Prediction of the rubberized concrete behavior: A comparison of gene expression programming and response surface method

Abstract

The use of rubber in concrete to partially substitute mineral aggregates is an effort to decrease the global amount of scrap tires. This study investigates the behavior of rubberized concrete (RC) with various replacement ratios (0–50%) by volume and replacement type (fine, coarse, and fine-coarse) using soft computing techniques. The uniaxial compressive strength (CS), elastic modulus (EM), and ductility (D) are measured, and the effect of rubber content and the rubber aggregate type on the properties of RC is investigated. Scanning electron microscopy and X-ray diffraction analyses are made to determine its microstructural and chemical composition. This article compares the efficiency of two RC models using a recently developed artificial intelligence technique, i.e., gene expression programming (GEP) and conventional technique, i.e., response surface method (RSM). Statistical models are developed to predict the CS, TS, EM, and D. The mathematical models are validated using determination coefficient (R 2) and adjusted coefficient (R 2adj), and they are found to be significant. Furthermore, both methods (i.e., RSM and GEP) are very well correlated with the experimental data. The GEP is found to be more effective at predicting the experimental test results for RC. The projected methods can be executed for any practical value of RC.

Keywords

• rubberized concrete
• gene expression programming
• response surface method
• statistical
• compressive strength
• ductility