Significance of Eco-Cement Constituents to Its Mechanical Properties, by Machine Learning Algorithms

Abstract

A novel proposed material for the potential replacement of cement in some of its applications was evaluated. This new material labeled Eco-Cement comprised of Biomass from the dairy and poultry industries, Urea, Cement Kiln Dust, Rice Husk Ash, Sand, and Water and was manufactured in a range of weight ratios. In this work, a comprehensive analysis of the ingredients in varying weight percentages of the novel material was manufactured and the corresponding strength and strains of the material were studied. A variety of concrete pastes using amounts of sought ratios were produced, molded into blocks, and allowed to cure under laboratory conditions. Unconfined compression tests were performed using a deformation control compressive strength machine. The strength and strains were evaluated from the initial zero load step incrementally, until the failure of each specimen. The resulting database was analyzed by utilizing Linear Regression, Random Forests, and the Gradient Boosting machine learning methods. Extensive sensitivity analysis with the machine learning algorithms, reveal certain patterns, which were established with three different methods. Furthermore, we present the analysis of the corresponding literature with Bibliometric techniques.