Towards a 3D Printed Strain Sensor Employing Additive Manufacturing Technology for the Marine Industry

Abstract

This study focuses on the successful fabrication of a cost-effective strain sensor using exclusively additive manufacturing Fused Deposition Modeling (FDM) technology, enabling fast on-site production, which is particularly advantageous in maritime settings, reducing downtime, and supporting a circular economy approach by minimizing inventory needs and environmental footprint. The principle of operation of the developed device is based on the piezoresistive characteristics of a carbon nanotube (CNT)-enriched building material, from which the main sensing element consists. The prototype exhibited reliable piezoresistive properties, and a clear correlation was observed between the thermal treatment of the printed piezoresistor and the resulting gauge factor, linearity, and hysteresis. Its robustness, simple design, and single-step manufacturing process, together with its ability to be integrated into the readout circuitry through standard soldering, enhance its reliability and durability. The key advantages of the proposed device include its low cost, simple design, and rapid remote production.

1. Introduction

Additive manufacturing, also referred to as three-dimensional (3D) printing, is based on melting, deposition, and solidification that leads to the production of a component with high dimensional accuracy and smooth surface finish [1], the degree of which is determined by the type of application. It is a rapidly evolving manufacturing process with broad application in a wide range of industrial and academic applications. The fields in which it is employed are medicine surgical planning [2], art, military, aerospace, food technology, and many other sectors [3]. Until now, additive manufacturing has mostly been used for health and medical purposes, but it is expected to expand in many other manufacturing sectors and alter production lines in general. The reasons for this are the continuously evolving range of materials with different properties that can be used in a way that serves better for each application, as well as the fact that employing 3D printing for on-site manufacturing of objects with complicated geometry, with the employment of available Commercial-Off-The-Shelf (COTS) equipment, is now feasible. The above of course give great flexibility to designers and engineers, as they can potentially alter their initial design on site, adapting better to the needs of each project, shortly after each trial, with low production cost. Furthermore, the combination of 3D scanning and 3D modelling can lead to the fast manufacturing of defective replacement parts, thereby prolonging the service life of mechanical equipment. Considering the above advantages and their contribution to longer service life, minimizing transport and production costs, and the environmental footprint, it is clear that additive manufacturing can be a part of the circular economy and a boost to the economic system [4,5].

The marine industry is an area of engineering where there is a delay in integrating additive manufacturing in comparison to other sectors (i.e., aerospace, automotive, and medical domains). According to Pasi Pukko, from the VTT Technical Research Centre of Finland [6], “each of these sectors has somewhat specific drivers and reasons to use this technology. For aerospace, it is lighter structures and fly-to-buy ratio; for automotive, fast prototyping and tooling; for medicine, patient specific instruments, guides, visual models i.e., personalization”. As the marine industry cannot clearly benefit from the above “drivers” (i.e., lighter components, patient-specific instruments, etc.), new relevant drivers should be explored. These can include better performing components, part consolidation, and digitalized spare parts [7]. Wartsila, a leading global company in this sector, has already employed additive manufacturing in related manufacturing procedures, that is, in the fabrication of special tools and structural and working component segments of engines [6]. It should be noted that printing working engine components was made practically feasible because of the significant reduction in the cost of 3D metal printing, as there were concerns initially regarding the durability/mechanical strength of plastic components (according to Andreas Hjort, Wartsila General Manager Smart Design). Nevertheless, Wartsila produced the world’s first 3D printed CE-certified lifting tool, capable of lifting a 240 kg engine piston with a maximum capacity of 960 kg, saving up to 100,000 euros in tooling costs. According to Giuseppe Sarago, Wartsila Hub’s Director of Manufacturing Excellence [8], the next step is from “an engine perspective to support new fuels, to embrace the freedom to design new geometries that can create products that normally cannot be achieved with castings or foundry solutions”, which contributes to the circular economy and sustainability. Finally, another drive regarding 3D printed parts in the shipbuilding industry lies in the continuously advancing era of artificial intelligence and condition-based maintenance, where 3D printed spare parts that have sensing capabilities (that is, 3D printed custom-made sensors) can play a key role, offering significant advantages compared to their traditional microelectronic counterparts, such as significantly reduced cost, remote and on-demand fabrication, simplicity, robustness, and customization [9].

