DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 02/27/2026 has been entered.
Response to Arguments
Applicant's arguments, see Remarks, filed 02/27/2026, with respect to the rejection(s) of claims 1, 11, and 12 under 35 U.S.C. 101 have been fully considered and they are persuasive. The claims now include extracting three-dimensional seismic prisms at a specific well locations and include features such as comparing the models in parallel, which reduce processing time, and produce more accurate reservoir property predictions. Therefore, the 35 U.S.C. 101 rejections of claims 1, 11, and 12 have been withdrawn.
Applicant’s arguments, see Remarks, filed 02/27/2026, with respect to the rejection(s) of claims 1, 11, and 12 under 35 U.S.C. 103 have been fully considered and are persuasive in light of the amendments. The combination of Alwon (US 20190302290 A1), Xu (US 20230176242 A1), and Skripkin (US 20220114302 A1) does not explicitly teach “extracting, from the input dataset, three-dimensional seismic prisms from the seismic data at one or more log locations or one or more interpretation locations to generate spatial context to one or more well locations;” Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of Bø (US 20210247534 A1) as modified by Skripkin (US 20220114302 A1).
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.
Claims 1, 2, 5, 8-13, 15, and 17-19 are rejected under 35 U.S.C. 103 as being unpatentable over Bø (US 20210247534 A1) as modified by Skripkin (US 20220114302 A1).
Regarding claims 1, 11, and 12,
Bø teaches,
A computer-implemented method for generating a model of a subsurface reservoir, the method comprising: generating an input dataset comprising seismic data associated with a subsurface reservoir; (Para. [0044] teaches “In an example embodiment, the simulation component 120 may rely on entities 122. Entities 122 may include earth entities or geological objects such as wells, surfaces, bodies, reservoirs, etc. In the system 100, the entities 122 can include virtual representations of actual physical entities that are reconstructed for purposes of simulation. The entities 122 may include entities based on data acquired via sensing, observation, etc. (e.g., the seismic data 112 and other information 114).” (i.e. associated with a reservoir.) Para. [0277] teaches “As an example, one or more computer-readable storage media can include computer-executable instructions executable to instruct a computing system to perform a method, which may be, for example, a method as described herein.”)
extracting, from the input dataset, three-dimensional seismic prisms from the seismic data at one or more log locations or one or more interpretation locations to generate spatial context to one or more well locations; (Para. [0179-0180] teaches “a seismic interpretation system that included receiving data from a seismic cube S and a set of points P as input, representing positions in the seismic data where a surface is located. Given the seismic data and the points, the following method actions include: [0180] 1. Extraction of training data.” Para. [0044] teaches “Entities 122 may include earth entities or geological objects such as wells, surfaces, bodies, reservoirs, etc. In the system 100, the entities 122 can include virtual representations of actual physical entities that are reconstructed for purposes of simulation.” Para. [0003] teaches “analyzing at least a portion of the digital seismic data using the trained machine model to generate results; and outputting the results as indicators of spatial locations of the structural feature of the geologic region.”)
providing the three-dimensional seismic prisms as an input to a deep learning computing technique; (Para. [0160] teaches “such that a point can be registered with respect to seismic data (e.g., a seismic cube in three-dimensions) where that point can be associated, for example, with a trace.” Paras. [0179] teaches “Various trials utilized a method implemented via a seismic interpretation system that included receiving data from a seismic cube S and a set of points P as input, representing positions in the seismic data where a surface is located.” (i.e. A cube is a prism.)
training, based on the input dataset and utilizing a deep learning computing technique, a plurality of reservoir models; (Para. [0262] teaches “In such an example, multiple machine models may be trained and utilized where each of the multiple machine models is trained to a particular structural feature (e.g., a particular event).).” Para. [0213] teaches “As an example, the CAFFE framework may be implemented, which is a deep learning framework developed by Berkeley AI Research (BAIR) (University of California, Berkeley, Calif.” (i.e. plurality of deep learning models.))
the deep learning computing technique including one or more image recognition algorithms that correlate the three-dimensional seismic prisms with the plurality of reservoir models; Para. [0095] teaches “As an example, seismic data may be processed in a technique called “depth imaging” to form an image (e.g., a depth image) of reflection amplitudes in a depth domain for a particular target structure (e.g., a geologic subsurface region of interest).” Para. [0161] teaches “method can involve subsequent robust tracking to provide additional training data while generating a more robust machine learning model to recognize the specific surface to which the point corresponds (e.g., a surface of a geologic feature)” Para. [0190] teaches “As an example, extraction of training data can be performed by taking a sub-image around each point in P (e.g., seismic trace data around each point in P), which can be positive examples. In such an example, amplitude values can be extracted at subsample precision with interpolation, since surface points may not fall precisely on the positions of voxels in a seismic cube or pixels in a seismic section (e.g., arrays of seismic data, seismic images, seismic volumes, etc.).” Para. [0248] teaches “as an example, a neural network may be trained and have an architecture to provide one or more types of outputs for analyzing seismic data, which can be in digital form (e.g., as a vector of trace, as a 2D array of an image/slice, as a 3D array of a volume, etc.)”
