Académique Documents
Professionnel Documents
Culture Documents
TURN IT DOWN
Reducing Operating Expenses through Condition Monitoring
Now Integrated with System 1* Software Photo on page 64 courtesy of Valero Energy
Company. Used by permission.
60 Relief Valve Leak Detection
TRADEMARKS
From The Desk of
The * symbol denotes a tradename,
14 Reducing OpEx Through Condition Monitoring trademark, or registered trademark
of General Electric Company or Bently
Back to Basics Nevada, LLC, a wholly owned subsidiary
of General Electric Company.
30 Asset Management 101: Part 1 – Maintenance Strategy Overview
Crystal Reports is a registered trademark of
Business Object S.A., an SAP company.
4 O R B I T Vol.29
Vol.28 No.1 2009
2008
CASE HISTORY
DARWIN
How accurate machinery data and in-situ trim
balancing of an aeroderivative gas turbine allowed
Darwin, Australia-based Power & Water Corporation
to meet the region’s electricity demands during the
critical wet season.
GE ENERGY
Chris Engdahl
Field Application Engineer,
Bently Nevada Asset Condition Monitoring
chris.engdahl@ge.com
Vo l . 2 9 N o. 1 2 0 0 9 OR B I T 5
CASE HISTORY
D
arwin is the capital city of Background
Australia’s Northern Territory
In mid-2006, Unit 7 was taken offline for general
(NT), also known as the “Top
maintenance and to address specific repairs needed on
End,” a vast outback region
the inlet gearbox (IGB). After the unit was brought back
rich in natural beauty and
to service, high vibration was measured at the turbine
mineral deposits. With a land
rear frame accelerometer filtered to LP shaft speed (TRF
area almost the size of Alaska,
LP channel) while operating within the 10-15MW range.
and a small, dynamic population, the Northern
While the unit remained in this load range, frequent
Territory currently enjoys economic growth driven
vibration alarms (above 1.5 inches/sec pk) occurred.
by investment in major infrastructure projects in
However, as the unit moved into higher load ranges, the
transportation, LNG, gas-to-liquids, minerals, and
TRF LP vibration decreased and remained below 1 inch/
power generation.
sec pk. Consequently, PWC restricted the machine from
The Power and Water Corporation (PWC) is the NT’s prolonged operation within the low-load range until
provider of electricity, water, and sewer services. the unit could be taken offline for further investigation
PWC is now one of the largest businesses in the during a planned outage later in the year.
NT with assets in excess of AUS$1.2 billion, 750
The planned outage called for a unit control upgrade,
employees, and more than 77,000 customers.
replacement of the thrust balance valve, and inspection
PWC’s current installed capacity is 399 megawatts of the complete engine package. It also provided the
with the largest gas turbine power station (232 window of opportunity that PWC’s diagnostic engineers
megawatts) at Channel Island, located on Darwin needed to investigate the low-load vibration problem.
Harbour’s Middle Arm. This power station consists Plans were made to connect temporary data acquisition
of five GE Frame 6 gas turbines, two of which are instrumentation, beyond the basic protection system
used in a combined cycle configuration with one (Figure 1) provided with the turbine-generator package.
Mitsubishi steam turbine, and one GE LM6000PC As the machine was not fitted with permanent online
aeroderivative gas turbine (see sidebar on page 7). condition monitoring capabilities, such as via System 1*
software, this temporary additional instrumentation
This GE LM6000 turbine, “Unit 7,” is an important
would capture transient (i.e., start-up and shut-down)
peaking unit on the PWC grid, supporting domestic
data as well as steady-state data, allowing the PWC
and industrial demand particularly through the
team to conduct advanced machinery analysis.
critical wet season periods when power demand in
the tropics hits its ever-increasing peak. The PWC engineering team already had a strong back-
ground in rotating machinery testing and troubleshooting.
When increased vibration levels threatened to
Many on the PWC staff had attended GE’s Bently Nevada*
load-limit the operation of this turbine, PWC
training courses, including Machinery Diagnostics. This
engineers needed to find a solution to keep the
training had equipped them with a solid foundation
machine in-service with minimum impact to
in vibration analysis techniques, rotating machinery
production. This case history discusses how the
condition monitoring methodologies, and rotor dynamic
PWC team carried out a quick and effective in-situ
fundamentals, providing them with the confidence to
trim balance of the LP rotor, minimizing downtime
tackle complex machinery problems in the field.
and delivering customer value during the peak
demand period. Continued on page 8.
LM6000
the
ENGINE
The General Electric LM6000 is derived from the CF6- sensor. The standard procedure whenever excessive 1X
80C2 turbofan, proven as one of the most economical vibration indicates an unbalance condition is to remove
aircraft engines currently in service, applied to 11 the engine and ship it to a GE-approved workshop for
wide body aircraft including Airbus’ A300 & A310, and overhaul and balancing.
Boeing’s 767. In its “LM” (Land and Marine) form it has
Balancing Considerations
been modified to make it more suitable for marine
In-situ balancing of the LM6000 is difficult, as the HP
propulsion and industrial power generation use.
rotor, running as a spool on the main LP rotor, is almost
These include an expanded turbine section to convert
inaccessible when the machine is in operation. No
thrust into shaft power, struts for deck mounting, and
Keyphasor* reference can be installed, and in fact the
reworked controls packages for power generation. It has
only speed-related signal is taken from the accessory
wide applications, including peaking power plants, fast
gearbox output, a multi-stage gearbox, linking starter
ferries and high speed cargo ship propulsion. To date,
motor, air compressor, and oil pump to the HP rotor. A
there are over 600 GE LM6000 units in power generation
proximity probe on the accessory gearbox gives a speed
service worldwide.
signal suitable for control and the HP vibration tracking
The LM6000 provides 40.7 MW from either end of the filter, but to get a steady phase signal requires a bank of
low-pressure rotor, rotating at 3,600 rpm. The twin spool three Bently Nevada TK16 “Keyphasor Multiplier Divider”
design allows the low pressure turbine to be directly units (now also embedded within the new ADRE* 408
coupled to a generator, although for 50 Hz generation, Keyphasor card). Consequently, trim balancing of the
a step-down gearbox is used. Its high efficiency and HP rotor cannot generally be done in-situ, although
installation flexibility make it ideal also for a wide variety accurate measurement of the HP shaft 1X vibration
of utility power generation and industrial applications, vector does provide important diagnostic information
especially peaking and cogeneration plants. on the machine.
In fourth quarter 2006, the outage was conducted as integrity, provided they could run the machine within the
planned. During re-commissioning, vibration testing OEM’s recommended limits for vibration amplitude.
was carried out by PWC personnel utilizing two ADRE*
Collaboration with GE engineers led to the conclusion
208 data acquisition instruments connected to the
that the LP rotor was unbalanced; however, the cause of
Bently Nevada 3300 series protection system. Because
the unbalance was not clear. Possibilities ranged from
GE Aviation also uses the ADRE system, and because the
blade damage to blade erosion to a mechanical- or
ADRE data formats are consistent with GE’s Installation
thermal-related incident that had affected the rotor
and Field Service Standard, PWC would be able to easily
during the previous overhaul. While an inspection and
share the collected data with GE Aviation specialists in
precision reassembly was the appropriate permanent
Cincinnati if a second opinion was required.
solution, with the wet season fast approaching,
unplanned downtime for Unit 7 was not feasible. After
conferring further with GE engineers, PWC decided to
perform an in-situ trim balance of the LP rotor as an
interim measure. If successful, this would reduce the TRF
LP vibration, thus removing the operational constraints
that precluded the unit from running in low-load condi-
tions when needed.
Introduced in the early 1980s as part of the Bently When ADRE Sxp software is connected to the 408
Nevada product line, the term ADRE is actually an instrument, real-time diagnostic display and on-the-fly
acronym meaning “Automated Diagnostics for Rotating configuration changes are available. The 408 unit can
Equipment.” The system automated the then labor- also operate completely independent of the laptop display.
intensive task of replaying rotating machinery vibration
A DC to 50 kHz bandwidth is available on all channels
data through analogue tracking filters to manually plot
simultaneously, with 24-bit data acquisition and
out each diagnostic data set.
configurable spectral resolution (up to 6400 lines). The
The ADRE 208 units were first introduced in 1993, 408 can be configured as a digital recorder, streaming
allowing diagnostic engineers to digitally capture up real-time data directly to its 130 GB internal storage for
to 16 channels of machinery data simultaneously. The post-processing as needed.
units acquired data automatically during machine
For more information, visit www.ge-energy.com/adre
transients or during steady-state testing, presenting the
data in the correct formats for analysis, reporting, and
balancing. Although the 208 has recently been super-
seded by the newer ADRE 408, many practitioners in
the machinery diagnostics field still rely on the 208 as a
dependable workhorse for rotating machinery analysis.
Together with GE Field Service personnel it was decided to position is determined. Nevertheless, this technique
install a temporary Keyphasor sensor at the Main Gearbox does provide good diagnostic information that might
input shaft. This would enable collection of synchronous, otherwise be unavailable. Due to the complex gear
transient vibration data for best diagnosis capability. ratios of the accessory gearbox, three TK16’s had to be
brought in from GE’s Singapore office for this purpose.
To better assess the HP spool shaft, a true Keyphasor
could not be fitted, but it was possible to use the Bently
Trial Weight Challenges
Nevada TK16 Keyphasor Multiplier/Divider to modify
The machine was started up on 3 November 2006 and
the speed sensor pulse according to the gear ratio, thus
ramped to its full load of 42 MW, remaining at this load
generating a steady simulated signal. The resulting
condition until 9 November when the opportunity arose
simulated Keyphasor signal only remains locked while a
to begin the trim balancing. The unit was shut down and
clean input signal is present to the TK16, so for each run
allowed to cool overnight, enabling PWC personnel to
of the machine, a new relative reference measurement
place the initial trial weight.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 9
CASE HISTORY
Figure 6. Graphical balancing methodology showing the TRF LP filtered vibration vectors. The change vector
(C) would ideally be rotated to eliminate the original vector (O), suggesting a clockwise move of the trial mass
of around 60 degrees to balance for low-load vibration. The high-load change vector, on the other hand,
suggests a counter-clockwise move of the mass for an ideal solution.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 1 1
CASE HISTORY
Figure 7. Comparison of TRF LP vibration before (left) and after (right) trim balancing. Prior to balancing, the
vibration amplitude consistently exceeded allowable limits when operating under low-load conditions (blue
curve), restricting operation to only high-load conditions. After balancing, the compromise solution enabled
the machine to remain within allowable limits under all load conditions.
