Multiphase Flow Design Guide

PART I - MULTIPHASE PIPELINE & SLUG CATCHER DESIGN GUIDE

CPTC NOVEMBER 1994 1

Multiphase Pipeline Design Guide



PART I

TABLE OF CONTENTS

SECTION 1.0 - INTRODUCTION1.1 Objective and Scope..................................................................................................................................................... 11.2 Definition of Terms........................................................................................................................................................ 1

SECTION 2.0 OVERVIEW OF MULTIPHASE FLOW FUNDAMENTALS2.1 Design Criteria.............................................................................................................................................................. 11

2.2 Velocity Guidelines ....................................................................................................................................................... 11

2.3 Flow Regimes............................................................................................................................................................... 13

2.4 Pressure Gradient ......................................................................................................................................................... 16

2.4.1 Frictional Losses .......................................................................................................................................... 16

2.4.2 Elevational Losses........................................................................................................................................ 17

2.4.3 Acceleration Losses...................................................................................................................................... 18

2.4.4 Allowable Pressure Drop............................................................................................................................... 20

2.5 Pressure Gradient Calculations...................................................................................................................................... 20

2.6 Section Highlights......................................................................................................................................................... 21

SECTION 3.0 STEADY STATE DESIGN METHODS3.1 Pipeline Design Methods............................................................................................................................................... 25

3.2 Steady State Simulators................................................................................................................................................ 26

3.2.1 Phase Equilibrium and Physical Properties.................................................................................................... 26

3.2.2 Pipeline Elevation Profile .............................................................................................................................. 28

3.2.3 Heat Transfer ............................................................................................................................................... 30

3.2.4 Recommended Methods for Pressure Drop, Liquid Holdup, and

Flow Regime Prediction................................................................................................................................ 33

3.2.5 Interpretation of Results................................................................................................................................ 35


SECTION 4.0 TRANSIENT FLOW MODELING4.1 Transient Flow Modeling (General) ................................................................................................................................ 414.2 Use of Transient Simulators........................................................................................................................................... 42




SECTION 5.0 SLUG FLOW ANALYSIS5.1 Slug Flow (General) ...................................................................................................................................................... 455.2 Slug Length and Frequency Predictions......................................................................................................................... 46

5.2.1 Hydrodynamic Slugging................................................................................................................................ 46

5.2.2 Terrain Slugging........................................................................................................................................... 51

5.2.3 Pigging Slugs............................................................................................................................................... 53

5.2.4 Startup and Blowdown Slugs........................................................................................................................ 55

5.2.5 Rate Change Slugs ...................................................................................................................................... 56

5.2.6 Downstream Equipment Design for Slug Flow............................................................................................... 56


SECTION 6 EXAMPLE PROBLEMS6.1 Example Problem 1 Low Gas/Oil Line Between Platforms .......................................................................................... 63

6.1.1 Line Size...................................................................................................................................................... 65

6.1.2 Slug Length Prediction ................................................................................................................................. 75

6.1.3 Slug Frequency and Length by Hill & Wood Method ...................................................................................... 80

6.2 Example Problem 2 Gas Condensate Line .................................................................................................................. 88

6.2.1 Table 1, Wellstream Composition ................................................................................................................. 89

6.2.2 Table 2, Pipeline Evaluation Profile ............................................................................................................... 90

6.2.3 Pipeline Simulation Comparison ................................................................................................................... 92

SECTION 7.0 REFERENCES .................................................................................................................................................... 106

FIGURES

I: 1-1 Flow Regimes in Horizontal Flow................................................................................................................................... 8

I: 1-2 Flow Regimes in Vertical Flow ...................................................................................................................................... 9

I: 2-1 Horizontal Flow Regime Map......................................................................................................................................... 23

I: 2-2 Vertical Flow Regime Map............................................................................................................................................. 24

I: 5-1 Taitel-Dukler Liquid Holdup Predictions.......................................................................................................................... 60

I: 5-2 Stages in Terrain Slugging ............................................................................................................................................ 61

I: 5-3 Pipeline Slugging.......................................................................................................................................................... 62

I: 6-1 Liquid Holdup for Example 1, Year 12............................................................................................................................ 101

I: 6-2 Inlet Pressure for Example 1, Year 12............................................................................................................................ 102

I: 6-3 Liquid Flowrate Out of Line, Example 1, Year 12............................................................................................................ 103

I: 6-4 Gas Flowrate Out of Line, Example 1, Year 12............................................................................................................... 104



I: 6-5 Liquid Holdup Predictions for Example 2........................................................................................................................ 105



SECTION 1.0 - INTRODUCTION

1.1 Objective and Scope

The simultaneous flow of gas and liquid through pipes, often referred to as multiphaseflow, occurs in almost every aspect of the oil industry. Multiphase flow is present in welltubing, gathering system pipelines, and processing equipment. The use of multiphasepipelines has become increasingly important in recent years due to the development ofmarginal fields and deep water prospects. In many cases, the feasibility of a designscenario hinges on cost and operation of the pipeline and its associated equipment.

Multiphase flow in pipes has been studied for more than 50 years, with significantimprovements in the state of the art during the past 15 years. The best available methodscan predict the operation of the pipelines much more accurately than those available onlya few years ago. The designer, however, has to know which methods to use in order toget the best results.

Part I of this guide consists of seven sections. The fundamentals of multiphase flow inpipelines are discussed in Section 2.0. The third section describes the use of steady statesimulation methods. This section of the guide helps the designer choose the best methodsfor the project, and it gives guidelines to use in designs. The fourth section of the reportbriefly describes transient flow modeling. The fifth section describes slug flow modeling,giving suggestions on the best methods to use in slug flow simulation. The sixth sectionincludes two sample problems, based on actual designs, which illustrate the design stepsused in analyzing the pipeline designs.

1.2 Definition of Terms

In discussing the design of multiphase pipelines, it is necessary to define several termsused repeatedly throughout this text.

Near Horizontal and Near Vertical Angles

The term "near horizontal" is used in this guide to denote angles of -10 degrees to +10



degrees from horizontal. The term "near vertical" denotes upward inclined pipes withangles from 75 to 90 degrees from horizontal.

Flow Regimes

In multiphase flow, the gas and liquid within the pipe are distributed in severalfundamentally different flow patterns or flow regimes, depending primarily on the gas andliquid velocities and the angle of inclination. Observers have labeled these flow regimeswith a variety of names. Over 100 different names for the various regimes and sub-regimes have been used in the literature. In this guide, only four flow regime names willbe used: slug flow, stratified flow, annular flow, and dispersed bubble flow.

Figure I:1-1 shows the flow regimes for near horizontal flow, and Figure I:1-2 shows theflow regimes for vertical upward flow. Descriptions of the flow regimes

1. Stratified Flow - at low flowrates in near horizontal pipes, the liquid and gas separateby gravity, causing the liquid to flow on the bottom of the pipe while the gas flowsabove it. At low gas velocities, the liquid surface is smooth. At higher gas velocities,the liquid surface becomes wavy. Some liquid may flow in the form of liquid dropletssuspended in the gas phase. Stratified flow only exists for certain angles of inclination.It does not exist in pipes that have upward inclinations of greater than about 1 degree.Most downwardly inclined pipes are in stratified flow, and many large diameterhorizontal pipes are in stratified flow. This flow regime is also referred to as stratifiedsmooth, stratified wavy, and wavy flow by various investigators.

2. Annular Flow - at high rates in gas dominated systems, part of the liquid flows as afilm around the circumference of the pipe. The gas and remainder of the liquid (in theform of entrained droplets) flow in the center of the pipe. The liquid film thickness isfairly constant for vertical flow, but it is usually asymmetric for horizontal flow due togravity. As velocities increase, the fraction of liquid entrained increases and the liquidfilm thickness decreases. Annular flow exists for all angles of inclinations. Most gasdominated pipes in high pressure vertical flow are in annular flow. This flow regime isreferred to as annular-mist or mist flow by many investigators.



3. Dispersed Bubble Flow - at high rates in liquid dominated systems, the flow is a frothymixture of liquid and small entrained gas bubbles. For near vertical flow, dispersedbubble flow can also occur at more moderate liquid rates when the gas rate is verylow. The flow is steady with few oscillations. It occurs at all angles of inclination.Dispersed bubble flow frequently occurs in oil wells. Various investigators havereferred to this flow regime as froth or bubble flow.

