DETAILED ACTION
Claims 1-10 are presented for examination. Claim 4 stands currently amended.
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Finality of Office Action
The following is a brief summary description of new ground(s) of rejection (if any) and the reason why those new ground(s) are made necessary by this amendment:
No new grounds of rejection are presented herein.
Response to Arguments
Applicant's remarks filed 8 May 2026 have been fully considered and Examiner’s response is as follows:
Applicant remarks page 9 argues:
Applicant submit that the Examiner correctly admitted that the Mathew does not disclose logging curve data, however the Examiners reliance on the AlTammar for a generic teaching that well log data may be used as inputs in a machine-learning model directed to determining mechanical properties for well planning is not justified and invalid as AlTammar only shows that well logs can be used as inputs in a different machine learning context but it does not show the claimed correspondence between the rock sample and the well from which the logging curve data are obtained. Accordingly, the cited combination does not sufficiently teach or suggest this limitation.
This argument is unpersuasive.
First, Applicant’s argument suggests the correspondence between the rock sample and the well is significant. However, the rock sample and well are merely a material upon which the claim is acting. MPEP §2115 states “’[i]nclusion of the material or article worked upon by a structure being claimed does not impart patentability to the claims.’ In re Otto, 312 F.2d 937, 136 USPQ 458, 459 (CCPA 1963); see also In re Young, 75 F.2d 996, 25 USPQ 69 (CCPA 1935).” Here, where the sample is coming from or the respective well of interest is merely the site acted upon by the claim. This correspondence relationship is given limited patentable weight in accordance with MPEP §2115.
Second, Mathew already teaches acquiring relative permeability. See Mathew page 780 fifth paragraph. AlTammar merely provides that this relative permeability data can be acquired from a well logging. AlTammar is only required to teach what Mathew is missing and is not required to further teach limitations already taught by Mathew. Accordingly, Examiner finds the combination of Mathew in view of AlTammar is not deficient as argued here.
Applicant remarks page 9 further argues:
Applicant respectfully submits that the cited combination of prior art does not disclose or suggest the claimed step of selecting a part of the relative permeability curve data in combination with a part of the logging curve data together as sample relative permeability curve data and sample logging curve data for use in machine-learning training. Mathew on page 785 (sixth paragraph) merely discloses parameters selected as features for an ML algorithm, while AlTammar generally teaches that well log data may be used as inputs. Neither reference alone or in combination teaches the claimed combined selection of portions of both relative permeability curve data and logging curve data as sample datasets.
This argument is unpersuasive.
The data which is selected as features for an ML algorithm is a selecting of respective data. That is, the input data which is used is a selecting of data. Because the references as discussed immediately above teach both the relative permeability curve data and logging curve data, it also teaches a selecting of respective parts of that data here.
Applicant remarks page 9-10 further argues:
Moreover, the Office action does not provide any motivation why a person of ordinary skill in the art would have modified Mathew to jointly select portions of both relative permeability curve data and logging curve data as corresponding sample datasets in the particular manner recited in claim 1.
This argument is unpersuasive.
As discussed immediately above, the actual claim language is taught by Mathew and AlTammar. Here, Applicant is arguing that a joint selection of data is required, but the claim does not recite this. Claim 1 clause 2 requires “selecting a part … as sample … data.” The claim language does not require a joint selection or other characteristic not actually recited by the claim.
Applicant remarks page 10 further argues:
Applicant respectfully submits that the given passage of Mathew on page 785 (sixth paragraph) merely discloses parameters used as features for an ML algorithm such as average water saturation and change in water saturation. Such disclosure does not teach or suggest using sample logging curve data as the input and a water saturation starting value in the sample relative permeability curve data as a marker for training a distinct relative permeability curve starting point model.
This argument is unpersuasive. Applicant is arguing against Mathew individually while Examiner’s rejection is based on a combination of references. Notably, Examiners rejection cites AlTammar regarding well logging.
Applicant remarks page 10 further argues:
Moreover, the statement that "without loss of generality some part of the permeability is a starting point" is not a teaching of Mathew and does not identify any disclosure of the specifically claimed water saturation starting value as a marker.
Without loss of generality is a phrase which in this context means any other selection/mapping could be made with equal validity. Here in context it means any of the respective data could be determined to be a “starting value” with equal validity. The phrase is an explanation of Examiners claim mapping and not a teaching from the art.
Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
Clearly, this cited portion of Mathew teaches a water saturation.
Applicant remarks page 10 further argues:
With regard to the Examiner's remarks regarding the limitation "obtaining a predicted water saturation starting value according to the first relative permeability curve starting point model", …. Such disclosure does not teach or suggest a distinct first relative permeability curve starting point model from which a predicted water saturation starting value is obtained as specifically required by claim 1.