SIDEBAR ARTICLE
is the change in mass causing the vibration change. While it is practical to perform manual calculations and
iterate to a good compromise solution when only one
A basic premise of the Influence Coefficient method
or two measurement planes and one or two solution
is that a linear relationship exists between the cause
points are involved, this quickly becomes unwieldy as
(change in mass) and effect (change in vibration vector).
more variables are added. To address such situations,
Thus, provided the trial mass doesn’t cause a vibration
GE’s Bently BALANCE* software is available, developed
amplitude response outside the linear range (and, of
specifically to address complex multi-plane balancing
course, doesn’t exceed permissible limits), a suitable
situations where numerous solution points must be
balance solution can be derived from the resulting data.
satisfied. In addition to performing the necessary calcu-
In the simple case, after the Influence coefficient is
lations, the software displays the predicted full transient
determined by the initial trial weight addition, the final
response using data imported from ADRE* or System 1*
solution can be easily calculated such that:
databases when testing various balance solutions.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 1 3
FROM THE DESK OF
I
recognize that our solutions are most often Consider a large process compressor with high
applied on only the highly critical turboma- vibration. Condition monitoring allowed the customer
chinery in a plant. However, an increasing to avoid a shutdown and inspection to determine the
number of customers are using our solutions cause. In this case, the software gave clear indication
more broadly to address all classes of assets of root cause as electrostatic discharge that had
across the plant. eroded part of the bearing and opened the clearances.
Although vibration was high, the cause was known
Why?
and the situation could be closely watched, allowing
Because they have discovered that the value of the machine to run until the next planned outage,
condition monitoring isn’t limited to just their critical nine months away. Maintenance expense savings?
turbomachinery. They have also discovered that the Essentially the labor costs that an inspection would
value of condition monitoring isn’t just in reducing main- have entailed. However, these costs were trivial in
tenance expenses, but in reducing operating expenses. comparison to the operating expense impact: three days
of lost production at a cost of several million dollars.
The operative word here is value. Customers use our
products and services because they find value in them. Consider these additional examples:
I’m continually reminded of this simple truth as I engage
• A customer uses System 1* software to monitor
in one of the most rewarding aspects of my job: visiting
performance and improve efficiency on a compressor
with customers. In most instances, our conversations
train, saving fuel. That’s an operating expense.
start with them sharing how our products and services
are delivering value. Obviously, I like that feedback. • A customer uses the Decision Support module in
There are also occasions when customers tell me that System 1 software to detect leaking rupture disks,
we have missed the mark. I appreciate that feedback as helping avoid regulatory fines and penalties. That’s
well because by taking the time to meet with me they an operating expense.
are indirectly sending the same message: they see value
• A customer uses our 3500 system to closely monitor
in condition monitoring—so much value that it’s worth
a problematic air blower, adjusting process conditions
insisting we get it right.
and loads to keep it running until a planned outage,
A moment ago, I mentioned that the value in condition avoiding the need to rent several portable units for
monitoring is not just the “M” in O&M expenses, but also make-up air. That’s an operating expense.
the “O.” Indeed, its impact on operating expenses is
• A customer uses our Machinery Diagnostics Services
often an order of magnitude larger than its impact on
(MDS) team to quickly diagnose the root cause of
maintenance expenses.
high vibration trips on a steam turbine as a thermally
Jeff Schnitzer – General Manager, Bently Nevada Asset Condition Monitoring – GE Energy – jeff.schnitzer@ge.com
1 4 O R B I T Vol.29 No.1 2009
…WHILE CONDITION MONITORING IMPACTS THE “M” IN O&M
EXPENSES, ITS IMPACT ON THE “O” IS OFTEN EVEN BIGGER.
induced rub, allowing them to meet dispatch You’ll also notice that several of these examples don’t
commitments rather than incur availability tariffs. even involve rotating machinery, which underscores
That’s an operating expense. another point I made earlier: our solutions are delivering
value on assets throughout the plant—not just on critical
• A customer uses our Trendmaster* Pro system to
turbomachinery. This is particularly encouraging because
monitor for leaks in relief valves, allowing rapid
we have worked very hard to develop solutions that will
intervention to minimize the loss of extremely expen-
also address the many other asset types in your plant.
sive finished product. That’s an operating expense.
As you consider where our condition monitoring
• A customer uses our remote service capabilities to
solutions can deliver value in your business, I encourage
upgrade the condition monitoring software on their
you to think as broadly as the customers showcased a
offshore platforms, saving weeks of labor, per-diem
few paragraphs ago. They think “plantwide,” not just
charges, and the costs of helicopter transportation
“critical turbomachinery.” And, they think of ROI for
to half a dozen separate facilities. That’s an
their condition monitoring investments in terms of
operating expense.
operating expenses, not just maintenance expenses.
These are just a few examples among dozens more I They have learned first-hand that while the savings in
could have cited. Nor are these contrived examples. Each maintenance expenses can be substantial, the savings
one reflects an actual success story—real customers, real in operating expenses can be even more substantial.
sites, real scenarios, and real operating expense savings. [Editor’s Note: The rupture disk and relief valve scenarios are discussed
in this issue of ORBIT on pages 60 and 68 respectively.]
RECIP TIPS
eciprocating compressor users frequently report that valve failures rank among the leading
causes of unplanned outages [1,2]. They apply a number of technologies to assess the condition
of the valve to better manage their compressors. One technique that has been around for years—
perhaps decades—is valve or valve cover temperature [3,4].
Properly understood and applied, this measurement provides valuable insight into reciprocating
compressor cylinder valve health. This article reviews the successes and limitations of this measurement
and discusses the three primary methods of monitoring valve temperature, comparing the advantages and
disadvantages of each.
Brian Howard, P.E. – Sr. Technical Manager – Reciprocating Compressor Condition Monitoring – GE Energy – brian.howard@ge.com
PROCESS WHERE IT IS
RE-COMPRESSION RESULTS
IN A TEMPERATURE
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 1 7
RECIP TIPS
200
100
0
08:56 08:56 08:56 08:56 08:56 08:56 08:56 08:56 08:56
12NOV2002 14NOV2002 16NOV2002 18NOV2002 20NOV2002 22NOV2002 24NOV2002 26NOV2002 28NOV2002
TIME : 12 Hours /div
TDC TDC
1500 1500
POUNDS PER SQUARE INCH GAUGE
1000 1000
100 psig/div
100 psig/div
500 500
0 0
0 20 40 60 80 100 0 20 40 60 80 100
5 %/div 5 %/div
Displaced Volume Displaced Volume
TDC TDC
1500
POUNDS PER SQUARE INCH GAUGE
1500
POUNDS PER SQUARE INCH GAUGE
1000 1000
100 psig/div
100 psig/div
500 500
0 0
0 20 40 60 80 100 0 20 40 60 80 100
5 %/div 5 %/div
Displaced Volume Displaced Volume
For an example of this phenomena consider a high- Over the next few days, the cover skin temperature of
pressure hydrogen cylinder instrumented with cylinder the distressed valve begins to drop. By 24 November,
pressure, discharge temperature, and valve cover skin the distressed valve cover skin temperature has fallen
temperatures. Figure 2 shows a valve failure progression to 215ºF. If valve temperature correlated accurately with
timeline for this cylinder. valve condition, one would expect the condition of the
valve to have improved.
The top left Pressure versus Volume (PV) curve shows
the cylinder pressure profile on 12 November. The plot In fact, as the PV diagram in the top right shows, valve
shows good agreement between the indicated cylinder condition has further deteriorated resulting in a signifi-
pressures and theoretical curves. Referring to the cant deviation between the indicated and theoretical
trend plot across the top of Figure 2, it can be observed curves as well as a further reduction in the compression
that on 12 November the discharge valve cover skin ratio of the cylinder
temperatures and the discharge temperature lie close
At this point, the rod load and rod reversals had dropped
to each other. Together, these observations indicate
near the limits recommended by the compressor OEM.
effective sealing by the piston rings and cylinder valves.
For this reason the plant shut the compressor down
On 13 November a leak develops in one of the crank end for overhaul.
discharge valves. This can be seen in the PV diagram
in the lower left of the plot where the actual pressure Secondary Temperature Effects
rises faster than the theoretical pressure. Valve cover of Valve Failure
skin temperature of the “LP Stage 2 Disch W” valve rises
The previous example focused the relationship between
quickly from 180ºF to 208ºF.
the temperature of the distressed valve cover and
At this point, the failure has a minimal impact on valve condition. The recirculation of gas at a particular
compression ratio. The valve failure did not adversely valve changes not only the temperature of the local
impact rod loads or rod reversals, so the plant decided valve cover, but also the temperature profile of other
to continue with operations. components of the cylinder.
By 18 or 19 November, the distressed valve cover skin A failing suction valve provides a good example of the
temperature reaches a maximum of 255ºF. The PV curve, secondary effects introduced by a valve failure. Figure
shown in the lower right of Figure 2, shows that the 3 shows the valve cover temperatures on the crank end
failure now begins to have a more noticeable impact in the left panes, and head end in the right panes. On all
on the compression ratio of the cylinder. The rod load trends, temperatures group together until the morning
and rod reversal of this cylinder and the other cylinders of August 19th.
servicing the compression stream were still acceptable,
so the plant continued to operate.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 1 9
RECIP TIPS
LP STG 1 Suct NW 14AUG2008 11:01:38 100 d LP STG 1 Suct NE 14AUG2008 11:01:38 108 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
LP STG 1 Suct W 14AUG2008 11:01:38 98 de LP STG 1 Suct E 14AUG2008 11:01:38 114 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
LP STG 1 Suct SW 14AUG2008 11:01:38 104 d LP STG 1 Suct SE 14AUG2008 11:01:38 103 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
LP STG 1 Suct Temp 14AUG2008 11:01:38 96 de LP STG 1 Suct Temp 14AUG2008 11:01:38 96 de
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
300 300
AMPLITUDE:
AMPLITUDE:
200 200
20 deg F/div
20 deg F/div
100 100
0 0
11:01 11:01 11:01 11:01 11:01 11:01
14AUG2008 18AUG2008 22AUG2008 14AUG2008 18AUG2008 22AUG2008
TIME : 12 Hours /div TIME : 12 Hours /div
LP STG 1 Disch NW 14AUG2008 11:01:38 175 d LP STG 1 Disch NE 14AUG2008 11:01:38 172 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
LP STG 1 Disch W 14AUG2008 11:01:38 184 d LP STG 1 Disch E 14AUG2008 11:01:38 177 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
LP STG 1 Disch SW 14AUG2008 11:01:38 186 d LP STG 1 Disch SE 14AUG2008 11:01:38 189 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
LP STG 1 Disch Temp 14AUG2008 11:01:38 208 d LP STG 1 Disch Temp 14AUG2008 11:01:38 208 d
Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3 Recip Compress From 14AUG2008 11:01:38 To 25AUG2008 11:01:3
300 300
SAMPLE FILTERING SAMPLE FILTERING
AMPLITUDE:
AMPLITUDE:
200 200
20 deg F/div
20 deg F/div
100 100
0 0
11:01 11:01 11:01 11:01 11:01 11:01
14AUG2008 18AUG2008 22AUG2008 14AUG2008 18AUG2008 22AUG2008
TIME : 12 Hours /div TIME : 12 Hours /div
System 1
Æ
Figure 3. LP stage 1 valve cover temperature trends.
