Академический Документы
Профессиональный Документы
Культура Документы
Figure 1
Left New Motor (pumps 1 & 2), Right Old Motor (pumps 3 & 4)
The new motor is a different design and weighs about 800 pounds more than the old
motor. A mounting plate was fabricated to connect the new motor to the top of the tube.
The vibration problem on Pump #1 began after its motor (an 800 HP, wound rotor,
Reliance motor with a liquid rheostat speed control) was replaced with a new U.S. motor.
The old motor was in service for many years. Since the new motor was installed, the
vibration on the machine has been extremely rough. Some other characteristics of the
problem are listed below.
The highest vibration is over .9 inches/second peak (IPS) at twice rotational speed.
The problem direction is in-line with the discharge piping.
The worst vibration occurs when the pump speed is above 820 rpm.
The pump cannot be operated at the desired speed.
The VFD was programmed to exclude the 820-890 rpm speed range.
Baseline Data Baseline vibration data was measured with the pump operating at 870
rpm (14.5 Hz.). The largest vibration was .7 IPS peak in the discharge direction on the
motor.
Vertical
.7 IPS @ 2x
Discharge
Cross Discharge
.88 IPS @
Discharge
.65 IPS @
Cross Discharge
.4 IPS @ 2x
Vertical
Discharge
.42 IPS @
Cross Discharge
Figure 2
Baseline Data
Copyright 2001, Computational Systems Incorporated. All rights reserved.
Coast-down Testing:
The pump was operated at maximum speed (899 rpm or 15 Hertz). A tachometer was
used to measure motor speed and accelerometers were placed at the top of the motor in
the discharge and cross discharge directions. The machine speed was decreased slowly
using the VFD control. Vibration and phase data was recorded during coast-down using
the Analyze | Monitor | Monitor Peak/Phase function on the 2120 Analyzer.
Accelerometer
Positions
Tachometer reading
reflective tape on shaft
The Monitor Peak/Phase function was configured to measure coast-down vibration and
phase at 2x turning speed. The screen below shows the set-up.
The Bode plot below is a trace of 2x turning speed vibration and phase during coastdown. It indicates that at speeds below 830 rpm, the vibration on the motor was smooth
(less than .1 IPS) in the discharge direction. Above 830 rpm, the 2x vibration increased
and peaked at 0.9 IPS when the motor speed was 876 rpm. The 2x vibration frequency at
this point was 1752 cpm or 29.2 Hertz. A phase changed of about 180o was noted
through the amplification area.
2x
Figure 5
2x Coast-down Bode Plot data (discharge direction)
The coast-down data for the perpendicular to discharge direction indicated increasing
amplitudes as well. The Bode plot (below) shows the 2x turning speed vibration during
coast-down. The vibration amplification peaks at a point above maximum speed. The
response in this direction was less than half the amplitude of the discharge direction.
Figure 6
2x Coast-down Bode Plot data (perpendicular to discharge direction)
Discharge
Figure 7
Pump1
A natural frequency is the frequency at which a part likes to vibrate. Resonant amplification results
whenever forced vibrations, from mechanical defects, coincide with the natural frequencies in a system. At
resonance, a small change in the excitation from mechanical defects can produce a significant change in
vibration. The amount of amplification depends on the system damping characteristics.
Impact Testing:
Once the pump was shut down, impact testing was performed on the motor. One sensor
was used to measure response to an impact made with a three pound impact hammer.
The impact and response were measured at the same position and direction on the motor.
Both the discharge and cross discharge directions were tested.
Impact & response direction
39
Natural
Frequencies at 5 and
30 Hertz
The pump has two vanes. Vane-pass frequency (pump speed x # vanes) and/or
misalignment are potential sources exciting the resonance. The coast-down data indicates
that the 2x vibration is smooth below 830 rpm. The amount of mechanical vibration
related to vane pass or misalignment is, therefore, very small and probably cant be
reduced any further.
The natural frequency at 5 Hertz is not coincident with any forced vibrations over the
operating speed range of the pump. Vibration at this frequency is noticeable only during
start-up or coast-down of the machine.
The natural frequencies in the cross-discharge direction are very close to the frequencies
in the discharge direction. The amount of 2x vibration in the cross-discharge direction
was less than the discharge direction.
