You are on page 1of 8

SPE 136931

A Comparative Study Between Empirical Correlations & Mechanistic Models of


Vertical Multiphase Flow
Anes Usman Yahaya and Abdallah Al Gahtani, King Fahd University of Petroleum & Minerals, Dhahran, Saudi Arabia
Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the 2010 SPE/DGS Annual Technical Symposium and Exhibition held in Al-Khobar, Saudi Arabia, 04–07 April 2010.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed by
the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members.
Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an
abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract

A number of empirical correlations and mechanistic models have been developed for predicting pressure drop and other fluid flow
characteristics during multiphase flow in vertical wellbore. At any particular time, engineers are often faced with the difficulty of
which particular empirical correlation or mechanistic model should be used to evaluate multiphase flow in vertical wellbore. This is
because; a particular correlation or model may be developed under some certain assumptions or for a particular region which may not
be applicable to other regions or under a new set of conditions.
This paper presents a study that was conducted to evaluate the predictive performance of both mechanistically based models and
empirical correlations in calculating pressure drop and other flow characteristics in vertical multiphase flow. Using 414 real field data
points covering pipe sizes of 2.375 in. to 7.0. in, oil flow rate of 280B/D to 23,200B/D, water cut up to 95%, gas-oil ratio up
927SCF/STB from the Middle East, the correlations and models are used to predict pressure drop in vertical multiphase phase flow for
several wells. The predicted performance is then compared with actual measured well pressure drop data.
The empirical correlations evaluated are those conventional correlations of Hagerdon and Brown, Duns and Ros, Orkiszewski, and
Beggs and Brill. While mechanistic models of Ansari et al, Aziz et al, Chokshi et al are also evaluated. The variations between the
predicted and the measured are analyzed using statistical error analysis.
Based on the results from the analysis, mechanistic model of Ansari et al. out-perform all the conventional empirical correlations
evaluated in this study for vertical multiphase flow for the Middle East region.

Introduction

Two-phase flow is commonly encountered in the petroleum industries. This frequent occurrence presents the challenge of
understanding, analyzing, and designing two-phase systems. Because of the complex nature of two-phase flow, the problem was first
approached through empirical methods. Recently, the trend has shifted towards the modeling approach. The fundamental postulate of
the modeling approach is the existence of flow patterns or flow configurations. The question “Which model?” has been asked since
modeling started, but it still brings about a lot of confusion and disagreement. Numerous models have been and are constantly being
proposed, almost each one of which claiming to add something to the knowledge of a phenomenon or process, but each inherently
inadequate to encompass the full complexity of the real world. The later, however, should not be and has never been seen as an
obstacle by the modelers, who continue to develop new models at varying levels of complexity, generality, and validity. There is a
constant competition and sometimes disagreement between the two respective schools of thought: the empiricists say that more
empirical, data-based models have a better practical value because of infinite complexity of the underlying phenomena; the rationalists
stress the (theoretically) better cognitive and predictive potential of mechanistically based models, which are able to generate new
knowledge.

Empirical Correlations

The empirical correlations are formulated by establishing a mathematical relation based on experimental data. Application of empirical
models is limited to the data range used in generating the correlation. In this section, each of the empirical correlations of two-phase
flow in vertical wellbore used in this study is reviewed so as to ascertain the assumptions and conditions under which they were
2

developed and their inherent limitations. The equations for pressure gradient calculations and flow pattern predictions are also
summarized.

Hagerdon & Brown Correlation (1964): This empirical correlation was developed using data from 1500-ft vertical wellbore.Tubing
diameters ranging from 1-2 in. were considered in the experimental analysis along with 5 different fluid types namely: water and four
types of oil with viscosities ranging from 10 to 110cp @ 80°F. The correlation developed is independent of flow patterns

Duns & Ros Correlation: The Duns & Ros correlation is developed for vertical flow of gas and liquid mixtures in wells. This
correlation is valid for a wide range of oil and gas mixtures and flow regimes. Although it was intended for use with 'dry' oil/gas
mixtures, it can also be applicable to wet mixtures with a suitable correction.

Orkiszewski Correlation: This correlation is limited to two-phase pressure drop in vertical wellbores and is an extension of Griffith
and Wallis work. The correlation is valid for different flow regimes such as bubble, slug, transition, and annular mist. It should be
noted that the liquid distribution coefficient (holdup) is evaluated using data from the Hagerdon and Brown model.

