Вы находитесь на странице: 1из 18

Rock Mechanics Models

The key to unlocking the full potential of GeoMechanics lies in determining rock

compressive strength, and that process begins with the rock mechanics models:

Lithology

Rock Strength

Shale Plasticity

Because all of the rock mechanics models focus on intrinsic rock properties,

rather than on any particular log, any log suite that is sensitive to porosity and
lithology can be used. This allows use of some of the most advanced logging
technology available including nuclear magnetic resonance (NMR),

photoelectric and neutron-density and puts GeoMechanics on an automatic


upgrade path as logging technology continues to improve.
Focusing on intrinsic rock properties has another benefit: It allows calibration of
well logs, an often overlooked, but important quality issue. Log calibration
significantly improves the accuracy of field logs, which can exhibit quite a bit of
error due to the many factors that influence the logging environment. A sonic
log, for example, which typically yields a high porosity, can be calibrated
against a measured core porosity or a more accurate nuclear log. The sonic log
might show an average porosity of 20% while the neutron-density yields only
10%. In such a case, the sonic log can be shifted lower until it overlays the
neutron-density. This calibration can remove significant error in the sonic log.
Such focus on achieving the best quality log data is a fundamental premise of
GeoMechanics at every level. As you are about to find out, the mathematical
concepts behind the rock mechanics models also ensure log data is soundly
manipulated so as to achieve the best, most accurate results.

Lithology Model
It is common practice in the oil industry to log a well, either while drilling or
after the well is completed, in order to evaluate the lithology of formations

penetrated by the drill bit. Knowledge of the earths mineralogical composition


the amounts of shale, sandstone and other components is important to
accurate rock strength analysis and petroleum reservoir modeling.
GeoMechanics Lithology Model

Page 1

The problem with conventional methods of estimating lithology is that they do

not correctly identify the range of solutions possible, and may even assume the
existence of a unique solution, when, in fact, there are infinite possible
solutions that lie within a quantifiable range of values.
Some existing methods attempt to determine a unique solution based on a
mathematical model involving use of simultaneous equations. Problems with
this type of approach include:

the assumption that each well log is equally accurate in distinguishing all
lithologic components

the assumption that all lithologic components must exist in pure form at
a reference value for a given log

the assumption that each lithologic component exists to some extent


over the entire range of possible log values (i.e., restricted ranges are not
possible)

Empirical data clearly shows these assumptions are wrong; a new lithology
model is needed.
Replacing Assumptions with Advantages
In creating a new model, first, limits are established to define the valid range of
all possible solutions to mineral composition; that is, what are the most and
least amounts of each component that may be present. This is done by

mathematically modeling lithology in two different ways: A pure component


model is needed to define the upper limit, and a proportional mixture model
is needed to define the lower. Since two models are employed to define the

limits correctly, the new lithology model is called a dual compositional model.
The solution identified within this established range can then be calibrated to a
mineralogical analysis of core samples, which makes the dual compositional

model a self-proving method. In addition, the new method acknowledges the


fact that some logs are more accurate than others in resolving particular

lithologic components. For instance, its well known that a gamma ray log is
the most accurate in resolving shale, while a neutron log typically is more
accurate than a sonic log for resolving coal. In short, the model recognizes that

GeoMechanics Lithology Model

Page 2

each log has unique strengths and weaknesses when it comes to identifying a
given lithology.

In summary, the dual compositional model offers the following advantages:

The valid range of all possible solutions to lithologic composition is

identified. This allows the maximum concentration of any component to


vary from 0 to 100%. Other methods generally assume the maximum
concentration must always be 100%.

A given log can be empirically weighted to a specific lithologic


component. For example, a gamma ray may be 90% more accurate than a
sonic log for shale resolution. On the other hand, the sonic log may be
90% more accurate in resolving non-shales such as sandstone or

limestone. Surprisingly, other methods generally overlook this key factor.

Methods that rely on simultaneous equations are prone to computational


problems, including division by zero errors and negative component

concentrations. Due to the nature of matrix algebra, the frequency of


these errors increases exponentially as the number of components

increases. Such methods are not tolerant of errors in the log data either.
With the dual compositional model, division by zero errors and negative
component concentrations are avoided altogether. Also, the new model is
tolerant of bad log data and will preserve the trend line of the log where
other methods would distort the trend or fail.

