Full-cycle prediction of crack healing in self-healing concrete using generalized polynomial chaos expansion

The crack healing capacity of self-healing concrete is crucial for enhancing structural durability, especially in aggressive environments where the dynamic progression of healing depth directly influences service life. This study introduces a modeling and prediction approach based on Polynomial Chaos Expansion (PCE) to quantitatively assess the crack cross-sectional repair rate throughout the full healing cycle. A foundational database is first established by statistically identifying key factors governing healing behavior. A first-order PCE surrogate model is developed to characterize the temporal evolution from early-stage variability to nonlinear saturation. Dimensionality reduction combined with order elevation enhances accuracy under limited data conditions. To overcome the constraints of conventional Hermite polynomials bound by Gaussian assumptions, a generalized PCE framework accommodating arbitrary distributions is formulated, enabling broad applicability across healing scenarios. Extrapolative validation on unmodeled healing ages confirms the model’s robustness and reliability throughout all healing stages. This work provides a reliable quantitative framework for predicting service life and optimizing repair strategies in engineering practice. Changhao Fu and colleagues developed a novel modeling framework based on Polynomial Chaos Expansion for predicting the complete process of crack healing in self-healing concrete. The research establishes a multidimensional model that correlates five key healing indicators with the crack cross-sectional repair rate.