CASE

HISTORY

2 DAYS, 200% ROI

TotalFinaElf A Machinery Management Success Story

Gear coupling 1 54ZT4502 1AD STEAM TURBINE 2 3 COMPRESSOR 4 54ZT4501 4AD 54ZT4501 4AD RESERVE

20 R 45 L
Y CCW 54XT4510 1VD 54XT4509 1HD Keyphasor 54KT4511 0 54XT4508 2VD 54XT4507 2HD X

45 R

45 L
Y X

45 R

45 L
Y X X

45 R

110 R
54XT4504 4VD 54XT4503 4HD

54XT4506 3VD 54XT4505 3HD

TRAIN DIAGRAM OF THE CRACKED-GAS COMPRESSOR | FIG. 1

Emmanuel Fix
Mechanical Engineer TotalFinaElf, Flanders Refinery

Didier Mouchie
Methods and Mechanical Engineering Service TotalFinaElf, Flanders Refinery

Frderic Le Souder
Solutions Specialist Bently Nevada SARL e-mail: fred.lesouder@bently.com

The TotalFinaElf group has six refineries in France. The Flanders refinery is located in Loon Beach, near the Channel tunnel in the north of France, and has a production capacity of nine million tons per year. At the Flanders refinery of the TotalFinaElf group, the Mechanical Team, led by Emmanuel Fix, wanted to increase the availability rate of five of their essential machine trains by using a machinery management system. This would give them results

in real time and would permit them to manage the machines proactively. The machines consisted of two cracked-gas, turbo compressors, an expander, a blower, and a turbo alternator. During a five week outage in October 2001 of one of the cracked gas compressors, Bently Nevada had just finished upgrading the existing monitoring system a day before the compressor was restarted. The upgrade consisted of the 3500 Series machinery protection system and the Data Manager 2000 machinery management software. During the startup, the 3500 system detected high vibration on a turbine bearing and shut down the machine, and the Data Manager provided the plots of the data that the Mechanical Team needed to evaluate the problem and identify possible corrective actions. Two days later, the plant information system generated a high temperature alarm for a compressor bearing,

50

ORBIT

1Q03

CASE HISTORY

and the Data Manager provided the data used to assess the validity of the alarm. In those two days, according to Fix, The return on investment was 200%,and the system allowed us to manage these problems by ourselves in a minimum of time. Vibration Incident The six-stage centrifugal compressor, driven by a six-stage, 3 MW steam turbine (Figure 1), runs at 7500 rpm. Pairs of proximity probes are mounted at the tilt-pad bearings and thrust bearings in both machines. The 3500 Series protection system transmits the data to the Data Manager software through a communication processor, TDXnet. On Thursday, 23 November 2000, the compressor was restarted after a five week outage. The Bently Nevada Product Services technician, Patrick Gaborieau, had just finished starting up the Data Manager system the day before. At that time, the system was only configured for alarms on the vibration data, and data was not being shared with the plant information system. The machine started but could not exceed 1100 rpm. The 3500 Series system generated alarms for high vibration on bearing 2 at the exhaust end of the turbine, and the machine was shut down. As soon as the alarm occurred, the Data Manager system stored the 40 data samples taken over the previous 160 second period. This allowed Didier Mouchie from the Process Mechanical Method Department to review the data to determine the cause of the problem. He found that the shaft centerline position plot (Figure 2) was particularly informative. When the

alarm occurred, the shaft was already well outside the bearing limits, and, by the time the machine had coasted down to just above 800 rpm, it had moved down about 800 m (31 mil) below the bottom of the bearing. Mouchie was confident of this value because, when the Bently Nevada MMS engineer, Asen Mokrani, had configured the system, he had set the reference point for the shaft location at the bottom of the bearing

clearance (indicated by the circle in Figure 2). This reference position is set when the rotor is stopped, so that the weight of the rotor holds it in the bottom of the bearing. The orbit at this bearing had grown from about 15 m (0.59 mil) to over 100 m (4.0 mil), almost 50% of the original clearance, and was a distorted ellipse (Figure 3). The vibration was primarily synchronous (1X) and forward precession. Along

THE MONEY-SAVINGS ASSOCIATED WITH THIS PERFORMANCE IS ESTIMATED AT $320,000 (INCLUDING LOSS OF PRODUCTION AND COSTS OF UNNECESSARY DISMANTLEMENT). THE

RETURN ON INVESTMENT IS 200%.