4. Slug Flow - for near horizontal flow, at moderate gas and liquid velocities, waves onthe surface of the liquid may grow to sufficient height to completely bridge the pipe.When this happens, alternating slugs of liquid and gas bubbles will flow through thepipeline. This flow regime can be thought of as an unsteady, alternating combinationof dispersed bubble flow (liquid slug) and stratified flow (gas bubble). The slugs cancause vibration problems, increased corrosion, and downstream equipment problemsdue to its unsteady behavior.

Slug flow also occurs in near vertical flow, but the mechanism for slug initiation isdifferent. The flow consists of a string of slugs and bullet-shaped bubbles (calledTaylor bubbles) flowing through the pipe alternately. The flow can be thought of as acombination of dispersed bubble flow (slug) and annular flow (Taylor bubble). Theslugs in vertical flow are generally much smaller than those in near horizontal flow.

Slug flow is the most prevalent flow regime in low pressure, small diameter systems.In field scale pipelines, slug flow usually occurs in upwardly inclined sections of theline. It occurs for all angles of inclination. Investigators have used many terms todescribe parts of this flow regime. Among them are: intermittent flow; plug flow;pseudo-slug flow, and churn flow.

Superficial Velocities

The velocities of the gas and liquid in the pipe are prime variables in the prediction of thebehavior of the multiphase mixture. Most multiphase flow prediction methods use thesuperficial gas and liquid velocities as correlating parameters. The superficial velocitiesare defined as the in situ volumetric flowrate of that phase divided by the total pipe cross-sectional area. In oil field units, this corresponds to:



Vsg = Superficial Gas Velocity, ft/sec

= (actual ft3/sec of gas) / (pipe cross-sectional area, ft2)

Vsl = Superficial Liquid Velocity, ft/sec

= (actual ft3/sec of liquid) / (pipe cross-sectional area, ft2)

Mixture Velocity

The mixture velocity (Vm) is the volumetric average velocity of the gas-liquid mixture. Itis equal to the sum of the gas and liquid superficial velocities.

V V Vm sg sl= +

Slip and Liquid Holdup

Liquid holdup is defined as the volume fraction of the pipe that is filled with liquid. It isthe most important parameter in estimating the pressure drop in inclined or vertical flow.It is also of prime importance in sizing downstream equipment, which must be able tooperate properly when the liquid holdup in the line changes because of pigging or ratechanges.

If there was no slip between the gas and liquid phases, both phases would move throughthe pipe at the mixture velocity. The liquid would occupy the volume fraction equivalentto the ratio of the liquid volumetric flowrate to the total volumetric flowrate. Inmultiphase flow terminology, this equates to the liquid holdup being equal to the ratiobetween the superficial liquid velocity and the mixture velocity:

Hlns = No-slip liquid holdup

= Vsl / Vm

Under most conditions, however, the liquid phase, which is more dense and viscous,moves more slowly than the gas. When this occurs, the liquid holdup (Hl) is greater thanthe no-slip holdup.



H Hl s> ln

Under these conditions, the actual gas velocity is greater than the mixture velocity, andthe actual liquid velocity is smaller than the mixture velocity. The expressions for theactual gas velocity (Ug) and actual liquid velocity (Ul) are:

UV

gsg

l=

1 H

UVHl

sll

=

For small diameter, low pressure piping, there is frequently a vast difference between Ugand Ul. For field piping, there is generally less slip between the phases, and the flow mayapproximate no-slip flow in dispersed bubble and annular flows.

It is possible to get conditions where the liquid holdup is less than no-slip, but this onlyoccurs over a small range of flowrates in downwardly inclined pipes.

Pressure Gradient

Two definitions of the term "pressure gradient" are used in the literature. In this guide, theterm "pressure gradient" will be used to describe the pressure drop per unit length of pipe,(Pin - Pout)/L. In many papers, the term "pressure gradient" is used to denote the pressurechange per unit length (dp/dx = (Pout - Pin)/L). The magnitude of the pressure gradient isthe same in either definition, but the sign of the pressure drop per unit length is usuallypositive, while the sign of dp/dx is usually negative. Most people prefer to work withpositive numbers, so the majority of people use the pressure drop per unit lengthdefinition. The choice of the definition is somewhat arbitrary, but it should be noted whenreading the multiphase flow literature, and working with some of the available software.

3-Phase Flow vs. 2-Phase Flow

In most of this guide, the discussion will consider 2-phase flow, or gas-liquid flow. In themajority of oil field applications, there will actually be 3 phases present (gas, oil, andwater). The rigorous prediction of 3-phase flow is in its infancy. 3-Phase flow methods



are not generally available, so most simulators use 2-phase models with a mixed liquidstream using averaged properties for the oil and water. The use of 2-phase models withaveraged properties generally gives acceptable results unless either: emulsions are present;or the flowrates are low enough to cause stratification of all three phases. These problemsare discussed in more depth in Section 3.2.1

Mechanistic Models vs. Correlations

The prediction of multiphase flow behavior has improved considerably during the 50+years that the subject has been studied. For many years, multiphase flow predictionmethods were correlations, based on curve fits of experimental data. The correlationsfrequently used arbitrarily selected variables and were based on limited databases,consisting almost entirely of low pressure, small diameter data. Extrapolations of theseprediction methods to field conditions frequently proved to be seriously in error. In the1960s and 1970s, several investigators undertook experimental studies to try tounderstand the fundamental mechanisms of the various flow regimes. In the past 15 yearsmodels have been developed, which are based on simulation of these mechanisms. Thesemodels, referred to as mechanistic models, have proven to extrapolate best to fieldconditions.

Newtonian vs. Non-Newtonian Fluids

Most condensates and crude oils follow Newtons law of viscosity, which is defined as:

yxxdv

dy=

where yx = shear stress

= viscosityvx = velocity

y = distance

Some liquids, however, exhibit behavior that is very different from Newton's law. Thesefluids are referred to as non-Newtonian. In the oil field, examples of non-Newtonian fluidsare drilling muds, polymeric additives, and crude oils below their cloud point.



Flowline simulators are based on Newtonian fluids. If a non-Newtonian liquid is present,the simulator must be tricked into giving a Newtonian viscosity equivalent to the actualviscosity at the given temperature and shear stress. The methods of doing this are beyondthe scope of this guide.



Figure I:1-1 Flow Regimes in Horizontal Flow



Figure I:1-2 Flow Regimes in Vertical Flow



SECTION 2.0 - OVERVIEW OF MULTIPHASE FLOW FUNDAMENTALS

2.1 Design Criteria

The majority of lines are sized by use of three primary design criteria: available pressuredrop; allowable velocities; and flow regime. In some cases, a more optimal line size maybe found that better suits the overall design of the pipeline system. These considerationswill be discussed later in the transient modeling section of the guide. A description of eachof the primary design criteria follows in Sections 2.2, 2.3, and 2.4.

2.2 Velocity Guidelines

The velocity in multiphase flow pipelines should be kept within certain limits in order toensure proper operation. Operating problems can occur if the velocity is either too highor too low, as described in the following sections.

It is difficult to accurately define the point at which velocities are "too high" or "too low".This section of the guide will try to quantify limits, but these limits should be consideredas guidelines and not absolute values.

Maximum Velocity

For the maximum design velocity in a pipeline, API RP-14E recommends the followingformula:

VC

ns

max = (Eqn. 2.1)

where Vmax = Maximum mixture velocity, ft/sec

ns = No-slip mixture density, lb/ft3

=

( ) ( ) g sg l slm

V VV+

g = Gas Density, lb/ft3



l = Liquid Density, lb/ft3

C = Constant, 100 for continuous service, 125 for intermittent service.

This equation attempts to indicate the velocity at which erosion-corrosion begins toincrease rapidly. Many people think this equation is an oversimplification of a highlycomplex subject, and as a result, there has been considerable controversy over its use.For wells with no sand present, values of C have been reported to be as high as 300without significant erosion/corrosion. For flowlines with significant amounts of sandpresent, there has been considerable erosion-corrosion for lines operating below C = 100.

The use of the API equation has been the subject of several research projects. It has beengenerally agreed that the form of the equation is not sophisticated enough, and shouldinclude additional parameters. Unfortunately, no other equation has been proposed whichhas gained acceptance in the industry as an alternative to the API equation. As a result,the recommended maximum velocity in the pipeline is the value calculated from Equation2.1 with a C value of 100.

