Gaoyuan Yu¹, Yunfeng Lou², Hang Dong³, Junjie Li¹, Xianlong Jin^1,*

CMES-Computer Modeling in Engineering & Sciences, Vol.136, No.1, pp. 135-155, 2023, DOI:10.32604/cmes.2023.025166 - 05 January 2023

Abstract Heterogeneous multicore clusters are becoming more popular for high-performance computing due to their great computing power and cost-to-performance effectiveness nowadays. Nevertheless, parallel efficiency degradation is still a problem in large-scale structural analysis based on heterogeneous multicore clusters. To solve it, a hybrid hierarchical parallel algorithm (HHPA) is proposed on the basis of the conventional domain decomposition algorithm (CDDA) and the parallel sparse solver. In this new algorithm, a three-layer parallelization of the computational procedure is introduced to enable the separation of the communication of inter-nodes, heterogeneous-core-groups (HCGs) and inside-heterogeneous-core-groups through mapping computing tasks to various… More >

M. Gomathy Nayagam^1,*, K. Ramar², K. Venkatesh³, S. P. Raja⁴

Intelligent Automation & Soft Computing, Vol.33, No.3, pp. 1485-1499, 2022, DOI:10.32604/iasc.2022.023953 - 24 March 2022

Abstract Moving object detection, classification and tracking are more crucial and challenging task in most of the computer vision and machine vision applications such as robot navigation, human behavior analysis, traffic flow analysis and etc. However, most of object detection and tracking algorithms are not suitable for real time processing and causes slower processing speed due to the processing and analyzing of high resolution video from high-end multiple cameras. It requires more computation and storage. To address the aforementioned problem, this paper proposes a way of parallel processing of temporal frame differencing algorithm for object detection More >

Yasunori Yusa^1,*, Tomoshi Miyamura², Tomonori Yamada³, Shinobu Yoshimura³

The International Conference on Computational & Experimental Engineering and Sciences, Vol.21, No.3, pp. 58-58, 2019, DOI:10.32604/icces.2019.05231

Abstract In a wind turbine blade, laminated plates consisting of fiber reinforced plastic materials are generally used due to its high specific strength. We have been developing a large-scale finite element solver to analyze the wind turbine blade structure. For such a structure, the laminated finite element is frequently used in modeling. Each laminated finite element has multiple layers, each of which is an orthotropic body in order to model the layered fiber reinforced materials with different fiber directions. Also, since a realistic wind turbine blade structure generally requires a large number of finite elements for… More >

Mizuma Takehito^1,*, Takei Amane¹

CMES-Computer Modeling in Engineering & Sciences, Vol.116, No.3, pp. 349-363, 2018, DOI:10.31614/cmes.2018.01714

Abstract A matrix equation solved in an eddy current analysis, A-ϕ method based on a domain decomposition method becomes a complex symmetric system. In general, iterative method is used as the solver. Convergence of iterative method in an interface problem is improved by increasing an accuracy of a solution of an iterative method of a subdomain problem. However, it is difficult to improve the convergence by using a small convergence criterion in the subdomain problem. Therefore, authors propose a method to introduce double-double precision into the interface problem and the subdomain problem. This proposed method improves More >

C.M.T. Tien¹, N. Pham-Sy¹, N. Mai-Duy¹, C.-D. Tran¹, T. Tran-Cong¹

CMES-Computer Modeling in Engineering & Sciences, Vol.104, No.4, pp. 251-304, 2015, DOI:10.3970/cmes.2015.104.251

Abstract This paper presents a high-order coupled compact integrated RBF (CC IRBF) approximation based domain decomposition (DD) algorithm for the discretisation of second-order differential problems. Several Schwarz DD algorithms, including one-level additive/ multiplicative and two-level additive/ multiplicative/ hybrid, are employed. The CCIRBF based DD algorithms are analysed with different mesh sizes, numbers of subdomains and overlap sizes for Poisson problems. Our convergence analysis shows that the CCIRBF two-level multiplicative version is the most effective algorithm among various schemes employed here. Especially, the present CCIRBF two-level method converges quite rapidly even when the domain is divided into… More >

