Strength prediction of mortar reinforced with babassu coconut fiber using artificial neural networks

Abstract

Using natural fibers as reinforcement in cement composites is a sustainable alternative in the construction industry. Nevertheless, the dosage of these materials employing traditional methods depends on laboratory tests and tends to be laborious and expensive. Thus, predicting composites characteristics can save time and reduce operating costs. In this study, Artificial Neural Networks (ANN) trained with Extreme Learning Machine (ELM) were used to predict the tensile and compressive strengths of mortar reinforced with babassu coconut fiber. Babassu coconut is an abundant product in the region where this research was carried out, and its use tends to bring socio-economic benefits. The data was obtained experimentally, generating a total of 51 samples. The ANN topologies have six parameters in the input layer (cement, sand, water/cement ratio, maximum fiber length, fiber percentage, and slump), one parameter in the output layer (tensile or compressive strength), and one hidden layer, which contains 6 and 7 neurons, respectively, in the tensile (ELM-T) and compressive (ELM-C) strength prediction models. Simulation results indicated that both models are promising tools for predicting the mechanical behavior of mortars reinforced with babassu coconut fibers.