Symbolic regression model for predicting the effective bending stiffness of composite GLB-concrete beams

Engineered bamboo products have stood out over the past decades as sustainable and high-performance alternatives to conventional construction materials. Among them, glued laminated bamboo (GLB) presents high specific strength and stiffness, being particularly promising for lightweight structural applications. However, in elements of greater slenderness or in applications involving considerable spans, GLB beams may exhibit flexural stiffness inferior to that of hybrid systems, which limits their use in certain design situations. The association between GLB and concrete in composite beams (CGC) has proven to be an efficient strategy for increasing the global stiffness and the load-bearing capacity of the system. However, traditional analytical methods, such as the γ-Method, present limitations in accuracy for this type of system, due to simplifications and the need for empirical calibration. To overcome this gap, after applying a three-dimensional finite element simulation model for the representation of the mechanical performance of an experimentally tested CGC beam (validation of the numerical model against experimental data reported in related research), a comprehensive parametric study was developed numerically with the objective of determining an equation, obtained through symbolic regression (SR - based on evolutionary algorithms), for estimating the flexural stiffness of CGC beams, as well as evaluating the significance of the varied parameters in the estimation of this property. The results obtained enabled the conclusion that the thickness of the concrete slab and the height of the GLB beam are the dominant parameters, and that the SR model for estimating the flexural stiffness of CGC beams presented excellent predictive performance (R 2 adj = 0.95; MAE = 3.26 MNm 2 ; RMSE = 4.27 MNm 2 ; MAPE = 8.63%), significantly outperforming the γ-Method.