COMPARATIVE STUDY OF CONCRETE COMPRESSIVE STRENGTH USING POTABLE AND RIVER WATER WITH MULTIPLE LINEAR REGRESSION PREDICTION

Abstract

Concrete strength development depends significantly on water quality, yet in many regions, reliance on potable water (PW) in construction is both unsustainable and costly. This study assesses the viability of river water (RW) as a substitute in concrete mixing and presents a predictive model utilizing Multiple Linear Regression (MLR). Testing was conducted on 36 concrete samples with 3 different mix ratios (1:2:4, 1:1.5:3, and 1:1:2) and water–cement ratios of 0.55, 0.60, and 0.65. The findings indicated that RW attained compressive strength with the 1:2:4 mix at W/C 0.55, satisfying the 15 MPa specification for structural concrete. A two-sample t-test (p = 0.1147) indicated no statistically significant difference between RW and PW specimens, confirming RW’s potential as an alternative mixing water. The MLR model, developed exclusively from Cavite RW data, demonstrated strong performance (R² = 0.877, RMSE = 1.57 MPa) and identified cement content and water–cement ratio as dominant strength predictors. While its predictive power is limited to the tested RW source, the framework demonstrates transferability to other contexts through recalibration. This study is novel in bridging empirical RW testing with statistical prediction, offering both sustainability insights and a decision-support tool for concrete mix design.