Computational Testing of a Unidirectional Carbon Fiber Composite: Micromechanical Simulations and Machine Learning Approach
Abstract
We present a computational method that incorporates micromechanical modeling through representative volume element theory simulations of a unidirectional carbon fiber / polymer resin composite to produce a dataset of 625 mechanical testing simulations. The focus was to first lay down the groundwork for subsequent, more sophisticated versions of the model; as such, effects such as temperature and humidity, and variable process parameters are not considered. Good agreement with room temperature, dry carbon fiber composite experimental data is found. Many constituent properties influence the final material properties of composite materials. An ML regression method that is especially useful for estimating feature importance in small datasets [1, 2] is applied to both a large dataset (N = 625 test points) and a smaller subset of this dataset (N = 30 test points) that is more representative of a dataset that could be generated through real-world experimental testing. A quantitative understanding of the influence of constituent material properties on the composite material response is achieved for both datasets. Through our empirical approach, the fiber volume fraction and fiber modulus emerge as important parameters in determining the Youngs modulus of the composite along the fiber direction, as anticipated from composite theory.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 18, 2024
- Accession Number
- AD1219143
Entities
People
- Anthony S. Tai
- Phillip H. Burnside
- Yannic J. Gagnon
Organizations
- Naval Surface Warfare Center