The research, led by Laoise Reynolds with Omid Sam-Daliri, Yadong Jiang and Pouyan Ghabezi, was motivated by a persistent problem in modern manufacturing. Conventional assembly relies on adhesives, mechanical fasteners or metallurgical joining, all of which carry inherent drawbacks. Adhesive bonds can trap voids and microcracks during manufacture, degrade under moisture, corrosive agents and temperature swings, and demand meticulous surface preparation. Fasteners and welds add weight, create stress concentrations, complicate recycling and make disassembly difficult. As industries push toward modular, repairable and environmentally responsible design, the researchers saw an opportunity in Sashimono, the traditional Japanese woodworking technique that achieves precise, load-bearing assembly with no extra joining elements at all.

Rather than copying traditional wooden joints exactly, the team designed three Japanese-joinery-inspired configurations suited to thermoplastic parts: a Dovetail joint, a T joint and a C joint. Each was printed in polylactic acid, or PLA, a biodegradable thermoplastic derived from plant sugars that is a staple of fused deposition modelling, the most widely used form of 3D printing. PLA offers a low processing temperature, good dimensional stability and a strong sustainability profile, making it an ideal candidate for prototyping interlocking concepts aimed at low-to-medium load-bearing applications. The parts were printed flat on a Prusa MK4 printer with a 0.4 mm nozzle, a 0.2 mm layer height, 100 percent infill and a plus-or-minus 45 degree raster orientation, with a dimensional tolerance of 0.05 mm on the interlocking features so the components could be pressed together by hand with a secure fit.

The experimental campaign was exhaustive. For each joint family, the researchers varied three geometric parameters: neck thickness, a shape-specific feature such as the dovetail angle, the C-joint radius or the T-joint neck length, and the number of interlocking notches, from one to three. Five specimens were printed and tested for every parameter combination, amounting to 270 printed pieces and 135 completed assemblies, roughly 40 hours of printing time. Standard dog-bone specimens were also printed as a continuous reference. Every assembly was pulled to failure on an Instron 4467 universal testing machine at a crosshead speed of 2 mm per minute, while a stereo camera system running GOM ARAMIS digital image correlation captured full-field strain maps on the specimen surface every two seconds, revealing exactly where deformation concentrated and where cracks began.

The continuous printed PLA reference achieved an ultimate tensile strength of 44.4 plus or minus 2.5 MPa with an effective elastic modulus of 2511 MPa, failing in a brittle manner from a localised strain zone that peaked near 20 percent. The interlocking joints, by contrast, reached only a fraction of that strength, a result the authors stress is expected for a proof-of-concept mechanical screening of geometric interlocks. The best performer was the C joint with a 5 mm neck thickness and a single notch, which achieved about 7 MPa, roughly 84 percent below the solid reference but the highest of all tested configurations. Notably, the C joint with a 3 mm neck was far weaker at 4.6 MPa, while thickening the neck to 7 mm also reduced strength to 5.0 MPa, showing that more material does not automatically mean better performance.

Digital image correlation explained why. In the thin-necked C joint, strain concentrated between 6 and 12 percent around the narrow neck, which snapped off in a sudden brittle fracture. The 5 mm version spread strain more evenly, reached about 15 percent local strain and failed progressively with only partial detachment of the neck. The 7 mm joint deformed the most of all, with recorded strains up to 37 percent, absorbing energy through plastic deformation before large chunks of material finally separated. Radius played its own role: increasing the C-joint radius from 3 mm to 5 mm boosted tensile strength by about 31 percent, but the joints also failed more abruptly, while the smallest radius allowed extensive plastic deformation without any visible cracking.

The number of notches produced opposite effects in different joint families, one of the study’s most intriguing findings. In the C joint, adding notches reduced tensile strength, from 5.8 MPa with one notch to 4.9 MPa with three, a drop of nearly 16 percent, while stiffness climbed sharply, with the effective elastic modulus rising about 66 percent. More notches restricted deformation, making the joint appear stiffer but weaker. The Dovetail joint behaved in exactly the reverse manner: three notches delivered 6.1 MPa, a 27.6 percent improvement over a single notch, with stiffness nearly doubling. DIC imaging showed the three-notch dovetail redistributing load after the first crack formed, sustaining additional force before a second brittle fracture appeared, a progressive failure pattern that contrasts with the single catastrophic break of simpler designs.

Dovetail geometry itself proved sensitive to angle and thickness in unexpected ways. Increasing the dovetail angle from 55 to 62.5 degrees raised tensile strength by about 21 percent, yet further thickening the neck from 5 mm to 10 mm cut strength by 38 percent while more than tripling the strain at break, from 1.7 to 5.3 percent. Thicker dovetails absorbed far more deformation, with local strains reaching 64 percent before damage spread, but they did so at lower overall strength and stiffness. The T joint emerged as the weakest family, with its single-notch configuration managing only 2.0 MPa, though adding notches transformed it dramatically: three notches lifted strength to 5.9 MPa, a 188 percent improvement, and produced the highest effective elastic modulus of any configuration at 1350 MPa. DIC revealed the middle notch bearing the brunt of the load, cracking first and then transferring strain to the remaining notches in a staged, progressive collapse.

