteristic models. Because of its excellent manufacturing and piezoresistive properties, an NCSS made of NCPC was investigated for use in the sensory part of a 3D PLC’s flexure. Our results indicate that NCPC and the 3D printing process may be promising tools in the development of novel load cells. A customized circuit and compensation model may be required to achieve an accurate performance. To develop a completely 3D-printed load cell, we are investigating a 3D-printed filament made of NCPC that allows both the flexure and sensory parts to be fabricated simultaneously through the 3D printing process. Further study of the flexure design is planned to determine a customized load capacity. We are also planning a sensing compensation study to obtain reliable signals based on the characteristic model, which will improve the NCPC-based 3D PLC for practical applications. The 3D PLC developed in this study may be a promising solution for the OEM/design-in load cell market and may also result in the development of a novel 3D-printed sensor.

Author Contributions: Conceptualization, K.-Y.J., S.-Y.K., I.K. and S.-H.C.; methodology, S.-Y.K. and K.-Y.J.; formal analysis, S.-Y.K. and I.K.; investigation, S.-Y.K. and K.-Y.J.; data curation, S.-Y.K. and I.K.; writing—original draft preparation, K.-Y.J.; writing—review and editing, I.K. and S.-H.C.; visualization, S.-Y.K. and K.-Y.J.; project administration, S.-H.C. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: Data available in a publicly accessible repository. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References 1. PCB Load & Torque Division. Load Cell Handbook: A Technical Overview and Selection Guide, 1st ed.; PCB: Depew, NY, USA, 2014; p. 4. 2. OMEGA. Available online: https://kr.omega.com/prodinfo/loadcells.Html (accessed on 15 March 2021). 3. Higson, G.R. Recent Advances in Strain Gauges. J. Sci. Instrum. 1964, 41, 405–414. [CrossRef] 4. Yang, S.; Lu, N. Gauge Factor and Stretchability of Silicon-on-Polymer Strain Gauges. Sensors 2013, 13, 8577–8594. [CrossRef] 5. Mangayil, R.; Rajala, S.; Pammo, A.; Sarlin, E.; Luo, J.; Santala, V.; Karp, M.; Tuukkanen, S. Engineering and Characterization of Bacterial Nanocellulose Films as Low Cost and Flexible Sensor Material. ACS Appl. Mater. Interfaces 2017, 9, 19048–19056. [CrossRef][PubMed] 6. Lim, H.-R.; Kim, H.-S.; Qazi, R.; Kwon, Y.-T.; Jeong, J.W.; Yeo, W.-H. Advanced Soft Materials, Sensor Integrations, and Applications of Wearable Flexible Hybrid Electronics in Healthcare, Energy, and Environment. Adv. Mater. 2019, 32, 1901924. [CrossRef][PubMed] 7. Pan, S.; Pei, Z.; Jing, Z.; Song, J.; Zhang, W.; Zhang, Q.; Sang, S. A Highly Stretchable Strain Sensor Based on CNT/graphene/ Fullerene-SEBS. RSC Adv. 2020, 10, 11225–11232. [CrossRef] Sensors 2021, 21, 3675 12 of 12

