Prediction of Compressive Strength of Concrete Using Machine Learning Techniques

Abstract

Compressive strength of concrete is an important parameter for designing any concrete structure. Compressive strength of concrete is a complex nonlinear function of its ingredients. Prediction of Concrete compressive strength plays a vital role in pre design phases of the structure and quality control of construction. The conventional methods of compressive strength determination are time consuming, so the use of data mining methods to predict the strength beforehand is helpful. Data mining techniques of Artificial Neural Networks (ANN), Support Vector Machine (SVM) have been used by the researchers for predicting strength of concrete. This paper discusses the use of tree based models like Random Forest (RF) and Model Trees (MT) for prediction of concrete strengths at various ages of concrete from 3 to 365 days. The model tree techniques give regression equations that are transparent as they give insight about the processes as compared to ANN. The paper also explores application of models to study the significance of various constituents of concrete using sensitivity analysis. Two ways to determine strength of concrete at various ages has been studied. In first method models have been proposed for different ages of concrete from 3 to 365 days. MT and RF have predicted compressive strength of concrete satisfactorily.
Keywords: Concrete compressive strength, model tree, random forest, prediction