Computational & Technology Resources
an online resource for computational,
engineering & technology publications |
|
Civil-Comp Proceedings
ISSN 1759-3433 CCP: 76
PROCEEDINGS OF THE THIRD INTERNATIONAL CONFERENCE ON ENGINEERING COMPUTATIONAL TECHNOLOGY Edited by: B.H.V. Topping and Z. Bittnar
Paper 13
Decomposition Method Applied to Orthogonal Polyhedra R. Obiala, B.H.V. Topping, D.E.R. Clark and G.M. Seed
HPC Research Group, Heriot-Watt University, Edinburgh, United Kingdom R. Obiala, B.H.V. Topping, D.E.R. Clark, G.M. Seed, "Decomposition Method Applied to Orthogonal Polyhedra", in B.H.V. Topping, Z. Bittnar, (Editors), "Proceedings of the Third International Conference on Engineering Computational Technology", Civil-Comp Press, Stirlingshire, UK, Paper 13, 2002. doi:10.4203/ccp.76.13
Keywords: boundary representation, orthogonal polyhedra, brinks, extreme vertices model, decomposition, parallel analyses.
Summary
A method of decomposition applied to orthogonal polyhedra is presented in this paper.
The process of decomposition of a solid model is carried out for two different purposes. The first is to reconstruct elements that were merged into a single solid. The second is the division of
a whole model into smaller elements as a pre-processing for subsequent parallel analyses. In the approach that is presented in the paper both purposes can be satisfied because the decomposed models are orthogonal polyhedra.
The paper shows decomposition of an orthogonal (pseudo) polyhedra based on an Extreme Vertices Model. Definition of an orthogonal polyhedra is presented followed by Aguilera [1]:
As an example of the whole procedure, let us analyse zyx decomposition. First the model is split according the z direction. The cutting plane is displaced from one to another extreme vertex along the z axis through the model. It starts at the point where the vertices have the smallest z value of the coordinates and finishes where the z value of the coordinates is the largest. At this point the model is split into two sub-models. All the brinks which create the cut model must be separated into two sub-models to complete the splitting. First the brinks in relation to the z direction are analysed by comparing their coordinates with the value of the cutting plane. Some of the brinks have to be cut into two. In the next step the brinks parallel to the y axis and the x axis are selected. The brinks which lies on the cutting plane are omitted for a while and the remaining ones are easily attached to the proper sub-model. Finally, the brinks laying on the cutting plane are separated, cut or new brinks are created if it is necessary. When the sub-models are completed in sense of a brink frame of the models the cutting plane moves to the next potential place of cutting. The same procedure is repeated until the plane reaches the maximal value in z direction. The number of sub-models are generated in that process, nevertheless, the decomposition is not finished yet. Now, each of the sub-model is treated as the basic model for decomposition according to the next direction. In the analysed example it is direction y. This analogue procedure is repeated for the y direction and subsequently for the x direction. The original model is cut into a number of parallelepipeds formed as frame models. To finalise the decomposition the surfaces for each block must be defined. It is clearly seen that the process enables the generation of six decomposed models. It is worth noticing the models can be created independently therefore parallel processing can be used to speed up the calculation. The decomposition is a part of the pre-processing required for a parallel finite element analyses. For the generated blocks a finite element mesh will be designed and then the blocks will be split and distributed to a number of processors. References
purchase the full-text of this paper (price £20)
go to the previous paper |
|