Re-Evaluation and Validation for Multiphase Flow Correlations in

Pipes
Mohamed A. Abd El-Moniem (Amal Petroleum Company)

Ahmed H. El-Banbi (Cairo University)

Abstract
This paper presents a new approach for selecting appropriate multiphase flow correlations in
pipes. Our target was to develop an expert system that can help petroleum engineers in selecting
multiphase flow correlation in the absence of actual bottom hole flowing pressure data.
The expert system is based on three main ideas: (1) identifying range(s) of input data where
certain correlations perform better than others, (2) quantifying the calculated pressure uncertainty
range for the best correlation(s) in their best range of applicability, and (3) validating the output
results from the system. This expert system will not only recommend the best correlation(s) for
use in certain well configuration, flowing conditions, and reservoir properties; but also will give
the engineer the uncertainty range in the pressure calculation (if any).
A large database of pressure drop data from many wells with different configurations, different
flowing conditions, and different fluid types and properties was collected and analyzed. The
database included more than 3,250 points covering wide range of oil production data (from few
to more than 30,000 bbl/D), gas-oil ratio (0 to 40,000 scf/bbl), and water cut (0 to 98%).
This comprehensive data set was tested against 14 of the most commonly used multiphase flow
correlations, and pressure calculation errors were calculated for each point and correlation. The
data were then clustered into groups with known inputs and best correlations were identified.
Also, sensitivity analysis was done on different parameters to study the degree of its
effectiveness on the selection of the multiphase flow correlation.
An expert system was developed with a set of rules to identify the best correlation for each range
of user input data. The expert system also gives the uncertainty range for each computed pressure
drop (derived from errors analysis for each correlation). In some clusters of data, more than one
correlation was found to yield good results, whereas in other groups, only one correlation was
clearly the best.
Finally we validated the expert system results by new data and the results matched with the
expert system results.
This expert system can then be used now to select the most appropriate multiphase flow
correlations in pipes in absence of flowing gradient surveys or downhole pressure sensor data.

Introduction
In the field of petroleum engineering, predicting multiphase flow pressure drop can sometimes
present a challenge for practicing engineers. This area received significant attention since the

1
1950s to unlock the secrets of pressure drop calculations in pipes. Many correlations by different
investigators were derived, and each correlation was derived for certain conditions.
Multiphase flow occurs due to the variations of phases that flow through one pipe. These phases
may be oil, water and gas. Due to the different properties of gas rather than oil and water for
viscosities and other properties, the gas tends to flow faster than oil and past the liquid in the
upward flow and that what we called gas slippage. This in-situ liquid volume in the pipe is
defined as the liquid holdup.
The pressure drop calculation depends on the properties of the mixture of liquid including “oil
and water” plus the gas. The total pressure drop is equal to the summation of the hydrostatic,
friction and acceleration pressure drops. The hydrostatic term depends on the mixture density
which is a function of the liquid holdup. The friction term depends on the friction factor for the
mixture. The acceleration term appears due to the change in velocity resulting from the change in
pressure.
A term called flow patterns was used to describe the physical phenomena of the phase’s
redistribution and it depends mainly on the superficial velocity of the phases.
The flow patterns for vertical wells can range from single liquid flow, bubble, slug, churn,
annular and mist flow with respect to reduction in pressure. It may include dispersed bubble,
annular, slug, stratified wavy and finally stratified smooth for horizontal wells.
Both empirical correlations and mechanistic models were derived to calculate the pressure drop
in pipes. The empirical correlations were derived based on some mathematical relations but they
were limited by the data used in the derivation process. Empirical correlations can be classified
into three categories3:
1- No slip with No flow patterns like Poettman and Carpenter, and Fancher and brown…etc.
2- Slip with No flow patterns like Gray, and Hagedorn and Brown……………..etc.
3- Slip with flow patterns recognize like Duns and Ros, and Beggs and Brill…………..etc
The mechanistic models depend on some physical phenomena with analytical relations to predict
the flow patterns firstly, and then calculate the different parameters for each flow regime.
Example mechanistic models include Ansari et al., and Aziz et al……………etc.
Although there are many different correlations to calculate the pressure drop in pipes, we lack a
clear criteria upon which we can select the best correlations for different conditions. Therefore,
the target of this paper is to develop such criteria to select the best correlation in case of absence
of actual bottom hole flowing pressure data.

