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 .
DETAILED ACTION
This action is final.
This action is in response to the amendments filed on 04/13/2026.
Claims 1-20 are pending and have been considered.
Independent claims 1, 11, 18 have been amended. No claims have been canceled.
The 35 USC 112(a) rejection of claims 1-20 has been withdrawn in view of the amendments.
The 35 U.S.C. 101 rejection of claims 1-20 has been withdrawn in view of the amendments.
The rejection of claims 1-5, 7-15, 18-20 under U.S.C. 102 as being unpatentable over DEN et al (US 2020/0183047) has been withdrawn.
In view of the amendments, a new ground of rejection under U.S.C. 103 is made with respect to independent claims 1, 11, 18 . The newly added limitations require application of different prior art than that previously relied upon. Accordingly, the rejection is properly made final in accordance with MPEP 706.07(a). In addition, dependent claims 2-10, 12-17, 19-20 which depend from the newly amended claims 1, 11, 18 are now also rejected under 35 U.S.C. 103 based on the updated prior art. Because the basis and the combination of references has changed, these rejections are likewise new grounds of rejection and are properly made final in view of the amendments.
Claims 1-3, 5, 9-13, 15 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over DEN et al (US 2020/0183047) (“DEN”), in further view of Wu et al Semi-Supervised Learning for Seismic Impedance Inversion Using Generative Adversarial Networks, Remote Sensing, 2021 (“Wu”)
Claims 4, 14 are rejected under 35 U.S.C. 103 as being unpatentable over DEN in view of WU in further view of Li et al (US 2022/0099855). (“LI”)
Claims 6-8, 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over DEN in view of WU, in further view of SMI, Two methods for voxel detail enhancement PCGames '11: Proceedings of the 2nd International Workshop on Procedural Content Generation in Games Article No.: 6, Pages 1 – 4. (“SMI”)
Response to Amendments/Arguments
The Examiner thanks the Applicant for the Interview of March 10, 2026 and for the Amendments and Arguments filed on 04/13/2026 which have been considered and which help clarifying the claimed invention and advance prosecution.
Claims 1-20 are pending and have been considered. Independent claims 1, 11, 18 have been amended. No claims have been canceled. In view of the Amendments and Arguments the rejections under 35 USC 112(a) are withdrawn.
The rejections under 35 USC 101 are also withdrawn in view of the Amendments and Arguments. The claim continues to recite abstract ideas, in particular the selection of well drilling operations, however the additional elements in the amended limitations moved the claim towards eligibility. 8. In view of the amendments, a new ground of rejection under U.S.C. 103 is made with respect to independent claims 1, 11, 18 . The newly added limitations require application of different prior art than that previously relied upon. Accordingly, the rejection is properly made final in accordance with MPEP 706.07(a). In addition, dependent claims 2-10, 12-17, 19-20 which depend from the newly amended claims 1, 11, 18 are now also rejected under 35 U.S.C. 103 based on the updated prior art. Because the basis and the combination of references has changed, these rejections are likewise new grounds of rejection and are properly made final in view of the amendments.
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 difference 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 the invention was made.
The factual inquiries set forth in Graham v. John Deere Co., 383 U.S. 1, 148 USPQ 459 (1966), that are applied for establishing a background for determining obviousness under 35 U.S.C. 103(a) are summarized as follows:
i. Determining the scope and contents of the prior art.
ii. Ascertaining the differences between the prior art and the claims at issue.
iii. Resolving the level of ordinary skill in the pertinent art.
iv. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-3, 5, 9-13, 15, 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over DEN in view of WU
Claims 1, 11, 18 share similar limitations, with system claim 11 having one additional limitation. The analysis is performed on Claim 11. Since all limitations of Claims 1 and 18 are similar to limitations of Claim 11, Claim 1 and 18 are rejected under the same rationale presented for Claim 11.