Beggs & Brill: The Beggs and Brill correlation was developed from experimental data obtained in a small scale test facility. the range
of parameters studied were; gas flow rate(0 to 300Mscf/D); (2) liquid flow rate(0 to 30 gal/min); (3) average system pressure(35 to 95
psia); (4) pipe diameter (1 to 1.5 in.); (5) liquid holdup (0 to 0.870); (6) pressure gradient (0 to 0.8psi/ft); (7) inclination angle (-900 to
+900); and (8) horizontal flow pattern. Air and water were the fluids used in the experiment.

Mechanistic Models

The mechanistic models are based on a phenomenological approach that takes into account basic principles (conservation of mass and
energy). The most important achievement of mechanistic models was that following flow regime determination, separate models for
predicting the vertical wellbore hydraulic characteristics, such as pressure drop, liquid holdup and temperature profile were also
developed. Mechanistic models are practical in extensive conditions.Various mechanistic models used in this study are reviewd so as
to ascertain the assumptions under which they were developed and their inherent limitations.

Ansari et al. (1994) Model: This mechanistic model was developed for upward two-phase flow in wellbores. The model predicts the
existence of four flow patterns, namely; bubble flow, slug flow, churn flow and annular flow as shown in Fig. 1. They then calculated
the flow variables by taking account the actual mechanism of the predicted flow pattern. The model was evaluated against a wide
range of experimental data and field data available in the updated Tulsa University Fluid Flow Projects (TUFFP) well data bank.

Aziz et al. Model: They proposed a new vertical flow pattern map was and also presented new equations for calculating the liquid
holdup occurring in the Bubble- and Slug-flow patterns. No new equations were proposed for the annular mist pattern and the Duns
and Ros equations were recommended for this flow pattern. The flow pattern was correlated with dimensionless numbers which
depend primarily on the gas and liquid superficial velocities.

Chokshi et al. (1996) Model: They developed a mechanistic model that considers three flow patterns: bubble, slug and annular flow,
and used drift flux modeling approach for bubble to slug transition. Measured data were gathered from 324 tests for widely varying
flow rates. Pressure drop predictions of the model were compared to eight correlations/mechanistic models using measured data and
independent data bank of 1712 data sets.

Flow Patterns

Flow patterns are a very complex hydrodynamic phenomenon. Flow pattern is characterized by the distribution of the liquid and gas in
pipe. The main flow patterns considered in this study are; buuble flow, slug flow, annular flow and churn flow. Flow pattern
information is given usually on a “flow pattern map” where transition boundaries are plotted. A brief description of each of the flow
pattern considered in this study is given below. A diagrammatic form of the flow pattern is shown in Fig 1.

Bubble Flow: The vertical well-bore is considered to be almost completely filled with liquid, and the free gas phase is present in small
bubbles. The bubbles move at different velocities and except for their density, have little effect on the pressure gradient.
Ansari et al., showed in the development of mechanistic model for two-phase upward flow in pipes, that the minimum data at which
bubble flow occurs is given by Taitel et al.10 as:

1/ 2
⎡ (ρ − ρ )σ ⎤
D = 19.01⎢ L 2 G ⎥ …………………….. (1)
⎢⎣ ρ Lg ⎥⎦
3

For pipes sizes larger than this, the basic transition mechanism for bubble to slug flow is coalescence of small gas bubbles into large
Taylor bubbles.

Slug Flow: Slug flow is ahighly complex flow with an unsteady nature. The prediction of pressure drop, void fraction and other slug
parameters is a difficult task. The purpose of “mechanistic modeling” is to simplify the flow configuration such that an nalysis would
be possible. The gas phase is more pronounced. Although, the liquid phase is still continuous, the gas bubbles coalesce and form plugs
or slug that almost fill the pipe cross section. The gas velocity is greater than that of the liquid. Both the gas and liquid have effects on
the pressure gradient.The first thorough physical model for slug flow was developed by Fernandes et al.11 A simplified version of this
model was presented by Sylvester12. The basic simplification made was the use of a correlation for slug void fraction.

