You are on page 1of 8

Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

Vibration Analysis – Knock Out Drum Gullfaks C


Author: S. Sandström
SSM Engineering, Strandvägen 40, SE-444 31 Stenungsund, Sweden
Stig@ssmengineering.se

Abstract: An eigenvalue as well as a harmonic in the flare stack, a so-called Knock Out drum
response analysis is performed on the (KOD) is installed directly upstream to the flare
Gullfaks C, Knock Out Drum (KOD) to separate the liquid from the gas stream.
including some of the connecting pipe In oil production systems there is sometimes
systems. The vibration behaviour of two new a need to depressure the oil well system (200-
choke valves and their specific connection to 300 bars). The depressurizing system is
the KOD are of special interest. The main connected to the flare system via so called
concern is the low frequency induced ‘choke valves’.
vibration as a consequence from the platform The existing KOD shall be modified by
movements. A frequency band in between 0– means of two new choke valves. These valves as
20 Hz has been investigated. The eigenvalue well as their long actuators are rather heavy in
analysis shows that in the low frequency band weight. The drum, the valves and the connected
(0-10 Hz), the vibration amplitudes are pipe systems are shown in figure 1.
dominated from the large-diameter pipes. In
the band 10-17 Hz, the dominant vibration
amplitudes are expected at the 6" and 8"
pipes. Especially, at the frequency of around
11-12 Hz the largest movements will occur in
the choke valve actuators. The higher modes -
over 17 Hz - are vibration modes on the
vessel’s shell with rather small amplitudes
compared to the overall vibration level. The
vibration behaviour of the internals - inlet gas
box and inlet device – shows rather small
movements compared to the overall vibration
amplitudes, while the Tecklok nozzles and the
reinforcement plate will vibrate with a Figure 1. The drum and some connected piping.
smaller amplitude. In order to crosscheck the
eigenvalue results, an additional harmonic The weight of the choke valves and their
response study has been performed. The connected actuators as well as the wide-stretched
driving vibration forces are the platform design of the actuators build up a system with a
prescribed accelerations at the structure high risk of vibrations due to blow-downs,
support taken from measured data. Also here, forced vibrations from the platform structure,
the analysis shows that the structure and other external excitation sources. Therefore
movements are small. Thus, we can conclude it has been decided to model and investigate the
that no large vibration will occur in the choke vibration behaviour of the drum, its internals (in
valves in the frequency band studied. the KOD documentation designated as inlet gas
box and inlet device), the connected pipe systems
Keywords: Vibration, Resonance, Eigenvalue, (30"-, 8"-, and the 6" piping lines) and the choke
Eigenvector, Harmonic analysis. valves with their actuators.
The drum - with the nominal thickness of 10
mm and 13 mm in the heads - has been
1. Introduction reinforced with a 30-mm thick plate (2600x2200
mm) where the inlet (Techlok Nozzle) tubes are
In many chemical process plants there is a located. These inlet tubes are quite thick-walled
need to have a flare system which acts as a (OD x Wall thickness - 480x65, and 508x58
‘safety valve’ in case of reactor or process respectively) and by it, they are rather stiff and
upsets. Since no phase is liquid phase is allowed heavy.
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

Of special concern is the low frequency ΦTi MΦ j = δ ij (4)