It should be noted that Equation 2.1 is also used by many people as an estimate of themaximum velocity for noise control.

For additional guidance on the use of the API equation, refer to Chevrons Piping Manual.

Minimum Velocity

The concept of a minimum velocity for the pipeline is an important one and should beconsidered in the design of the line. Turndown conditions frequently govern the design ofthe downstream equipment. Velocities that are too low are frequently a greater problemthan excessive velocities, so that the designers natural tendency to add "a bit of fat" to thedesign by increasing pipe diameter can cause severe problems in the operation of the lineand the downstream facilities.

At low velocities, several operating problems may occur:

a) Water may accumulate at low spots in the line. If there is an appreciable amount ofCO2 or H2S in the well stream, this water may be very corrosive.



b) Liquid holdup may increase rapidly at low mixture velocities. A large accumulation ofliquid may cause problems in downstream separators or slug catchers if the line ispigged or the rate is changed rapidly.

c) Low velocities may cause terrain induced slugging in hilly terrain pipelines andpipeline-riser systems.

It isnt possible to give a simple formula quantifying the velocity when the phenomenadiscussed above will occur. The minimum velocity depends on many variables, including:topography; pipeline diameter; gas-liquid ratio; and operating conditions of the line. Aball-park value for the minimum velocity would be a mixture velocity of 5-8 ft/sec. Theactual value of the minimum velocity can only be quantified by simulation of the systemusing the methods discussed in Section 5.2.2.

2.3 Flow Regimes

As discussed in Section 1, the gas and liquid in the pipe are distributed differently in eachof the four major flow regimes (stratified, annular, slug, and dispersed bubble flows). Theprediction of the correct flow regime is important for several reasons. The flow regimeprediction can show whether the line will operate in a stable flow regime or an unstableregime. The prediction of liquid holdup and pressure drop are highly dependent on theflow regime, with each regime exhibiting different behavior when the design variables arechanged.

The transitions between the flow regimes are frequently depicted in a flow regime map,such as that shown in Figure I:2-1. The flow regime map typically has the superficial gasvelocity (Vsg) on the X-axis and the superficial liquid velocity (Vsl) on the Y-axis. Asdiscussed later in this section, the flow regime map is only valid for a single point in thepipeline. As the angle of inclination, pressure and temperature change with position in thepipeline, the flow regime map also changes.

Some general comments, however, can be made about the flow regime transitions.Stratified flow occurs at low superficial gas and liquid velocities. Dispersed bubble flowoccurs at high superficial liquid velocities. Annular flow occurs at high superficial gas



velocities. Slug flow occurs at moderate superficial gas and liquid velocities. Figure I:2-2shows a typical flow regime map for vertical flow.

Many experimental studies of the transitions between the flow regimes for various systemshave been made, and many flow regime transition prediction methods have been published.Some of these methods work fairly well, but most are poor. The designer needs tocarefully choose the method that will work best for the set of conditions. The bestmethods are discussed in the remainder of this section.

Experimental studies of flow regime transitions have shown that each of the flow regimeboundaries reacts differently to changes in the system variables. The following table showsthe sensitivity of the transitions to changes in the major system variables:

Transition

Variable

Slug toDispersedBubble

Slug toAnnular

Slug toStratified

Stratified toAnnular

Angle ofInclination

Small Effect ModerateEffect

Strong Effect Strong Effect

Gas Density Small Effect Strong Effect Strong Effect Strong Effect

PipelineDiameter

Small Effect Small Effect Strong Effect ModerateEffect

Liquid PhysicalProperties

ModerateEffect

Small Effect ModerateEffect

ModerateEffect

Many people have attempted to develop simple flow regime maps, usually using somearbitrary dimensionless parameter on each axis (e.g. Baker, Beggs & Brill). Thesemethods are inherently inaccurate since no single parameter can model the sensitivityeffects shown in the previous table. The only flow regime map prediction methods thathave been effective for a wide range of conditions are those using mechanistic models toestimate the flow regime transitions.



In 1976, Taitel and Dukler published a landmark article describing a method of predictingflow regime transitions by modeling the mechanism of each transition. By modeling eachtransition, this method can show the same type of behavior observed in the experimentalwork. The original Taitel-Dukler paper covered flow regime transitions in near horizontalflow only, and one of the transitions (slug-dispersed bubble) is very much in error. Taiteland his co-workers at the University of Tel Aviv have subsequently published severalarticles that expand the range of angles of inclination and correct the errors in the originalpaper. The Taitel-Dukler paper and the latest paper from Tel Aviv model flow regimetransitions for all angles of inclination.

The Taitel, et al. methods give reasonably good predictions of the various flow regimetransitions, and the accuracy of the predictions has improved with each revision.

Another approach to the modeling of flow regime transitions is the method used in theOLGAS method. It employs mechanistic models of each flow regime and links the modelsby the assumption that the flow regime giving the lowest liquid holdup is the correct one.This assumption holds up well in practice. The OLGAS method predicts flow regimetransitions with similar accuracy to the Taitel, et al. models.

Within Chevron, there are several programs available for flow pattern prediction.Pipephase will print a flow regime map based on the Taitel-Dukler method for nearhorizontal flow and the Taitel-Dukler-Barnea model for near vertical flow. Unfortunately,these methods are the oldest and weakest of this family of methods. Two programs areavailable within CPTC that incorporate the latest versions of the Taitel, et al. models.These programs are FLOPAT, developed by Tulsa University, and FLEX, developed byAdvance Multiphase Technology. CPTC should be consulted if it is desired to use theseprograms.

As in many aspects of multiphase flow, the flow regime prediction methods are not exact.Errors of +/- 25% for the transition velocities are typical, even for the best predictionmethods. If the Taitel-Dukler map is used, the designer should be aware of the grosserrors in the slug to dispersed bubble transition. The errors for this transition can be1000%. The dispersed bubble to slug transition typically occurs at a superficial liquidvelocity of about 10 ft/sec. Taitel-Dukler frequently predicts this transition velocity to be50-100 ft/sec.



2.4 Pressure Gradient

In most pipelines, the pipeline diameter is determined by the allowable pressure drop in theline. The overall pressure gradient is composed of three additive elements:

a) pressure drop due to friction;b) pressure changes due to elevational effects;c) accelerational losses.

The calculation of the constituent parts of the pressure gradient will be discussed in thenext three sections.

The Chevron Fluid Flow Manual contains a good discussion of these pressure loss termsfor single phase flow and can be consulted as a reference.

2.4.1 Frictional Losses

In multiphase flow, frictional losses occur by two mechanisms: friction between the gas orliquid and the pipe wall; and frictional losses at the interface between the gas and liquid.The friction calculations, therefore, are highly dependent on the flow regime, since thedistribution of liquid and gas in the pipe changes markedly for each regime.

In stratified flow, there is wall friction between the gas and the pipe wall at the top of thepipe, and wall friction between the liquid and the wall at the bottom of the pipe. There isalso friction between the gas and liquid at the gas-liquid interface. The interfacial frictioncan be similar in magnitude to the wall friction if the interface is smooth, or it can beconsiderably higher if waves are present.

In annular flow, there is friction between the liquid film and the wall. There is alsoconsiderable interfacial friction between the gas in the core of the pipe and the liquid film.The interfacial friction is usually the larger component.

In dispersed bubble flow, friction occurs between the liquid and the wall. There isnegligible interfacial friction between the gas and liquid.



Slug flow has several frictional components. In the slug, the friction losses are caused bythe friction between the liquid and the pipe wall. In the gas bubble, the frictionalcomponents are the same as in stratified flow, namely gas and liquid friction with the pipewalls and interfacial friction between the gas and liquid. The friction loss in the slug isusually much higher than the losses in the bubble.

