Compositional Time–Temperature–Transformation and Post‐Processing Window Mapping of Epoxy Nanocomposites via Machine Learning

ABSTRACT

Time–temperature–transformation (TTT) diagrams are essential for defining processing windows in thermoset systems, but become experimentally impractical for nanofiller‐containing compositions. Here, DGEBA/TETA‐based epoxy systems containing graphene nanoplatelets (GN) or carbon nanotubes (CNT) (0–1 wt%) were investigated using DSC, rheological gelation, kinetic modeling, and machine‐learning‐assisted interpolation. Cure kinetics were described using a Kamal–Sourour model with diffusion control, combined with Di Benedetto vitrification analysis. Gel activation energies remained within 44–51 kJ·mol ⁻¹ , and gelation occurred at α _g ≈0.70–0.88 for all formulations, confirming preservation of the intrinsic single‐step autocatalytic epoxy–amine reaction mechanism. In contrast, CNT significantly intensified diffusion control, with log K _diff decreasing from 5.00 (neat epoxy) to −3.93 (1 wt% CNT). TTT analysis showed that vitrification shift, rather than gel displacement, governs the compression of the compositional cure‐time window. At 80°C, neat epoxy exhibited a post‐gel processing window (Δ t ) of ~40–50 min, GN systems showed Δ t  ≈ 17–28 min, while 1 wt% CNT reduced Δ t to ~5–9 min (≈60%–70% contraction). Machine‐learning models accurately reproduced TTT slices (MAE = 0.050 log ₁₀ [min]) and Δ t predictions (MAE = 0.037 log ₁₀ [min], ~9% deviation). This integrated framework converts discrete TTT measurements into continuous compositional cure maps for formulation design and processing optimization.