The authors are careful to frame the scope of their conclusions. The study sits within a Design for Assembly context but did not measure quantitative DFA metrics such as assembly time, assembly force, disassembly cycles, wear or manufacturing cost, and the reported comparisons are descriptive rather than statistically tested. The findings apply specifically to PLA, the tested geometries and static tensile loading. Even so, the message is striking: centuries-old joinery wisdom, translated into printed thermoplastic, can produce tool-free, adhesive-free, fully reversible connections whose strength and failure mode can be tuned almost at will through geometry. A designer can now choose a configuration that prioritises strength, like the multi-notch dovetail, or one that maximises stiffness, like the three-notch T joint, simply by adjusting parameters that a desktop printer can produce overnight.

That tunability is precisely what makes the work resonate beyond the laboratory. In an era of circular-economy targets, products that snap together, come apart cleanly and can be repaired or reconfigured component by component represent a genuine alternative to glued and bolted construction. The Irish team’s demonstration that digital image correlation can pinpoint exactly where printed interlocks concentrate strain gives future designers a practical map of weak points to avoid. The C joint, with its combination of the highest tensile strength and progressive, deformation-tolerant failure, is flagged as the most promising candidate for further development. If subsequent studies confirm that these printed joints survive repeated assembly cycles and real-world loading, the ancient art of Sashimono may find an unlikely second life inside everything from flat-pack furniture to modular drone frames, all built without a screw, a rivet or a bead of glue.

Subject of Research: Mechanical performance of 3D-printed PLA interlocking joints inspired by traditional Japanese joinery

Article Title: Design for assembly of 3D-printed structures using joinery-inspired interlocking joints

Article References: Reynolds, L., Sam-Daliri, O., Jiang, Y., & Ghabezi, P. (2026). Design for assembly of 3D-printed structures using joinery-inspired interlocking joints. Results in Engineering, 32, Article 113437. https://doi.org/10.1016/j.rineng.2026.113437

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113437

Keywords: 3D printing, Japanese joinery, PLA, interlocking joints, dovetail joint, digital image correlation, design for assembly, tensile testing, additive manufacturing, fused deposition modelling, sustainable design, mechanical testing

The study, published in the journal Sports Engineering, set out to answer a deceptively simple question: does a running shoe behave the same way when it is compressed straight down as it does when it is compressed at the angles produced by real foot strikes? Standard test methods, including the widely used ASTM F1614 protocol, prescribe uniaxial loading in which the shoe is squeezed perpendicular to its sole. Yet runners rarely land that way. Rearfoot strikers contact the ground heel-first with the sole inclined roughly 16 to 20 degrees above horizontal, while forefoot strikers land on the ball of the foot with the sole tilted about 15 to 19 degrees in the opposite direction. Those angular differences change which part of the midsole absorbs the load and how much foam is engaged.

To capture this, the team tested six women’s US size 8.5 running shoe constructions that differed in midsole material, hardness and stack height. Two commercial New Balance models, the 880 with a 34.5 millimetre stack and the 1080 with a 40.5 millimetre stack, provided the basis for constructions built on production tooling, meaning the midsoles were representative of retail footwear. The materials spanned the two foams that dominate the modern market: ethylene-vinyl acetate, or EVA, the long-standing workhorse valued for cushioning, durability and low cost, and polyether block amide, or PEBA, the high-resilience superfoam behind today’s performance racing shoes. PEBA was tested at two hardness levels, 37 and 50 Asker C, while EVA was available only at the harder 50 Asker C grade.

Each shoe was fitted over a rigid mechanical last shaped to a size 8.5 foot and mounted on an Instron ElectroPuls E3000 testing machine. The last could be rotated about its transverse axis, allowing the researchers to load the shoe at three orientations: 15 degrees of dorsiflexion to mimic a rearfoot strike, a neutral 0 degrees for a midfoot strike, and 15 degrees of plantarflexion for a forefoot strike. Every cycle ran from a 50 newton preload to a peak force of 1500 newtons, roughly 2.5 times the body weight of a 60 kilogram runner, applied as a sinusoidal waveform at 1.5 hertz under closed-loop force control. Because the waveform shape and peak force were held constant, orientation was the only loading variable that changed. Each shoe received 30 preconditioning cycles before testing, and 25 cycles were recorded at each orientation, with the first five discarded to let the foam settle after repositioning.

