The OPAL Discussion Forum

["Snuverink Jochem (PSI)" <jochem.snuverink AT psi.ch>]

Dear Sebastian,

Here are some thoughts and suggestions.

OPAL is not specifically designed to optimise optics, since as you have seen there are no "matching" routines. So instead you could try to do the matching in MadX and only then use OPAL.

However, if the MadX solution is not the one you want, or the OPAL and MadX simulations are too far apart (e.g. due to space charge), then here are a few more suggestions:

You can try the multi-objective optimisation (based on a genetic algorithm) built in with OPAL:

http://amas.web.psi.ch/opal/Documentation/master/index.html#chp.optimiser

With this you can define your constraints and objectives (target).

That should give you better solutions quicker than just a random grid search, especially if you can run in parallel on a cluster.

If you have a single objective, then instead of using a genetic algorithm, you could also try to use a more traditional optimisation method like Nelder-Mead. You could use the runOPAL python scripts to start the OPAL run in python and do the optimisation in python.

And sure we can have a zoom call about this. Just let me know.

Hope that helps,

Jochem

---

From: opal-request AT lists.psi.ch <opal-request AT lists.psi.ch> on behalf of Taubert, Sebastian <taubert AT uni-mainz.de>
Sent: Tuesday, January 26, 2021 12:28 PM
To: opal AT lists.psi.ch
Subject: [Opal] Optics Optimization in OPAL

 

Dear OPALers,

I have been struggling to optimize the optics of my beamlines for quite some time now. The beamlines include Quadrupoles, horizontal and vertical SBENDs as well as drifts. Once I import these beamlines from Elegant or MadX the optics matching is not working anymore and I have to tune the Quadrupoles in a way that fits the demands again so that I can start with the real investigations, that I want to conduct with OPAL.

The thing is, I have the feeling I am doing something wrong. I am missing something. How do you optimize your optics in OPAL? What is the recommended approach to match the envelope and dispersion of the beamline to desired values? At the moment I sample a hundred runs screening through the setting of a certain Quadrupole Triplett which takes a night for a high enough precision. Then I choose by eye the "best" solution and do another hundred runs with a closer screening of these values (like zooming in). And then I repeat that one more time. That accounts to three nights of simulation to optimize one of four Quradupole Tripletts in one of six arcs in MESA. That can't be close to what one is supposed to do. Most because the results are not good. They are not even passable. 

Maybe someone can help me to help myself in a short zoom call or something? I really really want to find out what OPAL is capable of, but I don't seem to be able to do a simple optics optimization with it. 

Any help is appreciated. Thank you very much in advance!

Cheers

Sebastian Taubert

Doctoral Student

Accelerator Physics

Institut für Kernphysik
Johannes Gutenberg-Universität Mainz
Johann-Joachim-Becher-Weg 45
D - 55128 Mainz

E-Mail: taubert AT uni-mainz.de
Office: Due to Covid-19, temporarily not in office
Mobile: +49 1515 0535622