This argument is unpersuasive.
Mathew page 786 first paragraph discloses:
the target was set as the entirety of Kr and Pc curves for two samples, whereas for another sample, the target was the Corey exponents nwd and nod, saturation endpoints Swc and
K
r
o
*
, and the fitting parameters for
P
c
-
c
w
d
, cod, awd, aod, and bd. Therefore, the input data set for an ML model comprises the discussed features and target strung together as a single file.
The permeability curve is an obtained first relative permeability. The saturation corresponds with obtained water saturation values. Without loss of generality some part of the permeability is a starting point. (Which is to say any part of the Kr may be considered a “starting” point.)
Applicant remarks pages 10-11 further argues:
With regard to the Examiner's remarks regarding the limitation "taking the sample logging curve data and the predicted water saturation starting value as an input, and a relative permeability in the sample relative permeability curve data as a marker, and training a relative permeability model with the machine learning algorithm to obtain a first relative permeability model", …. Such disclosure does not teach or suggest using a predicted water saturation starting value, obtained from a first model, as an input to a second relative permeability model, as specifically required by claim 1. In other words Mathew discloses general model features whereas claim 1 requires a specific sequential modeling framework in which a first model outputs a predicted water saturation starting value and a second model is then trained using that predicted value together with sample logging curve data.
This argument is unpersuasive.
The only sequential modeling required by this claim limitation is that the sample logging curve data and water saturation value is input training a relative permeability curve model with machine learning algorithm. Mathew paragraph 785 sixth paragraph teaches the respective data “were fed into the ML algorithm.” Applicant argument fails to establish that any particular data is missing or that the data is not used as input to the ML algorithm.
Applicant remarks page 11 further argues:
With regard to the Examiner's remarks regarding the limitation "obtaining a predicted relative permeability according to the first relative permeability model", the Applicant submit that the claim 1 does not merely recite obtaining a predicted relative permeability in general, whereas claim 1 specifically recites obtaining a predicted relative permeability according to the "first relative permeability model" which means the model trained in the immediately preceding step using the sample logging curve data and the predicted water saturation starting value as inputs. As discussed above, the cited combination does not teach or suggest that specifically claimed first relative permeability model.
This argument is unpersuasive.
Mathew figure 1 clearly shows “Predict Kr & PC” downstream of the train AI model. This corresponds with the disclosed “target” of Mathew page 786 first paragraph.
Applicant remarks page 11 further argues:
With regard to the Examiner's remarks regarding the limitation "plotting a relative permeability curve according to the predicted water saturation starting value and the predicted relative permeability corresponding to the predicted water saturation starting value", the Applicant submit that the Mathew may disclose on page 794 (figure 15) a plotted ML output in general but claim 1 recites two distinct model-training steps such as a first training step for obtaining a relative permeability curve starting point model, and a second training step for obtaining a relative permeability model, wherein the second training step uses the predicted water saturation starting value output from the first model as an input. Mathew do not disclose this sequential two-model training architecture. Accordingly, Mathew's plotted output does not teach or suggest the plotting limitation in the claimed context.
This argument is unpersuasive.
Applicant’s argument amounts to arguing that Mathew may disclose the cited limitation of plotting a relative permeability curve” as claimed, but the cited disclosure for this limitation does not teach an entirely different limitation for which Mathew figure 15 was not cited for. This is a non-sequitur. Regarding the relevance of this argument Examiner notes Mathew figure 15 does show plotting the relative permeability curve according to the respective prediction.
Applicant remarks page 12 further argues:
With regard to the Examiner's remarks regarding claim 2, Applicant respectfully submits that the cited combination of Mathew and AlTammar fails to teach or suggest the claimed limitation in the context of the method of claim 1. Claim 2 specifically requires the logging curve data comprise one or more of the recited logging curve types, and the Examiner has selected gamma-ray from the listed alternatives. AlTammar merely identifies gamma ray as one example of a generic well log type among several possible well log types, and is relied upon only for the broad proposition that well log data may be used as inputs in a model for determining mechanical properties for well planning.
This argument is unpersuasive. Applicant admits AlTammar teaches gamma ray as a well log type and input for a modeling. This is all that is required for AlTammar to teach in combination with the teachings of Mathew.
Applicant remarks page 12 further argues:
Such broad disclosure does not teach or suggest the specifically claimed step of performing preprocessing on the logging curve data and taking processed logging curve data as new logging curve data.
This argument is unpersuasive.
Mathew page 783 third paragraph discloses “once the data are secured, they have to be preprocessed depending on the requirements of the chosen algorithm.”