Regarding Claim 11, DEN discloses a system comprising one or more processors and a non-transitory computer-readable storage medium coupled to the one or more processors and storing programming instructions for execution by the one or more processors, the programming instructions instructing the one or more processors: { [0103…FIG. 11 is a diagram of an exemplary computer system 1300 that may be utilized to implement methods described herein. A central processing unit (CPU) 1302 is coupled to system bus 1304. The CPU 1302 may be any general-purpose CPU… while only a single CPU 1302 is shown in FIG. 11, additional CPUs may be present. [0104] The computer system 1300 may also include computer components such as non-transitory, computer-readable media..} One or more processors are interpreted as one or more CPUs. In BRI a non-transitory computer-readable storage medium coupled to the one or more processors and storing programming instructions for execution by the one or more processors, the programming instructions instructing the one or more processors is any non-transitory computer readable media.
receiving an input dataset that represents partial spatial information of an area of interest within a subterranean region, the partial spatial information comprising photographic images of a geographic feature type forming a portion of a dataset at a particular location in the dataset; { [0031] A 3-D geologic model (particularly a model represented in image form) may be represented in volume elements (voxels), in a similar way that a photograph (or 2-D geologic model) is represented by picture elements (pixels).
“[0072] Referring to the figures, FIG. 3 is a flow diagram 300 for generating multiple geological models using machine learning at one or more stages of the life cycle of oil and gas field (e.g., exploration, development and production). For example, machine learning may be used in any one, any combination, or all of: the petroleum exploration stage; the development stage; or the production stage. Exploration may include any one, any combination, or all of: analysis of geological maps (to identify major sedimentary basins); aerial photography (identify promising landscape formations such as faults or anticlines); or survey methods (e.g., seismic, magnetic, electromagnetic, gravity, gravimetric). Similarly, additional data may be generated in each of the subsequent stages of exploration; development (e.g., new densely-acquired broadband 3D seismic, well logs) or production (e.g., 4D or time-lapse seismic for monitoring reservoir).”
{Fig. 3(310), [0073] “At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. The life cycle of the oil and gas field may include any one, any combination, or all of: exploration; development; or production. As discussed above, various types of geophysical data (e.g., seismic data), various geological concepts (e.g., reservoir geological concepts, EODs or other concepts derived from experience or from the data), a set of interpreted surfaces (e.g., horizons or faults) or zones (e.g., strata, anticline structure and reservoir section), and various reservoir stratigraphic configurations (e.g., lithofacies learned from the well logs) may be used. In some or all embodiments, all of the available conditioning data relevant to the reservoir (or the target subsurface area) may be the input to a previously trained generative model.”}
In broadest reasonable interpretation and in view of the specification, this refers to partial spatial information of an area (of interest) within a subterranean region (i.e. below Earth surface), that includes any-image-like representation produced from data by an imaging process, not limited to optical photography. In the context of the application, this includes images generated from seismic data that resemble photographs (e.g cross-sectional images. It covers visual renderings of subterranean structures produced by wave reflection/refraction data, even if no light-based photography is involved. Thus the interpretation extends to analogous pictorial depictions of data.
PNG
media_image1.png
306
509
media_image1.png
Greyscale
in broadest reasonable interpretation and in view of specification the dataset that represents partial spatial information of an area of interest, receiving is interpreted as be accessed. DEN’s disclosure includes the photograph-like representation, which is not tied with being optical photography, but other types, including for example seismic imaging or seismic photography.
This interpretation is also consistent with the use of the term in the industry, where people loosely call non-visible (i.e. non light-based) imaging “photography”, e.g. infrared photography, X-Ray photography”, which means capturing images of the underground features with waves other than visible light – and that includes, for example, seismic imaging (acoustic, not optic waves). Furthermore, petroleum exploration even uses the term “seismic photography”. A few examples from the many that can be found in an internet search: “I also had the opportunity to participate in new research work in the field of seismic photography.” https://www.kaust.edu.sa/en/news/alumni-focus--hassan-al-ismail , or, “With the aid of three-dimensional seismic photography, directional drilling… “ “https://www.heraldtribune.com/story/news/2005/03/20/drilling-for-oil-best-energy-option/28837608007/”
The Examiner interprets the limitation maps to various types of data that can be represented in a similar way that a photograph (or 2-D geologic model) is represented by picture elements (pixels). This includes optical photography and seismic, and other photographic-like image obtained at exploration, development or production, of a seismic .
of a geographic feature type forming a portion of a dataset at a particular location in the dataset {[0069] Such training will enable the generative network to learn reservoir features or patterns that correspond with the particular concept. In this way, the GAN may process different sections of the subsurface in order to analyze the potential universe of geological structures and how they comport with the given data}. In BRI a geographic feature type forming a portion of a dataset at a particular location of a dataset is interpreted as local data corresponding to a feature of the ground, and is mapped to reservoir feature or partterns, in a section of the subsurface.