Optimization and
Page 1 Diagnostic Platform
2 2
0.5 g/div
0.5 g/div
POUNDS PER SQUARE INCH GAUGE
G'S
600 0
-2 -2
20 psig/div
20 psig/div
-4 -4
500 500
2 2
1 1
0.2 g/div
0.2 g/div
G'S
400 0
G'S
400 0
-1 -1
-2 -2
300 300
0 100 200 300 0 100 200 300
20 Degrees/div
20 Degrees/div
Crank Angle
Crank Angle
Figure 4. Cylinder pressure and crosshead Figure 5. Cylinder pressure and crosshead
acceleration waveforms, before valve failure. acceleration waveforms, after valve failure.
Figure 4 shows cylinder pressure curves and crosshead with this valve cover. The rise in the “LP STG 1 Suct E”
accelerometer signals for this cylinder, typical for the temperature, adjacent to “LP STG 1 Suct NE”, results
time period prior to the morning of August 19th. The from the re-circulating gas heat effect spreading to
close agreement between the theoretical and indicated other valve covers.
pressure signifies effective cylinder trim sealing. Further,
The 20°F plus rise in the head end discharge valve
the high frequency crosshead accelerometer signal
group, “LP STG 1 Disch NE/E/SE” deserves attention as
shows only discrete events associated with normal
well. Either one or more of the discharge valves has
valve opening and closing.
a leak, or there is something about the leaking suction
Referring back to Figure 3, the consistency across the valve that changed the operating conditions of the
trend line ends on the morning of the 19th. At this point, discharge valves.
the plots show relative changes in temperature trends.
Figure 5 shows the indicated cylinder pressure curves
The “LP STG 1 Suct NE” trend line in top right pane
and crosshead acceleration after the suction valve leak
displays the most significant change; however other
began. The slower rise in pressure during the compres-
points also show changes. For example, the “LP STG
sion stroke on the head end indicates a leak from the
1 Suct E” and valve cover temperature rises as do the
cylinder to a low-pressure reservoir, such as the suction
head end discharge valve cover temperatures, “LP STG 1
manifold. The high frequency content crosshead accel-
Disch NE/E/SE.”
erometer waveform, shown on the top, shows a rise in
The sudden change in relative temperature values amplitude as the difference between internal cylinder
indicates a change in the sealing ability of the cylinder pressure and suction valve manifold pressure increases.
trim components. As discussed above, this results in This rise in amplitude results from internal cylinder gas
recirculation of gases and a local increase in valve cover leaking across the valve into the suction manifold. The
temperature. Given the relatively high change in the features of this plot confirm that only a suction valve
“LP STG 1 Suct NE” temperature relative to the other leak exists at this time.
changes, one can reasonably associate the valve failure
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 2 1
RECIP TIPS
With the possibility of a discharge valve leak eliminated, Review of Valve Temperature Installation
only the scenario of a leaking suction valve causing the Arrangements
rise in the discharge valve cover temperatures remains.
Three main approaches in valve temperature monitoring
At first glance, it seems unlikely that the suction valve
have gained acceptance. These three approaches are:
could impact the performance of the discharge valves.
The connection lies in the re-circulating gases under- 1. Valve cover skin temperature
neath the suction valve cover. While some of this gas 2. Valve cover temperature
does stay local to the valve cover, large portions of the 3. Internal valve temperature
gas re-enter the cylinder to be compressed, resulting in
The following sections describe the measurements in
a higher effective suction temperature for that end of
detail along with the advantages and disadvantages
the cylinder. Since the compression ratios remain the
of each approach. Table 1 on the following page
same on both ends of the cylinder, the discharge gas
summarizes the discussion.
temperature for the head rises with respect to the crank
end valve cover temperatures.
1. Valve Cover Skin Temperature
In this temperature arrangement, a small hole drilled
Relying on Valve Temperature and tapped in the valve cover provides anchorage for a
Alone for Cylinder Condition fastener securing a washer-style thermocouple to the
Valve temperature, combined with a trending tool, valve cover. Figure 6 shows this type of arrangement.
can provide a good indication of a failing valve at Obviously, this arrangement provides ready access for
the onset of failure. As the failure progresses, valve maintenance and reduced retrofit effort.
temperature becomes a poor predictor of valve health.
The approach does limit temperature sensor options
Valve leaks may also result in secondary temperature
as only thermocouple temperature sensors have been
effects in other parts of the cylinder, making it difficult
offered in this configuration. Further, it is not possible to
to confidently pinpoint the leaky valve. Further, it does
install an explosion-proof housing around the element,
not provide any insight into the forces acting on the
if plant hazardous area requirements dictate such an
compressor (i.e., rod load and rod reversal), making it
arrangement.
difficult to understand the stress the failure places upon
the compressor. Nor does cylinder pressure provide The impact of the ambient environment has the poten-
sufficient information to pinpoint which valve on a tial to reduce the effectiveness of the measurement.
particular end of a cylinder has failed. For these reasons, For example, consider the valve temperature mapping
valve temperature measurement’s primary value is as shown in Figure 7. This end of the cylinder has three dis-
a supporting evidence tool in PV analysis, but is not charge valves. Two of the valves, “LP Stg Disch NE” and
sufficient by itself to fully understand and manage the “LP Stg Disch NE”, lay at an angle with respect to the true
cylinder’s condition. horizontal axis. The LP Stg Disch E valve is horizontal.
Valve cover skin temperature Valve cover temperature Internal valve temperature
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 2 3
RECIP TIPS
300 TDC
INVALID DATA
800
POUNDS PER SQUARE INCH GAUGE
250
700
20 psig/div
10 deg F/div
AMPLITUDE:
200
600
150
500
100
400
50 300
19:46 19:46 19:46 19:46 19:46 19:46 19:46
0 20 40 60 80 100
30MAY2006 06JUN2006 13JUN2006 20JUN2006 27JUN2006 04JUL2006 11JUL2006
5 %/div
TIME : 24 Hours /div Displaced Volume
Figure 8. Head end head discharge valve temperature trends (left side) and cylinder PV curve (right side).
300
700
20 psig/div
500
250
400
10 deg F/div
AMPLITUDE:
300
0 20 40 60 80 100
200 5 %/div
Displaced Volume
700
20 psig/div 500
400
50
01:00 09:00 17:00 01:00 09:00 17:00 01:00
02JUN2006 02JUN2006 02JUN2006 03JUN2006 03JUN2006 03JUN2006 04JUN2006 300
0 20 40 60 80 100
TIME : 2 Hours /div 5 %/div
Displaced Volume
Figure 9. Valve cover skin temperature (left side) and cylinder PV curve (right side).
2. Valve Cover Temperature Valve cover temperature has the advantage of not
The valve cover skin temperature installation approach requiring significant cover modification; however, the
can be modified slightly to allow explosion proof hous- installation—especially in the case of the explosion-
ings as well as to reduce the effects of exposure. Figure proof fittings—somewhat complicates maintenance
10 shows two examples of this valve approach, referred activities compared to valve cover skin temperature
to as valve cover temperature. installations.
In either case, a dimple or shallow hole receives the Figure 11 shows a photo of a typical non-explosion proof
temperature-sensitive portion of the transducer. The installation. In this installation, a compression-style
installation shown in the top pane does not require tube fitting threads into the valve cover and secures the
explosion-proof fittings allowing a bayonet connector temperature element rather than a bayonet connector.
with an armored cable style temperature transducer Although this installation requires more effort than the
to be used. In the case where the plant hazardous valve cover skin temperature approach, valve cover
area classifications require explosion-proof fittings an temperature typically experiences less influence from
explosion-proof head is installed into the bracket and orientation and environmental effects. The reduced
flexible conduit run from this head to the junction box. external influence can be demonstrated by consider-
ing the data provided by the sensor arrangement of
Figure 11 on a large hydrogen booster compressor in a
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 2 5
RECIP TIPS
Figure 11. Valve cover temperature installation. Figure 12. Throw 4 valve cover temperature maps.
Valve #55 90° Left Temperature 05DEC2006 09:53:56 185 deg F NA 400
20 psig/div
SAMPLE FILTERING
200
250 100
0
0 20 40 5 %/div 60 80 100
Displaced Volume
200
1stStage-HE4 Synch 399.8 psig
10 deg F/div
0%
AMPLITUDE:
Displaced Volume
TRAIN K-20 From 09DEC2006 10:06:13 To 09DEC2006 10:06:13 Historical MACHINE SPEED: 360 rpm
1stStage-HE4 From 09DEC2006 10:06:13 To 09DEC2006 10:06:13 399.8 psig
Displaced Volume 0%
TRAIN K-20 Reference MACHINE SPEED: 360 rpm
TDC
150
400
20 psig/div
100
200
100
50
11:00 11:00 11:00 11:00 11:00
05DEC2006 06DEC2006 07DEC2006 08DEC2006 09DEC2006 0
0 20 40 60 80 100
TIME : 4 Hours /div 5 %/div
Displaced Volume
Figure 13. 1st stage head end valve temperature trend and head end PV curves.
100
10 deg F/div
AMPLITUDE:
50
0
11:00 11:00 11:00 11:00 11:00
05OCT2006 19OCT2006 02NOV2006 16NOV2006 30NOV2006
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 2 7
RECIP TIPS
200
1st Stg CE Suct #2 90° Left 28DEC2006 10:11:41 76 deg F NA Recip Compressor Tra
TDC
Reference MACHINE SPEED: 327 rpm
SAMPLE FILTERING
800
20 psig/div
140
600
400
200
120
0
0 20 40 60 80 100
5 %/div
AMPLITUDE:
5 deg F/div
Displaced Volume
1st Stg CE Pres Synch 643.2psig
Displaced Volume 0%
100 Recip Compressor Tra From 22JAN2007 11:23:43 To 22JAN2007 11:23:43 Historical MACHINE SPEED: 327 rpm
1st Stg CE Pres 643.2psig
Displaced Volume 0%
Recip Compressor Tra Reference MACHINE SPEED: 327 rpm
TDC
1200
50 psig/div
800
600
400
60
200
Figure 17. Crank end suction internal valve temperature and PV curves.
