Development and Comprehensive Evaluation of 3D-Printed Prosthetic Feet: Modeling, Testing and a Pilot Gait Study

Abstract

Background/Objectives: The modern prosthetic foot market is characterized by a pronounced polarization between affordable but low-function devices and high-performance yet costly composite prostheses. The aim of this study was to develop and comprehensively evaluate cost-effective, functional prosthetic feet manufactured by fused deposition modeling (FDM). Methods: An iterative design methodology was employed, combining finite element analysis to optimize the biomechanical response of the device, the incorporation of user-specific requirements and experimental validation. Two TPU 95A-based 3D-printed prosthetic foot designs were designed and developed, and their strength and functional characteristics were assessed numerically under the ISO 22675:2024 normative loading cycle. Bench-top mechanical tests were conducted on the fabricated prototypes. Functional performance was evaluated by a transtibial amputee using an inertial motion capture system to analyze gait kinematics. Results: The results demonstrated that both designs operate predominantly within the elastic range with an adequate safety margin. The pilot feasibility gait assessment indicated feasibility and plausibility within the tested protocol and participant for both prototypes. Conclusions: The developed TPU 95A-based FDM prosthetic feet demonstrated promising structural integrity and functional feasibility, supporting the potential of low-cost additive manufacturing as a viable approach for producing affordable prosthetic feet. Further studies with larger participant cohorts and extended testing are needed to confirm clinical applicability and long-term performance.

1. Introduction

One of the key aims of rehabilitation for individuals with lower-limb amputation is to ensure access to prosthetic feet that are both functional and affordable. At present, markets in many regions of Asia, Africa, and other low- and middle-income settings are characterized by a pronounced polarization of available solutions. On the one hand, low-activity prosthetic feet manufactured using outdated technologies remain widespread, such as designs with a wooden core and polyurethane foam filling (e.g., SACH feet). Despite their low cost, their limited functional and biomechanical performance can substantially reduce users’ quality of life and increase the risk of secondary musculoskeletal disorders [1,2]. On the other hand, high-performance composite prosthetic feet (e.g., carbon fiber and related laminates) that better support natural gait remain inaccessible to a significant proportion of patients due to their high price [3,4]. Therefore, there is an urgent need for solutions that can bridge this gap by combining acceptable cost with high functional performance.

Additive manufacturing represents a promising approach for developing accessible yet functional prosthetic feet [5,6,7]. A key advantage of 3D printing in this context is the ability to precisely control the internal geometry and spatial stiffness distribution of the prosthetic structure [8,9,10]. This enables direct tuning of the biomechanical response and targeted design of prosthetic characteristics for specific gait phases and for individual users or user groups. In addition, 3D printing allows rapid and cost-efficient fabrication of functional prototypes for intermediate-stage testing, thereby accelerating iterative design cycles. Among the various 3D printing methods, fused deposition modeling (FDM) is particularly relevant for lower-limb prosthetic applications. This choice is motivated by two main factors. First, FDM is among the most cost-effective additive manufacturing technologies, which directly aligns with the goal of reducing the final product cost [11,12]. Second, the FDM ecosystem supports a wide range of polymeric materials, including thermoplastic polyurethanes (TPU)—a class of elastomers whose high impact toughness and ability to undergo large elastic deformations make them well-suited for mimicking key functional properties of biological foot tissues [13,14].

A review of recent studies confirms the potential of FDM for this task and illustrates the evolution of materials and modeling approaches. In one of the early works in this field [15], a conventional structural thermoplastic (ABS) was used to fabricate an energy-storing prosthetic foot prototype, demonstrating the feasibility of applying FDM in limb prosthetics. Subsequent research shifted toward purposeful tailoring of the biomechanical properties of printed prostheses. For example, to reproduce the nonlinear stiffness of the metatarsophalangeal joint during push-off, investigators combined FDM printing of carbon-fiber-reinforced nylon with advanced finite element analysis (FEA) and geometric optimization of auxetic cells [10]. This approach achieved characteristics close to those of the natural foot and highlighted the importance of controlling micro-geometry in printed prosthetic structures. In parallel, more compliant materials—particularly TPU—have been successfully applied to achieve the flexibility, strength, and comfort required in transtibial prosthetic designs [16].

A major step toward reducing the performance gap relative to composite commercial devices has been the integration of continuous-fiber-reinforced materials into the FDM process. Studies indicate that using composite cores with outer skins of continuous fiberglass can yield stiffness values within 5% of those of commercial carbon laminates [17]. Static testing of such 3D-printed composite prosthetic feet has confirmed compliance with international standards (ISO 22675:2024 [18]), the ability to sustain loads exceeding 4000 N, and differentiated stiffness profiles required for impact attenuation and energy return [19]. In addition, the broader field of additive prosthetics continues to advance through automated design methods based on 3D scanning [20,21] and topology optimization for weight reduction [10,16].

Overall, the current technological level reported in the literature supports the feasibility of producing functional prosthetic feet using FDM. Key success factors repeatedly emphasized include the use of elastomers (TPU) to enable elastic deformation [16] and rigorous biomechanics-driven design supported by FEA [10,20]. At the same time, additive prosthetics is expanding toward increased automation and optimization. Key problems for FDM elastomer feet include fatigue life under ISO cycling, environmental aging and systematic evaluation of user adaptation over extended acclimation periods.