executing a simulation on each of the plurality of reservoir models to generate an expected dataset for one or more holdout wells; (Para. [0163] teaches “Such a point or points and the seismic data provide a first basis for extracting training data for machine learning. As machine learning can involve both positive and negative examples, amplitude data surrounding points in example data can be extracted as positive examples. Negative examples may for instance be extracted around other locations than indicated by the input points.” Para. [0165] teaches “Given positive and negative example data, machine learning or statistical learning may be applied to learn a model M (e.g., a machine model) to predict whether an example seismic data profile is representing a surface of interest or not. In this context, profile can mean the surrounding data.”)
Bø does not explicitly teach,
determining, for each of the plurality of reservoir models, an accuracy score by comparing in parallel, the expected to a dataset of measured subsurface characteristics;
selecting, based on the accuracy score, an optimized reservoir model from the plurality of reservoir models, the optimized reservoir model having a lowest delta between the expected dataset and the dataset of measured subsurface characteristics;
and executing a prediction of one or more reservoir properties for the subsurface reservoir using the optimized reservoir model;
and transmitting, via a network, the prediction to a user device to cause the user device to output the prediction.
Nevertheless, Skripkin teaches,
` determining, for each of the plurality of reservoir models, an accuracy score by comparing in parallel, an expected dataset generated by each of the plurality of reservoir models, to a dataset of measured subsurface characteristics; Para. [0165] teaches “As mentioned, a method can include parallel processing where multiple simulations may be executed to generate results for a number of models, which may be assigned case index values.” (i.e. Simulations would include comparing expected values from the models to measured values.) Para. [0219] teaches “analyzing convergence of average value of correlation coefficients between predicted and actual values.” Para. [0200] teaches “a training block 1520 for training a suite of machine learning models using at least a portion of the model data (e.g., a training portion, a testing portion, etc.) to generate trained ML models (e.g., predictive ML models), a selection block 1530 for selecting one or more of the trained ML models (e.g., using one or more criteria, such as ability to train, test performance, etc.), a computation block 1540 for computing values of one or more defined accuracy metrics for the selected one or more of the trained ML models with respect to an index (e.g., to characterize the selected one or more models as to prediction accuracy, etc.),”(i.e. accuracy score) )
selecting, based on the accuracy score an optimized reservoir model from the plurality of reservoir models the optimized reservoir model having a lowest delta between the expected dataset and the dataset of measured subsurface characteristics; (Para. [0205] teaches “In such a method, there can be a suite of predictive models from which it is possible to select one predictive model that produces the most accurate predictions for a particular dataset.” Para. [0208] teaches “As explained with respect to the block 1540, a method can include defining an accuracy metric δ.sub.i(m) for each ensemble selected and computing values for the accuracy metric. In such an example, a method can include rendering a graph to a display for visualization by a user where the graph can include δ.sub.i(m) versus m for each reservoir output r.sub.i.” Para. [0247] teaches “FIG. 17 shows an example method 1700 that includes a determination block 1710 for determining one or more types of accurate predictive model, a sample analysis block 1720 for analyzing error associated with presence and/or absence of a sample, a determination block 1730 for determining a total information value, an identification and training block 1740 for identifying and training a most accurate predictive model,” A higher accuracy would indicate a lower delta value as seen in the Article Overfitting and Underfitting in Machine Learning by Rajbanshi (2020).)
and executing a prediction of one or more reservoir properties for the subsurface reservoir using the optimized reservoir model. (Para. [0132] teaches “As an example, a workflow can include a history period and a prediction period. For example, a simulator may be utilized in a history matching process to improve and/or to qualify a reservoir model using field data (e.g., field production data, etc.). In such an example, once an acceptable match is achieved, a prediction period may aim to predict future behavior of a reservoir. In some instances, a prediction may be rejected using one or more criteria (e.g., as deviating from historical data in an unlikely manner, etc.)” Para. [0253] teaches “As an example, the identification and training block 1740 can include identifying and training the most accurate predictive model μ”)
transmitting, via a network, the prediction to a user device to cause the user device to output the prediction. (Para. [0310] teaches “transmitting, via a network, the prediction to a user device to cause the user device to output the prediction.”
