Sei sulla pagina 1di 25

SPWLA

32nd Annual Logging Symposium,

June 16-19, 1991

An Adaptive Borehole Correct ion


Scheme For Array Induction Tools
G. P. Grove and G. N. Minerbo
Schlumberger

Abstract
Multiarray
induction tools, as well as all other induction conductivity
measuring
devices, are affected by the environment.
Correcting
for the borehole effects on
induction measurements
requires accurate knowledge of the borehole signal or conductivity contribution.
Available external measurements
of borehole geometry do
not have sufficient accuracy to adequately correct for borehole effect in the presence
of moderate-to-large
borehole-formation
conductivity
contrasts.
The AIT ilrray
Induction Imager Tool measures in-phase and quadrature signals from multiple arrays operating at several frequencies.
This multitude of information
allows the
formulation of a self-consistent
borehole correction scheme that can determine the
borehole geometry and correct for the borehole signal even in difficult cases, such
as irregula,r-shaped

boreholes

with caves.

The borehole signal is primarily a function of mud conductivity,


borehole size and
shape, tool standoff from the borehole wall and formation conductivity.
To develop
the borehole corrections, a large number of model calculations of the AIT response
were performed over a wide range of each of the input variables. These calculations
were valida,ted against laboratory
measurements
taken in cylindrical plastic tanks
of different diameters filled with salt water.
An algorithm

that rapidly

computes

the borehole

response

of each of the raw mea-

surements was developed by fitting the results of the model calculations


to closed
form mathematical
expressions.
The borehole correction algorithm includes a nonlinear least-squares
optimization
technique that matches the measured data from
the shortest spacings arrays to the model calculations.
Any of the variables of the
borehole signal may be combined and varied to achieve this match. Results obtained
using the adaptive borehole correction scheme for modeled cases and field test wells
have validated the accuracy and reliability of the borehole signal determination
and
correction.
Introduction
The recently introduced AIT tool,l measures in-phase and quadrature signals from
multiple arrays operating at several frequencies.
The tool is composed of SOTera1 simple, mutually balanced arrays with a single transmitter
coil, whose main
coil spacings range from a few inches to several feet. The short arra.ys mcasurc the
* Mark

of Schlumberger

-1-

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

environment while the long arrays measure deep into the formation.
Signal processing using deconvolution filters and weighting functions can give deep
log responses with little borehole effect; however, shallow logs are strongly influThe nature of induction response makes it impossible to
enced by the borehole.

near-borehole

remove the influence of the borehole from the short arrays with signal processing
alone. To address this problem, a borehole correction scheme to operate on the
raw array measurements
was developed to permit maximum flexibility in any subThis borehole correction scheme determines
the
sequent processing of the data.
borehole geometry and corrects for the borehole signal even in difficult cases, such
as moderate-to-large
borehole-formation
constrasts and irregular-shaped
boreholes
with caves. One of the main purposes of a multiarray induction tool is to provide an
accurate estimation
of the resistivity radially into the formation.
The importance
of correcting all measurements
for borehole effects cannot be overemphasized
when
both shallow and deep induction arrays are combined to describe a radial resistivity
profile. Since shallow induction arrays have a large borehole effect, care must be
taken to properly correct these measurements
prior to any further processing.
To develop the AIT borehole corrections,
a large number of model calculations
of
the tool response were performed, over a wide range of borehole conditions.
Results
were validated against laboratory
measurements
taken in cylindrical plastic tanks
of different diameters filled with salt water.
The adaptive borehole correction scheme developed for the AIT tool features full
boreholc corrections
derived from borehole geometry mea.surements
over a wide
The processing uses information from the short arrays
range of xl,/n,
contrasts.
t.o solve for effective borehole parameters
in difficult borehole conditions or when
cst,crnal mea.slu-cments are unavailable or are of poor quality. The borehole signal
is prima.rily a. function of the mud conductivity,
the borehole size and shape, the
tool standoff from the borehole wall and the formation conductivity.
From the set
of modeled cases, an algorithm was developed to compute the borehole response of
each of the raw a,rrny mea,surements.
This correction algorithm includes a look-up
table covering a wide range of each of the input variables, as well as a nonlinear lea,stsqriares optimiza,tion
Theoretical
The model

technique

to match

the data from the short-spacing

arrays.

Model
used to compute

the AIT

response

for different

borehole

conditions

assumes t,hat, the borchole is an infinite circular cylinder in a homogeneous formation, with t,he t.ool either centered or eccentered in the borehole.
iln eccentered
tool further complicates
the model, as there is no longer azimuthal symmetry.
A
finite-element
code wa.s developed for this model, which includes the finite size of
the coils, the sonde housing and other construction
details tha.t also account for
eccentrici t,y.3
The procedure for correcting borehole effect uses this modeled data to correct the
tool measurements
made while logging in a wellbore. The forward model assumes a
-2-

