DAΦNE POWER SUPPLY SYSTEM: 5 YEARS OF EXPERIENCE AND. STATISTICS. M. Incurvati, R. Ricci, C. Sanelli, LNF-INFN, Frascati, Italy. Abstract.
Proceedings of EPAC 2002, Paris, France
DAΦNE POWER SUPPLY SYSTEM: 5 YEARS OF EXPERIENCE AND STATISTICS M. Incurvati, R. Ricci, C. Sanelli, LNF-INFN, Frascati, Italy The analysis of faults and events covers 5 years of PS operation. Table 1: PS characteristics
Abstract The Power Supply (PS) System of the DAΦNE Accelerator Complex consists of about 500 PS ranging from 10 A to 3000 A and from 250 VA to 1.5 MVA. All these PS are running almost continuously since 1997. This paper, after a brief description of the entire system, reports a lot of statistics concerning about 5 years of operation, describing the main problems the different kinds of PS have been affected by, their causes and remedies. The time distribution of faults and their grouping by fault type, PS producer, etc., are reported in the form of plots and graphs.
PS Bending Dipoles Dipole Back Legs Pulsed Dipoles Wiggler Central Poles Wiggler End Poles Injection Septa H/V Steering “C” Steerings “Lambertson” Steer. Rectang. & Square Steer. Splitter magnets Quadrupoles Sextupoles Skew Quad.Currectors Solenoidal magnets Octupoles Solenoids for experiment
1 INTRODUCTION DAΦNE is an Accelerator Complex [1] consisting of: • e+- e- Linac • ≈ 180 m of Transfer Lines (T.L.) • e+- e- Accumulator/Damping Ring (A.D.) • two Storage Rings (S.R.) The PS in the DAΦNE Accelerators are used to supply different kinds of magnets working at room temperature. The global set-up, comprehensive of T.L., A.D. and S.R., of PS is listed in Table 1. According to the design specifications the main characteristics are: • Three phase, 50 Hz mains voltage (V) 380±10% • (SR Bending Dipoles) (kV) 20 • Room Temperature (°C) 0÷40 • Current Setting & Contr.Rng.(-100%) 0÷100 % f.s. • Normal Operating Range (100%) 70÷100 % f.s. • Current Setting Resolution 5÷20*10-5 • Current Readout Resolution 5÷20*10-5 10÷20*10-5 • Residual Current Ripple Different typologies of PS have been adopted according to the PS output current and power rates and to the specific experience of the builder as listed below. • SCR’s Graetz Bridge Converter with Active Filtering [HAZEMEYER, OCEM] • SCR’s Graetz Bridge Converter with Transistor Output Bank [DANFYSIK, EUTRON] • Diode’s Graetz Bridge Converter with Transistor Output Bank [DANFYSIK] • Series Double Resonant Switching Converter [OCEM] • Zero Voltage Switching Converter [DANFYSIK] • Hard Switching Converter [INVERPOWER] Linear Converter [DANFYSIK, • Bipolar HAZEMEYER, INVERPOWER, FUG] • Bipolar Switching Converter with 4 quadrant output chopper [INVERPOWER]
N. Imax(A) 24 100÷750 16 ±10 3 650 2 750 8 750 8 2300 102 ±10 ÷ ±215 16 ±215 16 ±215 64 ±10 8 750 143 100÷585 34 336 16 280 4 120 6 120 2 3000
Vmax(V) 25÷1250 ±20 1300 1250 120 8÷50 ±10 ÷ ±25 ±25 ±6 ±20 80 25÷80 25÷30 40 10 20 6÷10
Data have been collected in a DataBase from Control System Log-Files and Operators’ Log-Books where date and type of events have been recorded since 1997. The analysis has been limited to the equipment (4 builders) installed since the beginning of the commissioning. Data are not related to builder names for privacy. MATLAB software has been used for statistics.
2 STRUCTURE OF DATABASE The DataBase consists of 535 events (Section 3) concerning PS operation. Daily events have been collected on a suitable Time Base (TB, 14 days), to give a statistical meaning to the data. With such a grouping, 119 TB have been obtained. They are depicted in Fig.1 in steps of 30 days. A smooth curve, averaging the values over four TB (2 months), has been added. The histogram clearly points out the high first-life death rate during the operation initial period. In Nov-‘98, the KLOE detector was rolled in the IP1 interaction region of the S.R. After four months of machine shut-down, the graph shows a high increase of faults and the same holds for the short (1 week) time periods of shut-down for maintenance. Maintenance, necessary for long term good operation, seems in contrast with the tendency of PS to work their best with a continuous operating cycle.
