 |
Description  |
|
|
BACKGROUND OF THE INVENTION
The invention is in the field of well logging and relates specifically to
correlating the relative depth levels of different sets of well logging
data derived from separate passes of investigating devices through
boreholes or from the same pass of a device.
In well logging, measurements of various formation characteristics are
taken by an investigating device which is lowered into a borehole at the
end of a supporting cable extending from the surface of the Earth.
Measurements are taken at specified intervals as the device is drawn up.
Typically, the measurements are intended to provide indications of oil or
gas bearing strata, and may be measurements such as resistivity,
inductance, and the like.
It has recently become common to combine, through sophisticated data
processing techniques, different sets of measurements taken during passes
of investigating devices through the same borehole or through different
boreholes in order to produce computed measurements of various
characteristics of the investigated Earth formations. When combining sets
of measurements in this manner, it is important that they be accurately
correlated in depth, e.g., when combining two different measurements each
indicated as taken at depth of 2000 feet, it is important to be certain
that each of these measurements has been in fact taken at depth of 2000
feet.
The depth at which a measurement is taken is commonly indicated by a
sheave-wheel device which is located at the surface of the Earth and
provides a measurement of the length of cable that passes over the
sheave-wheel and suspends the investigating device in the borehole.
However, this device does not generally take into account cable stretch,
although the cable may stretch differently for different passes of the
investigating device and even for different portions of the same borehole,
because of factors such as changes in the borehole size, possible size and
shape differences between different measuring devices, and the like.
Because of the desirability of accurate depth correlation of different
logs, efforts have been made in the past to accurately measure the actual
depth of the investigating device. One prior art technique involves
accurate instantaneous cable length measurements, and is disclosed in U.S.
Pat. No. 3,497,958 granted to L. H. Gollwitzer on Mar. 3, 1970. The
Gollwitzer system measures the tension of the cable at the surface of the
Earth and at the investigating device and corrects the cable length
measurements indicated by the sheave-wheel device for changes in stretch
as reflected by the tension measurements. The Gollwitzer system also
corrects for sheave-wheel calibration errors and for temperature effects
on cable stretch. While the Gollwitzer system has been found to provide
extremely accurate depth measurements in most cases, there are factors
which can cause slight errors. For the usual situation, these errors are
insignificant and can be ignored. However, in certain of the recently
utilized highly sophisticated data processing techniques applied to
combining logs, even certain slight errors can become significant.
Another prior art approach to taking cable stretch into account involves
the use of data processing techniques to determine the depth match between
two or more logs derived from separate passes of investigating devices
through a borehole or even from the same pass of the device. This approach
is disclosed in U.S. patent application of David H. Tinch, Bruce N.
Carpenter and Elie S. Eliahou filed on Sept. 9, 1970 and assigned Ser. No.
70,709. It involves determining first assumption depth displacement values
for a number of log depth levels through use of a suitable correlation
function. The first assumption depth displacement values for a selected
number of depth levels are then analyzed to determine more accurate depth
displacement values. This approach works well, but the need still remains
for techniques and means for carrying out depth correlation that start
with conventional data, such as correlograms obtained from applying a
correlation algorithm to a pair of logs, reduce the amount of data that
must be processed, and reduce processing time and processing complexity.
SUMMARY OF THE INVENTION
This invention is in the field of well logging and relates specifically to
correlating the relative depth levels of logs. The logs may be obtained
from different passes of investigating devices through the same or through
different boreholes, or may be obtained from the same pass of single
investigating device.
An object of the invention is to utilize conventional data, such as
correlograms obtained from applying a correlation function to a pair of
logs, to reduce the amount of data that must be processed, and to process
the remaining data in an efficient and accurate manner in order to arrive
at results indicating exactly how the two logs of interest must be shifted
with respect to each other for an optimized depth correlation between
them.