P2
P3
P4
30.0
22.5
33.7
28.6
.011
.024
.008
.017
31.2
22.7
36.8
28.5
.011
.017
.007
.021
Note: All machines have the same pump configuration. Pumps 1 and 2 have US
Motors. Pumps 3 and 4 have the old style Reliance Motors.
The natural frequencies of pumps 1 and 2 should be similar since both pumps have U.S.
Motors. They are not. The natural frequencies of pumps 3 and 4 should be similar
because both pumps have Reliance motors. They are not. The variability in frequencies
and stiffness values of the machines is probably due to differences in the boundary
conditions at the machine base (i.e. the connection to the floor is different).
All of the natural frequencies are lightly damped as indicated by tall peaks with narrow
skirts. This means that a small amount of 2x vibration energy from misalignment or vane
pass frequency can cause large amplification in the 2x operating speed range. The
graphic below shows the proximity between natural frequency and 2x operating speed
range for each pump.
Operating S peed
P um p 1
P um p 2
2X
2D
4D 4X 1D 1X
3D
3X
P um p 3
P um p 4
.3
.0
.3
.7
38
36
35
33
.7
31
.0
30
.3
28
.7
26
.0
25
.3
23
.0
.3
.7
.0
.3
.7
21
20
18
16
15
13
11
.7
D = Disc h Dir
Direction
D = Discharge
X = P rpendic
X = Cross-Discharge
Hertz
Figure 12
Discharge and Cross-discharge Direction Natural Frequencies
near 2x Operating Speed Range for Pumps 1-4
10
The data indicates that Pump 1 (discharge direction) and Pump 4 (both directions) have
natural frequencies in the 2x operating speed range. Pump 1 has a vibration problem
related to resonance at 2x operating speed. Based on the impact data measured on Pump
1, amplification factors of 10-15 can be expected at the 30 Hertz natural frequency.
Pump 4s natural frequencies are below the upper end of the 2x operating speed range.
Pump 4 usually operates at full speed and 2x operating speed is far enough away from the
natural frequency that resonance will not occur.
Pumps 2 and 3 do not resonate because the natural frequencies fall outside of the 2x
operating speed range.
11
Motor
Motor
baseplate
Steel
Tube
Pump
Legs
Steel block
Concrete Pad
Floor
Figure 13
ODS Structure
An Operational Deflection Shape (ODS) is a non-intrusive test used to analyze the motions of rotating
equipment and structures under normal operating conditions. An ODS is an extension of phase analysis. In
an ODS, points on a computer generated model of the machine are animated using phase and magnitude
data measured during normal operation. If Frequency Response functions were measured using the
Advanced 2-channel DLP, the structure can be animated at any frequency of interest.
12
VibPro software is a CSI product used to download and analyze and print data collected
with the Advanced 2-Channel or Advanced Transient DLPs. VibPro is also used to
export data to spreadsheet format so it can be imported into the MEscope ODS software
and used to animate the structure drawing.
The graphic below shows the animation shape at 2x turning speed (29 Hertz). It indicates
that the pump tube is bending. The shape approximates the second bending mode of a
cantilevered structure. Click on the link to play the AVI file. ODS 29 Hz -2x.Avi
Figure 14
ODS Animation of Natural Frequency at 29 Hertz
In addition to bending of the tube, the ODS showed looseness on one leg of the pump
base. Another ODS study was made on the leg. Many additional points were measured
to get a better view of the problem. A new structure drawing of just the one leg was
made. It is shown below.
13
Tripod Leg
(discharge
side)
Tripod
Foot
Cast Steel
Spacer Block
Concrete
pad
Floor
Figure 15a
Picture of one Pump Foot
Figure 15b
ODS Structure of the Pump Foot
14
Look for global motions -- where the entire machine is moving together with no
relative motion between components. This could be a result of a machine mounted
on isolators or floor and building vibrations
Look for relative motions between bearing housings or shafts (if shaft data was taken)
-- an indication of misalignment
Look for phase lag and relative motion between bolted or welded joints -- indicating
looseness. Looseness problems will show similar motion at different frequencies.
Look for twisting of the machine base -- indicating torsional bending modes or
structural weakness
Look for bending of structural components an indication of resonance (note: ODS
does not prove resonance)
Look for localized motion on machine feet or bases an indication of soft-foot
Being a good vibration analyst means being a good detective. Study the ODS animations and
look for clues that will help solve the problem.