2478
Proceedings of EPAC 2002, Paris, France
45
Number of Faults/(14 days)
40
KLOE on
35
DAFNE day one
30
25
20
15
10
5
SD
SD
SD
SD
SD
0 09-Apr-2002
12-Mar-2002 12-Feb-2002
15-Jan-2002 18-Dec-2001
20-Nov-2001
23-Oct-2001 25-Sep-2001
28-Aug-2001
31-Jul-2001
03-Jul-2001
05-Jun-2001
08-May- 2001
10-Apr-2001
13-Mar-2001
13-Feb-2001
16-Jan-2001
19-Dec-2000
21-Nov-2000 24-Oct-2000
26-Sep-2000
29-Aug-2000 01-Aug-2000
04-Jul-2000
06-Jun-2000
09-May- 2000
11-Apr-2000
14-Mar-2000
15-Feb-2000
18-Jan-2000
21-Dec-1999
23-Nov-1999
26-Oct-1999
28-Sep-1999 31-Aug-1999
03-Aug-1999
06-Jul-1999 08-Jun-1999
11-May- 1999 13-Apr-1999
16-Mar-1999
16-Feb-1999 19-Jan-1999
22-Dec-1998
24-Nov-1998
27-Oct-1998
29-Sep-1998
01-Sep-1998
04-Aug-1998
07-Jul-1998
09-Jun-1998
12-May- 1998
14-Apr-1998
17-Mar-1998 17-Feb-1998
20-Jan-1998
23-Dec-1997 25-Nov-1997
28-Oct-1997 30-Sep-1997
Fig.1: Faults Time Distribution 50
143 45 40
Occurrences
35 30
≈
25 20 15 10 5 0
R TO O IS AF ES R R -T 5- W R H P TC 5- FI S WI X G E 5- US V- S -A U R F 5- H-O KE FO C A A 5- RE - TR B 5- L IM P A U 5- T T SE E UT 4- S O E R D 4- EA P SE R AL 4- AM T -F R 4- F R LE F O PL P 4- V- RIP M E IN - M G 4- X C CO E TA B 4- A D R OL K - AM B E 4- AT RV TIC RE W E S U 3- ND O A T C U M 3- R R - F IL E O TE P TY C 3- M RI GTE U N R 3- E C GI TE CH S G E D 3- L U M IT TE P W WR 3- LO R- S O C H L F 3- OO S- S X- P D 3- OIL - AU C 3- L IM A 3- I R - FC E A E S U 3- IR -F W 2- HY CH R T T 2- I TO SW IS 2- ES AR ION R E T 2- IN L A L 2- SU IN T 2- G B S I E 2- S FU 2- M I E T E D 2- IO EN OR G D R T E 2- UR CI E- R RE U C A 2- AP AG AS C T E 2- O L - M R S V P 1- EM STOO S T I L 1- ES ER S 1- HA P O 1- P T E O C S 1- C- F - FU LT U I D 1- ND FA G AR 1- ND S B O G E - T S 1- US E G OUOS F D 1- B -RE A E- L S E 1- B -RHA MP E P A T 1- - R R EB O L 1- B -N V -P RM L E N 1- B -I C E -A L E P E F 1- B - A LSRI P M - M E 1- B -F XC CO E E 1- B - AD E 1- B -B E 1- B E R EL E V NN 1- RI A D T -P 1- C E C L D O C C 1- AC TR T-F FA D C R 1- ON E E H C N NT 1- ON M -I OR ITC T C 1- OM CI -SW C A D 1- AP OGAR C L 1- NA BO FO C A - A L 1- P TR - P AM - X 1- IM AU R AL - X E 1- L IM- AU ILT N A -F TO 1- L IMV E A I UT 1- CT -B A Y IC Y 1- TB R NC S E E 0- E N G G R 0- M E E 0-
Fig.2: Faults divided by category
The average value of faults is: FAV = 4.47 faults / TB (1) In the last two years FAV was about 2.5 and rarely exceeded 5.