Applicant remarks page 12 further argues:
Such general disclosures do not teach or suggest the specifically claimed testing workflow for a first relative permeability curve starting point model.
This argument is unpersuasive.
Mathew page 786 second paragraph discloses “the remaining 15% was used for blind testing.” See also Mathew page 786 figure 2 showing “testing data.”
Applicant remarks page 13 further argues:
… and do not disclose taking a well-trained relative permeability curve starting point model as a new first relative permeability curve starting point model and returning to the claimed prediction step.
This argument is unpersuasive.
Mathew page 786 third paragraph discloses “The algorithm initializes the model with a first guess and in subsequent steps tries to minimize the loss function, thereby establishing a better relationship between the features and the target as the training process continues.” The teaching of subsequent steps corresponds to the iterative training of the machine learning algorithm with the training data. A subsequent step of minimizing the loss function is taking the new starting point model as a next starting point and returning to the start of the training iterations.
Applicant remarks page 13 further argues:
With regard to the Examiner's remarks regarding claim 6, … Such general disclosures do not teach or suggest the specifically claimed testing workflow for a first relative permeability model.
This argument is unpersuasive.
Mathew page 786 second paragraph discloses “the remaining 15% was used for blind testing.” See also Mathew page 786 figure 2 showing “testing data.”
Applicant remarks page 13 further argues:
With regard to the Examiner's remarks regarding claim 7, … Such disclosure does not teach or suggest the specifically recited machine learning algorithm to run the method of claim 1.
This argument is unpersuasive.
Mathew page 782 sixth paragraph discloses “the main algorithms analyzed in this study are gradient boosting (GB), random forest (RF), extreme GB (XGBoost), and SVM.”
Applicant remarks page 13 further argues:
With regard to the Examiner's remarks regarding claim 8, …. The Examiner relies on substantially the same disclosures as those applied to claim 1 and therefore suffers from the same deficiencies. Accordingly, for at least the reasons set forth with respect to claim 1, Applicant respectfully submits that the cited references, either alone or in combination, do not disclose or suggest claim 8.
This argument is unpersuasive. As discussed above, Examiner has not found the rejection of claim 1 as deficient. Accordingly, neither does Examiner find the rejection of claim 8 as deficient.
Applicant remarks page 14 further argues:
Applicant respectfully submits that Mathew's selected feature parameters are not the same as the specifically claimed selection of a marker layer for logging curve data and the cited references do not disclose or suggest this operation.
This argument is unpersuasive.
A general allegation that the disclosure of Mathew “are not the same” as the claimed selection does not specifically identify or point out any particular difference. Accordingly, Applicant’s argument here amounts to a general allegation of patentability, which is unpersuasive.
Furthermore, Examiner’s rejection of the marker layer for the logging curve data is based upon a combination of references and not only Mathew. Applicant’s argument is addressing Mathew individually while Examiner’s rejection is based on a combination of references.
Applicant remarks page 14 further argues:
Leseur's discussion of filtering preprocessing based on discrepancy between log and core porosity, optimization, and residual-based quality control does not teach or suggest the specifically claimed consistency correction performed on first logging curve data with a plotting tool.
This argument is unpersuasive.
Leseur column 16 lines 37-51 teaches:
For example, one or more embodiments provide a filtering preprocessing based on the discrepancy between log and core porosity that leads to a better prediction (Eq. 1-3). According to one or more embodiments, the input parameters traditionally arbitrarily fixed by the user are optimized through a constrained nonlinear optimization algorithm, and the individual weights of every input property influencing the output KNN-based permeability prediction type can be optimized (Eq. 5). According to one or more embodiments, residuals from a blind test on a one-by-one sample basis, as described in the BTP, are defined as the most effective and least biased objective function to minimize (Eq. 7); and/or the latter objective function sets the basis for measurable and easy-to-interpret quality criteria and QC of the permeability prediction.
The discrepancy is a measure of consistency. Accordingly, filtering preprocessing based on the discrepancy is a consistency correction.
Applicant remarks page 14 further argues:
Mathew's statement that the curves are to be in line with physics does not teach selection of an optimal logging curve according to a rock physical model. Further, the algorithms cited in Mathew are only as machine learning algorithms for prediction, and do not teach the specifically claimed deep learning method for selecting an optimal logging curve from second logging curve data according to known sample logging data.
This argument is unpersuasive.
Mathew page 780 fifth paragraph discloses “Relative permeability can be calculated from the pressure drop and the injected water/oil ratio using Darcy’s law assuming no change in differential pressure and saturation profiles along the core.” Darcy’s law is an empirical equation and rock physics modeling. Mathew page 784 first paragraph disclose “to ensure that the curves are in line with the physics.” Ensuring the curves are in-line with physics is ensuring the curves are according to a rock physical model.