In a specific embodiment of the invention, a plurality of normalized and
digitized correlograms are examined, and selected elimination parameters
are applied to them in order to reduce the amount of data which must be
processed further. For example, correlation coefficients which are below a
certain cutoff value, and hence less reliable, may be marked for
elimination, correlation coefficients which do not have a sufficient
number of defined neighbors may also be marked for elimination, and other
selected elimination parameters may be applied. A group of correlograms
which satisfy the elimination parameters are stored in a cluster array,
and additional elimination parameters are applied to the data stored in
the cluster array to dispose of the less reliable portion of it. If the
remaining data in the cluster array satisfies certain other parameters, it
is examined to detect changes in shift between successive groups of
correlograms where the correlograms of each group have the same shift
value and each group consists of a sufficient number of correlograms. Such
changes in shift are used to determine exactly how the logs that are under
consideration should be shifted with respect to each other to optimize the
correlation between them.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows an investigating device in a borehole along with apparatus at
the surface of the Earth for controlling the investigating device, for
recording the measurements by it and for processing these measurements.
FIG. 2 shows an example of logs derived from a pass (or passes) of an
investigating device of the type shown in FIG. 1 through a borehole (or
boreholes).
FIG. 3 shows segments of two logs and illustrates the application of a
correlation function thereto.
FIGS. 4, 5 and 6 show exemplary normalized correlograms in analog form.
FIG. 7 shows a plurality of exemplary digitized correlograms.
FIG. 8 shows a portion of FIG. 7.
FIG. 9 illustrates the major steps of the invented method.
FIGS. 10, 11, and 12 illustrate the detailed steps of an embodiment of the
invented method.
DETAILED DESCRIPTION
Referring to FIG. 1, an investigation device 10 is suspended within a
borehole 11 on a cable 13 to investigate subsurface Earth formations 12.
The cable 13 passes over a sheave-wheel 14 and is secured to a drum and
winch mechanism 15 which includes a suitable brush and slip-ring
arrangement 16 for providing electrical connection between the conductors
forming the cable 13 and a control panel 17. The control panel 17 operates
to supply power and control signals to the investigating device 10 and
includes suitable electronic circuitry for receiving well logging
measurements from the investigating device 10 and for preparing them for
recording by a digital tape recorder 18. The tape recorder 18 converts
analog signals from the investigating device 10 into digital signals and
records the digital signals. It is stepped as a function of depth by a
driving wheel 19 which engages the cable 13 and a mechanical linkage 20. A
detailed description of a recording system suitable for recording
measurements from the investigating device 10 in the embodiment shown in
FIG. 1 is disclosed in detail in U.S. Pat. No. 3,457,544 issued to G. K.
Miller et al. on July 22, 1969, and entitled "Method and Apparatus for
Recording Well Logging Data".
The digital tape recorder may record the data from a single pass of the
investigating device 10 through the borehole 11 or the data from two or
more passes of the investigating device 10 through the borehole 11. Two or
more logs may be obtained from a single pass of the investigating device
10 in certain cases. The digital signals recorded by the tape recorder 18
are then either transmitted or carried to a digital computer 23 for
processing in accordance with the subject invention, or to a suitable
special purpose device (not shown) embodying the subject invention. The
output of the digital computer 23, or the output of the special purpose
device is recorded on a tape recorder 24 for later use, or is applied to a
transmission channel 25 for transmission to another utilization device
such as a digital computer or a special purpose device. The tape recorder
24 may be of the same type as the tape recorder 18.
The investigating device 10 has a reference point 22 which forms the center
or recording point of the device 10. The depth of the investigating device
10 is registered by a depth recorder 21 which is driven via a mechanical
linkage 20a by the driving wheel 19 engaging the cable 13. The depth level
registered by the depth recorder 21 at each measurement by the
investigating device 10 may be recorded in the digital recorder 18
together with the corresponding measurement, for example by recording the
output of a suitable counter driven by the linkage 20 which steps the
recorder 18.
Because of the long and somewhat elastic cable 13, the investigating device
10 is subject to displacement due to cable stretch, with the result that
the true depth level of some measurements may be different from the depth
level registered by the depth recorder 21. (The depth level registered by
the depth recorder 21, which is identical with that registered by the tape
recorder 18 is designated "Z".) When attempting to depth match a log
produced by the investigating device 10 with the log produced from another
investigating device 10 passed through the same borehole 11 at a different
time, there is a possibility that the measurements recorded by the two
investigating devices at the same depth level will not be referenced to
the same true depth level. The same may occur when two logs are obtained
from two different boreholes 11, and when two or more logs are obtained
from the same pass of an investigating device 10 through a borehole 11.
