Choose timezone
Your profile timezone:
4th Quench Behavior Team meeting
→
Europe/Zurich
Minutes of the 4th meeting of the Quench Behavior Team
Date: 2015-09-24, 10.30-12.00
Place: 927
Presents: G. De Rijk, P. Hagen, P. Ferracin, J. P. Tock, M. Bajko, J. C. Perez, E. Todesco
Test of 3409 in SM18 [M. Bajko and G. Willering]
We have today three 3000 series magnet on surface. Previous analysis shows that two of them belong to the initial part of the production showing no training (probability of 0.1 quench / magnet to reach 6.5 TeV + 100 A). The third one (3409) is a candidate for further testing, belonging to a batch with a probability of 0.4 quench/magnet to reach 6.5 TeV + 100 A.
Marta shows a plot of Gerard summarizing the test of the magnet, which went under 5 powering phases in 2008-2015. There is a clear indication of a bad memory: the first quench went from 12.0 kA in the first to 11.5 kA in the second, and 11.3 kA in the third one. The two successive powering had first training well above nominal. On the other hand, 5 out of five powering cycles went to 11.1 kA with no training.
The proposal is to have a few thermal cycle where to power the magnet until 11.8 kA, to see if in some cases there is a training quench at current lower than nominal. A quench localization would be welcome.
Marta points out that the magnet is at the moment already on the test bench, so one could
- Power it in October and get some more statistics on the training “for free” but without quench localization, to understand if this magnet is really representative of the loss of memory in the 10-12 kA range before going for quench localization.
- Remove it from the test bench, install quench antenna and test it next year.
An additional complication is that the magnet is ready to be installed (with the beam screen) so to insert the quench antenna the beam screen has to be removed.
After some discussion it is proposed to make a test campaign (three thermal cycles) in October 2015 without quench localization. At the same time check (i) if the beam screen would be lost if removed and (ii) explore with Stephan the possibility of developing a quench antenna to be used with the beam screen.
Analysis of quenching aperture and follow-up [E. Todesco]
Ezio presents some additional analysis of the commissioning data.
Following the suggestion of Sandrine, he has included sector 45 in the analysis of the comparison between 2008 and 2015 training campaign, to understand if the training is longer (degradation) or shorter (memory). Both data set of sector 45 trained to 10300 A and of sector 56 trained to 11100 A, are compatible with the hypothesis that there is no degradation and no memory, i.e. at data are compatible with the hypothesis that after each long shutdown the situation will be the same.
The analysis to check the hypothesis that secondary quenches reduce training is being reviewed. After a first positive indication based on the statistical analysis presented by Ezio and Per in August, Arjan has suggested via email that the sample could be biased, so more effort should be done to understand this important feature.
Ezio carried out the analysis of which aperture has quenched to be able to test the hypothesis that the quench in one aperture is training the other one. If this hypothesis is not verified, also the beneficial effect of secondary quenches cannot be true. The quenching aperture will be handled to MP3 to be stored in the database.
As a byproduct, we can now test the symmetry between Ap1 and Ap2 quenches. Out of 170 quenches, 76 were in Ap1 and 94 in Ap2. All sectors are compatible with a 50-50 probability of Ap1 and Ap2 within a 2 sigma error, with the exception of 67, were we had 4 quenches in Ap1 and 16 in Ap2. This asymmetry probability is less than 5%.
Knowing the quenching aperture, we can now see how many apertures quenched twice. Among the 170 quenches, only 3000 series magnets had two quenches (9) and one 3000 series magnet had three. Among the double quench magnet, four magnets out of nine had quench in the same aperture. They are removed from the analysis since we have no statistics on the double quench of the same aperture, and it looks as negligible in this campaign (4 out of 170).
If the aperture were totally independent, one would expect an average of 16 magnets presenting a quench in Ap1 and Ap2 for the whole LHC, against an observed value of 5. A MonteCarlo shows that this discrepancy is statistically significant, well above two sigma. So the conclusion is that data are compatible with the hypothesis that the quench of one aperture reduces the training of the other one (in the analyzed range, i.e. 10-11 kA).
There are minutes attached to this event.
Show them.
