A comparative investigation using machine learning methods for concrete compressive strength estimation

Concrete compressive strength plays an important role in determining the mechanical properties of concrete. The determination of concrete compressive strength requires lengthy laboratory tests. The ability to predict concrete compressive strength with advanced machine learning algorithms speeds up these long experimental processes and reduces costs at the same time. In this study, using the compressive strength data of concrete samples cured for 7 and 28 days, concrete compressive strength was compared using Artificial Neural Network (ANN), Decision Tree (DT), Support Vector Machine (SVM) and Linear Regression (LR) algorithms. The research sought to determine the algorithm with the most successful performance. In the study, the input data were taken as the unit weight, water content, Schmidt hammer, ultrasonic pulse velocity and relative humidity of the hardened concrete, and the output parameter to be determined was concrete compressive strength. In the analyses, the best correlation coefficient (R-2) was 0.86, and the best mean absolute error was 2.59 using the DT algorithm. The data in the analyses with the best success were obtained from concrete samples cured for 28 days. As a result, it was determined that the DT algorithm had the least amount of error and is thus the most suitable for use in concrete compressive strength estimation.

Automated summary

The study compared the compressive strength of concrete samples cured for 7 and 28 days using different algorithms and found that the Decision Tree algorithm had the highest success rate and the lowest mean absolute error of 2.59.