Transactions Journal of Energy

of the ASME Resources Technology ®

Technical Editor Published Quarterly by The American Society of Mechanical Engineers

J. P. BRILL
Past Technical Editor
A. Ö. ARNAS VOLUME 122 • NUMBER 1 • MARCH 2000
Associate Technical Editors
M. ADEWUMI „2000… TECHNICAL PAPERS
F. C. CHEN „2000…
M. HACI „2001…
A. JACOBI „2002… 1 Hydraulic Hammer Drilling Technology: Developments and Capabilities
W. P. JEPSON „2000… Yuri Melamed, Andrei Kiselev, Michael Gelfgat, Don Dreesen, and
N. A. MOUSSA „2000… James Blacic
D. O’NEAL „2002…
C. SARICA „2000… 8 Evaluation of Inflow Performance of Multiple Horizontal Wells in Closed
S. SHIRAZI „2002… Systems
J. SMITH „2002… Suwan Umnuayponwiwat and Erdal Ozkan
K. S. SURANA „2002…
A. VALERO „2000…
14 Slug Dynamics in Gas-Liquid Pipe Flow
W. J. WEPFER „2000… Hong-Quan Zhang, Subash S. Jayawardena, Clifford L. Redus, and
A. K. WOJTANOWICZ „2000… James P. Brill
22 Investigation of Transition From Annular to Intermittent Flow in Pipes
BOARD ON COMMUNICATIONS Avni Serdar Kaya, X. Tom Chen, Cem Sarica, and James P. Brill
Chairman and Vice-President
R. K. SHAH 29 Sand Production Management
Mamdouh M. Salama
OFFICERS OF THE ASME
34 Pressure Transients Across Constrictions
President, R. E. NICKELL
Michael A. Adewumi, E. S. Eltohami, and W. H. Ahmed
Executive Director
D. L. BELDEN 42 The Nonoriented, Multiple Passageway Flowline Connection System
Treasurer, J. A. MASON Gary Galle and Rudolf Bond

PUBLISHING STAFF
Managing Director, Engineering
CHARLES W. BEARDSLEY
Director, Technical Publishing
PHILIP DI VIETRO
Managing Editor, Technical Publishing
CYNTHIA B. CLARK
Managing Editor, Transactions
CORNELIA MONAHAN
Production Coordinator
LYNN ROSENFELD
Production Assistant
MARISOL ANDINO

Transactions of the ASME, Journal of Energy Resources

Technology (ISSN 0195-0738) is published quarterly
(Mar., June, Sept., Dec.) for $195 per year by The American
Society of Mechanical Engineers, Three Park Avenue, New York,
NY 10016-5990. Periodicals postage paid at New York, NY
and additional mailing offices.
POSTMASTER: Send address changes to Transactions of the
ASME, Journal of Energy Resources Technology, c/o
THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS,
22 Law Drive, Box 2300, Fairfield, NJ 07007-2300.
CHANGES OF ADDRESS must be received at Society
headquarters seven weeks before they are to be effective.
Please send old label and new address.
PRICES: To members, $40.00, annually: to nonmembers,
$195.00. Add $40.00 for postage to countries outside
the United States and Canada.
STATEMENT from By-Laws. The Society shall not be
responsible for statements or opinions advanced in papers
or ... printed in its publications (B7.1, Para. 3).
COPYRIGHT © 2000 by the American Society of Mechanical
Engineers. For authorization to photocopy material for
internal or personal use under those circumstances not falling
within the fair use provisions of the Copyright Act, contact
the Copyright Clearance Center (CCC), 222 Rosewood Drive,
Danvers, MA 01923, tel: 978-750-8400, www.copyright.com.
Request for special permission or bulk copying should
be addressed to Reprints/Permission Department.
INDEXED by Applied Mechanics Reviews and Engineering
Information, Inc. Canadian Goods & Services Tax
Registration #126148048
 Yuri Melamed
Andrei Kiselev
SKB ‘‘Geotechnika,’’
2ya Roschinskaya str. 10,
Moscow 113191, Russia
Michael Gelfgat
Aquatic Company,
Letnikovskaya str. 7–9,
Moscow 113114, Russia
Don Dreesen
James Blacic
Los Alamos National Laboratory,
GeoEngineering Group, MS D443,
Los Alamos, NM 87545
Hydraulic Hammer Drilling
Technology: Developments
and Capabilities
Percussion drilling technology was considered many years ago as one of the best ap-
proaches for hard rock drilling. Unfortunately, the efficiency of most hydraulic hammer
(HH) designs was very low (8 percent maximum), so they were successfully used in
shallow boreholes only. Thirty years of research and field drilling experience with HH
application in former Soviet Union (FSU) countries led to the development of a new
generation of HH designs with a proven efficiency of 40 percent. That advance achieved
good operational results in hard rock at depths up to 2000 m and more. The most recent
research has shown that there are opportunities to increase HH efficiency up to 70
percent. This paper presents HH basic design principles and operational features. The
advantages of HH technology for coiled-tubing drilling is shown on the basis of test
results recently conducted in the USA. 关S0195-0738共00兲00101-1兴

R&D and Field Application Background
General. The application of percussion drilling methods to
hard rock results in the following advantages as compared to ro-
tary drilling: 1兲 the impact loads at the bit inserts in percussion
drilling are much higher than the load levels typically achieved in
rotary drilling; and 2兲 the time of total contact of inserts with the
rock is substantially less than during rotary drilling. Contact time
in percussion drilling is typically 2 percent of the total operational
time. This provides high efficiency rock destruction and decreases
the abrasive wear of the drilling tool.
The major feature of the percussion drilling is creation of a
crushed zone directly beneath the area of impact. Fractures are
initiated which allow shearing processes to remove the cuttings
easily and increase the rate of penetration. The most productive
method of rock destruction in this respect is percussion-rotary.
This method optimizes the amount of impact load in relation to
standard rotary drilling compressive and shear loads.
At present, some institutions and companies involved in the
drilling business are vigorously considering hydraulic hammers
for a variety of purposes, such as: coiled-tubing drilling; explor-
atory drilling for oil and gas, including extended reach boreholes;
geothermal drilling; exploratory drilling for hard minerals; and
offshore scientific and geotechnical drilling, coring and sampling
of soft, unconsolidated soils and formations at sea and lake bot-
toms 关1–4兴.
FSU Experience. The start of hydro-percussion drilling de-
velopment in the USSR dates from the late 1940s. The main ob-
jective was to increase rates of penetration and drill bit perfor-
mance both in geological prospecting, or ‘‘mining,’’ and slimhole
oil field drilling in hard formations. R&D work, including inves-
tigation of hydraulic machine operational processes and introduc-
tion of percussion drilling to the industry, were implemented in
several scientific research institutes and mechanical design bu-
reaus. This work has resulted in a great improvement in the per-
formance of percussion drilling systems 关5兴.
The All-Union Scientific Research Institute of Drilling Tech-
niques 共VNIIBT兲 made the major contributions in theoretical and
experimental studies of hydraulic percussion tools for oil field
application. Several designs of the reverse action-type hydraulic
hammers, VVO-5A with 130-mm outside diameter and VVO
6-5/8 with 168-mm diameter, were developed and field tested.
Field testing of these hammers started in 1960 in Bashkiria, West
Ukraine, and the Belgorod region. In 1963, testing started in the
Perm region as well. During the tests, more than 10,000 m of hard
formations were drilled as deep as 1400 m. The rate of penetration
in medium hard rock, like limestone, sandstone with siliceous
interlayers, was in the range of 4–10 m/h. That rate was two to
three times more than rotary drilling results in the same condi-
tions. During the field and bench tests, the application of
percussion-rotary drilling in oil and gas wells using different types
of bits 共cone, drag, and combined兲, was studied 关6兴.
The Special Design Bureau 共SKB兲 ‘‘Geotechnika’’ commenced
hydraulic hammer development in 1957, and at present is the only
enterprise in Russia continuing R & D work in that area of drilling
technology. The hydraulic hammers of direct action, double ac-
tion, diffuser types, and hydro-vibrators of different types, includ-
ing ones without moving parts, have been developed. More than
70 HH prototypes have been fabricated and tested both in the
laboratory and in boreholes. These include tools with outside di-
ameters from 42 to 145 mm. Twenty types went into batch pro-
duction. During that time, the theory, bench test facilities, and
measuring systems were improving continuously. Experience in
design, manufacture, and application was gained. The latest de-
signs provided wireline coring techniques, soil investigation, and
a core-type hammer for the continuous, reverse-circulation coring
system 关7,8兴.
The advantages of percussion drilling were confirmed by nu-
merous comparative tests and jobs performed in different geologi-
cal conditions. In 1988, the percussion-rotary and rotary-
percussion methods were used to drill over 3,000,000 m. The
latter amounted to 15 percent of the total drilling for hard mineral
deposit exploration by the USSR Ministry of Geology 关9兴.
Hydraulic percussion hammers were used to explore for all
types of minerals, from coal and iron to mercury, gold, and water,
in boreholes with depths down to 2000 m and diameters from 46
up to 220 mm. Penetration rates, as compared to rotary drilling
共depending on geological conditions兲, were increased from 30 to
100 percent; service life of the drill bit improved by 20 to 200
percent; and deviation of boreholes and their costs were drasti-
cally reduced.

Contributed by the Petroleum Division for publication in the JOURNAL OF EN- Rock Destruction by Impact Loads
ERGY RESOURCES TECHNOLOGY. Manuscript received by the Petroleum Division,
February 2, 1997; revised manuscript received September 27, 1999. Associate Tech- Basic Principles. A downhole hydraulic hammer generates
nical Editor: A. K. Wojtanowicz. an impact load, which is transmitted to the drilling tool 共drill bit,

Journal of Energy Resources Technology Copyright © 2000 by ASME MARCH 2000, Vol. 122 Õ 1
core bit, etc.兲 through an anvil. The hammer and bit 共Fig. 1兲 form
a mechanical system that consists of the jar-peen 共sometimes this
part of the hammer is named ‘‘hammer’’兲 and the intermediate bar
共anvil兲 with a length that is significantly greater than its diameter.
The latter is attached to the drill bit.
Transmission of load from the jar-peen to the drill is analyzed
by using stress-wave theory for metal bars undergoing collision.
This theory applies for flat parallel impact surfaces. In reality,
there are no flat impacts, because of misalignment of jar-peen and
anvil, as well as other manufacturing tolerances.
An applied theory of collision developed by Alexandrov and
Sokolinskyi assumed a bar with spherical ends and is useful in our
case. This theory takes into account the observed impact time
increase, in addition to the time predicted by the classical wave
theory.
It was established that the amplitude of the stress wave created
at the top of a bar decreases along the axis according to an expo-
nential law. The dampening depends on the number of thread
connections between the components and the length of the system.
There is a step change of the stress wave at each change in cross
section of jar-peen and anvil. This wave propagates down to the
bit through the anvil and other members, and then divides into two
waves at the cutter-rock contact point. The first is the main trans-
mitted wave and the second is the reflected wave. Experimental
studies show the reflected wave consists of two parts: tensile wave
and compressive wave. The first is smoothly transformed to the
second. In the case of rigid bottom connection, the stress in the
contact point increases until it is double the magnitude of the
down-coming wave. Hydro-percussion drilling in hard rocks is
relevant to the latter case with double the contact stress, which
was proven both by calculations and experiments 关5,10兴.
Rock Destruction Approaches. Impact energy is the major
parameter determining percussion drilling efficiency. This was
Fig. 1 The hydraulic machine of Bassinger, USA, 1948–1957
†5‡
2 Õ Vol. 122, MARCH 2000
verified by numerous long-term studies of rock destruction by
static and dynamic loads. Impact energy can be increased with
increased jar-peen mass or increased velocity, once a critical im-
pact speed has been exceeded. In practice, the rate of penetration
is directly proportional to the impact energy. Experiments also
show that for complete energy transmission to the rock, the length
of anvil has to be equal to or greater than the length of jar-peen.
Based on theoretical studies and industrial requirements, two
hydro-percussion drilling methods were developed: 1兲 rotary-
percussion with relatively high rotary speed and high-frequency
impacts, and 2兲 percussion-rotary with lower rotary speed and
lower frequency, but higher impact energy 关11兴.
The first method was useful for coring with diamond core bits.
Their relatively low-impact energy was very productive in some
fractured formations. Problems with core recovery caused by jam-
ming were overcome and the rate of penetration was increased.
Efficiency of the diamond percussion drilling with more than
50-Hz impact frequency increased with increasing rotary speed.
The percussion-rotary approach was used to match the drilling
method with solid bits, tungsten carbide crowns 共drag bits兲, or
cone bits. For cone bit drilling, the percussion-rotary method gave
the highest penetration rate, but the bit bearing design had to be
changed. Tungsten carbide crowns were designed for both meth-
ods. It has been demonstrated that core can be effectively frag-
mented by the formation of disks with the application of high-
frequency impact loads 关12兴.
It will be shown in the forthcoming that the Geotechnika hy-
draulic hammers 共GHH兲 develop impact loads and frequencies
sufficient for the two drilling methods.
Hydraulic Hammer Design Concepts and Operation
Conditions
Design Concept and Classification. Hydraulic hammers can
be divided into three groups determined by the method of energy
extraction from the drilling fluid as follows 关10兴:
1 Direct-action hydro-hammers 共DAHH兲 with a hydraulically
powered impact stroke and spring-powered return. Energy is ex-
tracted from the fluid when the jar-peen accelerates down, before
it strikes the anvil. Part of the hydraulic energy is used for impact
and the other part accumulated in the spring to provide jar-peen
return 共cock the peen兲.
2 Reverse-action hydro-hammer 共RAHH兲 with a spring-driven
hammer impact stroke and hydraulically powered return stroke.
Energy is extracted during the jar-peen reverse stroke and accu-
mulated in the spring, which is then applied to the impact itself.
3 Double-action hydraulic hammers 共DBHH兲: impact and re-
turn strokes are both hydraulically powered.
The hydraulic hammer as a self-sufficient and self-sustained
oscillating system can be operated in resonance. That characteris-
tic is usually observed in machines with a spring-loaded valve:
共DAHH兲 and 共RAHH兲. The forces acting at the jar-peen can be
divided into regular and irregular 共stochastic兲 forces. The latter
includes: 1兲 jar-peen rebound force, which depends on the bottom-
hole conditions; 2兲 drag forces; 3兲 forces activated by drill string
vibrations; and 4兲 forces induced by the reflecting hydraulic
waves coming into the working chamber. Reduction of the num-
ber of moving parts reduces the irregularity of jar-peen operation.
Simplification of design provides increased operating stability.
This approach, applied to the DAHH, has been the main trend of
GHH development.
DAHH Operational Concept. A schematic of the DAHH is
shown in Fig. 2 关10兴. The hammer is shown at the moment when
the drill bit is set on the borehole bottom. The housing together
with the valve is moving down and closes the hole in the jar-peen.
That action creates the hydraulic shock, and the pressure inside
the chamber above the valve increases rapidly. The pressure be-
low the valve is: 1兲 the same as that in the annulus, or 2兲 less than
Transactions of the ASME
 with the valve piston, thus closing the valve. This is the ‘‘idle
stroke’’ phase. Drag forces during this phase have to be overcome
as well. The speed of the jar-peen and valve interaction deter-
mines the time of build-up of the hydraulic shock pressure. As the
jar-peen has some inertia and the pressure build-up requires some
time, so the jar-peen and valve continue to move up together until
the forces are balanced. This phase is called the ‘‘floating phase.’’
At this time, the rarefaction occurs and cavitation bubbles possi-
bly form. Then, the cycle repeats.

GHH Operational Features and Parameter Calculations.

The hydraulic shock generates the pressure wave with specific
shape, amplitude, and duration. The wave propagates up the inside
of the drill string with dissipation and reflection at each point
where the cross-sectional area or slope varies 共at each joint, for
example兲, to the mud pumps, valves, and pulsation dampeners.
During percussion drilling in shallow boreholes without a damp-
ener, wave interaction with the mud pump can cause damage to
the pump. A hydraulic wave reflector can eliminate substantial
dissipation of wave energy. The reflector has the added advantage
that the hydraulic energy reflected back toward the drill bit 共and
away from the string and mud pumps兲, may increase the effi-
ciency of the rock destruction produced by the hammer 关10,11兴.
Elastic and hard reflectors were developed, as shown in Figs. 3
and 4, respectively. The use of hard reflectors doubles the ma-

Fig. 2 The direct action hydraulic hammer, general scheme

†10‡

the annulus pressure if a rarefaction is induced by deceleration of

the flow stream with valve closure when the jar-peen has not yet
started to move down. If the absolute pressure is insufficient, cavi-
tation occurs and results in increased differential pressure. The
differential pressure acts against the piston 共top of the jar-peen兲 to
accelerate the jar-peen downward, together with the valve. During
this movement, both valve spring and jar-peen spring are com-
pressed. When the stroke exceeds Xk, the valve movement is
stopped by the top shoulder. This latter event is named ‘‘valve
cut-off.’’ The previous operational phase is named the ‘‘accelera-
tion phase.’’ The jar-peen continues its movement down to strike
against the anvil. That distance is Xb, and the operational phase
name is ‘‘free jar-peen stroke.’’ During this phase the valve
moves to the upper position, as flow balances the pressure on the
valve.
During the accelerating phase, the energy extracted from the
flow is consumed to accelerate the jar-peen, compress the jar-peen
spring, and overcome both mechanical and hydraulic drag forces.
The external force stops acting on the jar-peen after ‘‘valve cut-
off’’ occurs, and the jar-peen continues moving down by inertia.
During this phase, the jar-peen spring is compressed. The phase of
impact starts at the end of the free jar-peen stroke. At this time,
the jar-peen kinetic energy is transferred to the anvil and distrib-
uted as follows: one part propels the jar-peen rebound and the
other drives the drill bit to impact against the rock, with the re-
flected and transmitted waves originating, as explained in the fore-
going.
At the end of the impact, the jar-peen starts moving up by the
forces of the spring and reflecting wave, which defines the re- Fig. 3 The submerged pneumatic elastic reflector PPO-70, GI,
bound. The jar-peen accelerates upward until it makes contact Ukraine †10‡

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 3

Fig. 4 Hydraulic wave hard reflector, SKB Geotechnika, 1984
chine efficiency from 8–10 percent to 16–18 percent, with a two-
fold reduction of flow rate. Similar results were obtained for elas-
tic reflectors.
These designs had definite drawbacks and did not solve the
basic problem of increasing the efficiency of percussion drilling.
SKB ‘‘Geotechnika’’ developed a new design for the reflector,
which was based on long-term studies at special test facilities.
This reflector provides significant reduction of the mud flow re-
quired for rock destruction, and the machine efficiency increased
to 40 percent 关8兴.
Cavitation is the other element that has to be considered when
developing hydraulic hammers. The jar-peen acceleration in-
creases when cavitation occurs, but with increase of the borehole
depth and hydrostatic pressure, the enhancement of the accelera-
tion stroke is diminished, and eventually eliminated. In this case,
the premature valve cut-off results in a short stroke, nonimpact
operating cycle. There are two ways to solve the problem. The
first, is to eliminate the conditions for the cavitation. The second,
is to control the duration of cavitation by adjusting parameters of
the hammer. The present GHH design eliminates cavitation. Some
preliminary experimental data supports the possibility of control-
ling cavitation.
With all the foregoing considerations, and accounting for the
drag forces in the valve motion, performance of GHH designs was
modeled. For each case, a model is developed with a system of
differential equations. The solutions are derived for each of the
operational phases described in the foregoing.
The current GHH designs are tailored to operate in both
percussion-rotary and rotary-percussion modes of drilling. GHH
designs are easily adjusted to operate in resonance, and have
25–40 percent efficiencies in borehole operations. Recent experi-
mental studies have shown the opportunity for a significant in-
crease in power for the GHH, and efficiency should approach 70
percent.
Drill Bits for Hydro-Percussion Drilling. Several types of
rock destruction tools have been developed in conjunction with
percussion drilling. First, there were solid bits and drag-type bits,
with tungsten carbide cutting structures. In these tools, both bits
and crowns are used, mainly for percussion-rotary drilling at rela-
tively shallow depths in medium and medium-hard rock. Impreg-
nated and surface-set diamond bits and crowns 共core drilling bits兲
were developed as well. These bits are best suited to the rotary-
percussion method in deep mining boreholes with hard, abrasive,
fractured formations. The cone-type bit was the main subject of
studies for percussion drilling.
During the percussion-drilling system development for oil and
gas field application, VNIIBT did some special studies of three-
4 Õ Vol. 122, MARCH 2000
Fig. 5 Drill bit bearing design scheme †13‡
cone bit performance. The experiments were conducted with a
6-in-dia milled-tooth bit while drilling blocks of granite and
Vuselemovsky limestone. Rotary drilling tests were conducted to
compare with the percussion drilling results. Some major trends
were observed:
1 Rate of penetration varied linearly with the impact power.
2 Above the minimum threshold WOB, a lower WOB resulted
in a higher percussion drilling rate. For example, in the rotary-
percussion mode of drilling, ROP of 3.3 m/h was achieved with
the 8-3/4-in. bit when the WOB was 4.5 tonne. To achieve the
same ROP in rotary mode required 18.5 tonne.
During the field tests in Bashkiria in hard limestone and dolo-
mites, it was found that the effect of WOB is less important for
percussion drilling 关6兴.
SKB ‘‘Geotechnika’’ developed a range of three and two-cone
bits for rotary and rotary-percussion drilling for 46, 59, 76, 112,
and 132-mm-dia boreholes in hard and superhard abrasive forma-
tions. In the early 1980s, R & D projects were conducted on the
bearing assemblies. Several designs of the sleeve bearings for the
small-diameter cone bits were tested. The main problem with
sleeve bearing is to develop a lock mechanism to prevent loss of
the cones in the hole. Five batches of 76-mm bits were prepared
with five types of lock units. The tests were conducted in granite
blocks. The segment lock was found to be the best one in terms of
bit life. This type of bit was field tested in the Krasnoyarsk city
region at 300–450 m depth in granite with quartzite layers. Aver-
age penetration per bit was 11.8 m at an average ROP of 1.15 m/h.
Some additional modifications in bearing lock design 共Fig. 5兲 and
drill tests were performed before these bits 共III76K-TsA兲 were
introduced for percussion-rotary drilling with GHH G-76U ham-
mer. Tests achieved 17.8 m per bit 共80 percent more than stan-
dard兲, and 2.7 m/h ROP 共34 percent more than standard兲 when
drilling very hard, fractured basalt. The important point was that
the magnitude of drilling parameters, WOB and flow rate in
percussion-rotary, were 40–50 percent less than for rotary drilling
关13兴. The results of this R & D work have been applied to other
bit sizes 共i.e., 46 and 59 mm兲.
SKB ‘‘Geotechnika’’ completed the development of III59K-
TsA and III46K-TsA bits in 1987 共Fig. 6兲. Both bit sizes have
never been manufactured outside of Russia. The 46-mm cone is
the only commercial bit this size in the world. Field tests in hard
and superhard abrasive granites included more than 800 m with
59-mm bits, and 350 m with 46-mm bits. The average penetration
per bit was: 12–21 m with 4–5 m/h ROP for 59-mm bits and 7–8
m with 1.5–2.3 m/h for 46-mm bits, respectively. These tests
were for the rotary mode of drilling only 关14兴.
Transactions of the ASME

Fig. 6 Three-cone 59 and 46-mm-dia drill bit, general scheme
†13‡
Testing of Existing GHH Prototype Tools
nated bit was not available from stock, so the surface-set bit 共de-
signed for hard formation drilling兲 was substituted.
For rotary drilling, a standard DRC test stand was used. MEI
modified a stand for percussion-rotary drilling. The stand was able
to record flow rate, pressure, displacement, WOB, and ROP. The
hydraulic motor used for assembly rotation provided only 150
rpm. The torque was not measured directly, but was estimated by
recording the oil pressure at the hydraulic motor. To determine the
impact frequency, an accelerometer was installed at the input hy-
draulic line. An oscilloscope and plotter were used for data pro-
cessing. The rock blocks were about 35 in. long, so each borehole
was 32–33 in. long, and two or three tests were conducted as each
borehole was drilled.
Percussion-Rotary Drilling Testing Performance. Eight
boreholes were drilled, but each bore included from two to four
tests, where WOB or flow rate were varied. The data are presented
in Table 2. Results of each test were presented on three charts.
Figure 8 shows the plots for test 527. The measured frequency
was 46–74 Hz. The pressure drop was used as the controlled
parameter along with the flow rate and the WOB magnitude,
which was specified for each test.
The cone bits showed considerable bearing wear during these
tests. Some axial play of the cones was clearly observed, and three
inserts of the two-cone bit were lost. Nevertheless, that failure did
not prevent additional testing.
The hammer start-up was very smooth in each test, but some-
times it was difficult to determine if the best operating conditions
were achieved. The reasons for that were: 1兲 difficulty in setting
the desired flow rate with the test stand pumps and control system,
and 2兲 the absence of an on-line frequency measurement system.
The last series of tests, 527, were performed with a ramping of the
pressure drop in an attempt to find the best operating parameters
for those conditions. The influence of the flow-rate/pressure-drop
increase on the ROP was demonstrated.

