“…Recently, Song et al and collaborators proposed the polygonal SBFEM 11,28,29 and the quadtree SBFEM, [30][31][32] in which the advantages of SBFEM in building element and conventional FEM framework is combined, and a series of successfully applications were achieved. However, when SBFEM works in FEM form, the computational cost on eigenvalue solution could greatly increase as the increase of the number of elements, particularly for the nonlinear problems, [33][34][35] large-scale problems or the problem with the reconstruction of stiffness matrix, such as uncertainty analysis, 19,[36][37][38][39] optimization, [40][41][42] inverse problem solving, 21,43 the computational expense may become a bottleneck.…”