The present study is positioned at the stage of engineering validation and early functional feasibility and experimentally evaluates the hypothesis that a clinically relevant prosthetic foot can be manufactured via FDM using an elastomeric TPU material. Two different structural strategies—monolithic and modular—were developed and analyzed with respect to strength and biomechanical performance. The study addressed the following objectives:

• To develop and validate a design-and-analysis methodology by comparing FEA predictions with results from mechanical bench testing of printed prototypes on a hydraulic testing machine, thereby assessing the adequacy of the computational model and its assumptions.
• To perform a comparative evaluation of the two design strategies across multiple criteria:

  (a)

  Strength performance via stress–strain state assessment under the ISO 22675:2024 standardized loading cycle;

  (b)

  Laboratory biomechanical metrics, including ankle and metatarsophalangeal flexion angles and heel damping behavior;

  (c)

  Real-world functionality via comparative gait kinematics assessment in a transtibial amputee using the developed prototypes and the user’s standard carbon-fiber prosthetic foot.

• To conduct a practical proof-of-concept evaluation in a high-body-mass user (108 kg), confirming that the TPU-based FDM prosthetic foot can withstand ISO 22675:2024-regulated loads while providing baseline functional walking performance.

The novelty of this work lies in the demonstration of an end-to-end development pipeline: CAD/FEA–FDM manufacturing–bench testing–pilot feasibility biomechanical gait study-implemented using widely available equipment and a common material (TPU 95A). This approach enables not only the assessment of structural integrity but also the linkage of design parameters to functionally relevant gait outcomes, which is critical for transitioning from prototyping to clinically meaningful solutions.

The relevance of the study is driven by the growing demand for prosthetic care alongside the need to reduce cost and manufacturing time without compromising functionality. FDM enables decentralized and/or local fabrication, rapid replication or replacement of components and personalization of stiffness as well as mass–geometry characteristics to match individual user needs and operating conditions.

2. Materials and Methods

2.1. 3D-Printed Prosthetic Feet Models

Two alternative geometric models of a prosthetic foot were developed for a comparative investigation of their strength and biomechanical characteristics. The baseline overall dimensions were adopted from commercial low-profile prosthetic feet with a height of 80 mm; the length of both models was matched to the tested patient’s foot size and set to 270 mm. The external shell, reproducing an anatomical shape and adapted for use with and without footwear, was designed using Blender 3.4 software. The internal load-bearing architecture governing the mechanical response was designed and modified in SolidWorks 2024 software. The designs were refined iteratively based on finite element analyses performed in ANSYS 2023 R1 Mechanical software, followed by evaluation of key performance indicators. The ankle behavior in both designs is realized through elastic deformation of the TPU structure and not through a discrete hinge mechanism. Accordingly, the primary functional degree of freedom is sagittal-plane dorsiflexion/plantarflexion, whereas inversion/eversion and axial rotation are limited and stiffness-coupled.

Model 1 is a monolithic design intended for additive manufacturing with horizontal layer deposition (Figure 1a). Under this printing strategy, the primary compliant elements are oriented predominantly parallel to the print layers, which reduces the likelihood of interlayer delamination under service loads and improves reliability under cyclic loading. However, this layer orientation constrains the feasible spatial arrangement and combinations of compliant features. In the functional region responsible for reproducing ankle-like kinematics and forefoot deformation, a lattice structure was implemented to provide the required compliance while maintaining increased torsional stiffness. Heel cushioning is achieved by a system of horizontally oriented flexible elements that deform mainly in bending with a secondary contribution from shear.

Model 2 is a modular (multi-part) design intended to be manufactured as separate components, with print layers oriented within the plane of the dominant service loading (Figure 1b). This strategy enables purposeful specification of the orientation and geometry of compliant elements with respect to the principal stress–strain directions, which can potentially increase energy efficiency by enhancing elastic energy storage and return during the gait cycle. Structurally, the model incorporates a heel damping element to attenuate impact loads during initial contact and a compliant toe segment to enable controlled rollover. The load-bearing section in the ankle region is implemented as an elastic structural system whose predominantly bending deformations provide a closer reproduction of the trajectory of the ground reaction force application point relative to the ankle center during stance and push-off, thereby making the overall kinematic-dynamic behavior more physiologically consistent.

The 3D-printed prosthetic feet were fabricated using an FDM printer Neptune 4 Max by Elegoo (Shenzhen, China) with the following process parameters: nozzle diameter—0.6 mm, extruder temperature—235 °C, heated bed temperature—60 °C, infill density—100% (solid), layer height—0.3 mm and infill line width—0.9 mm, print speed—100 mm/s, number of perimeters—3, raster angle—45°, and cooling fan activation starting from the second layer. The mass of the fabricated specimens was 929 g for Model 1 and 727 g for Model 2.