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Bø with determining, for each of the plurality of reservoir models, an accuracy score by comparing in parallel, the expected to a dataset of measured subsurface characteristics; selecting, based on the accuracy score, an optimized reservoir model from the plurality of reservoir models, the optimized reservoir model having a lowest delta between the expected dataset and the dataset of measured subsurface characteristics; and executing a prediction of one or more reservoir properties for the subsurface reservoir using the optimized reservoir model; and transmitting, via a network, the prediction to a user device to cause the user device to output the prediction such as that of Skripkin.
One of ordinary skill would have been motivated to modify Bø, because using a model that is the most accurate would lead to the most accurate predictions. Para. [0117] of Skripkin supports the importance of having an accurate model by including “As explained, reservoir flow simulation benefits from representing an underground environment accurately.” Also, a system that can perform parallel simulations reduces the runtime over systems that can’t as seen in Para. [0155] of Skripkin. Therefore, it would be obvious to combine the prior art in order to increase accuracy of the simulation and efficiency.
Regarding claims 2 and 13,
Bø further teaches,
wherein one or more image recognition algorithms include a three-dimensional image recognition technique. Para. [0190] teaches “As an example, extraction of training data can be performed by taking a sub-image around each point in P (e.g., seismic trace data around each point in P), which can be positive examples. In such an example, amplitude values can be extracted at subsample precision with interpolation, since surface points may not fall precisely on the positions of voxels in a seismic cube or pixels in a seismic section (e.g., arrays of seismic data, seismic images, seismic volumes, etc.).” Para. [0248] teaches “as an example, a neural network may be trained and have an architecture to provide one or more types of outputs for analyzing seismic data, which can be in digital form (e.g., as a vector of trace, as a 2D array of an image/slice, as a 3D array of a volume, etc.)”)
Regarding claims 5 and 15,
Bø further teaches,
further comprising: transmitting the plurality of reservoir models to a high-performance cluster of computing devices for training the plurality of reservoir models utilizing the deep learning computing technique. (Para. [0293] teaches “Such blocks generally include instructions suitable for execution by one or more processors (or cores) to instruct a computing device or system to perform one or more actions. As an example, a single medium may be configured with instructions to allow for, at least in part, performance of various actions of a workflow. As an example, a computer-readable medium (CRM) may be a computer-readable storage medium that is non-transitory, not a carrier wave and not a signal. As an example, blocks may be provided as one or more sets of instructions, for example, such as the one or more sets of instructions 270 of the system 250 of FIG. 2.”)
Regarding claims 8 and 17,
Bø further teaches,
displaying, on a user interface, a performance metric of the plurality of reservoir models. (Para. [0182] teaches “Output QC metrics concerning the accuracy of the model.” Para. [0298] teaches “information may be input from a display (e.g., consider a touchscreen), output to a display or both.”)
Regarding claims 9 and 18,
Bø further teaches,
receiving, via the user interface, a storage location of the input dataset. (Para. [0057] teaches “The model simulation layer 180 may be configured to model projects. As such, a particular project may be stored where stored project information may include inputs, models, results and cases. Thus, upon completion of a modeling session, a user may store a project. At a later time, the project can be accessed and restored using the model simulation layer 180, which can recreate instances of the relevant domain objects.” (i.e. If it accesses the data, it must receive a storage location for that data.)
Regarding claims 10 and 19,
Bø further teaches,
receiving, via the user interface, at least one of a training parameter, an optimizing parameter, or a prediction parameter. (Para. [0177] teaches “As indicated in the method 800, one parameter can be a score threshold. As an example, another parameter can be a window, for example, to be utilized for extracting training data based on one or more picked points. As yet another example, a parameter can pertain to “negative” examples, which may be to assure that a “negative” example is sufficient different from a “positive” example (e.g., of a picked point)”)
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Bø (US 20210247534 A1) and Skripkin (US 20220114302 A1) as applied to claim 1 above, and further in view of Daly (US 20230358917 A1).
Regarding claim 4,
Bø does not explicitly teach,
the method of claim 1,
further comprising: iteratively train the plurality of reservoir models by, for each of the plurality of reservoir models: generating, based on a corresponding reservoir model, a simulated; and generating, based on a comparison of the simulated dataset to the input dataset, a model error value.