SPWLA32ndAnnualLoggingSymposium,
June 16-19,1991

b orehole radius (T), tool standoff (x) and formation


known mud conductivity
(a,),
conductivity
(af); it then computes the tool response for those conditions.
The
first three parameters may be assumed known by other measurements;
however, af
cannot. Finding the conductivity of the formation is the goal of the AIT tool and is
determined after the borehole correction and subsequent processing.
The forward
model must be inverted, then, to solve for an effective formation conductivity
af
for a given set of measurements.
The inversion scheme used, minimizes in a leastsquares sense, the deviations of the model values (as a function of the formation
conductivity)
from the measured values.
Data tables were generated using this modeling code, giving the response of the tool
in different borehole geometries for both the in-phase and quadrature components
of each array, at all frequencies of that array. Each data table contains values that
vary with grn (0 to 68 S/m), r (0 to 10 in.), 2 (0 to 2.5 in.) and af (0 to 10 S/m),
making corrections possible in both water-base and oil-base muds. To practically
access this data, polynomial approximations
obtained by least-squares
fitting were
computed to generate all the data table numbers quickly, with minimum computer
storage requirements.
Finding

The Polynomial

Approximations

Contained in the data tables is the response of the AIT tool to a given formation
and borehole. The borehole signal response is determined by subtracting from the
total response the response of the tool in the same formation without the borehole.
1Ve attempt
to fit the borehole signal values for each of the arrays with the
polynomials.
In determining
the polynomial fitting, each of the four independent
and the dependencies of each must
variables (c m, r, 5, af) has its own functionality,
be found.
Mud Conductivitu

/a, ) Dependence

In a study of tool eccentricity on induction response, Gianzero derived an expression


for the apparent conductivity
measured by an eccentric tool in the absence of skin
effect.4 This expression has terms varying in urn as
12

um

(1-9 1

) rz =0,1,2.

flrn

This expression is not defined for urn = 0, and we are fitting the borehole signal
where data values must be zero when orn = af. To accommodate
this, the polynomial expansion P(c,)
chosen to do this fitting uses an overall multiplying factor
factor theory of
of (cm - af) and has additional terms to reflect the geometrical
induction5 of the type (0, - af)n. The resulting expression has the following form:

SPWLA 32nd Annual Logging Symposium,

June 16-19, 1991

where a; are the coefficients found when performing a least-squares


fit to the data,
a.nd K and L are the orders (number of terms) needed to obtain the desired accuracy
of fit.
The coefficients a; a.re uniquely determined in the least-squares
method of fitting
by minimizing the sum of the squares of the residuals.
The residual is defined as
the difference between a data point to be fitted and the corresponding
value from
the polynomial.
If there are n data points to be fitted and the polynomial chosen
1la.s m terms, this can be written

where Z is a vector of length m containing the coefficients to be determined,


c is a
vector of length m conta.ining the data values, and A is an m x n matrix containing
the terms of t,he polynomial.
The least-squares
solution for 2 can then be found by
minimizing the sum of the squares of the residuals, i.e.,

The solution

to this eqrlation

ca,n be expressed

as

2 = (ATA)-(AT;).

(5)

A routine written by Hanson from the SLATEC


software library7 wa.s chosen for
the evaluation in which Householder transformations
are used to perform a QR
factoriza.tion.
This routine can also detect and remove rank deficiencies from A
t,hnt, are not independent from the others within numerical uncertainty.
The desired a,ccura.cy of fit was the minimum of 1% relative error or 1 mS/m absolute
error. Figure 1 shows the excellent agreement when comparing the approximated
CT~,~
fit to the actual borehole signal data for a particular receiver coil configuration
with a 2.5-in. sta,ndoff in a 0.1-S/m formation.
The mud conductivity,
g,
versus the computed borehole signal for three different borehole radii.
Formation

Conductivitu

is plotted

fcr f 1 Dependence

To find the forrnation conductivity


dependence, the coefficients above (a;) were expanded in terms of af. The ability of the fitting routine to remove rank deficiencies
was .used to input several different types of terms, then to remove those terms providing the least improvement
to the overall fit. Removing rank deficiencies gives
the minimal number of terms with the maximum degree of accuracy.
Logarithmic
expansion terms in af as well as other types of terms were used to
obtain the best fit. In implementing
the fit, the expansions for both CT,, and aj
were used together to calculate the terms in the a;(af) expression used in Equation
6 directly.
-4-

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

For the adaptive borehole correction code to run in a reasonable amount of computer
time, the number of terms in the polynomial expansion must not be exceptionally
large.
This limitation
required a compromise between speed and acceptable
accuracy of fit, since the desired accuracy criterion of 1% or 1 mS/m could not be
satisfied for all points without using an excessively large number of terms. The final
terms chosen for each of the polynomials for the dependence on cm and af were a
subset of the terms in the following expression:

Borehole

Radius fr 1 and Tool Standoff 0)

Dependencies

The dependencies on T and x were analyzed in a similar manner. The representation (Q(r, x>) a g ain required inputing several different types of terms and using the
fitting routine to remove rank deficiencies.
As was done for the orn and af dependencies, the optimum number of terms for the polynomial were determined for the
r and 2 dependencies.
Fitting to each of the numerous data tables was approached
similarly.
Evaluating

the Full Fit

The full fit can be found now that the dependencies of the four input parameters
cnz, r, x, and af have been determined.
A polynomial can be written as a function
of the input parameters R(a,,
af, r, X) as follows:

