3-D-Printing and Reliability Evaluation of an Easy-to-Fabricate Position Sensing System for Printed Functional Wearable Assistive Devices

The aging population has increased the demand for advanced medical assistive devices with actuators and sensors to enhance independence in daily activities. However, the integration of sensors into rehabilitation devices in a direct and cost-effective manner remains a challenge. This study proposes a scalable linear encoder, fabricated using 3-D-printing technology. The encoder is a fully printed concept, including the electrical circuitry, and is benchmarked in a sensorized hand orthosis (sHO). It is constructed using two commercially available materials: one electrically conductive and one nonconductive. The encoder utilizes an encoder setup with a fixed scale and a sliding sensing head (SH). Encoder’s resolution and robustness are tested under various conditions, including actuation speeds, temperatures, printing repeatability, and fatigue. The results demonstrate the stable operation of the encoder, providing a resolution of 1.2 mm. The sHO incorporates three encoders with geometric shifts, along with a printed electrical circuit and inserted microcontroller for detecting finger flexion and movement direction. This setup further improves the resolution to 0.4 mm. The proposed linear encoder offers an effective and inexpensive approach for integrating sensors into medical assistive devices.


3-D-Printing and Reliability Evaluation of an Easy-to-Fabricate Position Sensing System for Printed Functional Wearable Assistive Devices
Paweł Michalec and Lisa-Marie Faller Abstract-The aging population has increased the demand for advanced medical assistive devices with actuators and sensors to enhance independence in daily activities.However, the integration of sensors into rehabilitation devices in a direct and cost-effective manner remains a challenge.This study proposes a scalable linear encoder, fabricated using 3-D-printing technology.The encoder is a fully printed concept, including the electrical circuitry, and is benchmarked in a sensorized hand orthosis (sHO).It is constructed using two commercially available materials: one electrically conductive and one nonconductive.The encoder utilizes an encoder setup with a fixed scale and a sliding sensing head (SH).Encoder's resolution and robustness are tested under various conditions, including actuation speeds, temperatures, printing repeatability, and fatigue.The results demonstrate the stable operation of the encoder, providing a resolution of 1.2 mm.The sHO incorporates three encoders with geometric shifts, along with a printed electrical circuit and inserted microcontroller for detecting finger flexion and movement direction.This setup further improves the resolution to 0.4 mm.The proposed linear encoder offers an effective and inexpensive approach for integrating sensors into medical assistive devices.Index Terms-3-D-printing, linear encoder, material extrusion, position sensor, printed electronics, sensorized hand orthosis (sHO), wearable assistive device.

I. INTRODUCTION
W ITH the increasing life expectancy, the demand for advanced wearable medical assistive devices is steadily rising [1].Several health conditions, including carpal tunnel syndrome [2], stroke [3], and various muscle disorders [4], can result in upper limb disabilities, significantly impacting the quality of life and independence of affected individuals.Consequently, they require rehabilitation and support with activities of daily living (ADLs) [5], [6].The aging population's growth also necessitates an increased number of therapy sessions.Therefore, therapists often rely on different types of wearable assistive devices for patient rehabilitation, assistance, and support, such as motorized exoskeletons and dynamic hand orthoses that utilize passive elements for actuation.These devices primarily focus on joint stabilization [7], movement assistance [8], or facilitation of human-computer interfaces [9].There are various approaches to achieving these functionalities, but sensors consistently play a critical role and are highly relevant [10], [11], [12].They are crucial in motor control, safety, and monitoring rehabilitation progress [13], [14].Furthermore, the collected data are utilized to visualize successfully accomplished rehabilitation goals, thereby increasing user motivation [15].Novak and Riener [16] emphasized that even when actuators have their own built-in sensors, motor control alone often proves insufficient for many use cases.Incorporating direct measures into the rehabilitation system enhances the detection of wearer intentions, leading to improved recovery outcomes.