vibration that may arise in the internals, the
choke valves, their actuators and the connected where T is the transpose and δij is the Kronecker
6" and 8" pipe systems due to the low frequency delta. The norm used by COMSOL Multiphysics
movements occurring in the platform structure is
itself. This low frequency range is usually in the 1 N
Uk = 1
2
interval 0-2 Hz. If the choke valves and their ∑
N k =1
(5)
actuators start vibrating heavily in this low
frequency band we might end up with a fatigue where Uk is the elements in the vector Φi, [4].
problem in the welds between the 30 mm thick Harmonic response analysis as briefly
penetration plate and the Techlok Nozzles. Other described in [1]-[3], is the solution to
possible problem that might arise is fatigue in the MU&& + CU& + KU = P (6)
internals or elsewhere, malfunctions of the where C is the damping matrix and P is the
valves etc. frequency dependant load vector which can be
To check the over all vibration susceptibility prescribed forces or displacements etc. Damping
we need first of all perform a resonance ratio coefficients could be found in [5].
frequency analysis. This kind of analysis is often Harmonic solutions to equation (6) for DOF no k
referred to an eigenvalue problem, which means now is of the form
that we are interested in the eigenvalues as well U k (ω ) = Ak (ω ) cos(ωt + α k (ω )) (7)
as the eigenvectors (eigenmodes).
where Ak is the amplitude and αk is a phase angle.
2. Theory
3. Calculation Model – Model Description
The theory of eigenproblems is since long
well established, see literature for more details, 3.1 General
[1]-[3]. In general, an undamped finite element
A number of assumptions have been made
system the eigenproblem could be described as
for the modelling of different parts of the
follows
calculation model. Reasons for making the
MU&& + KU = 0 (1) assumptions are partly lack of information, and
where M is the mass matrix, U the displacement partly to make as simple calculation model as
vector, and K is the stiffness matrix. The dots possible without losing accuracy.
indicate time derivatives. The system is illustrated in figure 2 showing
Eigensolutions could be found by assuming the complete model system i.e. the drum and the
harmonic response, namely connecting pipe systems. The length of the pipe
U = Φ e iωt (2) systems should at least pass by two fixes1.
where Φ is the eigenvector, i the imaginary unit,
and ω is the angular frequency. Equations (1)
and (2) yields
− MΦ λ + KΦ = 0 (3)
where λ=ω2. Solution of equation (3) now gives
the eigenpairs (λ1,Φ1), (λ2,Φ2), …, (λN,ΦN) where
N is the number degrees of freedom (DOF –
which could be explained as ‘the total number of
unknown elements in the vector U’). In practice
we only need to consider a few of the lowest
eigenpairs.
The eigenmodes Φi are orthogonal to each
other i.e. ΦiΦj=0 where i≠j. Also, the
Figure 2. The drum and the connected piping.
eigenmodes Φi are undetermined within a
constant which may be determined by
normalization. A frequently used norm is
1
A fix is a restraint in the three linear directions
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

Figure 3 shows the drum and its interiors, the plates that are bolted together see figure 4. The
inlet gas box and inlet device, the supporting bolts and the connecting flanges do not give any
beams, and the nozzles. The pipe systems as well contribution to the stiffness of the inlet gas box
as the beams are represented by means of the but the bolts and the flanges increase the mass of
coloured lines. Note the elements laid out in a the inlet gas box. In fact, the flange connections
cross at the end of the nozzle, which connects the slightly weaken the structure, which we
nozzle to the pipe. however, neglect here.
To simulate the increased mass due to bolts
and flanges we simply increase the plate density
in these parts. The steel plate density (true
density) in the inlet gas box was increased by
40%1 and the plate thickness was kept to 8 mm.
It was also decided to study the influence of a
variation in density by testing a case with a 50%
increase in steel density.

3.2 Inlet Device

The inlet device is illustrated in the right part


of figure 4. The top and bottom plates are
Figure 3. The drum and its interiors. modelled as ordinary 8-mm thick plates. Also, in
figure 4 we notice the blades, which intend to
3.2 Inlet Gas Box give the desired flow distribution. Modelling all
these blades is a rather time consuming and
The inlet gas box is illustrated in the left part difficult task – however, we are not interested in
of figure 3. Figure 4 illustrates the whole item detailed vibration behaviour of the blades. For
where we notice the supporting beams. Due to this reason the blades are simplified as 8-mm
lack of information many assumptions have been thick plates on each side of the device with an
taken about dimensions, weight etc both for the increase in plate density by 80%. Also here, the
inlet gas box as well as the inlet device. influence of the variation in mass is to be studied
by an increase in mass density up to 100%.
On the right hand side in figure 4, there is a
support arrangement at the top of the device.
This support is dedicated to act as a restraint in
the transversal direction of the drum (y-direction
in figure 3) and to behave as a slide support in
the longitudinal direction (x-direction in figure
3). The support is therefore modelled with a 3-
mm thick plate, which is rather stiff in the
transversal direction and weak in the longitudinal
direction.
Figure 5 illustrates the calculation domain of
the internals. The shell has here partly been
removed. The different colours designate plates
with different properties.
Figure 4. The inlet gas box and the inlet device.

In figure 3, we ca see the 30" nozzle to the


left, passing through the head of the drum.
Thereafter, the nozzle is connected to the inlet
gas box. The plate thickness in the inlet gas box
is 8 mm. However, the inlet gas box is actually 1
This means that the fictitious plate density is ρfict = 1.4 ρsteel
not a pipe as it is more or less curved and flanged
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

3.5 Techlok Nozzles

These nozzles are, as mentioned before, quite


thick-walled (OD x Wall thickness - 480x65, and
508x58 mm respectively) which mean that they
are rather stiff and heavy. They are connected to
the choke valves and are illustrated in figure 5.