References
[1] Leonard, Stephen M. “Increasing the Reliability of Reciprocating Compressor on Hydrogen Service,” Hydrocarbon
Processing, January 1996.
[2] Manurung, Togar MP, et. al. “Reliability Improvement of a Reciprocating Compressor in an Oil Refinery.”
[3] Smith, Tim. “Quantum Chemical Uses Reciprocating Compressor Monitoring to Improve Reliability,” Orbit Magazine,
June 1996, pp. 13-16.
[4] Silcock, Don. “Reciprocating Compressor Instrumented for Machinery Management,” Orbit Magazine, June 1996, pp.
10-12.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 2 9
BACK TO BASICS
A
s industries contend with global IN PART 3, the final installment of this series,
competition, unprecedented economic we explore PCM in more detail with an overview
conditions, regulatory demands, environ- of both offline and online condition monitoring
mental concerns, and other pressures, systems and their role in PCM. Particular attention
the ability to manage their physical production is devoted to explaining the role of scanning-type
assets has become increasingly important. The online systems (both wired and wireless) as they
emphasis that industry now places on the asset pertain to moderate- and low-criticality assets,
management function can be readily noted simply and specific threshold criteria is offered to help
by observing the number of individuals that users determine when to move an asset from an
currently carry the title “Asset Manager” on their offline approach to an online approach. Part 3
business cards compared to just ten years ago. concludes with a discussion of the impact that
The asset manager’s first and most fundamental wireless technology has had on moving the online/
task in establishing an asset management pro- offline threshold further down the criticality scale,
gram is to identify the appropriate maintenance and explores the categories of assets most suit-
strategy(ies) warranted by each asset. Accordingly, able for wireless condition monitoring.
we are devoting a series of articles to this
important topic.
PROGRAM IS TO IDENTIFY
IN PART 2, we explore the relationship
between P-F curves and asset criticality in
more depth, showing how an asset’s criticality THE APPROPRIATE
governs the level of analysis rigor necessary to
choose an appropriate maintenance strategy. The MAINTENANCE STRATEGY(IES)
various analysis methods are discussed, including
Reliability-Centered Maintenance (RCM), Root Cause
Failure Analysis (RCFA), and Failure Modes and WARRANTED BY EACH ASSET.
Effects Analysis (FMEA), with guidance offered on
when (and when not) to apply each method. We
then review the four basic maintenance strategies
introduced in part 1, and establish the condition
monitoring methodologies and system require-
ments that correspond to each.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 3 1
BACK TO BASICS
Asset Management and the P-F Curve P-F curve as possible. This results in strong planning
and scheduling programs, with condition monitoring
Figure 1 shows an example of a Failure (P-F) curve, with
technologies as one of the key work identification
P representing the point in time when the potential
systems driving those programs. The maintenance and
failure can be detected, and F representing the point in
reliability teams spend the right amount of time identify-
time the asset reaches functional failure. While not all
ing failures earlier for each asset, enabling improved
failures manifest themselves in this manner, most failure
coordination to better plan and schedule maintenance
modes do have technologies that can detect failures
activities. Today’s challenging economic environment,
early in their failure cycle. The intent is to manage assets
combined with the costs of HSE (Health, Safety, and
at the top of this curve.
Environmental) and regulatory compliance, makes it
While most companies strive toward managing their imperative to optimize the return on investment for
assets proactively, many plants often find themselves maintenance activities.
managing assets in a reactive mode. They are
continually reacting to assets reaching functional failure Maintenance Strategies
without warning. This situation often results in spare
Below, we summarize the four fundamental mainte-
parts shortages due to limited planning time, increased
nance strategies in use today. It is important to note
overtime and callouts, and poorer quality repairs
that multiple strategies may be (and often are) applied
and documentation. All of this can inhibit the plant
to a single asset. The strategy(ies) chosen for a par-
from having the time and resources to complete the
ticular asset are a function of its criticality—a concept
repertoire of maintenance routines and move into a
explored later in this article—and its failure modes and
more predictive mode.
consequences. Certain failure modes, for example, may
have serious consequences, but are not be detect-
Best-in-Class Practices able by any currently available condition monitoring
Best-in-class maintenance and reliability performers technology. Routine quantitative inspections must be
typically manage the majority of assets as far up the carried out instead. Other failure modes on the same
P6
Smoke
0–0 days
P7
Point where asset
stops doing what its
P–F Interval users want it to do is
Failure (F)
Figure 1. Potential Failure to Actual Failure (P-F) Curve.
asset may have equally serious consequences, yet can through a condition monitoring (CM) program. It allows
be readily detectable by a particular technology, such planners and schedulers to schedule maintenance
as thermography. Such an asset would employ a mix of when it is most cost-effective and precedes functional
preventive maintenance (e.g., time-based inspections) failure. A condition monitoring program monitors the
and predictive maintenance (e.g., thermography). health of the asset early in the P-F cycle and helps
identify the required maintenance work. Ideally, this
1. Preventive Maintenance (PM) allows reliability and maintenance professionals to
A PM strategy is often based on OEM recommendations conduct all required maintenance on a given piece of
for specific production assets, with preventive main- equipment simultaneously, saving costly downtime.
tenance performed at specified time-based intervals.
CM programs have often been characterized as
The intervals are generally based on the MTBF (Mean
“vibration monitoring” programs. However, while
Time Between Failure) data compiled by the OEM. PM
vibration monitoring is a critical aspect of most CM
includes intrusive time-based inspections and requires
programs, there are more than 75 different types of
taking the asset out of service and opening it to look
non-intrusive CM technologies, such as oil particulate
for worn parts or incipient failures. Often, since an
analysis, temperature, thermography, and ultrasonics.
asset is opened for inspection, wearable parts may be
The selection of CM technology(ies) and corresponding
replaced even though they do not show wear. Also, any
data collection/analysis frequency is done on an asset-
intrusive maintenance has the potential of imparting
by-asset basis by considering both the asset’s criticality
maintenance-induced failures, often called infant
and it’s failure modes.
mortality. Since asset failures can happen between
scheduled maintenance intervals, a strictly time-based The goal of predictive maintenance is to use condition
strategy may not be right for many assets and certain monitoring technology to detect future failures through
failure patterns. the evaluation of early warning indicators. A robust PdM
program has many benefits over other maintenance
Quantitative preventive maintenance incorporates
strategies (see Table 1).
non-intrusive predictive maintenance inspections into
a preventive maintenance program. Using gauges to
measure belt tension on a motor-driven blower is an Table 1. Attributes of PdM/CBM.
example of a non-intrusive inspection to detect pending Benefits Methods
failure. In this case, tension can be measured without • After initial hardware • Compares physical
opening the machine and is one indicator of excessive costs, less expensive measurements against
roller, bearing, or belt wear. recurring maintenance engineering limits and
costs than preventive specifications
While such inspections are non-intrusive, they are still maintenance or • Tracks trends to detect
reactive maintenance changes from normal
time-based. Further, they cannot be performed on an (see Figure 2) • Uses state-of-the-art
asset unless the asset is shut down for testing. Therefore, • Early detection of technology to discover
while it does not result in the typical problems inherent in failures generates failures
maintenance work • Allows alarm limits to
intrusive inspections, it still results in a loss of production
plans, resulting in more be established
while the asset is shut down. This, in turn, may reduce the planned work rather • Employs continuous
life cycle of those assets for which starting and stopping than unplanned online monitoring,
incurs greater wear than steady-state operation. • Failure identification intermittent online
results in less downtime monitoring, or manual
during maintenance data collection,
2. Predictive Maintenance (PdM)/ depending on the
Condition-Based Maintenance (CBM) asset failure cycle
PdM and CBM are synonymous. A CBM program evalu-
ates machinery via instrumentation, either periodically
or continuously, to determine its condition, usually
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 3 3
BACK TO BASICS
Once a robust PdM program is in place for an asset, PM Unfortunately, for many companies without good PM
routines can be reviewed and in many cases optimized and PdM programs, RM is not a deliberate strategy
or eliminated. As an example, assume that a particular applied to only selected assets; it is instead a vicious
OEM recommendation is to change a bearing after cycle where daily maintenance activities are dominated
30,000 hours of operation. With proper application of by unforeseen failures, hindering the transition to a
predictive technologies, the health of the bearing can be more proactive approach for managing assets.
monitored and managed to a high degree of reliability.
Indeed, RM may consume up to 80 percent of the total
By managing the health and condition of the asset,
time and budget of companies stuck in this mode.
time-based maintenance intervals can be extended or
eliminated altogether. Referring again to Figure 2, RM (i.e., RTF) also has the
dubious distinction of being the most expensive type of
Finally, PdM is not only more effective in driving early
maintenance when applied indiscriminately to all assets
warning and the ability to plan and schedule properly,
in the plant.
but it is also less costly than reactive and time-based
strategies (Figure 2). 4. Proactive-Centered Maintenance (PCM)
A one-size-fits-all approach utilizing RM has already
been shown to be the most expensive and least
effective maintenance strategy when indiscriminately
Cost of Maintenance applied to all assets. However, the same can be said
EPRI Power Generation Study (Modified)
for both PM and PdM. Simply applying any particular
18 strategy to all assets—independent of the asset’s
16
14 criticality—is non-optimal. PCM recognizes this and
$/HP/Year
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 3 5
CASE HISTORY
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 3 7
CASE HISTORY
Machine Description
The machine detailed in this case history is a 1547
kW back-pressure steam turbine (Figure 1) driving a
high-pressure multi-stage boiler feed pump at 4200
rpm. The turbine is forced lubricated, incorporates
tilting-pad thrust bearings at the steam end, and uses
a Woodward® Peak® 150 governor for speed control.
The entire train is monitored continuously via a Bently
Nevada 3500 Series Machinery Protection System
connected to System 1 software.
Considered one of the world’s leading petrochemical Diethylene Glycol (DEG), which are marketed by
producers, EQUATE Petrochemical Company is a MEGlobal, a separate PIC and Dow joint venture.
Kuwait-based international joint venture between
To meet ever-increasing worldwide demand
Petrochemical Industries Company (PIC), The Dow
for excellent petrochemical products, EQUATE’s
Chemical Company (Dow), Boubyan Petrochemical
shareholders have completed a multi-billion
Co m p any (BP C) an d Q urain Pet ro ch emical
expansion project which greatly increases the
Industries Company (QPIC).
existing production of PE and EG.
