Modelling the compressive strength of high-performance concrete containing metakaolin using distinctive statistical techniques

Abstract

Cement consumption is rapidly increasing around the world causing environmental concerns over higher CO2 emissions. Metakaolin obtained through the calcination of kaolinite clay is an eco-friendly substitution for clinker in portland cement production. The raw materials required to produce metakaolin are easily accessible and cheap. Addition of metakaolin in concrete alters mechanical properties, notably compressive strength. Prior knowledge of the strength characteristics of concrete play a vital role in determining the design aspects of the buildings. Robust prediction models that can effectively predict the compressive strength of pozzolan-based concrete are highly essential in the field of concrete technology in order to save time and reduce laboratory testing samples. In this regard, A total of 152 experimental datasets from the literature were collected and evaluated. Influencing parameters such as binder content (385–600 kg/m3), water to cementitious material ratio (0.25–0.45), metakaolin replacement levels (0%–50%) and SiO 2/Al2O3 ratio in the cement paste (1.05–12) were chosen as independent variables. The present study was aimed at developing prediction models using different statistical techniques such as Linear regression (LR), Multi-logistic regression (MLR), Nonlinear regression (NLR) and Response surface methodology (RSM) to predict the 28-day compressive strength of High-performance concrete containing metakaolin. Also, the performance of the proposed models was examined using different statistical metrics such as coefficient of correlation (R), root mean squared error (RMSE), mean absolute error (MAE) and scatter index (SI). The quadratic RSM model outperforms all the proposed models in the present study with R, RMSE, MAE and SI of 0.92, 6.5 MPa, 5.2 MPa and 0.08. The models were evaluated in strength ranges other than HPC, and the QRSM model showed a maximum residual error of ±20%. Furthermore, the results of sensitivity analysis indicate that binder content, w/b ratio and SiO2/Al2O3 are the most influential parameters controlling the compressive strength of concrete.