Figure 1 also shows photographs of the manufactured samples of both prosthetic foot configurations. To ensure compatibility with standard modular prosthetic components and to enable subsequent integration into a transtibial prosthesis assembly, the prototypes were equipped with a standard pyramid-type alignment and connection adapter (“pyramid” type). This interface allows precise alignment of prosthetic components and ensures reproducible mounting conditions during experimental testing and comparative evaluation of the two designs.

2.2. Prosthetic Foot Material

The primary material selected for fabrication and numerical modeling was thermoplastic polyurethane TPU 95A by Eryone (Shenzhen, China). This material was chosen due to a set of properties critical for prosthetic foot function, including high impact toughness, wear resistance, and the ability to sustain large elastic deformations, which collectively enable the required compliance and potential energy return.

It is well known that parts produced by fused filament fabrication exhibit pronounced mechanical anisotropy as a result of filament orientation and limited interlayer adhesion. Nevertheless, in the present study, a simplified isotropic material model was adopted. This assumption is justified by the fact that the geometry and print strategy of both designs were deliberately configured such that, in the most critically loaded regions, the principal service loads are oriented predominantly along the deposition direction. Under these conditions, the dominant deformation mechanisms are tensile and compressive loading of the extruded filament itself, whereas interlayer failure modes (shear and peel-off along adhesion planes), which typically exhibit reduced strength and stiffness, have a secondary influence on the global structural response. Therefore, the integral strength and deformation behavior in the considered loading regimes is primarily governed by the filament material properties under near-uniaxial loading, making the isotropic approximation acceptable for the intended comparative assessment of strength and compliance.

According to the literature [13,22,23], the tensile strength of TPU 95A exhibits substantial variability and typically falls within the range of approximately 20–40 MPa, while the yield strength (conventional elastic limit) is reported to be not lower than 8.6 MPa. Given that the mechanical properties of FDM/FFF-printed components can vary markedly depending on printing parameters, machine characteristics, and filament formulation/manufacturer, the elastic constants of the material were experimentally identified in this study. Specifically, the Young’s modulus E and Poisson ratio ν were determined on a printed TPU 95A test sample using a triaxial compression setup (Geotek GT 1.3.4).

Specifically, tensile tests were conducted on printed TPU 95A specimens of Types 1, 2, and 3 (Figure 2). Type 1 specimens were fabricated with printing filaments oriented parallel to the tensile axis, whereas Type 2 specimens featured filament orientation at 45° relative to the loading axis (within the printing plane). Type 3 specimens were tested with loading applied perpendicular to the printing layers.

The ultimate tensile strength exhibited dependence on the printing orientation, yielding approximately 24 MPa, 28 MPa, and 10 MPa for Types 1, 2, and 3, respectively. The effective Young’s modulus, E_xy, evaluated from the initial slope of the stress–strain curve, was found to be approximately 54 MPa for both Type 1 and Type 2 specimens, while the modulus in the Z-direction, E_z, was approximately 45 MPa. Although the material stiffness is inherently non-linear, it remains sufficiently close to linear within the strain range of 0 to 0.15. Consequently, for the numerical calculations, a linear elastic model was adopted with a Young’s modulus of E_c = 35 MPa and a Poisson’s ratio of ν = 0.47, the latter estimated as appropriate for a porous elastomeric structure.

2.3. Finite Element Model

Numerical simulations were performed in ANSYS 2023 R1 software. The geometric models of both prosthetic foot configurations were imported into ANSYS Mechanical and analyzed using the finite element method with the Large Deflection option enabled to account for geometric nonlinearities arising from significant deformations of the TPU material. In both cases, the computational domain was discretized with an unstructured three-dimensional mesh based on tetrahedral solid elements (Figure 3). Each model contained approximately 2.5 × 10⁴ volumetric elements, providing a practical balance between computational cost and the accuracy required to resolve the stress–strain state with the available computational resources.

In the area intended for coupling with the alignment adapter, a fully fixed boundary condition was imposed, constraining all translational and rotational degrees of freedom. The interaction between the plantar surface of the prosthesis and a rigid supporting plane was modeled using a frictionless contact formulation (large sliding). This assumption is justified by the fact that, during level walking, tangential (friction-related) forces are typically substantially smaller than the normal ground reaction forces and therefore have a secondary effect on the overall stress distribution within the structure. Moreover, the horizontal load component contributes only marginally to the bending moment generated by the ground reaction force about the ankle region; consequently, its influence on the global mechanical response of the prosthesis within the adopted modeling framework can be considered negligible.

2.4. Target Characteristics of Prostheses

Based on the finite element analysis, a comprehensive assessment of the strength and functional performance of the investigated designs was conducted. Strength was primarily evaluated using the distribution of von Mises equivalent stress. A design was considered to satisfy the strength requirements provided that the maximum predicted stresses did not exceed the allowable level for the employed material, taking into account the adopted safety factor.

The functional performance of the prosthetic feet was assessed by comparing the kinematic response of the structure against target requirements derived from the patient’s clinical needs and from gait analysis data acquired using an inertial motion capture system. The key functional metrics were the flexion angles in the ankle region and in the forefoot, reflecting the ability of the prosthesis to provide a physiologically plausible rollover and push-off during stance.