Developing the Expert System

The target was to collect as much reliable data as we can from many sources of both single point
and multi-points surveys.
We first collected around 3250 bottom hole flowing pressure data points from both the literature
and actual wells. The range of the data is shown in Table 1.

2
 Table 1 Range of Data Used in Developing the Expert System

Property / Fluid Type Oil Property / Fluid Type Gas

qo, bbl/D 5 – 36,800 qg, MMscf/D 0.24 – 157
GOR, scf/bbl 0 – 40,000 CGR, bbl/MMscf 0 – 616
W.C., % 0 – 98 WGR, bbl/MMscf 0 – 2,338
BHT, °F 82 – 370 BHT, °F 96 – 331
API, ° 8 – 60 API, ° 41 – 117
ɣg 0.57 – 1.81 ɣg 0.52 – 0.94
Depth, FT 250 – 14,358 Depth, FT 443 – 21,453

Tables 2 and 3 show more detailed statistics of the data and show that variety of well conditions
are covered (natural flow, gas lifted, and ESP).
Table 2 Statistics of Data Used in Developing the Expert System

Type Points No. of Surveys Survey Points Single Points

Natural vertical oil 1,599 174 1,305 294

Natural vertical gas 265 0 0 265

Natural deviated oil 499 43 330 169

Natural deviated gas 290 2 15 275

Natural horizontal oil 67 8 65 2

Natural horizontal gas 0 0 0 0

Gas lift vertical oil 135 12 120 15

Gas lift vertical gas 0 0 0 0

Gas lift deviated oil 96 6 69 27

Gas lift deviated gas 0 0 0 0

Gas lift horizontal oil 0 0 0 0

Gas lift horizontal gas 0 0 0 0

ESP vertical oil 184 2 15 169

ESP deviated oil 115 0 0 115

ESP horizontal oil 0 0 0 0

3
 Table 3 Statistics of Data Used in Developing the Expert System
Number of Pressure
Production Type No. of surveys survey points single points
Points

Natural Flow Oil 2,165 225 1,700 465

Natural Flow Gas 555 2 15 540

Gas Lift 231 18 189 42

ESP 299 2 15 284

All these data (around 800 wells) were modeled with a commercial pipe modeling software. We
performed around 1,600 cases and calculated the predicted bottom hole flowing pressure using
fourteen different correlations for each point. Results for one of the flowing gradient surveys
from the database are shown in Fig. 1.

Fig. 1 Calculated and Observed Pressure V.S. Depth for Different Fourteen Correlations

The correlations used in this work include:

1. Duns and Ros Modified

2. Hagedorn and Brown
3. Fancher and Brown
4. Mukherjee and Brill
5. Beggs and Brill
6. Petroleum Experts
7. Orkiszewski
8. Petroleum Experts 2
9. Duns and Ros Original

4
 10. Petroleum Experts 3
11. GRE
12. Petroleum Experts 4
13. Hydro
14. Petroleum Experts 5

We, then, calculated the error from each correlation for each point using Eq-1.

𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑒𝑑−𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑
𝐸𝑟𝑟𝑜𝑟 % = ∗ 100 ………………………………… (1)
𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑

After that we developed a Microsoft Access database, and collected the error data for all points
and all 14 correlations in the database. Eleven tables were constructed as shown in Fig.-2 for
rates, surface conditions, actual bottom hole flowing pressure, calculated pressure from different
correlations, type of fluid, type of production, type of well geometry, general and definition.

Fig. 2 Different Tables Used in Developing the Access Database File

Using the Access database file, we were able to extract the different statistics from the total
database.

Eqs. 2 and 3 were used for calculating both average error and absolute average error.

1
Average Error = ∑ni=1 Error ………………………………………………………………. (2)
n

1
Average absolute Error = ∑ni=1|Error| ……………………………………………………….. (3)
n

Table-4 shows the average and absolute average error for each correlation through the database
and Table-5 shows the minimum and maximum errors.