A. State-of-the-Art
Currently, rehabilitation faces many challenges, including economic affordability, lack of equal accessibility to professionals, the need for feedback and progress monitoring, and the decreased time of training sessions due to the high occupancy of therapists [17], [18].These necessities result in a growing interest of research in wearable rehabilitation devices [10].Additive manufacturing (AM), such as 3-Dprinting, is considered as a method able to solve some of the listed challenges.In this process, by placing a material layer by layer, the final object is built.Devices made by AM technologies [19], [20], [21] are widely used in the medical field and are considered to be cost and time effective.In addition, AM offers the possibility of a high degree of customization while providing significant design freedom [22].These features position AM as an excellent technology for producing sensors intended for medical applications [23].
There are no standards about the sensor type used in rehabilitation devices, and thus, many research prototypes utilize various commercially available sensors beginning from potentiometers, Hall effect sensors, strain gauges, magnetic encoders, and force sensors within hand exoskeletons such as those described in [24] and [25].Moreover, in other rehabilitation robots, researchers apply flex sensors [26], [27] and smart fibers [28].In many 3-D-printed rehabilitation devices, commercially available sensors are employed for feedback.They consist of encoders [29], [30], potentiometers [31], [32], rotary position sensors [33], and pressure sensors [34], [35], [36].These sensors are usually built into the actuator or are embedded or assembled after printing into a 3-D-printed structure [37], which results in additional workload or limits their usage [16].
Even though numerous works focus on 3-D-printed sensors, they are not used in 3-D-printed devices but are rather printed either on elastic flat surfaces or as stand-alone platforms, requiring an additional step to assemble them into a device [38].However, direct implementation of sensors to the device can be approached by usage of multimaterial printing [39].There are several types of 3-D-printing techniques; however, the most used commercially available printers for multimaterial printing are material extrusion printers based on three technologies: fused filament fabrication (FFF), PolyJet, and direct ink writing [40].Nevertheless, no work was found presenting rehabilitation device with directly printed sensor.The major drawbacks of many 3-D-printed sensors are complex manufacturing processes, poor surface quality, and lack of print repeatability, which strongly influence sensor performance [41].This still counteracts their general application in 3-D-printed rehabilitation devices.Strain gauges developed by Borghetti et al. [42], Kouchakzadeh and Narooei [43], or Xiang et al. [44], piezoelectric sensors [45], or eddy current position sensors [46] also suffer from the aforementioned difficulties.On the other hand, Herrojo et al. [47] developed an electromagnetic encoder produced via FFF, which is available at low cost and single process printed and does not require complex postprocessing, therefore addressing current challenges.However, the achieved resolution of 3.4 mm may limit its application in rehabilitation settings.For wearable hand assistive devices, it is important to provide a sufficiently high resolution, as inaccuracies can lead to health hazards and subsequently to injuries [48].
The need for rehabilitation accessibility for people who suffer from a variety of limb impairments [49] leads many researchers to focus on inexpensive 3-D-printed devices, which might still often have a complex structure, but can all be personalized to the user.Although it is generally necessary to incorporate sensors for control and feedback in such devices [14], [50], the choice of sensor type is not limited.However, there is a shortage of readily implementable solutions for sensorized devices that do not require complex postprocessing.This simplicity is essential for future implementation by nontechnical personnel.

B. Contribution
The main objective of this article is to develop a 3-D-printed sensor for rehabilitation applications, addressing the current challenges in this field.The aim is to create a sensor that does not require complex postprocessing and is resilient to printing geometrical variation while also offering improved resolution compared to existing devices.Therefore, we propose a single-process 3-D-printed linear encoder based on an electrical principle.The encoder is constructed using the FFF technology, which is well-suited due to its accessibility, cost-effectiveness, process simplicity, and capability for multimaterial printing.These characteristics lay the groundwork for future implementations in hospitals and rehabilitation clinics [51], [52].
In addition, several test stands are constructed to assess the resolution, fatigue, and temperature performance of the encoder.To demonstrate its functionality, the encoder is directly integrated into a 3-D-printed dynamic hand orthosis, referred to as the sensorized hand orthosis (sHO), and subsequently evaluated.The rest of this article is structured as follows.Section II presents the operational principle, design, and practical implementation of the proposed encoder.Section III provides a detailed description of the experimental procedure.The results are outlined in Section IV.Subsequently, the results are discussed in Section V, and the main conclusions are highlighted.

