Developing explicit design equations for hygrothermal durability of pultruded FRP composites using physics-guided symbolic regression

Pultruded fiber reinforced polymer (PFRP) composites are increasingly used in civil engineering. However, their long-term performance under hygrothermal aging remains challenging to predict quantitatively. Existing empirical methods lack physical interpretability, while physics-based models are often too complex for practical design. Although machine learning offers high accuracy, its “black-box” nature limits adoption in codified standards. To bridge this gap, this work employs physics-guided symbolic regression (SR) to develop explicit, interpretable design equations for predicting the hygrothermal durability of PFRP composites. Using a comprehensive database and embedding physical constraints, SR derives closed-form formulas for residual mechanical properties as functions of initial properties, aging duration, aging temperature, and material thickness. A unified mathematical framework is established, where the property-specific exponent α quantifies degradation sensitivity. Two model forms are derived: a best-fit model for accuracy (α₁) and a unified design model (α₂ = 1.33 for strengths and 1.22 for moduli) that simplifies practical application. Validation against natural weathering data confirms the reliability of the proposed equations, providing readily codifiable design tools for durable PFRP structures.