Gene expression programming based formulations for splitting tensile strength of concrete

In this study, gene expression programming (GEP) formulations for splitting tensile strength (fspt) of the cylinder specimens with 150 mm diameter and 300 mm height using compressive strength (fc) of concrete cube specimens with 150 mm dimension are developed. Two models, called as GEP-I and GEP-II, are developed for predicting fspt by using GEP. The database used in the GEP models is based on experimental data obtained from literature. In the GEP-I model, while fc is used as input variable, fspt is used as output variable. However, in the GEP-II model, as well as fc, water–binder ratio (WB) is used as input variables. The data sets used in training and testing stages are randomly selected among all experimental data. The GEP formulations are also validated with additional experimental data other than the data used in training and testing sets of the GEP models. Experimental fspt results of concrete specimens are compared with GEP formulations, proposed formulations by some national building codes and the developed regression-based formulation results. The results show that GEP formulations have strong potential as a feasible tool for prediction of the fspt from only 150 mm cube fc or WB and 150 mm cube fc of concrete.

► GEP formulations for fspt using only cube fc or WB and cube fc of concrete are developed.
► Experimental fspt results of concrete are compared with results of GEP and regression formulations.
► The results are also compared with the results of formulations proposed by some building codes.
► The results of GEP formulations are closer to experimental results according to the other results for predicting fspt.