Go to the Indico Home Page Oldest event Older event Up to category Newer event Newest event Printable version
  • Compact style
  • Indico style
  • Indico style - inline minutes
  • Indico style - numbered
  • Indico style - numbered + minutes
  • Indico Weeks View
    • Switch to the management area of this event
    Login

    11th Quench Behavior Team meeting

    → Europe/Zurich
    927/1-B23 (CERN)

    927/1-B23

    CERN

    35
    Show room on map
    Hide

    Date: 2016-07-12, 09.00-11.00

    Place: 927

    Presents: P. Ferracin, P. Hagen, J. P. Tock, F. Savary, O. Bruning, S. Le Naour, A. Verweij, G. De Rijk, M. Modena, E. Todesco

     

    Other arguments to train two sectors to 7 TeV [A. Verweij]

    Arjan presents a list of open issues regarding the training of the magnets towards 7 TeV:

    • Are there other slow trainer batches in the 3000 production?
    • Will the 1000 and 2000 series start to quench significantly more between 6.5 and 7 TeV, smearing out the difference between 1000, 2000 and 3000 series that we see at 6.5 TeV ?
    • Will the fast quench propagation (below a second) become more frequent at current larger than 11 kA, increasing the recovery time? Arjan notes that this sub-second propagation is present especially in the 1000 series, for unknown reasons. If this would become a showstopper, would a threshold increase work (topic for MP3)?
    • Is there a difference between a fast thermal cycle (weeks) and a long one (months, years)?

      Since we want to have a good sampling of the whole production, a matrix giving the repartition of production of each series in sector 12+45, 23+45 and 34+45 is presented. The other options are discarded since they give a much larger number of quenches according to the Gaussian extrapolation.

    Results are shown in slide 4, with Arjan conclusion that 23+45 and 34+45 would be preferable, since 12 +45 would not allow to test the first 100 magnets of each series, whereas with 23+45 we would miss only the first 100 of 1000 series, and with 34+45 we would miss only the last 100 of the 1000 series.

      A risk analysis is summarized in the table in slide 10. Different effects are analyzed, to each damage a frequency is estimated (we know, this is a difficult estimate of an order of magnitude) and a off time is give (this is much easier to guess). The product of the two gives the risk. For 30 quenches expected in one sector the most risky faults are (1) a single short-to-ground in the cold part, requiring cold mass replacement, and (2) a diode damage, both giving 3 day of risk.

      There is a general agreement on the fact that the risk estimate gives an order of magnitude, and that doubling the number of sectors would double this estimate, so it would not radically change the results of the risk analysis.

                                                                                       

    Training one or two sectors? [E. Todesco]

    Ezio estimates the error associated to testing one or two sectors. As a general remark, doubling the statistics should reduce the error by a factor Ö2, so decrease it by 30%. Indeed, this is what it is found for the 6.5 TeV case, considering the extrapolation of the information of one sector to the whole LHC. With one sector tested (and having removed the bias due to non-uniform repartition of the three series in each sector) estimate would range between 80 and 350 quenches, with a final value of the LHC of 172 quenches.

    The lowest estimate would be obtained by extrapolating 56 and the highest by extrapolating 45. This is also in agreement with the fact that 80-100 quenches were foreseen for the whole LHC, based on 56 experience in 2008, and we ended with twice this values.

    If the information on two sectors is used, the estimates range from 110 to 270, thus reducing the error to +60% -35%. So a second sector would significantly reduce the uncertainty on the training time estimate.

    There are minutes attached to this event. Show them.