For quantitative analysis, the coordinates of predefined control points were extracted from the simulation results before and after deformation. Changes in their spatial positions were used to compute the flexion angles corresponding to the ankle and metatarsophalangeal regions, denoted as α and β, respectively, for each loading phase (Figure 4). In addition, to characterize the damping behavior of the heel region, the vertical deformation dz at selected heel points was evaluated. This parameter quantifies the compliance of the structure and its capacity to attenuate impact loads during initial contact at the beginning of the gait cycle.

2.5. Load Cycle Description

As the normative basis for defining the loading regime in the numerical simulations, the load level corresponding to the test subject’s body mass was adopted in accordance with ISO 22675:2024. This load level represents one of the standard baseline conditions defined from an aggregated population of users, including individuals with body mass exceeding 100 kg. Application of this regime enables assessment of the structural integrity of the design under demanding service conditions representative of the combination of high body mass and elevated mechanical exposure, which is essential for demonstrating the practical applicability of the proposed prosthetic foot concept.

The standard specifies a dynamic loading profile intended to reproduce the characteristic stance phase of walking. The input parameters are provided as discrete values of the support-platform inclination angle γ(t) relative to the horizontal plane and the test force F(t) applied along the tibial axis, which is taken as vertical. Figure 5 presents the relationship between the prescribed test force and the platform inclination angle. For time normalization, the stance phase is expressed as the relative gait phase T: T = 0% corresponds to initial heel contact and T = 60% corresponds to toe-off, i.e., the end of the full stance phase.

Within the finite element analysis, all discrete points of the normative loading cycle were considered, except for states with zero load. For each force–inclination pair F(γ) specified in the standard, a separate quasi-static analysis step was defined, approximating the dynamic cycle through a sequence of equilibrium states. At each step, the rigid support surface was rotated by the prescribed angle γ about the corresponding axis, after which a vertically oriented force of magnitude F was applied to the support.

To improve numerical robustness and solution convergence, intermediate sub-steps were automatically introduced using linear interpolation of the loading parameters. Consequently, simulations were performed for both designs over the full normative cycle, which approximately reproduces service conditions for a user with a body mass of about 100 kg. This approach provides a detailed characterization of the structural response throughout the entire stance phase and enables identification of the most critical load cases in terms of both strength and the required compliance. Figure 6 also indicates the specific cycle points selected for subsequent experimental validation during bench testing of the manufactured prototypes.

2.6. Bench Testing and Validation of the Loading Model

2.6.1. Experiment Scheme

To validate the developed finite element model, a series of experimental tests was conducted using a hydraulic testing machine, Instron (Norwood, MA, USA) DX300. The test system comprises a base with a loading frame and guide columns, movable and fixed crossheads, specimen gripping fixtures, a hydraulic actuator driving the movable crosshead with an associated hydraulic power unit, and a measurement chain including a force transducer, a crosshead displacement sensor, and an electronic control unit (Figure 6a).

The load measurement error of the machine is ±0.5% of the applied force, while the displacement measurement error of the movable crosshead is ±0.25%, providing sufficient accuracy for capturing the mechanical response within the deformation range considered in this study.

To reproduce the support-surface inclination angles corresponding to specific phases of the gait cycle, a wedge fixture with a prescribed opening angle was manufactured. For each point of the test cycle, the wedge angle was measured in advance and fixed, enabling the following inclination values to be implemented: 32°, 28°, 20°, 12°, 0°, −7.5°, −15°, and −18°. The angle was set with an accuracy of ±1°. Negative angles corresponded to loading predominantly applied to the heel region of the prosthetic foot, whereas positive angles represented loading of the toe region.

For each prosthetic foot, eight load-deformation curves (“force-displacement” diagrams) were obtained, comprising the vertical displacement of the machine crosshead and the corresponding reaction force measured at the base. All tests were performed with a controlled crosshead speed of 20 mm/min. These experimental curves were used for comparison with the numerical predictions and for quantitative assessment of the validity of the adopted computational setup.

2.6.2. Interpretation of the Obtained Data

For validation of the numerical model, experimental measurements were compared with finite element predictions for the parameters most relevant to the biomechanical performance of the prosthesis, namely the vertical deformation of the heel region and the flexion angles in the regions corresponding to the ankle and metatarsophalangeal joints.

The coordinates of the control points used to compute these quantities were reconstructed based on known geometric relationships defined by the support-surface inclination angle γ and the actuator (piston) displacement h of the testing machine (Figure 6b). This procedure enabled reproducible determination of the kinematic metrics for each loading condition and ensured a consistent basis for comparison with the finite element results.

2.7. Methods for Testing the Prosthetic Feet on the Amputee

The developed prosthetic feet were evaluated both under laboratory bench conditions and during real-use walking. The tests included a qualitative assessment of structural integrity as well as a comparative gait kinematics analysis aimed at quantifying gait symmetry when using Model 1 and Model 2. Experimental testing was performed by a single participant—a 33-year-old male (height 179 cm, body mass 108 kg) with a right-sided transtibial amputation (TTA) resulting from a military injury. At the time of the study, the participant had completed rehabilitation and demonstrated stable independent walking without pronounced limitations; no chronic diseases or subjective complaints affecting mobility were reported.