5
 Table 4 Average Error and Average Absolute Error of the different fourteen correlations for the Entire
Database
Correlation Average Absolute Error % Average Error %
Duns and Ros Modified 19.98 12.21
Hagedorn and Brown 12.99 -3.23
Fancher and Brown 14.26 -9.38
Gray* 5.65 -1.51
Mukherjee and Brill 16.85 6.94
Beggs and Brill 13.67 8.01
Petroleum Experts 12.24 -1.04
Orkiszewski 22.76 4.81
Petroleum Experts 2 11.34 -1.15
Duns and Ros Original 14.04 4.91
Petroleum Experts 3 11.77 -4.79
GRE 12.77 0.86
Petroleum Experts 4 13.06 -1.35
Hydro 3P 50.51 42.75
Petroleum Experts 5 11.89 -0.21
* used for gas cases only.

Table 5 Minimum and Maximum Error for the Different Fourteen Correlations for the Entire Database

Correlation Minimum Absolute Error % Maximum Absolute Error %

Duns and Ros Modified 0.01 729.68

Hagedorn and Brown 0.00 924.66
Fancher and Brown 0.00 408.25
Gray* 0.00 40.71
Mukherjee and Brill 0.02 481.03
Beggs and Brill 0.00 712.03
Petroleum Experts 0.01 605.45
Orkiszewski 0.00 787.19
Petroleum Experts 2 0.00 568.50
Duns and Ros Original 0.00 321.14
Petroleum Experts 3 0.00 511.42
GRE 0.00 582.91
Petroleum Experts 4 0.00 320.65
Hydro 3P 0.00 920.55
Petroleum Experts 5 0.00 586.94

* used for gas cases only.

As shown in Figs-3 and 4 as samples, we determined the error for each correlation for all the
points in the database to show the correlation performance for different points. The large
variations in the error range of minimum and maximum error and from the lowest error from the
points in the correlation performance figures, we can see that there is a set of conditions in which
each correlation has the power to calculate bottom hole flowing pressure with minimum error.
Therefore, our objective was to cluster the data to find the best correlation(s) for each set of
conditions.

6
 HB Correlation Performance
100.00
90.00
80.00
70.00
60.00
Error%

50.00
40.00
30.00
20.00
10.00
0.00
328
110
219

437
546
655
764
873
982

2617

2835
1091
1200
1309
1418
1527
1636
1745
1854
1963
2072
2181
2290
2399
2508

2726

2944
3053
3162
1

Point#

Fig. 3 Hagedorn and Brown Correlation Performance Through the Entire Database

MB Correlation Performance
100.00
90.00
80.00
70.00
60.00
Error%

50.00
40.00
30.00
20.00
10.00
0.00
328
110
219

437
546
655
764
873
982

2617

2835
1091
1200
1309
1418
1527
1636
1745
1854
1963
2072
2181
2290
2399
2508

2726

2944
3053
3162
1

Point#

Fig. 4 Mukherjee and Brill Correlation Performance Through the Entire Database

7
We divided the data into clusters according to type of fluid, depth, well geometry and tubing size
as shown in Fig.-5

Clusters Map
Vertical Deviated Horizontal
Oil Gas Oil Gas Oil Gas
>= 2.99 < 2.99 >= 2.99 < 2.99 >= 2.99 < 2.99
Shallow

(A)

(B) (G) (R) (L)

Moderate

(C) (N) (H) (S)

(D) (O) (I) (T) (M)

(E) (P) (J) (U)

Deep

(F) (Q) (K) (V)

Fig. 5 Clusters Map

Shallow depth is considered to be less than 2000 ft, deep depth is more than 9000 ft and the
moderate depth is between. The clusters from A to M are for oil wells and from N to V are for
gas wells.

We, then, divided the data into sub-clusters according to production criteria. Table-6 shows the
criteria for oil wells according to oil rate, water cut and GOR.

8
 Table 6 Sub-Clusters Production Criteria for Oil Wells

qo, bbl/D W.C., % GOR, scf/bbl

< 2,500 <5 < 1,000

2,500 – 10,000 5 – 30 1,000 – 5,000

> 10,000 30 -70 > 5,000

> 70

Table-7 shows the criteria for gas wells according to gas rate, CGR and WGR

Table 7 Sub-Clusters Production Criteria for Gas Wells

qg, MMscf/D WGR, bbl/MMscf CGR, bbl/MMscf

< 10 0 < 30

10 – 30 0 – 100 30 – 100

30 – 100 > 100 > 100

> 100

We ran queries in the Access database to extract the different sub-clusters as shown in Fig. - 6

Fig. 6 Query Design for the Access Database

9
We could extract 121 sub-clusters for oil wells and 69 for gas wells. Then we calculated the error
statistics for each correlation in each sub-cluster.

