Prediction of residual tensile strength of glass fiber reinforced polymer bars in harsh alkaline concrete environment using fuzzy metaheuristic models

Abstract

The long-term durability of glass fiber reinforced polymer (GFRP) bars in harsh alkaline environments is of great importance in engineering, which is reflected by the environmental reduction factor in various structural codes. The calculation of this factor requires robust models to predict the residual tensile strength of GFRP bars. Therefore, three robust metaheuristic algorithms, namely particle swarm optimization (PSO), genetic algorithm (GA), and support vector machine (SVM), were deployed in this study for achieving the best hyperparameters in the adaptive neuro-fuzzy inference system (ANFIS) in order to obtain more accurate prediction model. Various optimized models were developed to predict the tensile strength retention (TSR) of degraded GFRP rebars in typical alkaline environments (e.g., seawater sea sand concrete (SWSSC) environment in this study). The study also proposed more reliable model to predict the TSR of GFRP bars exposed to alkaline environmental conditions under accelerating laboratory aging. A total number of 715 experimental laboratory samples were collected in a form of extensive database to be trained. K-fold cross-validation was used to assess the reliability of the developed models by dividing the dataset into five equal folds. In order to analyze the efficiency of the metaheuristic algorithms, multiple statistical tests were performed. It was concluded that the ANFIS-SVM-based model is robust and accurate in predicting the TSR of conditioned GFRP bars. In the meantime, the ANFIS-PSO model also yielded reasonable results concerning the prediction of the tensile strength of GFRP bars in alkaline concrete environment. The sensitivity analysis revealed GFRP bar size, volume fraction of fibers, and pH of solution were the most influential parameters of TSR.