From the force-displacement data the researchers extracted three outcomes. Vertical stiffness was calculated as the gradient of the loading curve between 70 and 90 percent of peak force, a range that avoids the nonlinear toe region at low loads and the densification zone near maximum compression. Peak displacement was the compression recorded at peak force. Energy return percentage was the energy recovered during unloading, obtained by trapezoidal integration, expressed as a fraction of the energy absorbed during loading. After removing 17 outlying cycles, or 4.7 percent of the dataset, by an interquartile range rule confirmed by curve inspection, 343 observations remained for statistical modelling.

The results were striking in their consistency. In every single shoe construction, stiffness was lowest under dorsiflexion, the rearfoot-strike orientation, and peak displacement was highest under the same condition. Compared with neutral loading, dorsiflexion reduced stiffness by 72 newtons per millimetre in the material-comparison model and increased peak displacement by 8.6 millimetres. Plantarflexion produced the same pattern of softer, deeper compression but with smaller magnitudes, a 37 newton per millimetre stiffness reduction and 2.4 millimetres of extra displacement. In the second model, comparing soft and hard PEBA, the orientation effects were smaller for stiffness but nearly identical for displacement, at 8.3 and 2.3 millimetres. All of these effects were statistically significant.

The physical explanation lies in contact geometry. When the shoe is tilted relative to the applied load, the region and volume of foam engaged during compression change, reducing the amount of material available to support the force. With less foam doing the work, the midsole compresses further under the same load. Dorsiflexion produced the larger effect, which the authors attribute to the reduced contact area associated with angled loading on the curved rearfoot geometry. The findings extend earlier work by Mohammadi and Nourani, who reported that compression angle affected strain energy in EVA soles, and they align with the observation that rearfoot and forefoot strikers load different regions of the midsole during running. In other words, the orientation effects observed in the laboratory are mechanically relevant to what actually happens on the road.

Energy return told a different story. Of the three outcomes, it was the least sensitive to loading orientation, and most of its variance was explained by differences between shoe constructions rather than by the angle of loading or the number of cycles. In the material comparison, the contrast between EVA and PEBA accounted for a large share of the variance, with PEBA showing a non-significant trend toward higher energy return and significantly greater peak displacement. In the PEBA hardness comparison, hardness itself explained very little, suggesting that the differences between soft and hard PEBA shoes reflect features other than durometer, possibly including geometry or microstructure that the study did not characterise. This supports the researchers’ hypothesis that energy return, being rooted in material-level properties, would remain relatively stable across orientations.

The study also documented a subtle but important drift across the 20 analysed cycles. Peak displacement and energy return changed slightly and linearly with successive loading cycles, meaning that the value measured depends on how many cycles precede the measurement. A standard single-orientation, fixed-cycle test captures neither this cyclic drift nor the orientation sensitivity, so it may not represent the stiffness and displacement a runner actually experiences. The authors are careful to note that the standard test remains a useful benchmark, but they argue that a fuller characterisation of running-relevant loading requires more than one orientation and a stated cycle count. Because rearfoot cushioning properties are known to influence vertical ground reaction forces during heel-toe running, the orientation-dependent differences could carry biomechanical consequences in actual use.

The researchers acknowledge several limitations. Only one shoe was tested per material-hardness-stack-height combination, so manufacturing variability could not be assessed, and with just two constructions per group the between-group comparisons are exploratory. The constant 1500 newton peak force does not capture the variation in ground reaction forces across running speeds and body masses, the three tested orientations approximate but do not span the continuous spectrum of real foot strike angles, and each material was represented by a single density, so material and density effects could not be separated. Foam cell structure was not characterised, and stack height was compared between two shoe models whose other differences cannot be fully disentangled. Future work, the authors suggest, should test whether the orientation effects translate into measurable biomechanical changes during running, examine loading frequencies given the viscoelastic nature of midsole foams, extend cyclic testing to see whether orientation effects persist as foam ages, and replicate across multiple shoes of the same construction. For now, the message for the footwear industry is clear: a shoe tested only straight down may not behave the way it does when a runner lands on the heel or the forefoot, and multi-orientation testing offers a more honest picture of how midsoles actually cushion the millions of angled impacts that running delivers.

Subject of Research: Effects of loading orientation on the mechanical cushioning properties of running shoe midsoles

Article Title: Mechanical cushioning properties of running shoe constructions across loading orientations

Article References: Scherrer, D., Legg, K. A., Rogers, C. W., Gottschall, J. S., & Cochrane, D. J. (2026). Mechanical cushioning properties of running shoe constructions across loading orientations. Sports Engineering, 29(2), Article 33. https://doi.org/10.1007/s12283-026-00567-2

Image Credits: AI Generated

DOI: 10.1007/s12283-026-00567-2

Keywords: running shoes, midsole, cushioning, EVA, PEBA, foot strike, stiffness, energy return, biomechanics, mechanical testing, stack height, sports engineering