8. Ali, A.; Ali, F.; Irfan, M.; Muhammad, F.; Glowacz, A.; Antonino-Daviu, J.A.; Caesarendra, W.; Qamar, S. Mechanical Pressure Characterization of CNT-Graphene Composite Material. Micromachines 2020, 11, 1000. [CrossRef] 9. Sheng, N.; Ji, P.; Zhang, M.; Wu, Z.; Liang, Q.; Chen, S.; Wang, H. High Sensitivity Polyurethane-Based Fiber Strain Sensor with Porous Structure via Incorporation of Bacterial Cellulose Nanofibers. Adv. Electron. Mater. 2021, 7, 2001235. [CrossRef] 10. Kang, I.; Schulz, M.J.; Kim, J.H.; Shanov, V.; Shi, D. A Carbon Nanotube Strain Sensor for Structural Health Monitoring. Smart Mater. Struct. 2006, 15, 737–748. [CrossRef] 11. Salvado, R.; Lopes, C.; Szojda, L.; Araujo, P.; Gorski, M.; Velez, F.J.; Castro-Gomes, J.; Krzywon, R. Carbon Fiber Epoxy Composites for Both Strengthening and Health Monitoring of Structures. Sensors 2015, 15, 10753–10770. [CrossRef] 12. Hu, N.; Itoi, T.; Akagi, T.; Kojima, T.; Xue, J.; Yan, C.; Atobe, S.; Fukunaga, H.; Yuan, W.; Ning, H.; et al. Alamusi Ultrasensitive Strain Sensors Made from Metal-Coated Carbon nanofiller/Epoxy Composites. Carbon 2013, 51, 202–212. [CrossRef] 13. Ferreira, A.; Mendez, S.L. Piezoresistive Response of Spray-Printed Carbon nanotube/Poly(vinylidene Fluoride) Composites. Compos. Part B Eng. 2016, 96, 242–247. [CrossRef] 14. Le, T.H.; Lee, D.H.; Kim, J.H.; Park, S.J. Polypyrrole/Graphene Quantum Dot Composites as a Sensor Media for Epinephrine. J. Nanosci. Nanotechnol. 2020, 20, 4005–4010. [CrossRef][PubMed] 15. Park, M.; Kim, H.; Youngblood, J.P. Strain-Dependent Electrical Resistance of Multi-Walled Carbon Nanotube/Polymer Composite Films. Nanotechnology 2008, 19, 5. 16. Li, C.; Thostenson, E.T.; Chou, T.-W. Dominant Role of Tunneling Resistance in the Electrical Conductivity of Carbon nanotube– based Composites. Appl. Phys. Lett. 2007, 91, 223114. [CrossRef] 17. Park, S.-H.; Hwang, J.; Park, G.-S.; Ha, J.-H.; Zhang, M.; Kim, D.; Yun, D.-J.; Lee, S.; Lee, S.H. Modeling the Electrical Resistivity of Polymer Composites with Segregated Structures. Nat. Commun. 2019, 10, 1–11. [CrossRef] 18. Choi, G.; Lee, J.W.; Cha, J.Y.; Kim, Y.-J.; Choi, Y.-S.; Schulz, M.J.; Moon, C.K.; Lim, K.T.; Kim, S.Y.; Kang, I. A Spray-On Carbon Nanotube Artificial Neuron Strain Sensor for Composite Structural Health Monitoring. Sensors 2016, 16, 1171. [CrossRef] 19. Kim, S.Y.; Choi, B.G.; Baek, W.K.; Park, S.H.; Park, S.W.; Shin, J.W.; Kang, I. Impact Paint Sensor Based on Polymer/Multi- Dimension Carbon Nano Isotopes Composites. Smart Mater. Struct. 2019, 28, 035025. [CrossRef] 20. Giovanni, P.; Gabriele, N.; Gianmarco, G.; Marinella, L.; Stefano, T. Conductive 3D Microstructures by Direct 3D Printing of polymer/Carbon Nanotube Nanocomposites via Liquid Deposition Modeling. Compos. Part A Appl. Sci. Manuf. 2015, 76, 110–114. 21. Zhang, D.; Chi, B.; Li, B.; Gao, Z.; Du, Y.; Guo, J.; Wei, J. Fabrication of Highly Conductive Graphene Flexible Circuits by 3D Printing. Synth. Met. 2016, 217, 79–86. [CrossRef] 22. John, M.H.; Godfrey, S.; Kim, J.W.; Roberto, J.C.; Russel, A.W.; Christopher, J.S.; Brian, W.G.; Dennis, C.W.; Emilie, J.S. 3-D Printing of Multifunctional Carbon Nanotube Yarn Reinforced Components. Addit. Manuf. 