Nevertheless, Daly teaches,
further comprising: iteratively train the plurality of reservoir models by, for each of the plurality of reservoir models: (Para. [0038] teaches “In block 107, the operations check whether the operations of 101 to 105 should be repeated for additional locations in the same reservoir. If so (e.g., for the case where sufficient training data has not yet been collected), the operations revert back to 101 to repeat the operations of 101 to 105 for additional locations in the same reservoir.”)
generating, based on a corresponding reservoir model, a simulated dataset; (Para. [0028] teaches “At a location where an estimate of the target variable is required, assumption observations of all the secondary variables for such location is obtained, and the external model can be used to make an estimate of the target variable at that location.”)
and generating, based on a comparison of the simulated dataset to the input dataset, a model error value. (Para. [0029] teaches “In cases where the ensemble of target variable predictions produced by the ensemble of machine learning models (e.g., decision trees) can be considered to give a good estimate of the conditional distribution at a target location—such as the Random Forest for example—the estimate of the conditional distribution at the target location can be used to provide at least one additional product selected from the group which includes: [0030] 1) an uncertainty estimate.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Bø and Skripkin further comprising: iteratively train the plurality of reservoir models by, for each of the plurality of reservoir models: generating, based on a corresponding reservoir model, an expected dataset; and generating, based on a comparison of the expected dataset to the input dataset, a model error value such as that of Daly.
One of ordinary skill would have been motivated to modify the combination of Bø and Skripkin, because according to Para. [0007] of Daly “Furthermore, the methods and systems can also calculate probabilistic results that demonstrate the likely uncertainty in the model(s) based on the quality/quantity of the initial input data. The methods and systems can be significantly quicker than traditional property modeling techniques due to the minimal user input required.” Para. [0008] teaches “The ensemble machine learning utilizes multiple machine learning models to obtain better predictive performance than could be obtained from any of the constituent learning models alone.” Therefore, modifying the combination of Bø and Skripkin with Daly would increase the speed, performance, and efficiency of the system.
Claims 6 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over Bø (US 20210247534 A1) and Skripkin (US 20220114302 A1) as applied to claims 1 and 12 above, and further in view of Zhang (US 11733414 B2).
Regarding claim 6,
Bø does not explicitly teach,
the method of claim 1,
wherein the input dataset comprises seismic data obtained from at least one of a far angle stack, a mid-angle stack, or a near angle stack.
Nevertheless, Zhang teaches,
wherein the input dataset comprises seismic data obtained from at least one of a far angle stack, a mid-angle stack, or a near angle stack. (Col. 10 Ln(s). [41-47] teach “The subsurface data may include a seismic angle stack change, a prestack seismic property change, a timeshift, a time strain, an amplitude-versus-offset (AVO) gather change, a compressional velocity change, a shear-wave velocity change, a density change, and/or a velocity ratio change.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Bø and Skripkin wherein the input dataset comprises seismic data obtained from at least one of a far angle stack, a mid-angle stack, or a near angle stack such as that of Zhang.
One of ordinary skill would have been motivated to modify the combination of Bø and Skripkin, because angle stack is a well-known measurement in the field of seismic data. According to MPEP 2143 an example of a rational that may support a conclusion of obviousness includes: “(E) "Obvious to try" – choosing from a finite number of identified, predictable solutions, with a reasonable expectation of success;”
Regarding claim 16,
Bø does not explicitly teach,
the system of claim 12,
wherein the input dataset comprises seismic data obtained from at least one of a far angle stack, a mid-angle stack, or a near angle stack.
Nevertheless, Zhang teaches,
wherein the input dataset comprises seismic data obtained from at least one of a far angle stack, a mid-angle stack, or a near angle stack. (Col. 10 Ln(s). [41-47] teach “The subsurface data may include a seismic angle stack change, a prestack seismic property change, a timeshift, a time strain, an amplitude-versus-offset (AVO) gather change, a compressional velocity change, a shear-wave velocity change, a density change, and/or a velocity ratio change.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Bø and Skripkin wherein the input dataset comprises seismic data obtained from at least one of a far angle stack, a mid-angle stack, or a near angle stack such as that of Zhang.
One of ordinary skill would have been motivated to modify the combination of Bø and Skripkin, because angle stack is a well-known measurement in the field of seismic data. According to MPEP 2143 an example of a rational that may support a conclusion of obviousness includes: “(E) "Obvious to try" – choosing from a finite number of identified, predictable solutions, with a reasonable expectation of success;”
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner
should be directed to JOSHUA L FORRISTALL whose telephone number is 703-756-4554. The examiner
can normally be reached Monday-Friday 8:30 AM- 5 PM.
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, Andrew Schechter can be reached on 571-272-2302. 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.
/JOSHUA L FORRISTALL/Examiner, Art Unit 2857
/ANDREW SCHECHTER/Supervisory Patent Examiner, Art Unit 2857
Prosecution Insights