Sensing devices play a vital role in ship operation and performance monitoring and are used to optimize the operation of specific subsystems to ensure acceptable performance, as well as for safety reasons, as in the case of gas leakage detectors on gas carriers, oil mist detectors, etc. Therefore, sensors installed in ships measure various quantities such as temperature, humidity, pressure, flow rate, exhaust gas purity, and shearing strain of the tail shaft as part of the power output calculation to achieve compliance with energy efficiency index regulations [10]; additionally, sensors can be utilized to calculate the displacement of the hull and the fatigue effect on the steel plates and brackets [11].

With regard to piezoresistive measuring devices, small-scale sensors have been proposed and employed in various applications over the years in marine and underwater environments where accurate pressure measurement is crucial [12]. These sensors are vital in tsunami detection [13], wave and tide gauges, platform levelling, and in-depth observation data from ROVs, profiling instruments, and towed arrays [14]. Additionally, piezoelectric sensors have been utilized to measure ocean wave heights and periods, as well as to detect underwater objects [15].

Furthermore, based on the work by Crawley and de Luis [16], who pioneered the development of the induced strain actuation mechanism for hull structure observation, the concept of vibration control for smart hull structures using optimally placed piezoelectric composite actuators was proposed [17]. Recent studies have suggested further advancements; for example, a piezoresistive pressure micro sensor, fabricated using Silicon On Insulator (SOI)-based micromachining technology, has been proposed as a novel solution for micro pressure sensors suited for harsh marine environments [18]. Additionally, a system of three triaxial propagation accelerometers has been employed to monitor crack propagation in wind turbines [19]. Lastly, piezoresistive sensors have potential applications as limit-switches [20], which are widely used in marine settings, particularly on cranes and davits where sensor hysteresis is a desirable property [21].

Over the past years, the size of all the above-mentioned sensing devices has decreased significantly because of the development of “microsensors” a product of the evolution of the semiconductor industry. However, this does come with some drawbacks. First, in the case of failure, on-site fabrication is not possible because of the complex structure and manufacturing procedure, which require high-end equipment and methods. Maintaining an adequate onboard quantity of spare sensors is very expensive; thus, only sensors required as “critical’’ spares are stored onboard (Section 10.3 ISM Code [22]). The type and quantity of critical spare parts to be kept onboard is imposed by classification societies, manufacturing companies, oil major forums (for oil and gas carriers), and ship-management companies, taking into account safety and acquired experience. However, in most cases, only metal parts and internal combustion engine hardware (bearings, liners, and cylinder covers) are included in the inventory, as parts directly connected with propulsion and electrical power generation that are considered as more critical [22,23]. In case of failure, the ship owner or the management company has to face the high cost of the sensor and its handling/packaging (as it is a sensitive material), while on many occasions, there is a waiting period of significant time. Additionally, when a ship is not new (e.g., over 15 years old), a significant percentage of the installed sensors is already obsolete, and many of them are out of stock [24].

The innovative approach of this research is focused on the design, manufacturing, and characterization of a new, state-of-the-art 3D printed strain sensor intended for use in the shipbuilding industry. The manufacturing process is simple, requiring only one 3D printer onboard. Building materials (filaments) are commercially available, while the sensing elements can be directly connected to the macroworld, overcoming the necessity of wire bonding. Note that employing available Commercial-Off-The-Shelf building materials and printer has a significant impact of the required manufacturing cost which is approximately equal to USD 0.21 per sensor (considering the cost of the building materials) which is notably lower than that of a typical commercial resistive strain gauge sensor. The above factors contribute significantly to the high availability (on-demand remote printing), ease of replacement in the case of defects, simplicity, and reliability. In addition, since the developed sensing element alters its initial geometry (passively, following the strain of the component on which it is applied), which in turn has a significant impact on its physical properties, it can be considered a device based on 4D printing technology [25]. More specifically, the printed strain sensor’s resistance changes due to mechanical deformation, which introduces a dynamic element to its behavior beyond its static 3D printed form. We should note that the proposed measuring device can potentially be employed in real-time health monitoring systems on vessels with composite hulls by integrating it into the hull during fabrication, opening the way for numerous potential applications.