15
Timber
Once the timber was wedged tightly in place, the pump was operated and another coastdown test was performed. The data indicated that the natural frequency of the structure
(near 2x operating speed) didnt change much, however the response at resonance was
cut in half. The Bode plot below shows that the vibration was reduced from .89 IPS to
.45 IPS. The vibration in the perpendicular to discharge direction also decreased. The
timber was left in place after the test and secured with a chain-fall. The speed restriction
was removed from the drive control and the Operating Department began to use the pump
in this speed range. Note: The timber should not be left in place as a permanent solution
to the resonance problem.
Copyright 2001, Computational Systems Incorporated. All rights reserved.
16
Figure 18
2x Coast-down Bode Plot of #1 Pump with a 8 x 8 Timber
Wedged Between the Building Wall and the Pump Tube
17
18
Modal Survey:
Modal survey3 data were collected on the #1 pump using the
2120-2 analyzer and the Advanced 2-Channel DLP. The
Advanced 2-channel DLP is required for modal data
collection.
The modal survey on the #1 Pump consisted of Frequency
Response Function (FRF) measurements at 194
positions/directions. At each measured point:
a) Natural frequencies were excited by striking a 12
pound impact hammer against the top of the motor
(discharge (X) direction)
b) An accelerometer was used to measure the 0-100
Hertz response at each measurement point and
direction. Four averages were acquired at each
measurement point.
Figure 19
Modal Structure
The figure below is a Modal Peaks plot. It is a summation of all 194 FRFs and makes it
easy to see the response of all points and directions in one view. Pump #1 had natural
frequencies at 4.8, 29.81 and 39.3 Hertz. Additional natural frequencies were found
above 50 Hertz. Each natural frequency has a unique shape called a mode shape. After
curve fitting the FRF data, the mode shapes were animated and analyzed.
29.81
4.817
56.3
39.3
68.8
39.9
51.4
24.6
Figure 20
Modal Peaks Plot is a Summation of all FRF Measurements
Modal Analysis is an experimental method of determining the natural frequencies, damping values and
mode shapes of a structure. A modal analysis is done with the machine off-line.
19
Figure 21
Frames View of 29.8 Hertz Natural Frequency
20
The first bending mode was 4.82 Hertz. No mechanical vibrations are exciting this mode
into resonance except during start-up and coast-down.
The frames view of the 4.82 Hertz mode shape is shown below. Click on the link to play
the AVI file.
Modal 1st bending mode - 4.82 Hz Quad View.Avi
Figure 22
Frames View of 4.82 Hertz Natural Frequency
21
Table 2 lists the Modal Survey results and characterizes each mode shape observed.
There were seven modes between 0-60 Hertz.
Table 2
Pump #1 Modal Frequencies
Frequency
(Hz)
4.82
25
29.81
39.3/39.9
51.4
56.3
Mode Description
First bending mode
First torsional mode not sufficiently excited in the modal survey
Second bending mode, Shaft/Support Pipe in Phase
Second bending mode, Shaft/Support Pipe out of Phase
Vertical mode
Third bending mode
22
Natural frequencies can not be eliminated. The effect of resonance may be diminished or
natural frequencies may be shifted up or down in the frequency range. Some methods of
correcting resonance are discussed below.
Reduce the exciting force Nothing resonates without an excitation force. Forcing
frequencies (mechanical vibrations) are most often the excitation for resonance.
Reducing the exciting force, by any amount, diminishes the effect of the resonance. Its
usually less expensive to reduce or eliminate the exciting force than it is to modify the
structure. Some examples of reducing the excitation include
Change the speed -- Move the exciting force away from the natural frequency. The rule
of thumb is to change the speed 10%-15% on either side of the natural frequency.
Depending on the damping value for a given bending mode, more or less speed change
will be required.
amplitude
23
Change the Mass -- Increasing the mass of a structure lowers its natural frequencies.
Consider this simplified natural frequency formula:
fn = 1/2 k / m
Where:
Fn = natural Frequency
K = Stiffness
M = Mass
= Constant
fn = 1/2 k / m
If K is increased and M remains constant then Fn, increases
If K is decreased and M remains constant then Fn, decreases
Examples of changing a structures stiffness include adding bracing or gussets to a base
frame and changing the thickness of components.
