You are on page 1of 8

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Relationships between permeability, porosity and pore throat size in carbonate rocks using regression analysis and neural networks

This article has been downloaded from IOPscience. Please scroll down to see the full text article. 2006 J. Geophys. Eng. 3 370 (http://iopscience.iop.org/1742-2140/3/4/008) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 171.64.171.238 The article was downloaded on 27/03/2012 at 15:26

Please note that terms and conditions apply.

INSTITUTE OF PHYSICS PUBLISHING J. Geophys. Eng. 3 (2006) 370376

JOURNAL OF GEOPHYSICS AND ENGINEERING

doi:10.1088/1742-2132/3/4/008

Relationships between permeability, porosity and pore throat size in carbonate rocks using regression analysis and neural networks
M R Rezaee1,3 , A Jafari1 and E Kazemzadeh2
1 2

University of Tehran, University College of Science, School of Geology, Tehran, Iran NIOC/Research Institute of Petroleum Industry

E-mail: mrezaee@khayam.ut.ac.ir

Received 5 May 2006 Accepted for publication 16 October 2006 Published 6 November 2006 Online at stacks.iop.org/JGE/3/370 Abstract Accurate estimation of permeability from other data has been a challenge for many years. The aim of this study was to establish relationships between permeability, porosity and pore throat size, and apply these relationships in a predictive sense. Regression analysis was utilized to achieve a set of relationships between permeability, porosity and pore throat size for 144 carbonate samples. These relationships can be used to estimate permeability from porosity and pore throat radii. Also in this study, a fully-connected multi-layer perceptron network was used to predict permeability from porosity and pore throat radii. An articial neural network, a biologically inspired computing method which has an ability to learn, self-adjust, and be trained, provides a powerful tool in solving complex problems. These characteristics have enabled articial neural networks to be more successful in predicting permeability when compared to regression analysis. This study also indicates that pore throat radii corresponding to a mercury saturation of 50% (r50) is the best permeability predictor for carbonates with complex pore networks.
Keywords:

permeability, porosity, pore throat radii, carbonates, regression analysis and articial neural networks

1. Introduction
Permeability, which indicates the capacity of a reservoir to deliver hydrocarbons, is one of the most important parameters for reservoir studies. Variation of permeability is related to a series of factors, and there have been many efforts to establish a relationship between these factors and permeability. McGowen and Bloch (1985), Atkinson et al (1990) and Bloch (1991) noted the relationship between facies and permeability. The inuence of composition on porosity and permeability is reported by several workers (Seeman and Scherer 1984, Scherer 1987, Pittman and Larese 1987,
3

Present address: Curtin University of Technology, Department of Petroleum Engineering, Technology Park West, Kensington, Western Australia 6151, Australia.

Marion et al 1989, Smosna 1989 and Ahmed et al 1991). The inuence of textural parameters and sedimentary structures on permeability was studied by Chilingar (1964), Berg (1970), Beard and Weyl (1973), Bloch (1991), Panda and Lake (1994) and Hurst and Rosvoll (1991). Many attempts have been made to establish a relationship between porosity and permeability (Berg 1970, Bloch 1991). Other parameters which have been used to estimate permeability include the surface area of the grain spaces (Carmen 1937, Johnson et al 1987 and Schwartz and Banavar 1989), formation resistivity factor (Archie 1942, Katz and Thompson 1986), capillary pressure (Dullien 1979, Swanson 1981, Serra 1984, Pittman 1992, Rezaee and Lemon 1997), nuclear magnetic resonance relaxation time (Ahmed et al 1991), sonic transit time (Iverson 1990, Vernik and Nur 1991) and pore throat characteristics derived from image
Printed in the UK

