Upload
nell-cantu
View
25
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Availability Workshop CERN 28 November 2013 Dependability calculations prior to 2008 and operational experience during the first LHC run. Jan Uythoven for the Reliability Sub Working Group (March 2004 – June 2005) & …. Reliability Sub Working Group. 13 th and last meeting on 17 th June 2005 - PowerPoint PPT Presentation
Citation preview
Availability Workshop CERN28 November 2013
Dependability calculations prior to 2008 and operational experience during the first LHC run
Jan Uythovenfor the Reliability Sub Working Group
(March 2004 – June 2005) & ….
2Jan Uythoven, Prior 2008 and experience run I
Reliability Sub Working Group
13th and last meeting on 17th June 2005 Many of the players are still around and here today
LHC Availability Workshop, 28/11/2013
3Jan Uythoven, Prior 2008 and experience run I
The model used – LHC Project Report 812 Simplified model of the LHC machine protection system Main players: PIC, QPS, BLM, BIS, LBDS Fast losses and slow losses, redundancy between systems
LHC Availability Workshop, 28/11/2013
60 %
15 %
15 %
P=0
10 %
4Jan Uythoven, Prior 2008 and experience run I
Redundancy between systems
LHC Availability Workshop, 28/11/2013
Presentation R.Filippini, 18/3/2005
P=0
15 %
C=1
15 %
5Jan Uythoven, Prior 2008 and experience run I
Resulting numbers from the study
Contributions from the different “PhD students” Result depends on operational scenario
LHC Availability Workshop, 28/11/2013
MP unsafety is 2.310-4 /year = 5.8 10-7 /hour = SIL3
42 false dumps is about 10 % of all dumps in the model200 days of 2 fills per day, 2 h between fills
6Jan Uythoven, Prior 2008 and experience run I
7 years later: Compare with Run I results
LHC Availability Workshop, 28/11/2013
2010
Estimated 10.5 %
7Jan Uythoven, Prior 2008 and experience run I
More from Ben @ Evian December 2012
Very good – global – agreement between prediction and what was measured
The QPS is the only system which systematically seems to behave worse than expected: major changes LS1
Availability ok model ok safety ok What is this trend on the BLM system ????
LHC Availability Workshop, 28/11/2013
8Jan Uythoven, Prior 2008 and experience run I
Reaction from Christos
How is the statistics done ? Only above 450.1 GeV: 2012 give 15 BLM faults If all interlocks: 31 BLM faults in 2012 If all faults: 70 BLM faults in 2012
But in all statistics the trend was spottedand the problem identified
Conclusions made on the equipment side The trend over the years was spotted The failure mode identified: optical link Action LS1: replace / clean optical fiber connections
LHC Availability Workshop, 28/11/2013
Statistics
Act
Identify
9Jan Uythoven, Prior 2008 and experience run I
Christos @ Annecy 2013
LHC Availability Workshop, 28/11/2013
10Jan Uythoven, Prior 2008 and experience run I
Machine Protection Machine Availability
LHC Availability Workshop, 28/11/2013
2010
Estimated 10.5 %Not addressed at the time
11Jan Uythoven, Prior 2008 and experience run I
Assumptions used by the model of RSWG The predicted number of dumps due to the MPS fit the model as
statistics only took into account all fills above 450.1 GeV The model assumed 2 fills per day of 10 h Only dumps in physics We did have about the 2 fills per day,
but they were not 10 h long …
However, down time at injection or without beam can be as important as downtime during physics The model should be extended to consider luminosity production and
not only the number of beam dumps The model should also be extended to all equipment and not only
machine protection elements MPS is only in a limited number of cases the cause of the down time
The model was really good for machine safety but not that useful for estimating machine availability
LHC Availability Workshop, 28/11/2013
12Jan Uythoven, Prior 2008 and experience run I
Where did we go after 2005?
LHC Availability Workshop, 28/11/2013
Predictions of the Sub-Working group
For a Sub-System of the LHC
LHC Run I StatisticsAgree with the predictions !
Machine is safe without too much impact on
availability
We need to have common metrics of
faults
For availability it needs to be extended to all
equipment
LHC Availability Working Group
To be applied to luminosity and not # faults
13Jan Uythoven, Prior 2008 and experience run I
LHC Availability Working Group
The Availability Working Group (AWG) has been created on 20/06/2012 after the endorsement of the LMC committee, with the purpose of creating common definitions and metrics for LHC dependability, modelling LHC availability using data from system experts and operations, identifying strengths and weaknesses for LHC machine availability and developing strategies for improving availability of LHC, LHC upgrades and future machines
LHC Availability Workshop, 28/11/2013
À la Ben:
14Jan Uythoven, Prior 2008 and experience run I
LHC Availability Working Group
8 meetings so far Synergy between equipment groups to measure and
improve availability of their systems Collaboration with the Maintenance Management Project
which also aims at measuring the same faults Workshop on Machine availability and dependability for
post LS1 LHC TODAY
LHC Availability Workshop, 28/11/2013
15Jan Uythoven, Prior 2008 and experience run I
Conclusions The simplified model of the machine protection system (2005 !) has
proven to be amazingly accurate Although the top model is very simple, it is based on very detailed and complex models
of the individual systems: a lot of work! One PhD per subsystem
Comparison with machine run I data gives us Confidence in the machine safety Handle on machine availability of some subsystems
Improve weak points, equipment which is underperforming Find trends and act
Fault statistics can be analysed at system level down to component level Compare with dependability models of the operational equipment Not all equipment has a dependability model
Extension of 2005 model – AWG Need to be able to measure properly, operational statistics Extend to outside the Machine Protection System Quantify the effect on luminosity
LHC Availability Workshop, 28/11/2013