General Concept. To obtain additional information for better Conclusions From the Test Results.
evaluation of the proposed microborehole coiled-tubing percus- 1 The G-59U共V兲O-type hydraulic hammer results confirmed
sion drilling system components, the following prototypes were the expected performance advantages of percussion-rotary drilling
recommended for lab testing at the Maurer Engineering Inc. Drill- in hard rock.
ing Research Center 共MEI DRC兲 in Houston, Texas.
1 GHH G-59U共V兲O type, 1996 design model: housing diam-
eter 54 mm, single impact energy exceeding 12 J, and fre-
quency range 40–80 Hz;
2 Three-cone bit, III 59K-TsA type;
3 two-cone bit, II 59TK-TsA;
4 diamond impregnated bit, 59-mm diameter;
Testing included a series of 1–3 ft boreholes drilled with vari-
ous assemblies in blocks of granite and marble rocks. Both rotary
and percussion-rotary methods would be used over a range of
WOB, RPM, and flow rates.
The main objectives of the test program were:
1 demonstrate rock bit suitability for horizontal coiled-tubing
drilling in hard rocks;
2 evaluate GHH tool efficiency for horizontal coiled-tubing
drilling in hard rocks;
3 determine the influence of the percussion drilling parameters
on the ROP.
SKB ‘‘Geotechnika’’ prepared a standard G-59U共V兲O hammer
assembled from components manufactured at the SKB factory in
1994 共Fig. 7兲. Table 1 shows standard hammers available 关9兴. The
smallest hammer was selected for testing, disassembled, checked,
adjusted for the expected drilling conditions, and reassembled.
Bench tests at the SKB facilities were performed to check the
hammer operating parameters. The assembled tools, the hammer,
the reflector and a set of spare parts, were delivered to the DRC.
Cone-type drill bits as specified, and a surface-set-type diamond
bit, were purchased from stock in Russia. The 59-mm impreg- Fig. 7 Unified hydraulic hammer, SKB Geotochnika †9‡

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 5

 Table 1 Geotechnika hydraulic hammers specifications

V-for rotary-percussion drilling; with fluid flow reduction unit; U-for percussion rotary drilling; O-with hydrodynamic wave reflector

2 For the coiled-tubing drilling 共CTD兲 application, an efficient
method of rotation needs to be developed. The simplest way
might be to adapt the existing low-speed PDM for that purpose
and conduct additional tests. The power required for the assembly
rotation was roughly evaluated on the basis of the hydraulic motor
performance data. A 1.6-kW PDM should be sufficient for CTD
drilling with the GHH assembly.
3 The modified test stand provided 150 rpm maximum assem-
bly rotation. It is well known from field drilling and laboratory
testing experience that ROP is linearly dependent on the rotary
speed for rotary drilling. For percussion-rotary drilling, this de-
pendency is supposed to be linear as well.
4 ROP increased noticeably, with increased WOB from zero to
1,500–2,000 lb. Further increases in WOB to 3000 lb showed
different results: in marble 100 percent improvement, and granite
15 percent improvement.

Table 2 General results of the percussion-rotary drilling tests at MEI DRC

6 Õ Vol. 122, MARCH 2000 Transactions of the ASME


Fig. 8 Plots with percussion-rotary test results at the MEI DRC
5 The GHH must be operated at the proper flow rate, pressure
drop, and frequency; these are more important for this drilling
method than appropriate WOB.
6 The direct comparison of ROP at the equivalent drilling pa-
rameters can be made on the basis of test 527A and the rotary
drilling test at the same conditions: 150 rpm and 1500 lb WOB.
Journal of Energy Resources Technology
The percussion-rotary method shows a 7.3-times higher ROP than
rotary. At the best operational conditions for both methods,
percussion-rotary still has a 2.3-times advantage in ROP over the
rotary method.
7 The major advantage of percussion drilling for CTD applica-
tion is the possibility of achieving good performance under low
thrust conditions.
Acknowledgments
This work was funded by the United States Department of En-
ergy through contract W-7405-ENG-36.
Thanks to Jody Benson, Los Alamos National Laboratory, for
final proofing and formatting of this paper.
References
关1兴 Knott, D., 1996, ‘‘Deep Thoughts on Drilling Technology,’’ Oil & Gas Jour-
nal, June 10, p. 29.
关2兴 ‘‘High ROP Hammer Drill Sets Australian Records,’’ 1996, Hart’s Petroleum
Engineer International, Sept., p. 13.
关3兴 Gelfgat, M., Gamsakhurdia, G., Geise, J., and Spierings, H., 1994, ‘‘Complete
System For Continuous Coring With Retrievable Tools In Deep Water,’’
IADC/SPE Paper 27521, Dallas, TX.
关4兴 Finger, J. T., 1984, ‘‘Investigation of Percussion Drills for Geothermal Appli-
cation,’’ J. Pet. Tech., Dec., pp. 2128–2136.
关5兴 Graf, L. E., and Kogan, D. I., 1972, Hydro-Percussion Machines and Tools,
Nedra, Moscow, Russia.
关6兴 Kichigin, A. V., Nazarov, V. I., and Tagiev, E. I., 1965, The Percussion-
Rotary Wells Drilling, Nedra, Moscow, Russia.
关7兴 Kiselev, A. T., and Krusir, I. N., 1982, Rotary-Percussion Drilling of Geologi-
cal Prospecting, Wells, Nedra, Moscow, Russia.
关8兴 Melamed, Y. A., 1993, ‘‘The Hydraulic Impact Technology: Large Possibili-
ties and Wide Application Spectrum,’’ Razvedka I Okhrana Nedr (Russian
Journal ‘‘Bowels of the Earth Prospecting & Protection’’).
关9兴 Operations Manuals & Technical Descriptions, 1988, ‘‘Hydro-Percussion
Drilling Techniques and Technology,’’ USSR Ministry of Geology, Moscow,
Russia.
关10兴 Yasov, V. G., 1977, The Theory and Calculations of the Hydraulic Machines
Operational Processes, Nedra, Moscow, Russia.
关11兴 Kiselev, A. T., and Melamed, Y. A., 1984, ‘‘Perspectives of the Rotary-
Percussion Drilling Development with High-Frequency Hydro-hammers with
Low Flow Rates,’’ Proceedings, High School: Geology and Prospecting, No.
11, pp. 85–94.
关12兴 Melamed, Y. A., 1995, ‘‘The Theoretical Ground And Techniques For Hydro-
Percussion Core Drilling Development,’’ doctor of science thesis, Geological
Prospecting Academy, Moscow, Russia.
关13兴 Smirnov, V. G., 1983, ‘‘The Research In The Design Elements Of Three-Cone
Drill Bits 76 mm Diameter With Sleeve Bearing,’’ Techniques for the Ad-
vanced Drilling Methods, Moscow, VPO ‘‘Souzgeotechnika,’’ pp. 108–122.
关14兴 Bodrov, S. M., Ivanovskaya, N. I., and Burinskyi, G. G., 1991, ‘‘Cone-Type
Drill Bits III59K-TsA and III46K-TsA,’’ Razvedka I Okhrana Nedr (Russian
Journal: ‘‘Bowels of the Earth Prospecting & Protection’’), No. 5, pp. 24–26.
MARCH 2000, Vol. 122 Õ 7
 Evaluation of Inflow Performance
of Multiple Horizontal Wells in
Suwan Umnuayponwiwat
Erdal Ozkan
Closed Systems
Petroleum Engineering Department,
Colorado School of Mines, This work presents a model to investigate the inflow performance relationships (IPR) of
Golden, CO 80401 horizontal and vertical wells in a multi-well pattern. The model can be used to compute
the overall and individual well performances. It is shown that stabilized IPRs may not be
sufficient for the evaluation of horizontal well performances due to prolonged transient
flow periods. The results presented in this paper clearly indicate that inflow performance
of wells in a multi-well pattern is a dynamic concept; and, especially in the prediction of
future performances, dynamic rather than static IPR models should be used.
关S0195-0738共00兲00801-3兴

Introduction Inflow Performance. The inflow performance of a well in a

closed drainage region producing at a constant rate, q, can be
The objective of this paper is to discuss the inflow perfor-
evaluated by the following pseudo-steady-state equation:
mances of multiple horizontal and vertical wells in a closed sys-
tem. In such systems, interaction between wells complicates the
estimation of the average pressure, true skin, and the correct ab-
solute open flow potential. Unequal lengths, different production
p̄ 共 t 兲 ⫺p w f ⫽
141.2qB ␮
2kh 冉
ln ␥
4A
e C A r w2 冊 (1)

conditions, and long transient flow periods of horizontal wells
exacerbate problems in the evaluation of the inflow performances
of individual wells and the overall field.
Rodriguez and Cinco-Ley 关1兴 and Camacho-V et al. 关2兴 analyti-
cally investigated the inflow performances of multiple wells in
closed rectangular systems. Both studies, however, only consid-
where p̄ is the volumetric average pressure of the reservoir, p w f is
the flowing wellbore pressure, A is the drainage area, ␥⫽0.5772
. . . , and C A is the Dietz shape factor 共Earlougher 关4兴兲. Equation
共1兲 indicates that the productivity index, PI, defined as
q kh

冉 冊
ered fully penetrating vertical wells and assumed that the produc- PI⫽ ⫽ (2)
p̄⫺p w f 4A
tion took place under boundary-dominated flow conditions 141.2B ␮ 0.5 ln
e ␥ C A r w2
共pseudo-steady state or exponential decline兲 for all times. In other
words, they assumed that the producing wells would reach is a constant and can be used to evaluate the inflow performance.
boundary-dominated flow condition as soon as the production be- For an M-well system where each well j produces at a sequence
gan and the changes in production rates and/or introduction of of N j constant rates, q jk , the counterpart of Eq. 共1兲 is given by
infill wells would not cause pressure transients. M Nj
In this paper, we present an analytical multi-well model that 141.2qB ␮ q jk
can take into account the pressure transients caused by intermit- p̄ 共 t 兲 ⫺p 共 x,y,t 兲 ⫽
kh 兺兺
j⫽1 k⫽1 q j
a (3)
tent production, production rate/pressure changes, and start of pro-
duction at different times in each well. The model is capable of where q is some reference rate, and a j is a constant depending on
considering a mixture of vertical and horizontal wells in the same the shape of the reservoir and the location of the well. For a
field. As in the previous models, we consider the flow of single- rectangular reservoir of dimensions x e ⫻y e , for example, a j is
phase liquids in homogeneous, isotropic porous media; but natu- given by 关3兴
rally fractured reservoirs and permeability anisotropy can be eas-
ily incorporated into the model along the lines suggested by 141.2qB ␮
aj
Ozkan 关3兴. In addition, because our model is a transient flow kh

冉 冊
model and the general solution is in the Laplace transformation
⬁
domain, wellbore storage effects can be easily added and the ye 1 y y 2 ⫹y w2 j x xwj
model can be used for simulating well tests in multi-well systems. ⫽2 ␲ ⫺ ⫹
xe 3 ye 2y 2e
⫹2
k⫽1 x e
兺
cos k ␲ cos k ␲
xe
In this study, however, we restrict our attention to the investiga-
tion of the interaction of wells producing from a common reser-
voir and to the discussion of the effect of multi-well production on
the individual and overall field IPRs. ⫻
cosh k ␲ 冉 xe 冊
y e ⫺ 兩 y⫺y w j 兩
⫹cosh k ␲ 冉
y e ⫺ 兩 y⫹y w j 兩
xe 冊 (4)
ye
k sinh k ␲
xe
Multi-Well Inflow Performance
where x w j and y w j are the coordinates of the well with respect to
Here, we first present the expressions to evaluate the inflow
the origin located at one of the corners of the rectangular
performances of single and multiple wells in closed drainage re-
reservoir.
gions and then comment on the differences in the evaluation of
Equation 共3兲 can be used to define the PI of the field similar to
multi-well performances. For simplicity, we assume that the wells
Eq. 共1兲, provided that the rate sequence of each well is known.
are vertical and fully penetrate the pay thickness.
Note, however, that the derivation of Eq. 共3兲 neglects the pressure
transients caused by the rate changes, and therefore assumes that
Contributed by the Petroleum Division and presented at the ETCE/OMAE2000,
New Orleans, Louisiana, February 14–17, 2000, of THE AMERICAN SOCIETY OF
the stabilized conditions are attained immediately after each
MECHANICAL ENGINEERS. Manuscript received by the Petroleum Division, October change in the constant production rate. It is well known, espe-
25, 1999; revised manuscript received January. Associate Technical Editor: M. Haci. cially for horizontal wells in highly productive formations, that

8 Õ Vol. 122, MARCH 2000 Copyright © 2000 by ASME Transactions of the ASME
the transient flow periods may be exceedingly long 共in the order
of months, even years兲, rendering the direct application of Eq. 共3兲
for performance predictions. In such cases, the use of a model that
can take into account the pressure transients is essential. In the
forthcoming, we propose such a model.
Multi-Well Transient Flow Model
The model we present assumes a rectangular reservoir with
arbitrary well locations. Although the general development does
not assume any specific well type, here we concentrate on vertical
and horizontal wells. Because of space restrictions, we will not
discuss the derivation of the model here. It should be noted, how-
ever, that the model is in the Laplace transformation domain and
the results are inverted into the time domain by using the Stehfest
关5兴 numerical inversion algorithm.
For convenience, we will use the following definitions of the
dimensionless variables in our discussions:
kh 关 p i ⫺p 共 x,y,z,t 兲兴
p D⫽ (5)
141.2qB ␮
where p i is the uniform initial pressure of the field
6.328⫻10⫺3 k
t D⫽ t (6)
␾ c t␮ l 2
␰ D ⫽ ␰ /l ␰ ⫽x, or y (7)
z D ⫽z/h (8)
and
q wD ⫽q w /q (9)
In Eqs. 共6兲 and 共7兲, l is a characteristic length in the system. We
will let f̄ (s) denote the Laplace transform of the function f (t D ), s
the Laplace transform variable with respect to dimensionless time
t D , and p DU the constant 共unit兲 rate solution. Note that p DU de-
pends on the shape of the reservoir and the type and location of
the well. The appropriate expressions for p DU for the cases dis-
cussed in this paper are given in the Appendix.
Let us now consider M wells in a closed reservoir each produc-
ing at a sequence of N j constant rates, q D jk , starting at t D jk⫺1 and
ending at t D jk 共with q D j0 ⫽0 and t D j0 ⫽0兲. Under the aforemen-
tioned assumptions, the pressure distribution in the Laplace do-
main is given by
M Nj
tinuities in Eqs. 共10兲–共12兲. We compared the results obtained
from Eqs. 共11兲 and 共12兲 with those reported in Rodriguez and
Cinco-Ley 关1兴 and Camacho-V et al. 关2兴 and obtained excellent
agreement in stabilized flow periods. In the forthcoming, we dis-
cuss example cases to demonstrate the use of the proposed model
and to highlight the characteristics of multi-well production with
rate or pressure changes.
Example Cases
Here we consider five examples. The data that are common to
all examples are given in Table 1. The reservoir is a closed square
and the total number of wells is three. Wells may be all horizon-
tals or a mixture of horizontal and vertical wells. Table 2 presents
the combinations of the wells and the other relevant information.
For all cases, the horizontal wells are in the x direction. We im-
pose the infinite-conductivity condition along the horizontal wells
by using the pressure averaging method 共Kuchuk et al. 关7兴兲.
Example 1. Here we consider three horizontal wells of equal
length of 8000 ft 共see Table 2兲. The constant production rates of
the wells are also equal and 12,000 stb/d. Well 3, however, starts
production 200 d after Wells 1 and 2. The skin factors of the wells
are S 1 ⫽5, S 2 ⫽0, and S 3 ⫽2.
Figure 1 shows the flowing wellbore pressures of the wells as a
function of time. We have also computed the long-time 共pseudo-
steady-flow兲 approximations and included in Fig. 1. The points at
which the pressure versus time curves computed from our model
collapse with the long-time approximations indicate the times at
the onset of the pseudo-steady state 共PSS兲 and are marked in the
figure. Table 3 summarizes the data and the results shown in Fig.
1.
We make two observations from Fig. 1. First, the transient flow
periods of horizontal wells are considerably long 共approximately
190 d兲 and may not be ignored in production design. Second, as
expected, the start of production at Well 3 introduces a new tran-
sient flow period 共again for approximately 190 d兲 in the entire
field and then a second PSS flow period is attained. Well perfor-
mances are different during the transient, first, and second PSS
flow periods 共introduction of a new producing well in the field
improves efficiency of reservoir depletion兲. As can be seen from
Table 1 Reservoir and fluid data for the example cases

p̄ D 共 x D ,y D ,z D ,s 兲 ⫽ 兺 兺 sp̄
j⫽1 k⫽1
DU j 共 x D ,y D ,z D ,s 兲

共 q D jk ⫺q D jk⫺1 兲
⫻ exp共 ⫺st D jk⫺1 兲 (10)
s
Note that in computing pressures at wellbores, skin effects should
be added to Eq. 共10兲 as follows:
p̄ wDi 共 s 兲 ⫽p̄ D 共 x wDi ,y wDi ,z wDi ,s 兲 Table 2 Well lengths and locations for example cases
Nj
共 q Dik ⫺q Dik⫺1 兲
⫹ 兺 ␴S
k⫽1
i i
s
exp共 ⫺st Dik⫺1 兲 (11)

where S i denotes the mechanical skin factor at well i and ␴ i

⫽h/L hi and 1 for horizontal and vertical wells, respectively. Simi-
larly, for a given sequence of N j constant pressures, p wD jk , 共with
p wD j0 ⫽0兲 at each well, the flow rates of the wells can be com-
puted from

N 共 p wD jk ⫺p wD jk⫺1 兲
兺 k⫽1
j
exp共 ⫺st D jk⫺1 兲
s
q̄ Di 共 s 兲 ⫽ (12)
j⫽1 sp̄ DU j 共 x wDi ,y wDi ,z wDi ,s 兲 ⫹ ␴ i S i
兺M
Note that the modification to the Stehfest’s algorithm suggested
by Chen and Raghavan 关6兴 is useful to evaluate the jump discon-

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 9

Fig. 1 Performance of three horizontal wells; Example
1—delayed production in one of the wells
Table 3 Summary of Example 1 data and results
Fig. 2 Well PIs as a function of time; Example 1
Fig. 2, if high PIs during the long transient flow periods are not
taken into account in well design, significant production loses may
occur. On the other hand, if transient flow information is used in
PI estimations, long-term production figures may not justify the
well design.
Example 2. The objective of this example is to examine the
effect of intermittent production. As in Example 1, we consider
three horizontal wells of equal length 共8000 ft兲. All three wells
10 Õ Vol. 122, MARCH 2000
Fig. 3 Performance of three horizontal wells; Example
2—intermittent production
Table 4 Summary of Example 2 data and results
start production at the same time, but Well 1 is shut in at the 200th
day for a period of 300 d. Other pertinent information is given in
Fig. 3 and Table 4.
Results shown in Fig. 3 and Table 4 emphasize two points.
First, intermittent production in a well significantly effects the
performances of the other wells in the field. In general, as Fig. 4
clearly displays, the inflow performances of the wells in this ex-
ample are strong functions of time and should be obtained and
interpreted accordingly. Second, care should be taken in the inter-
pretation of the pressure buildup data taken at the shut-in well if
the other wells keep producing. Figure 5 shows the Horner plot of
the bottom-hole pressures recorded at the shut-in well 共Well 1兲.
Also shown by the circular data points in Fig. 5 are the ‘‘true’’
buildup responses that would have been recorded at Well 1 if all
three wells had been shut in at the same time. Figure 5 indicates
that the standard Horner analysis can be applied before the pres-
sure responses start deviating from the true buildup behavior un-
der the influence of the producing wells. There are, however, two
problems. First, the confidence interval in the determination of the
straight line may be low. Second and more importantly, the drain-
age shape and area of the well must be known. The latter imposes
the additional requirement that all three wells produce at pseudo-
steady state prior to shut in. In Example 3, we present a ‘‘true’’
buildup case where all three wells are shut in.
Example 3. In this example, we simulate a ‘‘true’’ buildup
test to determine the average reservoir pressure. Again we con-
sider three horizontal wells. Well lengths, production rates, and
the skin values are noted in Fig. 6. All three wells start production
at the same time and shut in after 400 d of production. After
approximately 100 d, the pressures at all three wells stabilize at
Transactions of the ASME
 Fig. 4 Well PIs for Example 2 Fig. 7 Performance of mixed wells; Example 4—change in
production rate

of mixed well performances. Here, we have one horizontal and

two vertical wells in the field. Pertinent data are shown in Fig. 7
and Table 5. Similar characteristics to those discussed in Ex-
amples 1–3 can also be observed in Fig. 7. For completeness, Fig.
8 shows the productivity indices as a function of time for this field
and reemphasizes the ever-changing nature of the productivity
indices in a multi-well system. We also present the pressure maps
of this field at the 50th and 300th days in Figs. 9 and 10. These
figures clearly show the unfavorable depletion of the reservoir in

Table 5 Summary of Example 4 data and results

Fig. 5 Horner plot for the shut-in well „Well 1… in Example 2

1658 psia, which is in agreement with the material balance com-

putations. An interesting point to note here 共although not clearly
shown because of the scale of the figure兲 is the fact that the well-
bore pressure of Well 3 first rises above the average pressure of
the field then drops down to the level of the average pressure. The
pressures at Wells 1 and 2, on the other hand, reach the average
pressure from below. This indicates that Well 3 had been more
productive than Wells 1 and 2 causing less pressure drop in its
vicinity.
Example 4. The objective of this example is to demonstrate
the potential of the model developed in this work in the evaluation

Fig. 6 ‘‘True’’ buildup behavior; Example 3 Fig. 8 Well PIs for Example 4

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 11

 caused by this pressure change and its effect on the rate decline
characteristics of all three wells. We also note from Fig. 11 that
the horizontal well in this field displays a faster rate decline com-
pared to the vertical wells. In fact, the range of the rate change
during the decline period 共and when a pressure change is intro-
duced兲 is so large that it makes well and surface facility design a
real challenge in this field.

Fig. 9 Pressure map for Example 4; t Ä50 d
Conclusions
In this work, we have presented a transient multi-well produc-
tion model and, using the results of this model, demonstrated the
importance of dynamic evaluation of inflow performances in
multi-well systems. The results indicate that extensive care should
be taken in applying the single-well performance models in multi-
well settings. In addition, especially for horizontal wells, transient
flow effects resulting from intermittent production or rate/pressure
changes in one or more wells may have significant impact on
inflow performances. Therefore, care should be taken if the multi-
well performance models neglecting the transient flow periods
presented in the literature are to be used.

Nomenclature
A ⫽ reservoir area, L 2 , ft2
aj ⫽ constant used in Eq. 共3兲
B ⫽ formation volume factor, bbl/stb
CA ⫽ Dietz shape factor
D ⫽ dimensionless 共subscript兲
h ⫽ formation thickness, ft
k ⫽ permeability, L 2 , mD
Lh ⫽ horizontal well length, L, ft
M ⫽ no. of wells
Nj ⫽ no. of constant rates/pressures of well j
p̄ ⫽ reservoir volumetric average pressure, m/Lt 2 , psi
p ⫽ pressure, m/Lt 2 , psi
Fig. 10 Pressure map for Example 4; t Ä300 d p DU ⫽ constant 共unit兲 rate dimensionless pressure
pwf ⫽ flowing bottom-hole pressure, m/Lt 2 , psi
pi ⫽ initial pressure, m/Lt 2 , psi
the vicinity of Well 2. Such figures should be extremely useful for
PI ⫽ productivity index; L 4 t/m, stb/psi
production as well as reservoir engineering design purposes. q jk ⫽ flow rate k of well j, L 3 /t, stb/d
q ⫽ reference flow rate, L 3 /t, stb/d
Example 5. This example shows the use of our transient rw ⫽ wellbore radius, L, ft
multi-well model under constant pressure production conditions. s ⫽ Laplace space parameter
Here, we consider one horizontal 共Well 1兲 and two vertical wells. Si ⫽ skin factor of well, i
The horizontal well undergoes a wellbore pressure decrease at the t ⫽ time, days
200th day of production. Figure 11 clearly shows the transients Vp ⫽ pore volume, L 3 , ft3
x,y,z ⫽ physical coordinates
xe ⫽ reservoir boundary in x direction, L, ft
x w ,y w ⫽ wellbore coordinates, L, ft
ye ⫽ reservoir boundary in y direction, L, ft
␥ ⫽ Euler’s no. 共⫽0.5772...兲
␮ ⫽ fluid viscosity, m/Lt, cp
␾ ⫽ porosity
¯ ⫽ 共overbar兲 Laplace transform
SI Metric Conversion Factors

bbl⫻1.589 873 E⫺01⫽m3

cp⫻1.01 E⫺03⫽Pa•s
ft⫻3.048° E⫺01⫽m
ft3 ⫻2.831 685 E⫺02⫽m3
in.⫻2.541 E⫹00⫽cm
lbf⫻4.448 222 E⫹00⫽N
lbm⫻4.535 924 E⫺01⫽kg
mD⫻9.869 233 E⫺04⫽␮m2
psi⫻6.894 757 E⫹00⫽kPa

Fig. 11 Production rates of mixed wells; Example 5—constant

1
pressure production Conversion factor is exact.

