1. Methodology-selection by Fuzzy Analytic Hierarchy Process for Studying Net Pays
Pedram Masoudi1,3, Mohammad Ataei2, Tahar Aïfa1*, Hossein Memarian3*
1*: Géosciences-Rennes, CNRS UMR6118, Univ. de Rennes 1, France,
2: School of Mining, Petroleum and Geophysics Eng., Shahrood Univ. of Tech., Iran.
3*: School of Mining Eng., Univ. of Tehran, Iran,

2. Why methodology selection? Ex. of software selection
Software programs differ in:
Functionalities and performance
Precision
Being user-friendly/ complexity of parameters
Speed of calculations
Simulating geological conditions …
Projects goals vary:
Research oriented
Industrial applications
Project scale
Addressing to which organization …
6-8 November 2016 – Tehran International Exhibition Center

3. Definition of net pay and methods
Table of content
Definition of net pay and methods
AHP and FAHP
Comparison matrices
Results and conclusion
6-8 November 2016 – Tehran International Exhibition Center

4. What is net pay? Modified from Worthington (2010)
Modified from Bashari (2007)
6-8 November 2016 – Tehran International Exhibition Center

5. Net pay detection methodologies
Cut-off (Worthington 2010): conventional method
Diffusivity equation (Masoudi et al. 2011): MSc
Bayesian classifier (Masoudi et al. 2012): MSc
Dempster rule of combination (Masoudi et al. 2014): IOOC
Artificial Neural Networks (ANN) (Masoudi et al. 2014): IOOC
6-8 November 2016 – Tehran International Exhibition Center

6. Net pay detection: cut-off
If-then rules on shale, porosity, saturation, permeability, …
Sarvak Formation, the Abadan Plain (Masoudi et al. 2012)
6-8 November 2016 – Tehran International Exhibition Center

7. Net pay detection: Diffusivity equation
Calculating a continuous “productivity index” based on solving diffusivity equation
Sarvak Formation, the Abadan Plain (Masoudi et al. 2012)
6-8 November 2016 – Tehran International Exhibition Center

8. Net pay detection: Bayesian classifier
Training a based on well-test results of well 3
Sarvak Formation, the Abadan Plain (Masoudi et al. 2012)
6-8 November 2016 – Tehran International Exhibition Center

9. Net pay detection: Bayesian classifier
Training a based on well-test results of well 4
Sarvak Formation, the Abadan Plain (Masoudi et al. 2012)
6-8 November 2016 – Tehran International Exhibition Center

10. Net pay detection: Sugeno-Bayesian
Fusion of Bayes3 and Bayes4 by Sugeno integral
Sarvak Formation, the Abadan Plain (Masoudi et al. 2012)
6-8 November 2016 – Tehran International Exhibition Center

11. Net pay detection: Dempster rule
Combining the information of shale content, porosity and saturation
Mishrif reservoir, the Persian Gulf (Masoudi et al. 2014)
6-8 November 2016 – Tehran International Exhibition Center

12. Net pay detection: Artificial Neural Network
Combining the information of shale content, porosity and saturation
Backpropagation due to well-test
Mishrif reservoir, the Persian Gulf (Masoudi et al. 2014)
6-8 November 2016 – Tehran International Exhibition Center

13. Net pay detection: Dempster rule
Combining the information of shale content, porosity and saturation
Burgan reservoir, the Persian Gulf (Masoudi et al. 2014)
6-8 November 2016 – Tehran International Exhibition Center

14. Net pay detection: Artificial Neural Network
Combining the information of shale content, porosity and saturation
Backpropagation due to well-test
Burgan reservoir, the Persian Gulf (Masoudi et al. 2014)
6-8 November 2016 – Tehran International Exhibition Center

15. Table of comparison IOOC report (Masoudi 2013) Precision
Precision
Generalization
Fuzziness
Simplicity of Method
User-friendly
Speed
Cut-off
acceptable in carbonate, poor in sandstone
acceptable
discrete
very simple
too simple
time consuming
ANN
86% in carbonate; 94% in sandstone
not checked
complex
difficult
rather time consuming
Bayes
74% in carbonate; 81% in sandstone
fuzzy: too weak
crisp: very good
simple
quick
DST
very good
continuous
very complex
IOOC report (Masoudi 2013)
6-8 November 2016 – Tehran International Exhibition Center

16. Hierarchy: linking alternatives to the goals
6-8 November 2016 – Tehran International Exhibition Center

17. Analytical Hierarchy Process (AHP)
A decision making technique
Organizes human’s thoughts and psyches (Saaty 1987) by :
comparing alternatives due to criteria
comparing criteria due to goals
Algorithm:
Graph of hierarchy
Comparison matrices for each level: Subjective pairwise comparison (Saaty 1977,1987):
1- Reciprocal 2- homogeneity 3- dependency 4- expectations and 5- hypothetical syllogism/ cardinal consistency
Weighting: (logarithmic) least square method, eigenvector or approximate method
Multiplication and scoring
Inconsistency rate and modification of comparison matrices
Final score
6-8 November 2016 – Tehran International Exhibition Center

18. Fuzzy AHP
6-8 November 2016 – Tehran International Exhibition Center

19. Comparison matrices between alternatives: 6 matrices
Comparison matrix of precision with: IR=0.0076<0.10
Cut-off
ANN
Bayes
DST
[1/2,1,2]
[1/5,1/4,1/3]
[1/4,1/3,1/2]
[3,4,5]
[1,2,3]
[2,3,4]
[1/3,1/2,1]
weights
0.14
0.43
0.29
Comparison matrix of generalization with: IR=0.0000<0.10
Cut-off
ANN
Bayes
DST
[1/2,1,2]
[1/6,1/5,1/4]
[4,5,6]
weights
0.22
0.33
6-8 November 2016 – Tehran International Exhibition Center

20. Comparison matrices between criteria: 3 matrices
Precision Group
Application Group
Precision
Generalization
Fuzziness
Simplicity of Method
User-friendly
Speed
Comparison matrix for industrial mode with Inconsistency Rate: IR=0.0448
Preciseness
[1/2,1,2]
[2,3,4]
[8,9,9]
[4,5,6]
[6,7,8]
[1/4,1/3,1/2]
[1/9,1/9,1/8]
[1/8,1/7,1/6]
[1/6,1/5,1/4]
Weights
0.31
0.19
0.00
0.12
0.06
Comparison matrix for research mode with Inconsistency Rate: IR=0.0600
[3,4,5]
[5,6,7]
[1/5,1/4,1/3]
[1/7,1/6,1/5]
0.25
6-8 November 2016 – Tehran International Exhibition Center

21. Final scores for different modes
Precision
Generalization
Fuzziness
Simplicity of Method
User-friendly
Speed
Final Score
General mode
weights
0.24
0.09
Cut-off
0.14
0.22
0.50
0.43
0.00
0.23
ANN
0.17
0.25
Bayes
0.29
0.33
0.26
DST
Industrial mode
0.31
0.19
0.12
0.06
0.21
0.27
Research mode
6-8 November 2016 – Tehran International Exhibition Center

22. Thank you Conclusion None of the methods could be prioritized
But cut-off method is always ranked 3rd
Having different algorithms is not useless at all
Acknowledgement:
This work has been supported by the Center for International Scientific Studies & Collaboration (CISSC) and French Embassy in Iran through PHC Gundishapur Program.
Mahta Gholizadeh Ansari for sharing her experience about fuzzy AHP
Conclusion
General Mode
Industrial Mode
Research Mode
1st
Bayes, DST
ANN, DST
ANN, Bayes
2nd
ANN
Bayes
DST
3rd
Cut-off
Thank you