1742-2132/06/040370+07$30.00

2006 Nanjing Institute of Geophysical Prospecting

370

Relationships between permeability, porosity and pore throat size in carbonate rocks using regression analysis and neural networks

analysis of thin sections (Doyen 1988, Ehrlich et al 1991, Rezaee and Grifths 1996, Rezaee and Lemon 1996). Many studies have recently been targeted to predict permeability from well log data using articial neural networks methods (e.g., Mohaghegh et al 1995, Huang et al 1996, Kadkhodai et al 2005, Helle et al 2001, Zhang et al 2006). Core analysis under ambient or reservoir conditions is a common method for direct measurement of permeability. Another method of measuring permeability is well testing. Due to their high costs, only a limited number of core analyses and well tests are available in any eld (Mohaghegh et al 1994). Cuttings are available in almost all wells. The mercury injection technique may be used on well cutting or chips (Jennings 1987). As the reservoir properties such as porosity and permeability are controlled by the size and arrangement of pores and throats (McCreesh et al 1991), the mercury injection method is commonly employed to characterize pore size distribution and permeability in porous media (Kolodzie 1980, Swanson 1981, Katz and Thompson 1986 and 1987, Pittman 1992). Most of these studies have been for sandstones, and there is lack of comprehensive study for carbonates. In this study, regression analysis and articial neural network (ANN) computing techniques were used to predict permeability from porosity and pore throat sizes derived from mercury injection analysis for carbonate rocks. The results of both methods were then compared. This paper also briey discusses the differences in terms of permeability, porosity and pore throat radii relationships in carbonates and sandstones.

where Kair is air permeability (mD), SHG is the mercury saturation (%) corresponding to the apex of the hyperbola and PC is capillary pressure (psi). Katz and Thompson (1986, 1987) reported the following relationship: K = 1/226(lc2) (S/S0 ), (4) where K is air permeability (mD), lc is the characteristic pore size (e.g. the calculated pore size (mm) for the threshold pressure at which mercury forms a connected pathway through the sample), and S/S0 is the ratio of rock conductivity to the conductivity of the formation water. This approach, however, requires a rock sample, laboratory measurement of the threshold pressure and measurement of rock and formation water conductivity. Winland developed an empirical relationship between porosity, air permeability and throat size corresponding to a mercury saturation of 35% (r35). This relationship was published by Kolodzie (1980) as: Log r35 = 0.732 + 0.588 Log Kair 0.864 Log , (5) where r35 corresponds to the pore-throat radius (microns) at 35% mercury saturation from a mercury injection capillary pressure test, Kair is the uncorrected air permeability (mD) and is porosity (%). Pittman (1992) extended Winlands work and introduced a series of equations extracted from the multi-regression analyses of mercury injection, permeability and porosity data of 202 sandstone samples. He pointed out that the equation (6) yields the best correlation coefcient for permeability, porosity and mercury injection data. Log K = 1.221 + 1.415 Log + 1.512 Log r25 (6) where K is the uncorrected air permeability (mD), is porosity (%) and r25 is the pore throat corresponding to the 25th percentile of saturation on a cumulative mercury injection plot.

2. Background
Several workers have attempted to correlate capillary pressure data with permeability. Washburn (1921) expressed the relationship between mercury capillary pressures and pore throat radii as: PC = 2 Cos /rc , (1)

3. Data analysis and preparation


A set of 144 carbonate core samples with a wide range of porosity and permeability (gure 1) was selected from several oil elds in central and south-west Iran. Samples show a wide variety of textures including mudstone, wackestone, packstone, grainstone and dolostone, with different crystal sizes. Porosity types were generally ne intercrystalline, vuggy and moldic. Selected samples were cleaned by toluene in a Soxhlet apparatus and dried at 60 C for 24 h prior to any analysis. Porosity and permeability were measured using an UltraPorosimeter 200 A, an Ultra-Permeameter. Core porosity values ranged from 0.5 to 33.5% with a mean value of 11%. The range of permeability was between 0.006 and 414 mD with a mean value of 21 mD. Mercury-air capillary pressure tests were conducted to estimate pore-throat radius with maximum pressure of 1500 psi. Mercury injection capillary pressures were measured using a Micromeritics Autopore. A mercury injection curve is a curve which represents the increasing saturation of mercury as a function of pressure. Mercury injection curves can 371