12 Õ Vol. 122, MARCH 2000 Transactions of the ASME

Appendix The pseudoskin function F̄ can be expressed as
Here we present the expressions in the Laplace domain for the
constant unit rate responses for vertical and horizontal wells in F̄ 1 ⫽F̄ D1 ⫹F̄ DB1 ⫹F̄ DB2 (29)
rectangular reservoirs. The details and computational remarks can
be found in Ozkan 关3兴. where
⬁
Vertical Well in a Rectangular Reservoir. The solution for 2␲ cos共 n ␲ z D 兲 cos共 n ␲ z wD 兲
this case can be written as F̄ DB1 ⫽
x eD s n⫽1 兺 冑s⫹a

再 冉 兺 冊
p̄ DU v ⫽p̄ D1 ⫹p̄ DB1 ⫹p̄ DB2 (13) ⬁
where ⫻ E 共 s⫹a 兲 1⫹ e ⫺2m 冑s⫹ay eD
⬁ m⫽1

冉兺 冊冎
2␲ cos共 k ␲ x D1 兲 cos共 k ␲ x D2 兲 ⫺ 冑s⫹ay
p̄ D1 ⫽ 兺
x eD s k⫽1 冑s⫹a
e D1 (14)
⫹e ⫺ 冑s⫹ay D1
⬁

e ⫺2m 冑s⫹ay eD (30)

m⫽1
a⫽n 2
␲ 2 L D2 (15)

冉 冊
⬁
⬁ 4
p̄ DB1 ⫽
␲
E 共 s 兲 1⫹ 兺 e ⫺2m 冑sy eD (16) F̄ DB2 ⫽
s n⫽1 兺
cos共 n ␲ z D 兲 cos共 n ␲ z wD 兲
x eD s 冑s m⫽1

⬁ 1
2␲ cos共 k ␲ x D1 兲 cos共 k ␲ x D2 兲 sin共 k ␲ /x eD 兲 cos共 k ␲ x D1 兲 cos共 k ␲ x D2 兲
兺
⬁
p̄ DB2 ⫽ E 共 s⫹a 兲 k
x eD s k⫽1 冑s⫹a ⫻ 兺 冑s⫹b
冉兺 冊
k⫽1

再 冉 兺 冊
⬁
⬁
⫹e ⫺ 冑s⫹ay D1 e ⫺2m 冑s⫹ay eD (17)
m⫽1 ⫻ E 共 s⫹b 兲 1⫹ e ⫺2m 冑s⫹by eD
m⫽1

冉兺 冊冎
In Eqs. 共14兲–共17兲
⬁
y D1 ⫽ 兩 y D ⫺y wD 兩 (18) ⫹e ⫺ 冑s⫹by D1 e ⫺2m 冑s⫹by eD (31)
m⫽1
y D2 ⫽ 兩 y D ⫹y wD 兩 (19)
where
y D3 ⫽2y eD ⫺ 兩 y D ⫺y wD 兩 (20)
y D4 ⫽2y eD ⫺ 兩 y D ⫹y wD 兩 (21) b⫽n 2 ␲ 2 L D
2
⫹k 2 ␲ 2 /x eD
2
(32)

x D1 ⫽x D /x eD (22) 2␲
⬁
cos共 n ␲ z D 兲 cos共 n ␲ z wD 兲 ⫺ 冑s⫹ay
x D2 ⫽x wD /x eD (23)
F̄ D1 ⫽ 兺
x eD s n⫽1 冑s⫹a
e D1

E 共 s 兲 ⫽e ⫺ 冑sy D1 ⫹e ⫺ 冑sy D2 ⫹e ⫺ 冑sy D3 ⫹e ⫺ 冑sy D4 (24) 4

⬁

Horizontal Well in a Rectangular Reservoir. The solution,

⫹
s n⫽1兺cos共 n ␲ z D 兲 cos共 n ␲ z wD 兲
for a uniform flux well is given by
1
p̄ DUh ⫽p̄ D1 ⫹p̄ DB1 ⫹p̄ DB2 ⫹F̄ 1 (25) ⬁ sin共 k ␲ /x eD 兲 cos共 k ␲ x D1 兲 cos共 k ␲ x D2 兲
k
where ⫻ 兺 冑s⫹b
e ⫺ 冑s⫹by D1

冉 冊
k⫽1
⬁
␲
兺
(33)
p̄ DB1 ⫽ E 共 s 兲 1⫹ e ⫺2m 冑sy eD (26)
x eD s 冑s m⫽1
References
1
⬁ sin共 k ␲ /x eD 兲 cos共 k ␲ x D1 兲 cos共 k ␲ x D2 兲 关1兴 Rodriquez, F., and Cinco-Ley, H., 1993, ‘‘A new Model for Production De-
2 k
p̄ DB2 ⫽ 兺
s k⫽1 冑s⫹a
E 共 s⫹a 兲
cline,’’ Paper SPE 25480, presented at the Production operation Symposium,
Oklahoma, OK, March 21–23.
关2兴 Camacho-V, R., Rodriquez, F., Galindo-N, A., and Prats, M., 1994, ‘‘Opti-

冉 冊
⬁ mum Position for Wells Producing at Constant Wellbore Pressure,’’ Paper

兺
SPE 28715, presented at the SPE International Petroleum Conference & Ex-
⫻ 1⫹ e ⫺2m 冑s⫹ay eD hibition of Mexico, October 10–13.
m⫽1 关3兴 Ozkan, E., 1988, ‘‘Performance of Horizontal Wells,’’ Ph.D. dissertation, The

冉兺 冊
University of Tulsa, Tulsa, OK.
⬁
关4兴 Earlougher, R. C., Jr., 1977, ‘‘Advances in Well Test Analysis, Monograph
⫺ 冑s⫹ay D1
⫹e e ⫺2m 冑s⫹ay eD (27) Series, Society of Petroleum Engineers,’’ Dallas, TX, 5.
m⫽1 关5兴 Stehfest, H., 1970, ‘‘Algorithm 368, Numerical Inversion of Laplace Trans-
forms,’’ D-5, Commun. ACM, Jan 13, No. 1, p. 47.
1 关6兴 Chen, C. C., and Raghavan, R., 1994, ‘‘An Approach to Handle Discontinui-
⬁ sin共 k ␲ /x eD 兲 cos共 k ␲ x D1 兲 cos共 k ␲ x D2 兲 ties by the Stehfest Algorithm,’’ Paper SPE 28419, presented at the SPE 69th
2 k
兺
Annual Technical Conference and Exhibition, September 25–28.
p̄ D1 ⫽ e ⫺ 冑s⫹ay D1 关7兴 Kuchuk, F. J., Goode, P. A., Wilkinson, D. J., and Thambynayagam, R. K. M.,
s k⫽1 冑s⫹a 1991, ‘‘Pressure-Transient Behavior of Horizontal Wells With and Without
(28) Gas Cap or Aquifer,’’ SPE Formation Evaluation, pp. 86–92.

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 13

 Hong-Quan Zhang
e-mail: hong-quan-zhang@utulsa.edu

Subash S. Jayawardena
e-mail: subash-jayawardena@utulsa.edu
Slug Dynamics in Gas-Liquid
Pipe Flow
Clifford L. Redus
e-mail: redus@utulsa.edu The continuity and momentum equations for fully developed and spatially developing slug
flows are established by considering the entire film zone as the control volume. They are
used for the calculations of pressure gradient, slug frequency, liquid holdup in the film,
James P. Brill flow pattern transition, slug dissipation, and slug tracking. Comparison with available
e-mail: brill@utulsa.edu experimental results shows that these equations correctly describe the slug dynamics in
gas-liquid pipe flow. 关S0195-0738共00兲00701-9兴
TUFFP, The University of Tulsa,
600 South College Avenue,
Tulsa, OK 74104

Introduction the pressure loss in the film zone can be important, especially the
Previous investigators used the film segment ⌬z as the control component due to friction between the liquid film and the pipe
volume 共see Fig. 1兲 in developing or applying the momentum wall. A balance exists between all the forces acting on the liquid
equations for the liquid film zone of a slug unit 关1–3兴. From these film. The mixing term is actually the momentum exchange be-
equations, a differential equation can be derived for the height and tween the slug body and the liquid film and plays an important
velocity of the liquid film. It is not easy to integrate the differen- role in affecting the film height and velocity, and, in turn, the
tial equation, especially for an irregular liquid film. Some inves- pressure loss due to friction between the film and the pipe wall.
tigators neglected the convective terms, and assumed that the liq- In this study, the momentum equation will be developed using
uid film is uniform when solving the equation 关3,4兴. the entire liquid film as the control volume 共see Fig. 2兲 to incor-
If the convective terms are not considered, the momentum porate the momentum exchange term. The film is represented with
equations are the same as those used by Taitel and Dukler 关5兴 for
an average velocity and an average liquid holdup. The velocity
equilibrium stratified flow, except they are related to the proper-
ties of the slug body by use of continuity equations. If these mo- and liquid holdup are allowed to change spatially or temporally.
mentum equations are used to calculate the film thickness and This approach will be shown to be appropriate when describing
velocities in upward or downward flow, the results can be unreal- slug dynamics in gas–liquid pipe flows.
istic. For upward flow, they can predict a significant negative film
velocity, even at a small inclination angle, e.g., ⫹1 deg. For
downward flow, they give a film velocity larger than the slug
共mixture兲 velocity if the inclination angle is larger than a few
degrees. According to our experimental observations, neither of
these results is valid in inclined slug flow unless the film length is
quite long.
When a liquid slug travels in a pipeline, it picks up liquid from
the film in front and sheds liquid to the film behind. There can be
a significant difference between the momentum it picks up and the
momentum it sheds. Normally, the latter is larger. Dukler and
Hubbard 关1兴 used a mixing 共or acceleration兲 term when calculat-
ing the pressure gradient across a slug unit. Nicholson et al. 关6兴
and Kokal and Stanislav 关7兴 also considered the pressure drop due
to the acceleration of the slow moving liquid in the film to the Fig. 1 Control volume „⌬ z … used by previous investigators
liquid velocity in the slug body.
Taitel and Barnea 关2兴 found that it was insufficient to consider
just the acceleration of the slow liquid as the mixing term. After a
careful examination of the force balance on the slug body, they
concluded that the contribution due to the change in the liquid
level between the film zone and the liquid slug zone should also
be included in the mixing term.
Normally, the pressure drop in the film zone is negligible if the
film length is relatively short. Many investigators used the pres-
sure loss due to friction and gravity in the slug body plus the
mixing term to calculate the overall pressure gradient. However,

Contributed by the Petroleum Division and presented at the ETCE/OMAE 2000,

New Orleans, Louisiana, February 14–17, 2000, of THE AMERICAN SOCIETY OF
MECHANICAL ENGINEERS. Manuscript received by the Petroleum Division, October
28, 1999; revised manuscript received January 4, 2000. Associate Technical Editor: Fig. 2 Control volume „the entire film zone, l f … used in present
M. Adewumi. study

14 Õ Vol. 122, MARCH 2000 Copyright © 2000 by ASME Transactions of the ASME
Theoretical Development For fully developed slug flow, the input mass flow rate should
equal the output mass flow rate, and a continuity equation for the
Assumptions. In this study, slug flow is assumed to be a se- liquid film can be obtained as
quence of slug units travelling at a translational velocity U t . This
velocity can be expressed as a function of the mixture velocity of R s 共 U t ⫺U s 兲 ⫽R f 共 U t ⫺U f 兲 (4)
the slug U s in the form proposed by Nicklin 关8兴
Similarly, the continuity equation for the gas pocket is
U t ⫽CU s ⫹U d (1)
共 1⫺R s 兲共 U t ⫺U s 兲 ⫽ 共 1⫺R f 兲共 U t ⫺U G 兲 (5)
where U d is the drift velocity 共or the velocity of the elongated
bubble at the limit of U s →0兲. The parameter C is approximately The sum of Eqs. 共4兲 and 共5兲 gives
the ratio of the maximum to the mean velocity of a fully devel-
oped velocity profile. Nicklin 关8兴 proposed a value of 1.2 for the U s ⫽R f U f ⫹ 共 1⫺R f 兲 U G (6)
parameter C for turbulent flow. For laminar flow, C is about 2
For the liquid and gas in the entire slug unit, there are volumet-
共Fabre 关9兴兲. For the drift velocity in horizontal and upward in-
ric continuity equations
clined pipe flow, Bendiksen 关10兴 proposed the use of
l u U SL⫽R s l s U s ⫹R f l f U f
U d ⫽0.54冑gD cos ␤ ⫹0.35冑gD sin ␤
(7)
(2)
where ␤ is positive for upward and negative for downward flows. and
Since there is no correlation available for the drift velocity in
l u U SG⫽ 共 1⫺R s 兲 l s U s ⫹ 共 1⫺R f 兲 l f U G (8)
downward flow, the foregoing correlation is also used for the
translational velocity of downward slug flow in this study. Momentum Equations. In the conventional direction of z
A slug unit consists of the slug body and the liquid film. It is shown in Fig. 2, the momentum output at the left boundary of the
assumed that the separation between gas and liquid in the film liquid film per unit time is
zone is complete and that slip between gas and liquid in the slug
body is negligible. The liquid holdup in the slug body is assumed ␳ L R f A 共 U f ⫺U t 兲 U f
to be a function of the mixture velocity. Gregory et al. 关11兴 pro-
posed the correlation, The momentum input at the right boundary of the film per unit
time is
1

冉 冊
R s⫽ 1.39 (3) ␳ L R s A 共 U s ⫺U t 兲 U s
Us
1⫹
8.66 The force acting on the left boundary of the liquid film due to
the static pressure is
where U s is in m/s. This correlation will be used to estimate R s in
this study.
According to the analyses of Taitel et al. 关12兴 and Barnea and
Brauner 关13兴, there exists a critical slug length for slugs to be
R f A P 1 ⫹ ␳ L g cos ␤ 冕 0
hfe
共 h f e ⫺y 兲 bdy

stable. Therefore, the slug length is assumed to be constant for
either fully developed or spatially developing slug flow. A slug
length of 30D will be used for horizontal and inclined pipe flows,
except when there is an available slug length from measurement.
Consequently, the slug front velocity, the slug tail velocity, and
the slug translational velocity are all identical.
In fact, either slug length or slug frequency has to be given to
close the formulations. The modeling in this study is based on the
assumption of constant slug length. However, it is very easy to
where P 1 is the static pressure in the gas phase at the left bound-
ary, h f e is the film height at the film end 共assumed to be same as
the average film height兲, and b is the width of the film at a dis-
tance y from the pipe bottom. The force acting on the right bound-
ary of the liquid film due to the static pressure is
R s A P 2 ⫹ ␳ L g cos ␤ 冕 hfi

0
共 h f i ⫺y 兲 bdy

change the constant-slug-length mode to the constant-slug-
frequency mode by assuming that the slug frequency is a known
constant and the slug length is an unknown.
Although there is no difficulty to include gas compressibility in
the formulation, the fluids are assumed incompressible in this
work. The liquid film is represented with an average velocity and
an average liquid holdup. We assume that the film height and
velocity at the film tail is equal to the average values.
Equations for Developed Slug Flow
Continuity Equations. In this work, the entire liquid film and
the gas pocket in the film zone of a slug unit will be used as the
control volume, as shown in Fig. 2. Continuity and momentum
equations will be derived for them relative to a coordinate system
moving with the translational velocity U t .
The output liquid mass flow rate at the left boundary and the
input liquid mass flow rate at the right boundary of the liquid film
are
where P 2 is the static pressure in the gas phase at the right bound-
ary, and h f i is the front film height and can be calculated from the
liquid holdup R s . Also, there is a force acting on the curved
surface of the film in the z direction due to the static pressure.
Assuming the pressure on the curved surface is P 2 , the force
component in the z direction is
P 2 A 共 R s ⫺R f 兲
The frictional force acting on the film at the wall 共in the oppo-
site direction of z兲 is
⫺␶ fS fl f
The frictional force acting on the film at the interface 共in the same
direction of z兲 is
␶ iS il f
The gravitational force is

␳ L R f A 共 U f ⫺U t 兲 ⫺ ␳ L R f Al f g sin ␤

and All the foregoing forces should be in balance for fully devel-
oped slug flow. Therefore, the momentum equation for the liquid
␳ L R s A 共 U s ⫺U t 兲 film can be obtained,

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 15

 共 P 2 ⫺ P 1 兲 ␳ L 共 U t ⫺U f 兲共 U s ⫺U f 兲 ␶ i S i ⫺ ␶ f S f ␳ G 共 U G ⫺U f 兲 兩 U G ⫺U f 兩
⫽ ⫹ ␶ i⫽ f i (15)
lf lf RfA 2
␳ L g cos ␤ Due to the turbulent nature of most slug flows, turbulent friction
⫺ ␳ L g sin ␤ ⫺ factors will be used to calculate the wall shear stresses for the
lfRfA
liquid film and the gas pocket

⫻ 冉冕 hfi

0
共 h f i ⫺y 兲 bdy⫺ 冕0
hfe
共 h f e ⫺y 兲 bdy 冊 (9) f ⫽C Re⫺n
where C⫽0.046, n⫽0.2. The Reynolds numbers were defined as
(16)

The momentum equation for the gas pocket can be written as

4A f U f ␳ L
共 P 2 ⫺ P 1 兲 ␳ G 共 U t ⫺U G 兲共 U s ⫺U G 兲 ␶ i S i ⫹ ␶ G S G Re f ⫽ (17)
S f␮L
⫽ ⫺ ⫺ ␳ G g sin ␤
lf lf 共 1⫺R f 兲 A
4A G U G ␳ G
(10) ReG ⫽ (18)
共 S G ⫹S i 兲 ␮ G
Equating pressure drop in the two phases gives the combined
momentum equation for the liquid and the gas. Here, A f and A G are the cross-sectional
areas occupied by the liquid film and the gas pocket, respectively.
␳ L 共 U t ⫺U f 兲共 U s ⫺U f 兲 ⫺ ␳ G 共 U t ⫺U G 兲共 U s ⫺U G 兲 For the interfacial gas-liquid shear stress, a constant value f i
lf ⫽0.014 共Cohen and Hanratty 关14兴兲 will be used as the friction

冉冕 冕 冊
factor, assuming there are small-amplitude waves on the interface.
␳ L g cos ␤ hfi hfe
⫺ 共 h f i ⫺y 兲 bdy⫺ 共 h f e ⫺y 兲 bdy Equations for Developing Slug Flow. For developing slug
lfRfA 0 0 flow, we do not focus on a specific slug unit, but on a statistical

⫺
␶fSf
⫹
␶ GS G
R f A 共 1⫺R f 兲 A
⫹ ␶ iS i
1
⫹
1
冉
R f A 共 1⫺R f 兲 A 冊 slug unit. During slug dissipation, it is assumed that some slugs
disappear quickly, but others do not change in their length. The
slug unit length and film length will increase 共or decrease兲 with a
⫺ 共 ␳ L ⫺ ␳ G 兲 g sin ␤ ⫽0 (11) decrease 共or increase兲 in the slug frequency.

The first two terms on the LHS of Eq. 共11兲 are the momentum Continuity Equations. As assumed, the slug frequency of a
exchange between the slug body and the liquid film zone. They developing slug flow may change while the slug length remains
appear in the combined momentum equation as a result of taking constant. All the mass in the disappearing slug units, which in-
the entire film as the control volume. If the gas part of the first clude the slug bodies and the liquid films, will be distributed to
term is neglected, the momentum exchange term in Eq. 共11兲 is the each liquid film of the remaining slugs passing a specific location
same as the mixing term derived by Taitel and Barnea 关2兴. If the per unit time. Therefore, the mass input rate to the liquid film due
integration term 共the second term in Eq. 共11兲兲 is also dropped, the to the change rate of the slug frequency is
momentum exchange becomes the same as the mixing term used ␳ L 共 R s l s ⫹R f l f 兲 A d ␯
by Dukler and Hubbard 关1兴, Nicholson et al. 关6兴, and Kokal and ⫺
Stanislav 关7兴 in calculating the pressure gradient in slug flow.
␯ dt
In fact, these investigators obtained the mixing term by consid- where the slug frequency ␯ is given by
ering the slug body as the control volume. Therefore, it is natural
that the same momentum input 共with opposite sign兲 should appear Ut
␯⫽ (19)
in the momentum equation when the entire liquid film is taken as lu
the control volume. The characteristics of the liquid film are gov-
erned by the foregoing continuity and momentum equations. Then, the mass input rate becomes
The slug unit length given by
l u ⫽l s ⫹l f (12)
␳ L 共 R s l s ⫹R f l f 兲 A 冉 1 dl u 1 dU t
⫺
l u dt U t dt 冊
is not independent of the continuity equations, Eqs. 共7兲, 共8兲, for For developing slug flow, the mass change rate in the control
steady slug flow. Therefore, only two of them can be used at the volume should also be included in the balance equation
same time. In this work, we choose to use Eqs. 共12兲 and 共7兲 in d
solving different slug flow problems. 共␳ R l A兲
The continuity and momentum equations, Eqs. 共4兲, 共5兲, 共7兲, dt L f f
共11兲, and 共12兲, will be used for pressure drop, slug characteristics, Then, the continuity equation for the liquid in the film zone be-
and flow pattern predictions for fully developed slug flow. The comes
solution procedure is as follows: 1兲 Given an estimated value of
R f , U f and U G are calculated using Eqs. 共4兲 and 共5兲, respectively. d
l f is calculated using Eqs. 共7兲 and 共12兲. 2兲 Then, every term in Eq. 共 R l 兲 ⫽R s 共 U t ⫺U s 兲 ⫺R f 共 U t ⫺U f 兲
dt f f
共11兲 is calculated using the liquid holdups, liquid and gas veloci-
ties, and the film length. 3兲 If the residual of Eq. 共11兲 is smaller
than a given tolerance, the calculation is terminated. Otherwise, a
new R f is estimated according to the results of previous iterations
⫹ 共 R s l s ⫹R f l f 兲 冉 1 dl u 1 dU t
⫺
l u dt U t dt 冊 (20)

and the process is repeated. For a slug flow that experiences only a spatial development
The shear stresses in Eq. 共11兲 are evaluated as dU t
⫽0 (21)
␳ L U 2f dt
␶f⫽ f f (13)
2 d ⳵
⫽U t (22)
␳ G U G2 dt ⳵z
␶ G⫽ f G (14)
2 The continuity equation for the film then becomes

16 Õ Vol. 122, MARCH 2000 Transactions of the ASME

 ⳵ R s 共 U t ⫺U s 兲 ⫺R f 共 U t ⫺U f 兲 共 R s l s ⫹R f l f 兲 ⳵ l u ␳ L V t U SL ⳵ l u ␳ L 关 R s 共 U t ⫺U s 兲 U s ⫺R f 共 U t ⫺U f 兲 U f 兴
共 R l 兲⫽ ⫹ ⫽
⳵z f f Ut lu ⳵z R f l f dz Rflf
(23)
␶ iS i⫺ ␶ f S f 共 P 2⫺ P 1 兲
⫹ ⫺ ␳ L g sin ␤ ⫺
Since the slug length is assumed to be constant RfA lf

⳵l f ⳵lu
⳵z
⫽
⳵z
(24)
⫺
␳ L g cos ␤
lfRfA 冉冕 0
hfi
共 h f i ⫺y 兲 bdy

After rearranging, the continuity equation for the liquid in the film
zone becomes
⫺ 冕0
hfe
共 h f e ⫺y 兲 bdy 冊 (29)

⳵ R f R s 共 U t ⫺U s 兲 ⫺R f 共 U t ⫺U f 兲 l s 共 R s ⫺R f 兲 ⳵ l u Similarly, the momentum equation for gas in the film zone of a

lf ⫽ ⫹ (25)
⳵z Ut lu ⳵z spatially developing slug flow is

Similarly, the continuity equation for the gas in the film zone is ␳ G V t U SG ⳵ l u
共 1⫺R f 兲 l f ⳵ z
⳵ R f 共 1⫺R s 兲共 U t ⫺U s 兲 ⫺ 共 1⫺R f 兲共 U t ⫺U G 兲
⫺l f ⫽ ␳ G 关共 1⫺R s 兲共 U t ⫺U s 兲 U s ⫺ 共 1⫺R f 兲共 U t ⫺U G 兲 U G 兴
⳵z Ut
⫽
共 1⫺R f 兲 l f
l s 共 R f ⫺R s 兲 ⳵ l u
⫹ (26) ␶ iS i⫹ ␶ GS G 共 P 2⫺ P 1 兲
lu ⳵z
⫺ ⫺ ␳ G g sin ␤ ⫺ (30)
共 1⫺R f 兲 A lf
From Eqs. 共25兲 and 共26兲, we can also reach Eq. 共6兲, which was
obtained from the continuity equations, 共4兲 and 共5兲, for steady slug
By equating the pressure drop in the foregoing momentum equa-
flow.
tions for liquid and gas in the film zone, the combined momentum
Momentum Equations. For developing slug flow, the momen- equation for spatially developing slug flow can be obtained,
tum change rate in the liquid film as the control volume is
Ut ⳵lu
d 共 ␳ LU f ⫺ ␳ GU G 兲
共␳ U R l A兲 lf ⳵z
dt L f f f
␳ L 关 R s 共 U t ⫺U s 兲 U s ⫺R f 共 U t ⫺U f 兲 U f 兴
By use of Eqs. 共7兲 and 共19兲, this momentum change rate can be ⫽
Rflf
written in a different form, assuming U s R s is constant
␳ G 关共 1⫺R s 兲共 U t ⫺U s 兲 U s ⫺ 共 1⫺R f 兲共 U t ⫺U G 兲 U G 兴
⫺
d d 共 1⫺R f 兲 l f
共 ␳ L U f R f l f A 兲 ⫽ ␳ L A 共 U f R f l f ⫹U s R s l s 兲
dt dt ␶ iS i⫺ ␶ f S f ␶ iS i⫹ ␶ GS G
⫹ ⫹ ⫺ 共 ␳ L ⫺ ␳ G 兲 g sin ␤
dl u RfA 共 1⫺R f 兲 A
⫽ ␳ L AU SL

冉冕 冕 冊
dt ␳ L g cos ␤ hfi hfe
⫺ 共 h f i ⫺y 兲 bdy⫺ 共 h f e ⫺y 兲 bdy
␳ LA d␯ lfRfA
⫽⫺ 共 U f R f l f ⫹U s R s l s 兲 (27) 0 0
␯ dt
(31)
It is clear from Eq. 共27兲 that the momentum change rate in the
liquid film is equal to the momentum change due to the change The continuity and momentum equations, Eqs. 共7兲, 共8兲, 共12兲,
rate of the slug frequency. 共25兲, and 共31兲, for spatially developing slug flow will be used for
The momentum change rate in every liquid film should equal calculations of slug dissipation and slug tracking. The solution
the net force exerted on it. Accordingly, the momentum equation procedure is: 1兲 The entrance conditions are specified normally by
for liquid in the film zone can be written as solving the continuity and momentum equations for fully devel-
oped slug flow assuming the entrance section is sufficiently long
␳L d ␳ L 关 R s 共 U t ⫺U s 兲 U s ⫺R f 共 U t ⫺U f 兲 U f 兴 to ensure the fully developed flow state. 2兲 U f and U G are calcu-
共 R l U 兲⫽ lated using Eqs. 共7兲, 共8兲, and 共12兲. 3兲 Equations 共25兲 and 共31兲 are
R f l f dt f f f Rflf integrated using a finite differencing method to obtain values of
␶ iS i⫺ ␶ f S f 共 P 2⫺ P 1 兲 R f and l u at the next grid point. Steps 2 and 3 are repeated until
⫹ ⫺ ␳ L g sin ␤ ⫺ the downstream boundary is reached.
RfA lf

⫺
␳ L g cos ␤
lfRfA 冉冕 0
hfi
共 h f i ⫺y 兲 bdy

⫺ 冕 0
hfe
共 h f e ⫺y 兲 bdy 冊 (28)
Results and Discussions
Pressure Gradient. U f , U G , R f , and l u (l f ) are obtained
after solving the continuity and momentum equations for fully
Applying Eq. 共7兲 and assuming that the flow experiences only developed slug flow. Then, the pressure gradient in the film zone
spatial development, this equation becomes is calculated using

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 17

 冋 册
⳵P
⳵z f
⫽
␳ L R f 共 U t ⫺U f 兲共 U s ⫺U f 兲
lf

␳ G 共 1⫺R f 兲共 U t ⫺U G 兲共 U s ⫺U G 兲
⫹
lf

⫺
␳ L g cos ␤
lfA 冉冕 0
hfi
共 h f i ⫺y 兲 bdy⫺ 冕0
hfe
共 h f e ⫺y 兲 bdy 冊
␶ f S f ⫹ ␶ GS G
⫺ ⫺ 关 ␳ L R f ⫹ ␳ G 共 1⫺R f 兲兴 g sin ␤ (32)
A

The pressure gradient in the slug body is

冋 册
⳵P
⳵z s
⫽⫺
␳ L R f 共 U t ⫺U f 兲共 U s ⫺U f 兲
ls
Fig. 4 Computed overall À dp Õ dx , compared with Stanislav
et al. †16‡ data

␳ G 共 1⫺R f 兲共 U t ⫺U G 兲共 U s ⫺U G 兲
⫺
ls

冉冕 冕 冊
tum exchange term. It is apparent that a large difference from the
␳ L g cos ␤ hfi hfe
experimental data results if only the frictional term is considered
⫹ 共 h f i ⫺y 兲 bdy⫺ 共 h f e ⫺y 兲 bdy
l sA 0 0 in the pressure gradient calculation. More significantly, this dif-
ference shows conceptually the magnitude of the momentum out-
␶ sS s put from the slug body to the liquid film.
⫺ ⫺ ␳ s g sin ␤ (33) The calculations also show that the influence of the change in
A
liquid level between the film zone and the slug body is very small.
The slug body mixture density is calculated from The integration terms are only about 1 percent of the mixing term
共the momentum exchange term兲. Actually, the pressure drop due
␳ s ⫽ ␳ L R s ⫹ ␳ G 共 1⫺R s 兲 . (34) to the change in liquid level across the film can be recovered as
long as the film is connected with the slug body. Therefore, it may
The pressure gradient for a slug unit will be not be necessary to consider it in calculation of the overall pres-

冋 册
⳵P l f ␶ f S f ⫹ ␶ G S G l s ␶ s S s g sin ␤
⫽⫺ ⫺ ⫺
⳵z u lu A lu A lu
g sin ␤
⫻l f 关 ␳ L R f ⫹ ␳ G 共 1⫺R f 兲兴 ⫺
lu
The force due to the momentum exchange between the liquid film
and the slug body does not appear in the pressure gradient for the
slug unit. However, it plays an important role in determining the
velocity and liquid holdup of the film; and this, in turn, affects the
overall pressure gradient in the slug flow.
Figure 3 shows a comparison of the computed pressure gradient
in the slug body with the experimental results of Kouba 关15兴. In
Kouba’s study, kerosene and air were used as the liquid and gas
phases flowing in a 3-in-dia horizontal pipeline. The dashed line
in Fig. 3 is the calculated pressure gradient without the momen-
sure gradient.
Figure 4 shows a comparison of computed values of the overall
pressure gradient for slug flow with the experimental results of
Stanislav et al. 关16兴. They used a light refined machine oil and air
as the liquid and gas phases flowing in a 25.8-mm i.d. upward
l s␳ s (35) inclined pipeline.
Slug Characteristics. Figure 5 shows the computed liquid
holdup in the film region of slug flow compared with the mea-
surements of Kouba 关15兴. Figure 6 is a comparison between the
computed average film liquid holdups and those measured by An-
dreussi et al. 关3兴. The test fluids of that study were water and air at
atmospheric conditions and the test section was a 53-mm i.d. hori-
zontal pipe. Their measurements were made in the film region in
front of the slugs.
In Fig. 7, the computed slug frequencies at different flow rates
are compared with the experimental results of Andreussi et al. 关3兴.
Good agreement is observed in all these cases.

