The Use of Finite Element Method To Solve Vibration Problems On Vertical Pumps
Uploaded by
ashok pandiyanThe Use of Finite Element Method To Solve Vibration Problems On Vertical Pumps
Uploaded by
ashok pandiyanSESSION12 CASESTUDIES
THE USE OF FINITE ELEMENT METHOD TO SOLVE
VIBRATION PROBLEMS ON VERTICAL PUMPS
Massimo Scali
Giacomino Marenco
Weir Gabbioneta Srl, Sesto S.Giovanni (MI), Italia
ABSTRACf
Vertical pumpS afe often a matter as far
as regardsvibrations. Their design is very
simple but aften is cause af high level
vibrations.
Tbe only way tOpreventthis problem
is to know the dynamic behaviour of
tbe machine in design phase, by finite
element metbod calculations.
We made such a study on an our
pump. Many approximations were made
and the mode l is somehow poor,
neverthelessthe results were encouraging.
INT~ODUCTION: THE GENERAL
PROBLEM
In fluid pumping sometimes the available
Net Positive Suction Head (NPSH) isn 't
enough to avoid cavitation. One way to
solve tbe problem is tO .increase it by
setting tbe impeller severa! meters under
ground level. Therefore you must dig a
hole in the ground and put in it a ban-el
which will contain tbe fluid and the impeller,
maintaining tbe motor at, ground level.
Thus tbe power will be transmi~ to
tbe impeller by a long overhung shaft.
The depth of tbe ban-el depends by NPSH
available at ground level and by NPSH
required by ! 5t stage of tbe pump. The
number of impellers is variabl~ too
Fig.l - A VB type pump.
depending by required head (fig.!).
Al! theseconditions make it impossible
tO build a stiff shaft, because we should
ROTATING KACHINE DYNAMICS '92 STAFFORDSHlREPOLYTECHNIC
288
design a very large diameter. Thus the
pump is to ron among its resonancespeeds.
It's easy to understand that it's necessary
tO know exactly where these resonances
aIe.
There afe two way to know it: me
first is to build the machine and then
dynamically test it. Of course tbis soluti()n
is very expinsive and doesn't allow sttong
rrK>difications
on the originai design,without
implying high costs and long rimes.
The secondis tO calcolate them using
a finite element model of tbe entire pump
considering every single component like
motor, baseplate,as well as sleevebearings
and so on. This way aJlows tbe designer
to changetbe pump before it's built Besides,
il you canusea precise methodof calculation
for seals and bearings, you can use tbe
exact pumped fluid and not simply water,
which can bave a dynamic behaviourother
tban actual one.
A TROUBLE CASE
Our study was developed on a vertical
pump which was installed in Venezuela
two years ago. Its task is very imponant
and withòut it the whole plant stops. The
pumped flujd is isobutane. The operaring
condirions required a four stages pump,
running at 3550rpm, correspondingto nearly
60 Hz, (the actuaI Speed tumed out to
be 3480 rpm, i.e. 58 Hz). The NPSH
imposed to put the first stile at 5 meters
under ground level. Along me shaft afe
seI out four guide bushings at a distante
of 750 mm each other.
We tested it in our test room and
verified a perfectrunning, both for hydraulic
performancesand for dynamic behaviour.
Of course the test was carried out at 2950
rpm (50 HZ) and with water as pumped
fluido Neverthelesswhen it was staned on
actuai planI the result was very different Fig.2 - A view of the mesh used for
the vibration level was too high and for the calculation.
289
25-MAR-91 :8: 09: 03
UNITS : MM
OISPLAY: NASCOSTE (",Oditied)
Fig.3 - A closeup view of tbe supportmesh.
this reason some bushings were damaged making the vibrations rncreaseagain
(until 14 mm/s).
-mE APPROACHTO mE PROBLEM.
Customer was of course very impatient to salve the problem. In spite of this
we had poor info~tions frortl the pianto Thus we had to make a fuii simulation
of the phisical phenomenonon our computer. For this wk we decided to use
the finite element method (FEM). In fact if you make a precise mesh and the
boundarycondirions afe properly set, it's possible to reach the right solution without
any test on reaI structure.
We had to make some hypothesis to begin to move in one precise direction.
Our basic hypothesis was that the machine was runRing hear a criticai speed (we
didn't know which oDe). We knew that some shaft guide bushings were WOrD
and we knew the amount of the wear too: this infonnation was helpfui to detect
the moda! shape.
A finite elementS model of the pump was made using the FEM program
I-DEAS by SDRC: it's drawn in figg.2 and 3. It was composedby nearly 1200
elements and 1700 nodes. The elementSused were bricks (linear, 8 nodes) and
290
Fig.4 - Moda! shaperesonating
at S8 Hz.
(: ~ :"-r~ 't" r ~ :\1:'
'-,. I .! 1:3
-
Fig.S Fre.quencyresponse of shaft
Fig.6 - Frequencyresponscor the
shaft nooe on the scal.
node on motor join
joint.
bYt~mtJ;&,t.om;'IJ~'.$8i~fY8Jr'I~(-o,
_1Ji=~:!j ,,!, ~~:jJ~::II:::I ~
J
I
FiI.7 - MOOaI shape Il 45 Hz.