2.4.2 Elevational Losses

Elevational losses may be the major pressure loss component in vertical flow and flowthrough hilly terrain. The calculation of elevational losses is governed by the followingequation:

dpdx

elev

= mix

c

g sin144g

where: (dp/dx)elev = Pressure gradient due to elevation, psi/ftmix = Mixture Density, lb/ft3

= (l) (Hl) + (g) (1-Hl)Hl = Liquid Holdup

g = Acceleration due to gravity, 32.2 ft/sec2

= Angle of inclination

gc = Gravitational conversion factor, 32.2 lb-ft/(lbf-sec2)

In order to calculate the elevational gradient, the liquid holdup must be determined. Theholdup in each flow regime has its own sensitivity to the important operating variables. Asummary of the effect of the major operating variables on the liquid holdup is:



Slug Flow AnnularFlow

Stratified Flow DispersedBubble Flow

Superficial GasVelocity

Strong Strong Strong Strong

Superficial LiquidVelocity

Strong Strong Strong Strong

Gas Density Moderate Strong Strong None

Pipeline Diameter Moderate Weak Weak Weak

Angle ofInclination

Moderate Weak Very Strong None

Liquid Properties Moderate Moderate Moderate Weak

As seen in the previous table, the influence of the major variables on the holdup is verydifferent for each of the flow regimes. As a result, it is impossible to develop a generalholdup correlation that will apply to all the flow regimes. Unfortunately, almost all of thecommonly used holdup methods available in commercial software try to do this. Theywork poorly over much of the operating range as a result. The only way to accuratelypredict liquid holdup is to use mechanistic models for each of the flow regimes. Theaccuracy of available holdup methods is discussed further in Section 3.2.4.

2.4.3 Acceleration Losses

Although acceleration losses are present for all flow regimes, they are only significant fortwo flow regimes: annular flow and slug flow. The mechanisms for the losses in these twoflow regimes are very different and will be discussed separately.

In single phase flow, acceleration losses can be calculated from Bernoullis equation.Acceleration losses represent the change in kinetic energy as the fluid flows down thepipe. The expression for acceleration gradient is:

dpdx

Vg

dVdxaccel c

=

144



where: = Density, lbm/ft3

V = Velocity, ft/sec

For multiphase flow, the same type of relationship holds except that it refers to the flowof the mixed phase fluid. Most methods assume a no-slip mixture and use the no-slipmixture density (ns) and the mixture velocity (Vm) in the calculation of accelerationlosses.

The kinetic energy acceleration losses are small for most oil industry applications. Themain exception is high velocity flow through low pressure piping. Flare systems would bean example of piping that has high acceleration losses. Acceleration may account for 30-50% of the overall pressure loss in such lines. For a typical high pressure gathering systemline, acceleration is usually less than 1% of the total drop and is frequently ignored.

Please note that the present version of Pipephase, 6.02, does not properly account foracceleration losses, and, as a result, is not suitable for use in flare system design.

In slug flow, another source of acceleration contributes significantly to the total pressuredrop. As a slug propagates down the pipeline, it overruns and entrains the slower movingliquid from the film ahead of the slug front. Accelerating the liquid from the film velocityto the slug velocity can produce significant pressure losses. The acceleration loss may beanywhere from



flowrate, and composition that the gathering system must handle during the life ofthe field.

c) If it isnt feasible to do the rigorous simulations for a gathering system, the allowablepressure drop can be estimated from the initial wellhead pressure and the processingplant inlet separator pressure. A rule of thumb to use for this method is to take 1/3of the difference between the wellhead pressure and the separator pressure as theallowable pressure drop in the pipeline. The remainder of the difference would equalthe initial choke pressure drop. This approach would allow for future operation atreduced reservoir pressures.

d) A rule of thumb estimate of allowable pressure drop for long distancegas/condensate pipelines is to allow 10-20 psi per mile for frictional pressure drop atdesign rates.

2.5 Pressure Gradient Calculations

As indicated in sections 2.4.1 to 2.4.3, the calculation of the pressure gradient formultiphase flow is very complicated. Hundreds of methods have been proposed to predictpressure drops, but only a few methods work well over a wide range of conditions. Thebest available methods are discussed in Section 3.2.4.



2.6 Section Highlights

Points to remember from Section 2.0 - No other equation has gained acceptance in the industry like the API equaton. The

recommended maximum velocity in the pipeline is the value calculated from Equation 2.1with a C value of 100.

The Taitel et al. Methods give reasonably good predictions of the various flow regimetransitions. The accuracy of the predictions has improved with each revision.

The OLGAS method predicts flow regime transitions with similar accuracy to the Taitel etal. methods.

If the Taitel-Dukler map is used, the designer should be aware of the gross errors in theslug to dispersed bubble transiton.

Overall pressure gradient is composed of three additive elements: pressure drop due to friction pressure changes due to elevational effects accelerational losses

Frictional calculations are highly dependent on the flow regime, since the distribution ofliquid and gas in the pipe changes markedly for each regime.

Elevational losses may be the major pressure loss component in vertical flow and flowthrough hilly terrain.

Using mechanistic models for each flow regime is the only way to accurately predict liquidholdup.

Kinetic energy acceleration losses are small for most oil industry applications. The mainexception is high velocity flow through low pressure piping.

Pipephase 6.02 does not properly account for acceleration losses and is not suitable foruse in flare system design as a result.

For plant piping, rule of thumb values for pressure gradients, such as a frictional gradientof 0.2-0.5 psi per 100 ft. of equivalent length, are generally used.

The allowable pressure drop for a gathering system can be estimated from the initialwellhead pressure and the processing plant inlet separator pressure. The rule of thumb forthis method is to take 1/3 of the difference between the wellhead pressure and theseparator pressure as the allowable pressure drop in the pipeline.



The rule of thumb for estimating allowable pressure drop for long distance gas/condensatepipelines is to allow 10-20 psi per mile for frictional pressure drop at design rates.



Figure I:2-1 Horizontal Flow Regime Map



Figure I:2-2 Vertical Flow Regime Map



SECTION 3.0 - STEADY STATE DESIGN METHODS

3.1 Pipeline Design Methods

As stated in the previous sections, the pipeline designer needs to estimate the pressuredrop, flow regime, and velocities in the line in order to select the proper line size. Thecalculation of these parameters is laborious and is usually done by computer simulation.Line sizing is usually performed by use of steady state simulators, which assume that thepressures, flowrates, temperatures, and liquid holdup in the pipeline are constant withtime. This assumption is rarely true in practice, but line sizes calculated from the steadystate models are usually adequate.

Within Chevron, Pipephase and PIPEFLOW-2 are available for steady state pipelinesimulation.

For a more rigorous pipeline sizing, the simulations could be done using transientsimulators. Transient simulators allow changes in parameters such as inlet flowrate andoutlet pressure as a function of time, and calculate values for the outlet flowrates,temperatures, liquid holdup, etc. as a function of time. If the line is operating in slug flow,the line size calculated from the transient model may be different from that calculatedfrom a steady state simulator.

The principal uses of transient simulators are in the design of downstream equipment andthe development of operating guidelines. Transient simulators can model transientbehavior such as slug flow, pigging, rate changes, etc.

Transient simulators are quite new, developed in the last 10 years, and are not in generaluse. Chevron has used the OLGA program for transient flowline analysis on severalprojects, utilizing outside consulting services. CPTC developed an in-house transientsimulator, but it currently does not have as many features as the commercially availablecodes.

The use of steady state models will be further discussed in Section 3.2, and transientmodeling will be briefly discussed in Section 4.



3.2 Steady State Simulators

This section contains some general guidelines on the use of steady state simulators.Although there are several steady state programs available, the discussion will center onthe use of Pipephase, which is Chevrons currently recommended simulator. The topicscovered include: a) Phase Equilibrium and Physical Properties b) Pipeline Elevation Profile c) Heat Transfer d) Recommended Methods for Pressure Drop, Liquid Holdup, and Flow Regime Prediction e) Interpretation of Results

3.2.1 Phase Equilibrium and Physical Properties

Accurate prediction of the phase behavior and physical properties for the fluid flowingthrough the pipeline is essential to a good simulation of the pipeline operation. Theestimates of these parameters depend in large part on the quality of the input dataavailable.

During conceptual design work, the only data available may be an estimate of the oil rateand gas-oil ratio. After well tests have been performed, compositions of the wellstreamand PVT data may be available as well as projections of the flowrates of oil, gas and wateras a function of time. Obviously, as the accuracy of the input data improves, the quality ofthe pipeline simulation improves.

Pipephase has two fundamentally different models available within it for estimation ofphase behavior and physical properties. The black oil model estimates the phase behaviorand physical properties by use of a series of correlations that are based on operatingtemperature, pressure and some global parameters such as specific gravity of the oil andgas. Compositional models use an equation of state to estimate the quantity of liquid andgas at the operating conditions; then, correlations are used to estimate the physicalproperties.

The decision on whether to use the black oil model or compositional modeling depends onthe available information and the type of system that is being modeled.



