Accelerator Reliability Workshop ESRF, Grenoble, 04-06.02.02 High Power Cyclotrons P.A.Schmelzbach 1. The PSI Proton Accelerator Facility 2. Failure Analysis 2000/2001 3. The weak Points 4. How to improve the Reliability/Avialability 5. Conclusion
The PSI Cyclotrons as backbone of a multiuser facility
What means availability/reliability? Typical operating schedule in the average :4000 to 5500 hours of beam production per year
Our definition: Availability = Ratio of beam time over 1.5 ma on the pion production targets to the scheduled beam time. Also to optimize: delivered charge on target Routine production is 1.8 ma (1.2 ma on SINQ with 4 cm thick Target E) Availability (%) 100 90 80 70 60 50 40 30 20 10 1989 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 0 0 250 500 750 1000 1250 1500 1750 2000 Beam current (µa) Availability of the beam on the meson production targets as it developed since 1989. Plotted is the percentage of the scheduled beam time during which the beam current exceeded the values given on the x-axis
The good performance is shadowed by single events AND1 beam damage and Target E repair 500 Number of beamtrips per day 450 400 350 300 250 200 150 100 50 0 Beamtrips EEC repair 25.06 / W26 Average availabilit y 85.6 % Minishutdown AHB repair 7.10 / W40+41 18 23 28 33 38 43 48 EEC repair 22.11 / W47 100 6.3 Ah delivered charge Week 90 80 70 60 50 40 30 20 10 0 Charge * 10 [Ah]; Availablity [%] High Energy Operation at the PSI Accelerators in 2001: Availability, main failures, delivered charge and beamtrips due to discharges in the electrostatic elements inside the Ring cyclotron What reduces the availabilty? Short interruptions after beam trips: 20 s to ramp the intensity, more if operator intervention is needed. About 10000 per year = 1% of the beam time, but nevertheless disturbing for some users. Failures of components, infrastucture,, repairs, cause outage of >10% of the schedulded beam time
Which conditions cause beam interruptions? 2500 inputs in accelerator interlock/run permit system 140 beam loss monitors (ionization chambers, slits) users devices (IP, Medicine, SINQ) Personel protection 25.0 Failure distribution for years 2001/2000 % of total downtime 20.0 15.0 10.0 2001 2000 5.0 0.0 Magnets Septum repairs Beamtrips Miscellaneous Cooling Target High Frequency Power supplies Site power Control Vacuum PSA CW Characterization of the 590 MeV-accelerator downtimes. The total downtime was 440 hours
Distribution of beam interruptions > 1 minute in 2001. The weak points: A few events are responsible for large downtimes - e.g. Cooling: missing redundancy - e.g. Magnets: save on spare parts, repairs time consuming - e.g. RF: components used untill they fail - e.g. Target :used untill they fail - e.g. Septa: used untill they fail. Causes?
The electrostatic elements in the Ring Cyclotron are responsible for most of the beam trips. This problem seems to become particularily critical as the power increases (1.5 -> 1.8 ma?). Ways to improve the situation: In most of the failure categories a reduction of the down time rate to less than 0.5 % (or even 0.2%) of the schedulded beam time is not a technical but only a financial problem. The ongoing improvement program includes: New Ring central region with plug-in magnets Replacement of the 25 years old power supplies (also relevant for the beam stability) Fully assembled spares for some magnets (now coils only) Redundancies in the cooling water plant (?) Preventive replacement (Target E), modularity (e.g. RF) The reduction of beam trips remains a major problem. The number of trips due to interlocks from the RF-system has been dramatically reduced by a fast, on-line discharge analysis (see contribution by Peter Sigg). Part of the problem is however only shifted to another components (Extraction septum?). The behaviour of the electrostatic elements is by far not fully understood. R&D is needed (design, sensitivity to RF-leakage and X-Rays, surface physics, beam halo, ) The interlock system should be improved to allow for a better sequence analysis. More intelligence, e.g. by inclusion of transient detection and more sophisticated decision making procedures, can suppress unnecessary beam interruptions.
Conclusions The unschedulded outage is dominated by single events, in some case with an occurrence of once in 25 years. Stastitical analysis? Improving the present reliability of the utilities is not a technical problem. Components should be redesigned such that repair is fast (plug-in techniques, exchange of whole modules, Wechselflaschen ) The electrostatic elements are the talon d Achille of the Ring cyclotron. The elimination of this problem requires a considerable R&D effort. Alternative solution?