3.6 Knock Out Drum (KOD)

The drum itself is shown in figure 2. As


mentioned before, the shell thickness is 10 mm
and the head thickness is 13 mm. The two
Figure 5. The drum internals. Note the U-beams and stiffener rings consist of I-beams, height x flange
the longitudinals. width 172x100 mm, but are here modelled as a
simple plate, height x width 200x26 mm. These
3.3 Support Beams and Longitudinals dimensions give equal area moment of inertia.
As the rings are intended mainly to keep the
To support the inlet gas box and inlet device, roundness of the drum - and thus work mostly in
support beams (U-200x8x8) are located in the bending - this is a fair assumption.
transversal direction, see figure 5. The support The 30-mm thick reinforcement plate is
beams are connected to the shell simply by shown in figure 6 (green coloured). The plate is
passing through the shell. The small pieces of directly connected to the head.
beams sticking out on the outside of the shell
(they are hardly visible in the plots) are made
inactive in the calculation domain - simply
meaning that they are totally neglected (they do
not contribute to the stiffness nor to the mass
distribution). Modelling the connection between
the support beams and the shell according to the
drawings will be too complicated as we have no
interest in the detailed behaviour of these
connections.
The longitudinals, are running in crosses in
the longitudinal direction of the drum, see figure
5. Due to lack of information it is here assumed
that these beams are of size L-50x50x8. As the Figure 6. The drum and the reinforcement plate.
longitudinals are aimed as sideways stiffeners to
the U-beams only, the assumption of the Neither the welds between the ordinary and
dimension L-50x50x8 is not that critical. the thick shell plate are included in the
calculation model, nor the welds between the
3.4 Upper Vertical Nozzles Techlok Nozzles and the thick plate. This is in
line with a practical approach to avoid too small
In figure 5 we notice two nozzles on the and too any details in the finite element model.
upper part of the shell. These nozzles do not The drum is supported in saddle supports
have a stiff connection to the inlet gas box. The integrated to the stiffening rings. The support
nozzle pipes are simply thrusted into the inlet gas closest to the choke valves is - according to the
box. This is modelled by just removing a small drawings - restrained in the axial direction of the
piece of pipe in the connection nozzles / inlet gas drum. The support located at the second ring is a
box. In figure 5 one may notice two lines in the sliding support in the axial direction. Both
tangential direction on the nozzle pipe which supports are restrained in the transversal
‘shows’ the removed piece of pipe. direction.
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

3.7 6”, and 8” Pipes, Choke Valves, and Other The 6" and 8" pipe system are modelled as
Valves beams where a so-called flexibility factor is
applied to the bends. As a pipe bend is weaker in
In figure 6 the 6" and 8" pipe systems are bending than a straight pipe we need to apply a
shown. In the figure we clearly can see the flexibility factor to each bend.
connection between the shell and the pipe (beam) The flexibility factors are taken from the
systems with aid of rigid1 elements. The choke ANSI B31.3 Code. To get the correct stiffness in
valves are modelled not as rigid elements but as the bends, each bend area moment of inertia is
very stiff elements. The weights are 2851 and divided by its flexibility factor. In the 6" & 8"
2932 kg for the 6"x20" and 8"x18" valves pipe systems the flexibility factor is quite low,
respectively. around 2.
Due to lack of information two assumptions According to the Code the flexibility factor
made are; 1) the actuator is modelled as a 6" Sch in a pipe ‘Tee’ is 1, which means that the Tee is
160 pipe and a lumped mass of 222 kg is located as stiff as the straight pipe.
at the actuator tip; 2) the actuator is modelled as The gate valves are modelled as rigid
a 6" Sch 160 pipe, one lumped mass of 63 kg is elements. The drawings indicate the location of
located at the actuator tip and another lumped their centre of gravity. It is assumed that the
mass of 159 kg is simulating the flange between weight of the actuator is around 25% of the total
the valve body and the actuator. valve weight. If the valve body as well as the
In both cases, the total mass of the actuator is actuator are simply considered a of cylinder
kept to 320 kg. Assumption 1 gives a more geometry, then the rotational mass moment of
conservative result as the actuator itself is inertia acting in the centre of gravity point can
assumed to be heavier. The installed choke easily be calculated. This, together with the
valves is shown in figure 7 where we observe lumped linear mass are now applied in the centre
that the actuators are rather massive. of gravity. By doing so, it is in fact not necessary
to model the actual lengths of the actuators.
The flanges are manually input as lumped
linear masses as well as the rotational mass
moment of inertia in different directions.
The supports in the 6" & 8" pipe systems are
applied as given in the drawings. The supports
are assumed as rigid. However, the gaps
indicated in the drawings are not applied here.
Applying the gaps to the calculation model will
give a non-linear model for which it is not
possible to get an eigensolution.
All the pipe systems end at a support where
the pipe is also restrained in the axial direction.

