Prediction of Compressive Strength of Sustainable Concrete using various Machine Learning Techniques

Abstract

Machine learning and its applications have come out as a rescue for solving real-life cumbersome tasks as they imitate the functioning of a brain in interpretation and decision making. One of its possible applications in Civil Engineering is the prediction of compressive strength of concrete as it is one of the most important mechanical properties. In this work, has made an attempt to carry out a survey which was used as a stepping stone to hop onto the implementation of the concept of ML in concrete world. The models were produced and tested using different algorithms for their capability to learn and predict compressive strength based on dataset provided by UCI using various algorithms like Linear Regression, Lasso Regression, Ridge Regression, Decision Tree Regression, Random Tree Regression, Multi-layer Perceptron, Support Vector Machine and K Nearest Neighbour. The results prove that these algorithms can be used as a great tool for predicting compressive strength, with the Random Forest Algorithm proving to be the most accurate for the considered dataset with an accuracy of 0.91 and an error of 3.47. The voluminous disposal of demolition waste into landfills is proving to be an owing cause for environmental pollution. An essential outlook for economic construction is the processing of the waste procured from construction works to preserve an ecological balance. To contribute towards research to generate sustainable reusable construction products, an attempt was made to analyse the feasibility of recycled coarse and fine aggregates (RCA) as a substitute to coarse aggregate. Experiments have been conducted out to develop sustainable concrete grade M30 conforming to Indian Standards. The partial replacement was carried out using C & D waste extracted from the nearby dump yard in the vicinity of Ahmedabad city and processed at a nearby recycling plant. A total of 15 batches were produced with a 10 % increment of RCA and RFA and manufactured sand, upto 50 %. Cementitious materials like GGBS, Fly Ash and Metakaolin were also incorporated to improve the strength. Recycled aggregates at 10 % dosage provide the maximum compressive strength of 34.27 N/mm2 . The models produced using the combined dataset in 2 different sets performed pretty well with an error of 4.13 and accuracy of 0.90. The results prove that the algorithms are not much affected by the inclusion of non-dimensional parameters. Eventually, the Web Application was developed using the Django framework to cater to the user a friendly interface that can be utilized to predict the compressive strength of concrete.