As an illustration, logs A and B of FIG. 2 are of the same or similar
measurements and are made during different passes of the investigating
device 10 through the borehole 11. By observation, it can be seen that the
two logs have similar, though depth-shifted details. Considering the log A
as the "base log", it can be seen that at the depth level Z1, log B is
depth shifted by an amount X from log A. The same is true at depth level
Z2. Then, at depth level Z3, the two logs are depth matched, i.e., the
depth shift between them is zero. Then, at the depth level Z4, the depth
shift becomes 3X. (X is in units of length.) When attempting to combine
measurements from the logs A and B which are supposed to be taken at the
same depth Z, the results may be in error because of the poor depth
matching or depth correlation between the logs A and B.
One technique for obtaining a measure of the correlation between two logs
is the use of a correlation function whose value indicates the degree of
depth correlation. One technique of this type is disclosed in the U.S.
patent application of John P. Timmons and James J. Maricelli filed on July
31, 1972, and assigned Ser. No. 276,348. This technique involves comparing
a segment of a compare log with a segment of a base log by means of a
normalized correlation function, and providing a correlogram which is a
plot of the value of the correlation function versus displacement between
the two segments. It should be clear that the labels "base" and "compare"
may be interchanged.
As an example which is pertinent to the subject invention, assume that a 20
foot segment of the compare log is to be correlated with a 30 foot segment
of the base log (see FIG. 3). The start depths, i.e., the indicated
centers of the two log segments are at the same 60 foot depth. Assume that
a measurement is taken at each 6 inch interval to obtain the two logs. In
applying the normalized correlation function, a correlogram is derived by,
in effect, aligning the top ends of the compare log segment and of the
base log segment and correlating, then moving the compare log segment 6
inches down with respect to the base log segment and correlating again,
and repeating this procedure until the bottom ends of the two log segments
are aligned. This results in twenty-one correlation coefficients, i.e.,
twenty-one values of the correlation function. Successive segments of the
compare log may be similarly correlated with corresponding successive
segments of the base log to obtain a sequence of correlograms. Each
correlogram is for a defined starting depth, i.e., for a defined indicated
depth of the two segments which are correlated to derive the correlogram.
A correlogram may be plotted on a coordinate system of correlation function
value versus shift (displacement) between the segments. The normalization
may be such that the correlation function value ranges between +1 and -1,
with +1 indicating perfect correlation and -1 indicating anticorrelation.
A correlogram plot (in analog form) is shown in FIG. 4. The shift value of
the peak of the curve in FIG. 4 indicates how much and in which direction
the two segments (whose starting depths are the same) should be shifted
with respect to each other in order to depth-correlate them.
While the correlogram in FIG. 4 indicates unambiguously how the two log
segments from which it is derived must be shifted with respect to each
other, there may be correlograms, such as the one shown in FIG. 5, which
have two or more peaks. In such cases, it is not clear which one of these
peaks, if any, should be taken as the desired shift between the two
correlated segments. Additionally, the peaks of successive correlograms
may be such that no meaningful trends of change of shift are indicated.
Ambiguity in the peaks of correlograms and other ambiguities may be
resolved to a certain extent by looking for trends in a plurality of
correlograms. Referring to FIG. 6, which shows a plurality of correlograms
arranged along a third dimension Z, which is the indicated starting depth
of the correlograms, it is seen that there are certain trends which
persist and certain trends which are short-lived. Depending on the quality
of the correlograms, and on a number of other factors, it may be
reasonable to assume that trends which do not persist tend to reflect
transient events that could be disregarded, while trends which persist
could be taken as tending to be reliable indications of how the logs
resulting in these correlograms could be shifted with respect to each
other.
In fact, a highly skilled interpreter can examine by observation a group of
correlograms (of the type shown in FIG. 6, but comprising many more
correlograms), and can decide on the basis of long experience and perhaps
on the basis of some other knowledge regarding the starting logs, if there
are certain trends which appear reasonably reliable, and what these trends
may be said to mean in terms of desirable shifts between the starting
logs. Since there are hundreds or even thousands of correlograms between
any two typical logs, and since usually a number of logs are to be
correlated, it can be appreciated that it is difficult and not very
reliable to look for trends by observation and to interpret the meaning of
trends by observation. Additionally, it can be appreciated that
interpretation of correlograms by observation is time-consuming, that it
requires highly experienced interpreters, and that it is highly
subjective.