The dual compositional model can handle any number or combination of

well logs, and any number or combination of lithologic components. This


is a significant advantage over other methods, which generally restrict the
number of components to the number of well logs available.

The dual compositional model allows ranges of existence for each


lithologic component. These can be measured or inferred from laboratory
data. Other methods generally do not allow for component ranges.

With other methods it is difficult or impossible to calibrate a log-derived


lithology analysis to a laboratory measurement of mineral composition
from a core sample. The dual compositional model is designed to do so,
and delivers improved multi-component analysis as a result.

GeoMechanics Lithology Model

Page 3

Component Reference Values:


Limestone
Sandstone

Shale

700

Well Depth (ft)

Shale

800

Sandstone

Maximum sandstone
concentration

Limestone
900
47

55

M atrix Sonic Log

90

M atrix Sonic Log (ms/ft)

How the Model Works


The flow chart diagram at the end of this section illustrates the dual

compositional modeling process. Recall that the model can use any suite of
well logs sensitive to lithology. In addition, mud logs are necessary to identify
components that are physically present, and core samples are desirable though

not strictly necessary. Using these resources, log data is prepared for input to
the model: first, data is analyzed, then converted to matrix values where
possible, and finally correlated with mud logs. With this input, the dual

compositional model then can determine the concentration of each lithologic


component present, and output a final lithology based on multi-log analysis. As
usual, the process begins with porosity.
It is assumed that formation porosity can be extracted from the log suite using
any of several methods currently in use by the industry. A lithologyindependent porosity, such as the neutron-density, is preferred, since this

allows the log analyst to convert raw log data directly to matrix values, which
reflect properties of the solid rock matrix only. With most lithology logs, a

matrix transform is used to remove the effects of porosity, converting the raw
logs to matrix values.
GeoMechanics Lithology Model

Page 4

Matrix Transform
The matrix transform is based on Wyllies equation, shown here for the sonic
log, which essentially states that the speed
of sound through the sandstone matrix

and the speed of sound through the water


must add up to the sonic log value

Pore
Space

Sandstone M atrix

depending on the porosity.


where:
tlog

raw sonic log (micosec/ft)

tmatrix transit time of the formation


matrix, 55 (micosec/ft) for

55 m s/ft

189 m s/ft

Wyllies Equation:
tlog = ftpore + (1 - f)tmatrix

sandstone
tpore

transit time of the fluid occupying the pore space, 189 (micosec/ft)

formation porosity

for water

In Wyllies equation, a matrix value must be assumed in order to solve for

porosity. In this case, a pure sandstone is assumed. However, if porosity is


already known from another log, such as neutron-density, the raw sonic log
can be converted to a matrix log by re-arranging Wyllies equation and solving
for the matrix transit time:

tmatrix = ( t log - f tpore ) / (1 - f )


This form of Wyllies equation is known as a matrix transform. Similar matrix
transforms can be derived for any lithology-sensitive log including

photoelectric, density, sonic, and gamma ray logs. Note that the photoelectric
log is often converted to the volumetric cross section, U, before a matrix
transform is applied.
Mud Log Correlation
As mentioned earlier, mud logs are needed to identify the components

physically present within the formation. It is a common misconception that


lithology logs are able to sense the difference between, say, sandstone and
limestone. Unfortunately, they cant. Currently available logging technology

simply cannot discriminate between various non-shales with any certainty. In


GeoMechanics Lithology Model

Page 5

order to perform the log analysis correctly, the components physically present
in the formation must be known from a source other than the well logs. Mud
logs provide an excellent source since the components are verified by an
inspection of drill cuttings.

Overview of the Dual Compositional Model


With each log individually prepared as described above, the dual compositional
model then evaluates how each log sees the formation. To do so, the model
first determines the component concentrations for each log individually. The

concentrations are then normalized because all the components must always
sum to 100%. Finally, the well logs are combined by applying weighting factors

that accounts for the strengths and weaknesses of each log individually as well
as variations in log quality. The resulting output is a final lithology based on
multi-log analysis.

Component Concentration
The model determines the concentration of a given component in the solid
formation matrix based on a specific reference value for each component:

Sandstone, used in the example above, has a sonic reference value of about 55
microseconds/ft. At this value, sandstone reaches a maximum concentration in
the formation. Note that this concentration is not necessarily 100%, but can be
any non-zero value between 0 and 100%.
As shown in the figure below, the sandstone concentration decreases as the log
data moves away from the reference value, and toward values above and below
the reference value. These extinction limits, where the sandstone
concentration diminishes to zero, can be measured or inferred from laboratory
tests.