providing the input dataset to a spatial context generator, { Fig.3 (310) [0073] At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed.} in broadest reasonable interpretation and in view of specification conditioning data available and for use as input which may be accessed, in interpreted as providing the input dataset, and generative network is interpreted as the spatial context generator.
wherein the spatial context generator comprises the simulated machine learning model trained to generate, based on the partial spatial information, contextual spatial information for the area of interest comprising the dataset and an interpretation of the photographic images, the interpretation comprising a label of the geographic feature type, { Fig. 3(310), Fig. 3(320), Fig. 3(310); [0073] At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. …various types of geophysical data (e.g., seismic data), various geological concepts … may be used; [0031] A 3-D geologic model (particularly a model represented in image form) may be represented in volume elements (voxels), in a similar way that a photograph (or 2-D geologic model) is represented by picture elements (pixels). [0074] At 320, machine learning is performed using the accessed data in order to train a machine learning model. At 330, one or more geological models for the respective stage of the life cycle are generated based on the machine learning model. [0100] When the generative model is introduced with the different types of surfaces, their unique labels may either be removed, maintained, or changed to provide additional context to the model} in broadest reasonable interpretation generative network is interpreted as the spatial context generator, seismic data as the partial spatial information, geological models consisting of seismic data, with labels to provide additional context, are interpreted as the contextual spatial information; photographic images interpreted as seismic images.
the contextual spatial information providing geomorphological information identified as missing from the partial spatial information; { [0100] When the generative model is introduced with the different types of surfaces, their unique labels may either be removed, maintained, or changed to provide additional context to the model} in broadest reasonable interpretation contextual spatial information is interpreted are the labels; “identified as missing” from the partial spatial information is interpreted as being additional – additional interpreted as they were not there before, hence they were missing, which in is consistent with the specification “[0030]… where the contextual spatial information provides additional information (i.e., context) missing from the partial spatial information” (words “missing” and “geomorphological” were lacking in the recitation above)
receiving from the spatial context generator, at least one output dataset associated with the area of interest {“[0100] When the generative model is introduced with the different types of surfaces, their unique labels may either be removed, maintained, or changed to provide additional context to the model”; “[0102] FIGS. 9 and 10 illustrate respective sets of the interpreted surfaces, horizon and fault surfaces and automatically-generated reservoir model using the generative networks trained with the SEAM Foothill geological data… The corresponding outputs of the generative model trained with the paired samples from the structural framework and its seismic image (FIGS. 9(b) and 11(a) respectively of Regone et al. 2017) are shown in the second column of FIGS. 9 and 10 (1150, 1250).”} In broadest reasonable interpretation and in view of the specification the generative model is interpreted as the spatial context generator, the labels as the contextual spatial information, the outputs (1150, 1250) as the outputs sets associated with the area of interest.
wherein each output dataset comprises the simulated contextual spatial information for the area of interest; the simulated contextual spatial information a mapping of geological and petrophysical features of the area of interest within the subterranean region; {[0016] FIG. 1 is a flow diagram from seismic to simulations for building reservoir models. [0062] Thus, in some implementations, machine learning generates one or more geological models, such as one or more reservoir models or one or more stratigraphic models that are consistent with applicable geological concepts and/or conditioning data (e.g., seismic and other available information useful to infer the plausible reservoir geology). In particular, machine learning may generate reservoir models (or interpret stratigraphy) that are automatically conditioned with any one, any combination, or all of: (1) seismic data; (2) interpreted surfaces; (3) geobodies; (4) petrophysical/rock physics models; (5) reservoir property models; (6) well log data; and (7) geological concepts; [0079] GANs include generative models that learn mapping from one or more inputs to an output (such as y, G: z.fwdarw.y where y is output (e.g., reservoir model)}In broadest reasonable interpretation, reservoir model built from simulations is interpreted as the simulated contextual spatial information, the output generated by machine model as the output dataset. [0032] Subsurface model is a model (or map) associated with the physical properties of the subsurface (e.g., geophysical or petrophysical models) } the simulated contextual spatial information comprising an identification of geo/petro-physical features is interpreted as the petrophysical subsurface model which is a numerical representation of the parameters for subsurface region.