Fig. 3 Computed À dp Õ dz in the slug body, compared with Fig. 5 Computed film liquid holdup, compared with Kouba
Kouba †15‡ data †15‡ data

18 Õ Vol. 122, MARCH 2000 Transactions of the ASME

Fig. 6 Computed film liquid holdup, compared with Andreussi
et al. †3‡ data
Fig. 7 Computed slug frequency, compared with Andreussi
et al. †3‡ data
Flow Pattern Transition. When the transition from slug flow
to stratified flow occurs, the film length l f becomes infinitely long.
The momentum exchange terms in the combined momentum
equation, Eq. 共11兲, become zero. If the superficial gas velocity
U SG is given, the liquid holdup of the film and the gas velocity
can be calculated using the two continuity equations, Eqs. 共5兲 and
共8兲
R s 共 U t ⫺U s 兲 ⫹U s ⫺U SG
Rf⫽ (36)
Ut
and
U SGU t
U G⫽ (37)
U SG⫹ 共 1⫺R s 兲共 U t ⫺U s 兲
Then, U f can be calculated with the momentum equation, Eq.
共11兲, and finally U SL⫽R f U f . Several iterations are required to
obtain an accurate value of U SL . The curve of U SL versus U SG in
a flow pattern map defines the boundary between slug flow and
stratified flow.
This flow pattern transition prediction is compared with the
flow pattern boundary calculated using the TUFFP FLOPAT pro-
gram 共Triggia et al. 关17兴兲. For near-horizontal flow, this program
is based on the mechanistic model of Taitel and Dukler 关5兴. The
following fluid properties and pipe diameter are used for the cal-
culations: ␳ L ⫽837.9 kg/m3, ␮ L ⫽0.0176 kg/m-s, ␳ G ⫽2.3 kg/m3,
␮ G ⫽18.88⫻10⫺6 kg/m-s, and D⫽50.8 mm.
Figure 8 shows the comparison at an inclination angle of ⫺5
deg. It is seen that the predicted superficial liquid velocities at the
slug-stratified transition boundary are higher at low gas flow rates
Journal of Energy Resources Technology
Fig. 8 Flow pattern prediction compared with FLOPAT „À5
deg inclination…
and lower at high gas flow rates than predicted by FLOPAT.
However, the trends of both predictions are similar.
Developing Slug Flow. Using the continuity and momentum
equations for spatially developing slug flow, Eqs. 共7兲, 共8兲, 共12兲,
共25兲, and 共31兲, slug dissipation in the downward flow section of
the pipeline shown in Fig. 9共a兲 is calculated. The inclination
angles for the upward and downward sections are the same. The
upward flow is assumed to be fully developed at the top elbow.
The continuity and momentum equations for fully developed flow
are solved for the upward flow and the results are used as the
entrance conditions for the downward developing flow.
Equations 共25兲 and 共31兲 were solved using a first-order finite
difference method. In order to implement an explicit calculation, a
forward differencing scheme was used for the discretizations of
both ⳵ l u / ⳵ z and ⳵ R f / ⳵ z. The spatial domain over which the
simulation was carried out was discretized with a uniform spac-
ing, ⌬z⫽0.5D. Several ⌬z values from large to small were tested
to obtain this value that ensures there was no further change in the
results, even if a smaller ⌬z were used. According to the experi-
mental results observed, the slug length was set at 20D.
Figure 10 shows the changes of slug frequency in the down-
ward section at inclinations of ⫺5 and ⫺10 deg, respectively.
Also plotted are the experimental data. In the experiments 共Zhang
et al. 关18兴兲, a top bend was used to connect the upward and down-
ward sections and to form a U-shaped flow loop. The bend is a
half circle with a radius of 2 ft. Therefore, the integration of Eqs.
共25兲 and 共31兲 was carried out according to the gradual variation of
the pipe inclination angle. The effect of the centrifugal force in the
top bend was not considered. It is seen that the comparison of the
frequency change in the downward flow section is good.
Fig. 9 Pipeline configurations for slug dissipation and slug
tracking calculations; „a… transition from uphill to downhill flow,
„b… transition from horizontal to downhill flow
MARCH 2000, Vol. 122 Õ 19
Fig. 10 Comparison of predicted and measured frequency
versus dissipation distance
Fig. 11 Comparison of predicted and measured slug body to
slug unit length ratio versus dissipation distance at À5 deg
Fig. 12 Dependence of dissipation distance on gas and liquid
flow rates
It was found by Zhang et al. 关18兴 that the slug dissipation was
manifested by slug frequency and/or slug length decreases. In
both of these cases, the ratio of slug length to slug unit length
(l s /l u ) decreases. In fact, the ratio l s /l u is more meaningful than
the slug frequency or slug length as a parameter for determining
pressure gradient in slug flow. Figure 11 shows the comparison
between computed l s /l u and experimental results for two sets of
flow rates at ⫺5 deg. At U SG⫽1.05 m/s, U SL⫽1.0 m/s, both the
slug frequency and slug length changed when the flow turned
from upward to downward sections. At U SG⫽1.85 m/s, U SL
⫽1.41 m/s, the slug frequency did not change at the beginning of
the downward section and slightly decreased in the downstream
20 Õ Vol. 122, MARCH 2000
Fig. 13 Variation of slug frequency along an inclined pipeline
starting from horizontal flow „ U SLÄ U SGÄ1 mÕs…
development. However, there was an apparent drop in l s /l u after
the top bend, mainly due to a shrinking of the slug length.
With slug dissipation, the slug unit length l u increases. Theo-
retically, the slug flow becomes stratified flow as l u becomes in-
finitely large. A very long developing distance may be needed to
reach this infinite value. However, if a certain dissipation ratio
l u /l u0 共l u0 is the entrance slug unit length兲 and a developing dis-
tance Z are given, a boundary can be defined on the U SL versus
U SG plane. In Fig. 12, the dissipation ratio is 10, and the four lines
correspond to dissipation lengths of infinite, 500D, 200D, and
100D, respectively.
In Fig. 9共b兲, the slugs travel from a horizontal pipeline to a
downhill 共or uphill兲 pipeline of a specified inclination angle. The
flow is assumed to be fully developed at the end of the horizontal
pipe. Because of the change in gravitational force in the momen-
tum balance, the flow will experience a developing period. The
film velocity will increase and the slugs may dissipate in the
downward flow section.
Figure 13 shows the changes in slug frequency with distance
from the beginning of the inclined flow sections at inclination
angles of 0, ⫾2, ⫾5, and ⫾10 deg. The frequency is normalized
with the value at the entrance of the inclined segment. The super-
ficial gas and liquid velocities 共U SG and U SL兲 for the calculations
are all 1 m/s. The slug frequency and other parameters of the fully
developed horizontal flow are calculated by solving the steady
continuity and momentum equations. These parameters are used
as the entrance conditions for the calculations of developing slug
flows in the downward or upward pipeline segments.
It is seen from Fig. 13 that slug frequencies of downward flows
decrease rapidly within a short distance from the beginning of the
inclined section. Then, the decrease of the frequency slows down.
At ⫺5 and ⫺10 deg, the fully developed state is stratified flow.
Obviously, much longer distance is needed for the slugs to dissi-
pate at ⫺5 deg than at ⫺10 deg. The slug frequency at an incli-
nation of ⫺2 deg reaches a constant value after about 200D. The
fully developed state at this inclination angle is still slug flow. The
slug frequency increases as the slugs travel from the horizontal
section to the upward sections at different inclination angles. New
slugs are initialized as the inclination angle increases. The transi-
tion takes only a short distance, compared to downward flows.
Concluding Remarks
The momentum exchange between the slug body and the liquid
film in a slug unit was added to the combined momentum equa-
tions. This was accomplished by considering the entire liquid film
as the control volume. The continuity and momentum equations
for spatially developing slug flow were further developed by as-
suming that the slug frequency changes while the slug length re-
mains constant.
The continuity and momentum equations were used to compute
pressure gradient, liquid holdup in the film, slug frequency, flow
Transactions of the ASME
pattern boundary between slug and stratified flow, slug dissipa- G ⫽ gas
tion, and slug tracking in spatially developing flows. The com- i ⫽ interfacial
puted results were compared with available experimental results L ⫽ liquid
under different conditions and a good agreement was observed. s ⫽ slug
It should be noted that the equations are based on the assump- SG ⫽ superficial gas
tions of knowing the translational velocity of the interface U t , the SL ⫽ superficial liquid
liquid holdup in the slug body R s , the slug length l s , the interfa- t ⫽ translational
cial friction factor f i , etc. Some of these parameters, like f i , still u ⫽ slug unit
remain as an unresolved problem.
References
关1兴 Dukler, A. E., and Hubbard, M. G., 1975, ‘‘A Model for Gas-Liquid Slug
Acknowledgments Flow in Horizontal and Near Horizontal Tubes,’’ Ind. Eng. Chem. Fundam.,
The authors wish to thank the TUFFP member companies for 14, No. 4, pp. 337–347.
关2兴 Taitel, Y., and Barnea, D., 1990, ‘‘Two-Phase Flow,’’ Adv. Heat Transfer, 20,
supporting the research project ‘‘Slug Dissipation in Downward pp. 83–132.
Flow,’’ which resulted in this theoretical work. Special thanks are 关3兴 Andreussi, P., Bendiksen, K. H., and Nydal, O. J., 1993, ‘‘Void Distribution in
also due to Drs. Yehuda Taitel, Gene Kouba, Cem Sarica and Slug Flow,’’ Int. J. Multiphase Flow, 19, No. 5, pp. 817–828.
Tom Chen for their valuable suggestions and discussions. 关4兴 Taitel, Y., and Barnea, D., 1998, ‘‘Effect of Gas Compressibility on a Slug
Tracking Model,’’ Chem. Eng. Sci., 53, pp. 2089–2097.
关5兴 Taitel, Y., and Dukler, A. E., 1976, ‘‘A Model for Predicting Flow Regime
Nomenclature Transition in Horizontal and Near Horizontal Gas-Liquid Flow,’’ AIChE. J.,
22, No. 1, pp. 47–55.
A ⫽ cross-sectional area 关6兴 Nicholson, M. K., Aziz, K., and Gregory, G. A., 1978, ‘‘Intermittent Two-
b ⫽ width phase Flow in Horizontal Pipes: Predictive Models,’’ Can. J. Chem. Eng., 56,
C ⫽ coefficient pp. 653–663.
D ⫽ pipe diameter 关7兴 Kokal, S. L., and Stanislav, J. F., 1989, ‘‘An Experimental Study of Two-
phase Flow in Slightly Inclined Pipes—II: Liquid Holdup and Pressure Drop,’’
f ⫽ friction factor Chem. Eng. Sci., 44, No. 3, pp. 681–693.
g ⫽ gravity acceleration 关8兴 Nicklin, D. J., 1962, ‘‘Two-Phase Bubble Flow,’’ Chem. Eng. Sci., 17, pp.
h ⫽ film height 693–702.
l ⫽ length 关9兴 Fabre, J., 1994, ‘‘Advancements in Two-Phase Slug Flow Modeling,’’ SPE
27961, University of Tulsa Centennial Petroleum Engineering Symposium,
n ⫽ coefficient August 29–31.
P ⫽ pressure 关10兴 Bendiksen, K. H., 1984, ‘‘An Experimental Investigation of The Motion of
R ⫽ liquid holdup Long Bubbles in Inclined Tubes,’’ Int. J. Multiphase Flow, 10, pp. 467–483.
Re ⫽ Reynolds no. 关11兴 Gregory, G. A., Nicholson, M. K., and Aziz, K., 1978, ‘‘Correlation of the
Liquid Volume Fraction in the Slug for Horizontal Gas-Liquid Slug Flow,’’
S ⫽ perimeter Int. J. Multiphase Flow, 4, pp. 33–39.
t ⫽ time 关12兴 Taitel, Y, Barnea, D., and Dukler, A. E., 1980, ‘‘Modeling Flow Pattern Tran-
U ⫽ velocity sitions for Steady Upward Gas-Liquid Flow in Vertical Tubes,’’ AIChE. J., 26,
y ⫽ coordinate normal to interface pp. 345–354.
关13兴 Barnea, D., and Brauner, N., 1985, ‘‘Hold-Up of the Liquid Slug in Two Phase
z ⫽ coordinate along pipe Intermittent Flow,’’ Int. J. Multiphase Flow, 11, pp. 43–49.
␤ ⫽ inclination angle 关14兴 Cohen, S. L., and Hanratty, T. J., 1968, ‘‘Effects of Waves at a Gas Liquid
␮ ⫽ dynamic viscosity Interface on a Turbulent Air Flow,’’ J. Fluid Mech., 31, pp. 467–469.
␯ ⫽ frequency 关15兴 Kouba G. E., 1986, ‘‘Horizontal Slug Flow Modeling and Metering,’’ Ph.D.
dissertation, The University of Tulsa, Tulsa, OK.
␳ ⫽ density 关16兴 Stanislav, J. F., Kokal, S., and Nicholson, M. K., 1986, ‘‘Intermittent Gas-
␶ ⫽ shear stress Liquid Flow in Upward Inclined Pipes,’’ Int. J. Multiphase Flow, 12, No. 3,
pp. 325–335.
Subscripts 关17兴 Triggia, A. A., Caetano, E. F., and Shoham, O., 1986, ‘‘Gas-Liquid Two-
d ⫽ drift Phase Flow Pattern Prediction Computer Library,’’ J. Pipelines, 5, pp. 207–
220.
f ⫽ film 关18兴 Zhang, H.-Q., Yuan, H., Redus, C. L., and Brill, J. P., 2000, ‘‘Observation of
fe ⫽ film equilibrium Slug Dissipation in Downward Flow,’’ presented at ETCE/OMAE 2000, New
fi ⫽ film initial Orleans, LA, February 14–17.

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 21

 Avni Serdar Kaya
Petroleum Soft, Inc.,
Tulsa, OK 74133-1697
Investigation of Transition
X. Tom Chen
Texaco EPTD, From Annular to Intermittent
Humble, TX 77338
Flow in Pipes
Cem Sarica
Petroleum and Natural Gas Engineering, A new unified model is proposed for the transition from annular to intermittent flow
The Pennsylvania State University, patterns for the entire range of pipe inclination angles. Experimentally, it has been ob-
122 Hosler Building, served that the transition from annular flow takes place at a critical void fraction. To
University Park, PA 16802 obtain a transition boundary equation, conservation of momentum equations for gas and
the liquid film are combined and solved with the critical void fraction. The new model
captures the correct transition characteristics, agrees favorably with experimental flow
James P. Brill pattern data, and performs the best when compared with previous transition models.
Department of Petroleum Engineering, 关S0195-0738共00兲00601-4兴
The University of Tulsa,
Tulsa, OK 74104-3189

Introduction horizontal and downward flow. However, at more steep down-

ward inclination angles, another mechanism plays a role in the
Two-phase flow in the petroleum industry is a very common
transition from stratified flow to annular flow which can be pre-
occurrence because of the simultaneous production of gas and
dicted by the Barnea et al. 关7兴 model. The transition mechanism in
liquid from hydrocarbon reservoirs. When gas and liquid flow
the model reveals that, due to a high liquid flow rate, liquid drop-
simultaneously in a pipe, various flow configurations or patterns
lets are torn from the wavy interface and may wet the upper part
may form with different spatial distributions of the gas/liquid in-
of the pipe, forming the annular flow configuration.
terface. The existing flow patterns depend primarily on the liquid
In horizontal and downward flow, the mechanism for transition
and gas flow rates, the physical properties of the two phases, and
from annular flow based on film instability never exists; therefore,
the geometry of the conduit 共inclination angle and diameter兲. A
the transition boundary is predicted by a spontaneous blockage
knowledge of the flow pattern in a pipe is essential to the design
mechanism. Thus, the new model based on a critical void fraction
engineer because the flow hydrodynamics and the transfer mecha-
works for the transition to intermittent flow and other boundaries
nisms of mass, momentum, and heat differ significantly from one
identifying the annular flow region are predicted by aforemen-
pattern to another.
tioned methods in the new unified model.
Two-phase gas-liquid flow in pipes exhibits five basic flow pat-
terns: stratified, bubble, slug, churn, and annular flows. Annular
flow occurs at high gas flow rates and low to medium liquid flow
rates in upward flow. In downward flow, it occurs at low to mod-
erate liquid flow rates and low to high gas flow rates. Annular
flow is characterized by the liquid flowing as a film around the
pipe wall, surrounding a high-velocity gas core, which may con-
tain entrained liquid droplets 共Fig. 1兲.
Several mechanisms and physical models associated with the
transition from annular flow have been proposed. Turner et al. 关1兴
and Ilobi and Ikoku 关2兴 studied the minimum gas velocity required
for liquid removal from vertical wells. Taitel et al. 关3兴 used the
same concept to predict the annular flow transition boundary. Tai-
tel and Dukler 关4兴 presented a model for predicting the transition
to annular flow in near-horizontal pipes that requires unstable
stratified flow with a liquid level that is insufficient to form the
complete bridge, required for a slug flow configuration. Barnea
关5兴 developed a unified annular flow transition model based on
two different mechanisms: instability of the film at low liquid flow
rates and spontaneous blockage of the gas core at high liquid flow
rates. The Barnea model was later modified by Ansari et al. 关6兴 to
account for liquid entrainment in the gas core and rhombohedral
packing of gas bubbles in the liquid slug.
All annular transition models are terminated by the stratified
flow transition boundary wherever this flow pattern exists. This
boundary can be predicted by the Taitel and Dukler model for

Contributed by the Petroleum Division and presented at the 20th Annual Energy-
Sources Technology Conference and Exhibition, Houston, Texas, February 1–3,
1999, of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manuscript re-
ceived by the Petroleum Division, October 24, 1998; revised manuscript received
January 4, 2000. Associate Technical Editor: W. P. Jepson. Fig. 1 Schematic diagram of annular flow

22 Õ Vol. 122, MARCH 2000 Copyright © 2000 by ASME Transactions of the ASME
 In this study, we briefly review the existing transition mecha-
nisms and propose a new unified model for predicting the transi-
tion from annular flow to intermittent flow for the entire range of
inclination angles. Comparisons of the different transition models
against flow pattern data are presented.
Review of Existing Transition Models
Several models have been proposed for the annular flow pattern
Spontaneous Blockage. At high liquid flow rates, a thick film
is formed, which eventually bridges the core, resulting in inter-
mittent flow 共slug or churn flows兲. This occurs when
H LF⬎0.24 (10)
Ansari et al. Model. Ansari et al. 关6兴 modified the Barnea
model to account for liquid entrainment in the gas core. Core
properties with liquid entrainment have been incorporated in the
definition of the Lockhart and Martinelli parameters: Y M and X M .
They are given as

冉 冊
transition. Here, a brief description and assessment of each model
is given.
Taitel et al. Model. For vertical pipes, Taitel et al. 关3兴
XM⫽ 冑 共 1⫺FE兲 2
f F 共 dp/dL 兲 SL
f SL 共 dp/dL 兲 SC
(11)

claimed that annular flow cannot exist unless the gas velocity in g sin ␪ 共 ␳ L ⫺ ␳ C 兲
the gas core is sufficient to lift the entrained droplets. They devel- Y M⫽ (12)
oped the following transition criterion based on a balance between 共 dp/dL 兲 SC
gravitational and drag forces acting on the largest liquid droplet: For the interfacial friction correlating parameter, Ansari et al.

␯ SG⫽3.1 冋 g ␴共 ␳ L⫺ ␳ G 兲
␳ 2g 册 1/4
(1)
found that a correlation developed by Wallis 关10兴 worked well for
thin liquid films, while the Whalley and Hewitt 关9兴 correlation
performed well for thick films. Thus
This criterion implies that the annular flow transition boundary Z⫽1⫹300␦គ for FE⬎0.9 (13)
is independent of superficial liquid velocity.
Barnea Model. Barnea 关5兴 attributed the annular flow transi-
tion to blockage of the gas core. She postulated that the blockage
Z⫽1⫹24 冉 冊
␳L
␳G
1/3
␦គ for FE⬍0.9 (14)
might result from two possible mechanisms, instability of the liq- The superficial friction gradients for the liquid and the gas core
uid film and spontaneous blockage. are given in Eqs. 共7兲 and 共8兲. In these equations, superficial core
Instability of the Liquid Film. At low liquid flow rates, the velocity, core density, and core viscosity are defined as follows:
instability occurs because of the low core shear stress, resulting in ␯ SC⫽FE␯ SL⫹ ␯ SG (15)
downward flow of the film and blockage of the gas core. Instabil-
ity of the film occurs when ␳ C ⫽ ␳ L ␭ LC⫹ ␳ G 共 1⫺␭ LC兲 (16)
2⫺1.5H LF ␮ C ⫽ ␮ L ␭ LC⫹ ␮ G 共 1⫺␭ LC兲 (17)
Y M⭓ X 2M (2)
3
H LF共 1⫺1.5H LF兲 ␭ LC is the no-slip liquid holdup caused by the entrained liquid in
H LF is the fraction of the pipe cross section that is occupied by the the homogeneous mixture of gas and liquid in the core with re-
liquid film and can be expressed in terms of the minimum dimen- spect to the core cross section, and is given by
sionless film thickness, ␦គ , as
␭ LC⫽1⫺FE␯ SL / 关 ␯ SG⫹FE␯ SL兴 (18)
H LF⫽4 ␦គ 共 1⫺ ␦គ 兲 (3)
The entrainment fraction, FE is given by Wallis 关10兴 as
␦គ can be obtained from the following combined momentum equa-
tion: FE⫽1⫺exp关 ⫺0.125共 10000␯ SG␮ G / ␴ 共 ␳ G / ␳ L 兲 0.5⫺1.5兲兴
(19)
Z X 2M The friction factors f F , f SL , and f SC can be obtained from a
Y M⫺ 2.5 ⫹ ⫽0 (4)
4 ␦គ 共 1⫺ ␦គ 兲关 1⫺4 ␦គ 共 1⫺ ␦គ 兲兴 关 4 ␦គ 共 1⫺ ␦គ 兲兴 3 Moody diagram for the corresponding Reynolds numbers given in
the following:
Y M and X M are the modified Lockhart and Martinelli 关8兴 param-
eters, expressed as N Re,F ⫽ ␳ L ␯ SL共 1⫺FE兲 D/ ␮ L (20)

XM⫽ 冑 共 dp/dL 兲 SL
共 dp/dL 兲 SC
(5)
N Re,SL⫽ ␳ L ␯ SLD/ ␮ L
N Re,SC⫽ ␳ C v SCD/ ␮ C
(21)
(22)
g sin ␪ 共 ␳ L ⫺ ␳ C 兲
Y M⫽ (6) The mechanism of bridging, based on the minimum liquid
共 dp/dL 兲 SC holdup to form a liquid slug, is modified for rhombohedral pack-
The superficial frictional pressure gradients for the liquid and ing of bubbles and liquid entrainment as follows:
the gas core are given, respectively, as
共 dp/dL 兲 SL⫽ f SL␳ L ␯ SL
2
/ 共 2D 兲 (7)
冉 H LF⫹␭ LC
AC
A
⬎0.12 冊 (23)

共 dp/dL 兲 SC⫽ f SC␳ C ␯ SC

2
/ 共 2D 兲
Z in Eq. 共4兲 is a correlating factor for interfacial friction and
film thickness. Barnea used the Whalley and Hewitt 关9兴 correla-
tion for predicting Z
Z⫽1⫹300␦គ
In the model, Barnea assumed no entrainment in the gas core.
Therefore, the superficial core velocity, ␯ SC is the superficial gas
velocity and the core density and viscosity are the gas density and
viscosity.
(8) Proposed New Model
In annular flow, decreasing the gas flow rate reduces the void
fraction, and a critical void fraction may be reached for which
liquid bridges may be formed by the crests of interfacial waves.
Under this condition, the transition from annular flow to intermit-
(9)
tent 共churn or slug兲 flow takes place. By comparing void fraction
data with theoretical predictions for annular flow and intermittent
flow, Wallis obtained a critical void fraction of 0.80. Barnea sug-
gested that when the void fraction is lower than 0.76, annular flow
will change into intermittent flow because of the liquid blockage

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 23

of the gas core. Chisholm 关11兴 correlated extensive data for void
fraction. He found that, up to a specific void fraction of about
0.75, the void fraction varies almost linearly with the gas volu-
冉
64X 2M ⫹Y M ⫽162.2 1⫺3
␯ SL
␯ SG 冊 2
(32)

metric flow fraction 共no-slip gas holdup兲. In their experimental where the Lockhart and Martinelli parameters, Y M and X M , are
study, Jepson and Taylor 关12兴 also observed that liquid film void the same as given in the Barnea model by Eqs. 共5兲 and 共6兲.
fraction tends to a limit of 0.76 as the gas volumetric flow fraction
is increased in intermittent flow. After the void fraction exceeds Comparison and Discussion of Transition Models
0.75, it increases exponentially with an increase in gas volumetric
flow fraction. This can be interpreted as follows. When the void Performance of Models for Low-Pressure Systems. Figures
fraction is less than 0.75, liquid is the continuous phase, and gas is 2–10 compare the proposed new model and previous transition
distributed in the continuous liquid phase, whether the two-phase models with the Shoham 关14兴 data for air-water flow at 1.0 bar
flow is in bubbly flow, slug flow, or churn flow. Also, the void and 25°C in 5.1-cm-dia pipes having different inclination angles
fraction of these gas discontinuous flow patterns has almost the
same profile versus the gas volumetric flow fraction. When the
void fraction is greater than 0.75, annular flow occurs in the pipe,
and both the gas and liquid phases are continuous. The void frac-
tion profile of annular flow versus the gas volumetric flow fraction
is totally different from that of the gas discontinuous flow pat-
terns. Therefore, after Chen 关13兴, the transition criterion from an-
nular flow is set at
␣ ⭐0.75 (24)
Since it is normally assumed that the gas phase in intermittent
flow does not have entrained liquid droplets, the no entrainment
assumption can also be made at the transition from annular to
intermittent flow. Chen used this assumption and combined con-
servation of momentum in the gas core and the liquid film, which
yields

␶ is i 冉 1
⫹
1
AG AL 冊SL
⫺ ␶ L ⫺ 共 ␳ L ⫺ ␳ G 兲 g sin ␪ ⫽0
AL
(25)
Fig. 2 Comparison of annular flow transition models „90 deg,
5.1 cm pipe…
In Eq. 共25兲, ␶ i is the interfacial shear stress, ␶ L is the shear stress
at the pipe wall, A G and A L are the cross-sectional areas occupied
by the gas core and liquid film, respectively, and S i and S L are the
perimeters upon which ␶ i and ␶ L act, respectively.
The shear stresses are estimated in the conventional manner

␳ L ␯ L2 ␳ G共 ␯ G⫺ ␯ L 兲 2
␶ L⫽ f L , ␶ i⫽ f i (26)
2 2
The liquid/wall friction factor, f L , is evaluated from

f L ⫽C L 冉 ␳ LD L␯ L
␮L 冊 ⫺n
(27)

where D L ⫽4A L /S L is the hydraulic diameter of the liquid film.

