Coupled hygro-viscoelastic damage modelling of 3D-printed continous and short carbon fibre : reinforced nylon composites under aqueous aging

Abstract

Hygroscopic degradation in fibre-reinforced polymer composites couples matrix plasticisation, hydrolytic chain scission, and moisture-accelerated viscoelastic relaxation across kinetically separated timescales. This coupling is critical for fused-filament-fabricated (FFF) composites, where the layer-by-layer architecture and hygroscopic PA6-based Onyx matrix make moisture sensitivity particularly important, yet a constitutive framework capturing these simultaneous mechanisms remains unexplored for 3D-printed composites. This thesis presents an integrated experimental-computational framework for moisture-induced degradation in FFF carbon fiber/Onyx composites (V f = 21%). Aging studies in distilled water at 26◦C reveal a two phase mechanical response: an initial rapid strength loss of 11.7 % within 15 days (0.78 %/day) decelerates sharply to a cumulative 17.4 % at 90 days (0.05 %/day), accompanied by 26.1 % stiffness reduction and a non-monotonic failure strain peaking at 60 days. This indicates a mechanistic transition from mobile-moisture plasticisation to bound-moisture hydrolysis, a behavior that single-phase Fickian diffusion cannot reproduce.

A hygro-viscoelastic dual-damage constitutive model was developed and implemented as a user-defined material subroutine (UMAT) in Abaqus/Standard. The framework couples: (i) a Carter–Kibler transport model partitioning moisture into mobile (Cm) and bound fraction (Cb); (ii) two independent damage variables, reversible plasticisation driven by Cm and irreversible hydrolysis driven by Cb, each with its own kinetics; and (iii) a Generalised Maxwell viscoelastic model calibrated via dynamic mechanical analysis. The intrinsic timescale separation of the Carter–Kibler model (mobile saturation within days, bound accumulation over approximately 166 days) reproduces the experimentally observed fast-to-slow degradation. Coupled simulations show that the combined effect of moisture-accelerated relaxation on a damage-softened matrix exceeds the sum of isolated damage and viscoelastic contributions, indicating a non-linear amplification not capturable by uncoupled approaches.

Parameters identifiability analysis using bootstrap resampling (nboot = 1000) shows that the plasticisation coefficient is well-identified from the 90-day data (95% CI width ≈ 0.08), but the hydrolytic coefficient is practically non-identifiable within this window. Adding a single independent 365-day strength datum reduces the one-year prediction uncertainty from 22 to 7 percentage points without affecting short-term accuracy. These results establish a quantitative identifiability criterion for hygroscopic degradation models: at least one mechanical datum beyond the bound-moisture trapping timescale (t > τtrap) is required for operationally useful long-term predictions, a condition that conventional short-window experimental protocols are structurally unable to satisfy.