II. DESIGN AND FABRICATION A. Functional Principle
Linear encoders are commonly used to measure linear displacement [53], and one of their key benefits lies in their adaptability to be resized according to specific application requirements.There are various ways to construct these devices for encoding position or distance.In our approach, the sensor consists of two distinct components: a scale and a sensing head (SH).The scale is comprised of alternating electrically conductive and insulating stripes.All the electrically conductive stripes are connected together, and their output is read by a microcontroller.The SH includes an electrically conductive tip, which is connected to the voltage through the input pin [Fig.1(a)].As the SH tip slides along the surface of the scale, the voltage increases when it touches the conductive stripes and decreases when it comes into contact with the insulating stripe.This results in the implementation of an on/off working principle for the sensor.By knowing the distance between the stripes, i.e., resolution (res), and counting the number of spikes (NS), the distance traveled by the SH can be calculated and, consequently, its current position known [Fig.1(b)].The output of the encoder is connected to the voltage divider [Fig.1(c)] in order to read the signal from the sensor.The input of the SH is connected to the microcontroller's 5-V output.The output of the scale is further connected in parallel to AnalogRead and it is grounded with the resistor R vd before.Based on the resistance of the sensor, the resistance R vd of a resistor has to be adjusted in such a way that it is significantly higher than the sensor's resistance in order to distribute most of the voltage to AnalogRead while having the circuit grounded.
Various types of encoders and two test stands (encoder evaluation) are designed to better understand the dependence of the sensor performance in terms of its resolution, contact force (F), and dynamic behavior as well as the environmental temperature.Next, the type of encoder with the highest percentage of detected stripes is chosen and printed within the sHO and its performance is benchmarked in the addressed application.

B. Encoder Evaluation
To evaluate how the geometry influences the performance of the encoder, four types of scales (S 1 -S 4 ) and three types of sensing heads (SH 1 , SH 12 , and SH 2 ) are proposed (Fig. 2).It is identified that the signal readout quality depends on the sensor resolution and the quality of contact between both parts of the sensor.
One of the factors affecting the quality of contact, hence the signal quality, is the applied contact force, which depends on the position of the SH with respect to the fixed scale.This distance is defined as z and it is indicated in Fig. 1(d).By increasing z, the contact force of the tip on the surface of the scale increases, and thus, it is expected that also the electrical connection is improved.By increasing the thickness of the SH (d), higher contact force is also obtained.However, an excessive load may lead to increased wear of the tip and, consequently, result in a faulty readout.
In order to examine the effect of F on the readout, SHs of thickness 0.6 mm (SH 1 ) and 0.8 mm (SH 2 ), which is a difference of one printed layer height (0.2 mm), are proposed.Moreover, for SH 1 distance, blocks of 0.2 mm in height are added to manipulate the z value, hence F. This set is called SH 12 .The calculated contact forces with a formula for a beam deflection are as follows: F SH1 = 0.07 N, F SH12 = 0.10 N, and The resolution (res), which is a distance between the conductive stripes, is another variable influencing the encoder performance.It is investigated by designing four different patterns of scales.All the patterns have the same length (36 mm) and differ from each other by the width of the conductive and insulating stripes, which results in a different number of conductive stripes on the scale.The values are based on the minimal printing width, which is 0.4 mm.All the main pattern dimensions are shown in Table I.
The measurement quality of the proposed encoders is evaluated with a specifically developed platform, which is further referred to as a basic test stand, on which the sensors are mounted [Fig.1(a)].The movement of the SH over the scale is constrained by commercially available sliders (Igus Drylin NK-02-17), while the scale is fixed to the base plate.The sliding part is connected to the robotic arm by the slider mount.
The encoder is printed on the Pro2 Plus printer (Raise3D, Irvine, CA, USA) using two types of polylactic acid (PLA):  Premium PLA filament (Raise3D, Irvine, CA, USA) and Proto-Pasta electrically conductive PLA filament (Protoplant, Vancouver, WA, USA), which contains around 21% of carbon black particles and has measured conductivity of 30 /cm along the X Y layer and 115 /cm against Z layer [54].However, the filament's performance may exhibit variability attributed to environmental factors such as temperature or humidity, as well as being influenced by its geometry [55].Moreover, over an extended period, it gradually decreases its electrical properties, adding a layer of complexity to its long-term reliability and behavior [56].Nevertheless, due to their biocompatibility, the PLA materials are highly suitable for applications involving contact with the skin [57].The software used is ideaMaker (Raise3D, Irvine, CA, USA).Since both materials are PLA, there is a strong bonding between them and the printing process is similar.Both materials are fabricated using the same printing parameters besides the nozzle temperature, which is 205 All the parts of the test stand are printed with the 3-D-printer F170 (Stratasys, Rehovot, Israel) using their acrylonitrile styrene acrylate (ASA) material and QSR SR-35 soluble support material.The standard software (GrabCAD) of this printer is a closed system; thus, the standard printing settings are used.The layer height is 0.254 mm with a nearly full zig-zag infill called "Sparse-high density." Finding the operating temperature range of the encoder is a common procedure in sensors development.The encoders are mounted together with the basic test stand to the specifically designed platform (temperature test stand) powered by a stepper motor 17HD4001-53N (Moons Industries, Shanghai, China) and tested under different environmental temperatures.The temperature test stand is also printed with the F170 printer.