A modal survey alone does not indicate how to correct resonance. The SDM version of
MEscope software has some Finite Element Tools that can be used to estimate the
required structural modifications. The alternative is to have a Structural Engineer
complete a Finite Element Analysis.
24
Mass
2
IXX
IYY
2
IZZ
2
(lbf-sec /in)
(lbf-sec -in)
(lbf-sec -in)
(lbf-sec2-in)
23.8
21950
21950
13750
The principal mass moments of inertia are calculated at the motor center of gravity.
The pump assembly includes numerous bolted interfaces between the structural
components. Considerable effort was made to accurately model these connections. While
the bolts themselves are considered to be rigid, only assumed preloaded areas, which are
local to the bolts, are used to connect the parts. These preloaded areas are modeled using
special constraint elements that do not add stiffness to the bolted components themselves.
This technique was used to model the connections between the concrete and metal
blocks, the top of the metal blocks and the base plates of each leg of the pump base, and
the top flange of the pump base and the bottom flange of the support pipe.
4
A Finite Element Analysis (FEA) is an analytical technique that utilizes a mathematical model of a
structure to predict its natural frequencies and mode shapes. A resonance problem is corrected by using the
FEA model to evaluate the effectiveness of mass, stiffness and damping modifications. Sigmadyne, Inc.
(http://www.sigmadyne.com) completed the FEA on Pump 1.
25
The finite element model was created for use in MSC/NASTRAN, a commercial finite
element code. The model includes a total of 6575 beam, shell, and solid elements. Solid
base, support pipe and adapter plates were modeled using shell elements. The thickness
and internal design of the volute is not known. It is modeled as a uniform shell structure
with a thickness of 5/8 inch. The motor shaft was modeled using beam elements
connected to the model at the base of the motor, the top of the volute, and the top of the
conical structure mounted on the volute. The conical structure inside the legs of the pump
bases was modeled using a rigid element.
The entire model is constrained at the bottom surface of each of the three concrete blocks.
Physically this represents the interface between the concrete blocks and the floor beneath
the pump. The complete model is shown below.
Figure 24
FEA Structure
26
The FEA model is only as accurate at the information used in its construction. The FEA
prediction of the natural frequencies will not be accurate if any of the information is
incorrect. In addition, the estimates for structural modification may not be valid.
Examples of things that can lead to an inaccurate FEA include:
The assumptions used in the FEA model of the #1 Pump are listed below.
1. The weight of the motor, specified by the customer, was 9200 lbf. Since the mass
moments of inertia are unavailable, they are calculated from this weight based on
a diameter of 67.5 inches and a height of 92.5 inches. The mass moments of
inertia are calculated using the equations for a solid cylinder with uniformly
distributed mass.
2. The motor is assumed to be rigid over the frequency range of interest.
3. Concrete is assumed to be a homogeneous and isotropic material and to behave in
a linear elastic manner. The following mechanical properties are used in the
analysis:
Modulus of elasticity:
2.0 MSI.
Poissons Ratio:
0.2
Weight density:
145 lbf/ft3
4. The rebar used in concrete does not significantly alter its mechanical properties.
5. Steel is assumed to be a homogeneous and isotropic material and to behave in a
linear elastic manner. The following mechanical properties are used in the
analysis:
Modulus of elasticity:
30 MSI.
Poissons Ratio:
0.3
Weight density:
481 lbf/ft3
6. The pump is fixed at the bottom surface of the concrete block. Any additional
compliance due to the surrounding structure is ignored.
7. All dimensions used in the creation of the model were measured on Pump 1 as
part of the modal survey or were supplied by Plant personnel.
27
Natural frequency or modal analysis is used to analytically predict the natural frequencies
and mode shapes of vibration of a structure. As previously stated a modal survey is an
experimental procedure used to measure both the natural frequencies and mode shapes of
a structure. The accuracy and validity of the finite element can be evaluated and
enhanced by correlating the predicted natural frequencies with the measured values. This
correlation is based on comparison of the mode shapes of vibration rather than a simple
comparison of the frequency values. This ensures that similar modes (i.e. the first
bending mode of a cantilevered structure) are being appropriately compared.
Once the finite element model was correlated to the modal survey, a design optimization
analysis was performed using the model. The objective of this analysis was to determine
the changes in the natural frequencies due to alterations of the pump geometry.