Annular Flow: In annular flow, the gas flows through the central core of the pipe, while the liquid flow along the walls of the pipe as
a film. Therefore, the system may be looked upon as a single-phase flow of gas through a tube of slightly reduced diameter because of
the liquid film. A discussion of the hydrodynamics of annular flow was presented by Wallis13. Along with this, Wallis also presented
the classical correlations for entrainment and interfacial friction as a function of film thickness. Later, Hewitt and Hall-Taylor gave a
detailed analysis of the mechanisms involved in an annular flow. All models that followed later are based on this approach.

Churn Flow: This flow pattern exists in upward flow only and is very chaotic in nature and changes from a continuous liquid phase to
a gas phase occur. The gas bubbles may join and liquid may be entrained in the bubbles. Although the liquid effects are significant, the
gas phase effects are predominant.

Evaluation Procedure

Since a model calculates the pressure gradient and thus the pressure drop by integration of this gradient, their predictions are usually
compared with the total pressure drop over a flow string. Another factor to consider in comparison of pressure drops is the direction of
calculations. In this study, calculations are done against the flow direction. The calculations begin at lower pressures where the flow is
almost multiphase where the prediction method is less accurate. As calculations proceeds towards the higher pressure end, the single
phase flow occurs, where the gradients predictions are more accurate, and the empirical correlations or mechanistic models that predict
higher pressure gradients are obviously seen as poor performers.

In this study, measured or real well pressure drops are available for comparison purpose. Each of the empirical correlations and
mechanistic models considered in this study is evaluated using actual field data. The predicted pressure drop of each model is
compared against real pressure drop data from a field in the Middle-East. Also, the effect of tubing size or diameter, oil gravity, gas-
liquid ratio, two-phase flow with or without water- cut on the predicted pressure drop of each model and correlation are also
compared. Statistical error analysis is used in the evaluation.

Evaluation based on Pressure Gradient Predictions: A quadratic equation was found to fit the predicted pressure profile for each of
the empirical correlations and mechanistic models. The pressure gradient of each model can also be calculated by differentiation of
this quadratic pressure profile. A plot of the pressure gradient profile for each of the empirical correlations and mechanistic models is
shown in Fig. 4.

Empirical Correlations & Mechanistic Models Evaluation: First, each of the empirical correlations and mechanistic models is
evaluated by comparing the field measured pressure drop and the predicted pressure drop. The following four empirical correlations
and three mechanistic models were evaluated:
-Ansari model,
-Chokshi model,
-Aziz et al. model,
-Duns and Ros correlations,
-Orkiszewski correlation,
-Hagerdon and Brown,
-Beggs and Brill.

Model Comparison

First, each of the empirical correlations along with the mechanistic models considered in this study is evaluated by comparing the
measured (from real field data) and predicted pressure drop. The comparison is accomplished by comparing the statistical and
graphical error analyses. The statistical parameters used to determine the accuracy of the correlations are shown in Appendix A3.

Variation of Tubing Diameter, d: Figure 2 compares the average absolute percent relative errors of empirical correlations and
mechanistic models. Each empirical correlation and mechanistic model was evaluated against a range of tubing diameters to predict
vertical pressure drop in the vertical wellbore. Details of parameters used in the evaluation are shown in table 1. It can be confirmed
4

from the results displayed in Fig 2 that the mechanistic model of Ansari et al. performs best with an average absolute relative error of
8.29%, followed by the empirical correlation of Beggs & Brill with an average absolute relative error of 11.44%. All empirical
correlations and mechanistic models investigated showed an increase in average absolute relative percent error as the tubing diameter
increase above 3.5in.

Variation of Oil flow Rate: All the empirical correlations and mechanistic models evaluated in this were analyzed using oil flow rate
ranging from 4000STB/D to 21000STB/D. The result of the analyses in shown in Fig 3. It shows that at rate below 5000STB/D, all
empirical correlations and mechanistic models except that of Duns & Ros and Aziz et al. show average absolute relative error less than
10%. Above this rate, the superiority of mechanistic model of Ansari et al. becomes evident.
Overall, the mechanistic model of Ansari et al. gives the best performance with average absolute relative error less than 10%, followed
by empirical correlations of Beggs and Brill.

Conclusions

From Tables 2 – 3 and Figures 2 – 3, the performance of the empirical correlations and mechanistic models evaluated indicates that:

ƒ The overall performance of mechanistic model of Ansari et al. is superior to all the empirical correlations, with the
performance of correlation of empirical correlation of Beggs and Brill as the second best.

ƒ The superiority of mechanistic model of Ansari et al. over the empirical correlations is more evident if well data outside
the range of data used in the development of the empirical correlations are used in the evaluation.