(7)
where M and N are the number of terms needed.
are defined by the expansions,

The terms pi( gm, of)

Wh,

of)

and qj (r, X)

N
Q(r,x)

CniPi(gm!flf),
i=O

Ccjqj(r>x),

(8)

j=o

P
and the polynomials P and Q are truncated to the appropriate
terms for a given
table, selected as described above. The total number of terms for the polynomial
representing a table of data will be the number of terms used in the grn and af fit
times the number of terms used in the r and z fit.
Including all the terms from the polynomials P and Q in R may seem unnecessary,
since many of the resulting terms may not be needed to obtain the desired fit.
However, for efficiency in the calculation,
this is a useful polynomial representation
-5-

SPWLA 32nd Annual Logging Symposium,

June 16-19,1991

since some terms may be stored and not recomputed


change over consecutive depth levels.
Smoothness

if the associated

inputs do not

of Fit

Smoothness
of fit is a prime consideration
when fitting data to a polynomial exA
simple
quadratic
fit
to
a
set
of
points
may give a smooth fit, although
pression.
some of the points may not fit exactly.
A cubic fit to the same set of points may
fit the points exactly; however, oscillations can be introduced between data points
that do not reflect the intended functionality
or yield a smooth fit. To ensure that
unwanted artifacts from the fitting routine were not introduced into the data, the
polynomial fit was plotted with a density three times that of the computed dat,a
points and compared with a plot of the original computed data points. These threedimensional plots, shown in Fig. 2, illustrate the smoothness of fit in urn and af
obtained using the full polynomial expression in Equation 7. This example shows a
medium-spacing
coil configuration with a borehole radius of 4 in. and a tool standoff of 0.5 in. Similarly, the smoothness of fit in T and z, using the full polynomial fit
described above, is shown in Fig. 3. This example shows the same coil configuration with a mud conductivity of 10 S/ m and a formation conductivity
of 0.01 S/m.
Unphysical combinations
of borehole radius and tool standoff are plotted as zero in
this example. In general, the fit is excellent, although a slight artifact introduced by
the fitting routine is seen at values of small borehole radius and small standoff. This
artifact is the result of the availability of only a small number of valid computed
data points for borehole radii of 2 and 3 in. with standoff, giving the polynomia.1
fit more freedom between fewer data points. This effect may be practically ignored,
however, as it is restricted to a very isolated region of the total data space.
Accuracy

and Reliability

While determining the full polynomial fit, a compromise between accuracy and computational speed was necessary to produce a practical borehole correction scheme.
Consequently,
the accuracy of fit in some cases was slightly degraded, and these
cases were examined to determine the impact on the total borehole correction.
In
general, a.n accuracy of 1% or 1 mS/m is desired; however, in ma.ny cases, this high
degree of accuracy is unnecessary since errors caused by other input measurements
can be large enough to mask any error in the fitting accuracy. For example, when
large boreholes are encountered,
minute errors in the borehole radius or irregularities in the borehole geometry can generate errors much larger than those clue to
any degraded a,ccura.cy in the polynomial fit.
The majority of points unable to meet the accuracy of fit criterion were for a tool
This isolated corner of the
standoff of 0.0 in. and a small borehole radius.
data space where these points lie should not be encountered
in practice, since the
optimum st,andoff is on the order of 1 to 2 in. depending upon borehole parameters.
To identify the exact circumstances
under which a degrada.tion in the accuracy of
fit occurs in the isolat,ed a.rea, error maps were produced to identify the magnitude
-6-

SPWLA 32nd Annual Logging Symposium, June 16-19,1991

and percentage
of error for specific borehole parameter values. These maps were
generated for all cases unable to meet the desired accuracy of fit criterion.
Figure 4 shows two gray-scale plots representing
the magnitude
of error for the
medium-spacing
coil configuration
shown in Figs. 2 and 3, with a borehole radius
of 2 in. and a tool standoff of 0.0 in. The error is plotted as a function of grn and
af, as a gray scale in the z-axis. The absolute error is shown in the plot on the
left, and the percentage error is shown in the plot on the right. The dark shade of
gray (approaching black) indicates points where the accuracy of fit criterion was not
met. In this example, there are points in the low formation conductivity
- high mud
conductivity
region that did not meet the accuracy of 1% error, but were within
the 1 mS/m absolute error limit. All cases were examined to ensure that any errors
would not cause problems in performing the AIT borehole corrections.
Centered

