1. Using machine learning and experimental study to correlate and predict accelerated aging with natural aging of GFRP composites in hygrothermal conditions.
- Author
-
Wang, Jing, Karimi, Sajjad, Zeinalzad, Parastoo, Zhang, Jingyu, and Gong, Zhi
- Subjects
- *
HYGROTHERMOELASTICITY , *ARTIFICIAL neural networks , *MACHINE learning , *REGRESSION analysis , *AGING , *FILAMENT winding - Abstract
Composite materials are widely used in various applications, but their mechanical and physical properties can be significantly influenced by environmental factors such as temperature, humidity, and UV radiation during service life. This study aims to investigate the accelerated aging behavior of glass/epoxy composites under hygrothermal conditions and compare the results with naturally aged samples to understand the reliability of these materials in harsh environments. NOL GFRP samples were fabricated using the filament winding process and subjected to accelerated aging for varying time periods (ranging from 100 to 1200 hours in 50-hour intervals) under hygrothermal conditions. Additionally, naturally aged samples over several years (1, 1.5, 2, 2.5, and 3 years) were compared. Tensile strength measurements were conducted to assess the mechanical properties of the composites. Machine learning models, including linear regression, polynomial regression, Artificial Neural Network (ANN), random forest regression, and Support Vector Regression (SVR), were utilized to predict natural aging times from accelerated aging data. Hygrothermal aging led to significant matrix deterioration and fiber exposure, resulting in a notable reduction in tensile strength. The study observed a 35.60 % reduction in strength in three-year naturally aged samples and a 37.57 % reduction in 1000-hour accelerated aged samples. Among the machine learning models, the random forest regressor demonstrated the best performance in predicting natural aging times across different accelerated aging periods, while SVR exhibited poorer performance. Polynomial regression and ANN models showed moderate predictive capabilities. This study contributes to a deeper understanding of the aging process of composite materials, providing insights into the reliability and durability of glass/epoxy composites in various applications. By employing machine learning models, the research offers a novel approach to predicting natural aging times based on accelerated aging data, which can be beneficial for optimizing the design and maintenance of composite structures to enhance their long-term performance and minimize the risk of failure in service. • Investigated the impact of hygrothermal conditions on the accelerated aging of glass/epoxy composites over various time intervals. • Compared accelerated aging results with natural aging samples spanning multiple years to assess composite degradation. • Identified significant reductions in tensile strength due to hygrothermal aging, highlighting matrix deterioration and fiber exposure. • Employed five machine learning models to predict natural aging times from accelerated aging data, with random forest regression yielding the most accurate results. • Provided valuable insights into the aging behavior of composite materials through correlation analysis between natural and accelerated aging. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF