You are on page 1of 6

Identifying poor heat exchanger

performance
H
eat exchangers are designed with engineer-
ing tools that allow the designer to evaluate
the infuence of various parameters on
projected performance and select the optimum
combination. However, computer model results
are just calculations and only refect actual
performance if the model assumptions correctly
predict what actually happens inside the
exchanger. Consequently, when an exchanger
does not perform per design, the model assump-
tions are not correct. Heat exchanger performance,
just like that of any other refnery process equip-
ment, depends on specifc equipment design and
not on the ideals of the computer model.
Reboiler system
In 2000, National Petroleum Refners of South
Africa (PTY) Ltd (Natref, a joint venture of Sasol
and Total South Africa) commissioned a major
crude/vacuum unit expansion. As part of the
revamp, a new reboiler was installed on the
crude debutaniser to provide increased duty to
stabilise the increased production from the
prefash and atmospheric crude columns. Total
reboiler heat input needed to increase to remove
enough of the C
4
s to meet the light straight-run
(LSR) Rvp specifcation for gasoline pool blend-
ing. However, the debutaniser reboiler heat input
was not suffcient to meet the light naphtha Rvp
specifcations after startup, particularly when
processing light crude blends. The debutaniser’s
maximum throughput was limited to 90% of
design when the reboilers were clean. While in
operation, fouling resulted in a 3.5% reduction
in reboiler duty per month (equivalent to a
reduced maximum throughput of about 750 bpd
Mark Fernsby and Abe DuPont Natref
Tony Barletta and Steve White Process Consulting Services Inc
every month). At times, this began to limit the
overall crude unit throughput.
The revamped debutaniser column was
designed to operate with two reboilers. One used
gas oil product. Its heat input was therefore set
by product yield. This reboiler, which had not
been mechanically altered during the revamp,
was operating satisfactorily. The new steam
reboiler was designed to supply the balance of
the debutaniser heat requirements during normal
operation and the total column heat require-
ments, so the gas oil reboiler could be taken
offine for maintenance. This larger reboiler was
designed for 42 kg/cm
2
pressure steam (Figure 1)
to replace one using 16 kg/cm
2
pressure steam.
After commissioning, the exchanger had a calcu-
lated service overall heat-transfer coeffcient (U)
as low as 25% of design. Various feld tests were
www.digitalrefning.com/article/1000114 Revamps 2006 1
Identifying root causes of poor crude unit debutaniser reboiler performance
and low-cost solutions that help meet exchanger original design conditions
Unstabilized
naphtha
Stabilized
naphtha
Gas oil
product
42 kg/cm
2
steam
Figure 1 Crude debutaniser reboiler system
done to evaluate potential root causes. The
exchanger was eventually taken out of service
and inspected to ensure the actual design
conformed to the drawings and assess whether
the shell or tube sides were badly fouled.
However, only a marginal performance improve-
ment was noted when the exchanger was put
back into service. Even when the reboiler was
clean, its U value was less than 25% of design
clean performance. Although the reboiler was
never signifcantly fouled, its performance was
sensitive even to small amounts of fouling
compared to the other reboiler.
It is common to speculate on potential causes
when equipment does not meet expected
performance. In this case, theories included
inherently poor heat transfer when using U-
tubes instead of straight tubes, insuffcient
shell-side fuid circulation resulting in excessive
vapourisation and two-phase fow irregularities
in the reboiler return piping. Another was that
the exchanger simply was not big enough and a
new larger exchanger was needed. Many tests
were conducted to investigate these and various
other hypotheses that could explain the poor
performance.
Determination of root causes must begin with
a review of the specifc system and equipment
design, and the application of fundamental engi-
neering principles. As engineering tools become
increasingly complex, sophisticated and easier to
use, such as computational fuid dynamics (CFD)
models, engineers have a tendency to rely on
these models and search for complex causes and
solutions. This should not be a surprise given the
increasing focus on engineering tools rather than
on understanding fundamental equipment oper-
ating principles. Yet, understanding these
principles remains the single most important
factor in the design of process equipment and
troubleshooting when the equipment does not
perform as expected.
Reboiler basics
Potential reboiler problems can generally be
separated into three areas: steam-side, process-
side and exchanger design. Generally, there are
only a couple of steam-side problems, the most
common of which is fooding the exchanger with
condensate, which reduces the condensing
surface area. Potential process-side problems
include a high system pressure drop, excessively
2 Revamps 2006 www.digitalrefning.com/article/1000114
low and high circulation rates, and slug fow in
the reboiler return piping. Thermosiphon reboil-
ers circulate tower bottoms fuid through the
shell side of the exchanger based on system
hydraulics. The liquid level in the bottom of the
column and the globe valve at the inlet of the
shell side (if present) is used to control circula-
tion. Too low a circulation rate increases the
percentage of vapourisation, which reduces the
heat-transfer coeffcient, and too high circulation
reduces vapourisation, which can lead to slug
fow in the return piping. As for potential
exchanger design errors, many have caused low
duty.
Exchanger design problems can reduce the
service heat-transfer coeffcient by affecting
inside and outside flm coeffcients. Thermal
conductivity of the tube is small compared to the
flm coeffcients. Therefore, this term can be
ignored. Thus, the “clean” heat-transfer coeff-
cient (U
C
) is calculated from Equation 1:
1
=
1
+
1
U
c
h
i
h
o
Eq 1
where:
U
C
= heat-transfer coeffcient (clean)
h
i
= inside flm coeffcient
h
o
= outside flm coeffcient
Computer models calculate flm coeffcients
based on the specifc exchanger design, fuid prop-
erties, model assumptions and equations. For
example, heat exchanger models assume the shell-
side fuid fow is distributed so that the total
exchanger surface area is utilised for heat transfer.
In reality, the specifc exchanger design will deter-
mine whether the fuid entering the exchanger is
uniformly distributed or not. In the case under
consideration, the new steam reboiler design U
C

was 1355 kcal/h m
2
°C, but actually achieved only
300 kcal/h m
2
°C.
Clean exchanger coeffcients do not include
fouling resistance. The exchanger service heat-
transfer coeffcient includes fouling resistance on
the inside and outside of the tubes. Total resist-
ance to heat transfer is the sum of the shell-side
flm coeffcient, tube-side flm coeffcient and the
fouling resistance. For a crude unit debutaniser
steam reboiler, fouling resistance is generally 25–
35% of the total resistance, with shell-side fouling
more common. It is rare to have any signifcant
fouling inside the tube when using steam.
2 Revamps 2006 www.digitalrefning.com/article/1000114
In practice, inside and outside fouling resist-
ances are lumped together in an overall fouling
resistance commonly referred to as a fouling
factor. The dirty (or service) heat-transfer coeff-
cient can be calculated as shown in Eq 2:
1
=
1
+ R
F

U
D
U
C
Eq 2
where:
U
D
= heat-transfer coeffcient, dirty
U
C
= heat-transfer coeffcient, clean
R
F
= overall fouling factor
Fouling begins as soon as an exchanger is put
into service. Crude unit debutanisers can foul
badly on the shell side from corrosion products
and water in the feed from the crude unit over-
head receiver. At one point in the case study, the
reboiler U
D
was only 110 kcal/h m
2
°C.
Reboiler design
The new steam reboiler was a TEMA H-shell
type designed for a low pressure drop with no
vertical baffes (Figure 2). H-shell exchangers
have two inlet and two outlet nozzles, with a
horizontal baffe separating the inlet and outlet
nozzles. A partition baffe in the middle of the
exchanger essentially splits the shell side into
two separate sections, each taking fow from its
own inlet nozzle.
H-shells can be designed with or without verti-
cal baffes. The maximum allowable exchanger
pressure drop depends on the overall system
design. When the column liquid level above the
centreline of the exchanger is low, the maximum
allowable pressure must be low, as the available
head is limited. However, the maximum pressure
drop is sometimes intentionally specifed very
low because of conservatism or design guide-
lines. The consequences are a low heat-transfer
coeffcient, large surface area and increased like-
lihood of poor shell-side fow distribution. The
H-shell exchanger pressure drop must be
balanced against the resultant heat-transfer
coeffcient.
The reboiler maximum allowable pressure
drop depends on the overall system design.
Process-side hydraulics must be carefully evalu-
ated at the design stage so the exchanger design
is not compromised. Shell-side fuid circulation
depends on the available liquid level, exchanger
pressure drop, density difference between the
www.digitalrefning.com/article/1000114 Revamps 2006 3
liquid in and mixed phase outlet, and the piping
system pressure drop. Often, revamps are
constrained by the existing vessel skirt height,
liquid level and reboiler return nozzle location
(Figure 3). Ideally, shell-side circulation rates
should result in reboiler outlet conditions with
25–35% vapour in the mixed phase.
Balancing the exchanger pressure drop and
heat-transfer coeffcient is critical. Calculating
the shell-side pressure drop accurately requires
a model that predicts the phase change as the
shell-side fuid fows upward through the reboiler
and vapourises. Since this calculation is subject
to a number of inputs and uncertainty inherent
in the correlations, it is not unusual for the proc-
ess engineer to specify a conservatively low
maximum allowable pressure drop. Moreover,
the exchanger design engineer will also provide
some margin. Often, not surprisingly, the result
Outlet
Steam
in
Condensate
out
Intlet
Figure 2 H-Shell exchanger
Steam
h
1
Condensate
Figure 3 Reboiler liquid head
