Investigating & Improving the drooping curve

of a two-stage feed pump

Ng Tzuu Bin, Flowserve
 Presenter Biography
• Ng Tzuu Bin is the specialist engineer for Aftermarket
Services & Solution (AMSS) Engineering team based in
Flowserve Singapore. He has 10 years of work experience
in the pump and seal aftermarket business. He obtained
his PhD from the University of Tasmania Australia,
specializing in the unsteady Francis turbine operation for
hydroelectric plant.

2
 The Abstract
A two-stage feed pump exhibited a drooping head-flow
characteristic during its shop test. Impeller reworks were
done to improve the drooping curve. CFD study was
performed to examine the pump flow behaviour and a more
stringent test procedure was implemented. The key lesson
learnt from this case is not to overly push the efficiency of
the pump at a single best efficiency point, but to have a
more balanced design between achieving good pump
efficiency and attaining a stable curve.

3
 The Outline
• The Problem
• The Description of the Pump & the System
• The Rework & the Corresponding Shop Test Results
• The CFD Analysis
• The Refined Test for Shutoff Measurement
• Conclusion

4
 The Problem
• Two-stage between-bearing pumps of the upgraded
material were supplied to replace the existing machines.
• The pumps utilized a higher no. of impeller blades to
obtain higher efficiency & head at its design point, but
exhibited a drooping Q-H characteristic during shop test.
• Only eight serial numbers were found for this pump size.
No drooping curve was previously reported. Deviation in
cast geometry could be the probable cause.

5
 The Definition of Head Droop
• The head droop (Hd) occurs when the
pump TDH does not rise continually
when moving from BEP to shutoff.
• This could generate the static instability
for a pumping system with high static/
pressure head & negligible friction head.
• API 610: “pump with continuous HRTSO
is preferred for all applications & is
required for parallel operation.”

6
 The Pumping System
• The pumps would NOT operate in parallel and at low flow.
• Flow meter was installed at discharge line for monitoring.
• The system curve is made up of the moderately low
static/ pressure head and the non-negligible friction head
where drooping pump curve will NOT present problem.
• Customer requested OEM to investigate and improve the
amount of head droop on the test curve.

7
 The Pump Information

DE NDE

2 1
• BB2: 2-stage/ horizontal/ between bearing/ radially split.
• Pump material upgraded from CS to 316SS for corrosive crude.

8
 The Original Pump Impeller Design

Eye Bull Ring

Vane  Efficiency & Head 

• Stage 1 impeller: double suction, 5 vanes, eye bull ring.

• Stage 2 impeller: single suction, 8 vanes.
• Design specific speed, Ns ~ 1350.

9
 The Pump Impeller Effect on Droop

TDH

TDH

TDH
TDH

Q Q Q Q
Source: Yuan et al.

• Many design parameters affecting the droop: vane no. (Z),

exit angle (2), exit width (b2), exit vane thickness (Su),
diameter (D2), meridional shape, inlet vane position, etc.
• These design parameters are very difficult & expensive to
modify once the impeller has been casted. Minor rework is
more appealing here.
10
 The Initial Test Setup
PT for Suction Nozzle
PT for Discharge Nozzle

Flow Meter

• Test carried out in accordance with API 610 11th Ed.

• Pressures, flow, speed & motor power were measured.
• Pressure gauges were installed 2D away from pump nozzles.

11
 The Pump TDH Calculation
𝑉𝑑2 𝑉𝑠2
TDH = ℎ𝑑 − ℎ𝑠 + − + 𝑍𝑑 − 𝑍𝑠
2𝑔 2𝑔
Where ℎ𝑠,𝑑 = 𝑃𝑠,𝑑 Τ𝜌𝑔 = suction/ discharge pressure head
2
𝑉𝑠,𝑑 ൗ2𝑔 = 𝑄 2 ൗ2𝑔𝐴2𝑠,𝑑 = suction/ discharge velocity head
𝑍𝑑 −𝑍𝑠 = suction & discharge gauge height differential
𝑃𝑠,𝑑 = measured suction/ discharge pressure
𝑄 = measured flow
𝐴𝑠,𝑑 = suction/ discharge pipe cross sectional area

• Water density  was assumed constant.

• Flow velocity V was assumed uniform over pipe cross section.

12
 The First Shop Test Result

Hd ~ 7.3m

• TDH peaked at ~35% BEP flow in the first shop test. 7.3m
head droop was observed here.
• Head droop was not reported in the test done 22 years ago.

13
 The Rework 1: Impeller Oblique Cut & V-Trim
Flat
Flat
Oblique
V Trim Cut
Flat
Flat

Stage 1 Impeller Stage 2 Impeller

• Stage 1 Impeller (DS): V-trim angle 20.
• Stage 2 Impeller (SS): Oblique cut angle 13.
• This will reduce the secondary flow loss caused by streamline
differences between impeller shroud and hub.
14
 The Second Shop Test Result – Rework 1

Hd ~ 6.3m

• Minor improvement observed. 6.3m of head droop

remained in the test.
• Rated efficiency dropped ~1.1% (Compared to first test).