2016, 12, 38–44. 23. Rafael, M.C.; Cristiane, K.; Raquel, G.R.; Pãmyla, L.S.; Diego, P.R.; Paulo, R.O.; Bruno, C.J.; Juliano, A.B.; Eduardo, M.R.; Rodrigo, A.A.M. Additive-Manufactured (3D-Printed) Electrochemical Sensors: A Critical Review. Anal. Chim. Acta 2020, 29, 73–91. 24. Kim, K.; Park, J.; Suh, J.-H.; Kim, M.; Jeong, Y.; Park, I. 3D printing of multiaxial force sensors using carbon nanotube (CNT)/thermoplastic polyurethane (TPU) filaments. Sens. Actuators A Phys. 2017, 263, 493–500. [CrossRef] 25. Christ, J.F.; Aliheidari, N.; Pötschke, P.; Ameli, A. Bidirectional and Stretchable Piezoresistive Sensors Enabled by Multimaterial 3D Printing of Carbon Nanotube/Thermoplastic Polyurethane Nanocomposites. Polymers 2018, 11, 11. [CrossRef][PubMed] 26. Al-Rubaiai, M.K.; Tsuruta, R.; Gandhi, U.; Wang, C.; Tan, X. A 3D-Printed Stretchable Strain Sensor for Wind Sensing. Smart Mater. Struct. 2019, 28, 084001. [CrossRef] 27. Stano, G.; Nisio, A.D.; Lanzolla, A.; Percoco, G. Additive Manufacturing and Characterization of a Load Cell with Embedded Strain Gauges. Precis. Eng. 2020, 62, 113–120. [CrossRef] 28. Qu, J.; Wu, Q.; Clancy, T.; Fan, Q.; Wang, X.; Liu, X. 3D-Printed Strain-Gauge Micro Force Sensors. IEEE Sens. J. 2020, 20, 6971–6978. [CrossRef] 29. Kim, S.Y.; Kang, I. A Study on the Development of a Novel Pressure Sensor Based on Nano Carbon Piezoresistive Composite by Using 3D Printing. Trans. Korean Soc. Mech. Eng. A 2017, 41, 187–192. 30. 4thwave. Available online: https://4thwave3d.com/product/Resin/ (accessed on 23 May 2021). 31. Kordas, K.; Mustonen, T.; Toth, G.; Jantunen, H.; Lajunen, M.; Soldano, C.; Talapatra, S.; Kar, S.; Vajtai, R.; Ajayan, P.M.; et al. Inkjet Printing of Electrically Conductive Patterns of Carbon Nanotubes. Small 2006, 2, 1021–1025. [CrossRef][PubMed] 32. Christ, J.F.; Aliheidari, N.; Ameli, A.; Pötschke, P. 3D Printed Highly Elastic Strain Sensors of Multiwalled Carbon Nano- tube/Thermoplastic Polyurethane Nanocomposites. Mater. Des. 2017, 131, 394–401. [CrossRef] 33. Oliveri, A.; Maselli, M.; Lodi, M.; Storace, M.; Cianchetti, M. Model-Based Compensation of Rate-Dependent Hysteresis in a Piezoresistive Strain Sensor. IEEE Trans. Ind. Electron. 2018, 66, 8205–8213. [CrossRef] 34. Dehghani, S.; Moravvej-Farshi, M.K.; Sheikhi, M.H. Temperature Dependence of Electrical Resistance of Individual Carbon Nanotubes and Carbon Nanotubes Network. Mod. Phys. Lett. B 2012, 26, 1250136. [CrossRef] 35. Neitzert, H.C.; Vertuccio, L.; Sorrentino, A. Epoxy/MWCNT Composite as Temperature Sensor and Electrical Heating Element. IEEE Trans. Nanotechnol. 2011, 10, 688–693. [CrossRef] 36. International Organization of Legal Metrology. Metrological Regulation for Load Cells; OIML: Paris, France, 2000.