C. Sensorized Hand Orthosis
Typically, dynamic hand orthoses are used to assist in extending the fingers while restricting or controlling movement of the wrist.In the SaeboFlex glove [Fig.3(a)], an outrigger connects the fingers to a spring.This device utilizes the spring's tension to facilitate the opening of the hand in cases of spasticity.Our sHO employs the same concept with addition of developed encoders' system for detecting position of the fingers.This system can be utilized for monitoring the progress of recovery or improving rehabilitation by integrating gamification elements to boost user motivation and aid in the recovery process [59], [60], [61].
Based on the results from the tests performed during encoder evaluation, a set of S 3 with SH 1 is chosen for implementation in the demonstrator due to its ability to detect all of the stripes during the tests.In the proposed design [Fig.3(b)], the fingers are connected by a nylon line with a slider, which is pulled by the rubber band from the other side.The slider and rail are lubricated for a smoother movement.When the fingers flex, the nylon line pulls the slider over the encoder and this linear distance change can be translated to the rotation of the finger's joints and, thus, their position.In the fingers' extension, the rubber band pulls the slider back [Fig.3(c)].
For the final realization, all electrical traces and resistors are fully printed together with the scales [Fig.3(b)].The microcontroller is directly plugged into the device, and thus, only one wire is used to connect microcontroller's 5-V output with the SH.Moreover, the scale length is calculated based on the proposed device's geometrical dependence of the linear displacement due to flexion of fingers so that it can cover Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.the full movement of the fingers of the tested subject.The calculated scale length is 57 mm.
In order to achieve a further increase of resolution, the encoder system is used, which consists of three shifted, by 0.4 mm with respect to each other, encoders; thus, also the SH has one electrically conductive tip for each scale.This solution not only has the potential to increase the resolution from res S3 = 1.2 to res sHO = 0.4 mm but additionally offers the possibility of direction detection.At this point, it is worth adding that direction detection can be achieved also with only one sensor if every third conductive stripe would be connected together.
In the proposed sHO, accuracy varies due to position of the fingers, with the highest accuracy in fingers' extension.This behavior is caused by translation of the rotational movement to a linear scale.This can be calculated from geometrical dependences [Fig.4(a)] where where x 1 is an initial length of the string between the outrigger and fingers' connector in fingers' extension and x 2 in fingers' flexion.Furthermore, a and b are distances from the joint to the tip of the outrigger, c is a distance from a center of the central line of the finger to the fingers' connector, r is the distance of the fingers' connector to the finger joint, and α is the fingers' angle, with α = 0 • in full extension.Finally, a linear distance travel by SH caused by finger flexion can be calculated with the following equation d = Next, the sensitivity caused by inaccurate positioning of the sHO on the hand is evaluated [Fig.4(b)].In this design, the most critical is proper placement of the outrigger on the hand since it can lead to a high error in obtained results.
The impact of the resolution on the accuracy of the fingers' position detection is significant, e.g., res sHO = 0.4 mm gives uncertainty from 0.8 • to 1.4 • , while in the encoder proposed by Herrojo et al. [47], the error would be 6.4 • -13.3 • (Fig. 5).
The sHO is printed using PLA and conductive PLA with the same printing settings as in Section II-B.Next, it is assembled that Velcro straps are added, and the wire and microcontroller plugged-in.Finally, the nylon line is tied with the orthosis mounted on the hand for a proper fitting.
The evaluation of the performance of the printed encoders is based on a reference device, which is fabricated using a Stratasys printer, in the same manner as the other test stands, and is used together with a rotary potentiometer PTV09 (Bourns, Riverside, CA, USA) to directly measure the angle of the finger's joints while moving with a resolution of 0.28 • .

III. EXPERIMENTS SETUP
First, all possible combinations of sets of SHs and scales are evaluated.Next, after finding a set with the highest detection rate (DR), i.e., referring to the number of stripes detected by the encoder, the fatigue and temperature performance are determined.Finally, the use case of the sensors in an exemplary sHO is evaluated.
In all experiments, the data are acquired with a sampling rate of 115 200 bits/s and postprocessed in MATLAB (The MathWorks, Natick, MA, USA).The data consist of highvalue spikes, whenever there is an electrical contact, and zero or low-value segments, when there is no electrical contact between the SH and the scale.

