Using the two optimization algorithms (BBO and FDA) coupling with radial basis neural network to estimate the compressive strength of high-ultra-performance concrete
Affiliations:
School of Civil Architecture and Environment, Hubei University of Technology, Wuhan, Hubei, China
Correspondence:
[*]
Corresponding author. Mengmeng Wu, School of Civil Architecture and Environment, Hubei University of Technology, Wuhan 430068, Hubei, China. E-mail: [email protected].
Abstract: Using Ultra-High Performance Concrete (UHPC) as the highly resistant material is widely advised in constructing sensitive structures to enhance safety. The utilization of eco-friendly contents such as fly-ash and silica-fume replacing cement can decrease the pollution rate in the production process of concrete and improve the compressive strength (CS) factor. There are many ways to appraise the CS of concretes as empirically and mathematically Artificial Neural Networks (ANN) as the high-accurate model is used in the present study. In this regard, Radial Basis Function (RBF) coupling with Biogeography-Based Optimization (BBO) and Flow Direction Algorithm (FDA) created the two high-accurate frameworks: BBO-RBF and FDA-RBF. Enhancing the accuracy of RBF to predict the CS and decreasing the ANN net complexity leads to having better results evaluated by various metrics. Therefore, using the proposed frameworks, the correlation index of R2 to model the CS in the training phase for FDA-RBF was calculated at 0.9, although BBO-RBF could get 0.85, with a 0.5% difference. However, the RMSE of FDA-RBF was 9 MPa, and for BBO-RBF, this index was calculated at 10 MPa the former model has about three percent more accuracy than the latter.
