PSI - Issue 77

Boyu Li et al. / Procedia Structural Integrity 77 (2026) 316–322 Boyu Li et al. / Structural Integrity Procedia 00 (2026) 000 – 000

320

5

These observations suggest that the diffusivity of AM PLA was not homogeneous. In the vertical specimens, the inter-filament channels accelerated the moisture transport and, hence the water uptake, consistent with the FEA in Li et al. (2024). In contrast, moisture penetration in the horizontal specimens proceeded more slowly, leading to partial hydrolysis of the material in regions not yet saturated. This non-uniform diffusion process may account for the secondary decrease in mean weight observed in the fifth week. In addition, pH testing of the immersion water with the indicator paper confirmed a neutral pH throughout the experiment, suggesting that the observed weight variations were primarily due to moisture diffusion and hydrolysis. 3. Simulation study FE models of the horizontal and vertical specimens were developed using the FEA software ABAQUS. A model with homogeneous diffusivity was also introduced for comparison. Thanks to the geometric symmetry of the printed specimens, only a quarter of the geometry was modelled. The models were meshed with DC3D8 linear brick elements. The diffusivity and solubility parameters employed in the simulations are summarised in Table 2. Diffusion properties used in simulations were acquired from Banjo et al. (2022). The specimens were analysed under boundary conditions consistent with those applied in the experimental study. Also, following Wang (2024), the diffusivity along the printing direction was considered ten times higher than in the transverse directions (highlighted in blue in Tab. 1). The simulations were conducted for a total diffusion duration of 24 hours using a fixed time step of 60 s.

Table 2. Diffusion properties used in simulations. Homogeneous model Horizontal model

Diffusivity (mm 2 /s)

5.73×10 −6 5.73×10 −6 5.73×10 −6 0.9%

5.73×10 −5 5.73×10 −6 5.73×10 −6 0.9%

5.73×10 −6 5.73×10 −6 5.73×10 −5 0.9%

Vertical model

11 22 33

Solubility

In Fig. 3, the evolution of water diffusion in homogeneous, horizontal and vertical models is illustrated at 1200 s. For the homogeneous model, the water uptake proceeded in a centripetal manner, with the uniform saturated thickness due to isotropic diffusivity. As a result, the homogeneous model required the longest time to reach saturation. The vertical specimen exhibited a higher diffusion rate than the homogenous model, thanks to its larger 33 values (Fig. 3a and Fig. 3c). The moisture was transported along the printing direction to the outer side in the upper part of the specimen. For the horizontal model, diffusion initially accelerated in the in-plane direction when the portion of the model below the water surface was not fully saturated(Fig. 3b). Next, the evolution of the average concentration for each model was analysed (Fig. 4). The horizontal model had the water uptake rate similar to that of the vertical model until the water concentration reached 0.4% at 1600 seconds, when the immersed specimen part was close to the saturation state. All concentration-evolution curves had similar contours at the later stages of the diffusion process. The higher 11 values did not significantly influence the process at the later stages, after the immersed part was saturated and the main diffusion direction was along the Z-axis. The simulation results agree with the observation that the mass-change difference in the experiment was mainly due to the higher diffusion rate in the printing direction. Still, the anisotropic diffusivity of the printed material needs further fitting to the experimental data. Also, the diffusion rate perpendicular to the printed layer was assumed to be equal to the in-plane diffusivity. Further, in future developments, the internal structure of the AM polymers will be reconstructed to investigate the effects of the filament geometry and inherent defects (voids) on the diffusion process.