Multiscale modeling for accurate forecasting of concrete wear depth: a comprehensive study on mixture proportions and environmental factors

Accurately forecasting the depth of wear (Dw) of concrete is essential for ensuring the durability, safety, and efficiency of concrete structures while also promoting sustainability and cost-efficiency in construction projects. This study utilized systematic multiscale models to predict the concrete depth of wear, analyzing 391 samples with varying cement content (107–398 kg/m³), water-to-binder ratio (0.31–0.37), class C fly ash content (0–316 kg/m³), fine aggregate (607–705 kg/m³), coarse aggregate (1099–1266 kg/m³), plasticizer (2.6–2.9 l/m³), air entraining agent (0.3–1.4 l/m³), curing time (28–365 days), and testing time (5–60 min). Linear (LR), pure quadratic (PQ), interaction (IN), and M5P-tree models were used to identify key parameters affecting Dw. The models accurately estimated Dw about mixture proportions based on metrics such as R², MAE, RMSE, OBJ, SI, and a-20 index. The results indicated that concrete with up to 30% fly ash exhibited wear depth similar to conventional concrete, but abrasion resistance decreased slightly when fly ash content exceeded 30%.