Wallis proposed the following simple relation for evaluation of
the interfacial fraction factor:
Fig. 3 Comparison of annular flow transition models „70 deg,
f i ⫽ f G 关 1⫹75共 1⫺ ␣ 兲兴 (28) 5.1 cm pipe…
where the gas phase friction factor, f G , in the absence of the
liquid film is evaluated from:

f G ⫽C G 冉 ␳ L D G ␯ SG
␮G 冊 ⫺m
(29)

In the present work, the following coefficients in the friction fac-

tor correlation are used: C G ⫽C L ⫽0.046, m⫽n⫽0.20 for turbu-
lent flow and C G ⫽C L ⫽16, m⫽n⫽1 for laminar flow. The con-
figuration of the liquid film can be given as follows:
S i ⫽ ␲ D ␣ 0.5, S L⫽ ␲ D (30)

␲D2 ␲D2
A G⫽ ␣ , A L ⫽ 共 1⫺ ␣ 兲 (31)
4 4
Upon substitution of Eqs. 共26兲–共31兲 and the transition criterion,
Eq. 共24兲, into Eq. 共25兲, the following transition boundary equation Fig. 4 Comparison of annular flow transition models „30 deg,
can be obtained: 5.1 cm pipe…

24 Õ Vol. 122, MARCH 2000 Transactions of the ASME


Fig. 5 Comparison of annular flow transition models „1 deg,
5.1 cm pipe…
Fig. 6 Comparison of annular flow transition models „0 deg,
5.1 cm pipe…
Fig. 7 Comparison of annular flow transition models „À1 deg,
5.1 cm pipe…
for upward and downward directions. Figures 11–13 compare the
transition models with Shoham’s horizontal and vertical upflow
and downflow data in 2.54-cm pipe. Although Barnea’s 关5兴 spon-
taneous blockage criterion gives satisfactory predictions with the
correct trend, the film stability criterion displays an incorrect trend
of decreasing superficial gas velocity with increasing superficial
liquid velocity over part of the transition boundary. In Fig. 5, for
Journal of Energy Resources Technology
Fig. 8 Comparison of annular flow transition models „À30
deg, 5.1 cm pipe…
Fig. 9 Comparison of annular flow transition models „À80
deg, 5.1 cm pipe…
Fig. 10 Comparison of annular flow transition models „À90
deg, 5.1 cm pipe…
1 deg inclination angle, all models show inaccurate results with
respect to the data. In horizontal and near horizontal pipes, par-
ticularly for the low liquid flow rates, the poor performance of the
spontaneous blockage or critical void fraction mechanisms is not
surprising due to the low liquid supply. The mechanisms for the
Ansari et al. model and the Barnea model are almost identical.
However, the blockage criterion in the Ansari et al. model has
MARCH 2000, Vol. 122 Õ 25
Fig. 11 Comparison of annular flow transition models „0 deg,
2.54 cm pipe… Fig. 14 New model annular flow transition boundaries for dif-
ferent inclination angles „5.1 cm pipe…

Fig. 12 Comparison of annular flow transition models „À90
eter on the transition from annular flow. The new model also
displays the physically correct trend for both low and high liquid
flow rates.
Performance of Models for High-Pressure Systems. No
published flow pattern data for high-pressure, high-temperature
共HPHT兲 wellbore conditions could be found in the open literature.
The only way to evaluate transition models for HPHT conditions
is to test them against the expanded Tulsa University Fluid Flow
Projects 共TUFFP兲 well data bank and compare their pressure drop
results. The expanded TUFFP well data bank consists of 2052
well cases from various sources covering a wide range of flow
parameters.
Comparison Criteria. A variety of statistical parameters are
used to evaluate the model predictions using the entire data bank.
Following are the definitions of the statistical parameters used.
Average Percentage Error

冋 册
deg, 2.54 cm pipe… N
1
E1⫽
N 兺e
i⫽1
ri ⫻100 (33)

where
e ri ⫽ 共 ⌬ P i,calc⫺⌬ P i,mean兲 /⌬ P i,meas (34)
and N is the number of well cases that successfully converged. E1
indicates the overall trend of the performance relative to the mea-
sured pressure drop.
Absolute Average Percentage Error

E2⫽
1
N 冋 兺 冏 冏册
N

i⫽1
e ri ⫻100 (35)

The absolute average percentage error will eliminate the canceling

effect of E1. This parameter indicates the general percentage error
of the calculations.
Standard Deviation
N
Fig. 13 Comparison of annular flow transition models „90 deg,
2.54 cm pipe…
E3⫽ 兺 冑共 e
i⫽1
ri ⫺E1 兲
2
/ 共 N⫺1 兲 (36)

The standard deviation indicates the scatter of the relative error

been modified to incorporate liquid entrainment. Nevertheless, the
same trend is observed for both the Ansari et al. and the Barnea
models.
Figure 14 shows the sensitivity of the new model for different
inclination angles in 5.1-cm-dia pipe. The new model properly
represents the influences of both liquid flow rate and pipe diam-
about the average error.
The statistical parameters described in the foregoing are based
on the relative pressure drop error, rather than the actual pressure
drop error. Relatively small pressure drop errors in wells that
experience a small pressure drop may give a large percentage
error, even though the predicted pressure drop is not that far from
the actual measurement. To make the statistics independent of the

26 Õ Vol. 122, MARCH 2000 Transactions of the ASME

Table 1 Results of transition models’ pressure drop calcula- E3 ⫽ standard deviation of average error
tions against TUFFP well data bank E4 ⫽ average error
E5 ⫽ absolute average error
E6 ⫽ standard deviation of average error
f ⫽ friction factor
F ⫽ film
FE ⫽ fraction of liquid entrained in gas core
g ⫽ gravity acceleration
H ⫽ average holdup fraction
HF ⫽ hydraulic film
L ⫽ length
N ⫽ number of converged well cases
N Re ⫽ Reynolds no.
p ⫽ pressure
RPF ⫽ relative performance factor
S ⫽ wetted perimeter
v ⫽ velocity
X ⫽ Lockhart and Martinelli parameter
magnitude of the relative pressure drop, a set of statistical param- Y ⫽ Lockhart and Martinelli parameter
eters based on the actual pressure drop is used, defined as Z ⫽ correlating factor for interfacial friction
e i ⫽⌬ P i,calc⫺⌬ P i,meas (37) ␣ ⫽ void fraction
␦គ ⫽ ratio of film thickness to diameter
If we replace e ri with e i in Eqs. 共33兲, 共35兲, and 共36兲, the new ␭ ⫽ no-slip holdup fraction
parameters can be defined as ␮ ⫽ dynamic viscosity
N ␪ ⫽ angle from horizontal
1
E4⫽
N 兺e
i⫽1
i (38)
␴
␳ ⫽
⫽
density
surface tension
N
␶ ⫽ shear stress
1
E5⫽
N 兺 兩e 兩i (39) Subscripts and Superscripts
i⫽1
C ⫽ core

兺冑
N calc ⫽ calculated
共 e i ⫺E4 兲 2 F ⫽ film
E6⫽ (40)
i⫽1 N⫺1 G ⫽ gas
L ⫽ liquid
The performances of the new model and the Ansari et al. and
LC ⫽ gas core with liquid entrainment
Barnea transition models are compared. For these transition mod-
LF ⫽ liquid film
els, the Ansari et al. hydrodynamic model has been used. Table 1
M ⫽ mixture
reports the errors in predicting pressure drop calculations for those
meas ⫽ measured
well cases in the TUFFP well data bank for which annular flow is
mn ⫽ minimum
predicted to occur in the entire wellbore for the tested transition
n,m ⫽ power constant values for friction factor
model. Table 1 shows that the new model and the Barnea model
S ⫽ superficial
give higher error values than the Ansari et al. model. The Ansari
SC ⫽ superficial core
et al. model predicts that only 71 well cases have 100 percent
SL ⫽ superficial liquid
annular flow in the wellbore, while the new model and the Barnea
SG ⫽ superficial gas
model predict 162 and 206 well cases, respectively, with 100 per-
t ⫽ total
cent annular flow.

Conclusions
A new unified model for the annular flow transition is devel-
oped based on a critical void fraction mechanism. Available an- References
nular flow transition models and a new model are evaluated using 关1兴 Turner, R. G., Hubbord, M. G., and Dukler, A. E., 1969, ‘‘Analysis and Pre-
diction of Minimum Flow Rate for the Continuous Removal of Liquids From
the TUFFP well data bank for high pressures and Shoham’s flow Gas Wells,’’ J. Pet. Technol., Nov., 1475-82, p. 246.
pattern data for low pressures. Comparisons show that, for low- 关2兴 Ilobi, M. I., and Ikoku, C. U., 1981, ‘‘Minimum Gas Flow Rate for Continuous
pressure and temperature flow systems, the new model gives the Liquid Removal in Gas Wells’’ Paper SPE 10170 presented at SPE Annual
best results, while Ansari et al. model gives the best results for Technical Conference and Exhibition, San Antonio, TX, October 4–7.
关3兴 Taitel, Y., Barnea, D., and Dukler, A. E., 1980, ‘‘Modeling Flow Pattern
high pressure and temperature systems. Transitions for Steady State Upward Gas-Liquid Flow in Vertical Tubes,’’
AIChE. J., 26, pp. 345–354.
Acknowledgments 关4兴 Taitel, Y., and Dukler, A. E., 1976, ‘‘A Model For Prediction Flow Regime
Transition in Horizontal and Near Horizontal Gas-Liquid Flow,’’ AIChE. J.,
We thank the TUFFP member companies whose membership 22, pp. 47–55.
fees were used to fund part of this research project. 关5兴 Barnea, D., 1987, ‘‘A Unified Model for Predicting Flow-Pattern Transition
for the Whole Range of Pipe Inclinations,’’ Int. J. Multiphase Flow, 13, pp.
1–12.
Nomenclature 关6兴 Ansari, A. M., Sylvester, N. D., Sarica, C., Shoham, O., and Brill, J. P., 1994,
‘‘A Comprehensive Mechanistic Model for Upward Flow in Pipes,’’ SPE
A ⫽ cross-sectional area Prod. Facil., Trans. AIME, 297, May, pp. 217–226.
C ⫽ coefficient 关7兴 Barnea, D., Shoham, O., and Taitel, Y., 1982, ‘‘Flow Pattern Transitions for
D ⫽ pipe diameter Downward Inclined Two-Phase Flow; Horizontal to Vertical,’’ Chem. Eng.
Sci., 37, pp. 735–740.
e ⫽ error function 关8兴 Lockhart, R. W., and Martinelli, R. C., 1949, ‘‘Proposed Correlation of Data
E1 ⫽ average percentage error for Isothermal Two-Phase Two-Component Flow in Pipes,’’ Chem. Eng.
E2 ⫽ absolute average percentage error Prog., 45, Jan., No. 1, pp. 39–48.

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 27

 关9兴 Whalley, P. B., and Hewitt, G. F., 1978, ‘‘The Correlation of Liquid Entrain-
ment Fraction and Entrainment Rate in Annular Two-Phase Flow,’’ UKAEA
Report, AERE-R9187, Harwell.
关10兴 Wallis, G. B., 1969, One-Dimensional Two-Phase Flow, McGraw-Hill, New
York, NY.
关11兴 Chisholm, D., 1983, ‘‘Two Phase Flow in Pipelines and Heater Changers,’’
Charge Todwin, London, UK.
28 Õ Vol. 122, MARCH 2000
关12兴 Jepson, W. P., and Taylor, R. E., 1993, ‘‘Slug Flow and its Transition in
Large-Diameter Horizontal Pipes,’’ Int. J. Multiphase Flow, 19, pp. 411–420.
关13兴 Chen, X. T., 1994, ‘‘A Model for Transition From Annular Flow,’’ TUFFP
ABM Slide Copy, September 2.
关14兴 Shoham, O., 1982, ‘‘Flow Pattern Transitions and Characterization in Gas-
Liquid Two Phase Flow in Inclined Pipes,’’ Ph.D. dissertation, Tel-Aviv
Univ., Ramat-Aviv, Israel.
Transactions of the ASME

Mamdouh M. Salama
Conoco Inc.,
P.O. Box 1267,
Ponca City, OK 74603
Sand Production Management
Sand production may be inevitable in many fields that have a relatively low formation
strength. Sand erosion and settling predictions and sand monitoring are important ele-
ments of any effective sand production management strategy. Sand erosion predictions
are used to establish tolerable sand production rates, and, thus, well productivity, and to
develop cost-effective inspection frequency for critical components. Prediction of critical
flow rate to prevent sand settling is important for flowlines that are not designed for
pigging. Quantitative sand monitoring is essential in verifying the effectiveness of sand
control procedures and in generating an important input parameter for erosion and sand
settling predictions. This paper presents equations for predicting sand erosion rate and
sand settling flow rate, and assesses the accuracy of these equations. In addition, the
paper presents an assessment of the sensitivity of commercially available nonintrusive
acoustic and intrusive electrical resistance sand monitors. 关S0195-0738共00兲00201-6兴

Introduction Sand Erosion

Sand production may be inevitable in many fields that have a
relatively low formation strength 共⬍1000 psi兲. If sand production
is expected at the early life of the field, it will be prevented by
installation of downhole sand exclusion systems such as gravel
packs and screens. However, as would be the case for many res-
ervoirs, wells will not produce sand when they are first brought to
production; but they will produce sand at some later point in their
lives. In such a case, the operator has three options: 共1兲 complete
the wells with downhole sand exclusion systems from day one, 共2兲
recomplete the well by installing sand exclusion systems at a later
date when sand production begins; or 共3兲 do not install downhole
sand exclusion systems and manage sand production by designing
the facilities to handle sand if sand is produced. The last option,
sand production management, is preferred in many cases because
downhole sand exclusion systems can increase the risk of loss of
production or even the complete well by plugging and mechanical
damage.
There are three technical issues that are critical to sand produc-
tion management that need to be addressed in order to maximize
reservoir production while maintaining the integrity of the produc-
tion facility under conditions of sand production. These are sand
erosion, sand settling, and sand monitoring. Sand erosion predic-
tion models are necessary to establish tolerable sand production
rates, and, thus, allowable well flow rates, and to determine in-
spection strategy for the production facility. This will allow estab-
lishing the optimum production rate while maintaining acceptable
equipment integrity and safety levels. The second technical issue,
which is of paramount importance in sizing and operations of
infield flowlines and pipelines, is sand settling. In situations where
sand settling cannot be avoided and routine pigging is not fea-
sible, which may be the case for flowlines connecting subsea
wells and unmanned platforms, assessing corrosion under sand
beds is an important consideration. All analyses addressing the
effect of sand on reservoir management and equipment integrity
require quantifying sand production rates that may be feasible
using sand monitors. This paper will review sand erosion and
settling prediction models and present an assessment of the quan-
titative accuracy of commercially available sand monitors. Quan-
titative, rather than qualitative, accuracy of sand monitors is re-
quired to determine the effectiveness of sand control procedures,
and to provide input to predict sand erosion rate of critical com-
ponents and sand settling in flowlines.
Contributed by the Petroleum Division and presented at the Offshore Technology
Conference, Houston, Texas, May 4–7, 1998, of THE AMERICAN SOCIETY OF ME-
CHANICAL ENGINEERS. Manuscript received by the Petroleum Division, March 23,
1999; revised manuscript received December 1, 1999. Associate Technical Editor: C.
Sarica.
Unlike erosion in sand-free systems where erosion rate is re-
lated to two parameters, i.e., mixture density and flow velocity,
erosion due to sand is influenced by several factors including fluid
characteristics 共flow rate, composition, density, and viscosity兲,
sand characteristics 共concentration, impact velocity, impact angle,
number of particles hitting the surface, shape/sharpness, hardness,
size distribution, and density兲, component geometry 共bend, tee,
choke, and joint兲, and material properties 共hardness and micro-
structure兲. There exists an extensive database that can be used to
characterize erosion rates of different materials. These data are
generally presented using the following equation:
Er⫽AV np F 共 ␣ 兲 (1)
where Er is erosion rate measured as the ratio between the mass of
metal loss and the mass of sand hitting the target material. The
amount of sand hitting the target material is influenced by the flow
conditions, sand concentration, and the geometry of the compo-
nent. A and n are experimentally determined constants that depend
on the material properties. For ductile materials, the value of n is
in the range 2–3. For brittle materials, n can be as high as 6. V p is
the impact velocity of the sand particle on the metal surface. F( ␣ )
is a function whose value varies between 0 and 1, depending on
the impact angle. The function depends on the target material
ductile/brittle behavior. The value of F( ␣ ) is maximum for duc-
tile materials such as steel at impact angles of 20 to 40 deg, and
for brittle materials such as ceramics at 90 deg.
The difficulty in calculating erosion rate is in predicting the
proper values of particle impact angle, ␣, and velocity, V p , whose
values depend on: fluid density, fluid viscosity, sand particle di-
ameter, sand density, pipe diameter, and pipe geometry 共elbow,
tee, choke, etc.兲. One can account for these factors in predicting
particle impact angle and velocity through the use of particle
tracking simulation models.
There are six models that have been developed within the in-
dustry for predicting sand erosion in piping systems. These mod-
els are based on work done by Salama and Venkatech 关1兴 of
Conoco, Svendeman and Arnold 关2兴 of SouthWest Research,
Shirazi et al. 关3,4兴 of Tulsa University, Birchenough and Dawson
关5兴 of AEA, and Salama 关6兴 of Conoco. All models are, however,
limited to erosion predictions in simple pipe geometries such as
pipe bends and tees. A review of these models is provided in the
paper by Salama 关6兴. In this paper, Salama proposed the following
simple model for predicting sand erosion in pipe bends:
2
1 WV m d
ER⫽ (2)
Sm D ␳m
2

Journal of Energy Resources Technology Copyright © 2000 by ASME MARCH 2000, Vol. 122 Õ 29
 avoid excessive wear to the bottom of the pipe. In this case, sand
settling velocity is defined as the minimum velocity to keep solid
particles suspended in the flow, and thus avoid their drop-out. For
oil and gas transport, sand settling flow rate can be defined by the
transition between scouring and moving dunes 共i.e., sand is on the
bottom of the pipe, but moving along the pipe兲. The flow velocity
at this condition would be lower than the velocity to disperse the
sand, i.e., keep all the sand in suspension. Since the settling ve-
locity is defined as the transition between sand settling and sand
transport, the transport and settling velocities are almost the same.
This is not true for cases when the bottom sand layer has been
stationary for some time. A higher flow rate will be required to
initiate the movement of such a layer because sand tends, with
time, to become more compacted and more adherent to itself and
Fig. 1 Comparison between predicted and measured sand
erosion rate in bends
to the pipe surface. This level of flow rate increase is not well
established. Therefore, it is recommended that under conditions of
sand production, the periods of shutdown and production below
the sand settling flow rate should be kept to a minimum. Also, if
where ER is erosion rate in mm/yr. W is sand flow rate in kg/d.
the flowlines are operated for an extended period of time under
V m is fluid flow velocity in m/s. d is the sand size in micron
conditions below the sand settling flow rate, care must be taken
共effect of d becomes negligible above 400 microns; therefore, for
when the flow rate is increased to avoid the sudden transport of
d⬎400, the limit of 400 is used兲. D is pipe internal diameter in
large quantities of sand.
mm. ␳ m is fluid mixture density in kg/m3. S m is geometry factor
Prediction models for solid transport in pipelines are based on
that equals 5.5 for pipe bends.
two main approaches. The first approach is based on deriving an
The accuracy of this model 共Eq. 共1兲兲 is clearly illustrated by
expression for the forces necessary to roll a solid particle by in-
Fig. 1. Figure 1 presents a comparison between the model predic-
vestigating the forces acting on a single particle 共i.e., gravity,
tions and multi-phase flowloop erosion measurements on pipe
buoyancy, lift, and drag forces兲. This approach was used by
bends. The experimental data used in this comparison were gen-
Wicks 关11兴 to develop his model for predicting sand transport in a
erated by Weiner and Tolle 关7兴, Tolle and Greenwood 关8兴, Bour-
single-phase flow. Wicks’ model does not account for the effect of
goyne 关9兴, and Salama 关6兴. The details of the testing conditions
particle diameter on settling characteristics. Holte et al. 关12兴
are summarized by Salama 关6兴. The main data were developed
modified Wicks’ model to account for the effect of particle diam-
using a 1-in. medium pressure flowloop. The test section consisted
eter as reflected by experimental data. Holte et al. 关12兴 also ex-
of three parts: one inlet pipe section for simulation of erosion in a
tended Wicks model to two-phase flow and to account for tilted
vertical pipe, a 5-D pipe bend, and a horizontal outlet pipe section
pipes. Using the modified model, Anglesen et al. 关10兴 developed a
similar to the one in the vertical pipe. All pipes were fabricated of
sand settling subroutine which was integrated within DNV’s flow
duplex stainless steel. The test specimens were assembled ensur-
model and corrosion prediction package 共Corroline兲. The second
ing a smooth transition between the various pipe elements. All
approach is based on turbulence theory by considering the energy
tests were conducted using distilled water and nitrogen. Two sand
dissipated from turbulent eddies. Based on a detailed regression
sizes were used in the tests, 150 and 250 micron, and the sand was
analysis of more than 350 experiments, Oroskar and Turian 关13兴
not recirculated through the test. The sand was injected into the
derived a sand settling prediction equation, which was also veri-
test section via the liquid flow downstream of the liquid pump and
fied by Davies 关14兴 using simple turbulence theory. Both modified
was collected in a separator downstream of the test section.
Wicks 共Wicks/Holte兲, and Oroskar and Turian models can be
written in the following form:
Sand Settling
Below some minimum flow velocity in a horizontal pipeline,
solid particles in the fluid can form a bed on the bottom of the
line. Deposition of the solids can lead to partial or complete
V m ⫽Kd n 1 ␯ n 2 冉 冊
⌬␳
␳f
n3
D n4 (3)

blockage of flowlines, enhanced pipe bottom corrosion, and trap- where

ping of pigs. If small amounts of sand are settled in a pipeline, Vm ⫽ settling velocity
their removal can be relatively easy by pigging or by increasing d ⫽ particle diameter
the flow velocity above the sand settling flow rate. If, however, D ⫽ pipe diameter
large quantities of sand are deposited in a pipeline, their removal ⌬␳ ⫽ density difference between particles and fluid
can be both difficult and time-consuming. ␳f ⫽ fluid density
Sand transport in horizontal pipelines has four main patterns, n1–4 ⫽ constants
depending on the fluid flow rate 关10兴. As the flow velocity de- K ⫽ constant which is function of flow condition for two-
creases, the pattern of sand changes from dispersed to scouring phase flow
and from moving bed to stationary sand bed. A stationary sand Because of the similarity of the two models, a correlation was
bed is a stable bed with stationary sand particles at the bottom. developed to predict sand settling in single and two-phase flows
When a stable bed height is reached, the particles in the top are using Oroskar and Turian 关13兴 exponents. The constant 共K兲 was
transported downstream, increasing the sand bed length. As the derived based on correlations with both experimental data and
fluid velocity increases, the bed breaks up into characteristic slow- predictions made by the DNV Corroline software for sand settling
moving sand dunes with sand particles transported from the rear in both single and two-phase flows. The proposed model can be
to the front of the dunes. As the flow velocity increases further, presented by the following equation:
scouring of the sand along the bottom of the pipe occurs with
most particles moving along the pipe wall. When the flow velocity
increases further, sand becomes dispersed in the liquid phase and
in the gas phase in multi-phase flow pipes.
V m⫽ 冉 冊
V sl
Vm
0.53
d 0.17␯ 0.09 冉 冊
⌬␳
␳f
0.55
D 0.47 (4)

