Difference between revisions of "2016-Jul-10 Comp. Phys. Biweekly Report"
Jump to navigation
Jump to search
| Line 1: | Line 1: | ||
He Zhang: | He Zhang: | ||
| − | |||
<ul> | <ul> | ||
<li> I have made good progress in the heat load and trap rate optimization for CEBEF RF system. By now I have learned how to define the problem, handle the constraints, and apply the optimizer using the PyGMO/PaGMO library. I have tried four different algorithms: Non-dominated sorting GA (NSGA2), Non-dominated sorting PSO (NSPSO), Strength Pareto EA 2 (SPEA2), and S-Metric Selection EMOA (SMS-EMOA). The constraint of energy difference is handled by “death penalty” method. (The library provides other methods, which could be implemented if necessary.) All the above four algorithm works on our problem. The preliminary tests suggest, for this specific problem, GA works better than PSO. SMS-EMOA, although needs more generation to converge, takes less time than GA. | <li> I have made good progress in the heat load and trap rate optimization for CEBEF RF system. By now I have learned how to define the problem, handle the constraints, and apply the optimizer using the PyGMO/PaGMO library. I have tried four different algorithms: Non-dominated sorting GA (NSGA2), Non-dominated sorting PSO (NSPSO), Strength Pareto EA 2 (SPEA2), and S-Metric Selection EMOA (SMS-EMOA). The constraint of energy difference is handled by “death penalty” method. (The library provides other methods, which could be implemented if necessary.) All the above four algorithm works on our problem. The preliminary tests suggest, for this specific problem, GA works better than PSO. SMS-EMOA, although needs more generation to converge, takes less time than GA. | ||
| Line 6: | Line 5: | ||
<li> Did simulations for ion beam at 100 GeV and 250 GeV in the JLEIC collider ring under IBS effect with or without electron cooling, as Yuhong needs. Totally 16 parameter sets. | <li> Did simulations for ion beam at 100 GeV and 250 GeV in the JLEIC collider ring under IBS effect with or without electron cooling, as Yuhong needs. Totally 16 parameter sets. | ||
</ul> | </ul> | ||
| + | |||
River Huang: | River Huang: | ||
| Line 11: | Line 11: | ||
<li> River and He are trying to implement our FMM into the Hermite integrator. The Hermite integrator needs not only the field but also the derivative of the field. We have derived the formula to calculate the derivative of the field by FMM but hasn’t finish the code. In the meantime, River is study Rio’s parallel FMM code. We want to put our FMM kernel code into Rio’s parallel FMM frame, or at least understand how to use his code. | <li> River and He are trying to implement our FMM into the Hermite integrator. The Hermite integrator needs not only the field but also the derivative of the field. We have derived the formula to calculate the derivative of the field by FMM but hasn’t finish the code. In the meantime, River is study Rio’s parallel FMM code. We want to put our FMM kernel code into Rio’s parallel FMM frame, or at least understand how to use his code. | ||
</ul> | </ul> | ||
| + | |||
Yves Roblin: | Yves Roblin: | ||
Latest revision as of 13:28, 20 July 2016
He Zhang:
- I have made good progress in the heat load and trap rate optimization for CEBEF RF system. By now I have learned how to define the problem, handle the constraints, and apply the optimizer using the PyGMO/PaGMO library. I have tried four different algorithms: Non-dominated sorting GA (NSGA2), Non-dominated sorting PSO (NSPSO), Strength Pareto EA 2 (SPEA2), and S-Metric Selection EMOA (SMS-EMOA). The constraint of energy difference is handled by “death penalty” method. (The library provides other methods, which could be implemented if necessary.) All the above four algorithm works on our problem. The preliminary tests suggest, for this specific problem, GA works better than PSO. SMS-EMOA, although needs more generation to converge, takes less time than GA.
- I participated in the discussion and started working on the IMP experimental data processing. I developed a small program to automatically cut the BPM data into small pieces of half synchrotron motion period, and plot the raw data and integrated data, given the directory of the data and an initial time shift. The idea is each piece of the data should contain one ion bunch, for which other analysis could be implemented.
- Did simulations for ion beam at 100 GeV and 250 GeV in the JLEIC collider ring under IBS effect with or without electron cooling, as Yuhong needs. Totally 16 parameter sets.
River Huang:
- River and He are trying to implement our FMM into the Hermite integrator. The Hermite integrator needs not only the field but also the derivative of the field. We have derived the formula to calculate the derivative of the field by FMM but hasn’t finish the code. In the meantime, River is study Rio’s parallel FMM code. We want to put our FMM kernel code into Rio’s parallel FMM frame, or at least understand how to use his code.
Yves Roblin:
- Computational physics roundtables meeting with Amber Boehlein and co
- Richard Carlson (DOE/ASCR) visit. gave a presentation in the morning. Had a meeting in the afternoon in the context of the computational round table meetings whose goal is to define computational needs for accelerator+nuclear physics for the 2025 horizon.
- Organizing the efforts for the JLEIC modeling. Installing Synergia on the computing farm.
- Meeting/guiding student doing analysis of archived CEBAF data in the context of the CMB experiment.
- Ongoing meetings regarding the GHOST code development.
- Vacation: July 7th-8th and July 25th-August 5th.