Performance evaluation of compressive strength of concrete using different machine learning algorithms

Abstract

Accurately predicting the compressive strength of concrete is crucial for ensuring structural integrity, optimizing material usage, and reducing construction costs. Conventional experimental methods, though reliable, are often labour-intensive and time-consuming. To address these limitations, this study investigates the effectiveness of machine learning (ML) algorithms as efficient alternatives for predicting concrete compressive strength. Four ML algorithms—Linear Regression (LR), Multilayer Perceptron (MLP), M5 Rule-Based Model, and Support Vector Machines (SVM)—were evaluated based on their predictive performance. A comprehensive dataset comprising 350 concrete samples was prepared, with compressive strength tests conducted in accordance with Indian standard 516. The models were trained on experimental data and were tested using varying data splits of 50%, 40%, 30%, 20%, and 10% to assess their prediction accuracy. Among the evaluated models, the MLP demonstrated superior performance, achieving a correlation coefficient (CC) of 0.98 with a 20% testing split, outperforming the other algorithms. To further validate the predictive capability of the MLP model, multiple linear regression analysis was employed, confirming its robustness and generalization ability. The findings underscore the potential of machine learning techniques, particularly the MLP model, in providing accurate, reliable, and time-efficient predictions of concrete compressive strength. This study contributes to the growing body of research focused on leveraging machine learning for enhanced decision-making in construction material design, ultimately promoting more sustainable and cost-effective construction practices.