Tool

Case

When the tool is centered in the borehole, azimuthal symmetry exists that allows
the borehole to be described by the mud conductivity
and the borehole radius,
thus making the polynomial expression describing the borehole signal a function of
only three parameters P(am, T, ar). This makes the parameter dependencies much
simpler, and after determining a suitable set of terms applying the same methods as
in the decentered case, the fits were made using Chebyshev polynomials, which were
chosen for the recurrence relationship that can be used to develop simple, efficient
code. The complete set of terms for which all polynomials
were derived may be
written

as follows:

where Z~(X) is the Chebyshev polynomial of the first kind, and the argument 2
S ince the polynomial fits were obtained with the desired
lies in the domain [-l,l].
accuracy of fit using a relatively small number of terms, the extensive error analysis
described for the decentered case was not required.
Estimating

the Formation

Conductivity

The fitting polynomials determined above provide


tionally equivalent to the original modeling code
in a borehole to within the fitting accuracy. Since
af is unknown prior to performing the borehole
-7-

a fast forward model that is funcdeveloped for an eccentered tool


the true formation conductivity
corrections,
an inversion of this

SPWLA 32nd Annual Logging Symposium,

June 16-19,1991

fast forward model is required. This inversion model is based on the AIT raw measurements and the three known parameters,
cm, r and CC. From this information,
an optimum average formation conductivity af can be determined to calculate the
borehole correction directly from the forward model.
When external measurements
are provided, data for the mud conductivity
grn is
obtained from a mud conductivity
sensor, data for the borehole radius r is obtained from a mechanical caliper, and tool standoff z is estimated from geometrical
considerations
of the tool in the borehole.
In determining the optimum ;5f, a nonlinear least-squares solver is used to minimize
a penalty function E, which is the sum of the squares of a weighted difference
between the response measured by the tool and the response predicted by the model.
The penalty function has the following form:

is the complex
where Vi,,,
of the jth coil configuration.

signal (in-phase and quadrature) from the measurement


VAodel is the polynomial representation
of the forward

signal of the jth coil configuration,


V,lb,, is the signal that
coil
configuration
in
a
homogeneous
formation with
would be measured by the j
no borehole,
ej is the estimated
variance in the measured signal of the j coil
configuration,
and N is the number of coil configurations
used in the minimization.
model for the borehole

The estimated variance in the measured signal ej is comprised of a percentage of the


real part of the measured signal, which is the expected accuracy for the real part of
the model plus the expected absolute accuracy for the real part (to prevent 63 from
being zero) plus similar terms for the imaginary part. By putting more weight on
the small terms in the expression, the small measurements
are normalized with the
large measurements.
A subset of all the AIT measurements
is used in the minimization.
The conductivity just outside the borehole wall has the strongest effect on the borehole signal;
therefore, the shorter arrays are used to estimate the average conductivity
used in
Once the estimate of Y?f is made, the corrected input
the borehole corrections.
measurements

V&_,, d are determined


VcJ^orrld= VL,,

as follows for all array signals:


- V&del(gm,r,

Cf).

(11)

The Solver
The
The solver finds i7f, as described above, using the correction polynomials.
routine used to minimize the penalty function and optimize 5f is a nonlinear leastsquares solver,g modified to ensure successful convergence over the entire range of
CT,, r and x when solving for 7f.
-8-

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

Borehole

Correction

Simulation

Figure
5 shows a simulation of the response of the shortest spacing receiver coil configuration to a typical benchmark problem for the Gulf Coast, corrected for borehole
effect using the method previously described. The correction of this shortest spacing
array is the most difficult, because of the shallow depth of investigation.
The solid
square line in this figure gives the true formation conductivity,
while the dotted
square line gives the invasion conductivity.
The mud conductivity
is 2.5 S/m, and
the borehole radius is 4 in. The radius of invasion for the bed from 110 ft to 115 ft
is 20 in.; for the bed from 125 ft to 135 ft the radius of invasion is 40 in.; and for the
bed from 145 ft to 148.3 ft the radius of invasion is 20 in. The goal of the borehole
correction software is to generate a borehole corrected log equal to what the tool
would have measured in the same formation with the absence of a borehole. Ideally,
the no borehole curve and the borehole corrected curve on this plot would be
identical if the correction were perfect; indeed, they nearly are, with the exception
of the bed at 145 ft. In this thin bed, the highly conductive invasion causes an
error in the estimation of Zf, resulting in an imperfect borehole correction for this
shortest spacing array. The slight degradation in vertical resolution is due to the
Ff from the model failing to reflect the true af at bed boundaries in a thin bed a.s
well as it does in thicker beds.

Adaptive

Borehole

Correction

Applications

The solver used in the borehole corrections


is capable of optimizing more than
one variable and can adapt the borehole correction solution method to best reflect
the logging environment.
In most all cases, solving for borehole parameters
using
the adaptive borehole correction scheme rather than using inaccurate
input measurements is the preferred method of correcting array induction measurements
for
borehole effects.
In the adaptive

borehole

correction

mode,

the algorithm

used performs

a lesst-

squares optimization
to best match the data from the shorter-spacing
coil configIt is possible to solve for any combination
of the four borehole signal
urations.
Since the solution defines the
parameters
(urn, T-,X, or af) to match the data.
apparent borehole parameters
seen by the induction tool, this adaptive borehole
correction mode gives more reliable results when R,,/R,is large, when boreholes
are of irregular shape, when the mud resisistivity R, is unknown, or when either the
mud resistivity or mechanical caliper are inaccurate measurements
of the borehole
conditions.
Cave Effect