Mathew page 782 sixth paragraph discloses “the main algorithms analyzed in this study are gradient boosting (GB), random forest (RF), extreme GB (XGBoost), and SVM.” The machine learning algorithms are at least one deep learning method.
The respective relative permeability curves in conformance with Darcy’s law, compatible with random forest (RF), and in line with physics are considered to be “optimal” logging curves within the broadest reasonable interpretation of “optimal logging curves” as recited here. See MPEP §2111.
Examiner herein has interpreted “optimal” as being self-defined by the claim as those logging curves which are selected “with an empirical equation, a rock physical model and a deep learning method.” Otherwise, the claim term “optimal” would be an indefinite term of degree under §112(b). However, because the claim itself does provide a standard with which to determine whether or not a logging curve is ‘optimal’ this is the standard with which the claim is required to be interpreted. See MPEP §2111.
Applicant remarks page 15 further argues:
The amended limitation now makes clear that the claim is directed to completion of missing logging-curve data which is not taught or suggested by the cited references.
Examiners rejection has been accordingly updated as necessitated by the amended claim language.
Mathew page 783 sixth paragraph discloses “The missing features in relative permeability prediction were determined using RF regression, and by using pore-network-simulation-based synthetic data.” Determining missing features corresponds with completing missing curve data. The input for the random forest (RF) regression corresponds with respective optimal logging curve data.
Specification
The abstract has been appropriately corrected. Accordingly, Examiner's objection(s) to the specification is withdrawn.
Claim Rejections - 35 USC § 112
Claim 4 has been appropriately corrected to clarify the indefinite claim language. Accordingly, Examiner's rejection of claim 4 under § 112 is withdrawn.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-3 and 5-10
Claims 1-3 and 5-10 are rejected under 35 U.S.C. 103 as being unpatentable over Mathew, E.S., et al. “Artificial Intelligence Coreflooding Simulator for Special Core data Analysis” Society of Petroleum Engineers Reservoir Evaluation & Engineering, pp. 780-808 (May 2021) [herein “Mathew”] in view of US patent 11,828,168 B2 AlTammar, et al. [herein “AlTammar”].
Claim 1 recites “1. A method for predicting a relative permeability curve based on machine learning.” Mathew page 800 Conclusion discloses “Fast and accurate estimation of relative permeability (Kr) and capillary pressure (Pc) curves is critical in multiphase flow in porous media. This paper proposed an AI workflow that was developed to predict Kr and Pc curves.” Mathew page 783 Methodology section discloses “The prediction of relative permeability and capillary pressure curves.”
Claim 1 further recites “comprising: acquiring relative permeability curve data of a rock sample and logging curve data of a well where the rock sample is located, the relative permeability curve data comprising water saturations and relative permeabilities corresponding to different water saturations.” Mathew page 780 fifth paragraph discloses “Relative permeability can be calculated from the pressure drop and the injected water/oil ratio using Darcy’s law assuming no change in differential pressure and saturation profiles along the core.” The calculated relative permeability in the measurement section is an acquired relative permeability curve data of a rock sample. The core is a rock sample. The saturation profiles along the core are a water/oil saturations.
Mathew does not explicitly disclose logging data; however, in analogous art of reservoir data analysis, AlTammar column 2 lines 5-6 teaches “The instructions further include training a PCML model using the obtained well logs data as inputs.”
AlTammar column 1 lines 10-14 teach:
A well log is a detailed and sequential collection of one category of data (e.g., gamma ray, sonic, porosity, resistivity, density, etc.) for a geological formation by using a logging tool along the path of a well borehole in the ground.
Well log data of gamma ray, sonic, resistivity, and density are all well log data.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Mathew and AlTammar. One having ordinary skill in the art would have found motivation to use well log data into the system of artificial intelligence core data analysis for the advantageous purpose of “determining one or more mechanical properties for well planning.” See AlTammar column 2 lines 15-16.
Claim 1 further recites “selecting a part of the relative permeability curve data and a part of the logging curve data as sample relative permeability curve data and sample logging curve data.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data.
Mathew does not explicitly disclose logging data; however, in analogous art of reservoir data analysis, AlTammar column 2 lines 5-6 teaches “The instructions further include training a PCML model using the obtained well logs data as inputs.”
AlTammar column 1 lines 10-14 teach:
A well log is a detailed and sequential collection of one category of data (e.g., gamma ray, sonic, porosity, resistivity, density, etc.) for a geological formation by using a logging tool along the path of a well borehole in the ground.
Well log data of gamma ray, sonic, resistivity, and density are all well log data.