With the prospect of additive manufacturing expanding in spare parts and machinery hardware in the near future (as in the reported case of Wartsila [6,7,8]), these benefits can become significantly important and have a substantial effect on the operational costs and performance of vessels. Furthermore, an additional advantage that should be highlighted when employing additive manufacturing for marine spare parts is waste minimization, as the most common manufacturing procedure, Computer Numerical Control (CNC) machining, involves waste production of 40–70% while additive manufacturing utilizing layer-by-layer material deposition can lead to a reduction in waste production by 95%, vastly affecting production cost and sustainability.

In the following sections of the paper, a description of the developed sensing device along with the experimental setup employed will be presented. The measured characteristics of the sensor will be detailed, including the investigation of a necessary step of thermal treatment necessary to enhance sensor functionality. The obtained results will be presented and discussed, drawing the necessary conclusions regarding the specific research work.

2. Materials and Methods

2.1. Sensor Design and Fabrication

The underlying idea behind the design of the sensor was to develop a device that could be affixed to the surface of a structural component subjected to stress; thus, its design incorporated a non-conductive flat and flexible surface with an appropriate sensing piezoelectric element on top, as shown in Figure 1a. The non-conductive layer was made of Acrylonitrile Styrene Acrylate (ASA, BASF B.V.) while the conductive layer, which served as the sensing element, was made of Fiber Force Nylforce Carbon Nanotubes (CNT) Conductive, a Polylactic Acid (PLA) biopolymer compound enriched with carbon nanotubes. The piezoresistive element geometry was chosen to be as simple as possible because the goal of this initial investigation was to validate the feasibility of fabricating 3D printed strain sensors employing commercially available polymer-based building materials and their performance characteristics. The device was fabricated by Fused Deposition Modeling (FDM) additive manufacturing technology using a commercial desktop 3D printer, Ultimaker S3 [26]. First, the non-conductive layer was printed as a substrate, followed by the printing of the sensing element on top, which incorporated the appropriate electrical contact pads. Specifically, a 80 mm × 20 mm (L × W) layer with a thickness of 0.6 mm was printed as a substrate using ASA. On top of this layer, the sensing element with the geometry presented in Figure 1a was built employing CNT-enriched PLA. The sensing element comprises a printed resistor with dimensions of 50 mm × 4 mm (L × W) and two rectangular electrical pads situated at both ends, each with an area of 225 mm². The thickness of the sensing element, including the contact pads, is 0.2 mm. Figure 1a illustrates the geometry of the device, while Figure 1b presents two printed devices. It is significant to notice that the necessary connection with the readout circuit can be easily achieved through directly connecting suitably sized conductors using standard soldering. This is well demonstrated in reference [9] where the same building material was employed as contact pads

2.2. Experimental Setup for Sensor Characterization

The experimental setup employed for the sensor characterization is presented in Figure 2. It primarily includes an appropriate source meter and a mini-CNC device. Specifically, a commercial CNC Engraver (Woodpecker CNC 3018,TOOLOTS, CA, USA) was modified to provide a suitable platform for sensor support and strain application. This was achieved by replacing the engraving spindle with a bracket of appropriate geometry and attaching a second one fixed to the frame of the engraver, both 3D printed; the specific modification is presented in Figure 2. The sensor was placed between the two supports; during characterization, consecutive steps of displacement with an accuracy of 0.1 mm were applied to the movable bracket, by appropriately programming the engraver (G-code). Thus, one edge of the sensor was fixed while the other edge was displaced on the x-axis, as shown in Figure 3, causing the device to bend along the z-axis; note that movement is not permitted on the y-axis. The change in the initial resistance of the bent sensing element due to the piezoresistive properties of the carbon nanotubes-enriched filament was recorded by applying a constant current to the sensing element and measuring the corresponding voltage through an appropriate source meter (Kethley 2401, KEITHLEY, Solon, OH, USA).