where PC is capillary pressure, is mercury surface tension, is contact angle and rc is the radius of the pore throat being intruded by mercury. Considering = 480 (dynes/cm), = 140 for mercury/air, the equation becomes: PC = 107.6/rc , (2)

where PC is in psi and rc is in micrometers. In this equation, the contact angle commonly used is 140 (Ritter and Drake 1945, Juhola and Wiig, 1949). Purcell (1949) related capillary pressure empirically to air permeability through the graphical integral of the curve of mercury saturation versus reciprocal capillary pressure squared. Swanson (1977) noticed that the complete saturation of effectively interconnected pore spaces with a non-wetting phase (Woods metal) corresponded to the apex of the hyperbola of a log-log mercury injection capillary pressures curve. Swanson (1981) empirically expressed the relationship between permeability and the hyperbola of the log-log mercury injection capillary pressures curve by the following equation: Kair = 339(SH G /PC )1.691 , apex (3)

M R Rezaee et al
60

(A)
50
Number of samples

40

30

20

10

0 <5 5 --10 10 --15 15 --20 20 -- 25 25 -- 30 >30


Porosity (percent)

60

(B)
50
Number of samples

40

30

20 10

0 <0.1 0.1--1 1--10 Permeability (mD) 10 --100 >100

Figure 1. Histograms displaying the distribution of porosity (A) and permeability (B) values used in this study.

indicate the amount of porosity existing behind pore throats of various sizes (Purcell 1949). As mercury is a non-wetting phase, it must be forced to enter the pores of the samples. Pores with the largest throats will be the rst to be penetrated by mercury under increasing hydrostatic pressures. Pores connected with the smallest throats are the last parts of the effective porosity that may be invaded and lled by mercury, at the highest pressures. Mercury injection capillary curves vary signicantly in shape (gure 2), suggesting wide variety in the samples in terms of pore throat size distribution. Pore throat radii were calculated with a certain mercury saturation, r5 to r65, for the samples using Washburns equation (equation (1)). Since some samples were tight and were not completely saturated by the mercury with increasing pressure, the number of data points decreased at higher mercury saturations.

4. Regression analysis
Regression analysis has frequently been used as the main tool with which to correlate porosity and permeability values. With this approach, it is generally assumed that a linear or 372

non-linear relationship exists between permeability, porosity and depth. In the other words, it is assumed that a known linear or non-linear function is sufcient for modelling the relationship between permeability and other properties of the rock (Jamialahmadi and Javadpour 2000). Reservoir engineers and petrophysicists are interested in how permeability and porosity relate to pore throat size and pore throat distribution, primarily so they can estimate permeability (Pittman 1992). Mercury saturations from 5 to 65% were used to evaluate the relationship between permeability, porosity and pore throat radius at each saturation. Statistical software was used to establish various relationships between porosity and pore throat size as predictor and permeability as response. The approach was to develop empirical equations for calculating permeability from porosity and pore throat radius at several mercury saturation percentiles. Table 1 lists the equations developed from the regression analysis and the regression correlation coefcient between the core and the permeability calculated from the equations. The correlation coefcient increases from r5 to r45 and is the highest for r50 and then decreases at higher mercury saturations. Figures 3 and 4 compare calculated and measured permeability using the r50 equation of table 1.