The subject invention therefore provides a method of utilizing a digital
computer to process correlograms of the general type shown in FIG. 6 in a
novel manner and by novel techniques in order to first reduce the amount
of data which must be processed by eliminating data determined to be less
reliable and to leave only reasonably valid data, and to then process this
valid data to find and compute desirable shifts between the starting logs
and the depth at which these shifts should be made.
The starting data for the subject invention are correlograms which are of
the type shown in FIG. 6, but which are digitized. For example, when each
of the correlograms that appear as individual curves in FIG. 6 is
digitized on a scale of 10, it is converted thereby to a row of decimal
numbers each ranging in value from 0 to 9. In the correlation example
discussed in connection with FIG. 3, this would result in a correlogram
consisting of 21 correlation coefficients. It should be understood, of
course, that this is an arbitrary example, and that correlograms
consisting of other suitable numbers of correlation coefficients, and
scaled differently may be utilized in accordance with the subject
invention.
The starting data for the subject invention is thus a two-dimensional array
of digitized correlograms. The correlograms may be obtained with the use
of the invention disclosed in the copending Timmons and Maricelli patent
application cited above (See FIG. 6 thereof and the derived C.sub.K
values).
An exemplary portion of a correlogram array of this type is illustrated in
FIG. 7 of this specification, where the vertical dimension is the
correlogram depth, i.e., the start depth of the two log segments which are
correlated to produce the corresponding row of 21 correlation
coefficients, and the horizontal dimension is the shift (displacement with
respect to each other) of the central points (start depths) of the two
segments at which the shown correlation coefficient value is obtained. The
correlation coefficient is the decimal value given at the intersection of
a depth value and a shift value in the array of FIG. 7.
It is recognized in the subject invention that the amount of data, i.e.,
the number of correlation coefficients, which should be processed may be
substantially reduced by defining some elimination parameters. For
example, a cut-off parameter may be defined to eliminate as "invalid"
correlation coefficients whose value is 5 or below 5. In the example of
FIG. 7, the application of a cut-off parameter of this type would leave
only the three groups of correlation coefficients identified by the
reference numerals 30, 32 and 34. Each group consists only of "valid"
correlation coefficients.
An additional reduction in the amount of data which must be processed, and
an additional decrease in ambiguity of trends, may be obtained according
to the invention by defining and applying one or more "neighbor"
parameters. In the example of FIG. 7, one neighbor parameter may be that
each of the correlation coefficients within the groups 30, 32 and 34 must
have at least two valid neighbors on the left and two valid neighbors on
the right. This type of a neighbor parameter would eliminate group 30 and
would leave only groups 32 and 34 for further processing.
The correlation coefficients of the group 32 of FIG. 7 are shown alone in
FIG. 8. It can be appreciated that there are certain apparent trends
within the group 32. For example, the correlograms for the 50 feet through
70 foot depths each have the maximum correlation coefficient at the -1
shift position, the correlograms for the 75 foot through 95 foot depth
each have the maximum correlation coefficient at the +1 shift position,
and the correlograms for the 100 and 105 foot depths each have their
maximum coefficients at the -1 shift position. These trends are tested by
one more parameter in accordance with the subject invention. This
parameter reflects possible errors introduced by end effects, and is
applied by eliminating a selected number of correlograms from the
beginning depth and the end depth of a group of the type shown in FIG. 8.
The invention may be described in very general terms by reference to FIG. 9
which shows its major functional steps. Examples of the steps are given by
reference to FIGS. 7 and 8 discussed above.
The starting data for the functional steps of FIG. 9 is a two dimensional
array of digitized correlograms of the type shown in FIG. 7. A single
correlogram is read at step 36 of FIG. 9, for example, the correlogram for
the 50 foot depth level of FIG. 7.
