Experimental investigation and SVM-based prediction of compressive and splitting tensile strength of ceramic waste aggregate concrete

Abstract

Due to the brittle nature of ceramic, the ceramic and construction industry produces a large volume of waste that imposes a severe environmental threat due to its non-biodegradability. In this study, the suitability of ceramic waste as a replacement of natural coarse and fine aggregate in concrete has been investigated by evaluating engineering properties such as bulk density, water absorption, workability, etc. with respect to different concrete samples made with different mix proportions. Furthermore, a prediction model is introduced to predict compressive and splitting tensile strength using the machine learning tool support vector machine (SVM). A data set containing 108 records either for compressive or tensile strength was used for the training and testing purposes of the SVM model. A total of 9 mix proportions was selected and cast cylinders were cured for 7, 28, and 56 days. Four different kernel functions were used to optimize the results and different accuracy parameters like the value of R², mean absolute error, mean square error, root mean square error, etc. were compared to find the best kernel function for this study. By primarily evaluating the coefficient of determination (R²), SVM showed an acceptable result with an accuracy of over 90%. Moreover, in terms of other accuracy measurement parameters result indicates that the SVM is an effective tool to predict the compressive and splitting tensile strength of concrete comprised of different proportions of ceramic content.