Relationships between permeability, porosity and pore throat size in carbonate rocks using regression analysis and neural networks
1500 1400

3.0 R=0.78 2.0

1300

Log K (Regression)

1.0

1200 1100 1000

0.0

Capillary Pressure (psi)

-1.0

900 800 700 600 500 400 300 200 100 0 100

-2.0

-3.0 -3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

Log K (Measured)

Figure 3. Cross-plot of measured and calculated permeability using the r50 equation. Table 1. Empirical equations developed from regression analysis. Correlation coefcient between measured and calculated permeability varies for different pore throat sizes. Equations Log K = 2.030 + 1.930 Log + 0.288 Log r5 Log K = 1.920 + 1.930 Log + 0.379 Log r10 Log K = 1.710 + 1.830 Log + 0.378 Log r15 Log K = 1.610 + 1.770 Log + 0.479 Log r20 Log K = 1.460 + 1.730 Log + 0.484 Log r25 Log K = 1.460 + 1.800 Log + 0.612 Log r30 Log K = 1.390 + 1.790 Log + 0.630 Log r35 Log K = 1.310 + 1.740 Log + 0.702 Log r40 Log K = 1.270 + 1.770 Log + 0.825 Log r45 Log K = 1.160 + 1.780 Log + 0.930 Log r50 Log K = 1.020 + 1.720 Log + 0.861 Log r55 Log K = 0.972 + 1.730 Log + 0.855 Log r60 Log K = 0.995 + 1.780 Log + 0.754 Log r65 Correlation coefcient 0.676 0.698 0.696 0.714 0.720 0.754 0.743 0.755 0.758 0.787 0.770 0.740 0.692

90

80

70

60

50

40

30

20

10

Mercury Saturation (%)

Figure 2. Variation of mercury injection capillary pressure curves for the carbonate samples studied.

5. Articial neural network (ANN)


The ANN is one of the latest methods available to the petroleum industry. It is a biologically inspired computing scheme and can solve complicated problems which are not amenable to conventional numerical methods. The rst conceptual elements of neural networks were introduced in the mid 1940s, and the concept developed gradually until the 1970s (Al-Qahtani 2000). ANNs usually consist of many processing nodes that are connected by weights. Neural networks are used in many industries today to solve a range of

problems including pattern recognition, regression analysis and data clustering (Patterson 1996, Haykin 1999). In the oil industry, neural networks are now routinely used in seismic reservoir characterization and seismic pattern analysis (Wong and Nikravesh 2001, Van der Baan and Jutten 2000,

Log K (Measured) 3.0 2.0 1.0

Log K (Regression)

Log K

0.0 -1.0 -2.0 -3.0


1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141

Figure 4. A comparison between measured (dotted line) and predicted permeability (continuous line) using the r50 equation of table 1. In general, the predicted and measured permeability follows the same trend.

373

M R Rezaee et al
Hidden Layer Input Layer Output Layer
Log Log K Log r

3.0 R=0.90 2.0

Log K (Measured)

1.0

0.0

-1.0

Figure 5. Architecture of a back propagation ANN with two nodes in the input layer, six nodes in the hidden layer and one node in the output layer.

-2.0

-3.0 -3.0

-2.0

-1.0

0.0 Log K (ANN)

1.0

2.0

3.0

Mohaghegh et al 1995). Since ANNs process data and learn in a parallel and distributed fashion, they are able to discover highly complex relationships between several variables that are presented to the network (White et al 1995). The arrangement in an ANN is generally layer-based. An input layer consisting of input articial neurons or so-called processing elements (PEs), an output layer, and a number of hidden layers between the inputs and outputs. Though this architecture is almost always the same, the exact number of hidden layers and PEs in the layers varies from problem to problem. Neural networks can be classied by the way they are trained, using either supervised or unsupervised learning. In supervised learning the neural network starts with a training dataset for which we know both the input and output values. The neural network algorithm then learns the relationship between the input and output from this training dataset, and nally applies the learned relationship to a larger dataset for which we do not know the output values. In unsupervised learning, the neural network is presented with a series of inputs and then looks for patterns itself. Major applications of neural networks are in seismic inversion, log analysis and 3D reservoir modelling (Wong and Nikravesh 2001). For tasks like permeability estimation, supervised algorithms are generally preferred (Wiener 1995). For this reason we used a supervised back-propagation neural network (BPN). In this study, we dened a three-layer network (gure 5) with a sigmoid transfer function in the hidden layer and a linear function in the output layer. Then data set was normalized to range [1, +1] and divided to three parts, including training (50% of data), validation (25% of data) and testing (25% of data). In the multi-layer perceptron case, the training set is used to nd the optimal weights with the back propagation rule. The validation set was used to nd the optimal number of hidden layers or to determine a stopping point for the back propagation algorithm. The test set was in turn used to provide examples for the sole purpose of assessing the performance of a fully trained classier. Because of the vast distribution of the permeability data, logarithmic values were used. At each stage, the logarithm of porosity and pore throat radius in a spread of mercury saturation percentiles was fed as input and logarithm of the permeability as output. Table 2 lists the 374