Sensitivity Analysis on the Most Effective Input Parameters

Abd El-Moniem and El-Banbi2 stated that there were many parameters affecting the calculations
of the pressure drop, and consequently the selection of the best correlation. Several sensitivity
analyses were performed for the most important parameters to identify if they had any effect on
changing the best correlation(s) selection.

1- GOR for High Oil Rate Wells

It was found GOR for high oil rate well had minor effect in selecting the best correlation as
shown in Figs. 7 and 8. Changing the GOR value, although changed the pressure drop, did not
change the best correlation at high oil production rate.

Pressure V.S. Depth, qL = 15120 bbl/D

GOR = 370 scf/bbl
Pressure, Psi
2100 2200 2300 2400 2500 2600
7000

7100

7200
Depth, ft

7300

7400

7500

7600

DRM HB FB MB

Fig. 7 Correlation Comparison for GOR for High Oil Rate Wells (GOR=370 scf/bbl)

10
 Pressure V.S. Depth, qL = 15120
bbl/D
GOR = 323 scf/bbl
Pressure, Psi
2100 2200 2300 2400 2500 2600
7000

7100

7200
Depth, ft

7300

7400

7500

7600
DRM HB Observed FB
MB BEGG PET1 ORK
PET2 DRO PET3 GRE

Fig. 8 Correlation Comparison for GOR for High Oil Rate Wells (GOR=323 scf/bbl).

11
2- GOR for Low Oil Rate Wells

Similarly, a sensitivity analysis was performed for low oil rate wells. It was found that it had
significant effect in selecting the best correlation as shown in Figs. 9 and 10.

Pressure V.S. Depth, qL = 1407 bbl/D

GOR = 407 scf/bbl
Pressure, Psi
200 400 600 800 1000 1200 1400 1600 1800 2000
0

500

1000

1500

2000
Depth, ft

2500

3000

3500

4000

4500
Observed DRM HB FB MB
BEGG PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 9 Correlation Comparison for GOR for Low Oil Rate Wells (GOR=407 scf/bbl)

12
 Pressure V.S. Depth, qL = 1407 bbl/D
GOR = 460 scf/bbl

Pressure, Psi
200 400 600 800 1000 1200 1400 1600 1800 2000
0

500

1000

1500

2000
Depth, ft

2500

3000

3500

4000

4500
Observed DRM HB FB MB
BEGG PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 10 Correlation Comparison for GOR for Low Oil Rate Wells (GOR=460 scf/bbl)

13
 3- API for Oil Wells

Another sensitivity analysis was performed to test the effect on API for oil wells on best
correlation(s) selection. It was found that API had significant effect in selecting the best
correlation(s) as shown in Figs. 11 and 12.

Pressure V.S. Depth, API=29°

Pressure, Psi
1300 1400 1500 1600 1700 1800
3500

3550

3600

3650

3700
Depth, ft

3750

3800

3850

3900

3950

4000

Observed DRM HB FB MB
Begg PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 11 Correlation Comparison for API for Oil Wells (API=29°)

14
 Pressure V.S. Depth, API=25°

Pressure, Psi
1300 1400 1500 1600 1700 1800
3500

3550

3600

3650

3700
Depth, ft

3750

3800

3850

3900

3950

4000

Observed DRM HB FB MB
Begg PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 4 Correlation Comparison for API for Oil Wells (API=25°)

15
 4- Gas Specific Gravity for Wells

A sensitivity analysis was done, and it was found it has great effect in selecting the best
correlation as shown in Figs. 13 and 14.