As a world-class petro-
Emerging as Greater EQUATE,
c h e m i c a l p r o d u c e r, EQ UAT E
this $3 billion global-scale
currently provides markets in
venture added SM to EQUATE’s
the Middle Eas t, A sia, Africa
product por tfolio. Paraxylene
and Europe with high-quality
and Benzene are also produced
petrochemical products.
at the same location from a new
E Q U AT E p r o d u c e s — a n d i t s wo r l d - cl a s s A ro m a t ic s p l ant
commercial face, the EQUATE which is managed by EQUATE.
Marketing Company (EMC),
Hand-in-hand with being a
markets—a wide range of
leading market player, EQUATE is
p o lyet hy l en e (PE ), f ro m lin e ar l o w d en si t y
an empowered, competency-based organization
(LLDPE) to high molecular weight high-density
applying best practices and leading technology
resins (HDPE). Recently EMC has also begun
i n m a r ke t i n g , s a l e s , re cr u i t m e nt & ca re e r
selling Styrene Monomer (SM) produced from
development, environment, health & safety,
Greater EQUATE’s expanded facilities. EQUATE
industrial operations, and administrative affairs.
also produces Monoethylene Glycol (MEG) and
CASE HISTORY
Figure 8. The shaft centerline plot (a) at left depicts an almost instantaneous change in the shaft’s average
position as can be noted by the timestamps which are all within two seconds of one another. The amplitude
trend (b) at upper right provides another view of this sudden change, as does the timebase waveform (c) at
lower left which assumes the shape of a step change upon which is superimposed the AC vibration signal.
Figure 10. Trend of vibration data during final startup (bottom) shows vibration levels below 60 microns
pk-pk (2.5 mils pk-pk), consistent with normal levels. The shaft centerline plot at upper left also shows normal
response, with the centerline climbing up the oil wedge as the shaft speed increases from 0 to 4200 rpm. For
a shaft rotating in the counter-clockwise direction, it will slowly climb up the right wall of the bearing with
increasing speed, exactly as shown, before assuming a stable position at running speed.
To test this hypothesis, the turbine was started slowly by units can only run when this unit runs. Without dynamic
manually controlling the steam admission. As expected, data capture and a full complement of plot types such as
the turbine ramped slowly to full speed with very low shaft centerline, slow-roll timebase, and spectrum, the
vibration levels and a normal shaft centerline response problem would have been extremely difficult to diagnose
as shown in Figure 10. and would have taken considerably longer.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 4 3
APPLICATIONS
APPLICATION
Considerations for
EDDY-CURRENT
Proximity Probes
he plant’s rotating machinery engineer for suitability, mounting brackets would have to be
breathed a sigh of relief. After nearly half designed, and a host of other details would have to
a decade of having his requests turned be properly engineered for the application at-hand.
down, he had finally gotten capital Unfortunately, the rotating machinery engineer
budget approval to retrofit an aging assumed that his biggest hurdle was to obtain
compressor with proximity probes. He understood budget approval; the myriad of application details
the value of the data they could provide, as many were ignored simply because he didn’t know what
of the newer machines in the plant were fitted with he didn’t know. When the items finally arrived and
probes and he used the information regularly to plan were installed, it was discovered during system
maintenance and diagnose mechanical problems. check-out and commissioning that the transducers
were calibrated to the wrong target material,
However, unlike the machines in newer sections
that the probe holes were drilled and tapped
of the plant that had been supplied from the OEM
incorrectly – introducing crosstalk and counterbore
with probes already installed, the machine he
interference – and that the cable lengths were
was now retrofitting dated back to the 1960s
mismatched. While it eventually got sorted out, the
and did not have provisions for probes. As a result,
project took three times as long as it should have
bearing housings would have to be drilled and
and cost nearly twice as much to complete.
tapped, shaft surfaces would have to be checked
Want to avoid a similar scenario in your plant? You can, 1. Target material
but only when you have a firm grasp of the fundamen- 2. Target geometry
tals of eddy-current proximity probes and application 3. System length
considerations that can spell the difference between 4. Mounting considerations (e.g., counterbore,
failure and success when instrumenting a machine. This bracket resonances, installation convenience)
article is designed to help you do just that. 5. Environment (e.g., temperature, chemicals,
radiation)
Principle of Operation 6. Frequency response
7. Linear range
An eddy-current probe works by passing an alternating
8. Signal-to-noise ratio (e.g., device resolution,
current through a coil of wire and measuring the coil’s
electrical runout)
impedance. This impedance changes when the probe is
1
9. Field wiring
brought near an electrically conductive material and the
impedance change is proportional to the physical gap
1. Target Material
between the coil and the conductive target. The sensing
The impedance change of the probe is dependent
electronics in turn convert this impedance change to a
on the electrical properties of the target material. All
voltage, providing an electrical output directly
probe systems state the material to which they are
proportional to physical gap.
calibrated (typically AISI 4140 steel). Although some
While the operating principles are simple enough, that products claim to be insensitive to target material, these
is where the simplicity ends. A probe must be rugged, should be approached cautiously as there are usually
accurate, linear, repeatable, interchangeable, and performance compromises involved. The point here is
impervious to its environment. Building such a probe that you must know the target material and verify that
is daunting. Building millions of such probes is even the probe system is calibrated appropriately, or at least
more daunting. Advanced, proprietary manufacturing that the performance deviation is known and that the
processes involving precision winding and sealing transducer system can be compensated to account for
techniques have been perfected over more than 50 this deviation. Generally, the factory should be consulted
years to allow mass production of eddy-current probes to quantify this performance deviation and to offer
that possess all the necessary attributes demanded by advice on whether the deviation would be excessive
both industrial and laboratory applications. for the intended application. When the deviation is
unacceptably large, transducer systems calibrated
Likewise, applying eddy-current transducers requires
to the specific target material can be supplied.
careful attention to detail. Only by understanding the
factors that affect the probe’s performance and
applying the probe accordingly can one be assured
that the installation will perform as specified. This
article discusses each of the following nine basic
factors that the installer must consider:
APPLICATIONS
2
Typically, the eddy-current system is designed to frequently observe collars or ramped features on
operate with a flat target. Rotating shafts of sufficient the shaft. The size and shape of the collars must be
diameter introduce no significant error, but there can adequate or the probe will not function properly (Figure
be problems if the shaft is too small. The effect of an 1). The probe may have scale factor problems and can
undersized shaft is that the sensor scale factor will be also pick up radial vibration of the shaft if the probe is
reduced; consequently, vibration levels will be under too close. The probes used in these applications are
reported. The minimum shaft size or target diameter for typically larger (11, 25, and 50 mm tip diameters). The
flat targets should be stated on the product data sheet. data sheet and manuals for these products contain
Look for it and take note. the details for correct application.
Figure 1. Differential expansion probes observing a ramp (left) or collar (right) must be selected and installed
carefully to ensure that the probes observe only the intended portion of the shaft. If a radial vibration probe is
installed near a collar, it must be located adequately distant from the collar to ensure it observes only the shaft
—not the collar.
Crosstalk is another issue associated with the target The symptom is that a ‘beat frequency’ will be evident
geometry. For machines using probes in an XY configu- on the sensor output. This beat frequency is dependent
ration, the two probes can interfere with each other if on probe gap and will exhibit frequencies in the range of
they are too close. The mechanism that causes this is several hundred Hz to a few kHz.
when the magnetic fields emanating from the probes
are at approximately the same frequency (Figure 2).
3. System Length
The probe, Proximitor* sensor, and extension cable are
expressly designed to work in specific combinations.
The probe and extension cable have a very tightly
controlled impedance characteristic and the Proximitor
sensor will only work for that specific probe/cable com-
bination within a particular probe product family (e.g.
3300XL 11 mm). Mixing either the lengths incorrectly or
parts from different product families (such as a 3300XL
8 mm probe connected to a 3300XL 11 mm extension
cable) will result in problems. For this reason, the probes,
Proximitor sensors, and extension cables within a
particular product family are usually color coded [1].
3
meters. When connected to a 5-meter Proximitor sensor,
the output will be at least 20% higher than expected.
Similarly, a 5.5-meter system would have an output
lower than nominal.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 4 7
APPLICATIONS
4
that are too close to the probe can affect the signal. It is
common for probes to protrude directly through metal
to view the target shaft. In this case, a counterbore
should be employed to avoid metal too close to the
probe tip (Figure 4). Any metallic machine components
that protrude into the region defined by recommended
counterbore spacing requirements should be removed if
possible. Recommendations vary by transducer model.
• Bracket Resonances Figure 5. The probe on the top has a long unsup-
Brackets or probe mounting hardware should always ported length, resulting in mechanical resonance
be chosen to provide mounting and support that is and erroneous readings as the probe is no longer
stationary with respect to the shaft it observes.
as rigid as possible. In some instances, probes with a The probe on the bottom is identical, but has
long metal casing (for example, a probe with five or a mounting bracket to support and stiffen the
more inches of thread) can be prone to resonance at probe “stinger.” This ensures that the machine’s
vibration will not excite the probe mounting’s
transient or steady state operating speeds. This is due
resonance.
primarily to the physical behavior exhibited by a tube or
• Installation Convenience Ether Ether Ketone (PEEK) tips and are widely compatible
with steam and petroleum products. If the environment
For applications requiring long probes or installation
contains anhydrous ammonia (NH3) or other reactive
in tight spaces, a smooth body probe should be consid-
substances, probes manufactured with a ceramic tip
ered. As the name implies, smooth body probes have
can be used to resist corrosion or degradation. The
no threads and this eliminates the need to laboriously
ceramic tip probes are also excellent for highly pres-
thread the probe into a bracket. The probe may be
surized applications where the probe is required to seal
slid into place and secured in position using a special
differential pressure.
clamping-type bracket. For threaded installations,
brackets should be designed to match the thread For nuclear power plant or research applications involv-
6
tolerance of the probe in order to avoid binding. It is also ing exposure to radiation, specially designed probes are
good design practice to ensure that there is sufficient available which can withstand Gamma radiation.
threaded material for a solid joint. A good rule of thumb
is five full threads of engagement. 6. Frequency Response
Typical machinery vibration monitoring applications
5. Environmental Considerations require the probe to provide vibration data at one or
A key consideration when selecting probes is the two times the running speed of the machine. This is
5
temperature for each location on the machine that will normally well within the ability of the probe to respond.
be occupied by the sensors. The temperature ratings of Occasionally there are machines that run very fast
the probe tip, cable, and Proximitor sensor are all differ- or have a need to monitor at a higher multiple of the
ent and should be considered separately. The machine running speed (e.g., blade pass frequencies for pumps
temperatures must be within the rated temperature or turbines). The response of the proximity probe will be
limits of the probe, extension cable, and Proximitor sen- attenuated at higher frequencies. Check the data sheet
sor during idle, startup, shutdown, and (especially) base of the probe to find the maximum frequency response.
running conditions. This can be particularly important for speed sensing
applications when a sharp trigger is required and
The system must also be undamaged by any of the
multiple harmonics of running speed are required
process gases or liquids present. Standard probes
to give a crisp square wave output [2].
commonly feature PolyPhenylene Sulfide (PPS) or Poly
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 4 9
APPLICATIONS
7
The probe must stay in the linear range at all times in thrust or expansion measurements is severe as it will
order to provide a valid measurement. typically mean that the machine can fail in a manner
that allows rotating and non-rotating parts to contact,
The consequence of choosing a probe with insufficient
such as blades and casings.
linear range or gapping the probe incorrectly is that the
monitor system will enter a “NOT OK” condition or there
will be data errors instead of protecting the machine.
