Shape optimization of Hencky-type Reddy columns against buckling using genetic algorithms and artificial neural networks

The current investigation proposes a computationally efficient framework for optimizing the buckling performance of non-uniform beams by means of an enhanced Hencky bar-chain model (HBM) combined with Reddy’s higher-order beam theory. The model effectively captures shear deformation, making it appropriate for analyzing thick beams where classical models may fall short. To lessen the high computational cost typically associated with iterative buckling analyses, a surrogate model is developed utilizing artificial neural networks (ANNs), trained to forecast critical buckling loads with high precision. The trained ANN is then integrated into a genetic algorithm framework for both single- and multi-objective optimization, which leads to considerable acceleration in the design process. The approach is tested across different boundary conditions, including simply supported, clamped-free, clamped-sliding, and clamped-clamped cases. The obtained results demonstrate that the ANN-based predictions closely match those acquired from high-fidelity eigenvalue solvers while decreasing computational time. In general, the integration of the HBM with machine learning-assisted evolutionary algorithms leads to a dependable approach for the shape optimization of beams under buckling constraints. Such a framework provides precise outcomes with decreased computational effort and can be considered a solid foundation for future developments in structural stability.