Eric Schwabe Valerie, Eric J. Schwabe, Valerie E. Taylor et al. · 1995
This paper provides a novel insight into this tradeoff by providing analytical equations that can serve as quantitative guidelines for optimizing this tradeoff for parallel finite element applications. These equations include the following machine and problem parameters: the number of elements in the global domain (N ), the number of processors to be used for execution (P ), the ratio of message startup time to network bandwidth for the target machine (ff=fi), and the number of bytes per shared node communicated between subdomains (fl).
Cited by 0
Mincheol Jo, Woobin Park, Moonseong Kim et al. · IEEE Access · 2024
When addressing large-scale electromagnetic problems using the finite element method (FEM), the resulting matrices are typically sparse, necessitating numerous sparse matrix-vector multiplication (SpMV) operations. To handle this efficiently, research has focused on leveraging large-scale parallel processing with graphics processing units (GPUs). These GPUs can be controlled directly using NVIDIA’s Compute Unified Device Architecture (CUDA). In this paper, we analyze electromagnetic scattering for dielectric and dielectric-coated structures using iterative solvers with FEM. To accelerate the handling of the large-scale matrices generated during this process, we employ compressed sparse row (CSR) format, various preconditioners, and CUDA-based GPU parallelization. We verify the accuracy of our results by comparing them with those obtained using the commercial electromagnetic software High Frequency Structure Simulator (HFSS) and our custom-developed MATLAB-based FEM code. Performance improvements are assessed by comparing these results with those from MATLAB’s backslash direct solver under single-core processing conditions.
Cited by 3Open access
Amane Takei, Kouhei Murotani, Shin-ichiro Sugimoto et al. · Journal of Advanced Simulation in Science and Engineering · 2014
This paper describes large-scale full-wave analyses of electromagnetic fields using numerical human body models. This is achieved by the use of the parallel finite element method with iterative domain decomposition. Numerical examples of FDTD analysis using numerical human model by a High Performance Computing (HPC) are shown in previous researches. However, numerical examples and performance evaluation of the finite element method hardly exists. Therefore, we perform a feasibility study and a detailed performance evaluation of the high-frequency electromagnetic field analysis using the numerical human model on a super computer constituting Japan’s High Performance Computing Infrastructure (HPCI) as a numerical simulation infrastructure so as to extract problems for advancement of the future analysis. The numerical human body models provided by the National Institute of Information and Communications Technology (NICT) in Japan are composed of voxel data with all sides of 2mm and include skin layers, blood vessels, bones, internal organs, etc. distinguished by material properties. By using HPC, the user can evaluate the electromagnetic filed distribution inside the whole body model.
Cited by 6Open access
Rao, Rama Mohan A, Rao, Appa TVSR, Dattaguru, B · NOT FOUND REPOSITORY (Indian Institute of Science Bangalore) · 2003
Cited by 0
Yi Hu, Hong Lin Cao, Qiang Liu et al. · Lecture notes in civil engineering · 2023
Abstract As an important component of the tunnel support structure, the effective length and compactness of the anchors have a direct impact on the load-bearing capacity and safety. Anchor reinforcement is characterised by rapid construction and concealment. Traditional tension tests are difficult to meet the current requirements for quality control in engineering construction. In recent years, elastic wave anchor non-destructive technology has gradually emerged and achieved certain results, but still cannot solve the identification of anchor lengths in long anchor rods and complex surrounding rock conditions. In this study, the theory of parallel seismic methods is applied to the verification of anchor length by drawing on the technical means of pile foundation detection. The feasibility of the implementation of the method is verified by simulating the anchor-anchor solid-surrounding rock elastic wave conduction system using ABAQUS to analyse the anchor length detection process.
Cited by 0Open access
Jerry W. Rouse, Timothy Walsh, Garth Reese · The Journal of the Acoustical Society of America · 2011
This talk shall present an overview of SALINAS, a massively parallel finite element code for structural dynamics and acoustics analysis that is being developed at Sandia National Laboratories. SALINAS allows for prediction of both the time and frequency domain responses of complex structural, acoustic, and fully coupled structural acoustic systems having millions of degrees of freedom. An overview of SALINAS capabilities shall be presented including development history, solver and element types, quadratic eigenanalysis and frequency response, direct frequency response, nonlinear acoustics, implicit transient dynamic analysis, and infinite elements with focus given to structural acoustics capabilities. The application of SALINAS to structural acoustics problems shall also be presented as well as future directions of research for the development of the code.
