Novel metaheuristic-based type-2 fuzzy inference system for predicting the compressive strength of recycled aggregate concrete

Abstract

The use of recycled concrete aggregate (RCA) in the production of new concrete can provide several environmental benefits and reduce the pressure on natural resources. Due to the differences in the properties of RCA and natural aggregate, the properties of recycled aggregate concrete (RAC) (i.e., concrete containing RCA) are different from those of normal concrete. Compressive strength (CS) of the RAC, which is a key parameter in many design codes, can be determined either through expensive and time-taking laboratory-based procedures or using empirical relationships. In this study, two types of hybridized machine learning algorithms (i.e., type-1 fuzzy inference system (T1FIS) and interval type-2fuzzy inference system (IT2FIS)) were used to develop predictive models for the CS of RAC. Moreover, arithmetic optimization algorithm (AOA), as a novel optimization algorithm, was employed to optimize the parameters of FIS models. To develop the CS predictive models, a dataset containing information on 1868 data records was used. The results indicate that the IT2FIS model outperformed the T1FIS model. A comparison of the results of this study with those reported in previous studies also confirms the high accuracy of the IT2FIS model. The findings indicate that concrete age, natural fine aggregate to total natural aggregate ratio, and superplasticizer to binder ratio have positive impacts on the CS, while the remaining input variables have negative influences on the CS. Regarding the intensities of the variables, concrete age, total coarse aggregate to cement ratio, and water to binder ratio are in the first to third orders, respectively.