Prediction on compressive strength of Engineered Cementitious composites using Machine learning approach

Abstract

In recent times, Machine Learning Techniques are emerging continuously as an affordable and efficient in predicting how the property of the materials influences the properties, cost and time of the proposed mixes. In Civil engineering domain, the utilization of ML techniques are need to be strengthened with respect to the incorporation of Supplementary Cementitious Materials (SCM) to the conventional proportioning of mix. Hence to increase the sustainability and thereby reduce the environmental pollution that occurs due to disposal of the Industrial wastes which also holds the pozzolanic property. In this study one of the ML techniques namely Artificial Neural Network (ANN) are used to determine the 28 days compressive strength of the Engineered Cementitious Composite (ECC) incorporated with industrial pozzolans like Fly ash and Ground Granulated Blast Slag (GGBS) with respect to mix proportions and physical properties of polyvinyl alcohol fibres (PVA). The physical properties of fibre and mix proportions such as the ratio of cement, fly ash or GGBS, fine aggregate, water to binder (W/B) ratio, high range water reducer to binder (HRWR/B) ratio, length of fibres, diameter, tensile strength, density, modulus of elasticity and elongation were used to predict the compressive strength of ECC. Furthermore, the experimental investigations were conducted on the compressive strength of ECC with fly ash and GGBS separately and also verified with the ANN outputs. Hence, the present model (Levenberg-Marquardt algorithm) is validated by considering the standard benchmark with respect to the coefficient of determination (R2) which is highly correlated.