You are on page 1of 4

www.pumpengineer.

net 19
COLUMN: THE PUMP SURGEONS
The Pump Surgeons are back with a new series in which they delve into What Turns a Good Pump Bad.
In Part 1 of the series, pump experts Ian James and George J. Maddox lay the foundation of their series
by putting on their detective hats and outlining how they solved the mystery of why a previously reliable
process pump in a chemical plant suddenly went bad.
By George J. Maddox & Ian James, Best PumpWorks
Imagine owning a car, which
is regularly serviced and
maintained, and it lasts for 20
years which at an average of
12,000 miles a year is nearly 1/4
million miles. Oh, and this is not
a freak occurrence, every model
does this. They would outsell
every other car manufacturer, or,
they would make pumps because
the vast majority of pumps are
designed to last for over 20
years, and as API610 calls for
this, it happens consistently.
Correctly operated and
maintained, pumps are one of
the most long-lived and reliable
pieces of equipment. However,
the reliability of any piece of
equipment is directly affected by
understanding its requirements
and limitations. To assist in this,
we will illustrate several real
world examples that show, very
adequately, the phrase: a step
too far. We will give several
examples in this article, and will
follow with other examples in
future articles, to help develop a
pump troubleshooting matrix, as
the last article in this series.
Example 1: When good
pumps go bad
Some 14 years ago, several
heavy duty, overhung, API610
OH2 process pumps were
supplied to a chemical plant for
a batch type, stop/start process.
The pumps had been running
reliably for over six years,
with no problems, and nothing
required except to follow the
recommended maintenance Figure 1: The Good: The Original Batch System: A sketch of the original system.
routines. The batch process
was core to the operation of
that part of the plant, and based
on the batch process success
and equipment reliability, the
customer discussed plans to
expand the process and the
pumps that were required to
fulll this need. What we later
discovered, was that in order
to assess the economics and
operational philosophy of this
planned system expansion,
the system would be stretched,
by adopting low cost changes
to increase the batch operation
time, by increasing the liquid
volume levels in the tanks. As
it is with most problems, it is
not what was changed that
caused the problems, it was
how it was changed.
All was well, until shortly after
the batch process ow rate was
increased to full capacity, when
suddenly, the pumps started
suffering vibration issues,
cavitation problems, damaged
bearings, and seal issues. Not
just to an isolated pump, but
every pump and stand-by
N
E
W

s
e
r
i
e
s
:

W
h
a
t

T
u
r
n
s

a

G
o
o
d

P
u
m
p

B
a
d

-

P
a
r
t

1
PUMP
SURGEONS
The smoking gun:
Examples of
fault nding
Pump Engineer, September 2014 20
COLUMN: THE PUMP SURGEONS
Figure 2: The stretched system when full.
pump that was fed by this batch
process, suffered problems.
The good pumps suddenly
went bad; it was like Jekyll
turning into Hyde. We survive
and ourish as a race, by the
predictability of physical laws,
including hydraulic, corrosive,
and mechanical laws. So, if a
pump has been running reliably
for a long period of time, there
must be a step change cause
that suddenly starts to damage
the pump. For sure, a faulty
component within any pump can
result in an immediate change
in pump reliability. But in this
example, when the pumps were
stripped down and examined,
there was no evidence of a
faulty component. Of course,
if not regularly maintained,
any pump user will start to
see increasing trends in some
of the measured parameters,
such as higher vibration levels,
perhaps the onset of cavitation
(as would result from the use of
unmonitored suction lters), or
increased bearing temperatures.
However, they would be unlikely
to see the sudden step change
in these measured parameters,
as had suddenly occurred in this
installation.
The most likely cause of all the
parameters suddenly changing
from good (low) to bad, such as
high vibration levels, suction
surging and cavitation, is a
change in the liquid conditions
feeding the pump. We carried
out an in-depth and detailed
investigation on the system
and liquid to try to identify the
changes that may have occurred.
The customer was adamant that
the liquid characteristics were
as they had always been, and
his understandable frustration
became obvious with the
statement: The only change is
that all your pumps have stopped
being reliable. We took the
customer up on his invite to
immediately visit the site, with a
promise that all facts would be
made available to us. This was a
visit that we had planned as a
fact-nding exercise and to
rebuild the customers condence
in what he had claimed: the month
before these issues surfaced,
were the most reliable pumps in
his whole plant.
The essence of any good
detective story is the investigation
that gradually reveals the facts
that lead swiftly to the culprits
and causes. This analogy is used
because as our investigations
continued, it became evident that
every parameter that a pump
relies upon for good operation,
had changed from good to bad.
This batch process involved
pumping liquid from a large tank,
through a chemical batch process,
until the tank liquid levels
dropped to the minimum trip
level. Our investigations became
more focused when we were told
that the pump problems did not
occur when the tank was full,
but strangely, started to happen
when the liquid level dropped
to just below 40% full. This was
a liquid level at which, in the
original system, the pumps had
previously operated successfully,
without any event occurring.
Yet now, this relatively high
liquid level resulted in pumping
problems that went from bad
to worse as the levels dropped.
Prior to reaching the minimum
level setting, the pumps had to be
stopped since the bearings were
being damaged and mechanical
seals were leaking. This was their
rst occurrence of bearing and
mechanical seal problems. As
the full story evolved there were
several eureka moments.
Figure 1 shows a sketch of
the original batch system. In
this system, the liquid levels
were maintained within very
conservative levels, with the tank
not being very full at the start of
the process, and not draining
down to extremely low levels
by the end of the process. The
nitrogen pressure blanket was
a constant pressure design, so
as the liquid levels reduced, the
system fed a little more nitrogen
into the tank to maintain the
pressure. It was within this system
that the pumps had run reliably
for six years. In order to assess
the effects of pumping a larger
batch volume into their system,
the most economic method of
increasing batch volumes was
www.pumpengineer.net 21
COLUMN: THE PUMP SURGEONS
used, by moving the level settings
higher and lower. The engineers
were convinced the only effect
on the pumps is that they would
run for longer periods. This
modied, or stretched system,
as the customer preferred to
refer to it, is shown in Figure 2.
In order to maximize the liquid
level, the decision was made to
use an alternative anged liquid
inlet connection on the tank that
happened to be located on top
of the tank. The previous side
connection was just blanked off.
This of course, had no effect on
the pump reliability, as the pumps
just saw an increase in the inlet
or suction pressures due to the
higher liquid level.
Another change that was made
was that of increasing the pressure
of the nitrogen blanket, shown in
the vapor space in the tank. The
philosophy used was that since the
tank liquid levels were lower, to
help maintain the suction pressure
at the pump, the nitrogen pressure
would be increased in the tank. We
shadowed the process engineers
as they operated their stretched
system from full to empty, in
order to ensure we understood
all aspects of the system, and the
changes that were taking place
within it, to cause the problems.
Whilst the system was operating
from full to just below 50% full, all
parameters were good. The pump
ran quietly with low vibration
levels, and low noise levels at a
constant ow and head.
Then, from around 45% full, we
started to hear increases in noise
and vibration. As levels dropped
further within the tank, we saw
variations in both head and ow
as surging occurred, which
ended in sounds of cavitation and
high vibration levels. The events
occurring inside the tank, and the