Gait kinematics were recorded using an inertial motion capture system Perception Neuron 3 by Noitom (Beijing, China), comprising 17 wireless inertial measurement units (IMU), each integrating a triaxial accelerometer, gyroscope, and magnetometer [24]. In the Perception Neuron 3 system, the inertial sensor sampling frequency is 600 Hz, providing primary discretization of gyroscope, accelerometer, and magnetometer signals. The output data recording frequency in the Axis Studio software (version 2.12) by Noitom (Beijing, China) is fixed at 60 Hz, defining the temporal resolution of the final kinematic parameters available for analysis. The system latency between motion and the recorded sample does not exceed 25 ms. Comparative validation studies of this IMU system against high-precision optical motion capture systems indicate that the typical discrepancy in lower-limb joint flexion angle estimates during walking does not exceed 5° [25].

Data were collected in a standard indoor environment. Prior to each session, the Perception Neuron 3 system was calibrated in Axis Studio software following the manufacturer’s procedure, including sensor placement verification and T-pose calibration, heading alignment, and a short verification trial to confirm joint-angle plausibility. Magnetometer correction was enabled and potential magnetic disturbances were minimized by avoiding large ferromagnetic objects near the walkway and repeating calibration when required.

For comparative assessment, gait data were collected under three prosthetic conditions: the reference prosthesis Ottobock (Duderstadt, Germany) Triton (the participant’s habitual device), and the developed Model 1 and Model 2. To improve statistical robustness and reduce the influence of step-to-step variability (adaptation, cadence fluctuations, minor technique changes), for walking on each of the prostheses, 3 records of amputee walking were made. During the recording of the data, the patient made 4 direct passes through the office. Each straight passage contained only 8–10 steps. Thus, to analyze the bending angles of the hip, knee and ankle, data from about 54 steps per leg for each of the prostheses were analyzed. Before each recording condition, the participant completed a familiarization (acclimation) period of approximately 30 min of walking with the tested foot. Consequently, the gait data represent short-term adaptation, and longer acclimation periods (days–weeks) are required for definitive comparisons of functional performance.

The primary IMU outputs comprised a time series of lower-limb joint angles—hip, knee, and ankle flexion-extension. In addition, the vertical trajectory of the body center of mass (CoM) was analyzed as an indicator of vertical oscillations and as a proxy metric related to the energetic “economy” of gait.

The processing pipeline included several consecutive steps. For each pass, gait cycles were segmented by detecting local extrema (minima/maxima) in the knee flexion-extension signal, which served as a robust marker of gait phase. The extracted cycles were then filtered using allowable ranges of extrema to exclude non-representative segments associated with turning and transitional movement. Finally, the kinematic time series were averaged across all validated cycles by computing pointwise mean values, yielding representative ensemble profiles for each test condition.

The output of this processing procedure consisted of averaged time-series for hip, knee, and ankle flexion/extension angles, as well as the vertical CoM oscillation trajectory. These metrics were subsequently used for comparative analysis of gait symmetry and biomechanical performance during walking with the investigated prosthetic foot designs.

It should be noted that, given that the protocol involved a single participant (n = 1), the gait component of this study is intended as a pilot feasibility assessment aimed at verifying safe use and biomechanical plausibility of the printed feet, rather than establishing generalized clinical performance.

3. Results

3.1. Strength Performance

Figure 7 presents the von Mises equivalent stress distributions for the simulated load cases corresponding to 15, 30, 45, and 55% of the gait cycle. These time points were selected as the most demanding in terms of the functional response of the structures, as they coincide with the maximum metatarsal-region flexion α, toe-region flexion β, and heel vertical deformation dz.

The results indicate that local peak stresses occurring near geometric features (sharp transitions, edges, contact interfaces, and regions affected by boundary condition enforcement) may represent numerical stress concentrations and singularities and can therefore formally exceed the material strength limit. To obtain a representative estimate of stresses in structurally meaningful regions, a statistical post-processing procedure was applied whereby the top 1% of the maximum von Mises stress values were excluded from consideration. The resulting corrected maximum stress σ_max did not exceed 4.3 MPa for Model 1 and 7.5 MPa for Model 2. This indicates that, under the considered normative loading regime, the material response remains predominantly within the elastic range (σ_e < 8.6 MPa) and that the strength criterion is satisfied for both designs (Figure 8).

The corrected value σ_max = 7.5 MPa corresponds to an estimated margin of approximately 10–15% relative to the elastic limit of TPU, indicating that the developed prostheses operate predominantly as elastic energy-storage structures within the simulated cycle. This is important because even occasional excursions into the plastic regime may lead to the accumulation of residual deformations, alterations in rollover geometry, and accelerated damage growth under fatigue loading.

The obtained stress fields also enable a qualitative interpretation of the structural load-transfer mechanisms. For both designs, regions of elevated stress are primarily localized at transitions from relatively stiff components to compliant “spring-like” segments, as well as near attachment and contact areas where stress concentrations are inherently expected. At the same time, the majority of the material volume experiences moderate stress levels, suggesting an appropriate load path distribution and the absence of global overstressing throughout the stance phase.