3.8 30”, 24” and 16” Pipe Systems


Figure 7. The choke valves and their actuators.
These pipes are considered as large-diameter-
The lumped masses as well the rotational pipes. However, as our interest is mainly the
mass moment of inertia were manually inserted beam bending vibration behaviour in the pipe
in COMSOL at their locations. To calculate the systems, these pipes are modelled as beams. The
rotational mass moment of inertia the different flexibility factors applied to the bends are here
parts are assumed as a circular pipe with quite large - about 22-25 - giving quite weak
dimensions according to the drawings. bends. This has a large influence on the lower
eigenfrequencies.
The flanges are modelled according to the
1 earlier description on the 6" & 8 “ pipe systems.
In this paper, a rigid element is a very stiff element without
mass. The 30" line is supported on two spring
hangers. The springs are active only in the
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

vertical direction and they are modelled as a There is a small movement in the choke
force, -kw, where k is the spring constant, and w actuators in the 8.98 Hz-mode. Thereafter at
is the vertical displacement (z-direction). 11.1, 11.3, and 11.6 Hz respectively the choke
actuators vibrates heavily. At 15.9, 16.4, and
3.9 Finite Element Modelling 18.5 Hz respectively the 6" & 8" lines vibrates.
At higher modes we see that it is mostly the
The drum, the inlet gas box, the inlet device, drum shell that tends to higher vibration. No
and all the nozzles are modelled with shell large movements at the Techlok Nozzles are
elements. As the Techlok nozzles are rather observed.
thick-walled shell element might give a less
accurate result than solid elements. However, as
we are not interested in a detailed stress
distribution in these nozzles this assumption
seems fair enough.
The supporting beams U-200x8x8 and the
longitudinals L-50x50x8 are modelled with
beam elements. At the level where the inlet gas
box and the inlet device are resting upon the
supporting beams the nodes belonging to the
shell elements and beam elements are connected.
The chokes, the actuators as well as the pipe
systems are modelled as beam elements. Rigid
Figure 8. The lowest eigenmode 3.21 Hz.
elements have been utilized in the connection
between nozzles (shell elements) and the pipes
(beam element). Also, rigid elements are used to
model the 6" and 8" valves as the valves
themselves are very stiff. The masses of the
valves are input as linear mass as well as their
rotational mass moment of inertia in the centre of
gravity point.

4. Calculation Result

In this analysis the lowest 20


eigenfrequencies have been calculated. Some
eigenvalues and the corresponding eigenmodes Figure 9. Eigenmode no 12 at 11.6 Hz. Note the
are illustrated in figures 8-9. amplitudes at the choke actuators.
In general we observe that in the lower
frequency band - up to around 10 Hz - it is 5. Variation of Parameters
mainly the large-diameter-pipes that vibrate
(except for a few cases). In the interval from In the model description above, a number of
around 10 to 17 Hz there is a mixture between parameters mentioned were to be varied. These
involved components such as the choke valves parameters are 1) the accuracy of the finite
and its connecting 6" & 8" pipes, the shell, and element mesh; 2) weight of the flanges and bolts
the large-diameter-pipes that dominates the in the inlet gas box/inlet device (increase of plate
vibration. In the higher frequency range i.e. over density); 3) weight distribution in the choke
17 Hz shell vibrations dominate. actuators.
Studying the eigenmodes one can observe
that in the lowest eigenmode, figure 8, the 30" 5.1 Finite Element Mesh
line moves up and down swinging upon the two
spring supports. In the range up to 8 Hz we it is Several computer runs have been performed with
observed that the large-diameter-lines vibrate different mesh sizes. This is an important
mostly in the bends. parameter as it has a large influence on the
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

solution time. This parameter should thus be displacement amplitude (about 1 mm) is located
chosen before studying other parameters. at the frequency 0.34 Hz which is the lowest
In our case, the chosen accuracy of the finite eigenfrequency of the platform structure in the x-
element mesh results in a total number of DOF direction.
to 121184.
An increase in DOF to 172122 changes the
eigenvalues by less than 1%, which means that
the lower value of 121184 DOF is sufficient.
However, that increase in DOF increases the
solution time by a factor of around 3.
A lower number in DOF gives to low
accuracy.