Recommendations

Based on the above conclusions, the following recommendations are suggested.

ƒ The mechanistic model of Ansari et al. is recommended for use in predicting pressure drop in vertical multiphase flow in the
Middle East region.

ƒ For the region of which the data is used in this evaluation, the empirical correlation of Beggs and Brill is recommended for
use where mechanistic model of Ansari et al. is not handy.

ƒ To improve the performance of mechanistic models, the degree of empiricism in the development of the mechanistic models
should be reduced.

As a final remark, it should be mentioned that there may be other empirical correlations and mechanistic models in the literature that
were not evaluated in this study whose performance in predicting pressure drop in vertical multiphase flow could be superior to that of
Ansari et al. model. Hence, our drawn conclusions and recommendations were based on the empirical correlations and mechanistic
models evaluated in this study.

Appendix A. Statistical parameters

The following statistical means were used for evaluation in this study:

A. 1. Average percent relative error

nd
Er = n1
d ∑ Ei (A – 1)
1
where

Ei = ( X exp − X est
X exp
)x100(i = 1,2,..., n )d (A – 2)

A. 3 Average absolute percent relative error


5

nd
Ea = n1
d ∑| Ei | (A – 3)
1

Acknowledgments

The authors of this work will like to extend their appreciations to Department of Petroleum Engineering, King Fahd Universiy of
Petroleum & Minerals, Dhahran, Saudi Arabia, for providing the technical resources used in carrying out the study.

References

1. Ansari, A.M., Sylvester, N.D., Sarica, C., Shoham, O., and Brill, J. P.: "A Comprehensive Mechanistic Model for Upward Two-Phase Flow
in Wellbores," SPE Prod. & Facilities (May 1994) 143-152.

2. Chokshi, R.N.: Prediction of Pressure Drp and Liquid Holdup in Vertical Two-Phase Flow through Large diameter Tubing, PhD
Dissertation, University of Tulsa, Tulsa, OK (1994).

3. Aziz, K., Govier, G.W., and Fogarasi, M.:"Pressure Drop in Wells Producing Oil & Gas," J.Can. Pet. Tech. (July-Sept. 1972) 38-47.

4. Beggs, H.D., and Brill, J.P.: "A Study of Two-Phase Flow in Inclined Pipes," J.Pet.Tech. (May 1973) 607-617.

5. Duns, H. Jr., and Ros, N.C.J.: "Vertical Flow of Gas and Liquid Mixtures in Wells," Proc., Sixth World Pet. Cong., Frankfurt(1963) 451-
465.

6. Hagerdon, A.R., and Brown, K.E.: Experimental Study of Pressure Gradients occurring during Continuous Two-Phase Flow in Small
Diamter Vertical Conduits," J. Pet. Tech.(April 1965) 17, 475-484.

7. Orkiszewski, J.: "Predicting Two-Phase Pressure Drops in Vertical Pipe," J. Pet. Tech. (June 1967) 19, 829-838.

8. Hassan, A.R., and Kabir, C.S.: "A Study of Multi-phase Flow behavior in Vertical Oil wells: Part 1-Theorectical Treatment," Paper SPE
15137 presented at the SPE California Regional Meeting, Oakland, April 2-4, 1986.

9. H. Dale Beggs.:"Production Optimization Using Nodal Analysis,"OGCI Publications, Tulsa, OK.

10. Taitel, Y., Barnea, D., and Dukler, A. E.: " Modelling Flow Pattern Transitions for Steady Upward Gas- Liquid Flow in Vertical
Tubes,”AIChE J. (1980), 26, 345-354.

11. Fernandes, R. C., semait, T., and Dukler, A. E.:" Hydrodynamics Model for Gas-Liquid slug Flow in Vertical Tubes,"
AIChE J.(1986), 29, 981-989.

12. Sylvester, N. D.: a mechanistic Model for Two-PhaseVertical Slug Flow in Pipes,"ASME J. Energy Resources Tech. (1987), 109, 206-213.

13. McQuillan, K. W., and Whalley, P. B.: "Flow Patterns in Vertical Two-Phase Flow," Int. J Multiphase Flow (1985),
11, 161-175.

14. Ahmed J. Al-Muraikhi.,"Evaluation of Vertical Multiphase Flow Correlations for Saudi Arabian Field Conditions,"
Thesis (1989), Dept of Petroleum Engineering, King Fahd University, Dhahran, Saudi Arabia.