The choice of models for gas-condensate and volatile oil systems is clear. Compositionalmodels should be used for any gas-condensate or volatile oil system. This recommendationcovers gas-oil ratios above about 3500 SCF/bbl.

For lower gas-oil ratios, the choice of models is more difficult. Compositional modelsshould give more accurate phase equilibrium results, but the physical property estimatesfrom the compositional models may not be as good as the black oil model. (Section 6-1illustrates this point.) As a result, it cannot be stated categorically that either the black oilmodel or the compositional model is superior for low gas-oil ratio systems. Generalpractice with Pipephase has been to use the black oil model for lower gas-oil ratio streams.

The accuracy of compositional modeling depends, in a large part, on the characterizationof the heavy ends of the well stream. The materials heavier than hexane (C6+) are usuallycharacterized by use of pseudo-components or cuts. The heavy ends could becharacterized by one C6+ cut, or by a series of cuts corresponding to various boilingranges. In general, the accuracy of the predictions increases when more cuts are used.

Pipephase requires two of the following parameters in order to characterize a cut: specificgravity; molecular weight; or normal boiling point. In many cases, the mole fractions forcuts heavier than C6 may have been measured in the PVT analysis, but cut propertieswere not noted. In cases like this, the customary assumption is to use the properties of thecorresponding normal paraffin as the cut properties. This adds some error to the analysis,but it is unavoidable in many circumstances.

If tests of the phase equilibrium and physical properties have been done as part of thewellstream analysis, Pipephase allows the users of the black oil model to adjust the modelpredictions for solution GOR, densities, and liquid viscosity to match experimentalvalues. The pipeline predictions after PVT matching should be considerably better thanthose obtained with use of the standard correlations.

If the compositional model is used in Pipephase, the only variable that can be easilymanipulated to match experimental data is the liquid viscosity. Pipephase does not have anoption that will automatically adjust the phase equilibrium calculations to matchexperimental data. It is possible to manually modify the phase equilibrium calculations, but



it requires considerable effort, and the methods to do this are beyond the scope of thisguide.

Although it is possible to get good estimates of the phase equilibrium for 3-phase (gas-oil-water) systems, the available software does not allow rigorous simulation of threephaseflow. The models present in Pipephase can only do two-phase (gas-liquid) flowcalculations. Pipephase averages the properties of the liquid hydrocarbon and liquidwater, and uses those average in the two-phase flow methods. Volumetric averaging,however, may not give good values for the viscosity and surface tension of the mixture. Ifthe oil and water form an emulsion, the viscosity estimate may be off considerably usingsimple volumetric averaging, because the viscosity of an emulsion can be as much as 50times as high as the viscosity of the oil or water. If it is likely that an emulsion will form,the Woeflin method, which is available in Pipephase, should be used to estimate theviscosity of the emulsion.

3.2.2 Pipeline Elevation Profile

The pipeline elevation profile used in the simulation can have a significant impact on thecalculated pressure drop. Because the liquid holdup in upwardly inclined flow is greaterthan the holdup in downward flow, the elevational pressure drop in uphill legs is greaterthan the pressure recovery in downhill legs. As a result, elevational losses can account formuch of the pressure drop in hilly terrain pipelines, even if the inlet and outlet of the lineare at the same relative elevation.

If the velocities in the line are high, the uphill and downhill holdups may be close. As themixture velocity decreases, there will be an increasing difference between uphill anddownhill holdups.

The following table illustrates how sensitive the liquid holdup is to mixture velocity atvarious angles of inclination from horizontal. The feed stream is a gas-condensate withabout 4 bbl/mm SCF of liquid present. (The values shown are predictions of the OLGASmodel.)



MIXTURE VELOCITY FT/SEC

ANGLE,DEGREES

2.7 4.1 5.4 8.1 16.2

-2.0 0.0041 0.0053 0.0064 0.0091 0.0115

-1.0 0.0052 0.0068 0.0085 0.0108 0.0122

-0.5 0.0068 0.0087 0.0108 0.0124 0.0126

0.0 0.0224 0.0218 0.0198 0.0156 0.0131

0.2 0.5797 0.4134 0.2249 0.0179 0.0134

0.5 0.5961 0.4988 0.3846 0.0317 0.0135

1.0 0.5997 0.5000 0.4314 0.3023 0.0144

2.0 0.6009 0.5024 0.4337 0.3428 0.0158

Using the values in the above table, a comparison of two models for a given section of apipeline has been made. In the first model, the pipeline segment consists of two equallength sections of -0.5 degree and +0.5 degree each. The second model consists of asingle horizontal pipeline segment. The liquid holdups for the two models are:

MIXTURE VELOCITY,FT/SEC

HOLDUP FOR -0.5 DEGREE AND +0.5 MODEL

HOLDUP FORHORIZONTAL MODEL

2.7 0.3015 0.0224 4.1 0.2538 0.0218 5.4 0.1977 0.0198 8.1 0.0221 0.0156

16.2 0.0131 0.0131



The liquid holdups are far apart at low velocities and are the same at higher velocities.This comparison makes two points:

The pipeline profile must be realistic if the calculations of liquid holdup and pressuredrop are to be accurate.

Low velocities cause severe problems in prediction of the pipeline performance.

For very low velocities, it would be necessary to know the pipeline elevation profilewithin an accuracy of about one pipe diameter in order to get accurate holdup predictions.This is generally not practical.

In many cases, the pipeline topography is not known when the preliminary pipeline sizingcalculations are run. Frequently, in offshore pipeline design, the designer only knowswater depths at subsea wells or platforms. Instead of assuming a straight line pipelineprofile, it is recommended that the designer add some terrain features to the pipelineprofile to simulate hills and valleys that are inevitably present in the actual profile.

To improve the accuracy of the simulation, many calculation segments should be used insimulating the pipeline. Increasing the number of calculation segments always improvesthe accuracy of the simulation, but it increases the computer simulation time. The numberof segments required depends on how rapidly the temperature, pressure and holdup arechanging in the pipeline. For a system with rapid changes in pressure, e.g. flare systems,the number of calculation segments should be greater. If the temperature and pressure arechanging slowly, a more coarse grid can be used to simulate the pipeline.

3.2.3 Heat Transfer

The temperature profile along the pipeline is important in many circumstances. Theamount of condensation of liquids along a gas-condensate line, for instance, has a largeimpact on the pressure drop and liquid holdup in the line. Hydrate and wax deposition mayoccur in the line, requiring accurate estimates of temperatures. Corrosion is a strongfunction of temperature, so good heat transfer estimates are vital to corrosion prediction.

To properly model the heat transfer between the pipeline and the surroundings, it isnecessary to have information on the following:



thicknesses of the pipewall, pipeline coatings and insulation

whether the pipe is buried or exposed

the burial depth of the line

type of surroundings

ambient temperatures

thermal conductivities of the pipe, coatings and insulation.

With this information the programs can calculate heat transfer coefficients, which are thenused to calculate the temperature profile in the pipeline.

Values of the thermal properties for various materials can be read from the followingtable. Note that the Chevron Fluid Flow manual also has an extensive list of thermalconductivities for various types of materials.

Material ThermalConductivity,Btu/hr-ft-degF

Specific Heat,Btu/lb-degF

Density, lb/ft3

Carbon Steel 26 0.11 490

Stainless Steel 8-13 0.11 488

Concrete(Saturated)

0.75-1.2 0.10 147-200

Onshore Soil (Wet) 1.35 0.20 90-110Subsea Sandy Soil 1.25-1.50 0.30 105-115

Coal Tar Epoxy 0.20 0.35 92

Fusion BondedEpoxy

0.15 0.32 75-90

Neoprene 0.12-0.15 0.50 90

Polyurethane Foam 0.011-0.022 0.38 2-12



At the early stages of a project, there may not be enough information to enable rigorouscalculation of the heat transfer coefficient.. The following are rule of thumb values for heattransfer coefficients for subsea flowlines, which can be used in these instances:

Applications U Value, BTU/hr/ft2/degF

Wells 2

Risers 20-40

Buried Pipelines 1-3

Concrete Coated Nonburied Pipelines 3-5

Nonburied Pipelines without Concrete 5-10

For gas/condensate pipelines, temperature loss by the Joule-Thomson expansion (J-T)effect can be significant. In many gas pipelines, the temperature of the gas leaving thepipeline is less than ambient because of the J-T effect.