Figure 6. Cross-plot of laboratory measured permeability and ANN predicted permeability. The correlation coefcient is larger than that in the regression analysis. Table 2. List of input parameters, correlation coefcients between measured and predicted permeabilities and performance function (mse = mean square error). Inputs Log Log Log Log Log Log Log Log Log Log Log Log Log and Log r5 and Log r10 and Log r15 and Log r20 and Log r25 and Log r30 and Log r35 and Log r40 and Log r45 & Log r50 and Log r55 and Log r60 and Log r65 Correlation coefcient (R) 0.778 0.798 0.804 0.838 0.841 0.846 0.832 0.844 0.832 0.900 0.851 0.794 0.712 performance function (mse) 0.626 0.566 0.530 0.450 0.430 0.426 0.431 0.397 0.412 0.356 0.432 0.421 0.565

results of permeability prediction using the ANN. Figures 6 and 7 compare laboratory measured permeability with ANN predicted permeability using outputs of r50.

6. Discussions and conclusions


This study shows that in the absence of cores, using well cuttings and determination of porosity and pore throat radius derived from mercury injection in carbonate rocks, we will be able to estimate permeability using regression analysis and neural networks. The results from the neural networks method are better than those from regression analysis, and have proven to be successful in the prediction of permeability in carbonate rocks. Provided that sufcient and representative data is given to the neural network for training, then permeability can be reasonably predicted in uncored intervals located in the same geological setting as the wells from which the training data was derived.

Relationships between permeability, porosity and pore throat size in carbonate rocks using regression analysis and neural networks
Log K (Measured) Log K (ANN)

3.0 2.0 1.0


Log K

0.0 -1.0 -2.0 -3.0


1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 101 106 111 116 121 126 131 136 141

Figure 7. A comparison between measured (dotted line) and predicted (continuous line) permeability using the r50 equation. There is a good match between the permeability values.

This study shows that pore throat radius at 50% mercury saturation (r50) yields the best correlation coefcient for permeability, porosity and pore throat radii for carbonates (tables 1 and 2). Winland (Kolodzie 1980) and Pittman (1992) have proposed r35 and r25 respectively as the best permeability estimator for sandstones. A possible explanation for this difference is the pore network complexity of carbonates compared to sandstones. Generally speaking, the main type of porosity in sandstones is intergranular, whereas for carbonates porosity type varies widely (Choquette and Pray 1970). In carbonates, diagenetic processes cause very complex networks of pores and pore throats. For a simple network of pores and pore throats in sandstones, pore throat size values derived from lower mercury saturations (e.g. 25 and 35%) can be fair representatives for permeability estimation. But for carbonates with higher pore network complexity, pore throat size value at higher mercury saturations appears to be the best pore throat representative for estimating permeability.

Acknowledgments
The Research Institute of Petroleum Industry (RIPI), University of Tehran is acknowleged for its support. The authors acknowledge the RIPI for permission to publish this paper. The authors appreciate valuable comments from an anonymous referee of the Journal of Geophysics and Engineering, which improved the article.

