Modelling Mechanical Properties of Self-Compacting Rice Husk Ash Concrete Using Artificial Neural Network

Abstract

This study investigates the effects of Rice Husk Ash (RHA) on the mechanical properties and workability of Self-Compacting Concrete (SCC), focusing on compressive and flexural strengths. The mixes incorporated Hydroplast 260GR, a high-range water-reducing admixture (super-plasticizer), at dosages of up to 2.0% of the cement weight to achieve the high flowability required for self-compacting concrete (SCC) while minimizing overall water demand. Using a dataset of 56 concrete mix proportions developed using simplex lattice design method (N[6,3]) to systematically access the effect of varying Rice Hush Ash (RHA) content on the investigated performance parameters which include compressive strength, flexural strength and workability. The experimental analysis revealed a decline in compressive strength from 20 MPa to 11 MPa and flexural strength from 3.0 MPa to 2.2 MPa as RHA increased, though low RHA levels (5%–10%) occasionally matched the control due to pozzolanic effects. Workability remained stable (slump values 550–650 mm) up to 20% RHA, indicating minimal impact on flowability. Predictive models, including Linear Regression (LR) and Artificial Neural Network (ANN), were developed to estimate mechanical properties, with the dataset split into 80% training and 20% testing. The LR model outperformed the ANN, achieving RMSE values of 0.103 for flexural strength and 1.240 for compressive strength, compared to the ANN’s 0.374 (R² -0.89) and 2.074 (R² 0.09), respectively, highlighting the LR’s suitability for RHA-SCC predictions given the dataset’s linear tendencies and size constraints. Correlation analysis showed strong relationships between cement, water, and RHA, with RHA negatively affecting strength (r ≈ -0.48 to -0.51). The study recommends limiting RHA to 5–10% for practical applications, refining the ANN with a larger dataset, and exploring non-linear models to improve compressive strength predictions, providing a foundation for sustainable SCC design.