MECHANICAL DAMAGE TO PUMP
IMPELLERS CAUSED BY LOW FLOW
OPERATION. High-cycle fatigue damage imparted on a BB5 pump
impeller operating at low flow operation points was
analyzed via numerical methods.
A 1-way Fluid-Structure interaction approach whereby a
transient CFD subsequently mapped to a structural FEA
showed a significant increase in fluctuating stress
ranges as the flow decreased.
The authors present several mitigation options in the
case where low flow operation cannot be avoided.
1. SYMPOSIA: 24 – 26 MAY 2022
SHORT COURSES: 23 MAY 2022
MECHANICAL
DAMAGE TO PUMP
IMPELLERS CAUSED
BY LOW FLOW
OPERATION
Christopher Shages
F l o w s e r v e C o r p o r a t i o n
S t r u c t u r a l M e c h a n i c s G r o u p
Péter Tóth
F l o w s e r v e C o r p o r a t i o n
F l u i d D y n a m i c s G r o u p
2. Presenters
Christopher Shages
• BS & MS in Mechanical Engineering, Lehigh University
• 10+ years in Flowserve’s Pump Division R&D Group
• Currently working as an Applied Structural Mechanics
Specialist
Péter Tóth
• MS in Mechanical Engineering, Budapest University of
Technology and Economics, RM Von Kármán Institute
• 4 years at Flowserve, 13 years experience in
Computational Fluid Dynamics
• Currently working as an Applied Fluid Dynamics
Specialist
3. High-cycle fatigue damage imparted on a BB5 pump
impeller operating at low flow operation points was
analyzed via numerical methods.
A 1-way Fluid-Structure interaction approach whereby a
transient CFD subsequently mapped to a structural FEA
showed a significant increase in fluctuating stress
ranges as the flow decreased.
The authors present several mitigation options in the
case where low flow operation cannot be avoided.
Abstract
4. A boiler feedwater pump servicing a combined
cycle power plant began operating beyond
acceptable vibration levels and was pulled for
inspection.
• Problem was traced to 1 st stage impeller where
a section of the shroud wall had broken off.
• The incident led to more frequent inspections
which found cracks in the 1 st stage impeller
were appearing regularly.
Cracks were found only in the 1st stage
impellers.
• Appear between 3 months and 1 year of
operation.
• Same location: shroud side trailing edge fillet.
Background
5. Metallurgical Analysis Conducted
• Did not find any anomalies. Concluded material defects
were not the cause.
Review of End User’s Operational Data
• The pump was operated below prescribed minimum flowrate
for brief periods.
• No VFD, no min flow recirc valve.
• End user requested an investigation as to whether operation
below minimum flow could be a source of the cracks.
Numerical Analysis
• CFD Analysis: Transient, Full Wheel w/ front & back leak
paths, 30+ revolutions, 1° of rotation per time step.
• Flowrates of 100% of BEP, 50%, 25%, 17%, 9%.
• Pressure data transferred from CFD to FEA.
Root Cause Analysis Overview
6. Pressure Results from CFD Input to FEA Analysis FEA Results scoped to
area of interest
Modified Goodman
Diagram
Transient/Cyclic Stresses
Area of Interest
High Fidelity CFD Setup
Analysis Process Map
7. Bad Flow increases the variability of the local pressure at the
rotor-stator interface.
A flow visualization technique in the CFD post processing was
used to highlight the regions of Bad Flow.
• Velocity components were broken down into radial and tangential flow.
• Good Flow is where the radial velocity is outward from the rotating
axis, and tangential flow is in the same direction as rotation.
• Velocity plot legend set such that positive direction velocity is red
indicating Good Flow and negative direction is blue indicating Bad Flow.
Centerline Plane plots of
Tangential velocity at 100% BEP
flowrate (LEFT) and 10% BEP
flowrate (RIGHT).
Fluid Dynamics Analysis
8. Bad Flow
Velocity plots showing where flow
is moving radially inward (right)
or in the opposite direction of
impeller rotation (left).
The Bad Flow is colored Blue.
At 100% BEP, nearly all the flow is
moving in the proper directions.
At 9% BEP, 9 out of 12 diffusor
passages are fully stalled.
100%
BEP
Flowrate
50%
BEP
Flowrate
9%
BEP
Flowrate
9. 9% BEP Flowrate
Flow coming
off th e
imp eller
can n ot enter
th e d iffu sor
and is forced
b ac k into th e
ou tlet of th e
imp eller.
Diffu sor
passage’s flow
d irec tion is
b ac kward s!
F low enterin g
th e imp eller ’s
ou tlet
Bad Flow at Rotor-Stator Interface
10. Pressure Data at the Impeller OD
0
5
10
15
20
25
0 0.5 1 1.5 2 2.5 3
p
g
[bar]
rotor rev. [-]
50% BEP @ ps
0
5
10
15
20
25
0 0.5 1 1.5 2 2.5 3
p
g
[bar]
rotor rev. [-]
100% BEP @ ps
0
5
10
15
20
25
0 0.5 1 1.5 2 2.5 3
p
g
[bar]
rotor rev. [-]
25% BEP @ ps
0
5
10
15
20
25
0 0.5 1 1.5 2 2.5 3
p
g
[bar]
rotor rev. [-]
17% BEP @ ps
Location: ps
A Monitor Point in the
CFD analysis was used
to understand the
variability of the
instantaneous pressure.