References
Ahmed U, Crary S and Coates G 1991 Permeability estimation: the various sources and their interrelationship Journal of Petroleum Technology 43 57887 Al-Qahtani F A 2000 Porosity Distribution Prediction Using Articial Neural Networks, MSc thesis, Morgantown Virginia University Archie G E 1942 The electrical resistivity log as an aid in determining some reservoir characteristics Transactions of AIME 146 5462 Atkinson G D, Bloch S and Scheihing M H 1990 Sedimentology and depositional environments of the Oligo-Miocene gas-bearing succession in the Yacheng 131 eld, South China Sea, Peoples Republic of China AAPG Bulletin 74 601 Beard D C and Weyl P K 1973 Inuence of texture on porosity and permeability of unconsolidated sand AAPG Bulletin 57 34969

Berg R 1970 Method of determining permeability from reservoir rock properties Gulf Coast Association of Geologic Society Transaction 20 30317 Bloch S 1991 Empirical prediction of porosity and permeability in sandstones AAPG Bulletin 75 114560 Carmen P 1937 Fluid ow through granular beds Institute of Chemical Engineers Transactions 15 15066 Chilingar G 1964 Relationship between porosity, permeability and grain size distribution of sands and sandstones Deltaic and Shallow Marine Deposits ed L van Straaten (New York: Elsevier) pp 715 Choquette P W and Pray L C 1970 Geologic nomenclature and classication of porosity in sedimentary carbonates AAPG Bulletin 54 20750 Doyen P 1988 Permeability, conductivity and pore geometry of sandstone Journal of Geophysical Research 93 772940 Dullien F 1979 Porous media, uid transport and pore structure (New York: Academic Press) pp 4265 Ehrlich R, Crabtree S J, Horkowitz K O and Horkowitz J P 1991 Petrography and reservoir physics I: Objective classication on reservoir porosity AAPG Bulletin 75 157462 Haykin S 1999 Neural Networks: A Comprehensive Foundation Prentice-Hall, Inc Helle H B, Bhatt A and Ursin B 2001 Porosity and permeability prediction from wireline logs using articial neural networks: a North Sea case study Geophysical Prospecting 49 43144 Huang Z, Shimeld J, Williamson M and Katsube J 1996 Permeability prediction with articial neural network modeling in the Venture gas eld, offshore eastern Canada Geophysics 61 42236 Hurst A and Rosvoll K J 1991 Permeability variations in sandstones and their relationship to sedimentary structures Reservoir Characterization II ed L W Lake, H B Carroll Jr and T C Wesson pp 16696 Iverson W 1990 Permeability estimation from sonic versus porosity gradients Society of Exploration Geophysicists International Meeting BG52 1069 Jamialahmadi M and Javadpour F G 2000 Relationship of Permeability, Porosity and Depth Using an Articial Neural Network Journal of Petroleum Science and Engineering 26 2359 Jennings J B 1987 Capillary Pressure Techniques: Application to Exploration and Development Geology AAPG Bulletin 71 1196209 Johnson D, Koplik J and Dashen R 1987 Theory of dynamic permeability and tortuosity in uid-saturated porous media Journal of Fluid Mechanics 176 379402 Kadkhodai A, Rezaee M R and Moallemi S A 2005 A New Approach for Prediction of Porosity, Permeability and Rock Types Using Soft Computing Techniques, An Example from Southern Persian Gulf, Petroleum Geoscience Collaboration Conference, London 30 November1 December