Several concerns arise when using Pipephase for heat transfer calculations:

a) Pipephase only estimates temperature loss by the Joule-Thomson expansion coolingeffect if the compositional model is used. The J-T effect is ignored in black oilsimulations.

b) The default velocity of water flowing past a pipeline is 10 miles per hour inPipephase. This velocity is generally too high. More typical values are 1 to 3 ft/sec(0.7-2 mph).

c) The Pipephase viscosity routine does not estimate viscosities at temperatures below60 degrees F. At lower temperatures, it uses the viscosity at 60 degrees F. This canlead to errors for pipelines in deep water or cold climates.

d) The thermal conductivity for saturated concrete is much higher than that for dryconcrete. The saturated concrete value should be used for subsea pipelines withconcrete coating.



e) Unless a value is entered for Hrad, radiation is ignored in the heat transfer calculations.For subsea or buried pipelines, radiation is negligible, but it can be a significant effectfor surface flowlines.

f) The convective heat transfer routines in Pipephase are not very rigorous. Errors inheat transfer calculations can occur for systems in which convection is the primesource of heat transfer.

3.2.4 Recommended Methods for Pressure Drop, Liquid Holdup, and FlowRegime Prediction

There have been hundreds of multiphase flow design methods developed in the past 50years. Most computer programs contain dozens of options to select for pressure drop,liquid holdup, and flow regime predictions. Most of these methods only have small rangesin which their predictions are accurate. This section of the guide discusses this problemand gives some recommendations on which methods to use for certain applications.

Most of the methods available in Pipephase are correlations based on data taken in smalldiameter (0.5-2 inch) test loops having an air-water flow operating at nearly atmosphericpressure. The correlations developed from these data sets frequently do not include theeffects of all the key variables, such as pressure, because changes in these variables werenot studied in the experimental work. These correlations extrapolate poorly from fieldconditions.

In the past 10 years, the development of mechanistic modeling has created a markedimprovement in prediction capabilities. As noted in Section 1.2, mechanistic modelsattempt to model the physical phenomena associated with each flow regime. Mechanisticmodels solve a set of simultaneous equations developed for a specific flow regime.Correlations for a few key parameters are required in order to solve the equation set.Mechanistic models extrapolate to field conditions much better than correlations becausethe mechanistic models account for the effects of all the major variables.

Several mechanistic models have been developed in the past few years. Tulsa Universityhas developed models for near vertical flow (Ansari) and a general model covering allinclinations (Xiao). The physics in these models are good, but the correlations built intothem are based solely on small diameter, low pressure data. The OLGAS model is



currently the best available method for general multiphase flow calculations. OLGAS isbased on a wide range of data (diameter from 1 to 8 inches, pressures from atmosphericto 1400 psi), and it extrapolates best to field conditions.

OLGAS is a proprietary program that has not been available within Chevron. As thisguide is being written, however, negotiations are underway to add OLGAS to Pipephaseand several other programs as options. If OLGAS becomes available, it is therecommended method for prediction of pressure drop, liquid holdup and flow regime.Methods are available that are as good or slightly better than OLGAS in certain ranges,but they are not as good overall.

The following methods can be used in Pipephase as a check of OLGAS or as the designmethod if OLGAS is not available:

a) Pressure Drop1) Near Horizontal Low Gas-Oil Ratio - Beggs and Bril1 (Moody) is good.2) Near Horizontal Gas/Condensate - Eaton-Oliemans is good for relatively high

velocities. All of the methods are poor for low velocities.

3) Near Vertical Gas/Condensate - Both Gray and Hagedorn-Brown are good.4) Near Vertical Gas/Oil - Hagedorn and Brown is good.5) Inclined Up - Nothing is good; Beggs and Brill (Moody) is fair.6) Inclined Down and Vertical Down - Everything is poor. Use Beggs and Brill

(Moody), but answers may be suspect at times.b) Liquid Holdup _

1) Near Horizontal Low Gas-Oil Ratio - Beggs and Brill (Moody) is O.K.2) Near Horizontal Gas/Condensate Lines - All available methods are poor. The

Eaton holdup correlation is better than the other methods.

3) Near Vertical Gas/Condensate - The most accurate method is no-slip.4) Near Vertical Gas/Oil - Hagedorn and Brown is pretty good.5) Inclined Up - Beggs and Brill (Moody) is usable for low GOR lines, nothing is

accurate for gas/condensate. If gas velocities are high, use no-slip; otherwise use



Beggs and Brill (Moody). The user must be careful because the holdups can be afactor of 10 in error in some cases.

6) Inclined Down and Vertical Down - Everything is poor. Use Beggs and Brill(Moody), but answers may be suspect.

c) Flow Regimes1) The Taitel-Dukler flow regime map is as good as OLGAS for near horizontal

flow with the exception of the slug-dispersed bubble boundary. This boundary isvery poorly predicted. If this method is used, it is recommended that a value of~10 ft/sec be used as the superficial liquid velocity for the slug-dispersed bubbletransition rather than the Taitel-Dukler prediction.

2) The Taitel-Dukler-Barnea map for near vertical flow is also as accurate asOLGAS.

On occasion, the conditions for a simulation may cause otherwise good multiphase flowmethods to give erroneous results. It is usually a good idea to spot-check the results byuse of another method to ensure that the answers are reasonable. If there is a widevariance in results, CPTC should be contacted for guidance.

3.2.5 Interpretation of Results

When a multiphase simulator such as Pipephase is run, the interpretation of the results canbe difficult. The following section provides assistance in understanding Pipephase output,and ensuring that the design criteria for the line (velocities, flow regime, and allowablepressure drop) are met.

As discussed in Section 2.2, the velocity in the pipeline should be kept within a limitedrange. Calculation of the velocities from a Pipephase output is not straightforward. Thedesigners of Pipephase chose to include the actual gas and liquid velocities in their outputtable rather than the superficial gas and liquid velocities which are needed in the erosionalvelocity calculations. As discussed in Section 1.2, the superficial and actual velocities arerelated by simple formulas:

( )V Ug 1 Hsg l=



and V U Hsl l= 1

The liquid holdup is read from the "slip holdup" column. This calculation is made moredifficult by the poor formatting of the liquid holdup in the Pipephase output table. (Theliquid holdup is shown to only two decimal places in the table. For gas-condensate lines, ifthe liquid holdup is below 0.5 percent, the printout will show 0.00 for the holdup.)

A more accurate way of calculating the superficial velocities from the Pipephase outputtables which doesnt rely on reading the value for the liquid holdup is:

( )( )HU Vm

U Ul

g

g l=

V U Hsl l l=

V V Vsg m sl=

To calculate the C value in the API-RP14E equation, the value of the no-slip mixturedensity must be known. Pipephase apparently only calculates and tabulates this value inthe output table if the Beggs and Brill (Moody) method is used. If other methods are used,a value of 0.00 is given in the output table for the no-slip mixture density. The no-slipmixture density can be calculated, however, from the phase densities shown on the outputtable and the superficial velocities calculated above:

( ) ( )

ns

g sg l sl

m

V V

V=

+

Pipephase allows the user to print a flow regime map based on either the Taitel-Duklermap for near horizontal flow or the Taitel-Dukler-Barnea map for near vertical flow. Theflow regime map is printed only for the last "device" in a "link". If the "link" containsseveral pipes with different inclinations, the flow regime map for some of these sectionsmay be quite different from the map at the last "device". The only way to print the flowregime map at specific points along the line is to make these points ends of Pipephase"links".



The "link" summary tables print the flow regime predictions for each pipeline segment.The printout shows both the predictions of the multiphase flow design method (e.g. Beggsand Brill) and the Taitel-Dukler method. If OLGAS is available, the flow regimepredictions of OLGAS can be compared directly with the Taitel-Dukler prediction, and theuser can feel confident that the predicted flow regime is valid if the two methods match. Ifmethods other than OLGAS are used, disregard their flow regime predictions and onlyconsider the Taitel-Dukler predictions as reasonable.

Once the flow regime is determined, the designer needs to decide if this flow regime isacceptable. This decision is more difficult than it may appear. Ideally, the flow line shouldnot be in the slug flow regime. In practice, it may be very difficult to design a line to avoidslug flow under all anticipated flow conditions. The only variables the designer can changeare diameter and operating pressure; the changes in these variables required to avoid slugflow may be impractical. It should be pointed out that many pipelines operate successfullyin slug flow. As long as the pipeline and downstream equipment are designed with properconsideration of slug flow effects, they can be successfully operated.

