Proceedings of the ASME 2017 Pressure Vessels and Piping Conference

PVP2017
July 16-20, 2017, Waikoloa, Hawaii, USA

PVP2017-65875

A STATISTICAL APPROACH TO EVALUATING SINGLE-SIDE INLET VS DUAL

INLET TEMPERATURE DISTRIBUTIONS IN COKE DRUMS

Jorge PENSO Julian J. BEDOYA

Shell Global Solutions (US) Inc. Stress Engineering Services Inc.
Houston, Texas, USA Houston, Texas, USA

Alex STOLLER Richard S. BOSWELL

Stress Engineering Services Inc. Stress Engineering Services Inc.
Cincinnati, Ohio, USA Houston, Texas, USA

ABSTRACT INTRODUCTION
Slide valves used to unhead coke drums have had a Coke drums are pressure vessels used in refineries to
significant impact on the safety and efficiency of the unheading upgrade the residue produced by taking the incoming feedstock
process in these vessels. Therefore, many refiners have (residuum stream), and further separating (cracking) it into
changed to the inherently different inlet flow nozzle valuable fuel streams (lighter hydrocarbons) that are sent back
configurations that the slide valves have introduced. Single- to the plant for further processing [1]. What remains in the
side entry and dual-side entry have been common alternatives drum after most of the hydrogen molecules crack off eventually
used as a result of the implementation of slide valves. Both of solidifies, becoming petroleum coke (coke). Broadly speaking,
these configurations can depart from a centralized flow pattern coke drum cycles consist of a “switch-in” and filling phase,
and can create adverse flow and temperature distributions. which traditionally has taken about half the cycle, followed by
Furthermore, these changes manifest themselves throughout the a “switch-out”, cooling, cutting and decoking phase which all
vessel with measureable mechanical integrity consequences in account for the remainder of the cycle. Coke drums typically
the cone, skirt, shell, and piping. This paper analyzes historic operate in groups of two and the “switch-in” and “out” portions
measured skin thermocouple data as a function of elevation of of the cycle refer to the delayed coking process switching from
the coke drum. A total of three different refinery sites were one drum to the sister drum. The fill portion of the cycle is the
included in this study; two of them having dual-side inlets, and portion where the economic gain is realized since the lighter
one of them having a single-side entry. A statistical comparison cracked vapor is sent back to fractionation. The coking portion
was performed using the measurements focusing on the data of the cycle typically takes 50% of the total cycle time. The
that causes mechanical integrity problems in coke drums: coke is then subjected to steam stripping and is then cooled
temperature differences around the circumference and using water, “cut” using high pressure water jets, followed by
elevations, peak fill and peak quench rates. physically removing the coke from the pressure vessel, using
unheading technology. While one drum is actively filling, the
NOMENCLATURE other drum is quenching so that the batch process actually
works as a continuous process [1].
SSI: single-side inlet
DI: dual inlet
MCTD: maximum circumferential temperature difference Depending on the type of coke being produced (i.e.,
p: p-value, or statistical probability that an observed difference shot coke, sponge coke, needle coke), different combinations of
between groups could occur when the groups do not differ. p < temperature and pressure and “coking time” are required to
0.05 is generally defined as “statistically significant”. achieve the desired morphology. The required coking time will
typically be the strongest determinant factor on the total cycle
time. When refineries discuss “shortening the cycle” the

1 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 required coking time is typically off-limits in order to maximize bulges form depends on a number of factors, which broadly
economic gain; i.e., it takes a certain amount of time to crack speaking include stages of material degradation from plasticity
the heavy molecules and recover the vapors. In addition the and low-cycle fatigue, frequency of these thermal loads, and
coke is a product itself that is sold, and the type of coke other mechanical, geometrical, and material issues. The main
produced is dependent on the amount of refining that the concern with low cycle fatigue is the impact it has on the
feedstock (crude blend coming into the refinery) will require. circumferential seams, and how these are aggravated by
bulging in adjacent plates as a function of increasing severity
The typical unheading of coke drums prior to the early [6]. In addition, the formation of channels occurs regardless of
2000’s posed some challenges for safe operation, where the inlet configuration, and bulges in coke drums have been
refinery personnel had to physically open the top and bottom1 documented for over half a century [3]. Therefore, the required
headers of the coke drums in order to allow the coke to exit the time for the lack of centralized flow from traditional bottom
vessel, in preparation for the next coking cycle. This manual center feed inlets to manifest itself as an adverse effect took
process inherently represented a risk for exposure to noxious years. Tilting of drum shells from preferential thermal gradients
gases, hot water and vapor, and a substantial amount of coke from one side of the drum to the other – referred to as the
exiting the vessel onto catch basins, where coke is sold for “banana effect” in the industry, became more common,
other uses [2]. In recent times, automatic unheading devices in preferential damage, cracking and deformation to skirts at
addition to reducing the risk of exposing personnel to locations aligned with the inlet configuration were also
dangerous conditions on every cycle, the required time to becoming commonplace. However, the acknowledged random
decoke also went down since human interaction was and inconsistent nature of thermal loading in coke drums, and
minimized. the relatively long coking cycles contributed to these problems
having a slow and concealed evolution which ultimately
Therefore, the coking industry made considerable manifested itself as deformation and cracking several years
technological advances upon the implementation of automatic after the implementation of SSI.
unheading technology by not only increasing safety, but also
being able to reduce idle time (by previous standards) to a The dual feed configuration made its debut after the
minimum. Figure 1 shows a sketch of a traditional bottom feed SSI. The appeal for the DI was to merge two opposing feed or
inlet on a coke drum, a single-side inlet (SSI), and a dual inlet quench flows, and restore the traditional central flow that
(DI) configuration. In addition, the traditional bottom feed existed with the traditional bottom feed. Several refineries have
nozzle was a flanged connection and had been historically been adopted its use, and some owner-operators claim much better
used with longer coking cycles. reliability using this configuration that the SSI, while other
owner-operators have not reported such benefits.

This paper discusses from a statistical perspective, the

thermal distributions around the circumference of the cone and
lower shell of three different coke drums in three refineries.
One coke drum had an SSI configuration, while the other two
had the DI configuration. The variables that were analyzed
included temperature differences around the circumference, the
Figure 1. Illustration of bottom feed (left), SSI (middle), and DI temperature rate of change during the fill (fill rates), and the
(right) inlet configurations. temperature rate of change during the quench (quench rates).

Despite the obvious improvements to coking safety and METHODOLOGY

increased production, shortcomings of having a biased inlet This study originated as a request to analyze existing
flow (both from the hot feed and from the quench water) data to quantify the impact of SSI configuration vs. DI
directed to the opposite wall, in the case of the SSI, would configuration on thermal distributions. Three drums were
eventually appear. However, due to the very nature of coking identified with available data, similar construction, and
operations, as the coke bed solidifies, “coking channels” begin operational conditions. Key details for each drum are provided
to form and non-uniform temperature distributions begin to in Table 1. The data for all coke drums was collected using
manifest themselves [4]. Additionally, thermal distributions thermocouples, loggers, multiplexers, and other components
during coke drum cycles are notoriously unpredictable and used to monitor and control the coking process at each refinery.
variable from cycle to cycle. Over the course of many cycles, Due to the operators’ need for expediency, only existing
bulges eventually form in coke drums. How quickly these thermocouple data was utilized for this analysis; no new
sensors were installed specifically for this purpose. This
1
Only bottom unheading configurations are discussed in this paper.

2 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 resulted in instrumented locations that varied somewhat from processing temperature of approximately 900°F (482°C). The
drum to drum. peak quench rate, peak fill rate, and maximum circumferential
temperature difference (MCTD) were extracted from each coke
For the SSI coke drum, the available temperature data process cycle, at each thermocouple location. The absolute
were from thermocouples located on the drum body and skirt, value of the MCTD was utilized in the analysis. Statistical
within approximately 1 ft (0.3 m) of the drum-to-skirt weld comparisons were then made among the coke drums to
line. These included a total of 8 locations at the NW, NE, SW, determine whether the inlet configuration affected the
and SE quadrants of the drum. For the DI configuration coke measured thermal distributions.
drums, temperature data were available for 4 elevations,
including 2 elevations on the cone and 2 elevations on the drum Table 1. Coke drum details.
body. At each elevation, 2 thermocouples were located on Drum SSI DI 1 DI 2
ASME Design Code Year 1998 1980 1992
opposite sides of the drum, e.g. North and South. The cone Year of Construction 1999 1983 1994
locations were classified as “lower cone” (within 2-6 ft (0.6-1.8 Design Temperature 890°F (477°C) 900°F (482°C) 875°F (468°C)
Design Pressure
m) above the inlet elevation) or “upper cone” (9-12 ft (2.7-3.7 (at top)
50 psi (345 kPa) 50 psi (345 kPa) 50 psi (345 kPa)
m) above the inlet elevation). The drum locations were Base material SA-387-C SA-387-C SA-387-C
classified as “near skirt” (within 2 ft (0.6 m) above the drum- Cladding SA-240 410S SA-240 410S SA-240 410S
Shell ID 28 ft (8.5 m) 26 ft (7.9 m) 26 ft (7.9 m)
to-skirt weld) or “above skirt” (approximately 8-10 ft (2.4-3.0 Height
m) above the drum-to-skirt weld). The approximate (Bottom head to top head 89 ft (27.1 m) 68 ft (20.7 m) 73.5 ft (22.4 m)
instrumented locations are illustrated in Figure 2. Temperature seam)
Base Material Thickness
readings were recorded at 1 minute intervals on all sites. The bottom shell 1.267 in (32.2 mm) 1.079 in (27.4 mm) 1.082 in (27.5 mm)
data available for analysis included approximately 4 discrete cone 1.377 in (35.0 mm) 1.284 in (32.6 mm) 1.220 in (31.0 mm)
Cladding Thickness
weeks of operation for the SSI drum, collected during bottom shell 0.110 in (2.8 mm) 0.110 in (2.8 mm) 0.110 in (2.8 mm)
December 2010 to December 2013. Approximately 8 cone 0.125 in (3.2 mm) 0.134 in (3.4 mm) 0.125 in (3.2 mm)
continuous weeks of operational data were available for DI site Operating Conditions:
Typical total cycle length 34 hrs 32 hrs 42 hrs
1, collected during January – February 2014, and 4 continuous Typical Switch-in Temp 489°F (254°C) 563°F (295°C) 490°F (254°C)
weeks of data were available for DI site 2, collected during
February 2014. While a longer data collection period would
have been ideal, the observed thermal data were consistent 30
Fill Rate (ºF/min)
during the timeframe analyzed (see Figure 3) and the data are 20
believed to represent the random operating conditions for all 10
drums.
Quench Rate (ºF/min)
-10

-20
-30

Max Circumf. Temp. Diff. (ºF)

300
200
100

0 10 20 30 40 50
Cycle
Results over time for DI site 1.

Figure 3. Time history for the DI 1 drum.

RESULTS
Figure 2. Illustration of instrumented location zones. Histograms showing the percent distribution of peak
quench rate, peak fill rate, and MCTD are shown in Figure 4.
Thermal rates (quench and fill) were calculated by The distributions for SSI and DI drums shown in Figure 4 are
differentiating the recorded temperature data with respect to overall averages, grouping all available data at all instrumented
time. Circumferential temperature differences were calculated locations for each inlet configuration (combining both DI
by subtracting temperatures recorded at diametric opposite drums), regardless of location on the drum. When analyzed in
sides of the drum (North vs. South, East vs. West, etc). The this way, the SSI and DI inlet configurations appeared to be
time history was divided into coke process cycles, with the start similar for quench rate and MCTD, however the DI drums
of each new coke process cycle determined based on the showed a tendency for higher fill rate magnitudes. Analysis
observation of a sharp rise in drum inlet temperature to the of variance found a statistically significant difference (p <

3 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 0.001) in the average fill rate for DI drums vs. SSI drums. The skirt showed smaller thermal rates and MCTD compared to the
difference in quench rate was small but statistically significant locations on the drum body near the skirt, as would be
(p = 0.033). MCTD did not differ significantly between the expected. For the DI drums, fill rates and MCTD tended to be
groups (p = 0.43). However, it was recognized that the greater on the cone compared to the drum body, while quench
instrumented locations varied among the drums, and this rate showed no clear pattern. Differences between the 2 DI
variable must be included in the analysis. drums were also apparent, depending on elevation. Therefore,
a sub-analysis was conducted using only the locations common
to both SSI and DI configuration drums; namely, locations on
the drum body near the drum-to-skirt weld2.

Figure 5. Comparison of thermal rates and MCTD at different elevations.

Histograms of the peak quench rates, peak fill rates,

and MCTD near the skirt weld are shown in Figure 6.
Figure 4. Histograms illustrating thermal rates and differences for SSI and
DI configuration drums, grouping all instrumented locations per inlet
2
configuration. Note that the frequency is expressed in percent. See legend It is acknowledged that the instrumentation configurations for all three
for DI and SSI bars. Normal Distribution Fits are shown for each inlet sites were unfortunately not the same. However, the variables of interest; peak
type. rates during fill, quench and MCTD are still of interest. The more severe these
variables (higher magnitudes), the more aggressive the damage accumulation at
Therefore, the results were further analyzed to these locations will be. This premise has been verified by independent health
monitoring studies, and independent finite element analyses. Therefore, an
examine the effects of elevation on the measured thermal overall comparison while not as desirable and elegant as a deliberate research
distribution. A summary of mean thermal rates and MCTD effort would be in having identical instrumentation, is still useful in discerning
based on inlet configuration and elevation is shown in Figure 5. which sites are likely to have highest damage rates. Ultimately, the refineries
Location on the drum affected many of the temperature want to prevent cracks, and while locations of cracking matters, cracks still
represent risk and the potential for reduced production. Therefore, the refineries
distribution measurements. For the SSI drum, locations on the are still interested in reducing cracks, regardless of their location.

4 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Cumulative distribution plots were also generated for thermal
rates and MCTD, as shown in Figure 7. When limited to drum
locations near the skirt weld, some notable differences were
observed between the inlet configurations. Over 60% of coke
process cycles had a peak quench rate in excess of 30°F/min
(16.7°C/min) for the SSI site, vs. approximately 5% for DI site
1 and 33% for DI site 2. Nearly 50% of coke process cycles
included a circumferential temperature difference in excess of
300°F (166.7°C) for the SSI site, vs. less than 5% of cycles at
either of the DI sites. Fill rates at the SSI site were greater than
those recorded at DI site 1, but less than those at DI site 2.

The observed differences in thermal rates and MCTD were

statistically significant (p<0.001) based on analysis of variance,
however it should be noted that normality tests did not meet
goodness-of-fit criteria. Therefore a non-parametric
comparison of the median values (Mood’s median test) was
also performed. The non-parametric evaluation indicated that
the SSI drum had greater median quench rate compared to
either of the DI drums (p<0.001), and that the 2 DI sites did not
significantly differ (p=0.69). Likewise, the SSI drum had
greater median MCTD compared to the DI drums (p<0.001),
which did not significantly differ (p=0.14). With regards to fill
rate, significant differences were found for all 3 sites
(p<0.001), with fill rates at the SSI drum falling between those
at DI sites 1 and 2. All statistical analyses were performed
using Minitab v.17.3. Summaries of the thermal rates and
MCTD are provided in Table 2.

Figure 6. Histograms illustrating thermal rates and MCTD near the skirt
weld for SSI and DI configuration drums.

5 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 2. Summary statistics for all drums and elevations.

Peak Quench Rate (ºF/min)

Site Elevation
Mean St.Dev. Median Min Max
drum, near skirt -33.2 10.8 -33.3 -55.2 -4.4
SSI
on skirt -9.3 1.3 -9.3 -12.4 -5.3
cone, lower -26.2 13.3 -23.8 -64.6 -5.0
cone, upper -17.8 12.9 -13.3 -61.1 -4.3
DI 1
drum, near skirt -15.5 5.1 -14.8 -33.9 -6.0
drum, upper -21.1 12.7 -17.0 -68.6 -6.1
cone, lower -24.2 8.6 -23.5 -45.2 -10.3
cone, upper -36.9 20.9 -30.6 -87.8 -10.7
DI 2
drum, near skirt -20.9 13.9 -15.8 -45.0 -3.9
drum, upper -29.1 12.9 -26.1 -53.5 -10.8

Peak Fill Rate (ºF/min)

Site Elevation
Mean St.Dev. Median Min Max
drum, near skirt 12.8 4.6 12.3 6.3 23.4
SSI
on skirt 6.1 1.1 6.0 3.9 8.4
cone, lower 39.5 7.0 37.5 29.8 62.9
cone, upper 29.1 6.4 28.8 7.9 42.1
DI 1
drum, near skirt 10.3 2.9 9.6 7.0 28.2
drum, upper 9.1 2.0 8.6 6.4 16.5
cone, lower 24.0 3.9 23.7 17.6 32.1
cone, upper 21.9 3.7 22.4 15.8 28.8
DI 2
drum, near skirt 19.5 3.5 18.9 15.0 27.0
drum, upper 11.9 2.7 11.6 7.7 17.2

Max Circumf. Temp. Diff. (ºF)

Site Elevation
Mean St.Dev. Median Min Max
drum, near skirt 287 135 292 28 496
SSI
on skirt 157 81 162 30 379
cone, lower 281 101 290 98 511
cone, upper 294 102 270 131 539
DI 1
drum, near skirt 121 64 109 40 324
drum, upper 187 80 170 45 383
cone, lower 188 114 156 60 461
cone, upper 392 132 434 50 560
DI 2
drum, near skirt 147 56 140 80 262
drum, upper 251 99 245 58 394