Depending on the service conditions, any transition between where

these patterns may be considered critical. For slurry transport, the V m ⫽ minimum mixture flow velocity to avoid sand settling,
transition between scouring and dispersed is considered critical to m/s

30 Õ Vol. 122, MARCH 2000 Transactions of the ASME

 Fig. 3 Pitting under sand bed
Fig. 2 Comparison between predicted sand settling velocity
using the proposed model and experimental data

corrosion due to depletion of inhibitors or formation of biological

cells under sand bed. Tests that were conducted to evaluate the
V sl /V m ⫽ ratio between liquid superficial velocity and mixture
effect of sand settling on inhibitor effectiveness suggest that the
velocity at sand settling condition 共for single phase,
presence of sand does not affect corrosion rates and inhibitors can
ratio is 1兲
be effective in controlling uniform corrosion under sand beds.
d ⫽ particle diameter, m
However, the results raised a concern that some inhibitors may
D ⫽ pipe diameter, m
not prevent pitting formation under a sand bed, as shown in Fig. 3.
⌬␳ ⫽ density difference between particles and liquid, kg/m3
␳ f ⫽ liquid density, kg/m3
Sand Monitors
␯ ⫽ kinematic viscosity, m2/s
Sand monitors are required to determine the effectiveness of
The accuracy of Eq. 共4兲 is demonstrated by Fig. 2, which com-
sand control procedures and to provide input to predict erosion
pares predictions using the proposed model 共Eq. 共4兲兲 with experi-
rates of critical components. Currently available sand monitors
mental data. The experimental data were generated using a multi-
can be classified as intrusive and nonintrusive. The probes of in-
phase flow loop. The test section consists of a 12-m-long 4-in.
trusive monitors penetrate the component wall, while the probes
(i.d.⫽108.2 mm) duplex stainless steel pipe. For observations of
of nonintrusive monitors are mounted on the outside wall of the
the sand behavior, two transparent sections, each of length 500
component. The two basic principles for detecting sand are based
mm, were installed at the outlet of the 4-in. pipe. The sand was
on either measuring the noise generated by sand impact on a com-
injected at the inlet to the horizontal test section approximately 10
ponent wall or an intrusive probe, or by measuring a material loss
m upstream of where the observations were made. More than 50
due to sand erosion of a probe placed in the flow stream or of the
test conditions were evaluated covering the following main pa-
component wall. Intrusive monitors include CorrOcean electrical
rameters:
resistance 共ER兲 monitor, Cormon ER probe, Schlumberger Sandec
Sand particle size: 100, 280, 500 ␮m
acoustic monitor, and Sandia acoustic probe. The nonintrusive
Media: water, gas 共CO2, N2, air兲, oil, inhibitors, sand
monitors include acoustic probes marketed by Fluenta, ClampOn,
Water cut: 1, 10, 50, 100 percent
and Simrad. Sand may also be monitored by measuring the reduc-
Pressures: 4 and 8 bara
tion of pipe wall using ultrasonic, field signature analysis 共FSM兲,
Temperature: ambient
and thin layer activation 共TLA兲 methods. This approach becomes
Superficial liquid velocity: 0.03, 0.1, 0.2, 0.3, 0.4 m/s
complicated if metal loss can also be attributed to corrosion. The
Gas velocity: flow rate varied during tests
fundamental limitation with all commercially available sand
During testing, the superficial liquid velocity as well as the
monitors 共acoustic and ER兲, is that none of them directly measure
water cut was kept constant, while the superficial gas velocity was
the total amount of sand in the system; they only measure the
reduced stepwise. At each condition, the sand behavior and tran-
effect of the sand that impacts the pipe or the probe surface. There
sitions between each sand flow regime were observed. The mea-
are also several other methods that are expensive, but have the
sured scouring-moving dunes transition velocities were compared
potential of measuring sand directly. These methods include
with predictions using the model proposed by Eq. 共4兲 as shown in
Lasentec particle characterization monitor, flow vision particle
Fig. 2. The comparison shows that the model predicts the ob-
analyzer, and sand filters.
served transition fairly well.
For acoustic sand monitors, the acoustic transducer signal 共nV兲
The foregoing model can be used for horizontal and near-
is generated in response to the sand impact noise 共dB兲 which is
horizontal pipes; but for vertical pipes, a model developed by
integrated over a fixed time interval 共e.g., 1 s兲, and is compared to
Chein 关15兴 can be used. Using the same units as in Eq. 共4兲.
a predetermined value of a solid-free flow. The excess signal is
Chein’s model can be written as follows:
converted to sand rate using a calibration curve that is empirically

V m ⫽1.2 冉 冊冋 冑
␯
d
⫺1⫹ 1⫹0.073 冉 冊冉 冊 册
⌬␳
␳f
d
␯
2
(5)
developed and calibrated by injecting known amounts of sand
upstream of the sensor and measuring the transducer output sig-
nal. Therefore, the optimum monitor location is where sand im-
The foregoing models ignore the impact of condensate and added pact on the internal pipe wall is greatest, such as on the outside
chemicals on sand behavior and sand settling predictions. It is, curve of the first bend downstream of a well. The sensitivity of
however, expected that Eq. 共4兲 can lead to conservative results acoustic detectors can be influenced by factors that prevent sand
because oil-wetted sand should be expected to settle at a lower particles from striking the pipe wall such as sand settling and
velocity than water wetted sand. This situation may be reversed, buildup of deposits on the pipe wall. The sensitivity can also be
as previously discussed, when predicting sand lifting flow rates. influenced by interference from other noise sources such as hy-
In cases when the fluid composition is corrosive, an effective drate crystals, liquid particles, opening and closing valves, and
inhibition program is implemented when carbon steel pipes are operation of pumps and compressors.
used. The effect of sand beds on the effectiveness of inhibitors As part of an effort to evaluate the accuracy of several of these
should be an issue of concern because of the possibility of high commercially available monitors, they were tested on two test

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 31

Fig. 4 A comparison between injected sand and predictions Fig. 5 The effect of gas velocity of the sensitivity of acoustic
by acoustic sand monitors sand monitors

ERD 2 ␳ m
loops. Some of the tests were conducted in a 6-in. horizontal M ⫽S m 2 (6)
Vm dp
flowloop using single phase 共air or water兲, and two-phase 共air and
water兲. Acoustic detectors were mounted on 1 1/2D bends located where
at least 14 m of straight pipe from the sand injection point. The ER ⫽ reduction in probe element thickness due to erosion, nm
ER probes were located at 10D downstream of the bend. Other M ⫽ amount of flowing sand, kg
tests were conducted in a circulating flowloop using a 3-in. test dp ⫽ sand particle diameter, micron
section that contained a blinded Tee-bend, a horizontal pipe, an D ⫽ pipe diameter, mm
ordinary bend, and a vertical downward pipe. The tests were per- Vm ⫽ fluid mixture velocity, m/s
formed at a pressure of 8 bara and ambient temperature. The ER ␳m ⫽ fluid mixture density, kg/m3
probes were installed in the horizontal pipe downstream of the Sm ⫽ geometry-dependant constant
Tee-bend and in the vertical pipe downstream of the ordinary Equation 共6兲 is the same as Eq. 共2兲 that was proposed for pre-
bend. The acoustic probes were installed on the vertical pipe dicting erosion in pipe bends. The correlation between erosion
downstream of the ordinary bend. predictions of the ER probe elements using Eq. 共6兲 and the erosion
Figure 4 presents a comparison between the predictions of measured on the ER probe elements is shown in Fig. 5. Since ER
acoustic monitors and actual sand injected in a flowloop under the sand monitors are intended for use in long-term monitoring, it is
different test conditions of sand size, sand injection rate, and gas- more appropriate to compare the accuracy of both the monitor and
liquid ratios. To further understand the scatter in the data, the the calibration equation based on the cumulative sand production,
prediction accuracy was analyzed in terms of the gas velocity, as as shown in Fig. 6. Figure 6 compares the cumulative sand in-
shown in Fig. 5. The results show that the predictions greatly jected in the flow loop and the predictions using Eq. 共6兲 based on
improve for gas velocities greater than 15 m/s. the measurement of the probe erosion rate and selecting the value
Electrical resistance 共ER兲 sand probes, such as those produced of S m by calibration with the result of a single test. The value
by CorrOcean, monitor the change in the electrical resistance of derived in this set of experiments was 1/325. The value of the
sensing elements due to reduction in the element thickness by constant S m depends on the geometry of the piping system and the
sand erosion. Each probe has two to four sensing elements 共20– location of the sand probe. In field operation, the value of the
500 micron thick兲 manufactured of a corrosion resistance alloy constant S m can be determined using computational fluid dynam-
film to eliminate the possibility of material loss due to corrosion. ics 共CFD兲 simulations of the actual piping geometry, including the
The thickness of the sensing element is determined by passing a probe.
fixed current and measuring the resultant voltage drop. To com-
pensate for resistance change due to temperature variations, a ref-
erence element is installed on the backside of the probe holder so
that it will be shielded from the flow. Since element resistances
are low, inherent electrical noise generated by the system elec-
tronics can result in random fluctuations in the measured thickness
of the element. Typical element thickness variations due to elec-
trical noise, under conditions of no flow, for the 35 and 100-
micron-thick elements are ⫾4 and ⫾30 nm, respectively. These
limits can be translated into sand production resolution, i.e., mini-
mum quantity of sand required to obtain a significant response
from the detector. While thinner elements are desirable because
they are more sensitive because of their higher resistance than
thicker elements, they will have shorter lives.
The relationship between sand flow rate and reduction in ele-
ment thickness is based on an empirical calibration equation. The
accuracy of sand predictions depend on the accuracy of the sens-
ing element thickness measurement, the accuracy of the calibra-
tion equation, the accuracy of the input parameters, and the uni-
formity of sand distribution upstream of the probe. The empirical
equation relating erosion of probe elements to amount of sand is Fig. 6 A comparison between ER probe measurements and
given as follows: predictions

32 Õ Vol. 122, MARCH 2000 Transactions of the ASME


Fig. 7 Unexplained response of ER and acoustic sand moni-
tors under no-flow condition
The foregoing results clearly demonstrate that accurate quanti-
tative predictions of sand can be obtained using both acoustic and
ER probes. It is, however, the author’s view that acoustic probes
are more suitable for short-term assessment of sand production,
such as well testing, while ER probes are more suitable for long-
term monitoring. ER probes offer the advantage of being able to
validate the accuracy of sand production prediction by physically
examining the amount of erosion on the probe elements. In addi-
tion, erosion measurements from the ER probe elements can be
used to assess erosion at different locations in the piping system
by comparing the values of the constant S m for the probe with the
values for the different pipe locations. In spite of the advances
being made to filter out extraneous noises that may be wrongly
interpreted as sand production, unexplained noise signals could
sometimes be detected. Figure 7 presents unexplained responses
that were measured by both ER 共top兲 and acoustic sand monitors
共bottom兲 at a time when the flowloop and all neighboring equip-
ment were shut down. This signal, which is indicated by a reduc-
tion in the element thickness of ER probes 共top兲 and an acoustic
signal in the acoustic probe 共bottom兲, would have been wrongly
interpreted as sand production if it occurred when the flow veloc-
ity was not zero. Although the noise was registered in the 100-
micron ER probe, it was not registered by the 35-micron probe,
suggesting that thinner ER elements are less sensitive to spurious
noise.
Conclusions and Recommendations
1 Safe operations and proper equipment integrity can be
achieved under conditions of sand production, providing the tech-
nical issues of sand erosion, sand settling, and sand monitoring are
properly addressed.
2 For sand-laden fluids, erosion allowances and maximum flow
rate limit can be established using the following equation:
2 Flow, Paper D2, BHRA., Nice, France.
1 WV m d
ER⫽
Sm D ␳m
2 Pipes,’’ Advances in Solid-Liquid Flow in Pipes & Its Application, I. Zandi,
where ER is erosion rate in mm/yr; W is sand flow rate in kg/d;
V m is fluid flow velocity in m/s; d is sand size in micron 共effect of
d becomes negligible above 400 microns; therefore, for d⬎400,
the limit of 400 is used兲; D is pipe internal diameter in mm; ␳ m is
fluid mixture density in kg/m3; S m is geometry factor which
equals 5.5 for pipe bends.
3 The minimum flow velocity below which solid particles in
the fluid can form a bed on the bottom of a horizontal pipeline can
be predicted using the following equation:
Journal of Energy Resources Technology
V m⫽ 冉 冊
V sl
Vm
0.53
d 0.17␯ 0.09 冉 冊
⌬␳
␳f
0.55
D 0.47
where
V m ⫽ minimum mixture flow velocity to avoid sand settling,
m/s
V sl /V m ⫽ ratio between liquid superficial velocity and mixture
velocity at sand settling condition 共for single phase,
this ratio is 1兲
d ⫽ particle diameter, m
D ⫽ pipe diameter, m
⌬␳ ⫽ density difference between particles and liquid, kg/m3
␳ f ⫽ liquid density, kg/m3
␯ ⫽ kinematic viscosity, m2/s
4 Experimental results have shown that sand settling will not
have negative impact on the effectiveness of inhibitors to suppress
uniform corrosion. However, not all inhibitors are effective in
suppressing pitting. The possible bacteria growth under sand beds
is an area that requires more investigations.
5 Commercially available monitors have the potential of pro-
viding high accuracy for quantitative sand production, providing
the effects of flow conditions and geometry on sand transport are
properly accounted for.
Acknowledgment
The author would like to thank the management of Conoco for
their permission to publish this paper. The author would also like
to express his appreciation to Oddmund Kvernvold of DNV and to
Simon Dawson and Tim Locket of AEA for their input.
References
关1兴 Salama, M. M., and Venkatesh, E. S., 1983, ‘‘Evaluation of Erosional Velocity
Limitations in Offshore Gas Wells,’’ Proceedings of the 15th Annual Offshore
Technology Conference, OTC Paper 4485.
关2兴 Svendeman, S. J., and Arnold, K. E., 1994, ‘‘Criteria for Sizing Multi-phase
Flow Lines for Erosive/Corrosive Services,’’ SPE Prod. Facil., 9, No. 1, pp.
74–80.
关3兴 Shirazi, S. A., Shadley, J. R., McLaury, B. S., and Rybicki, E. F., 1993, ‘‘A
Procedure to Predict Solid Particle Erosion in Elbows and Tees,’’ Proceedings
of Codes and Standards in a Global Environment, ASME PVP-Vol. 259, pp.
159–167.
关4兴 Shirazi, S. A., McLaury, B. S., Shadley, J. R., and Rybicki, E. F., 1994,
‘‘Generalization of the API RP 14E Guidelines for Erosive Services,’’ Pro-
ceedings of SPE 69th Annual Technical Conf., SPE Paper 28518, pp. 583–592.
关5兴 Birchenough, P. M., and Dawson, S. G., 1993, ‘‘Design Procedure for Erosion
in Multi-Phase Flow,’’ private communication.
关6兴 Salama, M. M., 1998, ‘‘An Alternative to API 14 E Erosional Velocity Limits
for Sand Laden Fluids,’’ Proceedings of the Annual Offshore Technology Con-
ference, OTC Paper 8898.
关7兴 Weiner, P. D., and Tolle, G. C., 1976, ‘‘Detection and Prevention of Sand
Erosion of Production equipment,’’ API OSAPER Project No. 2, American
Petroleum Inst., Texas A&M Research Foundation.
关8兴 Tolle, G. C., and Greenwood, D. R., 1977, ‘‘Design of Fittings to Reduce
Wear Caused by Sand Erosion,’’ API OSAPER Project No. 6, American Pe-
troleum Institute, Texas A&M Research Foundation.
关9兴 Bourgoyne, A. T., 1989, ‘‘Experimental Study of Erosion in Diverter Systems
Due to Sand Production,’’ Proceedings of the SPE/IADC Drilling Conference,
New Orleans, LA, SPE/IADC 18716.
关10兴 Angelson, S., Kvernvold, O., Linglem, M., and Olsen, S., 1989, ‘‘Long Dis-
tance Transport of Unprocessed HC. Sand Settling in Multi-Phase Flow
Lines,’’ Proceedings of the 4th International Conference on Multi-Phase
关11兴 Wicks, M., 1971, ‘‘Transport of Solids at Low Concentration in Horizontal
ed., Pergamon Press, pp. 101–124.
关12兴 Holte, S., Angelson, S., Kvernvold, O., and Raeder, J. H., 1987, ‘‘Sand Bed
Formation In Horizontal and Near Horizontal Gas-Liquid-Sand,’’ The Euro-
pean Two-Phase Flow Group Meeting, Trondheim, Norway.
关13兴 Oroskar, A. R., and Turian, R. M., 1980, ‘‘The Critical Velocity in Pipeline
Flow of Slurries,’’ AIChE. J., 26, No. 4, pp. 550–558.
关14兴 Davies, J. T., 1987, ‘‘Calculations of Critical Velocities to Maintain Solids in
Suspension in Horizontal Pipes,’’ J. Eng. Sci., 42, No. 7, Pergamon Journal
Ltd, UK, pp. 1667–1670.
关15兴 Chein, S. F., 1994, ‘‘Settling Velocity of Irregularly Shaped Sand Particles,’’
SPE Paper 26121, Proceedings of Drilling and General Petroleum Engineer-
ing, SPE.
MARCH 2000, Vol. 122 Õ 33

Michael A. Adewumi
E. S. Eltohami
W. H. Ahmed
Petroleum Engineering Department,
The Pennsylvania State University,
102 Mineral Science Building
University Park, PA 16802
Pressure Transients Across
Constrictions
The modeling of pressure transient across constrictions is achieved by using a one-
dimensional, isothermal, noncompositional, single-phase representation of the Eulerian
model. A TVD scheme was used to solve the ensuing nonhomogeneous hyperbolic set of
first-order quasi-linear partial differential equations. Three types of constrictions were
modeled and in each case the behavior of the transient was analyzed. This analysis was
used to interpret the pressure response at the inlet resulting from the reflection of the
transient at the constrictions. The comparisons between the predicted and inputted data
are very good, suggesting that the technique has much promise.
关S0195-0738共00兲00301-0兴

Introduction this work is to explore the possibility of using pressure transient

analysis to detect the existence, location, and severity of block-
The use of natural gas is growing rapidly, making it one of the
ages. This method could serve as an early warning method. If the
major sources of energy for the 21st century. Its clean-burning
location, and severity of a blockage can be predetermined, this
attributes appeal to environmentalists, while its relative abundance
knowledge would enable the pipeline operator to carry out pig-
makes it attractive to energy policy makers, particularly those in
ging operations on unconsolidated depositions with minimum dis-
industrialized countries. Deep offshore sources factor prominently
ruption to production operations. Furthermore, early detection of
into gas availability considerations. For these sources in particu-
blockage occurrence, size, and location will enable the operator to
lar, flow assurance in pipeline transportation is a great challenge
devise an alternative plan for meeting their supply requirements
that must be overcome.
by adjusting their production schedule.
Natural gas pricing is directly influenced by its transportation
The basic strategy is to use pressure transient as a means of
cost. Pipeline remains the most efficient avenue for bringing this
detecting the existence of blockage in the pipeline. The transient
valuable commodity to the market. However, pipe flow assurance
is purposely initiated using inlet flow variation. This is achieved
can be severely curtailed by partial blockages, whose immediate
by momentarily altering the rate of fluid incursion into the pipe-
impact is loss of deliverability and higher compression costs.
line. This initiates transient propagation in the pipe. On reaching
These blockages can arise from condensation and solid deposition
the blockage, some portion of the transient will be reflected and
in pipelines. Early detection of these blockages is useful for op-
propagated upstream. The characteristics of this reflected tran-
erational purposes, since prompt remedial actions can be taken
sient, which can be monitored at the pipe inlet, should tell the
before catastrophic effects are incurred.
story of the pipeline internals. The monitored signal now becomes
Problem Definition. Pipeline blockages often result from a window for observing the pipeline’s interior. This study is an
deposition of hydrates, paraffin, and asphaltene. In the case of attempt to examine this concept through a series of carefully de-
hydrates, the deposition follows their formation, which can be signed numerical experiments.
hindered by the use hydrate inhibitors. Unfortunately, no inhibitor
The Mathematical Framework. The mathematical model
has been found that is fully effective in totally eliminating this
used in the development of this study is based on a one-
problem. Hydrate formation and deposition still occurs, and hence
dimensional representation of the isothermal, noncompositional,
the risk of blockage remains. The formation and deposition of
single-phase, Eulerian model, consisting of the continuity, and
hydrates has been found to occur under high pressures and low
momentum equations. Allowing for variation in the pipe’s cross-
temperatures. The rapid expansion of petroleum exploration into
sectional area along its length approximates the 2-D effects im-
deeper offshore waters, with the associated large capital outlay,
posed by the constrictions. Furthermore, the inertial force has
makes the protection of these investments a high priority. Risk on
been included as a forcing function in the momentum equation.
these investments must be lowered if these developments are to
continue at the current pace. Pipeline blockages pose serious haz-
ards to these investments. Early detection and accurate location of The Model
blockages will enable the operator to take prompt remedial actions
that will lower the risk significantly. In summary, apart from low- ⳵共 ␳ gS 兲 ⳵共 ␳ gv S 兲
ering deliverability, partial and/or total pipeline blockages can ⫹ ⫽0 (1)
cause other problems, some of which may be catastrophic, unless
⳵t ⳵x
proper remedial actions are promptly taken 关1兴. Pipeline blockage
problems have been reported worldwide. ⳵ 共 ␳ g v S 兲 ⳵ 关共 ␳ g v 2 ⫹ P 兲 S 兴 ⳵S
⫹ ⫽ P ⫺F f ⫺F grav (2)
⳵t ⳵x ⳵x
The Proposed Solution Strategy. Since total avoidance of
pipeline blockages is not possible by chemical injection, the most The friction factor f used in the inertial force F f was calculated
plausible way to handle pipeline blockages is early detection and using a friction factor correlation 关2兴. Compressibility factor cor-
prompt remediation. Early detection of pipeline blockage will im- relation 关3兴 was used to incorporate real gas effect.
prove the effectiveness of pipeline pigging operations. The goal of
zRT
P⫽ ␳ (3)
Contributed by the Petroleum Division and presented at the 20th Annual Energy- Mg s
Sources Technology Conference and Exhibition, Houston, Texas, February 1–3,
1999, of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manuscript re-
ceived by the Petroleum Division, March 15, 1999; revised manuscript received Since the flow is assumed to be isothermal, the acoustic speed of
December 3, 1999. Associate Technical Editor: M. A. Adewumi. sound through gas is given by

34 Õ Vol. 122, MARCH 2000 Copyright © 2000 by ASME Transactions of the ASME
 c⫽ 冉 冊
zRT
Mg
1/2
(4)
In order to minimize numerical dispersion, the foregoing equa-
tions were recast in the conservative form. This aids resolution of
sharp frontal discontinuities. The conserved variables are ␳ g and
m 共where m⫽ ␳ g v 兲. The governing equations are rearranged into a
set of 1-D, first-order, nonlinear hyperbolic partial differential
equations as follows:
៝ ⳵ F៝ 共 U
⳵U ៝兲
⫹ ៝ 共U
⫽Q ៝兲
⳵t ⳵x
冉 冊
(5)
冉 冊
m using the same procedure as that used by Ahmed 关8兴, which is
␳g
៝⫽ ៝ ⫽ m2
U F
m ⫹c 2 ␳
␳g
冉 冊
m ⳵S
⫺ ៝ 共U
S ⳵x
៝ 共 Ū 兲 ⫽
Q (6)
兩 m 兩 •m ⳵ S
⫺ ⫺F f ⫺F grav
␳ gS ⳵ x
Apart from the constitutive relationships already mentioned, the
model uses empirical correlations for estimating fluid properties
such as gas viscosity 关4兴, along with relations for the forces F f ,
and F grav .
The Initial Conditions. The integration of Eq. 共5兲 must pro-
ceed from an admissible starting point, the initial conditions.
There are three distinct options that can be specified as the initial
conditions for Eq. 共5兲. The best option when possible, is to start
up with real field data of the primary variables. However, in the
absence of field data, the other two, which are analytical assump-
tions, can be used as well. These are either a steady-state formu-
lation, which is in fact a late-time solution of Eq. 共5兲, or a line
packing or start-up specification where the mass flux everywhere
is set to zero and the pressure is set to a constant value every-
where. The former is of greater practical interest, since we would
like to predict a blockage without necessarily shutting down a
portion of the system, which the physical state described by the
former. However, the present study is aimed at discovering the
physical process that occurs and how the pressure response at the
inlet due to a pulse reflected by the pipe’s internal structure can be
analyzed in order to predict the internal structure. There are sev-
eral numerical and analytical late-time approximations to Eq. 共5兲.
An analytical technique developed by Zhou and Adewumi 关5兴 was
used to test the validity of the numerical scheme.
The Numerical Scheme. Equation 共5兲 constitutes a set of
first-order, nonlinear hyperbolic partial differential equations, to
which there is no known analytical solution. There is, however, a
large variety of numerical schemes that can be used to find an
approximate solution to Eq. 共5兲. Such numerical solutions are usu-
ally corrupted to varying degrees by either numerical dispersion
or numerical oscillations, or both.
The total variation diminishing 共TVD兲 scheme developed by
Harten 关6兴 is well suited to finding admissible solutions to Eq. 共5兲
since it is not only a monotonic scheme, but also second-order
accurate in space, thus prohibiting the creation of oscillations at
the wave front. This is achieved by limiting the value of the
second-order term by the use of a suitable flux limiter. However,
the scheme is only first-order accurate in time, and for increased
accuracy the Runge–Kutta multistage method of numerical inte-
gration of ordinary differential equations is applied to the time
derivative. Equation 共5兲 can be rewritten as
៝
⳵U
៝ 共U
⫽⫺R ៝兲 (7)
⳵t
where the primary variables are then updated to time level n⫹1
Journal of Energy Resources Technology
R៝ 共 U
៝ 兲⫽
⳵ F៝ 共 U
⳵x
៝兲
៝ 共U
⫺Q ៝兲 (8)
using the technique developed by Jameson et al. 关7兴, which for an
mth-order approximation is given by
៝ 0j ⫽U
U ៝ nj (9)
៝ kj ⫽U
U ៝ 0j ⫺ ␩ k •⌬t•R 共 U k⫺1
j 兲 k⫽1,m (10)
៝ n⫹1
U ៝ mj
⫽U (11)
j
when m⫽4 then ␩ 1 ⫽1/4, ␩ 2 ⫽1/3, ␩ 3 ⫽1/2, ␩ 4 ⫽1.
The numerical handling of the source term Q(U) is achieved
冉 冊
given in the following:
m j⫺1 共 S j ⫺S j⫺1 兲
Sj ⌬x
Q ៝ j 兲 ⫽⫺ (12)
m j⫺1 兩 m j⫺1 共 S j ⫺S j⫺1 兲
⫹F j ⫹F grav
␳ jS j ⌬x
It can be seen that while handling the jth block, the mass flux
used in the foregoing is that of the j⫺1th block. This was neces-
sary in order to maintain mathematical consistency. However, this
representation of the mass flux in the jth block causes the numeri-
cal artifact seen in Figs. 2 and 3.
The Boundary Conditions. The handling of boundary con-
ditions numerically differs from that of mathematical physics. If
an analytical solution of Eq. 共5兲 exists, the requirement from a
mathematical point of view would be that two boundary condi-
tions are needed in order to arrive at a unique solution to Eq. 共5兲.
However, numerically, the situation is different; in order to pro-
ceed with a numerical solution and arrive at a unique solution, we
require that all values of the primary variables, at the boundaries,
be known at the start of each time step. Again, as with the initial
condition, the solution that we arrive at is strongly driven by
where we start. Therefore, great care needs to be given to how the
other boundary variable is determined at the inlet, knowing the
value of one of the variables; the same is true about the outlet.
It should be noted that at the outlet, we only control the mass
flux, since from a practical point of view, this is what is required.
At the end of the day, an operator only knows how much gas is
needed at the outlet to satisfy a contractual obligation. Therefore,
the questions are:
• If the mass flux at the inlet is a constant or known function of
time, what is the pressure 共hence density兲 at the inlet?
• If the pressure at the inlet is a constant or known function of
time, what is the mass flux at the inlet?
• If the mass flux at the outlet is a constant or known function
of time, what is the pressure 共hence density兲 at the outlet?
In all the foregoing, the method of characteristics was used to find
the unknown variable 关9兴.
The Numerical Experiments. Before one could begin to
study the problem of pipeline blockage, a validation of the tran-
sient code is necessary. In doing this, we used similar experiments
to those used by Zhou and Adewumi 关9兴. This validation is nec-
essary since this code is going to be used to generate the propa-
gation of the transients through the pipeline; therefore, we need to
be sure that it can simulate this transient to a high degree of
accuracy. This was achieved to high level of satisfaction.
Three distinct sets of numerical experiments were carried out.
In all the experiments, a 1-mi long pipe, whose internal diameter
is 8 in. is divided into 200 blocks. The objective of each experi-
ment was to simulate the propagation of different transients in a
pipeline, which is partially blocked. Imposing different boundary
conditions at the inlet generated the different transients. As the
transient propagates through the pipe, the pressure at the inlet was
MARCH 2000, Vol. 122 Õ 35
Fig. 1 Inlet pressure response for continuous sudden area
change and no blockage

