Abstract
This chapter describes the design of a new kind of high-performance machine-learning model for predicting compressive strength (CS) of CFRP-CC (Carbon Fiber-Reinforced Polymer Confined-Concrete) specimens. This model allows for a remarkable reduction of construction waste, labor cost, and project duration. In modeling the case of predicting CS of concrete confined with CFRP under the tube and wrap confinement types, the capability of ANN (Artificial Neural Nets) has already been well-proven in the literature. The best optimization configuration of ANN strongly relies on the designer's intuition and experience rather than on a systematic procedure. Thus, the method is presented by integrating a metaheuristic algorithm with a specially tuned ANN. The proposed method optimizes the feed-forward backpropagation network through the SPO (Stochastic Paint Optimizer) algorithm in conjunction with ANN. Experimental results are assessed using R-squared, MSE, RMSE, and MAPE metrics. The resultant refined SPO-ANN model yields an R-squared of 0.92. This model can predict compressive strength more accurately, which reduces the time and cost associated with excessive lab tests.