Machine learning-assisted predictive modeling and hybrid optimization of buckling characteristics in porous laminated composite plates on nonlinear elastic foundations

Abstract

The paper presents a predictive modelling and optimization framework for analysing the buckling response of porous laminated composite plates resting on nonlinear Kerr-type elastic foundations. Higher-Order Shear Deformation Theory is utilized to accurately capture transverse shear deformation without requiring shear correction factors. A generic imperfection model is incorporated into the plate formulation to simulate both global and localized geometric imperfections. A comprehensive parametric study is conducted to examine the influence of ply-angle configurations, imperfection amplitudes, porosity gradients, and nonlinear foundation stiffness parameters on the critical buckling load. To improve computational efficiency and prediction accuracy, a Stacked Ensemble Surrogate Modelling approach is employed. It combines Gaussian Process Regression, Tri-layer Neural Networks, and Least Squares Support Vector Machines for robust and accurate response prediction. An integrated hybrid optimization strategy is implemented by coupling Teaching-Learning-Based Optimization, GPR-based surrogate models, and Sequential Quadratic Programming, with further enhancement using the parameter-less Rao algorithm. The proposed framework enables accurate, efficient design of porous composite plates by capturing the combined effects of geometric attributes, geometric imperfections, and nonlinear elastic foundation behaviour on buckling performance.