Fig. 2 The pressure profiles along a pipe with a sudden con-

tinuous blockage at 0.6 m „profiles are 427 ␮s apart…

Fig. 4 „a… Pipe with blockage type of sudden continuous area

change; „b… pipe with blockage of normal type; „c… pipe with
blockage type of limited sudden area change

Part of the energy introduced into the system becomes internal

Fig. 3 The mass flux profiles along a pipe with sudden con- molecular energy, while the rest goes into kinetic energy 共KE兲 to
tinuous blockage at 0.6 m from the inlet „profiles are 427 ␮s move the fluid along. Under the isothermal condition, any in-
apart… crease in the fluid’s internal energy must manifest itself in a rise in
the fluid’s pressure, which is caused by an increase in its density.
This is because the rise in pressure 共⌬ P inc in this case兲 is caused
by the agitation of the fluid molecules as they move about to
recorded every 1.78 ms 共as shown in Fig. 1兲. In addition to this, accommodate the incoming fluid. This creates a pressure differen-
both the pressure and mass flux profiles throughout the pipe were tial with the upstream blocks, hence creating a frontal discontinu-
recorded every 427 ms 共as shown in Figs. 2 and 3, respectively兲. ity. This frontal discontinuity then propagates downstream into
By analyzing these curves, deductions can be made about the the pipe 共Figs. 2 and 6, label 1兲 at the prevailing speed of sound.
internal structure of the pipeline. We will provide the basic phys- This propagation is driven by the difference in the potential head
ics for this analysis in the following paragraphs. of the fluids on either side of the discontinuity.
As the fluid propagates into the pipe, we consider the base case
The Square Pulse at the Inlet where there is no blockage in the pipe. By comparing this with
other results, one can infer the impact of pipe blockage on the
The first set of experiments uses a square wave at the inlet and mode of propagation of the imposed waveform.
two types of blockage were simulated. In the first case, the block-
age was simulated by reducing the cross-sectional area of the pipe
by a factor, ␭, beginning at a distance of x miles from the inlet
共see Fig. 4共a兲兲. In the second case, a deposition profile, which is Table 1 The initial start-up data
normally distributed along the pipeline, is used to simulate the
blockage. This is a more practical scenario. The initial flow con- Pipeline length 5280 ft
ditions imposed are given in Table 1. The pipe is closed at the Pipeline internal diameter 8 in.
Pipeline roughness 0.0006
outlet and the pressure is initially equal in all the blocks. Then, at Gas specific gravity 0.64
time t⫽0 ⫹ , the inlet is opened and a constant mass flux is ap- No. of nodes 200
plied, causing the mass flux at the inlet to jump from zero to the Inlet pressure at t⫽0 1000.0 psia
Initial pressure at all blocks 1000.0 psia
specified value instantaneously. Thus, the mass flux versus time at Inlet temperature 80 F
the inlet has a square waveform as shown in Fig. 5共a兲. This Inlet flow rate at t⫽0 0.0 MMScfD
boundary condition at the inlet implies that energy is continuously Inlet flow rate at t⫽0 ⫹ 10.0 MMScfD
being added into the system.

36 Õ Vol. 122, MARCH 2000 Transactions of the ASME


Fig. 5 „a… The square compression pulse at the inlet; „b… the
compression pulse at the inlet; „c… the expansion pulse at the
inlet
Fig. 6 The pressure profiles along a pipe with no blockage
„profiles are 427 ␮s apart…
No Blockage. From the analysis of Fig. 6, which shows the
pressure profiles along the pipe at different times when there is no
blockage, it is clear that the pressure at the outlet rises by 2 * ⌬ P inc
upon the arrival of frontal discontinuity there. This is because, as
the new fluid arrives at the outlet, it compresses the fluid in that
block thereby causing its pressure head to rise by ⌬ P inc . How-
ever, the boundary condition 共BC兲 of a closed value at the outlet
necessitates that the mass flux there be zero. This means that the
fluid there must come to a halt; hence, all its KE is transformed
into potential energy, causing its pressure head to rise further. It
can be seen from the figure that the fluid’s KE must have been
numerically equal to ⌬ P inc , since the pressure head at the outlet
doubles.
Journal of Energy Resources Technology
Fig. 7 The mass flux profiles along a pipe with no blockage
„profiles are 427 ␮s apart…
At this point, the potential difference between the outlet and the
next upstream block becomes zero. Therefore, there can be no
flow between these blocks. This causes the fluid in the upstream
block to decelerate to zero. This again means that the KE of its
fluid must be converted into a pressure head of ⌬ P inc . In this
manner, the pressure discontinuity, caused by flow stoppage, is
propagated upstream 共Fig. 6, label 2兲. Note that propagation of the
pressure front upstream occurs as the fluid flow is halted in se-
quential blocks and that it is the flow stoppage that propagates
upstream and not the fluid mass. As can be seen on the mass flux
profiles 共Fig. 7, label 1兲, it is the zero mass flux that propagates
towards the inlet.
Upon arrival of this frontal discontinuity at the inlet, the pres-
sure there rises by 2 * ⌬ P inc . This can be seen in both the inlet
response curve of Fig. 1, and the pressure profile curve of Fig. 6.
This rise in pressure head is caused by two different compressions
that occur simultaneously. First, the arrival of the 共zero兲 mass flux
front necessitates the pressure head to rise by ⌬ P inc as before.
This is due to the conversion of the fluid’s KE to a pressure head.
Secondly, the boundary condition at the inlet necessitates the mass
flux there to be at the prespecified constant value 共Table 1兲. There-
fore, the new incoming fluid compresses the fluid already in the
pipe further and sets it in motion 共hence KE兲, thus satisfying the
inlet BC. Now that a cycle has been completed, the new frontal
discontinuity begins to propagate downstream 共Fig. 6, label 3兲.
This cycle has a period of about 9.5 s. During this same period,
1100 scf of fluid is introduced into the system. Under the system’s
prevailing temperature and pressure, this is equivalent to 15 ft3, of
which only 0.075 ft3 actually arrives at the outlet, while the pres-
sure everywhere has risen by ⬃8 psia. So, it is the shock wave
that moves through the pipe so fast 共at the speed of sound兲, and
not the actual fluid mass. For this reason, great care must be
exercised during line packing, since if too much fluid is passed
into the pipe suddenly the pressure rise may cause the pipe to
rupture within an extremely short period of time.
Continuous Sudden Area Change Blockage. In this sce-
nario, the transient in Fig. 5共a兲 is imposed at the inlet of a pipe
with a cross-sectional profile of the type shown in Fig. 4共a兲. A
frontal discontinuity is created as before and begins to propagate
downstream in the same manner. This is evident from the pressure
response curve shown in Fig. 1 for a pipe with a constriction of 36
percent beginning at a distance of 3168 ft from the inlet.
Here, the flow stoppage occurs at the constriction first; whereas
for the no-blockage case, all the fluid is stopped at the outlet, since
there is nowhere for it to go. Now, only a fraction will be stopped
since the pipe is only partially blocked. The fraction of the flow
held up at the constriction loses all its kinetic energy 共Fig. 3, label
1兲, which is converted to a pressure head. This pressure head will
do work against the flow coming in from the upstream block, thus
causing an equal amount to lose its KE and be held up in that
block. A frontal discontinuity of ␣ ⌬ P inc is propagated upstream
as a result of the partial fluid stoppage 共Fig. 2, label 2兲. It is clear
MARCH 2000, Vol. 122 Õ 37
that the value of ␣ depends on the severity of the constriction. The
calculation of ␣ used in this study is presented in the Appendix.
From Fig. 2, label 3, it can be seen that when the frontal dis-
continuity, initiated at the constriction, arrives at the inlet, the
pressure there rises by 2 ␣ ⌬ P inc . The first pressure rise of ␣ ⌬ P inc
is due to the loss of the fluid’s KE, while the second ␣ ⌬ P inc is
due the compliance with the BC there. This creates a frontal dis-
continuity of ␣ ⌬ P inc that propagates into the pipe.
At this stage, the inlet response curve of Fig. 1 has developed to
the extent that we can deduce all we need to know about the
pipe’s internal structure. That is the location of the blockage as
well as the severity of blockage 共see the Appendix兲. It is worth
noting that the pressure discontinuity that occurs as a result of
partial fluid stoppage should have a perfectly square waveform,
however, because of the smearing affect of the numerical scheme
this was not realized.
Normally Distributed Blockage
In this set of the numerical experiments, the inlet mass flux is a
square waveform while the type of blockage simulated in the pipe
is that of a normally distributed deposition profile. This type of
blockage is completely defined by the mean and standard devia-
tion of a normal distribution. The mean value is the distance
where the deposition is greatest and about which the deposition is
symmetrical, henceforth referred to as the location of the throat.
The standard deviation is a measure of the lateral spread of the
deposition. A typical deposition profile of this type is shown in
Fig. 4共b兲.
A frontal discontinuity is created and propagated as before,
until it arrives at the blockage. Now the constriction in the area
occurs in small increments. As the front encounters this gradual
change in area, a small fraction of the fluid is stopped and its KE
is converted to a pressure head. This slight increase in pressure
head pushes the fluid into the reduced area at a faster rate. This
means that the mass flux in the reduced area has now been in-
creased along with the KE because of the increase in the fluid’s
velocity. It should be noted that the increased density associated
with the increased pressure also causes a further increase in mass
flux and the KE. The front then propagates into a further reduced
area and more fluid is stopped than in the previous block. This
causes a further increase in the pressure of the block; hence there
is a larger mass flux and fluid has greater KE. Each reduction of
area creates a similar pressure rise because of the increase in fluid
stopped.
The fluid stopped at each reduction, in turn, stops an equivalent
amount in the upstream block. In this manner the gradual increase
in pressure is propagated towards the inlet. Upon arrival of the
stoppage shock wave at the inlet, the pressure there is increased
due to fluid stoppage at the inlet block. In addition, the compli-
ance with the BC at the inlet causes a further pressure increase,
and sets the fluid in motion, thus imparting KE to it.
The pressure increase continues as the sequence of stoppage
waves 共caused by area decrement兲 arrives at the inlet. The maxi-
mum pressure rise corresponds to the stoppage wave initiated at
the throat. It can be seen from Fig. 8 that the pressure discontinu-
ity is more smeared than that in Fig. 1. The increased smearing is
due to the gradual reduction in the cross-sectional area as de-
scribed by the normal distribution of deposition along the pipe.
Therefore, the pressure response profile mimics the blockage’s
normal distribution and can be used to infer its mean and standard
deviation.
When the front arrives at the throat, the potential differential
between the fluids either side of it causes all the fluid in the throat
block to be pushed into the next downstream block. However,
because of the increased volume of that block, the density there is
reduced and the fluid is retarded leading to a reduction in the
fluid’s KE. This implies that the energy is reduced in that block;
however, in compliance energy conservation, more fluid is sucked
into the block. This causes the density there to rise, thus increas-
38 Õ Vol. 122, MARCH 2000
Fig. 8 Inlet pressure response for normally distributed block-
age and no blockage
ing both the pressure and KE in the block such that its energy
rises to the necessary value. The extra fluid required to do this
comes from the fluid held up in the blocks upstream from the
throat. In this manner, the fluid in these upstream blocks expands
and the pressure there is reduced. As the front propagates into
ever increasing cross-sectional areas, this is repeated until all the
fluid held up is exhausted. At this stage, the mass flux after the
constriction is consistent with that before it and that at the inlet.
This is why the inlet pressure response curve of Fig. 8 returns to
the original pressure increase of ⌬ P inc .
Therefore, the peak in Fig. 8 corresponds to the maximum held
up fluid, and hence it corresponds to the arrival of the front at the
throat, while the pressure drop thereafter corresponds to the in-
creased area after the throat. The information required to com-
pletely define the blockage profile is now available in Fig. 8 with
the standard deviation given by the rate of increase of inlet
pressure.
The Compression Pulse at the Inlet
Here, a compression pulse, corresponding to a single energy
packet, is introduced at the inlet by a mass flux variation over one
time step 共Fig. 5共b兲兲. The propagation of this energy packet
共pulse兲 is observed. A compression pulse is a more practical inlet
BC because during transient it is very difficult to keep the mass
flux constant at the inlet. Furthermore, the difficulty of realizing
phenomena that are superimposed on the constant inlet mass flux
is removed. This is because we only need to consider the single
energy packet introduced, and we need to account for it every-
where and every time.
The inputted energy packet compresses the fluid at the inlet
block and imparts to it KE simultaneously 共as before兲. The fluid
then flows out of the inlet block and the pressure there drops to its
original value, as the fluid is forced into the next block down-
stream. Now, the fluid in the second block is compressed and
forced into motion. This process is repeated from block to block
and the transient is propagated into the pipe’s internal structure.
No Blockage. The pulse arrives at the outlet where the pres-
sure head becomes the total energy introduced into the system at
the inlet. Hence, outlet becomes identical to the inlet, since now
the BCs are identical. The front then propagates to the inlet and a
cycle is now completed. In the absence of frictional forces, this
becomes a perpetual system, however, in practice the frictional
forces may dissipate the front before it completes any appreciable
distance. Therefore, in practice, care must be taken to ensure that
the pulse can do work against friction while completing the dis-
tance between the inlet and outlet twice. In these experiments, the
effect of friction was neglected so as to allow a more thorough
understanding of the physics first. Furthermore, in this type of
pulse, the problem of numerical dispersion is also present since
the peaks shown in Fig. 9 should have been triangular.
Transactions of the ASME

Fig. 9 Inlet pressure response for a sudden limited blockage
starting at 0.6 m from the inlet and 487 ft long „compression at
inlet…
Limited Sudden Area Change Blockage
Figure 4共c兲 shows the type of blockage simulated in these nu-
merical experiments. The inlet response curves shown in Fig. 9
corresponds to a limited sudden area change, which is approxi-
mately 450 ft in extent and the area available to flow at the con-
striction is 36 percent of the normal area 共heavy line兲 and 64
percent 共light line兲.
As the compression pulse arrives at the blockage, the pressure
builds up due to the conversion of part of the fluid’s KE into
potential energy. This converted energy is now reflected towards
the inlet and arrives as the first peak on the inlet pressure response
curve of Fig. 9. The fraction of energy reflected is proportional to
the blockage severity 共see the Appendix兲. The increased pressure
head causes the fluid pushed into the constriction to flow faster.
Thus, the mass flux in the constriction is increased along with the
KE. It should be noted that the increased density associated with
the increased pressure also causes a further increase in mass flux.
Furthermore, these are increases in energy per unit volume and
not the total energy, which must remain constant.
This compression pulse continues to propagate through the con-
stricted section until it arrives at the enlarged section again. When
this propagated fluid enters the enlarged section it expands and its
velocity is decreased. As a result, its energy will decrease, and
since this is not possible, then the excess energy will be used in
removing more fluid from the last block in the constricted area
than was originally pushed in at the first block. This means that
the pressure there will fall below the original value of 1000 psia
that was original everywhere in the pipe. This drop in pressure
initiates an expansion wave that then propagates towards the inlet.
The compression wave created in the enlarged section propagates
to the outlet where it is reflected into the pipe.
When the expansion wave 共with its pressure drop兲 encounters
the other end of the constriction the pressure differential there
causes fluid to be sucked into the constriction. The expansion then
continues to propagate towards the inlet where its arrival is
marked by the first minima in Fig. 9.
The fluid sucked into the constriction initiates a compression
wave that propagates towards the downstream end. Upon arrival
there, another expansion wave is created that travels into the con-
striction towards the inlet side, while a compression pulse is ini-
tiated in the enlarged section propagating towards the outlet.
When this expansion wave encounters the upstream end of the
constriction, a new expansion wave is created in the enlarged
section that arrives at the inlet as the second minima 共Fig. 9兲. Note
that this phenomena is repeated indefinitely, however the ampli-
tude of the waves created each time decreases as can be seen in
Fig. 9.
The next peak in Fig. 9 represents the arrival of the first com-
pression wave at the inlet. The arrival of further compression
waves at the inlet represents the compression waves formed each
time the foregoing phenomena take place within the constriction
共Fig. 9兲.
In Fig. 9, the time taken for the first compression wave to arrive
Journal of Energy Resources Technology
Fig. 10 Inlet pressure response for a sudden limited blockage
starting at 0.6 m from the inlet and 487 ft long „expansion at
inlet…
at the inlet corresponds to the distance between the inlet and the
blockage marked by x miles in Fig. 4共c兲. Furthermore, the time
between that peak and the next minimum corresponds to twice the
length of the blockage, while the ratio of the amplitude of these
two peaks can be used to calculate the severity of the blockage.
The Expansion Pulse at the Inlet. In this set of numerical
experiments, the goal is to see how the inlet response would
change if the input waveform were an expansion pulse rather than
a compression pulse. In the case of the compression pulse, we saw
that we needed to ensure that the pulse would be able to do work
against friction and still be detectable. This can only be achieved
if a large amount of pressure was applied at the inlet, which could
be a safety concern. For this reason, the expansion pulse may be
preferred at the inlet. Since mass would now be removed from the
system, there would be a pressure reduction almost everywhere.
As can be seen from Fig. 10, the response to the expansion pulse
at the inlet mirrors that of the compression pulse; whereas, in the
case of the compression pulse, there was compression, now there
is expansion, and vice versa.
Blockage Detectability. The objective of the preceding
analysis is to gain an understanding of the processes that cause the
generation of certain pressure response curves. These curves in-
clude such things as the increase and decrease in inlet pressure,
the creation of maxima or minima peaks, and the creation of pres-
sure plateaus. Once such an understanding was attained, we then
proceeded with writing a series of logical codes that can read the
inlet response data and then attempt to interpret the data.
These curves are generated by the internal structure of the pipe,
which manifests itself in the rate of change of the inlet pressure
response with respect to time. A study of these rates of change is
a good starting point from which we can make inferences about
this structure.
During the learning stage of this study, it was necessary to
incorporate data about the type of blockage. However, this would
be impractical since it must be assumed that this information is
not known when we try to analyze and interpret the inlet pressure
response curve. Having said that, in practice it is futile to assume
impractical types of blockages since they will make it impossible
to predict the internal structure from the inlet pressure response
curve alone. But this does not constitute a severe limitation on the
technique since, in practice, the blockage is most likely normally
distributed.
The results from the foregoing numerical experiments are given
as follows:
• For a square waveform, applied at the inlet to a pipe with
continuous sudden area change, the results show that from the
response generated by the transient, the location and severity of
the blockage can be accurately determined. In fact, all the results
were within 0.5 percent error 共see Figs. 11 and 12兲.
• For the normally distributed deposition, the results showed
that the throat of the blockage could be located with an error of no
more than 0.5 percent. The results for the spreads of the deposi-
MARCH 2000, Vol. 122 Õ 39
Fig. 11 Predicted versus actual location for continuous sud- Fig. 14 Predicted versus actual standard deviation for throat
den area change at 0.6 m from the inlet

Fig. 15 Predicted versus actual location for limited sudden

Fig. 12 Predicted versus actual blockage severity for continu- area change
ous sudden area change starting at 0.6 m from the inlet

Fig. 16 Predicted versus actual extent for limited sudden area

change

Fig. 13 Predicted versus actual location of throat for normally

distributed blockage

tion were found to be less accurate when the deposition is spread

over a small portion of the pipe. Even then, the error was less than
6 percent 共see Figs. 13 and 14兲.
• In the case of the compression and expansion pulses at the
inlet of a pipe with limited sudden area change, the results show
that all data could be determined to within 2 percent accuracy 共see
Figs. 15–17兲.

Conclusion
The primary goal of this work is to study the interaction of
pipeline transients with blockages in pipes. Having achieved this Fig. 17 Actual blockage severity versus percentage error for
goal, we tried to explore the possibility of using the understanding limited sudden area change starting at 0.6 m from the inlet

40 Õ Vol. 122, MARCH 2000 Transactions of the ASME

 冋 册
of these interactions for detecting and characterizing these block- S2
ages. The results suggest that this is a good possibility. It has been ⌬ P ref兩inlet⫺⌬ P inc 1⫺
S1
demonstrated, through a systematically designed suite of numeri- LHS⫽ ⫽2
cal experiments, that blockage detection could be achieved by ⌬ P inc S2
generating an appropriate transient pulse at the inlet. Since this 1⫹
S1
idea is yet to be verified in the field, various types of transient
configurations were used in the numerical experiments to generate Since all of the terms in the LHS can be recovered from the inlet
the inlet pressure responses required. pressure response curve, then the ratio S 2 /S 1 can easily be
The paper explains in detail the physical implication of each of calculated.
the peaks on the inlet pressure response curves that were gener-
ated. This was done so that a comprehensive system can be The Calculation of ␮ and ␴ for the Normal Distribution.
adopted to clearly identify each peak and to interpret the blockage This is done by normalizing the inlet pressure response curve up
inferred by that peak correctly. to the throat. Therefore
P共 t 兲
Appendix P norm共 t 兲 ⫽ ⫺1 where t⭐T throat
⌬ P inc
The Analogy With the Reflection of Sound Energy. The Then the time scale is converted to a distance scale by multiplying
severity of blockage can be determined from the analytical ex- by half the speed of sound. At this stage reflecting the normalized
pression for a plane sound wave reflected at constriction by using pressures about the throat completes a scaled deposition profile.
the ratio of the reflected pressure to that of the incident pressure. The values of ␮ and ␴ are then given by
From theoretical acoustics, the reflection coefficient R ␲ is given

冉 冉 冊冊
t⫽t end
by t•c
兺
冉 冊
P norm共 t 兲 •
S 1 ⫺S 2 2
t⫽0 2
R ␲⫽ ␮⫽
S 1 ⫹S 2 t⫽t end

Pr
兺 P norm共 t 兲
␣ ⫽ 冑R ␲ ⫽
t⫽0

冉 冉 冊冊
Pi t⫽t end

冪
2
t•c
where S 1 and S 2 represent the areas of the unconstricted and con-
stricted pipe section, respectively, and P i and P r are the incident
兺
t⫽0
P norm共 t 兲 •
2
⫺␮
␴⫽ t⫽t end
acoustic pressure and the reflected acoustic pressure at
constriction.
The acoustic pressure is defined as the excess pressure at any
兺
t⫽0
P norm共 t 兲

point, i.e., the actual pressure minus the pressure in the absence of
the sound wave at that point. Then, the incident acoustic pressure
is given by
P i ⫽ P inc⫺ P O
where P inc is the total incident pressure and P O is the initial pres-
sure before the propagation arrives. The total acoustic pressure in
the unconstricted pipe is now given by
P tot⫽ P i ⫹ P r ⫽ 共 1⫹ ␣ 兲 P i
Defining a pressure reflection ratio as
where t end⫽2L/c.
References
关1兴 Adewumi, M. A., 1997, ‘‘Natural Gas Transportation Issues,’’ J. Pet. Technol.,
49, No. 2, Feb., p. 139.
关2兴 Chen, N. H., 1979, ‘‘An Explicit Equation for Friction Factor in Pipe,’’ Ind.
Eng. Chem. Fundam., 18, p. 3.
关3兴 Dranchuk, P. M., and Abou-Kassem, J. H., 1974, ‘‘Calculation of Z Factor for
Natural Gas using Equation of State,’’ J. Can. Pet. Technol., 14, No. 3, p. 34.
关4兴 Lee, A. L., Gonzales, M. H., Eakin, B. E., 1966, ‘‘The Viscosity of Natural
Gas,’’ Society. Pet Eng., SPE 1340.