3 FAULTS CATEGORIES To point out the critical elements of the system, faults have been divided in categories [3] as follows: 0) Generic: No significant faults. 1) Electronic: Faults concerning logic electronic elements (DAC, Optocouplers, etc.) 2) Power Electronic: Faults concerning power electronic components (IGBT, power Fuses, etc.) 3) External: Faults caused by external events (Cooling Water, Ambient Temperature, etc.) 4) Control System: Faults caused by bad interaction with network elements (Readout, Reset, etc.) 5) Electric: Faults concerning typically AC elements (Breakers, AC-Fuses, etc.) As shown in Fig.2, faults (1), are particularly significant, even if concentrated in the first years of 2479
operation. Main troubles concerned optocouplers used in communication interfaces, electronic boards utilised for phase-loss detection, and fuses whose current rate was too stringent. Power electronic elements appear to be quite reliable. Linear components, the ones most thermally stressed, showed a higher fault rate. Also ventilation (Fans) and water cooling (water leakage) played an important role as causes of faults [2]. High external room temperature have caused the shutting down of the PS for self protection. Today, almost all PS rooms are air-conditioned and temperature is not a source of fault anymore. About Control System Interaction, troubles have been found for some PS that go to zero current without any reason (4_RESET). This is not serious problem because of a fast reset procedure. A reliability graph for any PS builder is shown in Fig.3. Three bars are presented: the first shows the total number of events occurred to each PS builder; the second relates to faults with no external events; the third shows a reliability index calculated as number of faults divided by total number of PS of the same builder.
Proceedings of EPAC 2002, Paris, France
1
300 Number of Shut-off
0.2 0.18
0.9
250
0.16
200
150 Number of Faults
50
0.5
Builder 4
Some statistics calculations have been done to determine the behaviour of the PS system and foresee its behaviour. To weight the contribution of external causes, the number of Total Faults has been plotted versus the PS Faults excluding external causes for each TB (Fig.4).
Number of PS in fault after Voltage Dip
Unity Slope Reference Line
25 20 15
60
0
0
20 40 Number of Faults
60
35 30 25 20 15 10 5
e1 9
e1 7
e1 5
e1 3
e9
e1
e7
0
Least Squares Linear Fit Slope=0.55
5
20 40 Number of Faults
40
40
10
0.02
Also an 8 months analysis of PS sensitivity to mains Voltage Dips has been done. Fig.6 shows the number of PS faults for each of the 19 recorded events. Tunings have been done on most sensitive PS to reduce the effects of Voltage Dips.
45
30
0.1 0.08
Fig.5: Cumulative and Density of Probability plots
4 STATISTICS
35
0.12
0.04
e5
Builder 3
e3
Builder 2
Builder 1
0.2 0
0.14
0.06
0.3
Fig.3: Builders Faults
PS Faults (No external events)
0.6
0.4
(Number of faults/Total PS of Builder)*100
0
0.7
e1 1
100
Probability Density
Cumulative Probability
0.8
0 -5 -5
0
5
10
15
20 25 Total Faults
30
35
40
45
Fig.6: Voltage Dips sensitivity
Fig.4: PS Faults (no ext.) vs. Total Faults The unity slope line represents the equality between Total Faults and PS Faults. The dotted line is the leastsquares interpolation of the scattered plot. The slope of this line can be taken as an incidence reference number of external faults. The percentile incidence of external faults can be evaluated as: 1 − 0.55 Ex% = ⋅ 100 = 45% (2) 1 To evaluate the statistical behaviour of the system, the number of faults in TB has been taken as chance variable (c.v.). Occurrences of c.v. have been summed and divided by total number (119) of measurements, obtaining the probability. Thus Cumulative Probability and Probability Density plots have been derived and are shown in Fig.5. The behaviour of the system seems to be well described by Poisson distribution (dotted line). Χ2 (Chi-square) goodness-of-fit test has been executed yielding Eq.(3), that, from probability tables, assures 5% statistical goodness-of-fit. 2 1 n (measurek − n ⋅ p fit ,k ) χ = ⋅ = 1.2 d k =1 n ⋅ p fit ,k 2
∑
(3)
5 CONCLUSIONS PS system of DAΦNE Accelerators has been analysed showing at present a high reliability after few years of running. All PS used at LNF have fully met specifications after an initial period of tuning. The analysis points out some critical points that are currently being optimised and will be useful for future PS specifications.
6 ACKNOWELDGEMENTS Authors whish to thank all the DAΦNE team with special attention the technicians B.Bolli, S.Ceravolo, F.Iungo, F.Sardone, for their kindness in helping to build up the database and their skill in operating the PS system.
7 REFERENCES [1] R. Ricci, C. Sanelli, A. Stecchi: “DAΦNE Magnet Power Supply System”, EPAC’98. [2] L. Pellegrino: “The DAΦNE Water Cooling System”, EPAC’98. [3] Dan Wolff, Howie Pfeffer: “Experience in Mantaining and Operating Power Supplies Used in Accelerators”, EPAC’96
2480