At step 38 of FIG. 9, a cut-off prameter and a neighbor parameter are
applied to the correlogram which was read at step 36.
The cut-off parameter is designed to eliminate correlation coefficients
which are considered insufficiently reliable to warrant further
processing. For example, of the cut-off parameter is 5, each correlation
coefficient of the correlogram read in at step 36 is compared with 5, and
only the correlation coefficients which are 6 or above are retained.
Additionally, if the correlogram contains no value which is above the
cut-off parameter, it is suitably flagged. The correlation coefficients
which are not eliminated by the application of the cut-off parameter are
called "valid" correlation coefficients in this specification. In the
example of the top correlogram of FIG. 7, the correlation coefficients
which are outside the group 32 are eliminated and the correlation
coefficients which are within the group 32 are valid correlation
coefficients.
The neighbor parameter may be the requirement that the maximum correlation
coefficient of a correlogram be flanked by two valid correlation
coefficients on the right-hand side and by two valid correlation
coefficients on the left-hand side. In the example of FIG. 7, the
correlogram at the 50 foot depth has a maximum correlation coefficient
which is 9, and it is flanked by two eights on the right-hand side and by
two sevens on the left-hand side. Therefore, this correlogram satisfies
the neighbor parameter. The correlogram for the 115 foot depth of FIG. 7
has a maximum correlation coefficient which is 7, and it has two valid
neighbors on the right-hand side, but only one valid neighbor on the
left-hand side. Therefore it does not satisfy the neighbor parameter.
It should be clear that the cut-off parameter can be any other suitable
value, and that the neighbor parameter may be any other suitable value, or
may be different in kind. For example, the neighbor parameter may
alternately be the requirement that each correlogram have a valid maximum
correlation coefficient which is flanked on the left and on the right, and
above and below by a suitable defined number of valid correlation
coefficients. Still alternately, the neighbor parameter may be
N-dimensional, and the dimensions may be attributes which may include the
number of valid neighbors on given sides of the maximum correlation
coefficient of a correlogram, but may also include other attributes of
correlograms.
At step 40 of FIG. 9, a sequence of correlograms which satisfy both the
cut-off parameter and the neighbor parameter is stored in a cluster array.
The clusterr array may be a two-dimensional array of a sufficient size to
store up to a selected number of correlograms. In the example of FIG. 7,
the correolograms for the 50 through 110 foot depth includes the group 32
which satisfies both the cut-off parameter and the neighbor parameter, and
would be stored in the cluster array, while the correlogram for the 115
foot depth does not satisfy the neighbor parameter and would not be stored
in the cluster array.
At step 42, an end effect parameter is applied to the group of correlograms
in the cluster array. this parameter may be, for example, a requirement
that the first and the last correlograms stored in the cluster array be
eliminated, but may be a different parameter of this type. For example,
referring to FIG. 7, the correlogram for the 50 foot depth and the
correlogram for the 110 foot depth may be eliminated in order to leave in
the cluster array only the more reliable central group of correlograms.
At step 44, a cluster array size parameter is applied to the correlograms
which are now stored in the cluster array. This parameter reflects the
recognition that trends in correlograms are valid only when a certain
minimum number of correlograms reflect these trends. Therefore, a cluster
array size parameter may be the requirement that the cluster array
contain, after the operation of step 42, a certain minimum number of
correlograms, such as three correlograms.
If the cluster array does not satisfy the test of step 44, new correlograms
are read in at step 36 for a new cycle of storing a sequence of
correlograms which satisfy the cut-off parameter and neighbor parameter in
the cluster array.
If the test at step 44 is satisfied, the correlograms which are retained in
the cluster array are examined for valid changes of shift. A valid change
of shift may be defined as a shift between two apparently reliable trends,
such as a shift between a set of at least three correlograms which have
their maximum correlation coefficient at the same position and another
adjacent set of at least three correlograms which have their maximum
correlation coefficient at the same position, with the positions of the
maximum correlation coefficients as between the two sets being different.
In the example of FIG. 8, the shift between the sets of correlograms
containing the correlation coefficient maxima identified by the reference
numerals 31 and 33 would be valid, while the shift between the sets
identified by the reference numerals 33 and 35 would not be valid because
the set 35 consists of only two correlograms.