The periodic pressure
signal become more
erratic as the flowrate
decreased.
11. Pressure & Head Analysis
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
9% 17% 25% 50% 100%
%
of
mean
value
Flow rate % of BEP
Stage Head RMS
Local pressure RMS @ ps
Local pressure peak to peak @ ps
Th e total d ifferential h ead ac ross
th is stage ( in let to ou tlet) d oes n ot
p rod u c e large flu c tu ation s ( < 3 %) ,
even at low flowrates.
Th e variab ility of th e R MS p ressu re
at th e Mo nito r Po int is gen erally low
as well ( < 1 2 %) , even at low flows.
Th e variab ility of th e in stantan eou s
p ressu re at th e Mon itor Point,
measured peak -to-peak is quite high
( >8 0 %) .
• Th is oc c u rren c e wou ld n ot b e ab le
to b e mon itored or d etec ted by
ty p ical p u mp in stru mentation .
Inlet
Outlet
12. Structural Analysis
Mapping the transient pressure and
adding a centrifugal load to the
structural FEA revealed a peak stress
location which coincides with the
likely area of crack initiation.
13. The peak stress for the
critical fillet location was
plotted at each of the 7
blades for various
flowrates.
The periodic nature of the
stress is a result of an
increase in local pressure
as the impeller blade
rotates past the diffusor
vane.
Structural Analysis
14. A s flowrate d ec reases, th e stress
ran ge in c reases sig n ificantly.
• 4 x at 1 8 % B EP vs . 1 0 0 % B EP!
Structural Analysis
15. Cyclic stresses were
extracted from these plots
for the purpose of fatigue
analysis.
Note: only 120 degrees of
rotation was considered
due to computational
constraints. It is quite
possible that higher peak
stresses were not captured.
Structural Analysis
16. Fatigue Analysis
Cyclic stresses were
extracted from these
plots for the purpose
of fatigue analysis.
Large reduction in
factor of safety with
respect to Infinite Life
with decreasing flow.
Material is CA6NM
17. Fatigue Analysis Summary
Although Modified Goodman fatigue analysis technically indicated
“infinite fatigue life” for the impeller at low flows, the variability
of the many important of inputs means we are not surprised by the
result of numerous but not all the seven impeller blade fillets
experiencing crack initiation and growth.
Potential sources of error in Fatigue Life Analysis calculations:
• Corrosion effec ts ig n ored
• Resid u al S tresses from castin g / q u en c h in g
• S tress risers from th e fillet b ein g rou n d ed by h an d g rin d in g
• S tress risers from u n d ersid e of fillet b ein g a ‘cast su rfac e’
• Fatig u e life of cast su rfac e h as h ig h scatter
• Pressu re d ata was on ly samp led for 1 2 0 d eg rees of rotation
• S p ec ified trailin g ed ge fillet valu e of R =.0 2 mm is c reated by h an d
g rin d in g op eration an d is imp ossib le to ap p ly a p rec ise toleran c e to.
18. The cracks on the impeller were highly likely to have
been caused by the increasing stresses at the critical
area when flow was reduced below 20% of the BEP
flowrate.
Fatigue life calculations have a high level of
uncertainty due to the variability of many of the
inputs.
• CFD may not have captured the highest pressure -
pulsation the impeller experiences due to the
complex nature of simulating chaotic flow.
• The actual geometry of the impeller at the critical
location comes from a hand grinding operation and
cast surface.
Mitigation efforts to reduce susceptibility to high
cycle fatigue are discussed in the subsequent slide.
Summary
19. • The typical reasons for avoiding operation at low flow are generally limited
to th ermal, vib ration al, an d cavitation con c ern s, as well as p oor effic ien c y
an d d iffic u lties arisin g from op eration on a ver y flat sec tion of th e
p erforman c e c u r ve.
• Th is stu d y sh ows cyclic fatigu e d amage can b e a con cern with res p ect to
OEMs specif ying a minimum flowrate .
• Mitigation efforts with resp ec t to imp eller d es ign in c lu d e:
• Red u c in g th e ou ter d iameter of th e imp eller in ord er to in c rease th e gap
b etween th e imp eller O D an d th e d iffu sor ID will red u c e th e mag n itu d e of
th e p ressu re p u lsation s.
• In c rease th e trailin g ed ge fillet rad iu s
• Imp rove su rfac e fin ish in c ritical area
• Mitigation efforts with resp ec t to p u mp op eration in c lu d e:
• D o n ot op erate b elow min imu m flow ( ~ 3 0 % B E P)
• Red u c e motor sp eed ( u tilize V F D ) to red u c e flow.
• A d d an au tomatic rec irc u lation valve.
Lessons Learned