A. Encoder Evaluation-Sets Performance
The test stand is fixed to the table and connected with the robotic arm IRB 120 M2004 (ABB, Zürich, Switzerland)

TABLE II SCALE TYPES TESTS ORDER
using a 3-D-printed robotic arm connection.The SHs are mounted in the following order: SH 1 , SH 12 , and SH 2 .Three specimens (scale copy) of each of the four scale types are printed to evaluate the variation between prints.The specimens are referred to by adding A, B, or C to the scale's name.The scales are tested in a random order one at a time.The order is shown in Table II.Both parts are connected to the prototyping and microcontroller board Joy-it Uno R3 in a manner as it is described in Section II.The used resistance is R vd = 820 k .The microcontroller is connected to the computer and all data are recorded using the CoolTerm freeware.All tests are conducted at standardized conditions (21 • C, 50% RH).
During tests, the robotic arm slides the SH by 45 mm over the scale in a loop containing five routines.Each routine is executed at different robot movement speeds (5, 20, 40, 80, and 150 mm/s).In each routine, data of 40 cycles are collected, each representing movements of the SH over a full length of the scale resulting in high and low signal values.Between each movement, there is a pause of 1 s for better data visualization.In this test, the voltage values of the sensor are captured.An example of acquired data is shown in (Fig. 6).

B. Encoder Evaluation-Wear
Since SH 1 with S 3 scale is found as the combination that has a full DR and the lowest F, the rest of the experiments are conducted with this set.This set has res S3 = 1.2 mm and F SH1 = 0.07 N.
The setup for fatigue tests contained three specimens of SH 1 in combination with the scale S 3C [Fig.7(a)].The scale is not replaced since it did not show any signs of wear during the tests.The specimens are tested one at a time with a speed of 150 mm/s in a loop of 20 000 cycles.In order to reduce the amount of data stored, the data record starts with the first movement and lasts for 15 s.After recording, there is a 5-min pause, and then, the process is repeated.The record captures around 13 full movements, while the pause lasts approximately 270 movements.The test is conducted at a standardized condition.

C. Encoder Evaluation-Temperature Performance
The encoder is mounted on the temperature test stand [Fig.7(b)] and placed inside a climatic cabinet VCL 4003 (Vötsch Industrietechnik, Balingen, Germany).The motor moves the SH 40 times under different temperature conditions at a speed of approximately 33 mm/s.The tests are carried out on three specimens of both, SHs and scales, starting at 40 • C and decreasing by 20 • C until reaching −20 • C. Due to the fact that the glass transition temperature of PLA is usually between 50 • C and 80 • C [62], it is not recommended to use these materials at such elevated temperatures.It is also not expected to reach such temperatures in the prospected application of wearable assistive devices, and thus, the tests are conducted in temperatures below these values.After reaching each of the temperatures, the specimen is kept in the climatic cabinet for additional 5 min before being tested.

D. Sensorized Hand Orthosis
The sHO together with the reference device is mounted to the hand of a test person and adjusted to match the finger joint axis.The reference potentiometer is read by the same microcontroller in order to be subject to the same sampling frequency.Next, the person twice flexes and twice extends all the fingers, while data are recorded.After the data are collected, it is smoothed and the distinct spikes are isolated.Finally, based on the geometrical dependence between linear translation and flexion of fingers, the fingers' tilt is calculated and compared to the reference tilt from the potentiometer.