Claim 1 further recites “taking the sample logging curve data as an input and a water saturation starting value in the sample relative permeability curve data as a marker, and training a relative permeability curve starting point model with a machine learning algorithm to obtain a first relative permeability curve starting point model.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data. Training the model is training with the machine learning algorithm. See further Mathew page 784 figure 1. Without loss of generality some part of the permeability is a starting point.
Claim 1 further recites “obtaining a predicted water saturation starting value according to the first relative permeability curve starting point model.” Mathew page 786 first paragraph discloses:
the target was set as the entirety of Kr and Pc curves for two samples, whereas for another sample, the target was the Corey exponents nwd and nod, saturation endpoints Swc and
K
r
o
*
, and the fitting parameters for
P
c
-
c
w
d
, cod, awd, aod, and bd. Therefore, the input data set for an ML model comprises the discussed features and target strung together as a single file.
The permeability curve is an obtained first relative permeability. The saturation corresponds with obtained water saturation values. Without loss of generality some part of the permeability is a starting point. See further Mathew page 784 figure 1.
Claim 1 further recites “taking the sample logging curve data and the predicted water saturation starting value as an input, and a relative permeability in the sample relative permeability curve data as a marker, and training a relative permeability model with the machine learning algorithm to obtain a first relative permeability model.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data. Training the model is training with the machine learning algorithm. See further Mathew page 784 figure 1.
Claim 1 further recites “obtaining a predicted relative permeability according to the first relative permeability model.” Mathew page 786 first paragraph discloses:
the target was set as the entirety of Kr and Pc curves for two samples, whereas for another sample, the target was the Corey exponents nwd and nod, saturation endpoints Swc and
K
r
o
*
, and the fitting parameters for
P
c
-
c
w
d
, cod, awd, aod, and bd. Therefore, the input data set for an ML model comprises the discussed features and target strung together as a single file.
The permeability curve is an obtained first relative permeability.
Claim 1 further recites “and plotting a relative permeability curve according to the predicted water saturation starting value and the predicted relative permeability corresponding to the predicted water saturation starting value.” Mathew page 794 figure 15 shows “Comparison between ML-predicted Kr curves and analytically estimated Kr curves for Reservoir Sample 1.” The figure is a plot of the relative permeability curve as predicted by the machine learning value. Mathew page 794 figure 15 shows “water saturation (fraction)” as the x-axis.
Claim 2 further recites “2. The method for predicting a relative permeability curve based on machine learning according to claim 1, wherein the logging curve data comprise one or more of a gamma-ray (GR), a depth, a diameter, a spontaneous potential (SP), a time difference, a neutron, an acoustic (AC), a shallow resistivity, a gradient resistivity, an induction conductivity (COND) and a density (DEN).” From the above list of alternatives the Examiner is selecting “a gamma-ray (GR).”
Mathew does not explicitly disclose logging data; however, in analogous art of reservoir data analysis, AlTammar column 2 lines 5-6 teaches “The instructions further include training a PCML model using the obtained well logs data as inputs.”
AlTammar column 1 lines 10-14 teach:
A well log is a detailed and sequential collection of one category of data (e.g., gamma ray, sonic, porosity, resistivity, density, etc.) for a geological formation by using a logging tool along the path of a well borehole in the ground.
Well log data of gamma ray, sonic, resistivity, and density are all well log data. See further AlTammar column 2 line 66.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Mathew and AlTammar. One having ordinary skill in the art would have found motivation to use well log data into the system of artificial intelligence core data analysis for the advantageous purpose of “determining one or more mechanical properties for well planning.” See AlTammar column 2 lines 15-16.
Claim 3 further recites “3. The method for predicting a relative permeability curve based on machine learning according to claim 1, after the acquiring relative permeability curve data of a rock sample and logging curve data of a well where the rock sample is located, further comprising: performing preprocessing on the logging curve data; and taking processed logging curve data as new logging curve data.” Mathew page 783 third paragraph discloses “once the data are secured, they have to be preprocessed depending on the requirements of the chosen algorithm.”
Claim 5 further recites “5. The method for predicting a relative permeability curve based on machine learning according to claim 1, after the obtaining a first relative permeability curve starting point model, further comprising: testing the first relative permeability curve starting point model, specifically comprising: respectively selecting remaining relative permeability curve data and remaining logging curve data except the sample relative permeability curve data and the sample logging curve data as test relative permeability curve data and test logging curve data.” Mathew page 786 second paragraph discloses “the remaining 15% was used for blind testing.” See also Mathew page 786 figure 2 showing “testing data.”