Cited by 9
문종근 · Seoul National University Open Repository (Seoul National University) · 2011
Cited by 0
Toshimitsu FUJISAWA, Genki YAGAWA · The proceedings of the JSME annual meeting · 2001
A parallel distribution technique of computing loads for compressible flow analysis is presented. In the present method, the optimal assignments of computing loads are obtained in a parallel way on any number of processors, by employing a graph partitioning library, parMETIS. The method is applied to a model of wind tunnel with a step, and it is confirmed that the number of 'independent nodes' in the NBN-FEM becomes substantially large by the parallel renumbering operations.
Cited by 0Open access
Kang Zhou, Zhirong Wu, Majid Cashany · 2016
Finite Element (FE) model calibration is an essential but very computationally-intensive task for real world civil structures, e.g., bridges, buildings and dams etc., which are usually large in size and represented by thousands of elements for achieving good analysis results. Although performing one analysis run of such a large FE model may take less than a minute with a well-developed FE model solver, but it is preventive to embed such a full-analysis FE solver into an iterative process for model calibration, which often requires for tens or even hundreds of thousands of FE model analysis runs. Therefore, it is of great importance and interest in developing an effective and efficient approach for FE model calibration of large structure systems. In this paper, an effective framework for the accelerated FE model calibration, based on authors’ previous work in FE model calibration, has been presented by integrating the substructure analysis or Component Modal Synthesis (CMS) with parallel genetic algorithm optimization provided by Darwin optimization framework. It enables the parallel solution evaluation processes on a many-core machine or in distributed manner across a cluster of computers. This is a powerful computation framework that has been tested on a FE model calibration of a highway bridge approach span in New York City. This study has proved that the developed tool is effectively tailored for the applications of many real world structures including but not limited to bridges. A calibrated FE model can well present the structural conditions and can be used for condition-based maintenance and facilitate structural life-cycle asset performance management.
Cited by 0
Jerry W. Rouse · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 2016
Abstract not provided.
Cited by 0Open access
Shang-hsien Hsieh, Glaucio H. Paulinos, John et al. · 1994
This paper studies and compares the domain partitioning algorithms presented by Farhat,1 Al-Nasra and Nguyen,2 Malone,3 and Simon4/Hsieh et al.5,6 for load balancing in parallel finite element analysis. Both the strengths and weaknesses of these algorithms are discussed. Some possible improvements to the partitioning algorithms are also suggested and studied. A new approach for evaluating domain partitioning algorithms is described. Direct numerical comparisons among the considered partitioning algorithms are then conducted using this suggested approach with both regular and irregular finite element meshes of different order and dimensionality. The test problems used in the comparative studies along with the results obtained provide a set of benchmark examples for other researchers to evaluate both new and existing partitioning algorithms. In addition, interactive graphics tools used in this work to facilitate the evaluation and comparative studies are presented. KEY WORDS: algorithms; domain partitioning; load balancing; mesh partitioning; parallel finite element analysis; partitioning algorithms
Cited by 0
Jingli Du, Chuanzhen Cui, Hong Bao et al. · Journal of Computational and Nonlinear Dynamics · 2014
Cable-driven parallel manipulator (CDPM) is a good solution to achieving large workspace. However, unavoidable vibrations of long cables can dramatically degrade the positioning performance in large workspace applications. Most work so far on cable-driven parallel manipulators (CDPMs) simply neglected the dynamics of the cables themselves. In this paper dynamic modeling of large CDPMs is addressed using a variable domain finite element method (FEM). A cable element with variable length is derived using the absolute nodal coordinate formulation to facilitate motion analysis of CDPMs. The effects of cable length variation and the resulting mass variation are also considered. Based on this element dynamics model of CDPMs can be readily obtained using the standard assembling operation in the FEM. Numerical results showed that the effect of the derivatives of cable length variation and that of the mass variation are trivial.
Cited by 38
Tomoshi Miyamura, Hirohisa Noguchi, Ryuji Shioya et al. · TRANSACTIONS OF THE JAPAN SOCIETY OF MECHANICAL ENGINEERS Series A · 1999
The hierarchical domain decomposition method (HDDM) is a solver for large scale algebraic equations in the finite element method. It is suitable for various kinds of parallel computers such as massively parallel processors (MPP) and workstation/PC clusters. In this study, the original HDDM is first modified to improve computation speed, and then applied to static elastic-plastic finite element analyses. Some key techniques employed in the static nonlinear finite element analysis are presented. As illustrative examples, the parallel elastic-plastic analyses are performed on a nuclear structure with 1.3 millions degrees of freedom and a cube with ten millions degrees of freedom. It is shown that both the conjugate gradient method used in the HDDM and the Newton Raphson method are successfully converged, respectively, in the two examples.
