Can we have a little bit of automation?


#1

I have a scenario where I want to analyse 30 different aerofoils across a range of angles of attack, the mesh will be corse and will take little time to compute. I want to be able to mesh the aerofoils set up a simulation and set a variable angle to be used by the simulation I then want runs to be created based on this variation. This will reduce my simulation setups/alterations from 750 to 30.

For instance inlet angle as a function of theta, lift and drag direction as a function of theta. theta=-5 to 20 create runs, and 25 runs are created with the described setups.

also being able to plot say all lift results on a single plot and downloading them as a CSV would also make this process a lot easier.

The only way I can think of doing this is to create some form of funky automated code to produce upload cases that I currently have no idea how to make :slight_smile:

I think in the short term ill have to use a panel method program as this seems to produce ok results and is capable
of batch solving however the results vary too much on input and input varies too much with aerofoil shape and wind speed. I had much better results using simscale (obviously). Any thoughts?

BTW the workshops at the moment are really great and I’m really enjoying them!

Kind regards,
Darren Lynch


#2

I think something like this is in the works. One thing that could help you some time with the set up is to make one simulation and then duplicate it 30 times. Change the geometry for each case so that your simulations are ready and then hit go. While one simulation is going you can always line up more simulations to be solved. I would run one simulation and while is going go back to my set up and change the angle and go again until i have covered all my angles and then do this for the 30 cases. I do not actually know from the top of my head but another alternative would be to explore if tables can be used for some of your vaiables or playing with rotation of the reference frame or mesh. This would get all your angles in one run but at a computational expense plus given that the rotation would always be happening you wouldnt really be seeing a steady developed scenario.

Good luck!


#3

Hi @oscarcorripio, yes I used this technique before for analysing 2 aerofoils and that is what put me off doing it for 30. Not only is it tedious and time consuming bu t if I want to change design I have to start again. And if something doesn’t save as it often doesn’t as runs are finishing/starting then it gives me false data.

For my wind turbine project I think I’m going to analyse in xfoil, this should give me fairly good accuracy.

Kind regards,
Darren


#4

Hello @1318980,

we are working on making the use case that you described a lot more comfortable in the future. On the one hand side by making it easier to run studies with varying parameters, on the other hand by adding more intelligence in the domain assignment, such that you only need to assign entities, that really changed in your design. It will take some time, but we are aware of the importance of those features. Thanks for sharing your feedback!

One additional question: You mentioned that saving is not working for you when meshes/runs start or finish in between. This is really unexpected. Could you share what exactly happens there? Do you mean that you are not able to click the “Save” button or does the “Save” get interrupted/ignored after you clicked it?

Best Alex


#5

Excellent looking forward to the changes :slight_smile:

The issue (although not tested it recently) when a run changes state (queing, solving finished) and I press save the save is prolonged for as long as the state change, if I’ve fired off say 5 simulations this can make it really hard to know what has saved and what hasn’t. I find my self setting of the next batch and realising some midifications haven’t saved thus causing me to have to cancel. @afischer.

Hope this discription helps,
Darren Lynch


#6

Thank you for the feedback. I will forward this to the engineering team.