Management of dynamic stability in Cu-Ni carbon composite structures via physics-informed neural network modeling

Dynamic stability management within multifunctional composite systems is vital for the development and structural reliability of engineering applications. The study focuses on Cu–Ni carbon composite structure management with an integrated framework of micromechanical modeling, higher-order shear deformation theory, and physics-informed neural networks (PINNs). Effective material properties are modeled by modified Halpin–Tsai models, allowing improved management of constituent interactions between Cu–Ni matrix and the carbon reinforcements. The equations of motion are formulated in accordance with Hamilton's principle and Hooke's law, establishing and maintaining a consistent variational formulation with three independent components for displacement. It is recognized that as substructural interactions are more effectively managed, the elastic foundation can consist of Winkler's and Pasternak's coefficients, incorporating both normal and shear-layer contributions. Higher-order shear deformation theory is applied to properly characterize the stress–strain state during representation, eliminating the need for shear correction factors, permitting better predictive management of moderately thick plates. A PINN-based solution procedure is developed in which the governing partial-differential equations, along with the boundary values called upon during learning, are embedded within the learning process. The machine learning framework allows efficient use of resources with the potential for more robust accuracy in predicting stability boundaries, critical buckling loads, and vibration responses. The comparison studies show that the proposed procedure offers advantages over an existing and historical finite element model. The results of the studies also illustrated that PINNs offered more effective predictive management of composite dynamic stability and represented a hybrid of material modeling, structural theory, and machine learning. Hence, this work contributes to the continuing advancements of materials development by providing a promising platform for the next generation of multi-functional composites.