A. Encoder Evaluation-Sets Performance
Before conducting the tests, the resistance of each scale and both SHs are measured with a digital multimeter (Table III).In addition, the current drawn by an encoder is determined by connecting a multimeter in series.On average, the encoder's current reads 4.6 µA when there is an electrical connection between a conductive tip and a scale.This information is used to calculate power using Ohm's law (P = V • I ), resulting in a Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.power value of 23 µW.The power value of the whole system is 237 mW, with the microcontroller being the main source of power consumption.
In total, four types of scales are tested.Each type has three copies to understand the influence of 3-D-printing on it.The tests are conducted using three SH types.
The DR is influenced by the following factors: the scale type, the scale copy, the speed, and the SH type.The results of each scale copy and SH type in dependence of the speed are shown in Fig. 8.The scales S 3 and S 4 , which are designed with the widest conductive stripes, detect all the stripes.The mean DR of S 2 is 55% and it is lower than the mean DR of S 1 , which is 86%.Each copy of S 1 and S 2 has different DRs.In S 1 , DR varies from 65% up to 100%.For S 2 , the variations of DR range from 4.5% up to 93%.DR is lower for the copies with higher resistance; however, with this sample size, the pattern cannot be confirmed as a correlation.Moreover, most of the specimens detect less stripes with increasement of speed and force.
The voltage values of spikes for different scales (V ), SH types, and speeds are presented in Fig. 9.There is no significant influence of the robot movement speed on these values.The mean voltage values of spikes for subsequent scale types are as follows: V S1 = 2.66 ± 0.49 V, V S2 = 2.01 ± 0.69 V, V S3 = 3.83 ± 0.20 V, and V S4 = 3.70 ± 0.24 V.The specimens with higher DR have also higher output voltage values.In all the specimens, the signal is clear with neglectable noise level with a maximal value below 0.05 V.

B. Encoder Evaluation-Wear
Fig. 10 presents data from a wear test.The last dataset, which detects all the stripes, is used as a fatigue point.One of the tests (specimen 3) was stopped before reaching 20 000 cycles.The number of cycles before wear is the following: 12 690, 9180, and 9720, which gives the mean value of 10 530 cycles.In all the specimens, there is observed wear of the conductive tip of the SH by 0.18, 0.31, and 0.07 mm, and no wear of the scale.In (Fig. 10), it is visible that in the first phase of operating, the voltage increases and then gradually decreases until reaching the fatigue point, after which there is a higher drop.

C. Encoder Evaluation-Temperature Performance
Fig. 11 shows an exemplary data from the temperature tests.In the temperature range from 40 • C to −20 • C, all tested sensors have 100% of DR.There is no significant difference between temperature and the output's voltage values.The following are the mean values for descending temperatures: 3.14 ± 0.36 V, 2.87 ± 0.35 V, 2.89 ± 0.56 V, and 2.92 ± 0.49 V.The output mean voltage of each specimen is: 3.00 ± 0.36 V, 3.43 ± 0.29 V, and 2.45 ± 0.37 V.In all tested Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.specimens, there is a noise level up to 0.5 V, with a mean value of 0.06 V.The occurrence of noise results from the utilization of a test stand equipped with a stepper motor that generates vibrations; however, this does not hinder the clear distinguishability of the signal.
At this point, an additional assessment of water resistance is performed by introducing a water droplet using a syringe onto a scale.This procedure results in a continuous flow of electricity through the water, resulting in a sustained and constant high sensor reading, which, in turn, precludes the sensor's ability to detect changes in position.

D. Sensorized Hand Orthosis
The data are collected from four movements of the fingers, two flexions and two extensions (Fig. 12).For all encoders, it is clearly distinguishable based on the occurring spikes and low noise level when there is a movement, whereas the encoders are in a static state when there is no movement.Spikes from different encoders occur alternately as expected.The average error in the angle between the encoder-based system and the reference potentiometer is 1.6 ± 1.2 • .In addition, the current is measured, and on average, the encoders' system current reads 7.4 µA, resulting in a power value of 37 µW.

V. DISCUSSION
The manufacturing of the encoder via FFF is a simple process, requiring no postprocessing except for connecting cables and pins.The sensor can be implemented to a device on a design level and fit to cover a specific range of motion, as presented in the sHO, or it can be manufactured as a separate part for further implementation, as seen in the encoder evaluation.There is a growing interest in material extrusion-based technologies for electronics due to low production costs of sensors [38].In particular, the material cost of the presented encoder is in the range of 0.06 e and is on the same cost level as comparable 3-D-printed linear encoders available in the presented literature.Moreover, the suggested approach does not require the material to possess a high degree of electrical conductivity, ensuring energy efficiency.It also does not necessitate the material to maintain consistent electrical properties, which can be challenging to attain in the 3-D-printing process.Instead, it only requires the material to establish a continuous electrical connection between traces.These features are achievable because the sensor primarily functions based on an on/off principle.