is an exchanger that is designed larger than
necessary without baffes. All this increases the
likelihood of shell-side fow irregularities and
poor performance. H-shell reboilers or condens-
ers designed without baffes are prone to
lower-than-expected heat-transfer coeffcients.
Following the initial startup of the unit, the
maximum observed duty was much lower than
the design. The steam fow-control valve was
nearly wide open, pushing the condensing
temperature to maximum. The original exchanger
data sheet clean heat-transfer coeffcient (U
C
)
was calculated at 1355 kcal/h m
2
°C, with a
design service heat-transfer coeffcient U
D
of 627
kcal/h m
2
°C. The highest average monthly heat-
transfer coeffcient, U
D
, achieved was only 391
kcal/h m
2
°C, with the exchanger operating at or
below 250 kcal/h m
2
°C much of the time.
Identifying the problem
When something does not perform according to
design, the most likely causes should be investi-
gated frst. Eq 1 shows that exchanger
performance depends on both shell- and tube-
side heat-transfer coeffcients. Since steam
4 Revamps 2006 www.digitalrefning.com/article/1000114
FC
TC
LC
LC
PC
42 kg/cm
2
steam
BFW
HP
condensate
LP
steam
Figure 4 Reboiler control system
reboilers generally have a high
inside heat-transfer coeffcient —
except when there is a large
amount of superheat or conden-
sate foods the tubes, effectively
reducing surface area — tube-side
problems are not the cause of
poor performance. Moreover, the
Natref system was designed with
a steam desuperheater, and feld
testing confrmed the desuper-
heater was performing per design.
(Figure 4).
Tube-side operation depends on
the specifc control system and
condensate drain system. In this
example, steam fow was control-
led to the reboiler based on
required duty. As duty goes up,
the steam rate increases, with the
fow-control valve opening to
allow more fow. As the valve
opens, the pressure downstream
of the control valve increases.
Once the valve is fully open and
the downstream pressure is at
maximum, no more heat can be
added. As pressure downstream of the fow-
control valve increases, the condensing
temperature increases, raising the exchanger
LMTD. The tube-side condensing pressure can
change from a minimum of approximately 15
kg/cm
2
condensate header pressure to a maxi-
mum of 39 kg/cm
2
steam pressure. Pressure
downstream of the fow-control valve is a good
indicator of exchanger performance, because the
higher the pressure needed for a given duty, the
lower the heat-transfer coeffcient. Even with the
fow-control valve wide open, the steam fow rate
was low and so was the exchanger duty.
Condensate fooding is the other common
problem. Condensate fows from the exchanger
into the condensate drum. As long as the
condensate level is in the external drum, the
exchanger is not fooded with condensate. But
once the external drum is full, the condensate
level is inside the exchanger, reducing the surface
area. Field tests showed the condensate level was
in the external drum and not fooding the
exchanger. This was also confrmed through
neutron baskscatter tests that were conducted on
the reboiler.
The exchanger bundle was removed from serv-
ice and visually inspected to determine if fouling
was the problem. Visual inspection and poor
performance after cleaning showed that fouling
was not the root cause of poor performance.
Exchanger design
Poor shell-side performance was therefore postu-
lated. The exchanger was a six-metre long U-tube
design with two horizontal baffes and no verti-
cal baffes. The exchanger design was evaluated
using Heat Transfer Research Institute’s (HTRI)
proprietary IST model. Using reasonable fouling
factors, the calculated performance was much
better than actual duty.
Review of the exchanger design showed
