1. Introduction
The fast multipole method (FMM) [1,2] is an efficient tool for reducing the computational time and memory requirement in the boundary integral equation (BLE) method, thus making possible scientific and engineering computations of large scale problems. An important study on FMM is improving the computational efficiency. One efficient way to improve the computational efficiency of the FMM is reducing the computational cost of the M2L translations. A new version of the FMM for Laplace problems, which can improve the efficiency of the FMM evidently, was proposed by Greengard and Rokhlin [3]. Then diagonal formed FMM for Helmholtz problems was presented [4]. Chen et al. [5] employed the FMM to accelerate the construction of influence matrix in the dual boundary element methods (DBEM). This separable technique promotes the efficiency in determining the influence coefficients. An adaptive FMM for 3D potential problems, which is several times faster than the non-adaptive one, is proposed by shen [6]. Bapat [7] proposed an adaptive tree structures based on a new definition of the interaction list. Some researchers also investigated variable orders of multipole expansions [8,9]. These methods can provide a significant improvement on the efficiency of the FMM. A tree data structure, which is more flexible matching the geometry, is proposed by Zhang [9]. In the tree data structure, the number of expansion terms in the multipole to local (M2L) translations can be determined according to the distance between the two interactions boxes in three dimensional potential problems [9]. The authors have demonstrated that the proposed adaptive tree structure with an adaptive selection of the expansion order could evidently improve the computational efficiency.
1. บทนำThe fast multipole method (FMM) [1,2] is an efficient tool for reducing the computational time and memory requirement in the boundary integral equation (BLE) method, thus making possible scientific and engineering computations of large scale problems. An important study on FMM is improving the computational efficiency. One efficient way to improve the computational efficiency of the FMM is reducing the computational cost of the M2L translations. A new version of the FMM for Laplace problems, which can improve the efficiency of the FMM evidently, was proposed by Greengard and Rokhlin [3]. Then diagonal formed FMM for Helmholtz problems was presented [4]. Chen et al. [5] employed the FMM to accelerate the construction of influence matrix in the dual boundary element methods (DBEM). This separable technique promotes the efficiency in determining the influence coefficients. An adaptive FMM for 3D potential problems, which is several times faster than the non-adaptive one, is proposed by shen [6]. Bapat [7] proposed an adaptive tree structures based on a new definition of the interaction list. Some researchers also investigated variable orders of multipole expansions [8,9]. These methods can provide a significant improvement on the efficiency of the FMM. A tree data structure, which is more flexible matching the geometry, is proposed by Zhang [9]. In the tree data structure, the number of expansion terms in the multipole to local (M2L) translations can be determined according to the distance between the two interactions boxes in three dimensional potential problems [9]. The authors have demonstrated that the proposed adaptive tree structure with an adaptive selection of the expansion order could evidently improve the computational efficiency.
การแปล กรุณารอสักครู่..