Enhancing predictive modeling of nano metal matrix composites with LEO-HDNN approach

This paper proposed an enhanced predictive modeling approach for nanometal matrix composites utilizing the Hamiltonian Deep Neural Network (HDNN) integrated with the lotus effect optimization (LEO) algorithm, referred to as the LEO-HDNN method. The objective was to optimize process parameters in the AA2219 composite reinforced with Al₂O₃ and Si₃N₄ to improve mechanical properties. The stirring squeeze casting method’s process parameters were optimized using the LEO methodology, while the HDNN predicted optimal parameters. The performance of the hybrid methodology was assessed on the MATLAB platform and contrasted with several existing methodologies, including heap based optimization (HBO), salp swarm algorithm (SSA), grasshopper optimization algorithm (GOA) and fire hawk optimizer-spiking neural network (FHO-SNN).Outcome indicated that the proposed technique exhibited high wear resistance, achieving values of 1.4 at a load of 4 N and reaching 2.9 at 16 N. The method demonstrated superior mechanical properties, with a tensile strength of 192 MPa and a hardness value of 77 HRB, significantly surpassing existing techniques: FHO-SNN (187 MPa, 72 HRB), GOA (182 MPa, 65 HRB), SSA (177 MPa, 55 HRB), and HBO (167 MPa, 48 HRB). These findings underscored the effectiveness of the LEO-HDNN method in improving the nanocomposites’ mechanical properties.