Figure 7. Cumulative distribution plots for thermal rates and

MCTD at the drum, near skirt elevation.
DISCUSSION
When considering all data in this study regardless of
location on the drum, the drum inlet configuration may appear
to have minimal net effect on the thermal rates and temperature
distributions with the exception of increased peak fill rate for
DI drums. However, the thermal parameters were also found to
be dependent on location or elevation on the cone and drum.
When comparing only the common location among the drums
(near the drum-to-skirt weld), the DI drums in this study were
found to have reduced peak quench rates and smaller MCTD
compared to the SSI configuration drum. The smaller thermal
rates and more uniform circumferential temperature distribution
would be expected to result in less severe cyclic thermal
stresses on the coke drum and prolonged thermo-mechanical
fatigue life, although strain data was not available for these
drums to confirm. It is not known whether the observed

6 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 differences in quench rates and MCTD would exist for been identified to be executed as a path forward, and phase 2 of
locations on the cone or at higher elevations on the drum body. this project.
Temperature readings for the DI drums in this study suggest
that thermal rates and gradients are more severe at higher ACKNOWLEDGMENTS
elevations, and variable results were found on the cone. The authors wish to recognize the contributions of Messrs.
Lucibar Davalillo, and Tom Meek of Shell Global Solutions
It is acknowledged that differences among the drums and (US) Inc. and the Stress Engineering Services Inc. staff who
process parameters may confound the results and make have contributed to this work.
comparisons difficult. Ideally this study would have involved
identical drum design and process controls along with REFERENCES
consistent instrumentation sites. Nonetheless, the comparative [1] Leffler, W.L., Petroleum Refining in Nontechnical
results were consistent at the near-skirt locations when other Language, 4th Edition, 2008, PennWell Corporation,
variables were relatively constant. Quench rates were higher Tulsa, OK.
for the SSI drum compared to either of the DI drums; this may [2] Lah, R, Curtiss-Wright Flow Control Corporation,
be due in part to longer cycle times at DI site 2, but the SSI “Coke Drum Bottom De-Heading System”, U.S.
drum cycle time was similar and slightly longer compared to DI Patent 6565714, Issued May 20, 2003.
site 1. Circumferential temperature differences were greater for [3] N.A. Weil and F. S. Rapasky, “Experience with
the SSI drum compared to either of the DI drums. Fill rates Vessels of Delayed-Coking Units”, Proceedings
were lower in the SSI drum vs. DI site 2 despite similar switch- American Petroleum Institute, 1958, pp. 214-232, vol
in temperature, however this may be due to the larger size of 38 [III], American Petroleum Institute
the SSI drum. Therefore, when other variables are relatively [4] Ellis, P.J., Hardin, E.E., “How Petroleum Delayed
constant, the results of this study suggest that the dual inlet Coke Forms in a Drum”, Printed in Light Metals 1993,
configuration may have advantages over the single-side inlet The Minerals, Metals and Materials Society, 1992.
with regards to thermal rates and temperature gradients near the [5] Levy, S., Evans, B., Boswell, R.S., Wright, B.W.,
drum-to-skirt weld. “Coke Drum Monitoring and Process Optimization”,
presented at Coking.com in Galveston, 2004.
However, ultimately, the most important measure of [6] Boswell, R.S., Wright, B., “State-of-the-Art
reliability is a concerted and intentional effort to characterize Improvements in Coke Drum Design and Life
features such as cracks that ultimately lead to lost production Extension Practices”, Proceedings of CREEP8, 2007,
activities. In addition, the use of strain gauges to characterize ASME, New York, N.Y
the unit damage per cycle would have also been useful in this
assessment. As an example, while the quench rates and MCTD
in the SSI site were higher than both DI sites, a path forward is
to perform phase 2 studies analyzing the actual strain
measurement response to coke drum operations.

By the authors’ significant experience in analyzing large

amounts of health monitoring data from many refineries (not
just the ones in this paper) using thermocouples and strain
gauges, the less frequent but highly damaging events can skew
the results of different statistical studies. Therefore, having the
damage information directly, and complementing the readily
available thermal data, could allow refineries to predict in
measureable monetary terms some notion of increased
reliability, production and production margins [5]. Moreover,
the use of cumulative probability density curves and histograms
further allow analysts to have some notion of forecasting
potential needed for inspections or repairs. Furthermore,
strategically coupling this data with diligent maintenance
practices and records, it is conjectured based on basic
engineering and physics principles that a measurably and
statistical significant difference in thermal gradients, rates and
strain measurements, can lead to better safety, planning and
more reliable operations. These stated concepts above have

7 Copyright © 2017 ASME and Shell Oil Company

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/ on 10/28/2017 Terms of Use: http://www.asme.org/about-asme/terms-of-use