Effectiveness of Rubberized Concrete by Using Adaptive Network-Based Fuzzy Inference System

Abstract

The building industry is always on the lookout for new materials to make infrastructure more durable and stronger. Rubberized concrete has emerged as a promising substitute by partially replacing coarse aggregates with reclaimed rubber aggregate. However, it is tough to refine its mechanical characteristics of rubber shreds that meet the structural requirements of the material because the performance output cannot be predicted with absolute certainty. This research conveys the neural network-based approach in evaluating the efficiency of rubberized concrete, which primarily concerns its compressive strength, flexural strength, and modulus of elasticity. A sample set of rubberized concrete composition was considered while training the neural networks by eight membership functions, which include trimf, trapmf, gbellmf, gaussmf, gauss2mf, pimf, desigmf, and psigmf.  Out of eight membership , four member ship function  such as trimf, trapmf, gbellmf, gaussmf has been used. The inputs of the ANFIS consists of five elements Rubber Volume Fraction, fibre volume fraction, Coarse Aggregate, Binder Sand Ratio and W/C Ratio. The targets are: compressive strength, flexural strength and modulus of elasticity In this regard, the model will be well under the equilibrium strength and sustainability with appropriate ratios for structural applications. Findings indicate that neural network predictions are close to empirical data, making it a valid and useful tool for improving rubberized concrete characteristics. This technique encourages an affordable solution for designing eco-friendly concrete materials with special characteristics, helping balance improvement in the building sector. Experimental data and those predicted by ANFIS models show good convergence. Gauss membership function provides good accuracy compare to remaining membership function