Abstract
This work focuses on addressing the pain points of poor generalization performance and difficulty in continuous learning that exist in the phenomenological and neural network surrogate models. Therefore, this study proposes a lightweight adaptive thermochemical information aggregation networks (ATANets) to overcome the gradient conflict challenge, and combines the generative knowledge distillation (GKD) algorithm to compress the model to capture finer-grained and enriched information on kinetics behaviors, which yields a thermochemical feature information extraction networks (FENets) with incremental learning capability. The experimental results demonstrated that as the complexity of the learning task deepened, the FENets models obtained by incremental training with ATANets and GKD still had excellent continuous learning capability, with the relative variations of the coefficients of determination and the mean square error being smaller than 6.7 × 10−3 and 0.3860 × 10−3, respectively. Meanwhile, the accurate characterization of cure kinetics behaviors was achieved in the thermochemical coupling analysis of CFRP, with the maximum values of the average and maximum temperature differences of 0.0176 °C and 0.2538 °C, respectively. Overall results show that our proposed incremental model is remarkably preferable to existing models and is beneficial in promoting the widespread reuse of the existing knowledge of cure kinetics behavior of resins in this domain.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.