selecting well drilling operations based on the geological and the petrophysical features of the at least one output dataset associated with the area of interest. { [0108] For instance, methods according to various embodiments may include managing hydrocarbons based at least in part upon the one or more generated geological models and data representations (e.g., seismic images, feature probability maps, feature objects, etc.) constructed according to the above-described methods. In particular, such methods may include drilling a well, and/or causing a well to be drilled, based at least in part upon the one or more generated geological models and data representations discussed herein (e.g., such that the well is located based at least in part upon a location determined from the models and/or data representations, which location may optionally be informed by other inputs, data, and/or analyses, as well) and further prospecting for and/or producing hydrocarbons using the well. [0032] Subsurface model is a model (or map) associated with the physical properties of the subsurface (e.g., geophysical or petrophysical models) } which selecting drilling operations based on the at least one output dataset associated with the area of interest, In BRI and in view of the specification (that recites “ drilling” only in “[Background, 0003] The images are interpreted. This interpretation…may include decisions about field development, such locations to drill future wells”, the limitation is interpreted as the interpretation method that helps make decisions about drilling taught by the underlined fragment above.
DEN further discloses [0040] Conditioning data refers a collection of data or dataset to constraint, infer or determine one or more reservoir or stratigraphic models. Conditioning data might include geophysical models, petrophysical models, seismic images (e.g., fully-stacked, partially-stacked or pre-stack migration images), well log data, [0043] Generative network model (also referred as a generative network to avoid the ambiguity with subsurface models) is an artificial network that seeks to learn/model the true distribution of a dataset giving it the ability to generate new outputs that fit the learned distribution data and reservoir structural framework.
This instant application uses the synthetic data obtained by the cGAN machine learning model to train a machine learning model. DEN discloses both types of learning models. DEN at lease implies that output of one can be used as input to the other and a person skilled in the are would find obvious to use the two.
DEN does not explicitly disclose, however WU discloses
performing a training for a machine learning model to generate based on partial spatial information, synthetic spatial data comprising simulated contextual spatial information, the training completing in response to determining that the synthetic spatial data is indistinguishable, by a discriminant model, from real spatial data;{ To alleviate this problem, we propose a semi-supervised learning workflow based on generative adversarial network (GAN) for acoustic impedance inversion. The workflow contains three networks: a generator, a discriminator and a forward model. The training of the generator and discriminator are guided by well logs and constrained by unlabeled data via the forward model. The benchmark models Marmousi2, SEAM anda field data are used to demonstrate the performance of our method.}
In addition, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to modify DEN to include elements of Wu. As mentioned DEN teaches all elements of the two machine learning models, the GAN and the one using the data, at least in different embodiments. A person ordinary trained in the art would be motivated to put the two together. Wu just provides this in a clear example. One would have been motivated to combine these to achieve training using the better and more data from cGAN. Examiner concludes the claimed subject matter is obvious over DEN/Wu.
Regarding Claims 2, 12, 19 – DEN/WU discloses the limitations of Claims 1, 11, 18. DEN also discloses:
wherein the machine learning model is a conditional Generative Adversarial Network (cGAN) {[0077] As discussed above, various machine learning methodologies are contemplated. As one example, a generative adversarial network (GAN) may be used, such as illustrated in FIGS. 6A-B. In this regard, any discussion regarding the application of GAN to generate and/or evaluate geological models may likewise be applied to other machine learning methodologies. [0078] Specifically, FIG. 6A is a first example block diagram 600 of a conditional generative-adversarial neural network}
Regarding Claims 3, 13, 20 – DEN/WU discloses the limitations of Claims 1, 11, 18. DEN also discloses:
herein the input dataset is a seismic dataset that represents the partial spatial information of the area of interest. {Fig. 3(310), [0073] “At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. The life cycle of the oil and gas field may include any one, any combination, or all of: exploration; development; or production. As discussed above, various types of geophysical data (e.g., seismic data), various geological concepts (e.g., reservoir geological concepts, EODs or other concepts derived from experience or from the data), a set of interpreted surfaces (e.g., horizons or faults) or zones (e.g., strata, anticline structure and reservoir section), and various reservoir stratigraphic configurations (e.g., lithofacies learned from the well logs) may be used. In some or all embodiments, all of the available conditioning data relevant to the reservoir (or the target subsurface area) may be the input to a previously trained generative model.”} in broadest reasonable interpretation and in view of specification seismic data is interpreted as the dataset that represents partial spatial information of an area of interest.