The flow regime analysis may show that the line is in stratified flow. In many instances,this is an excellent flow regime in which to operate. At low flowrates, however, sluggingmay occur in lines predicted to be in stratified flow, induced by the terrain. Terraininduced slugs are generally much longer than the slugs in normal slug flow and can causesevere operating problems. Terrain slugging is discussed in more detail in Section 5.2.2.

If the pressure drop and velocities for lines in dispersed bubble or annular flow are withinacceptable limits, these flow regimes are usually good regimes in which to operate.

The pressure drop in the line should be compared with the allowable pressure drop. Thepressure drop in the line can be read from the Pipephase "link summary" table. It should bepointed out that pressure drop is not always a maximum at the highest flowrate. If thepipeline contains inclined or vertical elements, it is possible that the highest pressure dropmay occur at a low flow condition due to high elevational losses at low flows.

It is worthwhile to emphasize the point that the pipeline design should be checked at off-design points as well as the nominal design point. For most pipelines, worst caseconditions for liquid holdup and flow regime occur at turn-down conditions.




Points to remember from Section 3.0 -

Compositional models should be used for any gas-condensate or volatile oil system.This recommendation covers gas-oil ratios above 3500 SCF/bbl.

General practice with Pipephase: use the black oil model for lower gas-oil ratiostreams.

If it is likely an emulsion will form, the Woeflin method (available in Pipephase)should be used to estimate the viscosity of the emulsion.

The pipeline profile must be realistic if the calculations of liquid holdup andpressure drop to be accurate.

Low velocities cause severe problems in prediction of the pipeline performance.

If OLGAS becomes available, it is the recommended method for prediction ofpressure drop, liquid holdup, and flow regime.

Mechanistic models extrapolate to field conditions much better than correlations,since the mechanistic models account for the effects of all the major variables.

The following methods can be used in Pipephase as a check of OLGAS or as thedesign method if OLGAS is not available:

1. Pressure Drop

a) Near Horizontal Low Gas-Oil Ratio - Beggs and Brill (Moody) is good.b) Near Horizontal Gas/Condensate - Eaton-Oliemans is good for relatively

high velocities. All of the models are poor for low velocities.

c) Near Vertical Gas/Condensate - Both Gray and Hagedorn-Brown aregood.

d) Near Vertical Gas/Oil - Hagedorn and Brown is good.e) Inclined Up - Nothing is good; Beggs and Brill (Moody) is fair.

f) Inclined Down and Vertical Down - Everything is poor. Use Beggs andBrill (Moody), but answers may be suspect at times.



Liquid Holdup

g) Near Horizontal Low Gas-Oil Ratio - Beggs and Brill (Moody) is O.K.h) Near Horizontal Gas/Condensate Lines - Nothing is accurate. The Eaton

holdup correlation is poor, but better than the other methods.

i) Near Vertical Gas/Condensate - The most accurate method is no-slip.j) Near Vertical Gas/Oil - Hagedorn and Brown is pretty good.k) Inclined Up - Beggs and Brill (Moody) issuables for low GOR lines,

nothing is accurate for gas/condensate. If gas velocities are high, use no-slip; otherwise use Beggs and Brill (Moody). Be careful because theholdups can be a factor of 10 in error in some cases.

l) Inclined Down and Vertical Down - Everything is poor. Use Beggs andBrill (Moody), but answers may be suspect.

Flow Regimes

a) The Taitel-Dukler flow regime map is as good as OLGAS for nearhorizontal flow with the exception of the slug-dispersed bubble boundary.This boundary is very poorly predicted. If this method is used, it isrecommended that a value of ~10 ft/sec be used as the superficial liquidvelocity for the slug-dispersed bubble transition rather than the Taitel-Dukler prediction.

b) The Taitel-Dukler-Barnea map for near vertical flow is also as accurateas OLGAS.

The flow line should, ideally, not be in the slug flow regime. In practice, it may bevery difficult to design a line to avoid slug flow under all anticipated flow conditions.

At low flow rates slugging may occur in lines predicted to be in stratified flow,induced by the terrain.

If the pressure drop and velocity for lines in dispersed bubble or annular flow arewithin acceptable limits, these flow regimes are usually good regimes in which tooperate.



SECTION 4.0 - TRANSIENT FLOW MODELING

4.1 Transient Flow Modeling (General)

Transient multiphase flow simulators have only been developed recently. The first widelyused commercial program, OLGA, began development in about 1983 and has beencommercially available since 1990. OLGAs only current competitor, PLAC, wasintroduced to the market at about the same time. Chevron currently does not own eitherprogram but has used OLGA for specific projects through consultants. Chevron internallydeveloped a transient code, Transpire, in the same time frame as OLGA. This programhas not been widely used, and it does not have as many features as the commercial codes.

Steady state simulators assume that all flowrates, pressures, temperatures, etc. areconstant through time. Inherently transient phenomena, such as slug flow, are modeled byuse of their average holdups and pressure drops. Transient models allow all the inputvariables to change with time. Transient programs can model phenomena such as slugflow and can show the variations in parameters such as outlet gas and liquid flowrates as afunction of time. Transient simulators, therefore, model the actual operation of pipelinescloser and with more detail than steady state simulators.

Transient simulators solve a set of equations for conservation of mass, momentum andenergy to calculate the liquid and gas flowrates, pressures, temperatures and liquidholdups. These calculations are done at each time step. The programs utilize an iterativeprocedure, which ensures that a set of boundary conditions (such as inlet flowrates andoutlet pressures as a function of time) are met while solving the conservation equations.

Steady state modeling can be used to size pipelines, but the predicted size may beinaccurate if the line is in slug flow. Transient simulators can size pipelines moreaccurately, and they are valuable in several other areas such as the design of downstreamfacilities, development of operating guidelines, and the diagnosis of operating problems.Steady state simulators cannot properly address any of these other concerns.

4.2 Use of Transient Simulators

Because of their power, transient simulators have been used for a variety of purposes.These uses include:

a) Slug flow modeling



b) Estimates of the potential for terrain slugging c) Pigging simulation d) Estimation of corrosion potential in low spots in the line e) Startup, shutdown and pipeline depressuring simulations f) Development of operating guidelines g) Real time modeling, including leak detection h) Operator training i) Design of control systems for downstream equipment j) Slug catcher design

A general guideline for the use of steady state and transient modeling would be to usesteady state modeling during the feasibility level design of a system but use transientmodeling in the final design of the pipeline and its associated equipment.

As transient simulators improve and computer power increases, it is likely that transientsimulators will eventually supplant steady state simulators.

Because Chevron does not own a transient simulator at this time, this guide does notcontain any guidelines for their use. Section 5.1 discusses the use of the OLGA programfor slug length prediction.


Points to remember from Section 4.0 - Transient simulators model the actual operation of pipelines much closer than steady

state simulators.

General guideline for the use of steady state and transient modeling: use steady statemodeling during the feasibility level design of a system, but use transient modeling inthe final design of the pipeline and its associated equipment.



SECTION 5.0 - SLUG FLOW ANALYSIS

5.1 Slug Flow - General

The formation of slugs of liquid can be caused by a variety of mechanisms:

a) Hydrodynamic Sluggingb) Terrain Sluggingc) Piggingd) Startup and Blowdowne) Flowrate Changes

Each of the mechanisms will be briefly discussed here, and will be further discussed inSection 5.2.

Hydrodynamic slugging refers to operating in the slug flow regime. In near horizontalflow, slugs are formed by waves growing on the liquid surface to a height sufficient tocompletely fill the pipe. When this happens, alternating slugs of liquid and bubbles of gasflow through the pipe, as illustrated in Figure I:1-1.

Terrain slugging occurs when a low point in the line fills with liquid. The liquid does notmove until gas pressure behind the blockage builds high enough to push the liquid out ofthe low spot as a slug. Terrain slugging can produce very long slugs in pipeline-risersystems. Although terrain slugging occurs at low superficial gas and liquid velocities, theactual velocities during slug release can be very high.

When a pipeline is pigged, most of the liquid inventory is pushed from the line as a liquidslug ahead of the pig.

When a line is shut down, liquid that is left in the line will drain to the low points in theline. When the flow is restarted, the accumulated liquid may exit the pipeline as a slug.



When the flowrate is increased, the liquid holdup in the line decreases. This change inholdup can either exit the line as a gradual increase in liquid flow, or it can come out inthe form of a slug, depending on the flowrate.