/ Irregular

Borehole

Geometry

Cave effect is seen when an induction tool encounters a washed out area, or cave,
in a borehole with a high formation resistivity to mud resistivity contrast.
Induction arrays normally have response peaks close to the tool that are very sensitive
to conductivity,
which cancel in smooth boreholes.
If a cave or other borehole
-9-

SPWLA 32nd Annual Logging Symposium,

June 16-19, 1991

when one of
irregularity
exists, large excursions can occur on the measurement
these sensitive areas is encountered.
The ability to use short array information
to
solve for effective borehole parameters in very difficult situations allows correction
to be made for any cave effect when the tool is run centered in the boreho1e.l This
also applies to irregular-shaped
borehole geometries that cannot be measured adequately by mechanical calipers. The adaptive correction scheme can then solve for
an effective borehole radius as seen by the induction tool. With external borehole
measurements
as inputs to the correction,
the short spacing-array
measurements
are usable when R,,/Rm < 10. The reliability of these measurements
is extended
to R,,/Rm
< 100 by using the adaptive borehole corrections scheme.
Computing

a Mud

The adaptive

correction

Resistivity

Calibration

has a unique feature

that allows for the optimization

of the

input mud conductivity prior to performing the borehole corrections.


In conductive
muds, or high formation resistivity to mud resistivity contrasts, this feature is very
important.
Borehole mud resistivity is generally a slowly varying value, increasing
very slightly as the temperature
in the wellbore decreases with decreasing depth.
Optimizing for the mud resistivity allows for an automatic self-consistent
calibration
of the external mud resistivity to the true mud resistivity seen by the multiarray
tool.
This optimization
is performed by computing a correction
to an external
mud conductivity measurement.
The external mud conductivity
is compared to the
computed mud conductivity
found by the adaptive borehole correction solver over
a short log interval.
The correction is determined by iterating over this interval
and isolating the minimum of a function to a fractional precision, using Brents
method,l
to compute a correction factor. By selecting a relatively short iteration
interval (approximately
30 to 50 ft) that is of fairly constant resistivity
and is
uninvaded by borehole fluids (i.e., a shale zone), the rapid process adds very little
to the total computation
time of the complete borehole correction.
If the entire
log is used as the iteration interval, there is little degradation
in the accuracy of
the correction; however, the computation
time for the complete borehole correction
rises rapidly as the total log interval increases.
The correction determined by the
described method is then applied to the external mud conductivity
measurement,,
and the borehole corrections are computed.
Mud Conductivitu

Sensors

Measuring the mud resistivity while logging with a downhole sensor can provide
a marked improvement
in accuracy over the mud pit or flowline mud resistivity
measurement.
In geothermal wells where a strong or variable temperature
gradient
may exist in the wellbore or when overpressured zones inject formation fluid into the
wellbore, a downhole measurement
is essential.
Mud resistivity sensors all show a
measurement error, however, and the accuracy of the measurement
can be improved
by using the adaptive borehole correction scheme of the AIT tool.

- 10 -

SPWLA32ndAnnualLoggingSymposium,
June16-19,1991

Optimum

Usage

of the Adaptive

Borehole

Correction

Scheme

To correct for both inaccuracy in the external mud conductivity


measurement
and
for cave effect or irregular borehole geometry, and to give the best possible borehole
signal correction to array induction measurements,
the adaptive borehole correction scheme is applied as follows. A suitable mud conductivity
correction iteration
interval is selected (for computational
time considerations),
and the borehole correction code computes and applies the appropriate correction to the external mud
conductivity
measurement.
The borehole corrections
are then performed solving
for the effective induction borehole radius in the adaptive mode. The result is a
self-consistent
borehole correction for an array induction tool that far surpasses the
capabilities of any previous induction borehole correction methods.
Summary

of Correction

Flow

At each depth level, the shallow arrays are used to determine the ?Yf value, and the
other borehole parameter values (gm, r, X) are either input from external measurements or computed along with the af in the optimization routine. The external grn
value may also be calibrated (as described above) prior to this borehole correction
Once these parameters
are determined,
each individual sensor measureprocess.
ment (in-phase and quadrature
signals for each array, at each frequency of that
array) is borehole corrected by using the polynomial representations
of the data
table for the particular sensor measurement
being corrected.
This is repeated for
each measurement recorded until all measurements
have been corrected for borehole
effect.
Field Examples
External measurements
to correct for borehole
external measurements

of the borehole are currently used in the oilfield industry


effect on induction measurements.
The accuracy of these
is decidedly insufficient in most cases to adequately correct

the complex measurements


of multiarray induction signals. This is particularly
true
for the shorter spacing (shallower reading) arrays. To use shallow array induction
measurements
for estimates of invasion profile and R,,, these measurements
must
be accurately corrected for borehole effects.
Figure 6 shows five of the many raw, shorter spacing array measurements
from the
AIT tool.
The increase in the measurement
number in the display corresponds
to an increase in the spacing of the coil configuration,
with CHANNEL #I having
the shortest spacing and CHANNEL #5 having the longest spacing.
The caliper
measurements
shown come from two consecutive
passes of a mechanical
caliper
run with the AIT tool. The difference in the two measurements
may result from
inaccuracies in the caliper measurement
as well as discrepancies in the caliper measurement of the irregular-shaped
borehole axes. This illustrates the inability of the
mechanical caliper to accurately describe the borehole geometry even in apparently
smooth borehole conditions.
The log section shown is impermeable,
and a gamma
- 11 -

