Estimating the Strength of Bi-Axially Loaded Track and Channel Cold Formed Composite Column using Different AI-Based Symbolic Regression Techniques

Abstract

In this research work, the strength of bi-axially loaded track and channel cold formed composite column has been estimated by applying three AI-based symbolic regression techniques namely “Genetic Programming (GP)”, “Evolutionary Polynomial Regression (EPR)” and “Group Method of Data Handling Neural Network (GMDH-NN)”. The collected numerically generated data entries containing global slenderness ratio (Column height / minor radius of gyration) (λ), local slenderness ratio of channel (bolts spacing S2 / channel thickness) (λc), local slenderness ratio of track (bolts spacing S1 / track thickness) (λt), relative eccentricity in the major direction (ex/D) and the relative eccentricity in the minor direction (ey/B) as the independent parameters and the normalized average normal stress at failure (Ult. load /Area) / yield stress (F/Fy) as the dependent parameter. The results of the models were validated using the R2, MAE and RMSE metrics. Both correlation and sensitivity analysis showed that the global slenderness ratio (λ) has the main influence on the strength, then the relative eccentricities (ex/D, ey/B) and finally the local slenderness ratios (λc, λt). Comparing predicted and calculated strengths showed that the three developed predictive models have the same level of accuracy (94%) with (R2 > 0.965), (MAE < 0.03) and (RMSE < 0.03).