Predicting mechanical properties of CFRP composites using data-driven models with comparative analysis.

Area of Science:

• Materials Science
• Polymer Science
• Machine Learning Applications

Background:

• Carbon fiber reinforced polymer (CFRP) composites are vital in engineering due to their high strength-to-weight ratio.
• Predicting mechanical properties of CFRPs is crucial for optimizing their application and design.
• Current prediction methods can be time-consuming and resource-intensive.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting CFRP mechanical properties.
• To identify key factors influencing these properties, including carbon nanotube (CNT) content and manufacturing parameters.
• To assess the efficacy of ridge regression, random forest, and support vector regression models.

Main Methods:

• Sixty-two distinct CFRP samples were designed and manufactured.
• Experimental testing was performed to obtain mechanical property data.
• Ridge regression, random forest, and support vector regression models were trained and compared.

Main Results:

• High prediction accuracy achieved for flexural strength (R2 = 0.966), flexural modulus (R2 = 0.871), and mode-II energy release rate (R2 = 0.903).
• Machine learning models effectively correlated input parameters with mechanical performance.
• The models demonstrated robust predictive capabilities across different CFRP types.

Conclusions:

• Machine learning offers a powerful, data-driven approach for predicting CFRP mechanical properties.
• This methodology can significantly reduce the reliance on extensive experimental characterization.
• The findings facilitate more efficient material design and development of advanced composites.