Hybrid Particle Swarm Optimization — Artificial Neural Network: Strength Prediction of Bambusa Blumeana Fiber and Recycled Coarse Aggregate for Multi-generation Recycled Aggregate Concrete

Abstract

Several studies on recycled aggregate concrete (RAC) have revealed percentage reductions in compressive strength (CS) and split tensile strength (STS). The addition of bamboo fiber, specifically Bambusa blumeana J.A. and J.H. Schultes, improves the mechanical properties of concrete. By combining these two materials, this paper prepared 175 concrete mixtures that produced 1050 specimens from the zeroth to third recycling generations of concrete. These specimens underwent CS and STS tests, and the findings were used to develop a hybrid Artificial Neural Networks and Particle Swarm Optimization (ANN-PSO) model. The results showed that the R values for the developed models for CS and STS were approaching 1 and the Mean Squared Error (MSE) were approaching 0 which were the ideal values. The findings suggest that the ANN-PSO model outperformed the Multiple Linear Regression (MLR) model as evident to the higher R value and lower Mean Absolute Percentage Error (MAPE) of the ANN-PSO models for CS and STS prediction. Moreover, using Garson’s algorithm (GA), the relative importance (RI) of the predictors used in the CS and STS models were calculated. The results showed that slump is the most influential parameter to the CS while that for the STS is the bamboo fiber % (BF%). This paper elucidates the potential of utilizing BF in RAC and the effectiveness of the ANN-PSO model in predicting CS and STS.