Modeling compressive strength of EPS lightweight concrete using regression, neural network and ANFIS

EPS concrete is an especial type of lightweight concrete made by partial replacement of concrete’s stone aggregates with lightweight expanded polystyrene beads (EPSs). This type of concrete is very sensitive to its constituent materials which complicate the modeling process. Considering the involved complexities, this paper dealt with developing and comparing parametric regression, neural network (ANN) and adaptive network-based fuzzy inference system (ANFIS) models for predicting the compressive strength of EPS concrete for possible use in mix-design framework. The results emphasized that the elite ANN model constructed with two hidden layers and comprised of three neurons in each layers, could be effectively used for prediction purposes. Moreover, ANFIS elite model developed by bell-shaped membership function was recognized as a proper model to this means; however, its prediction performances were evaluated to be diluted than ANN model. On the other hand, the prediction results of second-order partial polynomial regression model as elite empirical one showed the weakness of this model comparing ANN and ANFIS models.

► For the first time, models developed for prediction of the strength properties of EPS concrete.
► Robust ANN and ANFIS models proposed for predicting the compressive strength of EPS concrete.
► The overall performance of trained ANN is more accurate than ANFIS model.
► Such robust models could be easily utilized for EPS concrete mix proportioning as a problem with high complexities included.
► Higher accuracy of neural network is due to application of Levenberg–Marquardt backpropagation algorithm.