A. Resolution
The highest achieved resolution of the proposed encoder is res S3 = 1.2 mm.In comparison, the authors in the previously mentioned work [46] achieved much lower resolution of 3.4 mm.Moreover, the results from the scale S 1B showed that it is possible to have a full DR with res S1 = 0.8 mm;  however, it requires further investigation to provide devices with increased robustness.The resolution can be increased by using a smaller nozzle size, thus printing smaller and SH; however, this plead to faster wear.
The system in the sHO is designed to have a resolution of res sHO = 0.4 mm, as a consequence of using three encoders simultaneously.The tests prove that this system works and is able to detect the movement direction.The encoders' system accuracy error of 1.6 • is twice the calculated error (0.8 • ) for this system (Fig. 5).The first factor affecting the accuracy error is the positioning of the sHO on the hand, which could not be achieved exactly as designed, leading to different geometrical dependences and resulting in a deviating linear change for the angular change.It is recommended to calibrate system after each placing on the hand.The second factor affecting measurement is quality of the surface between the slider and the rail.The roughness can lead to a jerky movement.
Because of the encoder's incremental working principle, any errors resulting from missed steps accumulate and lead to a cumulative positional error.This issue can be resolved by incorporating a single conductive stripe as a reference point, effectively resetting this error and thereby improving long-term performance.

B. Performance
The encoder requires very low power, measured in microwatts, whereas the microcontroller's power consumption is significantly greater, measured in milliwatts.This indicates a potential for creation of a lightweight and low-power device.
An influence of different speeds, variation between the prints and contact force on DR in S 1 and S 2 , is observed.On the one hand, the speed can decrease the DR of spikes since the sensing head-SH-tip makes a contact with a conductive stripe for a shorter time duration.However, this can be alleviated by increasing sampling speed of microcontroller.It is assumed that differences in the DR are mainly caused by a quality of print, more precisely by the microstructure of the scale's surface.Variations in material flow and the position of the nozzles with respect to each other, the printed lines of PLA, and conductive PLA may overlap, creating uneven or inaccessible stripes, which then causes jumping of the tip or blocking contact between tip and conductive stripes (Fig. 13).This, in turn, leads to missing readout points.Moreover, the force generated by the SH, when combined with low-quality scales, results in jerky movements and premature wear.In the case of S 3 and S 4 , the influence of print quality on DR is eliminated since they have wider conductive stripes.It is proved that these scale types are robust against variations of speed, contact force, and quality.This renders them reliable for usage within any device.This robustness comes at the cost of lower resolution in comparison to S 1 .
The robustness against speed variation is an important factor since the hand movements can achieve high velocities.Nevertheless, considering ADL of healthy person, there are almost no situations when the hand joints exceed speeds of 300 • /s [63], which yields a speed of around 150 mm/s for the proposed device.
The sensors perform consistently between −20 • C and 40 • C, with no temperature-related impact on output voltage.Noise observed is likely due to test setup vibrations but does not significantly affect signal clarity.As a result, these sensors appear promising for use in everyday activities, given that typical daily temperatures fall within this range.

C. Endurance
The fatigue test results in over 10 000 cycles, which suffices for over 3000 grasps and releases.The test consists of a nonstop movement with a speed of 150 mm/s and can gener-Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
ate higher temperatures and consequently faster wear.These conditions are not expected when the sensor is used within a hand device.Moreover, using more robust materials for the electrical tip, the encoder lifetime can be increased.Finally, in the case of wear, only SH needs replacement since the scales show no signs of wear.
Furthermore, in all the fatigue tests, the voltage rises in the first phase of the test and drops in the last one.This can be caused by wear of the SH tip, which after the printing process exhibits minor geometrical imperfections and, when it wears, the contact area with the scale is increased.However, with further wear, the contact force decreases, leading to a worse electrical connection.
The sensor's performance is impaired by the presence of water, which raises concerns about its suitability in humid conditions.In real-world applications, protecting the sensor from moisture is crucial, as even a small water droplet can disrupt its operation.

D. Remarks
In this study, a functional linear encoder is developed using FFF.The dual-state operational principle of the sensor overcomes the limitations associated with low conductivity, making it compatible with both highly resistive and highly conductive materials.The sensor's function is not compromised by changing electrical properties of the material, such as during aging or wear.
While 3-D-printed parts typically necessitate postprocessing for improved surface quality, the presented sensor demonstrates satisfactory robustness, thanks to the low forces it encounters.The primary issue resides in the slider mechanism, leading to uneven movement.For future improvement, alternative systems, such as compliant mechanisms better suited for this printing technology, should be considered.
Accurate positioning of a sensorized orthosis on the hand is essential because it greatly affects the recorded position.The incorporation of soft materials in this application can substantially improve this precision by conforming the sHO shape to the user's hand, simultaneously enhancing comfort.