An analysis of the critical region in the vicinity of an abrupt transition in boundary conditions at the prosthesis attachment site was also conducted, for which calculations employing a locally refined mesh were performed (Figure 9). The region corresponding to loading values exceeding the conventional elastic limit yet remaining below the ultimate strength is highlighted in orange. Notwithstanding the potential for accumulation of fatigue-induced deformations within this region, the ultimate strength threshold is not exceeded. Furthermore, the potential influence of numerical errors in the vicinity of the examined region cannot be entirely excluded.

To assess the relevance of the isotropic model, the normal stress fields were analyzed (Figure 10). It can be seen from the figure that at the moment of maximum loading, the primary load in the working region is borne by the fibers oriented along the y-axis, while along the x and z axes, the loads do not exceed 0.8 MPa and 1.2 MPa, respectively. Accounting for the difference in the actual elastic moduli is impractical, since loading along the designated axis does not contribute significantly to the bending of the working element.

3.2. Biomechanical Performance

The computed flexion-angle profiles over the full stance phase are shown in Figure 11. Analysis of the α(T) and β(T) curves indicates that the most pronounced deformation of the metatarsal and toe regions occurs within T ≈ 45–60%, i.e., during the terminal part of stance corresponding to the transition toward push-off. This finding is physically expected and consistent with the functional role of the prosthetic compliant elements: as the load increases in late stance, elastic energy is stored primarily through bending deformation, and as the ground reaction force (GRF) application point progresses anteriorly toward the forefoot, this energy is released, thereby supporting rollover and enhancing push-off dynamics.

The results further show that the modular Model 2 exhibits higher flexion angles during the second half of stance, indicating increased compliance under the loading conditions most relevant to energy storage and return. From a biomechanical perspective, increased α and β for T > 45% can be interpreted as more pronounced elastic energy accumulation and more effective generation of the resultant moment about the ankle region. This behavior is consistent with an increased effective lever arm during toe-off and may potentially reduce the user’s energetic demands.

In the first half of the stance (T ≤ 45%), differences between the two designs are markedly smaller. This characteristic is critical for gait stability and controllability: during initial contact and mid-stance rollover, the prosthesis should provide limited and predictable compliance (in particular via heel damping and load redistribution), preventing excessive “sink-in” and undesirable oscillations of the ground reaction force. Consequently, the identified differences between the design strategies manifest primarily in the functionally critical terminal-stance region, where the prosthesis is expected to operate predominantly as an elastic energy-storage element rather than as a damper. These findings support the conclusion that the architecture and orientation of the compliant features mainly govern the behavior of the prosthesis during push-off, i.e., the phase most relevant to gait dynamics and energy efficiency.

Figure 12 presents the calculated profiles of heel vertical deformation dz, during the first half of the stance phase. It should be emphasized that this parameter is used as an integral indicator of the damping capacity of the posterior region of the prosthesis and reflects the overall compliance of the structure during initial contact with the support surface.

The presence of a pronounced yet amplitude-limited deformation within T ≈ 0–15% indicates that the required cushioning regime is achieved: the structure attenuates the impact component of loading and reduces the steepness of the rise in the ground reaction force, while avoiding excessive compliance (“heel collapse”). From a biomechanical perspective, this is of fundamental importance because a reduced loading rate of the vertical ground reaction force is associated with lower peak accelerations of the limb segments and, consequently, may decrease discomfort and impact transmission to the residual limb. Moreover, a smoother load transfer at the onset of stance may reduce compensatory overload of proximal joints (knee and hip) by mitigating impulsive load components and distributing forces more evenly over time.

Comparison of the two designs further shows that Model 2 exhibits a smoother evolution of dz(T), i.e., fewer pronounced inflection points and abrupt changes in deformation during early stance. This response can be interpreted as a more coordinated interaction between the damping and load-bearing elements, providing a more gradual redistribution of loads during the transition from heel contact to rollover through the midfoot region. In practical terms, such smoothness is beneficial for stable gait formation, as reduced abrupt variations in vertical deformation may lower the need for compensatory motions at proximal joints and potentially contribute to improved gait symmetry and reduced fatigue during prolonged walking.

3.3. Bench Testing

Figure 13 shows the experimentally measured piston (moving crosshead) displacement from initial contact to the load levels corresponding to the selected pairs of platform inclination angle γ and test force F.

The measured displacements at prescribed inclination angles represent the overall compliance of the “prosthesis–support surface” system and, within the adopted setup, can be interpreted as an integral indicator of the prosthesis’ equivalent stiffness at different stages of stance. Because the test protocol reproduces a sequential transition from negative to positive inclination angles, the resulting data make it possible to functionally separate the contributions of the main structural subsystems: the heel damping block, which mitigates impact during early stance, and the compliant elements of the forefoot, which govern deformation behavior and potential energy return in late stance.

The observed nonlinearity of the force–displacement response is expected for TPU-based structures and is attributable to both geometric nonlinearity (bending of slender compliant members) and the elastomer’s rheological behavior. This characteristic can be advantageous: under low loads, the prosthesis may provide greater comfort and cushioning, whereas with increasing load, it becomes effectively stiffer, improving support and enabling elastic energy storage.