With the extinction limits known, the concentration of sandstone may now be
modeled as follows, using the sonic log for illustration purposes. If a given
sonic value is greater than the sandstone reference value, then:
fSS = (( tlog - tSS ) / ( tSSmax - tSS )) a
On the other hand, if the sonic value is less than the reference value, then:
GeoMechanics Lithology Model

Page 6

fSS = (( tSS - tlog ) / ( tSS - tSSmin )) a


Concentration factors for other components may be derived similarly, using any
lithology log. Note that the sandstone concentration is reduced to extinction
when its concentration factor fSS = 1. When the sonic log value coincides with

the sandstone reference value, that is, when tlog = tSS , then the concentration
factor diminishes to zero and the sandstone concentration reaches a maximum.
This behavior can be mathematically modeled as follows:
cSS = cSSmax (1 - fSS )
where:
cSS

concentration of sandstone (fraction)

cSsmax maximum concentration of sandstone (fraction, 0 cSSmax 1)


where:
fSS

concentration factor of sandstone in matrix (fraction)

tSS

reference sonic value for sandstone, 55 (micosec/ft)

tlog

any sonic log value (micosec/ft)

tSSmax maximum extinction limit for sandstone (micosec/ft)


tSSmin minimum extinction limit for sandstone (micosec/ft)
a

mineralogy exponent, usually set to 1

GeoMechanics Lithology Model

Page 7

Concentration factors are shown graphically below:


Component Reference Values:
Limestone
Sandstone

Shale

700

Well Depth (ft)

fLS

800

fSH

fSS

Minimum
sandstone
extinction
limit
Maximum sandstone
extinction limit

M atrix Sonic Log


900
47

55

90

M atrix Sonic Log (ms/ft)


Component Concentration Factors and Extinction Limits

The concentration of other components can be modeled similarly.


Normalizing Component Concentrations
Recall that the range of concentration values is determined mathematically

using the dual proportional mixture and pure component models. These
models yield concentration values that are not normalized; that is, they do not
sum to one. The following equations describe the process for normalizing
component concentrations.

Proportional Mixture
GeoMechanics Lithology Model

Page 8

In the proportional mixture model, normalizing components is accomplished by


dividing each component by the sum of all components present. For instance, a
three component mixture of sandstone, limestone, and shale would be
normalized as follows:
cSS / ( cSS + cLS + cSH ) + cLS / ( cSS + cLS + cSH ) + cSH / ( cSS + cLS + cSH ) = 1
where:
cLS

cSH

concentration of limestone (fraction)


concentration of shale (fraction)

The normalized concentration for sandstone may now be expressed as:


VSS = cSS / ( cSS + cLS + cSH )
And the normalization equation may be simplified to:
VSS + VLS + VSH = 1
where:
VSS

sandstone concentration, proportional mixture model (fraction)

VLS

limestone concentration, proportional mixture model (fraction)

VSH

shale concentration, proportional mixture model (fraction)

The normalization equation above is referred to as a proportional mixture


model since it precludes the existence of any component in pure form, even at
that components reference value.
As shown in the diagram below, ranges of various components generally
overlap to some extent, so that even when the maximum concentration of each
component is 100%, the normalized concentrations reach a maximum that is

less than 100%. Stated another way, all components combined must always add
up to 100%, and even when a component reaches its reference value, there are
other components present. None is 100%. This effect is shown below:

GeoMechanics Lithology Model

Page 9

Component Reference Values:


D olomite
Limestone
Sandstone

Shale

Concentraion (% by Volume)

100

50

0
43

47

55

90

M atrix Sonic Log (ms/ft)


Proportional M ixture M odel Example

While the proportional mixture model allows the maximum concentration of a


given component to drop to zero, it does not permit a component to exist in

pure form. Therefore, a pure component model is needed to describe this latter
situation.
Pure Component Model
A pure component model is derived by multiplying the concentration of each
component by the concentration factors of all other components present.
Again, assuming a three component mixture:
cSSP = cSS fLS fSH
where:
cSSP

sandstone concentration, pure component model (fraction)

