Date

2012

Author

Kurç, Özgür
Polat, Mustafa Uğur

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

57
views

253
downloads

Cite This

The parallel computing systems became more affordable and available in consequence of the recent development in computer technology. Many institutions and engineers, however, can not utilize already available parallel computer hardwares due to the insufficiencies of the structural analysis softwares that they were using. Thus, one of the main objectives of this project is presenting a way to utilize the existing parallel computing hardwares without the need of additional cost and creating a considerable reduction in the analysis times by parallelizing the most frequently utilized finite element analysis techniques in structural engineering. In this project, a sigficant effort was spent on the main analysis methods of finite element method such as linear static, non-linear static, linear and non-linear time history analysis. As paralel solution techniques of linear systems of equations, two different solution approach, i.e. globnal and substructure based were implemented and their performances are tested with several structural models. Likewise, for time history analysis of structures, both implicit and explicit time integration techniques were implemented and their parallel efficiency were tested. Parallel non-linear time history analysis algoritms were also implemented utilizing the explicit integration technique. One of the main problems of developing a computational mechanics software is the difficulty of having the third parties other than the developers to use and further develop such softwares. Because of this reason, most of the academical softwares were being utilized only by a few researchers. Thus, the other important target of this project is to create an expandable software structure so that the framework can easily be utilized and further developed by other researchers. For this reason, an objectoriented data structure was carefully designed for such an analysis software and with the help of the state of the art ‘plug-in’ technolgy, external programs can be easily added to the analysis engine and utilized without any problems. In order to validate the extensibility of the developed analysis framework, finite elements and analysis methods for the heat transfer problems were developed and added to the framework as plug-ins. As a final step, the use of GP-GPU’s in finite element analysis were examined by developing several analysis methods. Even though fast solution times for direct sparse matrix solvers were not obtained when compared to the performance of multi-core CPUs, significant reduction in solution times for dense matrix operations and explicit time integration methods were obtained.

Subject Keywords

Structural analysis, Parallel computing, Finite elements, Large models, Partitioning, Substructure, PC clusters, Homogenous cluster, Shared memory, Distributed memory, GPU

URI

https://app.trdizin.gov.tr/publication/project/detail/TVRJM09EYzM
https://hdl.handle.net/11511/49989

Collections

Department of Civil Engineering, Project and Design