The investigation of the properties and performance of the developed sensor included the measurement of key characteristics that are considered to be significant [27] and Scanning Electron Microscope (SEM) characterization of appropriate printed samples to investigate the morphology of the structure, homogeneity, and chemical composition of the resulting piezoresistor. In more detail, the characteristics investigated were the following:

• Gauge Factor

Note that the gauge factor was calculated separately for each step of displacement, as well as for the total (maximum) range of displacement applied, employing the equation

$\mathrm{GF}={\displaystyle \frac{\frac{\Delta R}{\mathrm{R}0}}{\epsilon }}$,	where
R0	the initial gauge resistance in non-strain condition;
ΔR = RF – R0	the resistance change after applying strain;
ε = Δl/Lo	the strain applied;
Lo	the initial gauge length in non-strain condition (i.e., 80 mm);
Δl = LF − Lo	the change in gauge length after applying strain (maximum displacement).

b.

Linearity of the sensor

The linearity of the sensor was investigated to estimate the consistency of the obtained GF for a range of displacements. It was evaluated by calculating the standard deviation (σ_ΔR) of the recorded resistance change (ΔR) values that correspond to 0.1 mm singular displacement (which is the minimum that can be applied with the specific experimental setup) since a small value of σ_ΔR suggests a small variability of the deduced GF. Note that σ_ΔR was also calculated when no displacement was applied as an indication of resistance stability, as described in Section 3.2.2.

c.

Hysteresis

The change of the initial resistance (R_ο) was calculated for every cycle of compression and decompression; the average and maximum hysteresis were recorded.

d.

Repeatability

The changes in all of the above properties were recorded after applying a series of successive strain cycles to the same sample.

2.3. Post Processing Thermal Treatment

Relevant research on 3D printed silver nanoparticle-based inks has suggested that, to achieve optimum electrical conductivity of the printed elements, a post-printing sintering step is necessary, which usually requires thermal treatment at temperatures greater than 200 °C to enhance the diffusion and sintering of conductive atoms, creating more electrical paths [28]. Accordingly, an appropriate thermal treatment step was also employed in the manufacturing process of a 3D printed sensing device in which the same conductive material was employed as in the current work [9]. In particular, a significant total decrease of approximately 28% of the initial resistance value of the printed resistor was reported after performing 12 steps of two-hour thermal sintering at a temperature of 100 °C. We should note that ideally, a sintering temperature closer to the melting point of the employed building material (215 °C [29]) is desirable. Nevertheless, the supporting nonconductive material is Acrylonitrile Styrene Acrylate (ASA), with a glass transition temperature of 112 °C [30], limiting the maximum temperature for thermal treatment that can be applied. The highest sintering temperature without deformation of the substrate was experimentally deduced to be equal to 105 °C. Taking into consideration the above parameters, the thermal treatment employed for the printed element was performed at a temperature of 102 °C.

We should note that the mechanism of sintering does not rely solely on the temperature employed because other factors are also dominant, such as the conductive particle size and shape, as well as the printed pattern thickness [28]. Furthermore, although the applied temperature clearly has a significant effect on the degree of sintering, the applied time also affects the resulting outcome [31]. For the above reasons, the effect of thermal sintering, employing a temperature of 102 °C, on the resulting electrical resistance of the conductive 3D printed element, for different time intervals, was investigated.

3. Results

3.1. SEM Characterization and Elemental Chemical Analysis

As already mentioned, the current investigation included a SEM characterization of appropriate printed samples to investigate the morphology of the structure, homogeneity, and chemical composition of the resulting piezoresistor. Therefore, SEM images of 3D printed samples employing Fiber Force Nylforce CNT Conductive filament were recorded on a JEOL JSM-6510 LV-EDAX Scanning Electron Microscope (Oxford Instruments (Abingdon, UK), 10 mm² Silicon Drift Detector-x-act). The dimensions of the samples were 1 cm × 1 cm × 1 cm.