Claim 5 further recites “inputting the test logging curve data to the first relative permeability curve starting point model to obtain a predicted water saturation starting value.” Mathew page 786 figure 2 shows inputting the test data into the “test model.”
Claim 5 further recites “establishing a loss function with a mean square error (MSE) according to the predicted water saturation starting value and a water saturation starting value in the test relative permeability curve data.” Mathew page 804 Appendix A discloses “This loss function is quantified by the mean squared error.” Mathew page 786 figure 2 shows RMSE output from the test model which received the test data. Mathew figure 2 caption defines “RMSE=root mean squared error.”
Claim 5 further recites “training the first relative permeability curve starting point model completely in case of a minimum of the loss function to obtain a well-trained relative permeability curve starting point model.” Mathew page 786 third paragraph discloses “tries to minimize the loss function, thereby establishing a better relationship between the features and the target as the training process continues.” Minimizing the loss function corresponds with completing the training at a minimum of the loss function to obtain the trained model. See further Mathew page 786 figure 2.
Claim 5 further recites “and taking the well-trained relative permeability curve starting point model as a new first relative permeability curve starting point model, and returning to the step of "obtaining a predicted water saturation starting value according to the first relative permeability curve starting point model".” Mathew page 786 third paragraph discloses “The algorithm initializes the model with a first guess and in subsequent steps tries to minimize the loss function, thereby establishing a better relationship between the features and the target as the training process continues.” The teaching of subsequent steps corresponds to the iterative training of the machine learning algorithm with the training data. A subsequent step of minimizing the loss function is taking the new starting point model as a next starting point and returning to the start of the training iterations.
Claim 6 further recites “6. The method for predicting a relative permeability curve based on machine learning according to claim 1, after the obtaining a first relative permeability model, further comprising: testing the first relative permeability model, specifically comprising: respectively selecting remaining relative permeability curve data and remaining logging curve data except the sample relative permeability curve data and the sample logging curve data as test relative permeability curve data and test logging curve data.” Mathew page 786 second paragraph discloses “the remaining 15% was used for blind testing.” See also Mathew page 786 figure 2 showing “testing data.”
Claim 6 further recites “inputting the test logging curve data and the predicted water saturation starting value to the first relative permeability model to obtain a predicted relative permeability.” Mathew page 786 figure 2 shows inputting the test data into the “test model.”
Claim 6 further recites “establishing a loss function with an MSE according to the predicted relative permeability and a relative permeability in the test relative permeability curve data.” Mathew page 804 Appendix A discloses “This loss function is quantified by the mean squared error.” Mathew page 786 figure 2 shows RMSE output from the test model which received the test data. Mathew figure 2 caption defines “RMSE=root mean squared error.”
Claim 6 further recites “training the first relative permeability model completely in case of a minimum of the loss function to obtain a well-trained relative permeability model.” Mathew page 786 third paragraph discloses “tries to minimize the loss function, thereby establishing a better relationship between the features and the target as the training process continues.” Minimizing the loss function corresponds with completing the training at a minimum of the loss function to obtain the trained model. See further Mathew page 786 figure 2.
Claim 6 further recites “and taking the well-trained relative permeability model as a new first relative permeability model, and returning to the step of "obtaining a predicted relative permeability according to the first relative permeability model".” Mathew page 786 third paragraph discloses “The algorithm initializes the model with a first guess and in subsequent steps tries to minimize the loss function, thereby establishing a better relationship between the features and the target as the training process continues.” The teaching of subsequent steps corresponds to the iterative training of the machine learning algorithm with the training data. A subsequent step of minimizing the loss function is taking the new starting point model as a next starting point and returning to the start of the training iterations.
Claim 7 further recites “7. The method for predicting a relative permeability curve based on machine learning according to claim 1, wherein the machine learning algorithm comprises a random forest (RF), an adaptive boosting (AdaBoost), a gradient boosted decision tree (GBDT) and an extreme gradient boosting (XGBoost).” From the above list of alternatives the Examiner is selecting “a random forest (RF).”
Mathew page 782 sixth paragraph discloses “the main algorithms analyzed in this study are gradient boosting (GB), random forest (RF), extreme GB (XGBoost), and SVM.”
Claim 8 further recites “8. A system for predicting a relative permeability curve based on machine learning.” Mathew page 800 Conclusion discloses “Fast and accurate estimation of relative permeability (Kr) and capillary pressure (Pc) curves is critical in multiphase flow in porous media. This paper proposed an AI workflow that was developed to predict Kr and Pc curves.” Mathew page 783 Methodology section discloses “The prediction of relative permeability and capillary pressure curves.”