One embodiment of the invention is a method of operating a digital
computer, such as an IBM System 360/65. This embodiment of the invention
can be expressed in a standard high level programming language, such as
Fortran or PL/1, which can be translated by conventional compilers into a
machine language form executable on the computer. Applicants consider it,
however, more informative to describe this embodiment of the invention in
terms of a detailed flowchart, where each step of the flowchart is
directly expressible in a single high level language statement or at most
in a very low number of such statements. A detailed flowchart, therefore,
is shown in FIGS. 10, 11 and 12, which can be put together as indicated to
form a single flowchart.
It is noted that the procedure of FIGS. 10, 11 and 12 would normally be a
subroutine of a larger software system concerned with processing well
logging data, and would be entered and executed in a conventional way.
However, for the sake of simplicity, the procedure of FIGS. 10, 11 and 12
is described below as a free-standing system which starts with certain
input data and ends when certain results are generated. The conventional
steps which may be involved in running a software system (or a subroutine)
on a computer, such as partitioning a memory portion, providing disc and
tape space, etc. are omitted for simplicity, and only the actual invention
is disclosed in detail.
Referring to FIG. 10, entry is to step 48 at which certain parameters and
constants are defined. Thus, a cut-off parameter is defined as the decimal
number 5, a neighbor parameter is defined as the decimal number 2, an
index I is set equal to 1, a correlogram depth index D is set equal to 1,
and a constant D.sub.max is set equal to a selected number. The index I is
used to point to one of the twenty-one correlation coefficients forming a
single correlogram, the depth index D identifies a specific correlogram of
a set of N correlograms, and the constant D.sub.max is set to the number
of correlograms which are to be processed in accordance with the
invention. There is a one-to-one correspondence between each index D and a
selected depth Z.
At step 50 of FIG. 10, the correlogram which is identified by the current
value of the depth index D is read into a suitable register of the
computer carrying out the invented method. It is assumed that a set of N
correlograms of the type shown in FIG. 7 are stored on a medium such as
magnetic tape, and that the tape drive is advanced at step 50 to read out
one correlogram (whose depth index number is D) and to store the value
read off the tape into a register of the computer.
At step 51 of FIG. 10, a variable MAXVAL is set to 0. The next step of the
sequence of FIG. 10 is to find the value and the position within the
correlogram of the maximum correlation coefficient of that correlogram. To
do this, at step 52 the correlation coefficient CC(I) which is identified
by the current value of the index I is compared in value with the variable
MAXVAL. For the first run through step 52, MAXVAL is 0; since the CC(I)
value can be only a 0 or a positive number, the answer at step 56 at this
time must be "no". Then the pointer I is incremented at step 54, and the
next CC(I) is compared with 0, etc. When the CC(I) that is compared with
the initial 0 value of MAXVAL at step 56 is 1 or more, the answer at step
52 is "yes".
Then, at step 56, the current CC(I) becomes the current MAXVAL, a variable
MAXPOS is set to the current value of I, and a return is made to step 54.
The comparison at step 52 is now between the current CC(I) and the value
to which MAXVAL was set at step 56.
The runs through steps 52 and 54 (and possibly through step 56) are stopped
when a test at step 55 indicates that the last CC(I) of a correlogram has
been processed in the indicated manner. When this is the case, the current
MAXVAL is the value of the maximum correlation coefficient of the
correlogram, and the MAXPOS is set to its pointer I.
Referring to FIG. 7 for an example, the run through steps 52, 54, 55 and 56
for the topmost correlogram of FIG. 7 would be comparing the leftmost
correlation coefficient with 0, then moving one correlation coefficient to
the right, repeating the comparison, and repeating until the correlation
coefficient 4 is encountered. Then 4 is compared with 0, the answer at
step 52 is "yes", and MAXVAL =4 and MAXPOS =pointer I of 4. Then 7 is
compared with 4 at step 52, MAXVAL is now set to 7, and MAXPOS is changed
to the pointer of the leftmost 7, etc., until 9 is the MAXVAL and the
pointer I of 9 is the MAXPOS.
At step 58 of FIG. 10, the value of the maximum correlation coefficient
found previously is compared with the cut-off parameter, and if it does
not exceed the cut-off parameter, the correlogram is flagged at step 60.