Figure 6. Actual probe voltage versus gap response (red) overlaid on expected output (blue) showing that the
probe response is highly linear.
8
devices, but they do not have infinite resolution. There of a typical sensor is usually only a few volts. As such,
is always a noise floor for any electrical device below it is important to maintain a clean signal path to the
which any measurement is not reliable. Typically this monitoring system. Magnetic fields from power cabling
value is several tens of millivolts for a proximity probe and transformers can cause a significant voltage to be
system. Check the data sheet of the product and induced in the sensor system, introducing noise that
9
consider the resolution required of the system. can cause erroneous readings. Cabling for sensors
should always be routed away from power cables. If the
“Electrical Runout” is another problem that may be
instrumentation field wiring must cross power cables it
encountered in an installation. This phenomenon is
should cross perpendicular to the power cables in order
caused by variation in the material properties of the
to avoid coupling the electromagnetic field.
shaft appearing as a spurious vibration signal. The
magnitude of electrical runout can be quite high The frequency response of the sensor is also strongly
depending on the quality of the shaft material. The affected by the amount of capacitance in the wiring
causes and cures of electrical runout have been between the sensors and the monitor [8]. Extra
discussed in previous ORBIT articles [6,7]. capacitance on the output of the sensor acts as a filter
and reduces the magnitude of the high frequency
components of the signal. DC and low frequency
components are not affected. Reference Figure 7 to see
a graph showing the effects of added capacitance.
Figure 7. Added capacitance of long field wiring lengths will affect the frequency response of the probe system,
behaving essentially as a low-pass filter. This effect must be carefully considered, particularly when the probe
system will be required to detect very high frequencies, such as observing a toothed wheel or detecting vibra-
tion harmonics on a high-speed machine such as a turbo-expander.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 5 1
APPLICATIONS
The Mandate
Emerging European legislation will soon require offshore
operators to monitor and report gas turbine emissions,
with numerous other countries expected to follow suit
(see sidebar article on page 55).
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 5 3
NEW PRODUCTS AND APPLICATIONS
System 1 Display
— Real time and
historical data
Server PC
• Database
• Rules Engine
— Real time emissions
calculations
• System 1 PEMS Solution
• Bently PERFORMANCE
Solution Details
GE’s PEMS solution is implemented in System 1* A calibration feature allows customization to a specific
software and requires the following: installation along with post-installation refinement of
model outputs based on tuning adjustments derived from
• System 1 software, version 5.x or later
temporary CEMS and corresponding turbine data sets.
• Bently PERFORMANCE* System Extender module
Other Gas Turbines
• PEMS configuration, tuning, and installation services
In addition to the LM2500 SAC, a model specific to the
For users with existing System 1 and Bently
LM6000 SAC engine is also available. However, the flex-
PERFORMANCE software installations, the PEMS option
ibility of our PEMS solution allows us to accommodate
is a simple add-on module. In most cases, it can even be
other gas turbines as well. We invite you to contact us
installed remotely from an onshore location.
regarding your specific requirements and engine type(s).
The PEMS module employs a GE-developed “first
principles” emissions model for the LM2500 SAC gas Features and Benefits
turbine. This physics-based model uses fundamental • Lower Cost of Regulatory Compliance
engine parameters and conditions known to correlate to A PEMS solution reduces the installation and operating
emissions output, drawing on extensive test results from costs compared with a CEMS-only solution. The installed
the large installed base of LM2500 SAC units. It uses capital costs for CEMS on a remote platform can be as
ambient conditions, various turbine operating param- much as three times higher than PEMS.
eters, and fuel properties as inputs and is valid for both
gas and diesel fuels. The model computes emissions • Local and Remote Access
estimates for NOX, CO, and unburned hydrocarbons, The PEMS and System 1 Display client can be installed
while providing calculated values for CO2 and SO2. wherever (i.e., on or off platform) access to the required
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 5 5
NEW PRODUCTS AND APPLICATIONS
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 5 7
NEW PRODUCTS AND APPLICATIONS
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 5 9
NEW PRODUCTS AND APPLICATIONS
The Problem
PRVs are particularly susceptible to leaks and the
consequences of such leaks can be quite serious:
• Dangerous operating conditions, increasing the The Solution
risk to plant and personnel
Cost-effective and reliable online technology is now
• Damage to the environment
available to monitor critical PRVs for the presence of
• Regulatory fines/penalties
leaks. An online approach means that leaks are caught
• Negative public image/reputation
within minutes, incident severity (and associated
• Reduced process efficiency and profitability
consequences) is minimized, and costly routine inspection
• Loss of raw, intermediate, or finished materials
regimens are eliminated.
While most companies expend considerable resources
GE’s PRV Leak Detection System (see overview diagram
to detect and mitigate leaks, usually in the form of
on facing page) uses the Bently Nevada Trendmaster*
routine inspection or observation regimens, current
Pro architecture to gather periodic, online temperature
approaches have not proven entirely effective and
and acoustic measurements and automatically analyze
PRV leaks remain a significant challenge for industry.
them in System 1* software. By using both temperature
Additionally, PRVs are frequently subjected to unneces-
and acoustic measurements in complementary fashion,
sary maintenance. This occurs when a leak cannot be
more reliable detection is available than when using
confirmed, but to err on the side of caution, preventive
either technology alone. These measurements are
repairs, replacements, or overhauls are carried out “just
then processed through a special RulePak in System 1
in case.” This can drive the relative maintenance costs
software’s decision support engine. This RulePak contains
for PRVs very high.
sophisticated, embedded knowledge of PRV leak behavior,
automatically detecting the presence of a leak.
Continued on page 62.
Vent to atmosphere
FlexiTIM for
Ultrasonic
Inlet
temperature
ProTIM-C for
thermocouples
DSM supports up to 8
lines with up to 63
valves per line
Pressurized asset
Trendmaster Pro
platform
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 6 1
NEW PRODUCTS AND APPLICATIONS
PRV BASICS…
PRVs are designed to open at approximately Leaks can also result from other conditions,
10–20% above normal design pressures for not just a failure to reseat properly. For
the protected asset, venting excess pressure example, PRV internals can slowly corrode
to the atmosphere, a flare stack, lower- over time, eventually developing a leak. This
pressure containment system, or other means. failure mechanism is a particular concern
Once pressure returns to normal ranges, the when highly corrosive process fluids are
PRV is designed to reseat itself using a self- present. In such situations, a so-called “rupture
closing mechanism, typically incorporating disk” is often used to isolate the PRV from the
an internal preloaded spring or weight. process fluid, helping to preserve PRV integrity.
Unfortunately, PRVs can fail to reseat properly, [Editor’s Note: When a rupture disk is used in conjunction
with a PRV, online technology can be used to detect
resulting in leaks. While failure to reseat can be leaks in both the PRV and its associated rupture disk. For
caused by many conditions, one of the most more information on the role of rupture disks, and how
common is the presence of particulates (such an innovative customer is monitoring their disks, see
as rust or scale) suspended in the process gas the article “Using Decision Support to Reduce OpEx at
Valero Paulsboro” on page 64 in this issue of ORBIT.]
that becomes lodged between the valve
nozzle and the sealing disk.
www.bentlyuserforum.org
Vo l 2 9 N o. 1 2 0 0 9 ORB I T 6 3
COVER STORY
Op Ex
Valero’s Paulsboro refinery makes innovative
use of the Decision Support capabilities within
System 1* software to proactively manage
both rotating and non-rotating production
assets. As a result, operating expenses—not
just maintenance expenses—are reduced.
Jayesh Patel – Reliability Engineer – Valero Energy Company – Paulsboro Refinery – jayesh.patel@valero.com
Al Gomes – Manager – Reliability and Inspection – Valero Energy Company – Paulsboro Refinery – al.gomes@valero.com
John Kingham – Field Applications Engineer – Bently Nevada Asset Condition Monitoring – GE Energy – john.kingham@ge.com
V
alero’s Paulsboro Refinery, Managing Below the Alert Level
located on 950 acres in
One of the keys to Paulsboro’s success with condition-
Paulsboro, NJ, was first
based maintenance is their practice of managing
commissioned in 1917, and
machinery “below the alert level”—a euphemism
currently has a capacity of
describing the use of condition-based alarms that
195,000 barrels per day. The
are well below operator-level “alert” or “danger”
refinery has a fuels crude
alarms normally associated with pre-shutdown and
unit, lube crude unit, distillate hydrotreater, FCC
shutdown conditions respectively. Alarms set purely
complex, Delayed Coker, Reformer, Lube Plant and
to notify machinery specialists of impending problems
Alkylation unit. The refinery employs nearly 550
allow operations to continue uninterrupted while
individuals and is an OSHA VPP Star site.
appropriate actions are taken “behind the scenes,”
Throughout the refinery’s years of operation, such as scheduling appropriate maintenance, planning
its maintenance strategies and practices have an outage, or recommending changes to operating
continually evolved. Today, condition-based or process conditions that will slow or halt failure
maintenance is used extensively and employs progression on the affected asset. This level of “preemptive”
a mix of permanent and portable technologies, maintenance drives the Paulsboro refinery to the far left
depending on asset criticality. Low criticality limits of the P-to-F curve (Figure 1). In addition to fewer
assets are addressed by a portable data process interruptions, operating at this end of the P-to-F
collection system. High- and mid-criticality assets curve results in higher product quality, improved asset
are addressed by online systems. For its most availability, and increased operating profits.
critical assets, Paulsboro uses Bently Nevada*
continuous machinery protection systems in IN THE TWO YEARS SINCE
conjunction with System 1 software. The assets
in this category include the plant’s gas turbines THIS SIMPLE—BUT EFFECTIVE—
in power generation service, steam-driven and
motor-driven centrifugal compressors, hydrogen DECISION SUPPORT RULE
reciprocating compressors, utility air compressors,
and liquid ring compressors for flare gas recovery. HAS BEEN IMPLEMENTED,
Mid-criticality assets in the refinery’s coker unit
are addressed by a Bently Nevada Trendmaster*
IT HAS BEEN INVOKED TEN
system, a permanently wired “sensor bus”
TIMES, RESULTING IN TEN
architecture that takes condition monitoring
measurements several times per hour. Both PLANNED, ROUTINE, FILTER
the continuous monitoring systems and the
Trendmaster architecture are tied into System 1 REPLACEMENTS WITHOUT A
software for a unified online condition
monitoring environment. SINGLE “URGENT/EMERGENCY”
WORK ORDER BEING NECESSARY.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 6 5
COVER STORY
P6
Smoke
0–0 days
P7
Point where asset
stops doing what its
P–F Interval users want it to do is
Failure (F)
Figure 1. The so-called “P-F curve” showing qualitative time relationship between Potential Failure (P) and
Functional Failure (F). P0 represents the earliest time at which an impending failure could conceivably be
detected. P1 through P7 represent various detection methodologies and their typical capacity for advance
warning. The further to the left (closer to P0) one can operate for any given asset, the easier it is to plan
maintenance and lower the likelihood of “surprise” functional failures.