SPWLA32ndAnnualLogging Symposium, June 16-19, 1991

ray measurement

is also displayed

for lithology

identification.

7 shows the array measurements


from Fig. 6 after correction for borehole
effect with the standard approach of using external measurements
to describe the
borehole.
The mud resisistivity
labeled MRES in track 1, was obtained from a
mud sample at surface temperature
(R, = 1.36), corrected to downhole formation
Figure

temperature
and used as the external mud resistivity
measurement.
The curve
labeled CALIPER #I was used for the external borehole diameter measurement.
If
the borehole corrections were perfect, the corrected conductivities
from all arrays
should read the same conductivity
value throughout
the impermeable
zone. The
contrast in this zone is approximately
20, and from the separation between
R,,/Rm
the various array measurements,
it can be seen that the external measurements
do
not provide information accurate enough to correct the raw AIT signals for borehole
Further processing of this data will lead to erroneous radial resistivity
effects.
estimates.
Figure 8 shows the results obtained when the adaptive borehole correction scheme
is applied as described in the previous section. The external mud resistivity is optimized, and the mud resistivity displayed in this figure is the resulting optimized mud
The borehole radius was determined
resistivity used for the borehole corrections.
in the optimization
done by the solver and is shown as a diameter by the curve
labeled IND. BOREHOLE
DIA." The excellent overlay of these corrected
shallow array
induction measurements
throughout the entire impermeable
section illustrates the
effectiveness and accuracy of this borehole correction method.
value of
example in Fig. 8 was repeated using an external mud resistivity
R, = 0.1 at surface temperature. Results identical to those in Fig. 8 were obtained,
which illustrates
the ability of the method to correct even the most inaccurate
external mud resistivity measurements.
The

AIT

Processed

Logs

The AIT processing

chain takes the borehole

corrected

array measurements,

applies

deconvolution
filters to each measurement,
then combines these using a nonlinear
processing algorithm to produce five log curves. These log curves have median
depths of investigation
of 10 in., 20 in., 30 in., 60 in. and 90 in. The vertical
resolution of each of the curves is closely matched, and three different resolution
widths are available.
Due to tradeoffs in improved vertical resolution of deeper
widths
arrays and cave effect on shallower arrays, l-ft, 2-ft and 4-ft resolution
The nomenclature
for the logs is AOlO, ... A090 for the l-ft
can be produced.
resolution set, ATlO, ... AT90 for the 2-ft resolution set, and AFlO . . . AF90 for
the 4-ft resolution set. The 1-ft A0 set is designed for the best vertical resolution in
smooth boreholes; the 2-ft AT set gives good vertical
to cave effect; and the 4-ft AF set has extremely
but a slightly wider vertical
invasion parameter estimates

resolution with less sensitivity


low sensitivity
to cave effect

resolution.
With these five log curves, Rt,R,,,and
can be made without any built-in assumptions about
- 12-

SPWLA 32ndAnnualLogging Symposium, June 16-19,1991

the invasion profile.


The radial resistivity description from an array induction tool can be no better than
the borehole corrected inputs, regardless of the quality of any postprocessing.
When
inadequate borehole corrected array measurements
are processed to obtain a radial
resistivity profile, both the near borehole and deeper investigation
measurements
can be in error.
Figure 9 shows another section of the well where a hydrocarbon
zone is invaded by
the fresh borehole fluid. After the raw measurements
were borehole corrected using
external inputs (as in Fig. 7), the five AIT logs described above were computed
with a 2-ft vertical resolution.
Displayed on the figure are the ATlO,AT20, AT30
and AZ60 logs (the AT90 log is omitted because it overlays the AT60).
A gamma
ray curve and the input caliper and mud resistivity are displayed in track 1. A
strange invasion profile is predicted by the logs throughout the interval. The AT10
log is most affected by the shallow array measurements,
and the effect of using
the external inputs for the borehole corrections can easily be seen. The AT10 log
reads below the other logs in the shale zone at the top of the interval; the log then
meanders through the pay zone, giving erroneous indications of permeability
even
in tight streaks, and cannot correctly evaluate the thin beds in the carbonate
section at the bottom of the interval.
Figure 10 shows the same processed log curves; however, the borehole corrections
were made using the adaptive borehole correction scheme.
The results show the
correct identification
of the permeable and nonpermeable
zones starting from the
shale at the top of the zone, going through the invaded hydrocarbon
zone and the
carbonate

section

below.