Pressure V.S. Depth, Gas Sp.gr=0.896

Pressure, Psi
1300 1400 1500 1600 1700 1800
3500

3550

3600

3650

3700
Depth, ft

3750

3800

3850

3900

3950

4000

Observed DRM HB FB MB

Begg PET1 ORK PET2 DRO

PET3 GRE PET4 HYDRO PET5

Fig. 5 Correlation Comparison for Gas Specific Gravity for Wells (Gas Sp. Gr. =0.896)

16
 Pressure V.S. Depth, Gas Sp.gr=0.84

Pressure, Psi
1300 1400 1500 1600 1700 1800
3500

3550

3600

3650

3700
Depth, ft

3750

3800

3850

3900

3950

4000

Observed DRM HB FB MB
Begg PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 6 Correlation Comparison for Gas Specific Gravity for Wells (Gas Sp. Gr. =0.84)

17
 5- Pipe Roughness for Oil Wells

A sensitivity analysis was done, and it was found it has no effect in selecting the best correlation
as shown in Figs. 15 and 16.

Pressure V.S. Depth, ε = 0.0006 in

Pressure, Psi
400 600 800 1000 1200 1400 1600 1800
0

1000

2000
Depth,ft

3000

4000

5000

6000
DRM Observed HB FB MB
BEGG PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 7 Correlation Comparison for Pipe Roughness for Oil Wells (ε=0.0006)

18
 Pressure V.S. Depth, ε = 0.00015 in

Pressure, Psi
400 600 800 1000 1200 1400 1600 1800
0

1000

2000
Depth,ft

3000

4000

5000

6000

DRM Observed HB FB MB
BEGG PET1 ORK PET2 DRO
PET3 GRE PET4 HYDRO PET5

Fig. 8 Correlation Comparison for Pipe Roughness for Oil Wells (ε=0.00015)

19
 6- Pipe Roughness for Gas Wells

A sensitivity analysis was done, and it was found it has great effect in selecting the best
correlation as shown in Figs. 17 and 18.

Pressure V.S. Depth, ε = 0.0006 in

Pressure, Psi
2100 2200 2300 2400 2500 2600 2700
6000

6100

6200

6300

6400
Depth,ft

6500

6600

6700

6800

6900

7000

DRM Observed HB Gray MB

BEGG PET1 ORK PET2 DRO

PET3 GRE PET4 HYDRO PET5

Fig. 9 Correlation Comparison for Pipe Roughness for Gas Wells (ε=0.0006)

20
 Pressure V.S. Depth, ε = 0.00015 in
Pressure, Psi
2100 2200 2300 2400 2500 2600 2700
6000

6100

6200

6300

6400
Depth,ft

6500

6600

6700

6800

6900

7000
DRM Observed HB Gray MB

BEGG PET1 ORK PET2 DRO

PET3 GRE PET4 HYDRO PET5

Fig. 10 Correlation Comparison for Pipe Roughness for Gas Wells (ε=0.00015)

Approach
To make sure good correlations under certain conditions are not dropped from the selection, we
added ±1% over the best error for each point and then select the best correlation(s) for the new
error band. This step was done to overcome any misleading data that might affect the calculation.

Then a term called strength factor was introduced and calculated by Eq.-4. It is defined for each
sub-cluster as the number of times that each correlation appeared to be the best correlation or one
of the best over the total points in the sub-cluster.

21
 Number of matched Points
Strengh Factor = X 100 …………………………………(4)
Total Points

Results
The detailed results can be found in Abd El-Moniem thesis1. A sample of the results is shown in
Tables-8 and 9.

Table 8 - Sample of Complete Results of One Sub-Cluster

Correlation No. of Points Strength (%) Avg. Error (%) Avg. Abs. Error (%)
PET5 9 31 -0.87 1.86
GRE 9 31 -0.96 1.95
MB 8 28 0.15 2.10
PET3 7 24 0.49 2.38
HYDRO 7 24 -1.23 2.43
PET 7 24 -1.43 2.63
PET2 7 24 -2.17 2.67
DRO 6 21 -0.85 1.62
PET4 5 17 -0.07 1.32
DRM 4 14 0.23 0.96
ORK 4 14 -0.93 2.86
BEGG 4 14 -2.67 3.34
HB 3 10 0.44 0.44
FB 1 3 3.31 3.31
Total Points 29

Table 9 - Sample of Complete Results of One Sub-Cluster

Correlation No. of Points Strength (%) Avg. Error (%) Avg. Abs. Error (%)
DRO 15 65 0.17 1.24
ORK 12 52 0.39 1.80
PET2 10 43 -0.84 1.23
PET4 9 39 0.03 0.86
DRM 9 39 0.16 1.61
PET5 8 35 -0.20 0.77
HYDRO 8 35 -0.51 0.82
GRE 8 35 -0.13 0.88
PET 6 26 -0.10 0.75
HB 6 26 -0.10 0.75
PET3 6 26 -0.10 0.75
BEGG 6 26 -0.97 1.46
MB 4 17 -0.28 1.08
Total Points 23

Similar results were generated for all sub-cluster to identify the best correlation(s) and their
expected errors. Tables 10 and 11 show the general results for oil and gas wells.