While establishing a sufficient number of condition Historian into the System 1 database, Paulsboro is able
monitoring alarms can be very effective for the reasons to apply the Decision Support engine in innovative ways
noted above, managing too many alarm levels can such as:
become quite onerous and a balance must be found in
• Analyzing and detecting anomalies on assets for
the quest to move “farther to the left” on the P-F curve.
which only process measurements are available
One way to achieve this is by relying not only on level-
—such as filters, heater tubes, and pressure relief
type alarms, but also on technologies that automate
valve rupture disks.
the data analysis and anomaly detection processes that
human experts would use if manually reviewing data. • Addressing applications outside of conventional
Paulsboro has been particularly innovative in their use “condition monitoring,” such as stack emissions
of System 1 software’s Decision Support capabilities in monitoring.
this respect. The Decision Support module in System 1
• Detecting problems in non-rotating portions of
allows users to embed their subject matter expertise for
turbomachinery, such as gas turbine hot gas path
a particular asset (or class of assets), and detect asset
components.
problems automatically. While many users employ the
Decision Support module to detect anomalies with the To address this mix of conventional rotating machinery,
rotating machinery monitored by System 1 software, fixed equipment, and process-related applications,
what has set the Paulsboro facility apart is their use of Paulsboro uses both GE’s machinery expertise in the
the system on non-rotating assets as well. By bringing form of pre-configured RulePaks, and their own exper-
process data from the plant’s Distributed Control tise in the form of custom rules written by their resident
System (DCS), turbine control systems, and Process subject matter experts. In this way, Valero has created
Decision Support capabilities tailored specifically to Should the differential reach 15 PSI, an alert, or severity
the needs of their Paulsboro operations. As detailed in level 3 alarm is sent. At this level, the work order will be
the following six case history synopses, the system is upgraded to “urgent/emergency.” Finally, if the differential
delivering substantial value and has become an integral pressure reaches 30 PSI, a level 4 “Danger” notification
part of their operating and maintenance decisions. is sent, instructing the operator to immediately open the
bypass valve to prevent damage to the unit.
Case Histories In the two years since this simple—but effective—
Decision Support rule has been implemented, it has
CASE HISTORY #1:
Naphtha Hydrotreater Feed Filter been invoked ten times, resulting in ten planned, routine,
The filter at the front end of the Naphtha Reformer filter replacements without a single “urgent/emergency”
has a direct effect on the quality of the final product. work order being necessary.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 6 7
COVER STORY
To isolate the PRV from the corrosive process gas, • Even a defective PRV (such as one that does not
a rupture disk is used. The disk is a sacrificial, thin fully seal itself in its valve seat) will not leak to the
metal membrane designed to fail at a predetermined atmosphere under normal process conditions
pressure. Further, it is designed using materials that are
Under normal conditions, an intact rupture disk isolates
impervious to the corrosive process gases. As shown
the PRV from seeing process pressures; thus, PPRV ≈ 0.
in Figure 2, the disk is placed in series with the PRV. This
A fully ruptured disk results in PPRV = PProcess. A faulty
arrangement ensures the following:
disk, such as with a pinhole leak, results in PPRV ≠ 0. Thus,
• The PRV will not normally be exposed to the cor- anytime PPRV is non-zero under normal process condi-
rosive process fluid/conditions, thus preserving the tions, the disk is either ruptured or faulty (e.g., pinhole
PRV’s integrity leak, etc.) and corrective action needs to be taken.
Figure 2. Pressure relief system consisting of a rupture disk in series with a Pressure Relief Valve (PRV). Both
rupture disk and PRV are designed to open at over-pressure conditions, protecting the pressurized asset by vent-
ing to the atmosphere. By using a rupture disk made of corrosion-resistant materials, the PRV is isolated from the
corrosive effects of the process gases, preserving the integrity of the PRV. The rupture disk has the added benefit
of ensuring that process gases will not leak to the atmosphere at normal pressures, even if the PRV is leaky.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 6 9
COVER STORY
To help Paulsboro manage their rupture disks and ensure Further, using the Decision Support module’s ability to
regulatory compliance, a pressure sensor (or switch) is generate management-level alarms, violations are kept
installed in the spool piece to measure PPRV. A simple rule in check by notifying personnel whenever emissions
in the System 1 Decision Support engine detects when approach maximum allowable limits. At 80% of the
PPRV is non-zero, and sends an e-mail notification to the violation value, an alarm occurs, automatically notifying
engineers responsible for the affected area of the plant. the unit supervisor and process engineer via pager.
They then inspect the disk and replace it with a new one This warns them of a possible environmental violation
if found ruptured or otherwise defective. before it occurs, providing sufficient time to proactively
intervene by adjusting the process.
A failed rupture disk does not result in an atmospheric
leak unless the PRV is itself leaky. However, the PRV will
corrode unless the rupture disk is promptly replaced.
Thus, even if it is not now leaking, the PRV will eventually
leak as a result of contact with corrosive process gases.
For this reason, regulatory agencies require a ruptured
disk to be replaced within 15 days and appropriate
records kept. As such, the Decision Support rule also
Rule Testing and
sends an e-mail notification to operation supervisors who
are responsible for submitting an ICV (inter-lock control
valve) bypass form. The ICV bypass form allows continued
VALIDATION
operation until the rupture disk is replaced. These records The Decision Support module in
are audited by relevant regulatory agencies and Valero’s System 1 software allows user-
ability to show these agencies that a process is in place to
written rules to be tested against
constantly monitor the integrity of the rupture disk helps
avoid violations and associated penalties.
historical data. This feature greatly
aids the Valero team in validating
CASE HISTORY #6: new rules before they are deployed.
Stack Analyzer Monitoring
The Paulsboro facility is required by law to monitor flue
It also allows them to fine tune their
gas from their FCC (Fluid Catalytic Cracker) unit for CO, rules, ensuring that a rule triggers
SO2, and NOX—all environmentally sensitive emissions not just when needed, but only when
regulated by law. The emission levels must be recorded
needed. This reduces false notifica-
and kept within allowable limits at all times. Monitoring
is accomplished using a real-time gas analyzer that tions and inspires confidence in
computes the concentrations of these emissions. the system.
However, to avoid violations and associated fines, it is
necessary to periodically compare the gas analyzer’s
results with laboratory results, verifying that the
analyzer is not underreporting. Paulsboro uses System 1
software to capture both the gas analyzer outputs and
laboratory results, and compare them to one another.
Discrepancies are flagged and appropriate personnel
are notified using the Decision Support engine.
Summary and Best Practices Best Practice #3: Involve the Whole Plant
As these case histories have shown, Valero Paulsboro Another key to Paulsboro’s success is that a much wider
uses the flexibility and functionality of System 1 circle of plant personnel have been exposed to the
software and its Decision Support capabilities to reduce capabilities of System 1 software and encouraged to
both operating and maintenance expenses. The quality identify opportunities where it can be applied. As a result,
of results achieved is in direct proportion to Valero’s engineers throughout the complex are now making
deliberate and conscientious application of several “best requests for custom rules that address various failure
practices” which have been summarized in conclusion: modes, ensuring that the system’s usefulness does not
remain confined to a single individual’s perspective or
Best Practice #1: Allocate Necessary Resources
scope. Valero’s assigned Reliability Engineer is typically
While the technology of System 1 software is important, able turn these requests around in a matter of hours,
it is merely an enabling factor. Without the resources configuring the requested rules along with testing and
to actually use the system, it cannot deliver its full validation using historical and/or current data.
potential. Understanding the importance of this human
Best Practice #4: Customize Notifications, not just Rules
factor—not just the technology factor—Valero allocates
a full-time Reliability Engineer to use and optimize the While customized rules are an important part of Valero’s
system. This individual is responsible for monitoring the success, rules deal only with detection. Once a condition
health of the system, identifying new and innovative uses has been detected, no matter how simple or complex
of the system, and developing/deploying custom rules. the underlying rule, notifying the right people at the
right time with the right recommended actions is key
Best Practice #2: Think Beyond Vibration and Rotating
to whether the system is perceived as merely an alarm
Machinery
generator or as an indispensible tool. Valero gives
Paulsboro has been particularly innovative in broaden- deliberate thought to the optimal number of severity
ing their system beyond the normal confines of vibration levels that an escalating condition should have, to the
data and rotating machinery to include fixed assets and individuals that should be notified with each severity
even process-related applications. As a result, the value level, to the way in which they should be notified (e-mail,
they extract is correspondingly broader. text messaging, pager, etc.), and to the actions that the
recipient should take once notified. This is perhaps the
single biggest part of successfully “managing below the
alert level” and achieving widespread visibility of system
usefulness across the plant.
