基于GPU平台的大规模汽车结构重分析

中国机械工程 ›› 2014, Vol. 25 ›› Issue (15): 2117-2123,2129.

基于GPU平台的大规模汽车结构重分析

贺冠强;王琥;黄观新;李光耀

湖南大学汽车车身先进设计制造国家重点实验室，长沙，410082

出版日期:2014-08-10 发布日期:2014-08-13
基金资助:
国家自然科学基金资助项目(10902037)；国家高技术研究发展计划(863计划)资助项目(2009AA044501)；国家重点基础研究发展计划(973计划)资助项目(2010CB328005)

Reanalysis of Large Scale Vehicle Structure Based on GPU

He Guanqiang;Wang Hu;Huang Guanxin;Li Guangyao

The State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body,Hunan University,Changsha,410082

Online:2014-08-10 Published:2014-08-13
Supported by:
National Natural Science Foundation of China（No. 10902037）;National High-tech R&D Program of China (863 Program) （No. 2009AA044501）;资助项目National Program on Key Basic Research Project (973 Program)（No. 2010CB328005）

摘要/Abstract

摘要：

针对大规模结构重分析计算中，随着待分析结构规模的扩大，重分析计算效率大幅度下降的问题，采用CUDA并行编程模型并结合组合近似法，建立了基于GPU平台的重分析并行系统。分别对刚度矩阵组装、预处理共轭梯度法等重分析关键计算流程进行了GPU并行实现。通过对车架和车门的刚度分析对该系统的性能进行了测试。测试结果表明，所提出的重分析方法能够在确保重分析计算精度的同时，大幅度提升计算效率。

关键词: GPU, 并行计算, 重分析, CUDA

Abstract:

Combined approximation (CA) was commonly used in structural reanalysis. However, with the increasing complexity and scale of the engineering problems, the classical CA could not be used directly due to huge decrease in computational efficiency. To overcome such bottleneck for large-scale reanalysis problem, the CA procedure was reconstructed to accommodate the use of GPU by employing CUDA programming model, and then a GPU-based reanalysis system was established. Some important issues of the reanalysis, such as assembly of stiffness matrix, inverse of equilibrium equations were parallelized and optimized. To verify the performance of the suggested system and algorithms, stiffness analysis of frame and door of a vehicle were implemented successfully. Finally, the results demonstrate that the efficiency of the reanalysis is improved significantly and the accuracy of the simulated results is also promised.

Key words: graphic processing unit(GPU), parallel computation, reanalysis, computer unified device architecture(CUDA)

中图分类号:

TP391.9

贺冠强, 王琥, 黄观新, 李光耀 . 基于GPU平台的大规模汽车结构重分析[J]. 中国机械工程, 2014, 25(15): 2117-2123,2129.

He Guanqiang, Wang Hu, Huang Guanxin, Li Guangyao. Reanalysis of Large Scale Vehicle Structure Based on GPU[J]. China Mechanical Engineering, 2014, 25(15): 2117-2123,2129.

参考文献

[1]Kirsch U. Design-Oriented Analysis of Structures[M]. London: Kluwer Academic Publishers,2002.
[2]Kirsch U. Combined Approximations-a General Reanalysis Approach for Structural Optimization[J]. Structural & Multidisciplinary Optimization, 2000, 20(2): 97-106.
[3]Kirsch U, Panos Y P. Structural Reanalysis for Topological Modification-a Unified Approach[J]. Structural and Multidisciplinary Optimization, 2001, 21(5):333-344.
[4]杨志军，陈塑寰，吴晓明.结构静态拓扑重分析的迭代组合近似方法[J].力学学报，2004，36(5)：611-616.
Yang Zhijun, Chen Suhuan, Wu Xiaoming. An Iterative Combined Approximation Approach for Structural Static Reanalysis of Topological Modifications[J]. Acta Mechanica Sinica, 2004, 36(5):611-616.
[5]龙凯，左正兴，肖涛,等.组合近似法在结构优化中的应用[J].中国机械工程，2007，18(9)：1042-1046.
Long Kai, Zuo Zhengxing, Xiao Tao, et al. Application of Combined Approximation Approach in Structural Optimization[J]. Chinese Mechanical Engineering, 2007, 18(9):1042-1046.
[6]Papadrakakis M, Stavroulakis G, Karatarakis A. A New Era in Scientific Computing: Domain Decomposition Methods in Hybrid CPU-GPU Architectures[J]. Comput. Methods Appl. Mech. Engrg., 2011, 200(13): 1490-1508.
[7]Anderson J A, Lorenz C D, Travesset A. General Purpose Molecular Dynamics Simulations Fully Implemented on Graphics Processing Units[J]. Journal of Computational Physics, 2008, 227(100): 5342-5369.
[8]Zuo Wangda, Chen Qingyan. Fast and Informative Flow Simulations in a Building by Using Fast Fluid Dynamics Model on Graphics Processing Unit[J]. Building and Environment, 2010, 45(3):747-757.
[9]Dominic G, Robert S, Jamaludin M, et al. Exploring Weak Scalability for FEM Calculations on a GPU-enhanced Cluster[J]. Parallel Computing, 2007, 33(10): 685-699.
[10]Filelis-Papadopoulos C K, Gravvanis G A, Matskanidis P I, et al. On the GPUGPU Parallelization Issues of Finite Elements Approximate Inverse Preconditioning[J]. Journal of Computational and Applied Mathematics, 2011, 236(3): 294-307.
[11]Cris C, Adrian J L, Darve E. Assembly of Finite Element Methods on Graphics Processors[J]. Int. J. Numer. Mech. Engng., 2011, 85(5): 640-669.
[12]NVIDIA Corporation. NVIDIA CUDA C Programming Guide[M]. Santa Clara:NVIDLA Coorpoaration,2011.
[13]Filelis-Papadopoulos C K, Gravvanis G A, Matskanidis P I, et al. On the GPGPU Parallization Issues of Finite Element Approximate Inverse Preconditioning[J]. Journal of Computational and Applied Mathematics, 2011, 236(3):294-307.
[14]王勖成. 有限单元法[M]. 北京：清华大学出版社,2003.
[15]刘丽，沈杰，李洪林. 基于GPU的矩阵求逆性能测试和分析[J]. 华东理工大学学报(自然科学版)，2010，36(6):812-817.
Liu Li, Shen Jie, Li Honglin. Performance Testing and Analysis for Matrix Inversion Based on GPU[J]. Journal of East China University of Science and Technology (Natural Science Edition), 2010, 36(6):812-817.
[16]潘晓，徐友云，甘小莺. 基于Cholesky分解的可配置矩阵求逆FPGA实现[J]. 信息技术，2009(11)：141-143.
Pan Xiao, Xu Youyun, Gan Xiaoying. FPGA Implementation of Configurable Matrix Inversion Based on Cholesky Decomposition[J]. Information Technology, 2009(11):141-143.
[17]EM Photonics Inc. CULA Programmmer’s Guide. [2011-03-31]. http://www.culatools.com/cula_dense_programmers_guide.