A kinematically consistent second-order computational homogenisation framework for thick shell models

Abstract

This paper presents a kinematically consistent second-order computational homogenisation scheme for shear deformable shells. The proposed framework can accurately evaluate the membrane, bending, and transverse shear components of the shell resultants and tangent operators, whilst showing no size dependency on the fine scale model. To date, a proper extension of second-order homogenisation to a thick shell model, such as the 5-parameter formulation, remains non-trivial due to the difficulties in projecting the macroscopic transverse shear strains to the fine scale whilst satisfying the stress boundary conditions on the top and bottom faces. To overcome this, the paper proposes a novel volumetric constraint on the fluctuation moment field, that is used in conjunction with a set of constraints obtained through an orthogonality condition. The result is a consistent downscaling procedure that can properly downscale all the macroscopic shell strains. More notably, a pure transverse shear deformation can be achieved, thus producing a uniform parabolic shear stress distribution in homogeneous materials. The key equations for upscaling are obtained through the modified Hill–Mandel condition. In particular, the shell tangent operators are derived in closed form as a function of the Taylor upper bound and softening terms coming from the fluctuation matrices. The proposed framework is general and can incorporate full geometric and material nonlinearities across all the length scales of interest. Through a series of benchmarks, it is demonstrated that the constitutive tangents computed through the present framework correspond well with analytical solutions. Finally, excellent agreements in model responses, including through-thickness stress distributions are found between the multi-scale (FE2) and the full-scale models in the numerical benchmarks, featuring the nonlinear loading of thin and thick heterogeneous panels.