From an applied perspective, such a nonlinear response is functionally desirable, as it offers a compromise between enhanced damping at early stance (via higher compliance under small loads) and a progressively increasing effective stiffness with load, which improves stability and promotes elastic energy accumulation during the terminal stance phase preceding push-off.

3.4. Comparison of Numerical Predictions with Experimental Data

Using the piston displacements recorded during bench testing, the vertical displacements of the predefined control (support) points on the structure were reconstructed. The values obtained for the heel region are presented in Figure 14a. Subsequently, by combining these vertical displacements with the horizontal displacements taken from the numerical simulations, the inclination (flexion) angles of the metatarsal segment and the toe region were calculated (Figure 14b).

Overall, the comparison between the numerical predictions and the experimental results demonstrates satisfactory agreement in both the magnitude of deformations and the shape of the response curves for the key metrics—α, β, and heel vertical deformation dz (Figure 14c). This outcome supports the validity of the adopted computational setup: the imposed boundary conditions and simplifying assumptions (frictionless contact, isotropic material approximation, and quasi-static loading) capture the dominant deformation mechanisms of the prosthetic feet within the stance phase and provide sufficient predictive capability for comparative design assessment.

The remaining discrepancies, most pronounced at low load levels and during the initial contact stage, are likely attributable to factors that are not fully represented in the current model, including:

(1)

The viscoelastic behavior of TPU and the associated dependence of stiffness on loading rate and history (hysteresis), which is particularly influential at small strains and during transient regimes;

(2)

The presence of real friction and micro-slip at the “sole–support surface” interface, which can modify the effective load-transfer conditions and introduce additional tangential forces and moments;

(3)

Anisotropy inherent to FDM components and the effect of interlayer adhesion, leading to direction-dependent compliance and local stress–strain states that deviate from an isotropic approximation.

3.5. Patient Testing

The participant described in Section 2.7 performed a pilot feasibility observation of both developed feet. Figure 15 presents the processed time histories of gait kinematics acquired using the inertial motion capture system (see Section 2.7). Data obtained with the participant’s habitual prosthesis (Ottobock Triton) are also shown for reference.

The results indicate that, when using either 3D-printed foot, the hip and knee joint motion patterns remain qualitatively similar despite the structural differences between the two designs. This suggests comparable load-bearing capability during straight, level walking and the absence of pronounced compensatory strategies at proximal joints (hip/knee), at least within the scope of the applied test protocol.

The most notable differences between the prototypes are observed in the kinematics of the ankle–foot segment on the prosthetic side, which is expected because this region is directly governed by the compliance and damping characteristics of the prosthetic foot. Model 2 exhibits a more pronounced flexion–extension profile during stance, consistent with its design intent to promote elastic energy storage and return and to smooth the transitions between initial contact, rollover, and push-off. In contrast, Model 1 shows a comparatively “stiffer” response during certain phases, which may reflect higher local stiffness of the monolithic structure and a more limited redistribution of deformation among compliant features.

The Ottobock Triton reference foot demonstrates an overall “stiffer” kinematic pattern with smaller ankle-like deformation compared to the printed prototypes [9,26,27,28]. Nevertheless, both 3D-printed designs exhibit kinematics that can be considered biomechanically plausible: the hip and knee flexion/extension curves on both limbs retain the typical walking pattern with expected phase-related extrema and without abnormal drops or discontinuities, consistent with established descriptions of normal joint kinematics in gait analysis literature [29,30,31]. Importantly, the primary impact of changing the foot component occurs where it is biomechanically expected in prosthetic systems—at the prosthetic-side ankle–foot level. TPU-based passive feet exhibit a different elastic–damping response and therefore produce a distinct dorsiflexion/plantarflexion profile compared to a carbon-fiber energy-storage-and-return foot. This shift reflects differences in material behavior and structural architecture (greater compliance and damping in TPU versus higher effective stiffness and energy return in carbon), rather than “poor” kinematics per se, and aligns with observations that, in transtibial amputees, differences between prosthetic feet are typically most evident at the ankle–foot segment, whereas hip and knee profiles often remain broadly comparable [32,33].

Finally, it should be emphasized that the intact limb exhibited stable and consistent behavior for both prototypes: hip and knee curves preserved a normal pattern without an apparent increase in compensatory motions, and differences between Model 1 and Model 2 on the intact side were limited. Taken together, these findings support the conclusion that both 3D-printed feet provide adequate load-bearing function and reproduce a biomechanically consistent stance-phase progression.

As shown in Figure 16, the greatest stability (i.e., the lowest variability of CoM extrema) was observed when the participant used the Ottobock Triton prosthesis. Nevertheless, when using the 3D-printed feet, the CoM extrema remained within a comparable range, without pronounced abnormal shifts or systematic “breakdowns” of the trajectory. This indicates preserved, controllable vertical dynamics and sufficient load-bearing function for both printed designs. The apparent advantage of Triton in terms of CoM stability is likely driven largely by habituation: the participant had long-term experience with the carbon-fiber foot, whereas acclimation to the new prototypes within the pilot protocol lasted only a few minutes and was insufficient for establishing a stable motor pattern. Therefore, the observed CoM differences should be interpreted primarily as reflecting different levels of adaptation rather than evidence of functional inadequacy of the printed feet.

