Prediction of ANN, MLR, and NLR models for Compressive strength performance in fly ash based self compacting concrete

Abstract

Self-compacting concrete (SCC) blended with fly ash (FA) presents a promising low-carbon alternative to traditional concrete, enhancing both workability and long-term durability. Yet, the prediction of its compressive strength (CS) remains challenging due to complex mix interactions. This study presents a comparative modeling framework using Multi-Linear Regression (MLR), Nonlinear Regression (NLR), and Artificial Neural Networks (ANN) to estimate the CS of FA-modified SCC based on key input variables: cement (C), water-to-binder ratio (w/b), fly ash content (FA), sand (S), coarse aggregate (CA), and superplasticizer (SPA dataset of 270 mixes was statistically analyzed, divided into 70% training and 30% testing subsets, and validated using R², RMSE, and MAE. The results revealed that the ANN model outperformed both NLR and MLR, achieving superior accuracy (R² = 0.95, RMSE = 3.49 MPa, MAE = 2.45 MPa) and consistent residual behavior within (± 20%) tolerance bands. In contrast, the NLR and MLR models exhibited broader error ranges and lower predictive reliability. The ANN’s adaptability to nonlinear, multivariate Furthermore, residual error analysis and model robustness across low, medium, and high-strength ranges were evaluated. These findings demonstrate the usefulness of data driven advanced models to resolve the complexities in the modern cementitious materials and thus serve as scientific basis for the improvement of the design of SCC with high performance and high eco-efficiency.