Optimization of the hydroforming process of triple-layer sandwich cross-shaped tubes using genetic algorithm

Abstract

The hydroforming process enables the production of lightweight, high-strength tubular components with complex geometries, yet the nonlinear deformation behavior of three-layer steel–aluminum–steel sandwich tubes makes the process highly susceptible to localized thinning and ductile fracture. In this study, a comprehensive computational framework is developed to optimize the hydroforming of a multilayer cross-shaped tube by integrating finite element simulation, a fully connected neural network, and a genetic algorithm. The novelty of this work lies in the intelligent coupling of FEM-based process modeling with ANN-driven surrogate prediction and evolutionary optimization, enabling the simultaneous adjustment of steel and aluminum layer thicknesses together with the internal hydraulic pressure. This hybrid FEM–ANN–GA methodology introduces a unique optimization strategy that has not been previously applied to multilayer hydroforming systems. Using FEM results as training data, the neural network successfully captured the nonlinear relationship between the design parameters and the ductile fracture index, which was selected as the primary objective due to its strong correlation with failure initiation. The optimized configuration—0.22-mm steel layer, 0.77-mm aluminum layer, and 5 MPa internal pressure—reduced the ductile fracture index from approximately 0.35 to 0.09. The optimized simulation confirmed complete and defect-free forming with uniform stress distribution and no tearing. Overall, the proposed framework demonstrates an effective and computationally efficient approach for enhancing formability and structural reliability in complex multilayer hydroforming applications. So the integration of FEM– ANN–GA for three-layer hydroforming optimization is the main novelty of present research