Conclusions
With

the introduction

of multiarray

induction

tools having new shallow-array

mea-

surements, the influence of the borehole on the measurement


is more readily observed. The importance of correcting all array measurements for borehole effect cannot be overstated.
Irregular-shaped
boreholes, caves and high formation-borehole
resistivity contrasts make the borehole correction task more difficult.
While traditional methods of borehole signal correction are inadequate,
the new adaptive
borehole correction scheme, designed specifically for multiarray induction tools, can
correct for borehole effects even in very difficult wellbore conditions.
The shallow
arrays provide the means for accurately correcting the many array measurements
and for extending the range of reliability of the shallow logs to R,,/R,
< 100
by using the adaptive borehole corrections.
Further AIT processing gives reliable
radial resistivity profiling and Rt, R,, and invasion parameter estimates.

- 13 -

SPWLA 32nd Annual Logging Symposium,

June 16-19, 1991

Nomenclature
AIT

Array Induction

AF

AIT 4-ft resolution

log set

A0

AIT l-ft resolution

log set

ilT

AIT 2-ft resolution

log set

ai

Polynomial

coefficients

bi

Polynomial

coefficients

Polynomial

coefficients

Polynomial

coefficients

di

Polynomial

coefficients

dij

Polynomial

coefficients

Minimized

e;

Polynomial

II-

Number

of terms in a summation

Number

of terms in a summation

Number

of terms in the full fit summation

Number

of terms in the full fit summation

Pi

Polynomial

term dependent

on (a,,

Qi

Polynomial

term dependent

on (r, IC)

Borehole

Rrn

Resistivity

of the borehole

Rt

Resistivity

of the virgin zone (ohm-m)

R 20

Resistivity

of the flushed zone (ohm-m)

Ti

Chebyshev

polynomial

V corn

Corrected

Vlnea3

Complex

input measurements

Imode

Borehole

signal determined

Vhomo

Signal in a homogeneous

ci

Imager

quantity

Tool

in borehole

algorithm

penalty

function

coefficients

af )

radius (in.)
fluid (ohm-m)

input measurements

Distance
from the
(Tool standoff - in.)

by polynomial

formation

outside

surface

Value of a polynomial

expansion

&

Value of a polynomial

expression

Value of the full fit polynomial

- 14 -

with no borehole
of the

expansion

tool

to

the

borehole

wall

SPWLA 32nd Annual Logging Symposium,

cj

Estimated

variance

in the measured

Two-dimensional

;i

Vector

containing

the data values

Vector

containing

coefficients

grn

Conductivity

of the borehole

af

Conductivity

of the formation

?f

Average

matrix

conductivity

containing

June 16-19, 1991

signal of the jlh coil


the terms of the polynomial

determined

by minimization

fluid (mS/m)
(mS/m)

of the formation

(mS/m)

Acknowledgements
The authors would like to acknowledge John Miles for his important contributions
to
the original modeling and algorithm development as well as the processing algorithm
software; Yi-Ming Wang and Nina Zumel for their contributions
to the further development of the borehole correction algorithm; Elton Head for analyzing field test
results; Richard Chandler and John Hunka for managerial
support of this project;
and Tom Barber and Bob Freedman for helpful comments on the manuscript.
The
authors also thank the oil company that gave permission to publish the field examples.
References
System For Deep Formation
1. Hunka, J., et al.: A New Resistivity Measurement
Imaging And High-Resolution
Formation Evaluation,
paper SPE 20559 presented
at the 1990 SPE

Annual Technical

2. Miles, J. W. and Minerbo,


Schlumberger
Engineering

Conference,

New Orleans,

G. N.: The AIT Borehole


Report, January 1988.

September.

Correction

Implementation,

in Three Dimen3. Minerbo, G. N.: Numerical Solution of Maxwells Equations


presented
at
the
Fourth
Biennial
IEEE
Conference on
sions at Low Frequencies,
Electromagnetic
Field Computation,
Toronto, Canada, October 22-24, 1990, to be
published

in IEEE

4. Gianzero, S. C.:
Induction-Logging
October

Transactions

on Magnetics,

September

1991.

Effect Of Sonde Eccentricity


On Responses
Of Conventional
Tools, IEEE Trans. on Geoscience and Electronics,
GE-16(4),

1978.

To Induction Logging And Application


5. Doll, H. G.: Introduction
Wells Drilled With Oil Base Mud, J. of Pet., June 1949.
6. Lawson,
(1974).

C. L. and Hanson,

R. J.:

Solving Least

- 15 -

Sauares

Problems,

To Logging

Prentice-Hall

Of

I
SPWLA 32nd Annual Logging Symposium,

June 16-19, 1991


I

M., et al.: National Energy Softward Compilation Of Program Abstracts, Technical Report ANL-7411(Rev.
Suppl.
1), Argonne National Laboratory, Argonne, Illinois, September 1982. SLATEC Mathematical
Subroutine
Library, NESC Abstract 820.

7. Birgersson,

8. Abramowitz,
Publications,

M. and Stegun, I. A.: Handbook


Inc., New York (1970).