The views of the specimens along the different cross-sectional planes are presented in Figure 4. A fairly uniform picture arising from the carbon particles embedded in the plastic matrix was observed along every cross section (Figure 4a–c). This uniformity was enhanced upon annealing (Figure 4d–f and Figure 5). The morphology of the microstructure was the same and independent of the orientation of the cross-sectional plane. We should note that, as anticipated, no individual CNTs were detected at this observation level (nor at higher magnifications employing appropriately polished samples, reaching up to ×13,000) since the dimensions of individual CNTs are expected to be in the nanoscale. The purpose of SEM characterization, however, was not to observe individual CNTs but to detect (in combination with EDAX analysis results) potential non-uniform dispersion of CNTs (i.e., aggregates of CNTs) in the polymer matrix. From the obtained results, no distinctive particle formations were observed (both from SEM and corresponding EDAX carbon mapping), suggesting a uniform distribution of the CNTs within the polymer matrix. Therefore, a continuous and homogeneous solid material is built.

Elemental chemical analysis of the specimen was also performed through Energy Dispersive X-ray Spectroscopy (EDS)/EDAX. According to the obtained results, the specimen consisted mainly of C (81.4 ± 0.4% wt), O (7.2 ± 0.4% wt), and N (11.3 ± 0.1% wt). Small amounts of Al (approximately 0.1% wt) were also detected which can probably be attributed to impurities present in the specific filament. We should note that the recorded value for C is significantly larger than the corresponding one for a typical PLA filament which is not enriched with CNTs (see Section 4). A cross-sectional image of the specimen recorded by SEM is presented in Figure 6a, together with the EDAX carbon and oxygen elemental analysis (Figure 6b) and element distribution of C and O (Figure 6c,d). The presence of C is attributed to the conductive nanoparticles as well as the polymer matrix, whereas O and N originate from the matrix material. An isotropic behavior of the material is anticipated, based on the uniform distribution of the elements observed in the images.

3.2. Effect of Sintering

3.2.1. Reduction in Resistance of Sensing Element

As mentioned in Section 2.3, an appropriate investigation was performed with regard to the change in the initial resistance of the printed sensing element after thermal treatment. In more detail, a thermal treatment at 102 °C for a total time of 20 h was applied at consecutive time steps of two to four hours. At the end of each step, the resulting resistance was measured at room temperature. The corresponding results are presented in Figure 7; a total reduction of 44% in the resistance was observed (i.e., from 5400 Ω to 3020 Ω), with 28% of the reduction occurring within the initial three hours of the process. This finding indicates that the sintering process enhances the creation of conductive paths, as has been previously reported [9].

3.2.2. Reduction in Sensing Element Resistance Variation

In addition to the substantial reduction in the initial resistance of the printed sample, sintering also affects the resistance stability, an effect of high importance for the sensing element. In more detail, it was noticed that samples subjected to longer heat treatments were more stable during static (strain-free) I-V characterization (characterization performed by recording the developed electrical potential (V) vs. the applied current intensity (I). That is, a significantly decreased variation in their resistance value was recorded when no strain was applied. The deduced standard deviation σ_ΔR for the sample resistance is presented in

Table 1. Standard deviation during strain-free measurements for samples with increasing annealing time.

Sintering Time	No Sintering	11 min	60 min	20 h
Standard Deviation of Resistance change (σΔR)	6.02	3.84	1.69	0.53

; as we can notice, a tenfold decrease on σ_ΔR is recorded after sintering for 20 h. We should also note that the Temperature Coefficient of Resistance (TCR) for a CNT-enriched thermoplastic composite has been reported in the literature to have a value as high as −1.28 × 10⁻²/°C [32]. Therefore, even a small temperature increase due to self-heating can potentially result in a significant variation in its resistance, affecting the sensing element’s operation. With this in mind, the variation in the resistance of the sample was recorded (by calculating σ_ΔR) for different values of constant current applied during static characterization. The recorded standard deviation of resistance is presented in Figure 8. As can be seen, applying a current higher than 3 mA leads to approximately a threefold increase of σ_ΔR; this can be attributed to self-heating that modifies the conductive pathways within the sample combined with the resistance change due to the material’s high TCR. Therefore, to ensure accurate and consistent results, a current of 0.1 mA was utilized in all measurements performed.

3.3. Dynamic Characterization

3.3.1. One-Direction Bending

To investigate the key characteristics of the developed sensor, a series of dynamic measurements followed by applying appropriate steps of compression and decompression of the device were performed as described in Section 2.2. Sensors that underwent a heating treatment of 1 h or 20 h, as well as sensors that did not undergo sintering, were characterized for comparison purposes. Additionally, the piezoresistive layer was subjected to bending with consecutive compression and elongation steps as described in the following sections. The properties outlined in Section 2.2 have been identified for all devices.