In the example of FIG. 7, the correlogram whose depth is 120 feet has a
maximum of 5; this correlogram would not satisfy the test at step 58 of
FIG. 10, and it would be flagged at step 60.
If the answer at step 58 is "yes.revreaction., i.e., if the correlogram
which is currently considered has a maximum correlation coefficient which
exceeds the cutoff parameter, tests are made at steps 62 and 64 to see if
the correlogram satisfies the neighbor parameter. At step 62 the
correlation coefficient which is two correlation coefficient positions to
the left of the maximum correlation coefficient is tested to see if it
exceeds the cut-off parameter. If the answer is "no", the correlogram is
flagged at step 60; if the answer is "yes", a test is made at step 64 to
see if the correlation coefficient which is two positions to the right of
the maximum satisfies the cut-off parameter. If the answer is "no", the
correlogram is again flagged at step 60. Referring to FIG. 7 for an
example, if the first correlogram, at the top of FIG. 7, is examined at
steps 62 and 64, the test of step 62 would be if 7 exceeds 5 and the test
at step 64 would be if 8 exceeds 5. For the correlogram at the 115 feet
depth of FIG. 7, the test at step 62 would be if 5 exceeds 5, with the
result that the correlogram would be flagged at step 60 and the test at 64
would not be performed.
In step 66 of FIG. 10, the correlogram which was stored in the register at
step 50, and its MAXPOS and MAXVAL determined at step 56 and the flag from
60, if any, are stored in a correlation array, at a position identified by
the index D of that correlogram. The correlation array used in step 66 may
be an array in the core memory of the computer, or it may be any other
suitable storage medium such as a tape or a disc.
At step 68 a test is made to determine if the D index of the current
correlogram is equal to the D index of the last correlogram which is to be
processed (D.sub.max). If the answer is "no", a return is made to step 50
to read into the register (destructively) the next correlogram from the
tape. If the answer at step 68 is "yes", the sequence continues to the
flowchart in FIG. 11.
After the sequence of FIG. 10, each of a selected number of correlograms
has been examined to determine the position and the value of the maximum
correlation coefficient in it, whether the maximum correlation coefficient
exceeds the set cut-off value, and whether the maximum correlation
coefficient has at least the set number of valid neighbors on the left and
on the right-hand side. The information determined in this sequence has
been stored, together with the correlogram, in a correlogram array, at a
location identified by the index number D for the correlogram. The
procedure of FIG. 11 then processes the information which has been stored
in the correlogram array as a result of the procedure of FIG. 10.
In FIG. 11, certain parameters are set at step 72. The parameter LEVLS is
set equal to 1, the parameter LEVHIS is set equal to 3, the correlogram
index D is set at 1, and the index D.sub.max of the last correlogram which
is to be processed is set equal to a selected number. The parameter LEVLS
is the end effects parameter discussed earlier. Referring back to FIGS. 7
and 8, the procedure of FIG. 11 would read into a cluster array the
correlation coefficients of group 32, which are shown in FIG. 8, and the
parameter LEVLS specifies how many levels from the beginning and end of
the cluster array contents should be dropped. In the example of FIG. 8,
and assuming that LEVLS equals 1, the lines for the 50' and the 110' depth
of FIG. 8 would be eliminated in the course of the procedure of FIG. 11.
The parameter LEVHIS indicates the minimum number of levels that should be
in the cluster array to give a meaningful result. If LEVHIS is 3, a
cluster array must contain at least five (that is, 3+1+1) levels before
the parameter LEVLS has been applied. In the example of FIG. 8, the levels
from 55' through 105' inclusive remain after the parameter LEVLS is
applied, so that the cluster of FIG. 8 would satisfy the parameter LEVHIS.
It should be clear that other suitable values may be selected for the
parameters LEVLS and LEVHIS.
At step 74 of FIG. 11, a cluster array of a suitable size is cleared, and
an index LEVDX is set equal to 1. The index LEVDX points to the first
storage location in the cluster array cleared at step 74. Each storage
location in the cluster array can contain a level, i.e., a correlogram
identified by its D in | | |
