Evaluating hydrothermal degradation of GFRP composites under sustained loading using explainable machine learning

The understanding of GFRP composites under hydrothermal conditions and sustained loading offers valuable insights into their performance in challenging environments such as coastal areas, deep-sea structures, and environmentally friendly and long-lasting infrastructure solutions. This study examined the mechanical response of GFRP composites subjected to synergic sustained loading and hydrothermal degradation. A prediction application based on the XGBoost machine learning model was developed to estimate the residual mechanical response. The developed model was used to calculate the conversion factor accounting for moisture and temperature-based degradation in FRP composites. The SHAP analysis corroborated the experimental findings such that GFRP-based composites experience an initial rapid decline in mechanical properties when exposed to harsh environments, followed by a slower degradation rate over time. The pultruded vinyl ester-based GFRP composites depict less degradation than polyester-based composites and composites made via vacuum infusion. It was inferred that sustained loading below 30% has no negative impact on the mechanical characteristics of hydrothermal-aged GFRP composites. The degradation became worse for the sustained loading beyond 30% of the ultimate strength of the GFRP composite. Comments are also made on the current recommendations in technical specifications by the European Committee for Standardization CEN/TS 19101 related to moisture and temperature conversion factors. The current work is limited to the mechanical investigation of GFRP composites subjected to hydrothermal degradation. It needs to be extended to other composites such as CFRP and BFRP.