The deduced variation in the resistance of the sensing element when applying consecutive steps of compression and decompression is depicted in Figure 9a,c,e. The corresponding increase in the initial resistance (%) with respect to the displacement for both the compression and decompression cycles is presented in Figure 9b,d,f.

It should be noted that after performing extensive testing with varying degrees of displacement, it was determined that the optimal range of displacement beyond which the hysteresis and linearity of the sensor began to decline was 0 to 0.6 mm. The degradation of these characteristics is possibly attributed to the fracture of the conductive paths within the piezoresistive material. The values of the deduced gauge factor and hysteresis (mean value) of the characterized devices are shown in Figure 10. Additionally,

Table 2. Deduced sensor properties for different sintering times.

Sintering Time	Gauge Factor	Hysteresis (Mean Value, Ω)	σΔR Corresponding to a Singular Displacement Step (Ω/0.1 mm)–Linearity
No sintering	0.54	29.14	4.38
1 h	0.10	9.92	4.00
20 h	0.95	8.03	2.12

provides a summary of the specific results along with the corresponding standard deviation for resistance change per step of displacement, which is a characteristic measure indicating the linearity of the sensor, as previously mentioned (see Section 2.2). As we can observe from the specific data, both the gauge factor and linearity of the sensor were significantly improved for samples sintered for 1 and 20 h; this finding is analyzed in the Section 4.

A correlation analysis was performed to examine the relationship between the gauge factor and hysteresis, and sintering time using the Pearson correlation method. The specific method is a widely used approach for numerical variables, assigning a value between −1 and 1, that indicates the strength of the relationship between two variables. A value of 0 indicates no correlation, a value of 1 suggests a strong positive correlation, and a value of −1 suggests a strong negative correlation [33]. The findings presented in

Table 3. Pearson correlation for key sensor characteristics vs. thermal treatment time.

	Gauge Factor	Hysteresis	σΔR Corresponding to a Singular Displacement Step (Ω/0.1 mm)–Linearity
Pearson factor for sintering time	0.94	−0.75	−0.99

indicate a strong positive correlation between heat treatment time and gauge factor (0.94), a negative correlation between heat treatment time and hysteresis values (−0.75), and a strong positive correlation between sintering time and linearity, which aligns with previously reported results.

3.3.2. Consecutive Alternating Bending

To evaluate the ability of the developed sensors to measure strain under different types of bending stresses, cyclic alternating bending was applied. In more detail, the device was subjected to hogging and sagging modes of deformation to investigate its response with respect to the direction of bending. In this way, the conductive layer (on top of the ASA substrate) undergoes either compression (sagging) or tension (hogging) depending on the induced deflection of the sensor (downwards or upward, respectively; see Figure 11). The recorded resistance variation is presented in Figure 12a. It should be noted that thermal treatment was employed (20 h) while for comparison, the corresponding result for a sensor sintered for one hour is also presented (Figure 12b). As illustrated in the specific figure, different sensor responses are recorded for hogging and sagging, which will be discussed in the Section 4.

4. Discussion

From the above results, we can safely assume that thermal sintering plays an important role not only in the obtained resistance of the conductive pattern that is fabricated using FMD 3D technology, as in the cases of inkjet printing or screen printing [28,31], but also in the main characteristics of the resulting sensor, that is, gauge factor, hysteresis, and linearity (σ_ΔR). In more detail, as shown in Figure 9a,b, in the devices that did not undergo sintering, a notable hysteresis is observed since the final value of the recorded resistance is decreased by approximately 0.6% compared to the initial value. Note that the level of hysteresis is significantly reduced (that is, 0.4%) with the application of sintering for even one hour, while a twenty-hour thermal treatment completely eliminated hysteresis. Moreover, as shown in Table 2, sintering clearly affects the resulting resistance change ΔR for a given displacement (therefore, the gauge factor), while its duration notably influences the resulting linearity (σ_R) of the sensor.

In general, the mechanisms by which conductivity is achieved in polymer composites enriched with conductive fillers (such as CNTs) are closely related to the conductive filler loading and can be classified into two main types: tunnelling effect and ohmic conductance [34]. The tunnelling effect primarily occurs in composites with lower conductive filler loading, where most filler particles are covered by a polymer layer during mixing. According to Morteza et al. [35], in CNT-filled polymers, such as the conductive filament employed, the CNTs are entangled and folded in complex networks inside the polymer matrix. When an external strain is applied, the unfolding of entangled elastic nanomaterials is more probable than sliding in their axial directions, changing the tunnelling resistance; as a result, the conductivity of the polymer also changes.

When the concentration of the conductive filler is increased beyond a specific point, referred to as the “Percolation Threshold”, conductive networks are formed within the polymer matrix, providing relatively low-resistance paths for the movement of free electrons. As a result, ohmic conductance is the primary conductive mechanism. According to the results presented in Section 3.1 (SEM and elemental chemical analysis), the conductive layer of the sensor primarily consists of C (81.4 ± 0.4% wt), O (7.2 ± 0.4% wt), and N (11.3 ± 0.1% wt). Taking into consideration that a typical C concentration for pure PLA material is under 60% wt [36], we can safely assume that Ohmic conductance is the main conductive mechanism for the particular filament and not the tunnelling effect since the filler concentration (CNTs) is over the typical percolation threshold; note that a typical percolation threshold of 0.1 wt% might be sufficient for nearly any CNT/polymer system with optimized dispersion of the conductive nanofillers [37]. In addition, according to [34], polymers enriched with CNT usually possess a very low percolation threshold because of their nanosized diameter and large aspect ratio.

The piezoresistor’s conductive filament is made using PLA enriched with CNTs, according to the manufacturer’s specifications [29]. A. Bouamer et al. [38] reported that annealing of PLA at a temperature of 95 C for 4 h resulted in the modification of its microstructure, which changed from an amorphous state to a semi-crystalline state, exhibiting two different crystal morphologies. Furthermore, a noticeable hysteresis occurs in the conductance change when strain is applied to a polymer composite that is enriched with a conductive filler [39,40]; this has been attributed to the viscoelastic nature of the polymer and the interaction between the conductive nanomaterial fillers and polymers. Our findings indicate that thermal treatment significantly affects the measured hysteresis (see Section 3.3), in accordance with the research results mentioned above [39,40], showing that annealing alters the polymer microstructure, which changes from an amorphous state to a semi-crystalline state, affecting its viscoelastic behavior. Based on the above, we believe that this change in the polymer microstructure is the main cause of the observed decrease in σ_ΔR and the corresponding increase in GF.

With regard to consecutive alternating bending, a nonmonotonic strain–sensing cycle was observed; that is, a different sensor response was recorded for the hogging and sagging modes of deformation. In more detail, as we can notice in Figure 12c, the minimum sensor resistance (ΔR = −0.8%) occurs during sagging (i.e., when it undergoes compression) at a displacement of 0.2 mm. Further displacement in the same direction results to an increase in the resistance. In addition, the sensor resistance reaches its maximum value (ΔR = 2.5%) at hogging (i.e., when subjected to tension) at a maximum displacement of 1.2 mm. It has been previously reported [34] that the resulting entangled CNT structure (due to the van der Waals forces between individual CNTs) usually results in a nonmonotonic behavior during the strain cycle, arising from competition between the destruction and construction of the conductive network (percolation pathways); this is often observed for various types of CNT/polymer systems. Furthermore, J.F. Christ et al. [41] also noted a double-peak (nonmonotonic) behavior during the strain cycle, which can be attributed to the restructuring of percolation pathways owing to competition between the formation of new pathways and the destruction of old ones; in general, the resistance decreases when the number of conductive nanofiller interconnections increases. According to the authors, the following mechanism applies: the interconnections between rod-like conductive fillers are maximized at a slight degree of alignment and not at a fully random state of orientation or at the maximum orientation state. In addition, increasing the applied strain leads to a corresponding increase in the alignment of the CNTs in the direction of stretching. Thus, it is reasonable to assume that the resistivity initially decreases at slight alignments until the number of interconnections reaches its maximum, beyond which further straining starts to disconnect the nanotubes, and consequently, the resistance increases. Based on the above, the minimum resistance value can appear near the mid-strain levels, in agreement with our findings (as shown in Figure 12c), and corresponds to the maximum number of CNT interconnections at an optimum strain, rather than the minimum or maximum applied strain.