VI. CONCLUSION
This work presents a concept for an easy-to-fabricate linear encoder manufactured via FFF.The encoder is used for angular position sensing and consists of two parts: an SH and a scale.This work proves that the sensor can be printed via FFF together with the wearable assistive device without necessity of complex postprocessing.
The effect of four types of scales, different speeds, variation between prints, and contact force of the SHs on the performance of the sensor is investigated.The performances of two types of scales are not influenced by the listed factors.In these tests, the highest resolution achieved by the printed encoders is res S3 = 1.2 mm.In addition, the fatigue resistance and the temperature performance of the devices are investigated.The average lifetime of the sensor is around 10 000 cycles, and its operating temperature ranges from −20 • C to 40 • C.
Finally, an sHO with three implemented encoders and almost fully printed electrical circuit is proposed, printed, and tested.The usage of this system results in an increase of the resolution to res sHO = 0.4 mm and offers the possibility to detect the direction of movement.The proposed sensor can be simply and cost-effectively manufactured using the FFF technology and integrated into wearable assistive devices on the level of designing or assembled afterward.This gives a possibility to apply it in patient progress monitoring or to actuators control.

Fig. 1 .
Fig. 1.Proposed linear encoder: (a) mounted in the basic test stand, (b) position calculation based on the distance between the stripes (res) and the NS, (c) connection to the microcontroller, and (d) main factors affecting the signal quality.

Fig. 4 .
Fig. 4. (a) Variables needed for calculation: initial length of the string between the outrigger and fingers' connector-in extension (x 1 ) and flexion (x 2 ), angle of fingers' bending (α), mounting distance of the fingers' connector from the joint (r), and distances from the finger joint to a tip of the outrigger-(a) vertical and (b) horizontal, distance from fingers to fingers connector c; and (b) sensitivity of a readout due to position misalignment.

Fig. 5 .
Fig. 5. Computed influence of the encoders resolution (res) and the fingers' position on the resolution (res sHO ) of the proposed sHO based on translation of rotational movement into linear displacement.

Fig. 6 .
Fig.6.Example of acquired data from testing S 3C with SH 1 with different speeds.Low signal value represents the contact between a tip and insulating stripes, while spikes (marked with red) occur when there is an electrical contact with conductive stripes.

Fig. 7 .
Fig. 7. Test stands.(a) Basic test stand for different sets and fatigue tests.(b) Temperature test stand.

Fig. 8 .Fig. 9 .
Fig. 8. Influence of different speeds and scales on the DR of the encoder in combination with (a) SH 1 , (b) SH 12 , and (c) SH 2 .

Fig. 10 .
Fig. 10.Fatigue test with marked fatigue points.(a) DR and (b) average voltage values of spikes and standard deviation.

Fig. 11 .
Fig. 11.Temperature test-acquired data at −20 • C with occurring noise due to stepper motor vibration.

Fig. 12 .
Fig. 12. Data from testing sHO with reference device.(a) Two flexions and two extensions.(b) Zoomed-in view on data from flexion.(c) Smoothing data and spikes detection.(d) Translation of spikes to fingers' position with usage of three encoders as a whole encoders' system.

Fig. 13 .
Fig. 13.Macrostructure of a scale S 3 with visible uneven top surface.

TABLE I MAIN
VARIABLES FOR ALL SCALE PATTERNS: SCALE TYPE, WIDTH OF CONDUCTIVE STRIPES, WIDTH OF INSULATING STRIPES, NUMBER OF STRIPES, AND POSITION RESOLUTION (RES) • C for PLA and 210 • C for conductive PLA.The extrusion rate is 95% and the heated bed temperature is 60 • C. The parts are printed with an extrusion speed of 50 mm/s and 100% fan speed.Based on the nozzle size, which is 0.4 mm, the linewidth is set to 0.4 mm and the layer height is set to 0.2 mm.The prints have shell numbers equal to two and grid infill set to 100%.The retraction distance is set to 0.3 mm with a speed of 40 mm/s.The multimaterial prints use wipe tower.The only postprocessing activity for the [58]ders is connecting wires to the input and output pins.The overall dimensional accuracy of the encoder depends on the specific 3-D-printer employed and, in this instance, it results in an accuracy of approximately 0.1 mm[58].

TABLE III SCALES
RESISTANCES OF THE NEAREST (R Sb ) AND THE FURTHEST (R Se ) STRIPE Both SHs have similar resistance of around R SH = 14 k .