Metaheuristic Based AI Approach Towards Strength Prediction of Strain Hardening Cementitious Composites (SHCC)

Abstract

This research aims to create new empirical prediction models to assess the mechanical properties of strain hardening cementitious composites (SHCCs), to avoid the costly, and hectic experimental procedures that need enough time and require skilled investigators. Soft computing method adopted in this research is gene expression programming (GEP). This study compute five outputs, i.e., compressive strength (CS), first crack tensile stress (TS), first crack flexural stress (FS), first crack tensile strain (TST), and first crack flexural strain (FST). Wide-ranging records were considered from available literature with twelve parameters selected as the predictor variables. Important inputs of the study were cement percent weight (C%), fine aggregate percent weight (Fagg%), fly ash percent weight (FA%), water to binder ratio (W/B), super plasticizer percent weight (SP%), fiber amount percent weight (Fib%), length to diameter ratio (L/D), fiber tensile strength (FTS), fiber elastic modulus (FEM), environment temperature (ET), and curing time (CT). Correlation coefficient (R), and regression coefficient (R2) were used in the deduction of the model’s performance. In addition to this, the performance of the models was also established using relative root mean square error (RRMSE), mean absolute error (MAE), root squared error (RSE), root mean square error (RMSE), objective function (OBF), performance index (PI) and Nash-Sutcliffe efficiency (NSE). The resulting mathematical GP-based equations disclose the originality of GEP model. In addition to this, these equations are easy to understand and consistent. The objective function and performance index are also in accordance with the literature references. Consequently, all the proposed AI approaches has high generalization. The sensitivity analysis showed cement percentage, fine aggregate percentage and environmental temperature to be the most sensitive and significant variables for all the five models developed (CS, TS, FS, TST and FST). The result of this research can assist researchers, practitioners and designers in freely assessing SHCC; consequently, limiting environmental exposures. It will lead to sustainable, faster and safer construction from environment-friendly waste management point of view.