Estimating compressive strength of concrete using multiple regressions based on principal component analysis

Abstract

In this paper, experimentally generated data on four variables, namely, water-cementations ratio, fine aggregate-cementitious ratio, coarse aggregate-cementitious ratio and cementitious content has been employed to estimate compressive strength of concrete using Ordinary Least Square (OLS) regression and Principal Component Regression (PCR) techniques. Separate models have been developed to estimate compressive strength of concrete for fly ash replacements (0 and 15 percent), zones of aggregates (A, B and C) and curing ages (28, 56 and 91 days). In each case, predictors are highly correlated to each other indicating the presence of multicollinearity in the data. Comparison of OLS regression models and PCR regression models reveal that the PCR technique successfully tackles the problem of multicollinearity with a slight reduction in the predictive capability of the models.