Improving analytical models of circular concrete columns with genetic programming polynomials

This study improves weighted genetic programming and uses proposed novel genetic programming polynomials (GPP) for accurate prediction and visible formulas/polynomials. Representing confined compressive strength and strain of circular concrete columns in meaningful representations makes parameter studies, sensitivity analysis, and application of pruning techniques easy. Furthermore, the proposed GPP is utilized to improve existing analytical models of circular concrete columns. Analytical results demonstrate that the GPP performs well in prediction accuracy and provides simple polynomials as well. Three identified parameters improve the analytical models—the lateral steel ratio improves both compressive strength and strain of the target models of circular concrete columns; compressive strength of unconfined concrete specimen improves the strength equation; and tie spacing improves the strain equation.