POINT: 2VD 54XT4508 45 Left REF: -10.505 V 23NOV2000 20:12:03 POINT: 2VD 54XT4507 45 Right REF: -9.032 V 23NOV2000 20:12:03 From 23NOV2000 20:12:03 to 23NOV2000 20:14:38 Delta Time (not orbit or polar plot)

THE AVERAGE SHAFT POSITION PLOT SHOWS THE RAPID DROP OF THE SHAFT IN BEARING 2 OF THE TURBINE. FIG. 2 -600 -400 -200 0 200

TOP

854 892 923 1104 1105 1097 1076 1029 991 956 915 887 874 -400
50 m/div

-800 -200

200

400

Rotation: X to Y (CCW)

1Q03

ORBIT

51

CASE HISTORY

with the shaft position information, this indicated to Mouchie that there may have been a significant increase in bearing clearance, the result of significant bearing damage. The decision was made within an hour to dismantle just the bearing as soon as possible. When the bearing was dismantled, it was discovered that the bearing pads were seriously damaged (Figure 4) due to lack of lubrication. The lack of lubrication

minimized the consequences of this incident. Not only was the rotor undamaged, but the downtime necessary for the repair was strikingly shortened. The value of Data Manager was demonstrated again just two days later on the same machine. Temperature Incident On Saturday, 25 November, after the machine been running at low speed for several hours, it was brought

ually shut down. Could the increase have been caused by a rub? Was this the same problem as before? Mouchie and Fix immediately consulted the dynamic data recorded by the Data Manager 2000 system to see if it could help them identify the problem. The orbits recorded during the alarm did not show any abnormal loading of the machine; they were forward, circular, and with forward precession. The vibration
FIG. 4

Y: 2VD 54XT4508 45 Left Waveform Pk to Pk: 107 m pp X: 2HD 54XT4507 45 Left Waveform Pk to Pk: 99.9 m pp 23NOV2000 20:14:38 Delta Time Direct

TOP

FIG. 5
10 m/div Rotation: X to Y (CCW) 876 rpm

THE SILICONE PLUG IN THE OIL LINE RESULTED IN OIL STARVATION AND DESTRUCTION OF THE BEARING.

FIG. 3

was caused by silicone that had oozed into the oil line and formed a plug (Figure 5). Mouchies diagnosis was confirmed. Mouchies quick and accurate diagnosis, based on the information supplied to him by Data Manager,

up to running speed. Suddenly, the plant information system indicated a temperature alarm and showed that the temperature on bearing 4 on the compressor had increased to 115 C (239 F) (Figure 6). As a precaution, the machine was man-

levels were lower than 10 m at all the bearings. The shaft position plot showed that the rotor was operating well inside the limits of the bearing clearances. The rotor was located in the correct position in the bearing during the alarm; it did not show

52

ORBIT

1Q03

CASE HISTORY

the effects of any radial constraint that might result from a rub. The Bode plot of synchronous data recorded during the shutdown did not show any change in the resonance speed. A rub during the high vibration associated with resonance would have increased the stiffness of the mechanical system and been obvious on the Bode plot. None of the dynamic data supported a mechanical problem. Mouchie and Fix then consulted the alarm list from the Data Manager: no alarm related to vibration appeared during the time of the temperature alarm. So they reasoned that only an instrument failure could have produced the temperature alarm. Confident of their diagnosis based on their interpretation of the vibration data, they decided to restart the machine. Indeed, the vibration levels remained normal, even though the temperature changed up and down dramatically, and the machine returned to production at 6600 rpm. About this second event, Fix said, The system strongly accelerated and improved the reliability of our decision process. A few months ago, the same event would have obliged us to consult an outside company to identify the incident and to restart the machine to take more data (potentially more stress on the machine). Without any more data on the nature of the incident, we would have decided to open up the machine. Furthermore, this problem happened during a weekend. We would not have been able to easily manage this heavy operation before the next Monday. We would not have been able to solve the problem within 24 hours. It would have taken at least 2 days.

Conclusion These two incidents demonstrate the versatility of Data Manager. In the first case, it allowed the problem to be precisely identified so that the repair could be well organized. In the second case, it quickly confirmed the mechanical condition of the rotor and efficiently directed the investigation to other causes. The

also allowed us to discover an instrument error by validating the good condition of the machine. For us, the value of this system is threefold: 1. It lets us shut down the machine in time to limit potential damage, 2. It helps us decide if we can restart the machine, and 3. It helps us determine if we need to shut the machine down. Data Manager 2000

300
BEARING TEMPERATURE (C)

10000 9000

BEARING TEMPERATURE (C)

250

MACHINE SPEED (RPM)

8000 7000 6000 MACHINE SPEED (RPM)

200

150

5000 4000

100

3000 2000

50 1000 0 25/1 1 /0 0 0:00 25/1 1 /0 0 2 :24 25/1 1 /0 0 4:48 25/1 1 /0 0 7:1 2 25/1 1 /0 0 9 :36 0 25/1 1 /0 0 12:00

T I M E

FIG. 6

results were obtained by a simple, but in-depth study that accelerated the search for the most likely cause, thanks to the recorded data and its presentation in effective formats: Bode for transient data, and orbits and shaft centerline for steady state data. Fix concluded that, This system allowed us to manage this problem by ourselves in a minimum of time, and to enable the maintenance operations required (dismantle and change one bearing only). It

allowed us to completely fulfill these objectives. Since these incidents, the 3500 system has been configured to alarm on shaft centerline position, and the Data Manager configured to import all the process values and temperature via a Modbus link and to export all the vibration values via an OPC link.

1Q03

ORBIT

53