375

M R Rezaee et al

Katz A J and Thompson A H 1986 Quantitive Prediction of Permeability in Porous rock Physical Review B 34 817981 Katz A J and Thompson A H 1987 Prediction of Rock Electrical Conductivity from Mercury Injection Measurments Journal of Geophysical Research 92 599607 Kolodzie S Jr 1980 Analysis of pore throat size and use of the Waxman-Smits equation to determine OOIP in Spindle Field, Colorado Society of Petroleum Engineers, 55th Annual Fall Technical Conference Paper 9382, 10p Ligtenberg J H and Wansink A G 2001 Neural Network Prediction of Permeability in the EL GARIA Formation, Ashtart oil eld, offshore Tunisia, Journal of Petroleum Geology 24 389404 Marion D, Nur A and Albert F 1989 Modelling the relationship between sonic velocity, porosity, permeability, and shaliness in sand, shale and shaly sand SPWLA Thirtieth Annual Logging Symposium Proceeding paper G 122 McGowen J H and Bloch S 1985 Depositional facies, diagenesis and reservoir quality of Ivishak sandstone, Prudhoe Bay eld (abs) AAPG Bulletin 69 276 Mohaghegh S, Are R, Ameri S and Rose D 1994 Design and Development of an Articial Neural Network for Estimation of Formation Permeability SPE 28237 Proc. of SPE Petroleum Computer Conf, Dallas, July 31-August 3 Mohaghegh S, Balan B and Ameri S 1995 State-of-The-Art in Permeability Determination from Well log Data: part 2Veriable, Accurate Permeability Predictions, the Touch-Stone of All Models SPE Eastern Regional Conference and Exhibition Sept 1721, West Virginia, SPE 30979 Patterson D W 1996 Articial Neural Networks: Theory and Applications Prentice-Hall, Inc Pittman E D 1992 Relationship of Porosity and Permeability to various Parameters Derived from Mercury Injection-Capillary Pressure Curve for Sandstone AAPG Bulletin 76 1918 Pittman E D and Larese R E 1987 Experimental compaction of lithic sands SEPM Annual Midyear Meeting 4 66 Purcell W R 1949 Capillary Pressures-their measurement using mercury and the calculation of permeability there from American Institute of Mechanical Engineers, Petroleum Transactions pp 3948 Rezaee M R and Grifths C M 1996 Pore Geometry Controls on Porosity and Permeability in the Tirrawarra Sandstone

Reservoir, Cooper Basin, South Australia AAPG Annual Convention, San Diego, May 1922, 1996 A119 Rezaee M R and Lemon N 1997 Permeability Estimation from Mercury Injection Capillary Pressure Data, a case study in the Tirrawarra Sandstone, Cooper Basin Australian Petroleum Production and Exploration Association Journal 37 824 Scherer M 1987 Parameters inuencing porosity in sandstones: a model for sandstone porosity prediction AAPG Bulletin 71 85491 Schwartz L and Banavar J 1989 Transport properties of disordered continuum systems Physical Review B 39 1196570 Seeman U and Scherer M 1984 Volcanoclastics as potential hydrocarbon reservoirs Clay Mineralogy 9 45770 Serra O 1984 Fundamentals of Well-Log Interpretation (New York: Elsevier) Smosna R 1989 Compaction law for Cretaceous sandstones of Alaskas North Slope Journal of Sedimentary Petrology 59 57284 Swanson B F 1977 Visualizing pores and non-wetting phase in porous rocks Society of Petroleum Engineers Annual Fall Technical Conference, SPE Paper 6857 10p Swanson B F 1981 A Simple Correlation between Permeability and Mercury Capillary Pressures Journal of Petroleum Technology Dec 2488504 Van der Baan M and Jutten C 2000 Neural networks in geophysical applications Geophysics 65 103247 Vernik L and Nur A 1991 Lithology prediction and storage/transport properties evaluation in clastic sedimentary rocks using seismic velocities (abs) AAPG Bulletin 75 384 White A C, Monlar D, Aminian K, Mohaghegh S, Ameri S and Esposito P 1995 The Application of ANN for Zone Identication in a Complex Reservoir SPE 30977 Wiener J 1995 Predict Permeability from Wire line Logs using Neural Networks Petroleum Engineering International May 1824 Wong P M and Nikravesh M 2001 Field applications of intelligent computing techniques Journal of Petroleum Geology 24 38187 Zhang J, Li L and Ping S 2006 Neural Approach for Calculating Permeability of Porous Medium Chinese Phys. Lett. 23 100911

376

You might also like