22
 Table 10 - General Results for Oil Wells

Best
Production
Type of Well qo bbl/D W.C. % GOR scf/bbl Correlation
Type

GRE & PET 5

≤ 10,000 <5

< 1,000 ORK

Tubular < 2,500 5 - 30
Vertical
MB & BEGG
< 2,500 30 - 70

DRM
Annular ≥ 2,500 <5 < 1,000

FB & ORK
≤ 10,000 <5 < 1,000

DRO
< 2,500 30 - 70 < 1,000
Deviated Tubular
DRM
< 2,500 <5 1,000 – 5,000

2,500 – HB
<5 1,000 – 5,000
10,000

Table 11 - General Results for Gas Wells

Production
Type of qg MMscf/D WGR bbl/MMscf CGR bbl/MMscf Best Correlation
Type
Well

≥ 30 0 ≤ 30 GRAY

Vertical
GRAY, MB &
Tubular Any 0-100 ≤ 30
BEGG

Deviated Any 0-100 30-100 PET 2 & FB **

Quality Check Using Survey Points Only

A quality check on the expert system results was performed by using the survey points only
rather than the total points. This was done to overcome any misleading results due to using single
pressure point data (in case of error in measured data).

We used 1,904 flowing points in this process. These points covered 25 sub-clusters. We repeated
the work above using these survey points rather than the entire database. 24 sub-clusters where
matched between the results from the survey points and the total database with a 96% success
rate. Also, an enhancement in the strength factor and in the average absolute error could be

23
obtained when we used the survey points alone. This step gave us more confidence in the results
we obtained in this research and increased the confidence in the expert system results.
A sample of the comparison between total points and survey points results can be found in
Tables 12 and 13.

Table 12 - Comparison Between Total Points and Survey Points Results For Cluster “C”

Survey Cluster Survey Cluster

Improvement Improvement
Avg. Avg.
Sub- Survey Total Strength Strength in Avg. Abs.
Correlation Percentage Abs. Abs.
Error
Status
Cluster points points Factor Factor Error Error
% %
% % % %
C1 49 99 PET3 37 29 8 3.96 5.81 1.85 Ok.
C2 6 30 ORK 100 47 53 3.27 6.63 3.36 Ok.
BEGG 63 54 9 0.56 4.57 4.01 Ok.
C3 16 28
MB 56 54 2 16.53 14.28 -2.25 Ok.
GRE 72 58 14 2.14 2.14 0 Ok.
C13 29 36
PET5 72 58 14 2.48 2.48 0 Ok.

Table 13 - Comparison Between Total Points and Survey Points Results For Cluster “I”

Survey Cluster Survey

Cluster
Improvement Avg. Improvement
Sub- Survey Total Strength Strength Avg. Abs.
Correlation Percentage Abs.
Error
in Avg. Abs. Status
Cluster points points Factor Factor Error Error %
% %
% % %
I1 39 84 HYDRO 33 35 -2 1.01 2.22 1.21 Ok.
PET3 60 48 12 1.08 1.08 0 Ok.
FB 55 44 11 1.73 1.73 0 Ok.
I2 42 52 HB 52 42 10 1.62 1.62 0 Ok.
PET1 50 40 10 0.86 0.86 0 Ok.
PET2 45 37 9 0.99 0.99 0 Ok.
I3 74 108 BEGG 43 35 8 1.61 2.35 0.74 Ok.
I4 10 25 DRM 50 48 2 4.92 6.82 1.9 No
I5 27 69 DRM 74 67 7 2.05 2.28 0.23 Ok.
I8 4 7 DRM 50 43 7 2.36 2.14 -0.22 Ok.
I13 13 24 FB 62 50 12 0.48 1.54 1.06 Ok.