28
FEA Results:
Both predicted and measured values for the first eight natural frequencies of Pump 1 are
summarized in Table 4. Natural frequencies of a symmetric structure occur in orthogonal
pairs. The physical significance is that the pump can actually vibrate (bend) in any
direction based on the direction of the applied excitation. It is difficult to experimentally
measure multiple modes with nearly identical frequencies. During the modal survey of
Pump 1, the frequency resolution was insufficient to individually measure the two first
bending modes and the first set of second bending modes. The frequency separation of
the next mode pair (motor shaft and support pipe oscillating out of phase) is sufficient
and both were identified in the test.
Table 4
Correlation of Measured and Predicted Natural Frequencies of Pump 1
Measured Value
Mode Shape Description
(Hz)
Analytical
Prediction
(Hz)
st
1 Bending
4.8
st
1 Bending
4.4
4.4
Torsion
24-25
23.5
29.8
29.2
29.5
39.2
41.8
39.9
42.2
Vertical Extension
51.4
60.3
The correlation between the analytical prediction and the measured frequency values was
very good. A picture of the mode shape at 29.2 Hertz is shown below. Click on the link
below to play the FEA animations.
FEA 2nd bending mode 29.2 Hz.mpg
FEA 1st Bending mode 4.3 Hz.mpg
29
Figure 25
Second bending mode of the Pump1 structure
showing the drive shaft vibrating in-phase with the support pipe at 29.2 Hz.
Since the natural frequency at 29 Hertz is at the upper end of the 2x operating speed
range, the structural engineer recommended increasing the machine stiffness. Increasing
stiffness results in a higher natural frequency. To accomplish the change, several
potential solutions were evaluated using the FEA software including:
The only option that was feasible was to reduce the length of the pump tube. The
customer agreed that changing the tube length was an acceptable solution. Tube length
reductions of 48, 62 and 72 inches were evaluated. All three models produced acceptable
results. Table 5 shows the change at each natural frequency based on pipe length
reductions. The stiffness of the support pipe (Tp) was considered by treating the support
pipe thickness as a design variable. It was found that this parameter has little effect on
the modes near 29 Hz.
30
Table 5
Effect of Support Pipe Length on the Predicted Natural Frequencies of Pump 1
Length Reductions
Baseline
Tp=0.75
48
62
72
(Hz)
(Hz)
(Hz)
(Hz)
(Hz)
1st Bending
4.4
4.8
5.4
5.8
6.2
1st Bending
4.4
4.9
5.4
5.9
6.3
Torsion
23.5
25.5
24.6
25.1
25.5
29.2
30.0
37.1
39.9
41.8
29.5
30.2
37.3
40.2
42.2
41.8
39.2
49.6
54.1
58.5
42.2
39.6
49.8
54.2
58.6
Vertical Extension
60.3
64.9
64.4
66.0
67.3
Decreasing the length of the support pipe and correspondingly, the overall height alters
the natural frequencies of the pump. A length reduction of 48 inches increases the
frequency of the first pair of second bending modes (shaft and support pipe bending in
phase) to approximately 37 hertz. An important assumption in this prediction is that all
other stiffnesses remain unchanged. This is especially important given the number of
bolted connections in the pump assembly.
Operating Speed
31
.
33
.
35
.
io
n
re
du
ct
io
n
re
du
ct
le
ng
th
72
en
gt
h
62
l
0
29
.
43
.
0
27
.
0
25
.
41
.
0
23
.
0
21
.
39
.
0
19
.
48
l
en
gt
h
0
17
.
.0
0
15
.
37
0
13
.
11
.0
Cu
rre
nt
c
on
di
tio
n
re
du
ct
io
n
Figure 26
Pump 1 Second bending mode Frequency Shift
With Pipe Length Reduction
The vibration measurements, made during operation of the pump, did not indicate any 3x
vibration that might excite the (new) second bending mode natural frequency.
31
Summary:
Resonance testing and Operational Deflection Shape studies are useful tools for
analyzing vibration problems. When resonance is identified as the problem, a Modal
Survey is necessary to identify all of the natural frequencies and evaluate each mode
shape. Without knowing the mode shape, it is impossible to know how to correct the
resonance.
Resonance is best corrected by using FEA tools to evaluate the effectiveness of potential
structural modifications.
The recommendations from this job were well received by the customer. To date, the
modification has not been implemented on Pump.
32