Large-scale topology optimization using parallel computing based on GPU

来源 :7th China-Japan-Korea Joint Symposium on Optimization of Str | 被引量 : 0次 | 上传用户:jonh0521
下载到本地 , 更方便阅读
声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架
论文部分内容阅读
  Topology optimization problems require the repeated solution finite elements problems, for a large-scale problem the solution time is typically very high.Driven by the insatiable market demand for real-time, high-definition 3D graphics, the programmable Graphic Processor Unit (GPU) has evolved into a highly parallel, multithreads, manycore processor with tremendous computational horsepower.Nowadays, it has gained more and more attention as a kind of general parallel processor, followed by various general purpose GPU computing technologies represented by NVIDIA CUDA.In this paper, a parallel topology optimization method is proposed to deal with large-scale structural design problems based on GPU.To minimize the data transfer between host and device, the entire parallel process of analysis and topology optimization has been implemented on GPU.The parallel algorithms presented here are based on the solid isotropic material with Bi-directional Evolutionary Structural Optimization (BESO) formulation of the topology optimization problem and used a soft-kill method.Considering the characteristic of GPU computation, the conjugate gradient iterative solver with a nodal based visual finite element assembly scheme is introduced for the finite method problem.This strategy can reduce the GPU memory usage during the optimization, and it can easily be made in parallel by mapping one node to one GPU thread.The programming details to realize the parallel optimization on GPU is presented in this paper.Several high API offered by NVIDIA CUDA is adopted to reduce the parallel programming difficulty.To obtain higher computational efficiency, the original serial algorithm is optimized.Finally, several numerical examples are considered, more than 100 times speedup can be obtained in handling large-scale design problems on a personal computing with a NVIDIA GTX 580 GPU.It shows the scalability and effectiveness of the present parallel approach.
其他文献
In these days, there are more and more requirement to make sure to keep schedule in any types of product in Japan.Partly because of supply-chain management becomes popular, and partly because keeping
会议
会议
The railway vehicle industry faces increasing pressure to enhance performance and safety while minimizing weight, manufacturing cost and fuel consumption.The cost can also be increased when satisfying
会议
In recent years, compliant mechanisms have been paid to attention as new mechanisms to replace traditional rigid link mechanisms.Compliant mechanisms achieve a specified motion by deforming the struct
会议
This research presents a new synthesis method named the constraint force design method allowing topology optimization for planar rigid-body mechanisms.In conventional mechanism design procedures, the
会议
Reflector antenna is a kind of typical electronic equipment with integration of mechanical and electromagnetic parameters.The ultimate goal of design is to have both of the excellent electrical and st
会议
Traditionally, a reduction of noise can be obtained by minimizing sound pressure or sound power.Thus, the objective of design optimization of an acoustic system is to reduce physical acoustic properti
会议
This paper presents a study on the integrated optimization of heat pipes and heat conductive continua.A numerical method is developed for determining the optimal heat pipe locations and continuum topo
会议
The topology optimization design of the continuum structures is one of the most challenging research topics in the field of the structural optimization.Most formulations for structural topology optimi
会议
Topology optimization of continuum structure is a current hot research field of structural optimization.The structures composed with plate or/and shell elements are important structural forms of vario
会议