Regarding Claims 5, 15 – DEN/WU discloses the limitations of Claims 1 to which 5 depends, and 11, to which 15 depends on. DEN also discloses:
wherein the input dataset is a photographic image dataset that represents the partial spatial information of the area of interest. { [0031] A 3-D geologic model (particularly a model represented in image form) may be represented in volume elements (voxels), in a similar way that a photograph (or 2-D geologic model) is represented by picture elements (pixels); [0072] Referring to the figures, FIG. 3 is a flow diagram 300 for generating multiple geological models using machine learning at one or more stages of the life cycle of oil and gas field (e.g., exploration, development and production). For example, machine learning may be used in any one, any combination, or all of: the petroleum exploration stage; the development stage; or the production stage. Exploration may include any one, any combination, or all of: analysis of geological maps (to identify major sedimentary basins); aerial photography (identify promising landscape formations such as faults or anticlines); or survey methods (e.g., seismic, magnetic, electromagnetic, gravity, gravimetric). Similarly, additional data may be generated in each of the subsequent stages of exploration; development (e.g., new densely-acquired broadband 3D seismic, well logs) or production (e.g., 4D or time-lapse seismic for monitoring reservoir).”
{Fig. 3(310), [0073] “At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. The life cycle of the oil and gas field may include any one, any combination, or all of: exploration; development; or production. As discussed above, various types of geophysical data (e.g., seismic data), various geological concepts (e.g., reservoir geological concepts, EODs or other concepts derived from experience or from the data), a set of interpreted surfaces (e.g., horizons or faults) or zones (e.g., strata, anticline structure and reservoir section), and various reservoir stratigraphic configurations (e.g., lithofacies learned from the well logs) may be used. In some or all embodiments, all of the available conditioning data relevant to the reservoir (or the target subsurface area) may be the input to a previously trained generative model.”}
In broadest reasonable interpretation and in view of the specification, photographic image dataset that represents the partial spatial information of the area of interest this refers to partial spatial information of an area (of interest) that includes any-image-like representation produced from data by an imaging process, not limited to optical photography. In the context of the application, this includes images generated from seismic data that resemble photographs (e.g cross-sectional images). Thus, it covers visual renderings of subterranean structures produced by wave reflection/refraction data, even if no light-based photography is involved. Thus the BRI is analogous pictorial depictions of data.
PNG
media_image1.png
306
509
media_image1.png
Greyscale
Receiving is interpreted as be accessed. DEN’s disclosure includes the photograph-like representation, which is tied with a number of image sensing modalities, including for example seismic imaging or seismic photography. Thus, Examiner interprets the limitation maps to various types of data that can be represented in a similar way that a photograph (or 2-D geologic model) is represented by picture elements (pixels), and which includes optical photography and seismic, and other photographic-like image obtained at exploration, development or production, of a seismic .
Regarding Claim 9. – DEN/WU discloses the limitations of Claims 1 to which 9 depends. DEN also discloses:
generating a model of the area of interest based on the at least one second seismic dataset.
{[0076] Thereafter, responsive to obtaining additional data responsive to reservoir development, an updated set of applicable conditioning data (e.g., second stage data) may be used in addition to the available prior conditioning data from exploration stage by the machine learning methodology in order to generate the geological models; [0073] “At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. The life cycle of the oil and gas field may include any one, any combination, or all of: exploration; development; or production. As discussed above, various types of geophysical data (e.g., seismic data); [0073] In some or all embodiments, all of the available conditioning data relevant to the reservoir (or the target subsurface area) may be the input to a previously trained generative model to generate one or more geological models in the respective stage.}. Generating the geological model interpreted as generating a model, target area is interpreted as the area of interest,
an updated set of conditioning data (second stage data), which is seismic, interpreted as based on at least a second seismic data.
Regarding Claim 10 – DEN/WU discloses the limitations of Claims 1. DEN also discloses:
wherein the area of interest is at least one of a surface or subsurface. {[Abstract] machine learning may be used to generate one or more reservoir models that characterize the subsurface. [0007] (i) seismic data 110 is processed to generate a geophysical model 120, which may define one or more geophysical properties (e.g., compressional and shear wave velocities, density, anisotropy and attenuation) of the subsurface. [Claim 10]. The method of claim 9, wherein the one or more input geological models of the subsurface comprise simulated reservoir models of the subsurface. [0073] In some or all embodiments, all of the available conditioning data relevant to the reservoir (or the target subsurface area) may be the input to a previously trained generative model to generate one or more geological models in the respective stage.} Target area is interpreted as the area of interest, which is of the subsurface.