5.2 Variation of the Mass of the Inlet Gas Box


and the Inlet Device

To investigate the sensitivity of the weight of Figure 10. Acceleration spectra in the x-direction.
the inlet gas box/inlet device the plate density
was increased to 1.5ρsteel in the inlet gas box and
to 2.0ρsteel in the inlet device respectively.
No significant changes were found except for
slightly larger movements of the inlet device in
the higher modes (over 20 Hz).

5.3 Variation of the Weight Distribution in the


Choke Actuators

The two assumptions regarding the weight


distribution in the choke actuators were
discussed in the model description above.
We now find that some of the Figure 11. Displacement spectra in the x-direction
eigenfrequencies in the 9 - 14 Hz band have at the 6” choke valve actuator tip.
increased as much as 10% comparing the first
case to the second. On the other hand, the shape 7. Experimental Results
and the relative magnitude of the eigenmodes are
more or less unchanged. In parallel to this study vibration
Thus, we can conclude that if we keep our measurements and tests were performed [6]. The
first model (assumption 1 - mass 222 kg) we are vibration measurements were performed in
more conservative. different weather conditions such as hard winds
and rough sea. Also, tests such as blow-downs
6. Harmonic Analysis and pressure release (of the oil wells) were made.
The measured movements of the support of the
Measurements of the platform movements tank shows low frequencies, the lowest
also was performed, [6]. Figure 10 shows the resonance frequency is 0.34 Hz.
acceleration spectra of the tank support in the x- The structure was exited to measure the
direction. resonance frequencies. In general, the measured
The measured acceleration spectra in the x-, resonance frequencies were 10-20% higher than
y-, and z-directions respectively were applied as the calculated frequencies which implies that the
input text-files to COMSOL. Calculated calculation model is either too weak or too
harmonic x-displacement amplitudes at the 6” heavy, or both. The calculated result is thus
choke valve actuator tip are illustrated in figure conservative. Also, measured stress (strain
11. As seen in the figure the movements are very
low and we notice that the maximum
Excerpt from the Proceedings of the COMSOL Users Conference 2007 Grenoble

gauges) showed low values so the design is


considered as safe. 10. Acknowledgements

8. Conclusions The author wishes to thank STATOIL A/S


for the permission to publish this work. Also, the
The frequency study shows that in the low author gratefully thanks Nils-Petter Dalstø,
frequency band (0-10 Hz) the eigenmodes show Dedico A/S for his contribution with response
larger amplitudes in the large-diameter-pipes. In spectra and other measurement data.
the frequency band 10-17 Hz the vibration in the
6" & 8" pipes dominate. Specially, around 11-12
Hz we find the largest movements in the choke
valve actuators. The higher modes - over 17 Hz -
are shell vibration modes with fairly small
amplitude.
The vibration behaviour of the internals -
inlet gas box and the inlet device - shows rather
small movements in the investigated frequency
band.
Also, the Tecklok nozzles and the
reinforcement plate show small vibration
amplitudes.
As the main concern is the forced vibrations
arising from the platform movements, which are
in the range 0-2 Hz, and their influence upon the
internals and the choke valves we can draw the
conclusion that no large vibration will occur in
these items in the low frequency band
mentioned.
A harmonic analysis confirms that the
movements in general are small.

9. References

1. Bathe, K. -J., Finite Element Procedures in


Engineering Analysis, Prentice-Hall, Englewood
Cliffs, New Jersey 1982
2. Cook R., D., Concepts and Applications of
Finite Element Analysis, 3:rd ed., John Wiley &
Sons, New York 1989
3. Zienkiewicz, O., C., and Taylor, R., L., The
Finite Element Method, vol 1: The Basis & vol
2: Solid Mechanics, 5:th ed., Butterworth
Heinemann, Oxford 2000
4. COMSOL, Multiphysics User’s Guide,
COMSOL AB 2007
5. Blevins, R., D., Flow-Induced Vibration, 2:nd
ed., Krieger Publishing Company, Malabar 2001
6. Dalstø, N. -P., Vibration Measurement Report
(in Norwegian), Technical Report, Dedico A/S
2006

You might also like