N. Pham-Sy¹, C.-D. Tran¹, N. Mai-Duy¹, T. Tran-Cong¹

CMES-Computer Modeling in Engineering & Sciences, Vol.100, No.5, pp. 363-397, 2014, DOI:10.3970/cmes.2014.100.363

Abstract In this paper, a high performance computing method based on the Integrated Radial Basis Function (IRBF), Control Volume (CV) and Domain Decomposition technique for solving Partial Differential Equations is presented. The goal is to develop an efficient parallel algorithm based on the Compact Local IRBF method using the CV approach, especially for problems with non-rectangular domain. The results showed that the goal is achieved as the computational efficiency is quite significant. For the case of square lid driven cavity problem with Renoylds number 100, super-linear speed-up is also achieved. The parallel algorithm is implemented in More >

N. Pham-Sy¹, C.-D. Tran¹, T.-T. Hoang-Trieu¹, N. Mai-Duy¹, T. Tran-Cong¹

CMES-Computer Modeling in Engineering & Sciences, Vol.92, No.1, pp. 1-31, 2013, DOI:10.3970/cmes.2013.092.001

Abstract Compact Local Integrated Radial Basis Function (CLIRBF) methods based on Cartesian grids can be effective numerical methods for solving partial differential equations (PDEs) for fluid flow problems. The combination of the domain decomposition method and function approximation using CLIRBF methods yields an effective coarse-grained parallel processing approach. This approach has enabled not only each sub-domain in the original analysis domain to be discretised by a separate CLIRBF network but also compact local stencils to be independently treated. The present algorithm, namely parallel CLIRBF, achieves higher throughput in solving large scale problems by, firstly, parallel processing More >

Yan Gu¹, Wen Chen^1,2, Xiao-Qiao He³

CMC-Computers, Materials & Continua, Vol.29, No.2, pp. 129-154, 2012, DOI:10.3970/cmc.2012.029.129

Abstract This paper applies an improved singular boundary method (SBM) in conjunction with domain decomposition technique to stress analysis of layered elastic materials. For problems under consideration, the interface continuity conditions are approximated in the same manner as the boundary conditions. The multi-layered coating system is decomposed into multiple subdomains in terms of each layer, in which the solution is approximated separately by the SBM representation. The singular boundary method is a recent meshless boundary collocation method, in which the origin intensity factor plays a key role for its accuracy and efficiency. This study also introduces More >

Ju’e Yang¹, Hongying Huang², Dehao Yu³

CMES-Computer Modeling in Engineering & Sciences, Vol.85, No.4, pp. 347-366, 2012, DOI:10.3970/cmes.2012.085.347

Abstract In this paper, a new domain decomposition method is suggested for the stationary Stokes equations on unbounded domain and its convergence is proved. We draw an artificial boundary to make the domain into two parts: one is bounded, in which we use the mixed finite element method; the other is unbounded, in which we apply the natural boundary reduction. Then we change the sub-problem on the unbounded domain onto a one in a bounded domain and we use the Dirichlet to Neumann(DtN) alternating algorithm to solve the resulting mixed system. The theoretical results as well More >

D. Ngo-Cong^1,2, N. Mai-Duy¹, W. Karunasena², T. Tran-Cong^1,3

CMES-Computer Modeling in Engineering & Sciences, Vol.83, No.3, pp. 311-352, 2012, DOI:10.3970/cmes.2012.083.311

Abstract In this study, local moving least square - one dimensional integrated radial basis function network (LMLS-1D-IRBFN) method is presented and demonstrated with the solution of time-dependent problems such as Burgers' equation, unsteady flow past a square cylinder in a horizontal channel and unsteady flow past a circular cylinder. The present method makes use of the partition of unity concept to combine the moving least square (MLS) and one-dimensional integrated radial basis function network (1D-IRBFN) techniques in a new approach. This approach offers the same order of accuracy as its global counterpart, the 1D-IRBFN method, while More >