Claims 4, 14 are rejected under 35 U.S.C. 103 as being unpatentable over DEN/WU in further view of LI
Regarding Claims 4, 14 – DEN/WU discloses the limitations of Claims 1 to which 4 depends, and 11, to which 16 depends on. DEN does not disclose, however Li discloses:
wherein the input dataset is an input seismic cube that has a first dimension {[0186] FIG. 11 shows an input block 1110 for receipt of seismic data as a “cube” with appropriate dimensions in x, y and z; noting that they need not be equal (e.g., the term “cube” as applied to seismic data is to mean volumetric and not necessarily of uniform x, y and z dimensions). While the seismic data may be “raw”, it may also be or include seismic data subjected to some amount of processing such as, for example, seismic attribute processing, filtering, normalization, etc.}
wherein each output dataset is an output seismic cube that has a second dimension larger than the first dimension. {[0186] FIG. 11 shows an input block 1110 for receipt of seismic data as a “cube” with appropriate dimensions in x, y and z; noting that they need not be equal (e.g., the term “cube” as applied to seismic data is to mean volumetric and not necessarily of uniform x, y and z dimensions). While the seismic data may be “raw”, it may also be or include seismic data subjected to some amount of processing such as, for example, seismic attribute processing, filtering, normalization, etc.; [0231] As mentioned, a method can include generating a series of outputs of 2D stratigraphic units based on a slice of seismic image data from a seismic cube. In such an example, the method can include interpolating between the series of 2D stratigraphic units to generate a 3D model of stratigraphic units. As to interpolation, linear and/or nonlinear approaches may be implemented. As an example, a spline fitting approach may be implemented where constraints may be imposed, for example, based on output from a slice that may be orthogonal to the series of 2D stratigraphic units. As an example, a method can include generating a series of 2D stratigraphic units along a first dimension and generating a series of 2D stratigraphic units along a second dimension, which may be orthogonal to the first dimension. In such an example, a 3D model of stratigraphic units may be built using the two series (e.g., or more series), optionally using interpolation.}
In addition, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to modify to include elements of Li. DEN/WU teaches taking seismic data as an example of spatial data from the subsurface, to be used as input for machine learning training. DEN/WU uses the training to obtain mode context for the model. One would have been motivated to use a volumetric data with cube shape (rectangular cross-sections more precisely) since it offers advantages in computer processing, for example in structured data storage, as cube-based (or rectangular grid) aligns well with memory structures and processing algorithms used in computing; moreover, it is efficient for matrix-mased numerical solvers, allows fast spatial interpolation, and has consistent resolution in all directions. The advantage of having one additional dimension for the outputs is that it allows locations where to add the context which is what is aimed for. One would have had a reasonable expectation of success, thus predictable results, since it is a most suitable form of data format for computing and data storage. Furthermore, both prior art elements, of DEN/WU and Li are in the same or similar context of obtaining seismic data for training machine learning models for interpretation. Examiner concludes the claimed subject matter is obvious over DEN/WU/Li.
Claims 6-8, 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over DEN/WU in further view of SMI
Regarding Claims 6, 16 – DEN/WU discloses the limitations of Claims 1 to which 4 depends, and 11, to which 16 depends on. DEN/WU does not disclose, however SMI discloses:
wherein the partial spatial information comprises a single 16x16x16 seismic volume and the simulated contextual spatial information for the area of interest comprises one or more 256x256x256 seismic volumes. {Figure 1. Enhancing a 16x16x16 input voxel map (left) to produce a detailed, 256x256x256 output fragment map (right) }
In addition, it would have been obvious for one of ordinary skill in the art before the effective filing date of the claimed invention to modify DEN/WU to include elements of SMI. DEN/WU teaches three-dimensional (geophysical) augmentation and generating such simulated contextual spatial information. SMI is relied upon to provide dimensions for the volumetric data, One would have been motivated to use cube sizes of lateral dimension 16 (2^4) and 256 (2^8) since these are binary friendly dimensions that align with radix-2 algorithms (e.g. FFT) and memory addressing, yielding predictable speed-ups and simple indexing. A POSITA would naturally select powers of two for block and volume sizes to optimize performance. Similarly POSITA would chose 16^3 tiles since it is a cash-friendly working set for CPUs/GPU fitting common shared memory strategies. One would have had a reasonable expectation of success, thus predictable results, since it is a most suitable form of data format for computing and data storage. Furthermore, both prior art elements, of DEN/WU and SMI are deal with methods of image enhancement/reconstruction. Examiner concludes the claimed subject matter is obvious over DEN/WU/SMI.
