Choose timezone
Your profile timezone:
17th Quench Behaviour Team meeting
→
Europe/Zurich
Date: 2017-04-18, 09.00-10.30
Place: 927
Presents: G. Willering, P. Hagen, E. Todesco, M. Modena, G. De Rijk, P. Ferracin, S. Le Naour
Expected quenches per sector [E. Todesco]
Ezio presents the estimate of quenches in each sector to reach 7 TeV. This estimate is needed as input for the LS2 schedule, namely to understand if some sectors should be ready in advance, to allocate more time for training.
The model splits the production in five batches: 1000 series, a slow (2001-2200) and a fast (2201-2246) batch of 2000 series, and a slow (3126-3300, 3376-3416) and a fast (3001-3125, 3301-3375) batch of 3000 series. These separations of slow and fast batches are based on the hardware commissioning data of 2015, where the whole LHC has been trained to 6.5 TeV.
For the 1000 series, and for the 2000 fast batch, a 10% quench probability is assumed for 7 TeV (no extrapolation is possible due to the lack of data). The outcome (slide 8) gives ~550 quenches, including an estimate of 100 second quenches in the slow 3000 series batch. Note that second quenches are associated to the 3000 series slow batch, even though data show that they are all localized in a smaller batch of ~120 magnets only (3151-3230, 3375-3416)– so this estimate could be pessimistic.
The slowest sector is 45, as expected, with ~110 quenches. Then we have three slow sector, with ~80 quenches (56, 67 and 81). Then we have three fast sectors, with ~60 quenches (23, 34 and 78). Finally, the superfast sector 12, with ~25 quenches. As suggested by Gijs, the evaluation of the statistical error gives ±10 quenches (Gaussian sum of the errors) or ±20 quenches (linear sum of the errors) for each sector estimate. The systematic error due to the estimate of the parameters of the fit is unknown.
Correlation between maximum current reached in SM18 and training in the tunnel [E. Todesco]
Ezio presents the first study concerning the correlation between the maximum current reached in SM18 and the training during hardware commissioning. Even though all magnets were planned to be brought to 12850 A, during the production a relevant fraction has been trained to lower values: 57% of the 1000 series were not trained to12850 A, and 38% of the 2000 and of the 3000 series.
This different test procedure is not correlated to the series performance at 6.5 TeV (1000 series trains much faster than 2000, and 2000 much faster than 3000 series). So the different performance of the three manufacturers is not related to the test strategy but it is intrinsic of the manufacturer.
Since the fraction of magnets that were trained until 12850 A on the test bench varies considerably along the production, one has to be careful when making an analysis of the performance within the same firm. For the slow batch of the 2000 series, we observe that 21% of the magnets quenched in the tunnel among the magnets trained below 12500 A in SM18. This percentage reduces to 10% for magnets that were trained above 12500 A, so there is a factor two improvement, but the statistical error is large: 21%±13% versus 10%±5%. A similar situation holds for the fast batch of the 3000 series: 39%±23% for the magnets trained below 12500 A and 9%±4% for the magnets trained above. For the slow batch of 3000 series the impact is marginal: 60%±15% for the magnets trained below 12500 A and 47%±8% for the magnets trained above. For the second quench, we have 9%±8% for the magnets trained below 12500 A and 4%±3% for the magnets trained above.
In conclusion, the magnets that were trained above 12500 A in average have a faster training in the tunnel, but within the statistical error of the sample. Neglecting the statistical error, one can estimate that if all the LHC magnet were trained individually to 12500 A, we would have saved only 15 out of 175 the number of quenches required during the hardware commissioning campaign.
There are minutes attached to this event.
Show them.