Validation Process
As a final step, we validated the expert system by using a new set of data that had not been used
in developing the expert system. 138 flowing pressure data points were obtained from 64 new
wells to test the expert system. The new data came from both single point and multi-point
flowing surveys. According to their input data, the validation data covered 40 sub-clusters.

24
We constructed models and ran 69 cases using the same commercial software to obtain predicted
bottom hole flowing pressure from all correlations. We used the expert system to predict the best
correlations(s) for each data point and we tested the prediction against actual best correlation for
every point.

As shown in Table-14, validation results reached 72%. The rest of the points (28%) were not
predicted properly by the expert system. However, 38% of the not valid points predicted best
correlation to be one of the medium strength correlations, while the rest 62 % failed to identify
the best correlation(s).

Table 14 - Statistics of Total Points Resulted From Different Simulated Correlations

Percentage
Total S imulated Point S tatus Number of Points
%
M atch 154 72
No 61 28
Total 215

Conclusions
We utilized a large dataset to test the accuracy of 14 most widely used multiphase flow
correlations. We also used an approach to take advantage of the fact that each correlation has
application strength under certain conditions of well geometry, flow conditions, and PVT
properties. The approach is based on data clustering. The following conclusions can be made:

1. The data clustering approach allows the engineer to select the best multiphase flow
correlation that will provide the lowest error. In certain clusters, few correlations are
found superior (and not only one correlation). The error is also quantified for every
correlation in each data cluster so the engineer would expect the accuracy of pressure
drop prediction when utilizing this approach.
2. The overall error is around 11.34% for the best correlation in the entire dataset, while the
range of error for best correlation(s) in different data clusters ranges between 0.01% and
3% for most cases with accurate PVT. This error comparison proves the usefulness and
applicability of the clustering approach presented here.
3. At very high rate oil wells, best correlation selection is not affected by changing GOR
and temperature. However; in low rate oil wells, best correlation selection is affected
significantly by GOR due to rapid change in flow regimes with increased GOR.
4. Best correlation selection is sensitive to gas specific gravity and stock tank oil API.
5. PVT data is probably the most important factor affecting the accuracy of the correlation.
Care should be taken in providing PVT input to best correlations identified in this work.
6. For the dataset tested here, tubing roughness was not found to have significant impact on
changing the best correlation(s) for different clusters of oil data. However, tubing
roughness has significant effect on selection of best correlations for gas wells.
7. In practice, an accurate gas rate should be determined through production test separator
especially for low rate oil wells as GOR has important effect on pressure drop results.

25
Acknowledgement
The authors would like to thank Prof. Dr. El-Sayed El-Tayeb and Mr. Samir Zayed for their
ideas and valuable discussions on the subject.

Nomenclature
API = American Petroleum Institute (Density Measurement)
bbl = Barrel
Begg = Beggs and Brill correlation
BHT = Bottom hole temperature
CGR = Condensate gas ratio
d = Diameter
D = Day
DRM = Duns and Ros Modified Correlation
DRO = Duns and Ros Original Correlation
ESP = Electrical Submersible Pump
FB = Fancher and Brown Correlation
FT = Foot
°F = Degree Fahrenheit
GOR = Gas oil ratio
GRE = GRE Correlation
HB = Hagedorn and Brown Correlation
HYDRO = HYDRO Correlation
i = iteration Number
in = Inch
ID = Inner diameter
MB = Mukherjee and Brill Correlation
MM = Million
n = Number of Points

26
 OD = Outer Diameter
ORK = Orkiszewski Correlation
PET = Petroleum Experts Correlation
PET 2 = Petroleum Experts 2 Correlation
PET 3 = Petroleum Experts 3 Correlation
PET 4 = Petroleum Experts 4 Correlation
PET 5 = Petroleum Experts 5 Correlation
qg = Gas flow rate

𝑞𝐿 = Liquid Flow Rate

𝑞𝑜 = Oil Flow Rate

scf = standard cubic feet
T.S. = Tubing Size
W.C. = Water Cut
WGR = Water to gas ratio
Ɣg = Gas specific gravity
ε = Pipe Roughness
% = Percentage