Each of the slug flow mechanisms is highly transient in nature. Steady state models cannotproperly simulate slug flow behavior and are very limited in their ability to predict slugcharacteristics such as slug length and frequency. The next two sections of the guidediscuss the slug flow mechanisms in more detail, discuss available methods of predictingslug flow behavior and give some recommendations on sizing of slug catchers andseparators.

5.2 Slug Length and Frequency Predictions

Although estimates of slug length and frequency are of prime importance in design ofpipeline system facilities, most of the prediction methods available are poor. Developmentof prediction methods has been hampered by the difficulty of the problem and the meageramount of available test data. This section discusses each of the mechanisms for slug flow,discusses the available test data, and give recommendations on the best availableprediction methods.

5.2.1 Hydrodynamic Slugging

Experimental measurements of the slug length in hydrodynamic slug flow show severalinteresting results:

a) The slug length is not constant. At a given point in the line, the slug length variesgreatly around an average value. Different investigators have characterized the sluglength distribution as log normal, truncated Gaussian, inverse Gaussian, or fractaldistributions. The maximum slug length may be several times greater than theaverage.

b) The average slug length and the slug length distribution change with the positiondown the pipe. Slugs may grow, dissipate, or merge as the flow continues down thepipe. As a result, the average slug length usually increases with the position in thepipe, while the standard deviation of the slug length distribution decreases.



c) Slugs in vertical pipes are much smaller than slugs in horizontal pipes.

d) The slug length in laboratory experiments can be fairly well correlated. These testsshow that the average slug length (in feet) is approximately 32 times the PipeDiameter (in feet) for horizontal pipes.

e) In the data base of published pipeline field test results, the average slug length ismuch higher than the results observed in the laboratory. The field tests results showaverage slug lengths of 300-2000 times the Pipe Diameter, with some slugs as longas 10,000 times the Pipe Diameter.

The differences between laboratory and field data shown in points d) and e) above are dueto factors such as:

- terrain features have a large effect on the slug length and frequency;

- slug flow in the field can be combination of mechanisms such as hydrodynamicslugging causing terrain slugging;

- field pipelines are much longer, allowing more time for slug growth.

Average slug length is a complex function of many variables: the diameter and length ofthe pipeline; the topography of the line; the gas and liquid superficial velocities; the liquidphysical properties; and the gas density.

Several correlations have been presented for the prediction of slug length and slugfrequency for horizontal piping and pipelines. Most of these correlations are based solelyon laboratory data, which means they are of limited use in the design of pipelines in thefield.

A few correlation methods have been presented based on field data. Two of thesemethods, the Brill, et al. Correlation and the Hill & Wood method, have been widely usedfor slug length prediction. Both methods will be discussed in detail.

Brill, et al. took several sets of data on 12 and 16 inch pipelines at Prudhoe Bay in about1978. They were the first experimentalists to report the wide disparity between theextrapolation of lab results and field data. They developed a simple correlation for slug



length based on the Prudhoe Bay tests and small diameter laboratory data. Theircorrelation is:

ln(Ls) = -2.663 + 5.44 (ln(D))0.5 + 0.059 ln(Vm)where Ls = Average slug length, ft.

D = Pipeline inside diameter, inches

Vm = Mixture velocity, ft/sec

The Prudhoe Bay test data appeared to be a log normal distribution around the mean sluglength. Log normal distributions were fit to each of the Prudhoe Bay tests, and the meanslug length and variance were calculated. Their method assumes that the variance for anypipeline is the same as the average value of the variance from these tests. With thisassumption, the slug length distribution is the same for all pipelines. The distribution is:

Percent Probability Slug Length/Mean Slug Length

50.0000 1.00

84.1300 1.65

97.7200 2.72

99.8600 4.46

99.9900 6.42

99.9999 10.76

The Brill method is easy to use, and it has been used as the design basis for manyfacilities. The Brill et al. model is the basis for the slug length predictions in Pipephase.Unfortunately, the Brill method is very inaccurate. It gives poor predictions for almostevery data set that was not included in the original correlation. A comparison of the Brillpredictions against measured average slug lengths from a 16" laboratory line and twofield pipelines shows:



Actual Predicted

Data Source Avg. Slug Length, ft. Avg. Slug Length, ft.

BHRG 30 700

Field Data 1000 120

Field Data 400 1800

The inaccuracy of the Brill et al. method is due to the simplicity of their formula. Theslug length is predicted to be almost exclusively a function of the line diameter. As notedpreviously, the slug length is a complex function of many variables, and a simple formulalike this cannot approximate reality.

There are two other methods based primarily on the Prudhoe Bay data, namely, theNorris correlation and the Scott, et al. correlation. Their performance, while a bit betterthan the Brill method for larger pipelines, is also weak. Both of these methods assumethat the slug length is only dependent on the line diameter.

In 1990, Hill and Wood of BP published a paper proposing an alternate way of modelingslug frequency. Their work was based on both laboratory data and a large number of fieldmeasurements. The model is more sophisticated than the Brill approach and attempts toaccount for the many of the major variables. Their model correlates the slug frequencywith the diameter, gas-liquid slip velocity, and the equilibrium holdup at the beginning ofthe pipeline. The model assumes a horizontal pipe and uses the Taitel-Dukler stratifiedflow model to estimate the slip velocity and the equilibrium holdup at the beginning of thepipeline. In order to calculate the Taitel-Dukler holdup, the value for the Lockhart &Martinelli X factor is calculated from the following equation:

XVV

slsg g g

=

0 91

0 41

01. . .

where: 1 = Liquid Viscosity, cp

g = Gas Viscosity, cp



Once X is known, the value for the liquid holdup (Hle) for stratified flow can be read fromFigure I:5-1. The actual gas and liquid velocities can be calculated from the superficialvelocities and the stratified flow liquid holdup by the formulas:

UV

Hgsg

le=

1

and UVHl

slle

=

Their equation for slug frequency is:

( )( )F

H U U

D Hsle g l

le=

2 74

1

.

where Fs = Slug Frequency, slugs/hour

D = Pipe Inside Diameter, ft

Ug = Actual Gas Velocity, ft/sec

Ul = Actual Liquid Velocity, ft/sec

Hle = Liquid Holdup (based on stratified flow at the inlet of the pipe)

To calculate the slug length, the slug fraction, which is defined as the slug length dividedby the sum of the slug and bubble lengths, must be known. The slug fraction can becalculated rigorously by use of mechanistic slug flow models, such as the Hubbard-Duklermodel or the Nicholson, Gregory, and Aziz model. A simplified alternative to the use ofthese models is the following procedure.

First, the liquid holdup in the slug will be estimated. The Gregory, Nicholson, and Azizmethod is the easiest method to use, and it is as accurate as most of the methods. Theliquid holdup in the slug is given by:



HVls m

=

+

1

128 4

139

.

.

where Hls = Liquid Holdup in the slug

The liquid holdup in the bubble (Hlb) can be assumed to be ~0.20 for much of the slugflow range. From a material balance, the slug fraction (SF) can be calculated from thesevalues and the overall liquid holdup prediction:

SFH HH H

l lbls lb

=

The slug length can be calculated from:

LSF VFs

m

s=

3600 ( )

where Ls = Mean Slug Length, ft

An example of the use of the Hill & Wood model is shown in Section 6.1.

To use the Hill & Wood model for slug catcher designs, the design slug length is neededinstead of the mean slug length. A rule of thumb, based on the limited amount ofexperimental data available, is that the maximum slug length is approximately 6 times themean slug length.

The Hill & Wood model is more sophisticated than the Brill model and is based on abroader data base. If a correlation based model is used for slug length prediction, the Hill& Wood model is probably the best available method.

Another alternative approach to hydrodynamic slug length prediction is the method usedin the OLGA transient program. The slug tracking model in OLGA attempts todynamically track each slug in the system, from its inception to its dissipation or exit fromthe pipeline. This method models the growth or dissipation of each slug and modelsphenomena such as the merger of two slugs. The model accounts for terrain effects and



the effects of all of the major variables. This model, which has only been available to theindustry since early 1994, shows promise in prediction of slug lengths.

5.2.2 Terrain Slugging

Terrain slugging refers to slug

Multiphase Flow Design Guide

Documents

pipeline design methods

design criteria

steady state design

pipeline elevation profile

section highlights

pressure gradient calculations

frictional losses

elevational losses