Regarding Claims 7, 17 – DEN/WU, SMI discloses the limitations of Claims 6 to which Claims 7 depends on, respectively 16 to which 17 depends on. DEN also discloses:
wherein the machine learning model is a conditional Generative Adversarial Network (cGAN) {Claim 5. The method of claim 4, wherein the machine learning model comprises a generative adversarial network (GAN) including a generator and a discriminator. [0021] FIG. 6A is a first example block diagram of a conditional generative-adversarial neural network (CGAN) }
wherein training the machine learning model comprises: {Claim 5. The method of claim 4, wherein the machine learning model comprises a generative adversarial network (GAN) ; [0069] For example, during GAN training… }
training a generator network of the cGAN to generate the contextual spatial information for the area of interest based on the partial spatial information { [0049] Generative Adversarial Network (GAN) is an artificial network system including generator (or interpreter) and discriminator network used for training the generative network model; [0021] FIG. 6A is a first example block diagram of a conditional generative-adversarial neural network (CGAN); [0069] For example, during GAN training, a section from the mask volume may be extracted. There may be multiple potential concepts (e.g., different potential geological templates) associated with the extracted section.; Fig. 3(310), Fig. 3(320), Fig. 3(310); [0073] At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. …various types of geophysical data (e.g., seismic data), various geological concepts … may be used; [0074] At 320, machine learning is performed using the accessed data in order to train a machine learning model. At 330, one or more geological models for the respective stage of the life cycle are generated based on the machine learning model. [0100] When the generative model is introduced with the different types of surfaces, their unique labels may either be removed, maintained, or changed to provide additional context to the model} in broadest reasonable interpretation generative network is interpreted as the spatial context generator, seismic data as the partial spatial information, geological models consisting of seismic data, with labels to provide additional context, are interpreted as the contextual spatial information.
wherein the generator network is trained based on feedback received from a discriminator network of the cGAN, {Claim 5. The method of claim 4, wherein the machine learning model comprises a generative adversarial network (GAN) including a generator and a discriminator. [0046] Training (machine learning) is typically an iterative process of adjusting the parameters of a neural network to minimize a loss function which may be based on an analytical function (e.g., binary cross entropy) or based on a neural network (e.g., discriminator); [0078] Specifically, FIG. 6A is a first example block diagram 600 of a conditional generative-adversarial neural network}} In broadest reasonable interpretation the generator network is the generator network and what the discriminator neural network provides to minimize the loss function, which is part of the training, is the feedback from the discriminator network. The discriminator network is a component of the GAN, which can be a cGAN.
wherein the discriminator network is configured to distinguish between real data and simulated data generated by the generator network. {[0083] The generative model G may be trained iteratively by solving an optimization problem which may be based on an objective functional involving discriminator D and a measure of reconstruction loss (e.g., an indication of the similarity of the generated data to the ground truth) and/or adversarial loss (e.g., loss related to discriminator being able to discern the difference between the generated data and ground truth).} Ground truth is interpreted to be real data, generated data be simulated data generated by generator network.
Regarding Claims 8 – DEN/WU discloses the limitations of Claim 7 to which Claim 8 depends on. DEN also discloses:
wherein the real data and the simulated data are conditioned on training partial spatial information. { {[0083] The generative model G may be trained iteratively by solving an optimization problem which may be based on an objective functional involving discriminator D and a measure of reconstruction loss (e.g., an indication of the similarity of the generated data to the ground truth) and/or adversarial loss (e.g., loss related to discriminator being able to discern the difference between the generated data and ground truth); [0073] At 310, various conditioning data, available for a respective stage of the life cycle of an oil and gas field and for use as input to the generative network, may be accessed. The life cycle of the oil and gas field may include any one, any combination, or all of: exploration; development; or production. As discussed above, various types of geophysical data (e.g., seismic data)…}. Ground truth is interpreted to be real data, generated data be simulated data generated by generator network. Conditioning data used in training, which is seismic data as the training partial spatial information.