References
1. Abd El-moniem, Mohamed A. 2016. Evaluation and Proper Selection of Multiphase
Flow Correlations, M.S. Thesis, Cairo University, Egypt.
2. Abd El-Moniem M.A., El-Banbi A.H. 2015 “Proper Selection of Multiphase Flow
Correlations,” SPE 175805 Presented at North Africa Technical Conference and
Exhibition, Cairo, 14-16 September.
3. Ansari A.M., Sylvester N.D., Shoham O., and Brill J.P. 1990. "A Comprehensive
Mechanistic Model for Upward Two Phase Flow in Wellbores," SPE 20630 Presented at
the 65th Annual Technical Conference and Exhibition, New Orleans, 23-26 September,
151-165.
4. Ashiem H. 1986. "MONA, An Accurate Two Phase Well Flow Model Based on Phase
Slippage," SPE 12989, SPE Production Engineering, May, 221-230.

27
5. Aziz K., Govier G.W. and Fogarasi M. 1972. "Pressure Drop in Wells Producing Oil and
Gas," J. Can. Pet., Montreal, July-September, 38-48.
6. Baxendell P.B. and Thomas R. 1961. "The Calculation of Pressure Gradients in High
Rate Flowing Wells," J. Pet. Tech., October, 1023-1028.
7. Beggs H.D. and Brill J.P. 1973. "A Study of Two Phase Flow in Inclined Pipes," J. Pet.
Tech., May, 607-617.
8. Brown K.E., Beggs H.D. 1977. The Technology of Artificial lift Methods.
9. Chierici G.L., Ciucci G.M. and Sclocchi G. 1974. "Two Phase Vertical Flow in Oil Wells
Prediction of Pressure Drop," J. Pet. Tech., August, 927-938.
10. Cornish R.E. 1976. "The Vertical Multiphase Flow of Oil and Gas at High Rates," J. Pet.
Tech., July, 825-831.
11. Duns H.Jr. and Ros N.C.J. 1963. "Vertical Flow of Gas and Liquid Mixtures in Wells,"
Section II, Paper 22-PD 6, Proc. Sixth World Pet. Congress, Frankfurt, 19-26 June, 451-
465.
12. Fancher G.H.Jr. and Brown K.E. 1963 "Prediction of Pressure Gradients for Multiphase
Flow in Tubing," Soc. Pet. Eng. J., March, 59-69.
13. Gould T.L., Tek M.R. and Katz, D.K. 1963. "Two Phase Flow Through Vertical,
Inclined, or Curved Pipe," J. Pet. Tech., July, 915-926.
14. Hagedorn A.R. and Brown K.E. 1964. "The Effect of Liquid Viscosity in Two Phase
Vertical Flow," J. Pet. Tech., February, 203-210.
15. Hagedorn A.R. and Brown K.E. 1965. "Experimental Study of Pressure Gradients
Occuring during Continuous Two Phase Flow in Small Diameter Vertical Conduits," J.
Pet. Tech., April, 475-484.
16. Hill T.J. and Wood D.G. 1994. "Slug Flow: Occurrence, Consequences, and Prediction,"
SPE 27960 Presented at the University of Tulsa Petroleum Engineering Symposium, OK,
U.S.A., 29-31 August, 53-62.
17. Kleyweg D., Tiemann W.D. and Dalziel S.G. 1983. "Gas Lift Optimization in the
Claymore Field," SPE 11885-MS Presented at Offshore Europe, Aberdeen, United
Kingdom, 06-09 September.
18. Minami K. and Brill J.P. 1987. "Liquid Holdup in Wet Gas Pipelines," SPE Production
Engineering, February, 36-44.
19. Mukherjee H. and Brill J.P. 1983. "Liquid Holdup Correlations for Inclined Two Phase
Flow," J. Pet. Tech., May, 1003-1008.
20. Orkiszewski J. 1967. "Predicting Two Phase Pressure Drops in Vertical Pipe," J. Pet.
Tech., June, 829-838.
21. Peffer J.W., Miller M.A. and Hill A.D. 1988. "An Improved Method for Calculating
Bottomhole Pressures in Flowing Gas Wells with Liquid Present," SPE Production
Engineering, November, 643-655.

28
22. Reinicke K.M., Remer R.J. and Hueni G. 1987. "Comparison of Measured and Predicted
Pressure Drops in Tubing for High Water Cut Gas Wells," SPE Production Engineering,
August, 165-177.
23. Prosper Software Help Manual, Petroleum Experts, 2013.

29