9. More, Jorge J.: The Levenberg-Marquardt


ory, Numerical

Analysis,

Springer-Verlag

(1977).

G. A. Watson,

About

Gregory

Functions,

Dover

Algorithm, Implementation And Thecd., Lecture Notes in Mathematics 630,

10. Brent, Richard P.: Ahzorithms For Minimization


Hall, Englewood

of Mathematical

Without

Derivatives,

Prentice-

Cliffs, N. J.: (1973).

The Authors

P. Grove

is a senior project

engineer in Electrical

Interpretation

with

Schlumberger in Houston.
Since 1976, he has held various field assignments and
positions involving computerized log analyses and interpret ation development.
He
holds a BS degree in electrical engineering from California State University at
Northridge, and is a member of SPWLA and SPE.
G. N. Minerbo

received a B. S. in Physics at New York Polytechnic

University and

a Ph. D. in Theoretical Physics at Cambridge University, England. His professional


employment includes Brookhaven National Laboratory, A.E. R.E Harwell, England,
Adelphi University. He served as a consultant for Vitro Corporation of America
and the Ballistic Research Laboratory, Aberdeen Proving Ground. He was a Staff
Physicist for ten years at Los Alamos National Laboratory. He joined Schlumberger
in 1981, and is now a Senior Research Scientist at Schlumberger-Doll
Research. He
is married to Grace Moffat, M. D.; they have two children.

-16-

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

o 6 Radius Exact)
A 8 Radius (Exact)
x 10 Radius (Exact)

6 Radius (Approx)
8 Radius (Approx)
....._.._._.
_ _ _ 10 Radius (Approx)

Figure 1:Compalison of the o, polynomial fit to the data for a medium spacing receiver
coil configuration, with a 2.5 in. stamiqff in a 0.1 Slmformation

- l?-

SPWLA 32nd Annual Logging Symposium,

June 16-19, 1991

Figure 2: Comparison of the o,,, and uf polynomial fit to the data for a medium spacing coil
configuration and r = 4 inches, x = 0.5 inches, showing the smootness of the fit.

Figure 3: Comparison of the r and x polynomial fit to the data for a medium spacing coil
configuration and a, = 10 Slm, of= 0.01 S/m, showing the smoothness of the fit.

- 18 -

SPWLA 32nd Annual Logging Symposium,

105

June 16-19, 1991

110

115

120

125

130

135

140

145

150

Depth (Ft.)

_
Uncorrected
.. . .. No Borehole
Borehole Corrected
--Figure 5: Plot of the borehole correction for the shortest spacing receiver coil configuration.
The borehole radius is 4 inches, and the mud conductivity is 25 Slm

- 20 -

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

1
6

10

100
I

I2

14

1.0

16

J I11111

I40

I Illllll

10.0

100.0

I I I ,I111

1000.0

I I I II111

l70

230

260

290 L
CAUPER #l
(3upER

#2

CHANNEL #i
_ _ _ _____

__

GR

aANNa

#2

CHANNEL #3

Figure 6: In-phase
receiver

____________

(3jANNEL#4

me-_-___-_

mANNn#s

__----_____

------

array measurements
coil configurations.

- 21 -

from five different

shorter

spacing

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

L--L-i

100
I

10

12

16

1 I40

230

260

290

CALJpER#2

_ _________
GR
),jRs

Figure

CHANNEL #4 __-_-_____

CHANNEL #l

CALIFER #I
_

CHANNEL#z

____________

CHANNEL #3

------

CHANNEL #3 __-

________

______

7: Borehole corrected
inputs

to describe

array

the

measurements

wellbore.

- 22 -

using

inaccurate

external

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

d
OJ

lo

12

I4

16

260

IND. BOREHOLE

DIA. - GR
MRs

wANNFJ#4

CHANNEL #I

--mm_-

Figure 8: Borehole

..

correctaon

mANNEL#2

____________

CHANNEL #3

------

corrected

array

measurements

scheme.

- 23 -

(&jANNn

using

__-______-

#5 _____-_--_-

the adaptive

borehole

SPWLA 32nd Annual Logging Symposium, June 16-19, 1991

L---L-d

0
I
6

10

100
I

I2

14

1.0

16
260

10.0

lllll

1111111

100.0
I

I1111

1000.0
I

1111111

360
WpER

AT20 ________

GR
),&Es

Figure

AT60 ._____

AT10

mm__---

______

-_-_-_-_

__________

AT30 ____-----

9: AIT logs using inaccurate

external

corrections.

- 24 -

measurements

for borehole

SPWLA 32nd Annual Logging Symposium,

June 16-19, 1991

l--d

100
I

10

I2

14

16
260

260

300

320

340

360
IND. BOREHOLE

DIA. - -

AT20 __________________

GR
MRm

AT60.__-_--__-__-_--_--_

AT10

______

Figure 10: AIT

logs using

ATSO--_--_-_-_---

the adaptive

- 25 -

borehole

correction

scheme.

Potrebbero piacerti anche