Cure process evaluation of CFRP composites via neural network: From cure kinetics to thermochemical coupling

This work focuses on the comprehensive modeling of the cure kinetics and the thermochemical coupling to investigate the cure process of CFRP composites. Neural network (NN) is instead of the general cure kinetics model to find the relationships between the cure kinetics parameters where the temperature rate is also involved. Large data sets from non-isothermal differential scanning calorimetry (DSC) are used for the network training and validation. For comparison, the NN model and general model are respectively written as user subroutine and combined with the thermochemical coupled model to implement the finite element (FE) cure analysis. Compared with the general model, the NN model significantly improves the prediction accuracy of cure behavior. The reaction rates at various temperature rates from the NN model are also coincide with the corresponding experiments. Furthermore, it is also found that the NN model can capture the local trends of the composites during cure by taking into account the effect of the temperature rate on the cure kinetics. The temperature and degree of cure (DoC) gradients of the composite during cure process are reduced than the general model, which provides a new insight for the cure and cure related characteristics analysis.

Automated summary

This work focuses on the comprehensive modeling of the cure kinetics and the thermochemical coupling to investigate the cure process of CFRP composites. By using neural network (NN) model, the prediction accuracy of cure behavior is significantly improved and the local trends during cure process are captured.