Composites are widely used in aerospace structures, wind turbine blades, and automotive lightweight components due to their excellent strength-to-weight ratio. However, curing large composite parts requires careful control of both temperature and process planning. Limited through-thickness thermal conductivity and insulation from vacuum-bagging materials impede heat dissipation, leading to thermal degradation, cure shrinkage, distortion, and void formation. Reducing the heating rate can mitigate these issues but it leads to increased energy consumption and reduced overall productivity.
To address these challenges, composite cure optimization can be viewed as a three-stage workflow: (1) develop accurate cure kinetics models that capture exothermic behavior, (2) construct reliable thermochemical simulation frameworks to complement or replace extensive experimental trials, and (3) implement optimization strategies that design cure cycles robust to spatial temperature gradients and degree of cure variations.
Projects
Improving the Curing Efficiency of Composite Aerostructures Through a Physics-Informed Multiple Neural Network Based Optimization Framework
The goal of this project is to develop and validate a computationally efficient Physics-Informed Multiple Neural Network (PIMNN) modeling and optimization framework. This framework will be used to improve the efficiency of manufacturing composite aerostructures through optimization of the composite curing process inputs such as the temporal and spatial distribution of the air temperature profile to reduce the expended energy and cycle time while meeting the specified process requirements (e.g., degree of cure) that yield acceptable quality parts. Our group’s role is to characterize the material properties, develop the finite element models, and experimentally validate the optimized cure cycle generated by PIMNN framework.
Publications
Humfeld, Keith D., Geun Young Kim, Ji Ho Jeon, John Hoffman, Allison Brown, Jonathan Colton, Shreyes Melkote, and Vinh Nguyen. “Co-training of multiple neural networks for simultaneous optimization and training of physics-informed neural networks for composite curing.” Composites Part A: Applied Science and Manufacturing 193 (2025): 108820.
Geun Young Kim, Shreyes Melkote, and Jonathan Colton. “Ensemble cure kinetics network (ECK-Net): A method to derive cure kinetics of thermosetting resin.” Composites Part A: Applied Science and Manufacturing, (Accepted,11/2025).
Contact information
- Geun Young Kim: gkim439@gatech.edu