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 |
Predicting effective properties of unidirectional non-crimp fabric composites with manufacturing-induced defects using a multiscale ANN model |
Author(s) |
Yu Zeng, Khizar Rouf, John Montesano |
On-Site Speaker (Planned) |
Yu Zeng |
Abstract Scope |
Fiber-reinforced plastic (FRP) composite materials, incorporating non-crimp fabrics (NCFs), are increasingly considered for electric vehicle structures requiring energy absorption during impact events. However, microstructure variations and manufacturing-induced defects in NCF-FRP composite materials pose challenges in simulating and optimizing their properties. To address this, robust tools that capture critical manufacturing-induced defects are needed to predict mechanical properties. This study proposes a multiscale finite element (FE) modeling approach to predict in-tow (micro) and lamina-level (meso) effective properties of NCF-FRP composites, with use of artificial neural networks (ANNs) models to reduce prediction time. Coupled microscale and mesoscale FE models of FRP, incorporating manufacturing-induced defects within representative volume elements (RVEs), were developed to generate reliable training and testing data for ANN models based on constituent properties, geometric parameters, and defect parameters. Micro and meso-scale ANN models were then developed, trained, and tested to predict relations between inputs (constituent properties and material structure parameters) and outputs (in-tow and lamina-level effective properties). Trained ANN models significantly reduced estimation time from hours in multiscale FE modeling to seconds with good accuracy. |
Proceedings Inclusion? |
Definite: Post-meeting proceedings |