mechanical features that caused poor shell-side
fuid fow distribution through the bundle. Poor
fow patterns can cause shell-side fuid to bypass
portions of tube surface. The inlet nozzle had an
impingement plate, which restricted fow into
the tube bundle. Furthermore, Figures 5 and 6
show the exchanger was designed with skid bars
running its complete length. The reduced inlet
area and skid bar design were causing a signif-
cant portion of the feed to actually bypass much
of the exchanger surface area (Figure 6). The low
calculated coeffcient was a result of a signifcant
bypass exacerbated by a low bundle pressure
drop design (no vertical baffes).
www.digitalrefning.com/article/1000114 Revamps 2006 5 4 Revamps 2006 www.digitalrefning.com/article/1000114
Skid bars
Impingement
baffle
Plan view
(top)
Skid bars
on bottom
Inlet nozzle
on bottom
Figure 5 Exchanger bundle
Figure 6 Original bundle
Plan view
(top)
Full diameter
baffle
Full diameter
baffle
Double
segmental
baffle
Double
segmental
baffle
Figure 7 Bundle modifcation
Figure 8 New bundle
Solution
Many solutions were considered including a new
larger exchanger shell and bundle, and new
reboiler return piping, both of which were expen-
sive options. Furthermore, these options did not
address the root cause and may not have worked.
Instead, a new exchanger bundle was designed
and installed in the same shell.
A new bundle was designed with the intent to
improve the fow distribution to fully utilise the
tube surface area (Figures 7 and 8) and maxim-
ise the tube outside coeffcient h
o
. Vertical
double segmental baffes were installed to
improve the fow distribution. The baffe design
was optimised to ensure shell-side fuid was
forced through the entire bundle. Thermosiphon
hydraulics were checked to ensure the higher
pressure drop would not reduce circulation and
cause problems with high percentage
vapourisation. The skid bars were redesigned
and impingement plates elimi-
nated. In addition, full-diameter
vertical baffes were installed
between the inlet nozzles to help
with fow distribution. After the
new bundle was installed, the
service heat-transfer coeffcient
improved from 250 kcal/h m
2
°C
or less to 950 kcal/h m
2
°C (Figure
9). The debutaniser was now able
to process light crudes while
meeting LSR product RVP
specifcation.
The authors would like to acknowledge the
support of the Natref personnel who were
involved in this project, particularly the CDU
operations division.
Abe du Pont is the manager for process engineering at Natref. Du
Pont obtained a BSc Eng (Chem) at the University of Pretoria in 1984.
Mark Fernsby is a process engineer at the Natref oil refnery in
Sasolburg, South Africa. Fernsby has BSc and MSc degrees
in chemical engineering from the University of Cape Town.
Tony Barletta is a chemical engineer with Process Consulting
Services in Houston, Texas, USA.
Email: tbarletta@revamps.com
Steve White is a chemical engineer with Process Consulting
Services in Houston, Texas, USA. Email: swhite@revamps.com
6 Revamps 2006 www.digitalrefning.com/article/1000114
10
5
15
20
25
30
0 0
200
400
600
800
1000
1200
1400
6
0
0
2

g
u
A

8
2
6
0
0
2

g
u
A

1
3
p
e
S

2
6
0
0
2

p
e
S

4
6
0
0
2

p
e
S

6
6
0
0
2

p
e
S

9
6
0
0
2

p
e
S

1
1
6
0
0
2

p
e
S

3
1
6
0
0
2

p
e
S

5
1
6
0
0
2

p
e
S

8
1
6
0
0
2

p
e
S

1
2
6
0
0
2

p
e
S

3
2
6
0
0
2

p
e
S

6
2
6
0
0
2

r
h
/
J
G

,
y
t
u
D
,
U

t
n
e
c
i
f
f
e
o
c

r
e
f
s
n
a
r
t

t
a
e
H
m

.
h
/
|
a
c
K
2
.
o
C
)NSTALLATION
OFNEW
TUBEBUNDLE
GJ/hr C11068 duty
kca|/h m
2
CU
Figure 9 Improvement in service heat-transfer coeffcient after new bundle
installation

Links
More articles from: Process Consulting services
More articles from the following categories:
Revamps, shutdowns and Turnarounds
Crude Vacuum Units
Process Modelling & simulation