ABSTRACT Understanding the cure kinetics of epoxy prepregs is essential for controlling the curing process. In this study, the cure kinetics of a carbon fiber (CF)/epoxy prepreg were investigated using isothermal and non‐isothermal differential scanning calorimetry (DSC). Six phenomenological kinetic models were fitted and compared. A Friedman isoconversional analysis was used to provide initial estimates and constraints for the Arrhenius parameters, which were then refined by nonlinear least‐squares estimation with a trust‐region reflective (TRR) solver. Two global‐to‐local pipelines were also evaluated: TRR with particle swarm optimization (PSO) initialization and TRR with genetic algorithm (GA) initialization. Across both isothermal and non‐isothermal datasets, the piecewise model consistently demonstrated excellent fitting accuracy, achieving coefficients of determination ( R 2 ) exceeding 0.98 and 0.99, respectively. Global initialization prior to TRR yielded modest accuracy improvements compared to TRR alone, albeit at a higher computational cost. Models calibrated with non‐isothermal DSC produced superior predictions for a standard cure cycle compared with models calibrated on isothermal data. These results provide practical guidance for selecting effective combinations of experimental protocol, kinetic model, and parameter‐estimation strategy to support reliable thermochemical simulation and process optimization of epoxy prepregs.
Zhang et al. (Sun,) studied this question.