Claim 8 further recites “comprising: a sample acquisition module configured to acquire relative permeability curve data of a rock sample and logging curve data of a well where the rock sample is located, the relative permeability curve data comprising water saturations and relative permeabilities corresponding to different water saturations.” Mathew page 780 fifth paragraph discloses “Relative permeability can be calculated from the pressure drop and the injected water/oil ratio using Darcy’s law assuming no change in differential pressure and saturation profiles along the core.” The calculated relative permeability in the measurement section is an acquired relative permeability curve data of a rock sample. The core is a rock sample. The saturation profiles along the core are a water/oil saturations.
Mathew does not explicitly disclose logging data; however, in analogous art of reservoir data analysis, AlTammar column 2 lines 5-6 teaches “The instructions further include training a PCML model using the obtained well logs data as inputs.”
AlTammar column 1 lines 10-14 teach:
A well log is a detailed and sequential collection of one category of data (e.g., gamma ray, sonic, porosity, resistivity, density, etc.) for a geological formation by using a logging tool along the path of a well borehole in the ground.
Well log data of gamma ray, sonic, resistivity, and density are all well log data.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Mathew and AlTammar. One having ordinary skill in the art would have found motivation to use well log data into the system of artificial intelligence core data analysis for the advantageous purpose of “determining one or more mechanical properties for well planning.” See AlTammar column 2 lines 15-16.
Claim 8 further recites “a sample data selection module configured to select a part of the relative permeability curve data and a part of the logging curve data as sample relative permeability curve data and sample logging curve data.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data.
Mathew does not explicitly disclose logging data; however, in analogous art of reservoir data analysis, AlTammar column 2 lines 5-6 teaches “The instructions further include training a PCML model using the obtained well logs data as inputs.”
AlTammar column 1 lines 10-14 teach:
A well log is a detailed and sequential collection of one category of data (e.g., gamma ray, sonic, porosity, resistivity, density, etc.) for a geological formation by using a logging tool along the path of a well borehole in the ground.
Well log data of gamma ray, sonic, resistivity, and density are all well log data.
Claim 8 further recites “a first relative permeability curve starting point model training module configured to take the sample logging curve data as an input and a water saturation starting value in the sample relative permeability curve data as a marker, and train a relative permeability curve starting point model with a machine learning algorithm to obtain a first relative permeability curve starting point model.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data. Training the model is training with the machine learning algorithm. See further Mathew page 784 figure 1. Without loss of generality some part of the permeability is a starting point.
Claim 8 further recites “a water saturation starting value prediction module configured to obtain a predicted water saturation starting value according to the first relative permeability curve starting point model.” Mathew page 786 first paragraph discloses:
the target was set as the entirety of Kr and Pc curves for two samples, whereas for another sample, the target was the Corey exponents nwd and nod, saturation endpoints Swc and
K
r
o
*
, and the fitting parameters for
P
c
-
c
w
d
, cod, awd, aod, and bd. Therefore, the input data set for an ML model comprises the discussed features and target strung together as a single file.
The permeability curve is an obtained first relative permeability. The saturation corresponds with obtained water saturation values. Without loss of generality some part of the permeability is a starting point.
Claim 8 further recites “a first relative permeability model training module configured to take the sample logging curve data and the predicted water saturation starting value as an input, and a relative permeability in the sample relative permeability curve data as a marker, and train a relative permeability model with the machine learning algorithm to obtain a first relative permeability model.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data. Training the model is training with the machine learning algorithm. See further Mathew page 784 figure 1.
Claim 8 further recites “a relative permeability prediction module configured to obtain a predicted relative permeability according to the first relative permeability model.” Mathew page 786 first paragraph discloses:
the target was set as the entirety of Kr and Pc curves for two samples, whereas for another sample, the target was the Corey exponents nwd and nod, saturation endpoints Swc and
K
r
o
*
, and the fitting parameters for
P
c
-
c
w
d
, cod, awd, aod, and bd. Therefore, the input data set for an ML model comprises the discussed features and target strung together as a single file.
The permeability curve is an obtained first relative permeability.
Claim 8 further recites “and a relative permeability curve plotting module configured to plot a relative permeability curve according to the predicted water saturation starting value and the predicted relative permeability corresponding to the predicted water saturation starting value.” Mathew page 794 figure 15 shows “Comparison between ML-predicted Kr curves and analytically estimated Kr curves for Reservoir Sample 1.” The figure is a plot of the relative permeability curve as predicted by the machine learning value. Mathew page 794 figure 15 shows “water saturation (fraction)” as the x-axis.
Dependent claim 9 is substantially similar to claim 3 above and is rejected for the same reasons.
Dependent claim 10 is substantially similar to claim 5 above and is rejected for the same reasons.