Other components may be modeled similarly. Normalizing the components


yields the following expression for the pure component model:
cSSP / ( cSSP + cLSP + cSHP ) + cLSP / ( cSSP + cLSP + cSHP ) + cSHP / ( cSSP + cLSP + cSHP ) =1
GeoMechanics Lithology Model

Page 10

where:
cLSP

limestone concentration, pure component model (fraction)

cSHP

shale concentration, pure component model (fraction)

The normalized concentration for sandstone may now be expressed as:


VSSP = cSSP / ( cSSP + cLSP + cSHP )
And the normalization equation may be simplified to:
VSSP + VLSP + VSHP = 1
where:
VSSP

sandstone concentration, pure component model (fraction)

VLSP

limestone concentration, pure component model (fraction)

VSHP

shale concentration, pure component model (fraction)

The pure component model guarantees that a given component will be 100%
pure at its reference value. As shown below, this forces the other components
to zero.

GeoMechanics Lithology Model

Page 11

Component Reference Values:


D olomite
Limestone
Sandstone

Shale

Concentraion (% by Volume)

100

50

0
43

47

55

90

M atrix Sonic Log (ms/ft)


Pure Component M odel Example

Concentrations that lie between the pure and proportional limits can be
modeled by taking a weighted average of the two models. For instance, 90% of
the pure model and 10% of the proportional mixture model would yield an

impure concentration between the two. In this fashion the dual compositional
model can be calibrated to a mineral analysis of an actual core sample as shown
below. Once this is accomplished, the model is self-proving since it is in
agreement with measured core data.

GeoMechanics Lithology Model

Page 12

Component Reference Values:


D olomite
Limestone
Sandstone

Concentraion (% by Volume)

100

Shale
Core Calibration
Points

50

0
43

55

47

90

M atrix Sonic Log (ms/ft)


D ual Compositional M odel Example

For instance, the dual compositional model may be applied to sandstone as


follows:
VSSC = VSS ( 1 - B ) + VSSP B
where:
VSSC

sandstone concentration, dual compositional model

VSS

sandstone concentration, proportional mixture model

pure component model weighting factor (empirical)

VSSP

sandstone concentration, pure component model

Concentrations for other components may be modeled similarly.


Multi-Log Analysis
Up to this point, weve applied the dual compositional model to a single well
log only. Multiple logs may be analyzed by simply applying the model to each
log individually.

GeoMechanics Lithology Model

Page 13

In a computerized application, it is important to visually inspect the lithology

derived from each log, since every log has its own particular view of formation
composition, as well as its own strengths and weaknesses. Because different
logs see the formation differently, there is also a need to assign an empirical
weighting factor to each component of each log. If the quality of an individual

log is poor, then low or zero weighting factors can be applied to minimize the
impact of the log on the overall analysis.
For example, a gamma ray log might show a massive shale section, while sonic
and density logs show only trace amounts of shale in the same depth interval.
As every log analyst knows, the gamma ray is essentially a shale indicator; thus,
it should be given a high weighting factor, even 100%. The sonic and density
logs would get a low or zero weighting since they are poor shale indicators.
In the same depth interval, suppose there are sandstone, limestone, and coal

sections. If the density and sonic logs are in reasonably good agreement, then
50% of the sandstone concentration could be taken from each log, with the
same for limestone.
However, when logs disagree significantly, the analyst must exercise greater
judgement. If, after carefully examining the mud log, he feels that the density
log was significantly in error, he might give the sonic log a 100% weighting for
both sandstone and limestone.

For coal, he would probably give high weighting, even 100%, to the density log,
since coal is very light and would register strongly. The sonic log should also

sense coal, but is far less accurate. Because a sonic log sometimes sees coal
as limestone or dolomite, the analyst might give it a zero weighting factor.
Although some logging experts might disagree, the gamma ray is a poor

differentiator of non-shales, and should be given a low or zero weighting for


non-shales. There is one notable exception to this rule: when there is only one
non-shale, the entire lithology analysis can be done with the gamma ray log.
The following table shows how the weighting factors are applied:
Components
Coal
Gamma Ray

Limestone

Sandstone

Shale

VCL

WCL

VLS

WLS

VSS

WSS

VSH

WSH

(0.1)

(0)

(0.1)

(0)

(0.1)

(0)

(0.7)

(1.0)

