You are on page 1of 11

usm

024223 Supplement t o SPE 19399, Improved Data-Analysis Method Determines A r c h i e Parameters From Core Data R.E. Maute, M o b i l R&D Corp.; W . D . L y l e , M o b i l R&D Corp.; E.S. Sprunt, M o b i l R&D Corp.

Copyright Society of Petroleum Engineers This manuscript was derived directly from SPE which appeared this year in a Society of Petroleum Engineersjournal. The material in this Supplement passed SPE peer review with the published paper. Permission to copy is restricted to an abstract of not more than 300 words. Write SPE Book Order Dept., Library Technician, P.O. Box 833836,Richardson, TX 75083-3836 U.S.A. Telex 730989 SPEDAL.

L ,

SPE 24223

SUPPLEMENT TO SPE 19399 "Improved Data Analysis Method of Determining Archie Parameters 'm' and 'n' from Core Data"

by R. E. Maute, W. D. Lyle, and E . S. Sprunt This supplement contains the detailed mathematics of CAPE with equations that will implement the method, as well as program flow diagrams. Please note that other mathematical methods of solving the problem could have been used. The basic problem is to determine the Archie parameters by minimizing the mean-square enor between the measured saturations and the saturations computed using the Archie equation. As noted in the main text, there are two separate cases to be considered. In the first case, the parameter a is set equal unity, and the mean-square saturation error defined by
P @

=p tlt L=1 [ 5 i j

-p@brn) ]
lln
2

ij

is to be minimized, where the jindex sums over the P cores that were measured, and the i index sums over the number of measurements Qj made on each core. If a is not fmed at unity but is allowed to be a fitting parameter, the mean-squared error % to be minimized is defined by

where the indices are as described above. In both equations (A.1) and (A.2), SGdenotes the ith measurement of water saturation of core j, RGdenotes the corresponding elecmcal resistivity, and R,, denotes the water resistivity of core j. Minimum values of the and % occur when the appropriate partial derivatives are set equal to zero. In the case, the requirements are

and in the E;! case the requirements are (A.4.)

Considering first the

case, the two partials in equation (A.3) lead to

and

The notation for this problem as well as the problem for E;? is considerably simplified by introduction of new functions g and h defined by

and

Using equation (A.7) along with the observation that the 2/n term does not affect the solution for m and n, the simplified form of equations (A.5) and (A.6) becomes

(A.9) (A. 10)


where F1 and F2 are defined in the above equations. Equation (A.10) can be simplified by noting that

Consequently, the portion of the summation of equation (A.10) involving In* is zero from equation (A.9). This leads to an equivalent expression for F1 as (A. 12)

The above expression for F1will be used in the remainder of this appendix. Inspection of equations (A.9) and (A. 12) reveal their obvious nonlinear nature. Therefone, an analytical solution form and n is not possible. A numerical solution can be obtained by linearizing F1and F2about some point near the true solution. The linearization is accomplished by expanding the functions in a fxst order Taylor series. For F1the series becomes

With a similar expression for the expansion for F2. The derivatives in equation (A.13), as well as the corresponding derivatives in the F2expansion are evaluated at the point mk, nk. Since the solutions of equations (A.9) and (A. 12) require that both F1and F2be equal zero, equation (A. 13) and the corresponding equation for F2form the basis for an iterative numerical solution by observing that if F1and F2were approximately zero at the point mk+l, nk+l then

(A. 14) and (A. 15) The above two equations can be easily solved to yield for mk+l and nk+l the expressions

(A. 16)

SPE
4

2422-3

Equation (A. 16) above is the basic equation to solve for the m and n that minimizes of equation (A. 1). A starting point mo,no is required along with a stopping rule of the form

and

for some selected value of 61 and &. In practice the value of m 0 and n 0 = 2 works well with the values of c$ and 4 chosen to be .001. The partial derivatives required in equation (A. 16) can be determined from equations (A.9) and (A.12). These derivatives have been evaluated and found to be

(A. 19)

aF1 ---am and

n ZZ (Sij- 2gij ) gij

'

Rwj
'J

Again, it is worth noting that in equation (A. 16) the derivatives of equations (A. 19) through (A.22) are evaluated at the point m = mk and n = nk.

In the case of e,there are three parameters (a,m,n) that are solved for in order to minimize equation (A.2). In exactly the same manner as was done for the first case, an iterative algorithin can be developed by defining the new functions HI, H2, and H3 that are determined from the t!hree partial derivatives of equation (A.4). These are

The new functions H1, H2, and H3 are thus defined by

and

H 3 = Z Z ( S Q -h i j ) h i j = O ,
where he was defined in equation (A.8). Again, as was the case for F1, there is a considerable simplification possible in the above expressions. This simplification is possible by noting that in equation (A.26) the term involving hi can be written as

using the definition of hi from equation (A.8). Defining a new function

zqby

equation (A.29) becomes hii= a l / n rl h.

SPE

24223

Inspection of equations (A.26) - (A.28) reveals that using equation (A.3I), a constant factor of alln can be factored out. Since the HI, H2, and H3 are each set equal to zero, the constant factor alJn will not influence the solution. Consequently, in terms of Eij the simplified form of equations

(A.26) - (A.28) become

A three-variable iteration algorithm similar to equation (A. 16) can be developed by expanding each of the functions H1, H2, and H3 of equations (A.32) - (A.34) in a linear series of three variables as was done for F1and F2. Canying out these operations results in the algorithm for solution of the a,m,n that minimizes as

In equation (A.35) the partial derivatives were evaluated using the H1, H2, and H3 of equations (A.32) - (A.34) and found to be

SPE

24223

and

Equation (A.35) along with the derivatives in equations (A.36) - (A.44) evaluated at mb nk, and akwill solve for the minimum of E;?.A stopping rule that has been found to work in practice is

as was the case for the solution for E

~ . S1,

6,and 63 typically have values of .001.

The initial point m,, no, and a, that works in practice is to set a, equal one with m,, no being two, or the values obtained from the solution for the E;? case.

A final point to be noted is that for the three-parameter case there is a requirement that the number of cores be greater than one. Otherwise, equation (A.35) will not converge since the mat r i x of partial derivatives is singular. This can be seen from equations (A.33) and (A.34) for the single core case. In this case, equation (A.33) becomes

Clearly, in this special case, the equations for H2 and H3 are identical since equation (A.34) for H3 becomes

which is identical to equation (A.47). Thus, in order to apply the three-parameter a,m,n algorithm, it is necessary to have data from two or more cores.

FLOW CHART - m AND n DETERMINATION

WE

24223

0
START

SET POR, SW BOUNDS ON DATA TO BE USED (DEFAULT, USE ALL DATA)

SET 6

6 21TERATION CUTOFF

I
USE INITIAL CUES S mo= 2, no= 2. SET ITERATION COUNTER k = 1.

YES

PRINT FINAL m, n

FLOW CHART

- m, n, AND a DETERMINATION

SFBE 24 2 2

0
START

SET POR, SW 'BOUNDS ON DATA TO BE USED (DEFAULT, USE ALL DATA)

SET S 1, S2, S31TERATION CUTOFF

USE INITIAL GUES S (e.g., m0=2, n0=2, ao=l). SET ITERATION COUNTER k=l.

COMPUTE mvl, n

~ a@,

PRINT FINAL m, n, a

You might also like