PREDICTION OF COMPRESSIVE STRENGHT OF CONCRETE WITH RICE HUSK ASH (RHA) AS PARTIAL REPLACEMENT OF CEMENT USING NON-LINEAR REGRESSION MODEL.

ABSTRACT This study investigates the effect of Rice Husk Ash (RHA) as a partial replacement of cement on the compressive strength of concrete and evaluates three predictive models—Spline Regression, Polynomial Regression, and a Physics-Based model. Experimental samples with RHA replacement levels of 2.5%, 5%, 12%, and 15% were tested at curing ages ranging from 7 to 28 days. Results showed that the incorporation of RHA enhanced strength at lower replacement levels due to improved pozzolanic activity, while higher percentages slightly reduced strength because of increased water demand and reduced density. Predictive modelling revealed that the Spline model performed best, achieving an R-squared value of 0.8310, followed closely by the Polynomial model with the same R-squared but higher error values. The Physics-Based model recorded a lower R-squared of 0.5946, indicating weaker predictive accuracy. The Spline model also achieved the lowest Mean Absolute Error (6.0425) and MAPE (18.52%). The findings show that RHA can significantly improve concrete strength when used in appropriate proportions and that the Spline model provides the most reliable prediction technique for compressive strength assessment.