Publication

Abstract : In recent years machine learning is considered as a highly effective and widely used tool to predict the behaviour of complex and heterogeneous problems. In this paper, the behaviour of superplasticised cement paste is assessed by XGBoost, which is accepted to accomplish the state-of-the-art results on many machine learning challenges. The data required for the development of model is formulated experimentally by conducting rheological tests on cement pastes using a temperature controlled Coaxial Cylinder Viscometer. Various parameters like amount of cement, superplasticiser, water and test temperature are taken as input parameters and the behaviour is assessed by taking rheological characterises like yield stress and plastic viscosity as output parameters. Out of the 252 data formulated experimentally 85 % are used for training and the remaining is for testing the efficacy of the network. From the results it is observed that the model developed using XGBoost is a promising tool for the solution of highly complex and heterogeneous civil engineering problems, which otherwise is highly tedious and time consuming in nature.