Prior art made of record
The prior art made of record and not relied upon which, however, is considered pertinent to applicant's disclosure:
Wei et al. US 2021/0293983 FACILITATING HYDROCARBON EXPLORATION AND EXTRACTION BY APPLYING A MACHINE-LEARNING MODEL TO SEISMIC DATA
Abstract: Hydrocarbon exploration and extraction can be facilitated using machine-learning models. For example, a system described herein can receive seismic data indicating locations of geological bodies in a target area of a subterranean formation. The system can provide the seismic data as input to a trained machine-learning model for determining whether the target area of the subterranean formation includes one or more types of geological bodies. The system can receive an output from the trained machine-learning model indicating whether or not the target area of the subterranean formation includes the one or more types of geological bodies. The system can then execute one or more processing operations for facilitating hydrocarbon exploration or extraction based on the seismic data and the output from the trained machine-learning model.
Lai, S-H et al, US 20190236759 A1, METHOD OF IMAGE COMPLETION
Abstract: A method of image completion comprises: constructing the image repair model and constructing a plurality of conditional generative adversarial networks according to a plurality of object types; inputting the training image corresponding to the plurality of objective types such that the plurality of conditional generative adversarial networks respectively conduct a corruption feature training; inputting the image in need of repair and respectively conducting an image repair through the plurality of conditional generative adversarial networks to generate a plurality of repaired images; and judging a reasonable probability of the plurality of repaired images through a probability analyzer, choosing an accomplished image and outputting the accomplished image through an output interface.
Baumstein, A. et al 20210318458 A1 Methodology For Enhancing Properties Of Geophysical Data With Deep Learning Networks
Abstract: A method for enhancing properties of geophysical data with deep learning networks. Geophysical data may be acquired by positioning a source of sound waves at a chosen shot location, and measuring back-scattered energy generated by the source using receivers placed at selected locations. For example, seismic data may be collected using towed streamer acquisition in order to derive subsurface properties or to form images of the subsurface. However, towed streamer data may be deficient in one or more properties (e.g., at low frequencies). To compensate for the deficiencies, another survey (such as an Ocean Bottom Nodes (OBN) survey) may be sparsely acquired in order to train a neural network. The trained neural network may then be used to compensate for the towed streamer deficient properties, such as by using the trained neural network to extend the towed streamer data to the low frequencies.
Liu, W.D. et al US 11397272 B2 Data Augmentation For Seismic Interpretation Systems And Methods
Abstract: A method and apparatus for machine learning for use with automated seismic interpretation include: obtaining input data; extracting patches from a pre-extraction dataset based on the input data; transforming data of a pre-transformation dataset based on the input data and geologic domain knowledge and/or geophysical domain knowledge; and generating augmented data from the extracted patches and the transformed data. A method and apparatus for machine learning for use with automated seismic interpretation include: a data input module configured to obtain input data; a patch extraction module configured to extract patches from a pre-extraction dataset that is based on the input data; a data transformation module configured to transform data from a pre-transformation dataset that is based on the input data and geologic domain knowledge and/or geophysical domain knowledge; and a data augmentation module configured to augment data from the extracted patches and the transformed data.
Fuchey, Y. et al US 20240077642 A1 GEOLOGIC ANALOGUE SEARCH FRAMEWORK
Abstract: A method can include, responsive to receipt of a search instruction that includes one or more search criteria, accessing a data structure for subsurface geologic regions categorized at least in part according to parameters that describe depositional environments, where the data structure includes one or more includes virtual distances between the parameters; generating a search result using the one or more search criteria and the data structure, where the search result represents an organization of at least a portion of subsurface geologic regions as closest analogues to the one or more search criteria; and transmitting search result information for graphically rendering the search result to a display as part of an interactive graphical user interface.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). 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.
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 ADRIAN STOICA whose telephone number is (571) 272-3428. The examiner can normally be reached Monday to Friday, 9 a.m. -5 p.m. PT.
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, Ryan Pitaro can be reached on (571) 272-4071. 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.
/A.S./Examiner, Art Unit 2188
/RYAN F PITARO/Supervisory Patent Examiner, Art Unit 2188
Prosecution Insights