Learning Forum

[peteroznewman]

Hello Ana,

There are two ways to reduce the elapsed time to get to the end of the simulation. One is to take fewer iterations, the other is to reduce the time per iteration.

Look at the Force Convergence plot under the Solution Information folder.  Is the purple Force Convergence line spending a lot of iterations below the aquamarine Force Criterion line? If so, you can take larger time increments and complete the solution with fewer iterations. Another question is if there were any Bisections during the solution as shown by vertical red lines in the plot. Those are a waste of iterations and better selection of time steps can avoid them. Please reply with that plot.

To reduce the time per iteration, reduce the number of nodes in the mesh. Remesh the stent with hexahedral elements. This is done by slicing the stent into six-sided pieces in SpaceClaim then use the Share button on the Workbench tab to connect the mesh using Shared Topology. Then you can mesh with 3×3 linear hex elements in the cross-section.

I can help with this slicing if you reply with a link to the stent geometry .scdoc file. Upload that file to a sharing site such as Google Drive or OneDrive and put the link to that in your reply. If using Google Drive, select the option that allows anyone with that link to download the file.

One way to reduce the elapsed time is to use a computer with more cores. How many cores are on the computer you are solving this on?  Does the univerity have Ansys installed on an computer with more cores? For this type of problem, there is only a limited benefit. Doubling the cores doesn’t halve the elapsed time. It nearly cuts in half going from 2 to 4, then less benefit going from 4 to 8 and then very little benefit going from 8 to 16 cores assuming you have a 20 core computer. I’m talking about physical cores, not threads. Many computers can run 2 threads on each core, but that isn’t helpful for the solver which consumes all the resources of each physical core it is given.

The clock speed of the CPU has an inverse linear relationship to solution time. If the computer you used has a 2.0 GHz clock, and you switch to a computer with a 4.0 GHz clock, the solution time will be cut in half (approximately).

Another hardware question is the amount of RAM installed on the computer and the amount of RAM needed by the solver. Copy the text from the Solution Output in the Solution Information folder and paste into a text editor. That shows whether the solution solved “incore” or “out-of-core”, the latter meaning that the computer did not have enough RAM to hold the whole model during the solution, and spent extra time paging smaller parts of the solution out to the slower disk drive while it solved. It may happen that remeshing the stent with fewer nodes will result in the solution being able to solve incore in the computer you used before, so you will get a double benefit from the remesh as the solution will now will be done entirely in memory. You can put this text file on the file share if you want me to look at it.