About this Abstract |
Meeting |
2024 ASC Technical Conference, US-Japan Joint Symposium, D30 Meeting
|
Symposium
|
2024 ASC Technical Conference, US-Japan Joint Symposium, D30 Meeting
|
Presentation Title |
Thermal modeling for autoclave curing of composite structures: From high-fidelity thermal CFD to physics-informed machine learning |
Author(s) |
Ze Zhao, Yao Sun, Kalyan Shrestha, Dianyun Zhang, Jim Lua, Jinhui Yan |
On-Site Speaker (Planned) |
Jinhui Yan |
Abstract Scope |
Autoclave curing is widely used in composite manufacturing. As a heat treatment technique, modeling and controlling temperature in autoclave processes plays a significant role in mitigating the residual stress and distortion of cured composite structures. However, due to the complexity of curing processes involving coupled aerodynamics, heat transfer, and chemical reactions, it is challenging to obtain the spatiotemporal full-field temperature in cured structures. In this paper, we aim to accurately predict the full-field temperature directly from autoclave loading conditions without utilizing empirical parameters. Firstly, a high-fidelity model coupling thermodynamics, computational fluid dynamics (CFD), and curing kinetics (thermal-CFD) is developed to simulate the aerodynamics and temperature distribution for the entire autoclave system, including both circulating air, tools, and composite structures. The high-fidelity thermal-CFD model is digitized using an immersed boundary method, allowing simulations of structures with arbitrary orientation without invoking sophisticated meshing processes. We perform experiments to validate the high-fidelity model in terms of temperature and degree of curing. We show that the high-fidelity model can accurately predict the temperature distribution in the cured structure and capture the discrepancy between the actual temperature history on the structure and the desired temperature cycle. Through the numerical experiments, we find that modeling curing kinetics is important in autoclave thermal modeling. Without using an appropriate curing model and a careful coupling with the thermal-CFD model, significant errors in temperature prediction can be introduced. Finally, leveraging the high-fidelity simulation and experimental data, a physics-informed machine learning-based reduced-order model is developed for rapid full-field temperature construction. |
Proceedings Inclusion? |
Definite: Post-meeting proceedings |