4. Discussion

The TPU 95A FDM prosthetic feet developed in this study show that an appropriately designed compliant architecture can meet strength requirements while providing a meaningful stance-phase mechanical response. Under the ISO 22675:2024 loading cycle, finite element results indicate that both designs operate predominantly within the elastic range, with elevated stresses confined mainly to expected stiffness-transition and attachment/contact regions rather than broadly distributed overload.

Functionally, the two design strategies exhibit distinct behavior: both store and release elastic energy primarily in late stance, but the modular Model 2 demonstrates larger late-stance flexion (α, β) and a smoother early-stance heel deformation profile (dz), consistent with controlled damping at initial contact and a more pronounced spring-like response during push-off. Model 1 shows a comparatively stiffer response in several phases, which may reduce perceived energy return despite adequate heel cushioning.

Agreement between simulations and bench tests for α, β, and dz supports the proposed CAD–FEA–prototype—testing workflow as a practical tool for comparative design and iterative refinement. Remaining differences, particularly at low loads and during initial contact, are plausibly attributable to effects not captured in the current quasi-static, isotropic, frictionless formulation (TPU viscoelasticity, contact friction/micro-slip, and FDM anisotropy).

The pilot feasibility gait study further indicated that both 3D-printed feet enable biomechanically plausible gait patterns: hip and knee kinematic profiles remained stable, with no clear signs of pronounced compensatory strategies, whereas the primary differences were—as expected—localized to the prosthetic-side ankle–foot kinematics, reflecting the distinct elastic–damping behavior of TPU compared with carbon-fiber energy-storage-and-return designs. Analysis of vertical center-of-mass (CoM) oscillations showed greater stability with the habitual prosthesis; however, CoM extrema with the printed feet remained within a controlled range without abnormal shifts. The most plausible explanation is the short acclimation period to the new devices, which was insufficient for the formation of a stable motor pattern. In addition, it should be noted that, given the single-participant design and limited acclimation, the above findings should be interpreted as feasibility data demonstrating that design parameters measurably affect ankle–foot kinematics, rather than as evidence of clinical advantage.

Future work should focus on transitioning from proof-of-concept to engineering and clinical refinement. From a mechanics standpoint, priorities include long-term fatigue testing according to ISO 22675:2024, assessment of TPU property degradation due to aging and environmental exposure (temperature and moisture), and refinement of the numerical model by incorporating viscoelasticity, parameterized friction, and an orthotropic description of FDM material behavior. Also, future work should include a systematic screening of alternative elastomeric FDM materials beyond TPU 95A (e.g., TPU grades with different Shore hardness, TPE/TPC blends, and reinforced flexible filaments) to explore stiffness–damping trade-offs and to assess durability under cyclic fatigue and environmental aging. From a biomechanical perspective, the study should be expanded to a larger cohort (at least n ≥ 10) stratified by body mass, activity level, and amputation level, with a mandatory acclimation period prior to comparative measurements. Finally, from an engineering standpoint, the most promising direction is to parameterize the design for rapid tuning of stiffness and mass to individual users and to investigate hybrid solutions—multi-material printing, local reinforcement, and modular replaceable components—which is particularly compatible with the modular architecture of Model 2. In summary, these steps can advance elastomer-based FDM prosthetic feet toward practical deployment as an affordable and customizable alternative to more expensive commercial counterparts.

5. Conclusions

Two geometrically and structurally distinct low-profile prosthetic foot designs were developed, and an iterative “CAD–FEA–prototype—testing” workflow was implemented for their comparative evaluation. Finite element simulations under the ISO 22675:2024 normative loading cycle indicated that both designs satisfy the strength criterion. Functional analyses showed that the dominant metatarsal and toe deformations occur at 45–60% of stance, consistent with elastic energy storage and return during the transition to push-off. Compared with Model 1, Model 2 exhibited larger terminal-stance flexion angles and a smoother early-stance heel-deformation profile, reflecting a more compliant mechanical response and smoother rollover mechanics; potential implications for energetic economy and comfort remain inferential and were not directly measured.

Bench testing confirmed the adequacy of the numerical framework: the experimental trends for the key metrics agreed with the simulations in both shape and order of magnitude, supporting the validity of the adopted boundary conditions and simplifying assumptions. A pilot patient evaluation (n = 1) demonstrated biomechanically plausible gait patterns with both 3D-printed feet: hip and knee kinematics remained stable, while the main differences were expectedly localized to the prosthetic-side ankle–foot kinematics. Analysis of vertical center-of-mass oscillations showed higher stability with the habitual prosthesis; however, both printed feet maintained a controlled range of oscillations without signs of functional failure, likely reflecting the short acclimation period. Overall, the present findings support prototype-level feasibility of TPU-based FDM prosthetic feet and motivate further work on long-term durability, refined material/contact modeling, and larger-cohort gait studies with controlled acclimation before clinical competitiveness can be assessed.