Cited by 5Open access
Ammar T. Al Sayegh, Elisa D. Sotelino · International Journal of Earthquake and Impact Engineering · 2020
A parallel scheme is devised to efficiently parallelise all steps of parallel finite element analysis in this study. In addition, this scheme is based on a row-wise matrix distribution. A new row-wise parallel finite element analysis algorithm that exploits the nature of distributed compressed row sparse matrices and multivectors to improve concurrency is developed. A new dynamic load balancing technique has also been devised. The dynamic load balancing technique has been designed specifically to balance the computational workload among processors suitable for the analysis of nonlinear structures. This new algorithm has been implemented in ParaStruc, which is a parallel structural analysis system. Trilinos, a set of parallel numerical libraries developed by researchers in the Sandia National Laboratory has been used to build this algorithm. ParaStruc is a lightweight fully parallelised parallel finite element analysis system, which contains only three classes and a pre-processor. It is shown that this approach produces superior performance in terms of speedup, efficiency, and isoefficiency in the analysis of nonlinear structure response ranges when compared to parallel ABAQUS. The performance and efficiency of this algorithm has been verified with numerical simulations of a 200-metre 50-story 10-frame 10-bay 3D structure subjected to various load levels.
Cited by 3
A. Rama Mohan Rao, T. V. S. R. Appa Rao, B. Dattaguru · Computers & Structures · 2003
Cited by 22
B. Patzák, D. Rypl, Jaroslav Kruis · Computational Technology Reviews · 2013
Cited by 0
Ammar T. Al Sayegh, Elisa D. Sotelino · International Journal of Earthquake and Impact Engineering · 2020
A parallel scheme is devised to efficiently parallelise all steps of parallel finite element analysis in this study. In addition, this scheme is based on a row-wise matrix distribution. A new row-wise parallel finite element analysis algorithm that exploits the nature of distributed compressed row sparse matrices and multivectors to improve concurrency is developed. A new dynamic load balancing technique has also been devised. The dynamic load balancing technique has been designed specifically to balance the computational workload among processors suitable for the analysis of nonlinear structures. This new algorithm has been implemented in ParaStruc, which is a parallel structural analysis system. Trilinos, a set of parallel numerical libraries developed by researchers in the Sandia National Laboratory has been used to build this algorithm. ParaStruc is a lightweight fully parallelised parallel finite element analysis system, which contains only three classes and a pre-processor. It is shown that this approach produces superior performance in terms of speedup, efficiency, and isoefficiency in the analysis of nonlinear structure response ranges when compared to parallel ABAQUS. The performance and efficiency of this algorithm has been verified with numerical simulations of a 200-metre 50-story 10-frame 10-bay 3D structure subjected to various load levels.
Cited by 1
James G. Malone, Nancy L. Johnson · International Journal for Numerical Methods in Engineering · 1994
Abstract A parallel implementation of the contact algorithm discussed in Part I of this paper has been developed for a non‐linear dynamic explicit finite element program to analyse the response of three‐dimensional shell structures. The parallel contact algorithm takes advantage of the fact that in general only some parts of the structure will actually be in contact at any given time. Special interprocessor communication routines and a method which enables individual processors to dynamically build local contact domains during execution have been developed. The performance of the parallel contact algorithm has been studied by executing the program on 128 processors of a distributed‐memory multiprocessor computer.
Cited by 32
Takuto NIMURA, Rei SHIBATA, Yasunori YUSA · Keisan Rikigaku Koenkai koen ronbunshu/Keisan Rikigaku Kouenkai kouen rombunshuu · 2024
In finite element analysis (FEA) that takes into account nonlinearities such as large deformation and elastic-plasticity, a solution is often not obtained depending on the input parameters. Therefore, we proposed the parallel search method for analyzable input parameters in nonlinear FEA using Bayesian optimization. However, an issue remained that the search would be adversely affected as the number of parallel processes increased. Therefore, in this study, we proposed a parallel search method that reflects the progress of FEA being executed in each processors. The proposed method was applied to a simple rectangular body torsion analysis and achieved a 24% computational time reduction compared to conventional Bayesian optimization in 120 parallel executions.
Cited by 0Open access
Osamu OKUDA · Keisan Rikigaku Koenkai koen ronbunshu/Keisan Rikigaku Kouenkai kouen rombunshuu · 2004
Cited by 0Open access