20.07.2015 Views

PhD Thesis - staffweb - University of Greenwich

PhD Thesis - staffweb - University of Greenwich

PhD Thesis - staffweb - University of Greenwich

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>PhD</strong> <strong>Thesis</strong> by John Ewer.4.2.3.3 VisualisationVisualisation, as previously mentioned, is relatively costly in terms <strong>of</strong> compute resourcesparticularly for 3D displays. This performance overhead <strong>of</strong> visualisation has been mitigated bythe choice <strong>of</strong> 2D slice visualisation and the provision <strong>of</strong> options to limit the frequency <strong>of</strong> visualupdates. Another problem is the question as to whether intermediate (part processed) solutionstatus displays have any meaning. This depends on the context <strong>of</strong> the display and theinterpretation by the user. Any visualisation <strong>of</strong> a non-converged solution status has NO realphysical meaning other than as an indication <strong>of</strong> the current trends within the data. At the end <strong>of</strong>a time step or when a converged solution is obtained (for steady state problems) the visualisationis a true representation <strong>of</strong> the simulated physical behaviour. It was not clear at this stage as towhether other variables could give more meaningful and indicative data and status visualisations,e.g. error values. Any visualisation tools built into the CFD code will serve a dual purposebecause they can be used to monitor the intermediate solution status and they can be used aspost processing data explorer when the simulation has finished. It is possible that there is anovice user role for pattern recognition for dynamic KBS reasoning. It is envisaged that evenif planned attempts at automated pattern recognition fail then it will still be possible for users tovisually detect patterns and to thus select appropriate options based on known examples andadvice from the UI. In general, visualisation techniques can be costly to develop and to tailorto the underlying data architecture particularly for the unstructured mesh class <strong>of</strong> codes wherethere is no regularity or predictability in the layout <strong>of</strong> the data or the navigation betweencomputational cells.4.2.3.4 GraphsGraphs can be highly informative particularly for data trends (continued movement in onedirection by similar amounts) and instabilities (oscillatory behaviour). They are inexpensive interms <strong>of</strong> compute resources and are generally easy to develop. Graphs may, however, overemphasisecertain solution features inappropriately (e.g. spot value graphs are not very usefulunless used with experimental data or a known analytic solution).4-76

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!