GeoMechanics Lithology Model

Page 14

Sonic

(0.1)

(0)

(0.2)

(0.6)

(0.5)

(0.6)

(0.2)

(0)

Density

(0.7)

(1)

(0.1)

(0.4)

(0.1)

(0.4)

(0.1)

(0)

Weighted Totals

0.7

0.16

0.34

0.7

where:
VCL

coal concentration

VLS

limestone concentration

VSS

sandstone concentration

WCL

coal weighting factor

WLS

limestone weighting factor

WSH

shale weighting factor

VSH

WSS

shale concentration

sandstone weighting factor

Also note that the following constraints apply:


VCL + VLS + VSS + VSH = 1
( WLS ) GAMMA RAY LOG + ( WLS ) SONIC LOG + ( WLS ) DENSITY LOG = 1
The concentrations, V, must sum to 100% for each component. The weighting
factors, W, also must sum to 100%, but for each log rather than each

component. In the table, concentrations are summed horizontally by row, and


weighting factors are applied vertically by column. This may be mathematically
expressed as follows:
VSHW = ( VSH WSH ) GAMMA RAY LOG + ( VSH WSH ) SONIC LOG + ( VSH WSH ) DENSITY LOG
VSSW = ( VSS WSS ) GAMMA RAY LOG + ( VSS WSS ) SONIC LOG + ( VSS WSS ) DENSITY LOG
VLSW = ( VLS WLS ) GAMMA RAY LOG + ( VLS WLS ) SONIC LOG + ( VLS WLS ) DENSITY LOG
VCLW = ( VCL WCL ) GAMMA RAY LOG + ( VCL WCL ) SONIC LOG + ( VCL WCL ) DENSITY LOG
where:
VCLW weighted coal concentration
VLSW

weighted limestone concentration

VSSW

weighted sandstone concentration

GeoMechanics Lithology Model

Page 15

VSHW weighted shale concentration


Because of the empirical weighting factors, the weighted concentrations do not
sum to one; they must be normalized. However, since the weighted shale
volume is always correct, it needs no further adjustment. For consistency of

terminology, the final shale volume is set equal to the weighted shale volume:
VSHF = VSHW
However, the non-shales must be normalized again. This is accomplished by
multiplying each of the non-shales by a normalizing factor:
k = ( 1 - VSHF ) / ( VCLW + VLSW + VSSW )
The final concentrations of the non-shales are now given by:
VCLF = VCLW k
VLSF = VLSW k
VSSF = VSSW k
Garbage In, Garbage Out
Under certain unusual conditions, some final concentrations may be in error. In
such cases, it is desirable to suppress the incorrect lithologic component in
essence, eliminating the garbage.

For instance, a coal stratum often is resolved by the density log, but poorly or
incorrectly resolved by the sonic log, which may see coal as limestone due to

the high coal porosity. In a computerized application, the following logic might
be used to suppress the erroneous limestone: If the volume of coal seen by the
density log exceeds a threshold volume of 5%, then set the volume of limestone
to zero.

Or, in pseudocode: If VCOAL > 5%, then VLS = 0.


Similar logic may be applied to other components as necessary.
Summary of Lithology Model
With the strength of a dual compositional model, the Lithology model of the
GeoMechanics rock mechanics system provides powerful multi-component

analysis that can be verified by calibration with core samples. This self-proving
model, diagrammed below, incorporates advanced mathematical computations
GeoMechanics Lithology Model

Page 16

that account for the myriad of factors that influence log data, resulting in a

reliable, accurate estimation of formation lithology. Yet, beyond the math, there
remains a need for subjective evaluation of log data, so that once again, it is
the skill and experience of the log analyst which bring the capabilities of
GeoMechanics to the highest level.

GeoMechanics Lithology Model

Page 17

Dual Compositional M odel Flowchart


Well Logs

D etermine porosity from


log suite

Convert logs to matrix values


if possible

M ud Logs

D etermine which lithologies


are physically present

Process lithology for


each log individually

Core samples
(optional)

D etermine concentration of
each lithologic component
using dual compositional
model. Compare lithology
to mud log and adjust
concentration factors as
necessary.

All well logs processed?

Process loop
for each log
individually

No

Yes
D etermine multilog weighting
factors for each log

O utput final lithology


from multilog analysis

GeoMechanics Lithology Model

Page 18

Вам также может понравиться