Nomenclature 1980

Er = Average percent relative error.


nd = Number of data points
D= Tubing diameter, ft
σ = Surface tension, dyne/cm
ρ = Density of fluid, Kg/m3
Xexp = Measured value2004
of data 2004
Xest =Predicted value of data
6

Fig. 1: Flow Pattern in a vertical oil well

50
Hagerdon & brown
45 Beggs & Brill
Duns & Ros
Orkizeskwi
40
Average Absolute Relative Percent Error

Chokshi et al.
Aziz-Govier-Foga
35 Ansari et al.

30

25

20

15

10

0
0 1 2 3 4 5 6 7
Tubing Diam eter, in

Figure 2: Comparison of Average absolute Percent Relative Error for Empirical Correlations &
Mechanistic Models as a function of Tubing diameter.

30
Hagerdon & Brown
Orkizeskwi
Duns & Ros
Beggs & Brill
Average Absolute Percent Relative Erro

25
Chokshi et al.
Aziz-Govier
Ansari et al.
20

15

10

0
0 5 10 15 20 25
Oil Rate , M s tb/d

Figure 3-Comparison of Average absolute Percent Relative Error for Empirical Correlations &
Mechanistic Models as a function of Oil rate.
7

Predicted Pressure, psia

0 500 1000 1500 2000 2500 3000


0
Ansari Model
1000 Chokshi Model
Aziz et al. Model
2000
Duns & Ros Correlation
Measured Depth, ft

3000 Orkizewski Correlation


Beggs & Bril Correlation
4000 Hagedorn & Brown Correlation
Well
5000

6000

7000

8000

Fig. 4 -Predicted Pressure Profile for 3.816in Tubing, 12.5% water cut, liquid rate of 11650 STB/D

12.5

Ansari Model
12 Chokshi Model
Aziz et al. Model
Duns & Ros Correlation
Orkizewski Correlation
Vsl, ft/sec

11.5
Hagerdorn & Brown Correlation
Beggs & Brill Correlation
11

10.5

10
1 10 100

Vsg, ft/sec

Fig. 5 -Flow Pattern Map

Diameter(in) Oil Rate(STB/D) Water Rate(STBW/D) GOR(SCF/STB)


1.992 300-3100 0-1119 525-999
2.436 368-4200 0-2974 3-928
2.988 3207,7480 0-2146 516-858
3.816 280-17076 0-7900 120-896
3.96 240-20200 0-7790 32-999
6.36 18400 0 500
Table 1- Range of Well Data
8 [Paper Number]

Average Absolute Percent Relative Error


Correlation/ Hagerdon & Orkizeskwi Duns & Beggs & Chokshi Aziz- Ansari et al.
Tubing Brown Ros Brill Govier
Diameter(in)
1.992 9.54 5.71 15.28 9.12 19.40 25.71 8.17
2.436 14.01 16.17 13.84 12.44 20.55 31.84 10.84
2.988 7.38 4.27 6.97 4.65 6.24 29.45 2.54
3.816 8.80 16.96 7.10 5.70 10.40 45.46 3.59
3.96 24.6 24.63 30.34 19.40 34.10 41.43 11.99
6.36 31.19 17.85 23.54 17.31 20.47 39.27 12.62
Table 2- Average absolute Percent Relative Error for Empirical Correlations & Mechanistic
Models as a function of Tubing diameter. (Calculations against the Flow Direction)

Average Absolute Percent relative Errors


Correlations/Oil Hagerdon & Duns & Beggs & Chokshi et Aziz- Ansari
rate, mstb/d Brown Orkizeskwi Ros Brill al. Govier et al.
4.137 9.54 8.12 15.28 5.71 7.40 20.71 4.17
7.535 12.19 10.31 9.54 8.85 6.47 18.27 6.62
9.963 10.01 11.45 17.84 7.17 8.55 25.84 6.84
15.526 8.38 9.65 9.97 8.27 6.24 20.45 3.54
17.076 7.80 8.70 8.10 7.16 10.40 17.46 2.59
20.2 15.63 16.40 16.34 13.63 15.09 19.43 11.99
Table 3- Average absolute Percent Relative Error for Empirical Correlations & Mechanistic
Models as a function of Oil rate. (Calculations against the Flow Direction)

You might also like