Additionally, with regard to the different maximum variation of resistance (ΔR) for the hogging and sagging modes of deformation, it is important to consider that the 0.2 mm thick piezoresistive layer is situated on top of the 0.6 mm thick non-conductive substrate as presented in Figure 1. This configuration affects the location of the neutral axis during bending of the whole device. Specifically, the different thicknesses of the piezoresistive layer and non-conductive substrate create an asymmetric bending response, resulting in different strain distributions in the conductive layer during the two different modes of testing (hogging and sagging). Consequently, the mechanism involving CNT network restructuring described above is not solely responsible for the observed difference in the corresponding maximum variation of resistance.

Future work can be categorized into two primary directions. Firstly, it includes investigating the use of different building materials with appropriate characteristics (e.g., conductive filaments with different polymer matrices) and exploring their application in the fabrication of sensing devices exclusively with FDM technology. With regard to the developed sensor in this work, appropriate durability testing should be performed in order to fully investigate the long-term usability of the developed sensor. Consequently, the integration of the sensor into a Glass Reinforced Polymer (GRP) composite will be investigated. The goal of this investigation is to employ the developed device for real-time health monitoring in vessels with composite hulls. By successfully embedding the proposed strain sensor into a GRP hull, a sensing network can be established during the vessel’s construction. This network will enable strain measurements at specific locations through a real-time monitoring system, thereby identifying potential fatigue cracking or damage during operation.

5. Conclusions

This study demonstrates the feasibility of fabricating a low-cost strain sensor by employing additive manufacturing technology using a commercially available 3D desktop printer and filaments. The building materials employed were Carbon Nanotube (CNT)-enriched Polylactic Acid (PLA) and Acrylonitrile Styrene Acrylate (ASA). The key advantage of the proposed sensor is its ability to be manufactured onsite aboard a ship without the need for external facilities. This on-demand production capability is particularly beneficial in remote or maritime environments, where access to specialized manufacturing services is limited. By enabling the on-site production of spare parts and sensors in general, downtime is reduced, enhancing operational efficiency, which results in a reduction in the need for extensive inventories and long supply chains, extending the lifecycle of equipment, and decreasing the overall environmental impact.

Preliminary characterization of the prototype demonstrated reliable piezoresistive properties, which are essential for accurate strain measurements. This characterization included the determination of the gauge factor, sensor linearity, hysteresis, and cyclic alternate bending response measurements. The deduced experimental results show a clear correlation between the thermal treatment of the device and the aforementioned characteristics, which can be attributed to previous research; overall, reliable and consistent characteristics of the prototype sensor were deduced, indicating the effectiveness of the sensor in measuring the strain.

The sensor design is both robust and simple, rendering it advantageous for practical applications, whereas the developed one-step manufacturing process, in which the piezoresistive element, substrate, and contact pads are fabricated simultaneously, enhances the reliability and durability of the sensor. The deduced gauge factor for the optimum fabrication parameters investigated (0.95) is comparable to those of typical commercial resistive strain gauge sensors [42]. Moreover, direct communication with the necessary readout circuitry can be easily achieved through standard soldering using the integrated contact pads of the device which can be applied directly to the working structural component. This has immediate benefits in terms of the device’s complexity, mechanical reliability, process time, and cost. In addition, its simple design and manufacturing enhances its durability under harsh conditions often encountered in marine environments.

In summary, the successful development of the proposed 3D printed strain sensor demonstrates a viable and effective alternative for fabricating measuring devices for the marine industry, offering numerous benefits, including low-cost, on-demand, remote rapid production, robustness, and environmental sustainability. Its performance in preliminary characterization suggests that it can reliably measure strain, making it promising for applications, such as structural health monitoring and maintenance. Future work will focus on optimizing the design of the prototype and exploring other 3D printing materials with enhanced physical properties (such as low resistivity and hydrophilicity) that can be employed for the development of novel 3D printed sensing devices.