291
quadrilateral shells (linear, 4 nodes) far the column and the casing, while linear
bearns were used far the shaft. Lumped massessimulated the motor, the impellers
and bushings inenia. The shaft and the column were connectedby spring element,
whose stiffnesses were calculated by a self-made program far evaluaring dynamic
characteristicsof seals and oil-fJlm bearings.
The whole structure was restrained by four springs simulating ten'ain effect,
whose stiffness is not infmitive. The stiffness used far these elements was read
by literature and was assumed to be 9.806e9 N/m.
We had an unce11aintyab9ut the value of shaft guide bushings stiffness,
becauseour method of evaluating oil-film dynamic behaviour is quite precise if
there is a pressure gradient across the film, but is not if pressure is nearIy
constant, and that's our case.
We founded our process on the fact that alI parameterswere well known,
but dynarnic behaviour of shaft guide bushings. Therefore we fixed alI known
parametersand varied only the unknown 'one. Several calculations were made to
see how the struc~ would react to the variations of bushings stiffness. We used
at this piont the nypothesis we made: the pump was operating in resonance.The
infonnations regarding the wom bushings' too were useful in this process phase
to detect the moda! shape and by this to detennine the right value of the stiffness
far shaft guide bushings.
Here are the values used for every spring stiffness in the model:
1. seal on tbe casing: 5.394e5N/m;
2. bushingsalong the shaft: 4.903eSN/m;
3. impellersseal.s:3.383c6N'in
4. ball bearings:9.807e8N/m.
The next step Wa$to modify the structureonly for tbe parameterswe were
able to evaluate witb good precision and tben to ana1ysetbc results to understand
whetber tbc modi.fications made could reduce vibration levels enough.
1HE CALCULA110N RESULTSANALYSIS AND n:IE MODIFICAnON.
Following the previously described steps we were able 10 say that the pump
resonanceexcited during normal running was in corrispondenceof the 4th frequency
of the sh~t (tbe column and tbe suppon resonancefrequency were far from
operdting one). In fig.4 is the mode correspondingto the resonating frequency,
while in fig.5 you can see the frequency responsefor the shaft node on the
moto,l'joint and in fig.6 the frequency responsefor ~e shaft node on the seal.
Observingthese diagrarnswe can see that under the operatingfrequency severals
resonancesexist artd they belong 10 column, to suppon and, of course, to shaft.
The :Iower modes or the shaft are in a little range near operating speed and for
this reason we couldn't make drastical modifications on the strutture because it
was.possibletbat one of thesemodesincreasedits frequencyuntil operatingODe.
In fig.7 it's shown the 3rd mode of the shaft, tbc nearer to ronDini Speed.These
292
Fig.8 - Moda!shapc n.4 for I1'K>djfied
pump (65 Hz).
t.t, . l' 1 J . f n' J'. .- . I
-IN.t. . \. - \ft/. \. . I :lJ. I
,:""'. . . . I . +. . _I . .- '-1
Fig.9 - Frequencyresponse
or shaft Fig.)o - Frequencyresponse
or thc
n<xle on motor joint shaft node on me scal.
Fig.11 - MOOaJshape n.3 for m<X1ifiedpump (50 Hz).
293
considerarionsmake it clear that we could neither make large errors in computation,
because a wrong evaluation could bring 10 criticaI sitUations for the machine.
We thought that the besl and fastest way to solve the problem was to shift
the frequency that was found to coincide with running one by shonening the
shaft and the column. The measures taken on the plant showed an available
NPSH high enough, thus it was possible to make this modificàtion without
compromise the good hydraulic operation. This solution was even quite fast to
implemento requiring few modifications on the originai designo Thus we decided
tO shonen the pump by 500 mm.
The caIculations results were encouraging: the 4th shaft mode was icreased
to 65 Hz (fig. 8), while the 3rd shifted 10 50 Hz (fig. l 1); the frequency response
diagrams for the same nodes we examined previously show that the pump is
going to run in a minimum point as far as regard vibrations amplitude (fig.9
and lO).
Therefore this solution was implemented. The test on the plant confirn1ed the
calculation, in fact the vibration levels decreaseduntil 7 mm/s and mantained this
value.
CONCLUSIONSAND FUTURE DEVELOPMENTS
We know tltat many uncertainties affected the whole study. On the other side
we didn't get any infonnations from tbe plant, but tbe vibration level (no impact
tes~ nor order n-acking). Anyway the results were satisfactory and encouraging.
We're acquiring tbe moda! analysis package for I-DEAS, and by this we
think to develope a method to analyze systematicalIyour vertical pumps before
they afe manufactured.Of course the results obtained will be compared :.vith tests
on actual pumps in our plan~ by impact test and order tracking using a dynamic
signal analyzer.
We intend to develope a more precise and wider application rangc program
too for evaluating tbc dynamic behaviour of scals and bearings.
Anyway wc think it's today possfule to reach a vcry high computational
precision by finite element analysis using a powcrful workstation. and it's in our
opinion that by these rneans you can fully predict tbc dynamic behaviour of a
machine when it is stilI in designer's mind.
