Two-dimensional Hermitian numerical manifold method. (March 2020)

• Record Type: Journal Article
• Authors: Liu, Zhijun; Zhang, Peng; Sun, Cong; Liu, Feng
• Is Part Of: Computers & structures. Volume 229(2020)
• Publication Date: 2020-03
• Subjects: Numerical manifold method -- Hermitian numerical manifold method -- C1 continuity -- Convergence -- Accuracy -- Linear independence; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2019.106178
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: A new numerical manifold method is proposed based on the Hermitian interpolation and adoption of rectangular meshes. C 1 continuity is obtained in the proposed Hermitian numerical manifold method, and the convergence, accuracy and efficiency are considerably improved. Strains at mathematical nodes inside the physical domain are directly available without extra postprocessing. The linear dependence problems in most high-order numerical manifold methods are circumvented. Abstract: Numerous approaches have been proposed to enhance the accuracy and convergence of the numerical manifold method (NMM) in recent years, but most, if not all, of these approaches cannot ensure C 1 continuity. Hermitian interpolation is an effective approach for obtaining high-order approximations. However, the requirement of rectangular meshes hinders the application of this approach in the finite element method. Taking advantage of the freedom in meshing in NMM, Hermitian interpolation is incorporated into NMM to obtain the C 1 approximation. In contrast to the common high-order NMM, the Hermitian NMM (HNMM) improves the accuracy and convergence without causing the linear dependence problem. Moreover, the degrees of freedom (DOFs) of the mathematical nodes inside the physical domain have physical meanings, and the strains at nodes can be obtained directly without the need for extra postprocessing. The proposed HNMM is verified by solving numerous benchmark linear elastic problems, and the results are compared against those of linear and cubic Lagrangian NMMs. The numerical solutions for these examples confirm the remarkable superiority of the HNMM over the Lagrangian NMMs in terms of accuracy, convergence and efficiency.