* denotes a trademark of Bently Nevada, LLC, a wholly owned
subsidiary of General Electric Company.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 7 1
APPLICATIONS
RELATIONSHIP
of Influence Coefficients Between
Static-Couple and Multiplane
Methods on Two-Plane Balancing
This article was originally published in Vol. 131, Issue 1 of the Journal of Engineering for Gas Turbines and Power of the American Society of Mechanical
Engineers (ASME) International. It is reprinted here with the permission of ASME, which retains all copyrights.
nbalance accounts for the majority synchronous vibration levels. Topics on balancing
of high vibration problems in rotating have been of great interest to rotor dynamic
machines. High synchronous forces researchers and engineers [1,2]. Typically a turbine,
and vibration amplitudes due to mass compressor, or generator section is supported
unbalance produce excessive stresses by two bearings. This often requires two-plane
on the rotor and also affect bearings and casing, balancing for most cases where cross-effects
thus reducing the life span of the machine. The among different sections through couplings
source of unbalance may be imperfect manu- are trivial. There are a few papers discussing
facturing processes including assembly variation two-plane balancing with amplitude [3] or phase
and material nonhomogeneity. Though rotors [4] only. These approaches would often require
are typically balanced by manufacturers before more runs in the field and may increase both the
they are installed for service, unbalance may still time and the cost for users of rotating machinery.
occur afterward for various reasons. These include The influence coefficient method is typically used
deposits or erosion on (and shifting of) rotating for field trim balancing. There are basically two
parts, as well as thermal effects. Therefore, in approaches to apply this method. The first one is to
many cases, field balancing is required to reduce treat it as a multiplane balance problem involving
John J. Yu, Ph.D. – Senior Engineer, Machinery Diagnostics – Bently Nevada Asset Condition Monitoring – GE Energy – john.yu@ge.com
a 2X2 matrix of complex influence coefficients, as relationship of influence coefficients was given
Thearle [5] first presented in 1934. In this approach, between these two approaches. It was also some-
two direct influence coefficients along with two times believed that static-couple balance could
cross-effect influence coefficients are generated not reduce both static and couple vibration vectors
so that correction weights at two balance planes successfully because static (couple) weights
can be determined. The second one is to treat affect couple (static) response. In this paper, the
it as two single-plane balance problems using multiplane approach with a 2X2 influence coef-
static and couple components, respectively. The ficient matrix is first presented, followed by the
latter approach has been used extensively in the static-couple approach. In the latter approach,
field [6,7]. Having valid influence coefficient data cross-effects between the static (couple) weights
makes balancing much easier. Influence coefficient and the couple (static) component are introduced.
data can be employed to save trial runs for many Then, an analytical relationship of influence coef-
machines of the same design or for future balanc- ficients between these two approaches is derived
ing on the same machine. For two-plane balancing for two-plane balancing. Real examples are given
with influence coefficients, either static-couple or to verify the developed analytical conversion
multiplane approaches can be used. However, no formulas as well as to show their application.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 7 3
APPLICATIONS
Figure 1. Diagram of vibration and weight vectors when the Keyphasor pulse occurs.
However, cross-effects of static weights to the couple where ∆SS is the static vibration component with static
component or couple weights to the static component weight(s)—static vibration component without static
have often been neglected when performing balancing. weight(s) and ∆CS is the couple vibration component
A nonsymmetric rotor with respect to its two ends, or with static weight(s)—couple vibration component
strongly influenced by its adjacent section via coupling, without static weight(s).
might have significant cross-effects.
Similarly, having vibration data before and after couple
This article introduces the following static-couple weight placement WC (without static weights) yields
balance model to include these cross-effects:
(7)
(4) and
(8)
Figure 2. Diagram of static/couple vibration and weight vectors when the Keyphasor pulse occurs.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 7 5
APPLICATIONS
(18)
(10)
where
∆A1,S = A1 with static weight(s) − A1 without static (19)
weight(s)
(20)
∆A2,S = A2 with static weight(s) − A2 without static
Note that all the above equations apply to cases where
weight(s)
static weights are placed either at the middle balance
Similarly, the couple weight influence to probes near weight plane only or at two end balance weight planes
planes 1 and 2 can be given by with the same amount of weights in the same orienta-
tion. Couple weights are always defined throughout the
paper as placement at two end balance planes with the
same amount of weights in the opposite orientation (180
deg apart).
7 6 O R B I T Vol.29 No.1 2009
APPLICATIONS
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 7 7
APPLICATIONS
Subtraction of Equation (25b) from Equation (25a) Note that all the above equations in this section apply
followed by subtraction of Equation (26b) with to cases where static weights are placed at two end
application of Equation (3) yields balance weight planes. In case the static weight is
defined as placement at the middle balance plane,
(28)
Equations (27), (28), (39), and (40) are still valid.
Figure 4. Polar plots and vibration vectors at approximately 4800 rpm for initial run, and first and second trial
runs with weight placements.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 7 9
APPLICATIONS
Thus, the above four values within the matrix are the It is shown from this real example that influence vectors
influence coefficients for the multiplane method at for the static-couple method can be calculated from
this running speed. It is noted that the above two sets known influence coefficients h11, h12, h21, and h22 in a
of weight placement were also just for couple and 2X2 matrix for the multiplane method, without having to
static weights, respectively. Therefore, the influence place trial static or couple weights. Since Eqs. (29)–(32)
coefficients for the static-couple method can be directly are equivalent to Equations (23), (24), (27), and (28), and
computed. Using Equation (3), static and couple vibra- Equations (33)–(36) are equivalent to Equations (37)–(40),
tion vectors for the initial run without weight placement, Equations (29)–(36) also hold true in this example.
the first trial run with couple weights Therefore, influence coefficients h11, h12, h21, and h22
can also be obtained from influence vectors for the
static-couple method without having to place two sets
and the second trial run with static weights
of trial weights.
Note that
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 8 1
APPLICATIONS
Synchronous vibration amplitudes were higher on These large amounts of weight at two planes were
Y-probes than on X-probes at the two bearings on unable to be placed into empty holes or achieved by
the generator. Balance calculations were therefore adjustment of existing weights. An alternative needed
conducted on Y-probes only. In order to use the to be found. The study of influence data was then
same nomenclature and subscripts for the equations performed. Influence coefficients for static and couple
developed earlier, probes and weight plane at bearing weights were calculated based on known h11, h12, h21,
No. 5 are denoted as 1 while those at bearing No. 6 are and h22 values without placing static or couple trial
denoted as 2. As shown in Figure 6, Y-probe readings at weights. Note that the Y-probe at bearing No. 6 was
bearing Nos. 5 and 6 were not parallel to the Y-probe at bearing No. 5. In order to
evaluate static and couple effects better, the synchro-
nous vector at bearing No. 6, as though it was measured
by a proximity probe at 45 deg left, needed to be known,
and h11, h12, h21, and h22 needed to be applicable to this
change. Although the above-mentioned synchronous
vector at bearing No. 6 could be determined by using
The previous influence coefficients used for the
vectors from both X and Y probes, h11, h12, h21, and h22
multiplane method were given by
might not fit the new defined vector. Therefore, the origi-
nal vector was used as the new vector except its phase
was lagged an addition 15 deg. Thus, the two vibration
vectors referenced to 45 deg left became
where h11, h12, h21, and h22 were applied to Equation (1)
in which synchronous vibration vectors were defined as and the influence matrix with both vibration and weight
original ones from the two Y-probes 1 was referenced vectors referenced to 45 deg left became
to 45 deg left and 2 was referenced to 60 deg left, while
weights at both ends were all referenced to 45 deg left.
The balance plane radius where weights were placed
was about 0.254 m (10 in.) with the one at bearing No.
5 slightly larger than that at bearing No. 6 (about 1%
difference). Note that the radius difference between the Table 2 shows calculated influence vectors for static
two weight planes would not affect the validity of all and couple weights from known influence vectors
the equations developed in the paper. Weight planes h11, h12, h21, and h22 used for the multiplane method,
at bearing Nos. 5 and 6 had 44 and 36 holes for weight without having to place static or couple trial weights.
placement, respectively. Their weight sizes were also The direct couple influence vector HCC was the most
different between two planes. sensitive one (0.0111 mil pp/g 131 deg), indicating that
appropriate couple weights would effectively reduce the
Using Eq. (1), the required balance weights at two planes
current synchronous vibration level, especially to bear-
appeared to be
ing No. 5 (h1,C=0.0135 mil pp/g 131 deg). Static weights
appeared not to be sensitive to synchronous vibration
vectors at the running speed for this generator, as
shown in Table 2. The current static and couple vibration
vectors were as follows:
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 8 3
APPLICATIONS
Nomenclature Greek
Superscripts
References
[1] Ehrich, F. F., 1999, Handbook of Rotordynamics,
Krieger, Malabar, FL.
Vo l . 2 9 N o. 1 2 0 0 9 ORB I T 8 5
ANNOUNCEMENT
ORBIT
Reader
Survey
As ORBIT magazine nears its 30th year
of publication, we remain as committed
as ever to our mission: delivering deep
expertise that helps you understand,
manage, and optimize machinery.
Our Reader Survey is specifically designed to ensure that ORBIT is meeting your needs with
the right topics, the right level of detail, the right delivery mediums, and the right frequency.
The survey is short—just 5 to 10 minutes—and if your time is even more limited, you don’t
have to answer every question.
www.ge-energy.com/orbitsurvey
If you don’t have Internet access...
We’ll be happy to send you a hardcopy of the survey that you
can complete and return via fax or mail. Contact us at:
Phone: (775) 215-2291 | Fax: (775) 215-2864 | Email: orbit@ge.com
www.ge-energy.com/orbitsurvey
GIVE US 5 MINUTES, WE’LL GIVE YOU…
Pre-release access to “virtual ORBIT” A chance to win a gift basket containing
State-of-the-art “virtual magazine” technology will GE Olympic Games merchandize valued
soon be offered to our subscribers, with the option to at more than $500
receive both the conventional printed version and the One lucky survey respondent will be randomly selected
online version, or to go entirely digital and receive the to receive a gift basket containing dozens of items
magazine in electronic format only. from the GE Olympic Games merchandize collection,
commemorating the 2010 Winter Olympic Games in
Slated for release in mid-2010, this exciting new online
Vancouver, BC.
format will be available to all of our subscribers.
However, as a survey respondent, you don’t have to • 2 matching three-in-one jackets
wait. When you submit your survey, you’ll be provided • Collectors 8-sport pin set in maple box
with a special link, giving you immediate pre-release • Computer backpack
access to this enhanced online format. It’s our way • 2 thermal travel mugs
of saying “thanks” for taking the time to provide your • 2 sports bottles
valuable feedback. • 2 boxes of Taylormade® Noodle® golf balls
(1 dozen per box)
What is a “virtual magazine”?
• 2 compact travel umbrellas
Flash animation technology
• 2 leather passport wallets
is used to display magazine
• International travel plug adapter set
pages on your screen much
as the pages of a printed copy
would look in your hands. You
also interact with the magazine
much as you would a printed
copy: simply click on the page
corner to flip to the next page
(you can even enable the sound
of pages turning). The result
is a format that blends the
simplicity of the conventional
hard copy reading experience
with all the benefits of digital
documentation.
Vo l 2 9 N o. 1 2 0 0 9 ORBI T 8 7