Computational & Technology Resources
an online resource for computational,
engineering & technology publications |
|
Civil-Comp Proceedings
ISSN 1759-3433 CCP: 93
PROCEEDINGS OF THE TENTH INTERNATIONAL CONFERENCE ON COMPUTATIONAL STRUCTURES TECHNOLOGY Edited by:
Paper 332
Implementing Modal Analysis Software on Multi-Core Computers: With Application to Seismic Analysis of Space Trusses R.I. Mackie
Civil Engineering, School of Engineering, Physics and Mathematics, University of Dundee, United Kingdom R.I. Mackie, "Implementing Modal Analysis Software on Multi-Core Computers: With Application to Seismic Analysis of Space Trusses", in , (Editors), "Proceedings of the Tenth International Conference on Computational Structures Technology", Civil-Comp Press, Stirlingshire, UK, Paper 332, 2010. doi:10.4203/ccp.93.332
Keywords: component-oriented, object-oriented, eigenproblems, modal analysis, seismic analysis.
Summary
The work is motivated by ongoing research into the behaviour under seismic loading of space
trusses. Changes in the design of the space truss alter its dynamic characteristics. The interaction
between the dynamic characteristics of the structure and the frequency spectra of the earthquake
are an important factor in determining the response of the structure to seismic loading. These
calculations can be very time consuming, and therefore it is desirable to speed things up, and to
enhance the design of the software to facilitate engineering design. The work described in this
paper forms the preliminary stages in an ongoing project to meet these objectives. The work
currently focuses on the modal superposition approach to solving dynamic problems. The first part
of this is the determination of the vibration modes of the structure in question, once this has been
done the mode superposition method is used. The paper will look first at the implementation of
eigensolution on muti-core computers using the subspace iteration method. Earlier work by Heng
and Mackie [1] looked at the component mode synthesis method, and others have used the
Lanczos method. In this paper attention is limited to the eigensolution phase. Attention is given to
software design aspects; execution speed; and using the methods to enhance usability.
The subspace iteration method is used to calculate the modes. Parallelisation of the algorithm itself and the use of the domain decomposition approach to achieve parallelism are considered. On the software engineering side, component oriented design is used. The subspace solver object uses a component to represent the data (i.e. the mass and stiffness matrices etc), thus making the algorithm independent of the data storage scheme used. The use of object and component oriented software design allowed significant code reuse, including reuse of equation solvers and node optimisers. Parallelisation of the subspace iteration method is implemented using a beta version of the .NET Task parallel Library (TPL). The software is tested on two space trusses, with 1623 and 9831 degrees of freedom, with the code being run on a quad-core processor. Parallelisation of the subspace algorithm itself yielded a speed-up of 2.59 for the larger problem, whereas the domain decomposition version achieved 3.16. Future work will use these in conjunction with the mode superposition method. References
purchase the full-text of this paper (price £20)
go to the previous paper |
|