Buckling of sandwich panels with transversely flexible core: Correction of the equivalent single-layer model using thick-faces effect

Abstract

Modeling a periodic structure as a homogeneous continuum allows for an effective structural analysis. This approach represents a sandwich panel as a two-dimensional plate of equivalent stiffness. Known as the equivalent single-layer, the method is used here to analyze bifurcation buckling of three types of sandwich panels with unidirectional stiffeners in the core: truss-core, web-core and corrugated-core panels made of an isotropic material. The transverse shear stiffnesses of these panels can differ by several orders of magnitude, which cause incorrect buckling analysis when using the equivalent single-layer model with the first-order shear deformation theory. Analytical solution of the problem predicts critical buckling loads that feature infinite number of half-waves in the direction perpendicular to the stiffeners. Finite element model also predicts buckling modes that have non-physical, saw-tooth shape with infinite curvature at nodes. However, such unrealistic behavior is not observed when using detailed three-dimensional finite element models. The error in the prediction of the critical buckling load is up to 85% for the cases considered here. The correction of the equivalent single-layer model is proposed by modeling the thick-faces effect to ensure finite curvature. This is performed in the finite element setting by introducing an additional plate with tied deflections to the equivalent single-layer plate. The extra plate is represented with bending and transverse shear stiffness of the face plates. As a result, global buckling is predicted accurately. Guidelines are proposed to identify the sandwich panels where ordinary model is incorrect. Truss-core and web-core sandwich panels need the correction. Corrugated-core panels without a gap between plates in the core have smaller shear orthotropy and do not need the correction. Modeling the thick-faces effect ensures correct results for all cases considered in this study, and thus one should resort to this approach in case of uncertainty whether the ordinary equivalent single-layer model is valid.