Dependent Claim 4
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Mathew and AlTammar as applied to claim 3 above, and further in view of US patent 9,229,127 B2 Leseur [herein “Leseur”].
Claim 4 further recites “4. The method for predicting a relative permeability curve based on machine learning according to claim 3, wherein the performing preprocessing on the logging curve data specifically comprises: selecting a marker layer for the logging curve data to obtain first logging curve data.” Mathew page 785 sixth paragraph discloses:
the final list of parameters that were fed into the ML algorithm as features are core dimensions, porosity (
ϕ
), permeability (K), pressure drop (ΔP), average water saturation of the core, change in water saturation (Sw) along the core in selected gridblocks, and injection rates of both oil and water. The pressure drop and the average water saturation points included in the features were equal in number and correspond to the specific timesteps mentioned earlier.
The parameters selected to input into the ML algorithm are selected data. Without loss of generality, any of the selected data is a marker layer within the data.
Mathew does not explicitly disclose logging data; however, in analogous art of reservoir data analysis, AlTammar column 2 lines 5-6 teaches “The instructions further include training a PCML model using the obtained well logs data as inputs.”
AlTammar column 1 lines 10-14 teach:
A well log is a detailed and sequential collection of one category of data (e.g., gamma ray, sonic, porosity, resistivity, density, etc.) for a geological formation by using a logging tool along the path of a well borehole in the ground.
Well log data of gamma ray, sonic, resistivity, and density are all well log data.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Mathew and AlTammar. One having ordinary skill in the art would have found motivation to use well log data into the system of artificial intelligence core data analysis for the advantageous purpose of “determining one or more mechanical properties for well planning.” See AlTammar column 2 lines 15-16.
Claim 4 further recites “performing consistency correction on the first logging curve data with a plotting tool to obtain second logging curve data.” Mathew does not explicitly disclose a consistency correction; however, in analogous art of reservoir analysis, Leseur column 16 lines 37-51 teaches:
For example, one or more embodiments provide a filtering preprocessing based on the discrepancy between log and core porosity that leads to a better prediction (Eq. 1-3). According to one or more embodiments, the input parameters traditionally arbitrarily fixed by the user are optimized through a constrained nonlinear optimization algorithm, and the individual weights of every input property influencing the output KNN-based permeability prediction type can be optimized (Eq. 5). According to one or more embodiments, residuals from a blind test on a one-by-one sample basis, as described in the BTP, are defined as the most effective and least biased objective function to minimize (Eq. 7); and/or the latter objective function sets the basis for measurable and easy-to-interpret quality criteria and QC of the permeability prediction.
The discrepancy is a measure of consistency. Accordingly, filtering preprocessing based on the discrepancy is a consistency correction.
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to combine Mathew, AlTammar, and Leseur. One having ordinary skill in the art would have found motivation to use preprocessing into the system of artificial intelligence core data analysis for the advantageous purpose “for measurable and easy-to-interpret quality criteria and QC of the permeability prediction.” See Leseur column 16 lines 37-51.
Claim 4 further recites “selecting an optimal logging curve from the second logging curve data with an empirical equation, a rock physical model and a deep learning method.” Mathew page 780 fifth paragraph discloses “Relative permeability can be calculated from the pressure drop and the injected water/oil ratio using Darcy’s law assuming no change in differential pressure and saturation profiles along the core.” Darcy’s law is an empirical equation and rock physics modeling. Mathew page 784 first paragraph disclose “to ensure that the curves are in line with the physics.” Ensuring the curves are in-line with physics is ensuring the curves are according to a rock physical model.
Mathew page 782 sixth paragraph discloses “the main algorithms analyzed in this study are gradient boosting (GB), random forest (RF), extreme GB (XGBoost), and SVM.” The machine learning algorithms are at least one deep learning method.
The respective relative permeability curves in conformance with Darcy’s law, compatible with random forest (RF), and in line with physics are considered to be “optimal” logging curves within the broadest reasonable interpretation of “optimal logging curves” as recited here. See MPEP §2111.
Claim 4 further recites “and completing a missing logging curve by predicting the missing logging curve according to the optimal logging curve.” Mathew page 783 sixth paragraph discloses “The missing features in relative permeability prediction were determined using RF regression, and by using pore-network-simulation-based synthetic data.” Determining missing features corresponds with completing missing curve data. The input for the random forest (RF) regression corresponds with respective optimal logging curve data.
Conclusion
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Jay B Hann whose telephone number is (571)272-3330. The examiner can normally be reached M-F 10am-7pm EDT.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Renee Chavez can be reached at (571) 270-1104. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/Jay Hann/Primary Examiner, Art Unit 2186 16 July 2026
Prosecution Insights