15
 The Rework 2: Impeller Vane Squaring

90 cut

Stage 1 Impeller Stage 2 Impeller

• Both Impellers: Cut the exit edge of vanes radially.
• This could reduce the wake shedding area due to traversing
flow at the exit. Further trimming is needed, as the rework
will rise the TDH at all capacities.
16
 The Third Shop Test Result – Rework 2

Hd ~ 4.2m

• Substantial improvement obtained, but 4.2m of head droop

remained in the test.
• Rated efficiency dropped ~1.3% (Compared to first test).

17
 The Rework 3: Inlet Guide Vane at Casing

• Install the guide vanes at inlet of stage 2 impeller. The guide

vanes extend right into the suction eye.
• This will reduce the hydraulic loss due to inlet pre-rotation.

18
 The Fourth Shop Test Result – Rework 3

Hd ~ 3.7m

• Minimal improvement at low flow. Effect was more obvious

at high flow. A 3.7m of head droop recorded in the test.
• Rated efficiency dropped ~1.5% (Compared to first test).

19
The Rework 4: More Impeller Oblique Cut & V-Trim
Flat
Flat
More
MoreV Trim Oblique
V Trim Cut
Flat
Flat

Stage 1 Impeller Stage 2 Impeller

• Stage 1 Impeller (DS): Increase V-trim angle to 29.
• Stage 2 Impeller (SS): Increase oblique cut angle to 32.
• Positive outcome obtained for this method in the past.

20
 The Fifth Shop Test Result – Rework 4

Hd ~ 4.1m

• Head droop increased to 4.1m against the expectation. Flow

recirculation overwhelmed the impeller streamline difference.
• Negligible change on rated efficiency.

21
 Summary of the Rework
• Impeller V-trim/ oblique cut generated minor recovery on
head droop. Higher cut angle could have negative impact.
• Impeller vane squaring showed significant improvement
in head droop, but further trim is required for rated TDH.
• Extended inlet guide vanes at stage 2 casing produced
very minimal impact on head droop recovery.
• Improvement on the head droop was obtained at the
expense of pump efficiency. None of these well known
methods could completely remove the head droop here.
22
 CFD Shows Potential Problems
Inflation
layer

Flow Domain Meshing

• Inlet/ outlet boundaries extended 10D away.
• 30M unstructured mesh elements with inflation layer.
• Frozen rotor approach & SST turbulence model applied.

23
 CFD Result – Pump TDH prediction

• CFD over-predicted the TDH rise below the minimum flow.

• CFD did not clearly reproduce the head droop at the test.

24
 CFD Result – 2D Streamline Plot
5% BEP 83% BEP 5% BEP 83% BEP 5% BEP 83% BEP
(Rated)

Stage 1: 5-Vane Impeller Stage 2: 8-Vane Impeller Crossover

• Flow recirculation & separation in the volute & crossover.
Such phenomena become more severe at low flow and will
extend further into the suction/ discharge pipe.
• Highly transient flow observed below 35% BEP.
25
 Summary of the CFD Analysis
• CFD provides a good qualitative assessment, although it
doesn’t clearly reproduce the head droop effect from test.
• Flow recirculation & separation in the casing cause huge
static and total pressure losses, especially at low flow.
• Pump flow is highly transient below 35% BEP. More time
is needed for pump to settle down at low flow.
• Impeller rework will not likely resolve the head droop.
Further casing modification is very costly.

26
 Testing at Pump Shutoff
• Two problems:

 Flow is highly transient as learnt from CFD. More time is

required for pump to settle down during shutoff test.

 Continuous pumping temperature rise at shutoff.

Calculated temperature rise at shutoff is 4-9C/min.
Static pressure head will be affected as the fluid density
is temperature dependent.

27
 The Modified Test Setup
RTD Probe

High Speed
Data Logger

• Fast response RTD probe was installed at casing vent to

monitor the temperature rise.
• Higher sampling rate was set in the data logger.

28
 The Pump Shutoff Test Result

• Water temperature rise effect must be corrected at shutoff.

• TDH fluctuated more at the reduced flow.

29
 The Pump Shutoff Test Result

• No hunting curve effect was observed during the low flow test.
• Unsteady dynamic pressure head was not accounted in the
TDH calculation.

30
 The Pump Shutoff Test Result

0 Rated Point

Drooping Point

Shutoff Point

• Standard deviation of the measured TDH  as Q .

• The measured head droop was statistically insignificant.
• Customer witnessed the test & accepted the result.

31
 Conclusion – Lessons Learnt
• A balance between achieving good pump efficiency and
attaining a stable curve must be considered at early
design stage. Fixing head droop at the test stand is
expensive (>$100K) and time consuming (4-6 months).
• Minor rework could improve the head droop but at the
expense of pump efficiency.
• CFD provides good qualitative assessment on the head
droop.
• Test setup must be refined for pump showing <5% HRTSO.
32
 References
• De Neumann, B., “ The interaction between geometry and
performance of a centrifugal pump”, Mechanical
Engineering Publications UK, 1991.

• Lobanoff, V.S. & Ross, R.R., “Centrifugal pumps design &

application”, 2nd ed., Gulf Publishing USA, 1992.

• Yuan et al., “Design method of obtaining stable head-flow

curves of centrifugal pumps”, Pumping Machinery
Symposium, ASME FED Vol. 154, ASME, 1993.

33