problem they caused become
obvious when you look at the
sketch in Figure 3 that shows
the stretched system when at
a lower liquid level. As the tank
contents dropped to below 50%
the engineers had arranged
feedstock from another part of
the plant to start to replenish
the tank by owing into the inlet
connection. The intent was that
the inlet liquid ow was around
50% of the pumpout ow, and so
the batch process was extended
even further. Ironically, in using
the inlet connection at the top of
the tank to increase the liquid
levels and batch process times,
the engineers had created a
shower or waterfall effect which
had a progressively greater
effect after the tank levels
had dropped to only half full.
So although the liquid volume
was greater, the volume in that
problem setup resulted in a lower
useable or pumpable volume.
Sequence of Events
Here are the sequence of events
that occurred as this stretched
batch process started:
Event 1: Frothing
As the full tank contents started
to reduce to around 50% and
the feedstock inlet ow started,
Figure 3: The Bad and the Ugly: The stretched system when at lower liquid.
for every foot of reduced liquid
height in the tank, the inlet liquid
had a greater free-fall height,
and so frothing started to occur.
There was no diffusion of the inlet
stream to minimize this effect
there was only free-fall.
Event 2: Vortexing
As the liquid levels fell even lower,
not only did this effect increase,
but the engineers had made no
changes inside the tank to prevent
exit vortices from forming. There
were no vortex breakers in place
at all. As levels fell even further,
free surface vortices started to form
until a well-dened vapor core
developed at lower liquid levels,
resulting in bubbles and vapor
being drawn into the pump suction
stream, and so entering the pump.
Event 3: Nitrogen
As if these events were not bad
enough, a detailed investigation
into the characteristics of the
pumped liquid revealed that it had
a slight afnity to absorb small
Pump Engineer, September 2014 22
COLUMN: THE PUMP SURGEONS
amounts of nitrogen. But as the
nitrogen pressure was increased,
this slight afnity developed into
a major problem, resulting in the
liquid containing frothing bubbles
and dissolved nitrogen.
Event 4: Strainer
The nal nail in the poor pumps
cofn, was that the engineers
were so busy and focused in
investigating their stretched
system that they omitted to monitor
the suction strainer that was in
place to protect the pumps, since
the strainer had only previously
needed cleaning every few weeks.
However, the combination of the
liquid bubbles and the turbulence,
which caused some of the dissolved
nitrogen gas to come out of
solution, effectively blocked the
lter in a matter of days and led to
a cavitating pump.
The Solution
Once all the causes were known,
the solution was relatively
straightforward.
Changes to the tank:
The top liquid inlet connection
was blanked off and a new inlet
connection was cut into the side
of the tank ve feet below the
top of the tank. Also, a diffuser
was welded within the tank, such
that the incoming liquid stream
was directed to ow against
the walls of the tank where it
owed down to the liquid with
a much lower velocity and less
turbulence. This diffuser was
cone-like in shape and effective.
The lowest liquid level was
raised by two feet, and a cross-
type Vortex breaker was welded
over the exit port from the tank.
The Nitrogen:
The Nitrogen pressure was
reduced to 50% of its previous
level, to ensure it maintained its
pressure blanket effect, without
incurring the bad effect of being
absorbed by the liquid. This
was aided by the much reduced
splashing and turbulence within
the tank.
ABOUT THE AUTHORS:
George J. Maddox & Ian James
George J. Maddox, P.E,
is the Engineering/Global
Hydraulics Manager at Best
PumpWorks located in
Tyler, Texas with more
than 32 years experience
in a variety of engineering
design and engineering
management posions.
He is currently involved in
the design, remanufacture,
repair, re-rang, and
packaging of centrifugal
pumps. His responsibilies
include the hydraulic and mechanical design and design
evaluaons of centrifugal pumps and pump units. Mr. Maddox
has a BSME degree from the University of Texas At Arlington
and an MBA from Amber University and is a registered
Professional Engineer in the State of Texas.
Ian James is the
Engineering Manager at
Best PumpWorks located
at the Hobby facility at
Hobby, Texas. He has
more than 42 years
experience in various pump
engineering mechanical
design, applicaons and
engineering management
posions, covering both
mechanically sealed and
seal-less pump sectors. Mr.
James is involved in pump design, manufacture & applicaons,
and leads a team of Design, Project & CAD Engineers. He is also
a member of the API610, ASME B73 & API685 commiees.
The Suction Strainer:
The suction strainer was monitored
by pressure transducers both
upstream and downstream of the
strainer, and was moved further
upstream, from its previous
location of just three pipe-
diameters away from the pump
inlet, to 10 pipe diameters. This
ensured any turbulence from the
strainers is given time to dissipate
before the liquid enters the pump.
The effectiveness of these system
corrections was so benecial that
the customer decided that what
began as his stretched system,
would be maintained as the new
larger batch process system. Also
after these corrections, the pumps
reverted to their previous exalted
position as the most reliable
pumps in the plant.
Pumps are very reliable, and long
lived machines, and to ensure this
continues, it is essential to know
What Turns a Good Pump Bad
and these series of articles will
help in this process.

You might also like