冉 冊
关5兴 Zhou, J., and Adewumi, M. A., 1995, ‘‘The Development and Testing of a
P tot P ref⫺ P O ⌬ P ref 2 New Flow Equation,’’ Proc., Pipeline Simulation Interest Group, 27th Annual
R P⫽ ⫽ ⫽ ⫽ Meeting. Albuquerque, NM, October 19–20.
Pi P inc⫺ P O ⌬ P inc S2 关6兴 Harten, A., 1983, ‘‘High Resolution Schemes for Hyperbolic Conservation
1⫹ Laws,’’ J. Comput. Phys., 49, p. 357.
S1 关7兴 Jameson, A., Schmidt, W., and Turkel, E., 1981, ‘‘Numerical Solutions of the
Euler Equations by Finite Volume Method Using Runge-Kutta Time-Stepping
Now, because of the doubling effect described in the text, we Schemes,’’ AIAA, Paper No. 81–1259.
know that 关8兴 Ahmed, W. H., 1996, ‘‘Location of Partial Blockage in Natural Gas Pipelines
Using Transient Modeling,’’ M.S. thesis, The Pennsylvania State University,
⌬ P ref兩inlet⫺⌬ P inc⫽2 共 ⌬ P ref⫺⌬ P inc兲 University Park, PA.
关9兴 Zhou, J., and Adewumi, M. A., 1996, ‘‘Simulation of Transients in Natural
Dividing through by ⌬ P inc , we get Gas Pipelines,’’ SPE Prod. Facil., 11, No. 4, p. 202.

Journal of Energy Resources Technology MARCH 2000, Vol. 122 Õ 41

 The Nonoriented, Multiple
Passageway Flowline Connection
System
Gary Galle The remote, diverless, simultaneous connection of multiple subsea flowlines has histori-
Rudolf Bond cally been a difficult and costly operation. Difficulties and intensified operational activi-
ABB Vetco Gray Inc., ties stem from the need to orient the multiple flow paths prior to the final connection of the
12221 North Houston Rosslyn Road, tubulars. This rotational orientation has proven to be very difficult for randomly prelaid
Houston, TX 77086 flowlines or flowlines which must be rotated a relatively large amount prior to makeup. A
new, innovative method has been developed whereby multiple flowlines can be connected
without the need to rotationally orient the multiple tubulars. This new connection system
technology has undergone full-scale subsea testing and will be implemented in a major
development in the North Sea. The significance of this achievement is assessed in this
paper with a focus on the design philosophies used, the principles of operation, the
overall system reliability, the projected amount of operational cost reduction, and the
full-scale testing results. Additional comments are made concerning the applicability of
this technology in various other subsea applications. 关S0195-0738共00兲00501-X兴

Introduction In theory, conventional multiple passage connections are gen-

This paper covers the pull-in and connection of two or more
large-diameter tubulars simultaneously in a diverless, marine en-
vironment. The scope of this paper does not cover the connection
of ‘‘pod-type’’ systems which contain a plurality of smaller con-
trol lines, as these connections are well understood.
The nonorienting multiple passageway flowline connection sys-
tem consists of a diverless metal-to-metal sealing connector and
hub combination designed to remotely connect multiple rigid to
flexible flowlines and umbilicals without the use of divers 共see
Fig. 1兲. The design of the connector is based on a field proven
trunnion and drive-screw mechanism which actuates a three-piece
clamp 共see Fig. 2兲. By applying a torque to the end of the drive-
screw, the clamp can be opened or closed. The hubs, which attach
to the multiple flowlines, incorporate an integral guidance profile
which provides the fine alignment during pull-in operations and
include the necessary passageways to conduct the plurality of tu-
bulars in a single connection.
The paper begins with a general description of the operational
features along with the benefits stemming from this new technol-
ogy. This is followed by an overview of the design methodology
and analysis techniques used in the development of the multiple
passageway system. The paper concludes with a summary of the
full-scale testing program performed, as of this writing, on the
described technology.
Operational Features
This section details both the current methodology of connecting
multiple flowlines and a new, improved method. It further pre-
sents the new technology by describing the physical characteris-
tics of the design and the operational advantages.
Current State-of-the-Art, Multiple Connections. Currently,
multiple passage connections are made by incorporating several
flow passages into two solid hub housings. Each discrete flow
passage contains its own seal element which is energized between
the two housings, as shown in Fig. 3.
Contributed by the Petroleum Division and presented at the 20th Annual Energy-
Sources Technology Conference and Exhibition, Houston, Texas, January 27–31,
1997, of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manuscript re-
ceived by the Petroleum Division, February 7, 1997; revised manuscript received
January 4, 2000. Technical Editor: J. P. Brill.
erally used to reduce cost by:
• Eliminating the Number of Connectors Required. For ex-
ample, if a field requires twenty connections, only 10 two-passage
connections are required, thus eliminating 10 connectors.
• Reducing Operation Expenditures. Considering the preceding
example, the remote intervention activities will be reduced by
approximately half that of single passageway connections. This
will substantially reduce the connection activities, and thus the
overall operational cost of the pull-in activities.
• Reducing Overall Project Risk and Cost. In many areas of the
world, there is a limited ‘‘window’’ in which pull-in and connec-
tion operations can take place. In some parts of the world, the
operational window can close with very short notice, such as in
the Gulf of Mexico where hurricanes are prevalent. Thus, the
reduction of time required to complete field operations will add
more flexibility to work schedules centered around the acceptable
operational season and/or reduce the likelihood of mid-operational
abandonment.
In practice, however, the actual cost savings using conventional
multiple passage connections may not be fully realized. There are
a number of factors which may lead to complications and unsuc-
cessful connections. These factors include:
• Difficulty and/or Inability to Orient the Tubulars and Pas-
sageways Prior to Making the Connection. In most cases, the tu-
bulars 共for second-end connections兲 are placed on the seabed in a
somewhat random orientation. This means that it is not known
what orientation the laid multiple passageway housing will be
relative to the usually fixed second hub. Thus, the free end must
be rotated 共about the tubular axis兲 until the individual seal ele-
ments align with each other. Most systems today passively align
the tubulars, meaning that the axial load supplied to bring the
tubulars together also is partially converted to a torque to rotate
the tubulars into alignment. This conversion is usually done by
means of a cam or helical slot and pin arrangement. Depending on
the amount of initial rotational misalignment, the torque required
to rotate the hub into alignment can be high, which can lead to
stalled pull-in operations and/or damaged cam systems.
• Improperly Laying of the Flowlines Which Lead to
‘‘Kinked’’ Flexibles. For numerous reasons during the flowline
laying operation, it is quite possible for the flowlines to be laid

42 Õ Vol. 122, MARCH 2000 Copyright © 2000 by ASME Transactions of the ASME

Fig. 1 GSR marine connector
improperly due to the incorrect length of flexible remaining 共i.e.,
too long or short兲, improper approach to the target area, or diffi-
culties stemming from unexpected surface geography. These
problems compound the difficulty during rotational alignment and
have caused kinks to be created in the flexible lines. This problem
can be solved only by extensive operational activity to reposition
the flexible lines.
• Additional Axial and Cross-Axes Rotational Misalignment
Adds to the Rotational Misalignment Which can Damage Seal
Elements. Most marine connections require the use of metal-to-
metal seals to ensure life-of-field service and virtually zero per-
meation of gas across the seal body and interface. The use of
metal seals, however, place an additional requirement on the
pull-in and connection system by necessitating near-perfect align-
ment in the axial, cross-axis rotational, and rotational directions.
Fig. 2 Three-piece clamp assembly
Fig. 3 Conventional multi-bore hub
Journal of Energy Resources Technology
Thus, for conventional multiple passage connections, difficulty in
obtaining rotational orientation 共as described previously兲 coupled
with axial and cross-axes rotational misalignment may cause the
fine alignment of the tubulars relative to the metal seals practically
impossible. If not aligned within the acceptable envelope, the
metal seals and/or the seal pockets may be damaged—causing
leakage.
These problems, either separately or combined, can cause sig-
nificant unanticipated operational and possibly additional capital
costs. These unanticipated costs can make conventional multiple
passageway connections less attractive than using an equivalent
additional number of single connections.
Ahead-of-the-Art, Nonoriented, Multiple Connections. A
new method of connecting multiple flowlines 共see Fig. 4兲 has been
developed which attempts to prevent many of the experienced
problems associated with multiple connections. This new method
eliminates many of the current problems by:
• Precluding the Need to Rotationally Align Prior to Making
the Connection. The general seal arrangement is shown in Fig. 5.
For the two-passageway system shown, there are three concentric,
metal-to-metal seal elements. The first innermost seal element,
called the primary seal, seals the inner tubular as in a conventional
connection. The outer two seals, the inner and outer gallery seals,
provide the flow passage for the second tubular. The inner gallery
seal is a pressure-energized dual-lipped seal element similar to the
standard inner bore seal, except that the seal has been reversed.
The outer gallery seal is a conventional pressure-energized metal
seal. Thus, with the geometry shown, no rotational orientation is
required during the connection process. This drastically reduces
the operational difficulties normally associated with multiple pas-
sageway connections.
• Maintaining a Large Alignment Envelope Using Concentric
Seals. Since concentric seals are used throughout the design, the
alignment requirements are reduced to the same alignment re-
quirements of single passageway systems. This fact dramatically
increases the total alignment envelope and makes fine alignment
consistent with existing proven technologies.
• Maintaining a Uniform Preload Path Throughout the Connec-
tion. Since the seal elements are arranged in a concentric manner
共see Fig. 5兲, the connection preload path is nearly identical to
standard single bore applications. In conventional multiple pas-
sage systems, the preload paths are usually asymmetric, which
makes the total connection preload path impossible to predict with
two-dimensional finite-element techniques. This path can also
vary from connection to connection due to its inherent geometri-
cal tolerance sensitivity. Concentric seal elements eliminate asym-
metric load paths and provide easily predictable connection per-
formance using standard two-dimensional finite element
techniques.
As a result of the inherent features of the concentric multiple
passageway connection, the operational costs approach that of a
single passageway connection. Since there is no rotational orien-
tation required, conventional and existing pull-in techniques can
be employed. Thus, true cost savings can be realized.
Design Methodology
In a remote pull-in operation, there are various loading and
operational conditions which must be included in the design of the
connector/hub system to ensure both safety and performance. The
addition of multiple flowlines in a single hub further increases the
complexity of this analysis, as the interaction of multiple seals and
the behavior of a more complex flow path must be taken into
account. The starting point for the design of the nonorienting
multibore flowline connection system begins with the conceptual
development of the hub. The design criteria for the multibore hub
is summarized as follows:
MARCH 2000, Vol. 122 Õ 43
 Fig. 4 The nonoriented, multi-bore connection
• Nonorienting Design. A nonorienting hub significantly sim-
plifies the pull-in operation by eliminating large, complicated
tools designed solely to rotationally align the two tubulars.
• All Metal Sealing. To insure a reliable seal throughout the
life of the field, a metal sealing system is used.
• Provision for the Remote Verification of the Metal-to-Metal
Sealing System. The metal-to-metal seal rings used must be ex-
ternally pressure tested by ROV. This ensures that the seals are
fully energized and will function properly.
• Remote Retrieval of the Metal-to-Metal Sealing System by
ROV. The seal plate/seal ring assembly can be remotely retrieved
and replaced if necessary. This eliminates the need to retrieve the
manifold and/or flowbase in order to perform a seal change-out,
which, in many cases, is both time intensive and cost prohibitive.
• Misalignment Correction Capability. The multibore hubs in-
corporate an alignment profile integral to the hub to correct the
misalignment conditions present during the final stages of the
flowline pull-in.
• Minimal Pressure Loss Across the Connection. The design of
the system must minimize the pressure losses across the connec-
tion.
A nonorienting design is achieved through the use of a concen-
tric metal sealing system for the hub. The first system developed
consists of a conventional 8-in. inner bore for the production fluid
and a 2.5-in. bore for gas lift. The production bore utilizes a
standard metal seal to isolate the fluid from the environment. The
Fig. 5 Cross section of the nonoriented, multi-bore connection
44 Õ Vol. 122, MARCH 2000
eccentric 2.5-in. bore is isolated from both the external environ-
ment and the production bore through the use of two metal seals
which are concentric to the inner production seal. The two metal
seals for the gas lift line form an annular cavity or ‘‘gallery’’ into
which the gas will flow through prior to exiting the corresponding
mating hub. A flow analysis was performed in order to optimize
the geometry associated with the transition of the 2.5-in. line into
the gallery area as well as the gallery area itself.
The three metal seals used on the multibore connection system
are held together in an assembly which allows the sealing system
to be removed as a single unit 共see Fig. 6兲. The assembly, starting
from the inside, consists of the production seal which is directly
attached to the inner gallery seal through a simple breach-locking
system. The breach-lock system consists of three tabs which pro-
trude from the outside diameter of the inner gallery seal and
which mate with three corresponding slots cut on the rib of the
seal ring 共see Fig. 6兲. Once the tabs are engaged, the seal-rings are
rotated 180 deg relative to each other to interlock them. Next, an
intermediate seal-plate is used to attach the inner gallery seal to
the outer gallery seal. This seal plate incorporates three slots on
the face of the plate to allow the gas coming from the 2.5-in. line
to the gallery area to cross-over to the adjacent hub. Finally, an
outer seal plate is used to secure the outer gallery seal to the plate
which retains the assembly in the hub. To secure the entire seal-
plate assembly to the hub, the outer seal plate incorporates several
beam-type segments around its circumference 共integral to the
Transactions of the ASME
 Fig. 6 General multi-bore seal arrangement
plate兲 which collapse radially during its installation. When the
seal plate reaches the retaining groove in the hub, the beam seg-
ments expand outward to hold the seal plate and seal ring inside
the connector inboard hub. An axial push/pull action is all that is
required to install/remove the seal plate assembly from the hub. In
addition, elastomer face seals are used on both sides of the seal
plate to create a pressure cavity which isolates the metal seal
when the clamp is fully made-up. The external pressure test,
which is performed through an ROV hot stab coming from the
hub, injects a pressure of 500 psi into this cavity to ensure that a
metal-to-metal seal has been achieved prior to the retrieval of the
connection tool.
The design of the concentric sealing system required the devel-
opment of two new seals for the gallery area. These seals were
designed to work in conjunction with the existing bore seal. In
order to reliably set the three seals simultaneously, several key
variables in the design of the gallery seals were optimized through
the use of finite element analysis. These variables included the
seal stand-off, the rib thickness, and the radial interference. Dur-
ing connector make-up, it is normal for the hubs to elastically flex
as the initial preload is generated. In a similar fashion, the hubs
respond to the conditions of internal pressure and external load-
ing. What is important is that the hubs remain in contact with the
rib of all three metal seals under all loading conditions while
maintaining the central load path through the inner seal ring. In
addition, the contact band and the bearing pressures between the
seal and seal seat on the hub must be sufficient to effect a metal
seal. To analyze the seal performance, bearing pressure plots were
made which show the contact pressures mapped along the length
of the seal leg. An internal metal seal design criteria is then used
to evaluate the seal performance depicted in the analysis.
The design of the metal sealing system has a direct effect on the
design of the alignment profile used in the multibore hub. The
Table 1 Typical misalignment capabilities
Journal of Energy Resources Technology
alignment profile, which is integral with the hub 共see Fig. 5兲,
provides the misalignment correction during the final pull-in por-
tion of the flowline installation. The performance requirements for
the internal alignment profile are summarized in Table 1 for the
three misalignment conditions.
A three-step profile was developed 共see Fig. 5兲 to accommodate
the alignment requirements. The geometry of the profile was ana-
lyzed mathematically to ensure that the hub would not prema-
turely contact any of the metal seal rings prior to contacting the
alignment profile. The geometric model was examined under both
nominal and worse-case tolerance conditions. For both of these
conditions, the gaps between the outboard hub 共the hub being
pulled in兲 and the seal rings was calculated as the out-board hub
contacted each portion of the alignment profile. The geometric
analysis showed that the worse-case misalignment condition re-
sulted from the axial misalignment condition alone.
Flow Analysis
The flow path associated with the 2.5-in. gas lift line as it trav-
els through the multibore hub gallery area needed to be analyzed
in order to minimize the pressure drop across the connection. The
pressure drop was determined by first partitioning the flow path
into several discrete regions which contribute to the pipe friction
and minor losses. The flow area and wetted perimeter 共see Fig. 7兲
for each region is determined from a solid modeling design pack-
age. Next, using information on the volume flow of the gas, a
calculation is performed to determine whether the flow is subsonic
or supersonic. Once the flow characteristic is known, the pressure
losses due to choking and Mach waves 共for supersonic flow only兲
and for pipe wall friction and flow path discontinuities in each of
the flow path regions can be calculated. The individual head
MARCH 2000, Vol. 122 Õ 45

Fig. 7 Transitional flow regime
losses are then summed in order to determine the total pressure
loss 关1兴. The flow path used to model the pressure drop is sum-
marized as follows:
• a long entrance to the milled slot;
• four sections of the milled slot on the lower hub area 共area
increases with flow, see Fig. 7兲;
• the interhub region;
• four sections of the milled slot on the upper hub 共area de-
creases with flow兲;
• a long exit from the milled slot.
The volumetric flow rate in the gas lift line of 1.97 m3/s was
determined to be subsonic. Knowing this, the only pressure drops
experienced will be due to friction on the transition cone walls
and flow discontinuities. Gravity effects are negligible given the
small changes in elevation. The minor losses are much more dif-
ficult to quantify since they are usually determined experimentally
for a small number of geometric profiles. Estimates of minor
losses are made by selecting conservative minor loss coefficients
for geometry with similar characteristics. For example, the en-
trance and exit regions are modeled as reentrant orifices with the
flow path through the inner hub modeled as a flow expansion. The
flow from the inner hub to the gallery region and from the gallery
region to the outer hub is modeled as a branch flow tee, and the
flow path through the upper hub is modeled as a flow constriction.
The losses which occur in each of these regions are summed to
determine the total pressure head loss due to minor losses. The
final iteration of the gas lift geometry produced a change in the
gas lift pressure of approximately 6.5 psi. This small drop in pres-
sure across the connection means that there is little energy loss for
the gas lift operations.
Finite Element Analysis
Table 2 provides the load cases that were analyzed using finite
element techniques to determine the structural integrity of the de-
sign per API 6A 关2兴.
To determine the design loads for the connector, the equivalent
tension method is used. The design loads are defined by the fol-
lowing equation:
4M ␲ ␲ hoop stiffness to simulate the hoop discontinuity present in the
T DESIGN⫽T⫹ ⫹ P i •SD2 ⫺ P e •d 2o
do 4 4
T DESIGN ⫽ total coupling design tension
T ⫽ applied tension
M ⫽ bending moment
do ⫽ outer diameter of pipe
Pi ⫽ internal pressure
46 Õ Vol. 122, MARCH 2000
Table 2 Design cases considered
P e ⫽ external pressure
SD ⫽ sealing diameter
The allowable stresses for strength evaluation of the connector
are based on the API 6A design criteria. The allowable stress
methodology involves separating the total stress at a cross section
into stress categories. The stress distribution across a section of an
axisymmetric model can be categorized into three types of
stresses: net section stress, local bending stress, and local peak
stress. Local peak stresses are primarily used for fatigue life
evaluation. The strength evaluation involves primarily the net sec-
tion stresses and local bending stresses. Linearized local bending
stress through a section at a given direction is calculated for the
same local bending moment generated from the actual component
stress contribution. Local peak stresses are caused by stress con-
centrations at geometric discontinuities. The local peak stress
combined with local bending stress and net section stress forms
the total actual stress at a section. For static strength evaluation,
the linearized component stresses are used to calculate linearized
von Mises stress, and stress intensity. These linearized combined
stresses are then checked against the allowable stresses set forth in
the appropriate design code. For the evaluation of the connection,
the allowable stresses are summarized as follows:
For design
␴ m ⭐ 32 ␴ y
␴ m⫹b ⭐ ␴ y
For test
␴ m⭐ ␴ y
␴ m⫹b ⭐1.35␴ y
where
␴ m ⫽ membrane stress
␴ b ⫽ bending stress
␴ m⫹b ⫽ membrane plus bending stress
The finite element models were built with PATRAN. The finite
element analyses was performed using ANSYS and ABAQUS.
ANSYS was used to verify the structural integrity of the equip-
ment, while ABAQUS was used for its nonlinear capabilities to
design the gallery seals. The finite element analyses consisted of
two-dimensional models. The clamp was modeled with reduced
three-piece clamp. Two-dimensional interface/gap elements were
used to model the interface between the hubs, seals, and clamp.
Tensile loads were applied to the model using negative pipe pres-
sure on the pipe wall at the in-board hub end, while restrained at
the outboard hub end in the axial direction only.
The finite element models for ABAQUS are identical to
ANSYS, except that the contact at the seal interfaces was modeled
Transactions of the ASME
using slide line elements in lieu of gap elements, allowing for
more accurate modeling of the bearing pressures and the frictional
effects at these locations. All load cases in the model included
static friction between the mating interfaces using a static friction
coefficient range of 0.07 to 0.18.
There are three load cases that are of primary concern for the
connection evaluation. They are 1兲 preload only, 2兲 working con-
ditions 共design conditions兲, and 3兲 test pressure loading. The first
nonoriented multibore connection was designed to:
• Preload only⫽1,453,000 lb
• Preload plus working pressure
• Internal pressure⫽2320 psi
• Gas lift 3000 psi for
• End load⫽467,326 lb
• Preload plus test pressure load
• Internal pressure⫽3480 psi
• Gas lift 4500 psi for
• End load⫽465,283 lb
A set of secondary load cases were applied to predict the per-
formance of the multi-bore hubs under various operating condi-
tions. These include working pressure in bore without pressure in
gas lift sector, and gas lift pressure with no bore pressure. Both
the secondary and primary load cases were analyzed for three
different tolerance variations on the seal rib thickness, given as
follows:
• Standoff at gallery seals after preload—dimensional mini-
mum
• Standoff at gallery seals after preload—dimensional nominal
• Standoff at gallery seals after preload—dimensional maxi-
mum
Through the rigorous analyses performed, the multi-bore con-
nection was optimized to meet all anticipated loading conditions.
Prototype Testing Program
The extended factory acceptance test 共EFAT兲 qualification for
the 8⫻2.5-in. multi-bore connection involves a hydrostatic test at
ambient temperature per API 6A, Hydrostatic Body Test—
Individual Equipment, and a gas test at ambient temperature per
API 6A, 关2兴 Gas Body Test—Individual Equipment. The accep-
tance criteria for the two tests are per Appendix F in API 6A, 16th
Edition. This testing suite is performed to produce a load condi-
tion worse than any anticipated conditions in operation. In addi-
tion to the testing program, the following additional tests were
performed:
• Back Pressure Test. This is designed to verify the sealing
integrity of the elastomer back-pressure test seals 共which create
the pressure cavity兲 and test circuit. In this test, 500 psi of nitro-
gen gas is introduced in the cavity area behind each metal seal.
The connector assembly, which is submerged in a water tank, is
monitored for bubbles. The connector passes this test only if no
bubbles are observed for the duration of the holding period. This
test shows that sufficient bearing pressure has been generated by
the metal seals and that there are no substantial flaws in the seal/
seat surfaces.
• Seal Plate Removal Load Test. This is used to confirm the
calculated load required to remove the seal-plate assembly from
the inboard hub. In this test, a calibrated load cell is attached to an
overhead crane which pulls the seal plate assembly from the hub.
The load test is performed 10 times.
• Alignment Verification Test. This is used to verify the maxi-
mum anticipated geometric misalignment condition between the
Journal of Energy Resources Technology
two mating hubs. For this test, the inboard hub 共IBH兲 is placed
face up on a level surface with a shim placed on its base to rotate
the hub by 1.5 deg. The outboard hub 共OBH兲 is then lowered over
the IBH. The OBH is positioned over the IBH to produce the
maximum axial misalignment condition. The OBH is then low-
ered down over the IBH and the funneling action of the alignment
profiles observed. A successful test aligns the OBH to the IBH
with no premature contact or interference between the outboard
hub and the metal seals during the operation.
• Misalignment Test. This is used to evaluate the misalignment
correction feature in the multibore hubs under load. In this test the
IBH is positioned horizontally, as it will be in the field. The OBH,
which is supported by a crane, is loaded with a counterweight to
produce a bending moment. The OBH is then brought into the
IBH until both alignment profiles come into contact. The connec-
tor is then closed and the moment correction and final alignment
sequences are observed. Once the connector has been preloaded,
the connector is then pressurized to verify that the seals have
energized properly.
Conclusions
This paper addresses the development of a multiple passageway
connection which precludes the requirement for rotational orien-
tation. It has been shown that this new development can greatly
simplify the operational tasks required to perform a remote, diver-
less connection. The following benefits can be realized using this
technology:
• reduction in the number of connectors required for a given
field application;
• enhanced seal reliability through remote verification of the
metal sealing system and remote seal placement/retrieval;
• true operational benefits of combining multiple tubulars into
one connection;
• reduction in the pull-in system hardware complexity and cost;
• reduced overall project risk due to simplified operations and
reduced weather window requirements.
This technology can be used in a variety of applications includ-
ing sales/gas lift combinations, dual sales, multiple passageway
mid-line connections, and sales/control combinations. It is also
planned to expand this technology to connect three, four, and five
passageways simultaneously without rotational orientation.
It is hoped that the aforementioned technologies will continue
to reduce overall system costs while providing enhanced reliabil-
ity and safety for future developmental fields and projects.
Acknowledgments
• Shyam Kumar, ABB Vetco Gray, Houston, Texas. Performed
classical flow analysis of multi-bore hub system.
• Justin Whitehead, ABB Vetco Gray, Houston, Texas. Per-
formed editing and technical publishing.
• ABAQUS is a registered trademark of Hibbitt, Karlsson, and
Sorensen, Inc.
• ANSYS is a registered trademark of Swanson Analysis Sys-
tems, Inc.
• PATRAN is a registered trademark of MacNeal Schwendler
Corporation.
References
关1兴 Kumar, S., 1996, ‘‘Multibore Flow Analysis,’’ ABB Vetco Gray Internal Re-
port.
关2兴 API 6A, 1989, 16th Edition, ‘‘Specification for Wellhead and Christmas Tree
Equipment,’’ API Production Department, Dallas, TX, pp. 49–59.
MARCH 2000, Vol. 122 Õ 47