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 .
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-15 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (abstract idea) without significantly more.
Under Step 1 of the 2019 Revised Patent Subject Matter Eligibility Guidance, the claims are directed to a process (claim 1, a method) or a machine (claim 9, an apparatus), which are statutory categories.
However, evaluating claim 1, under Step 2A, Prong One, the claim is directed
to the judicial exception of an abstract idea using the grouping of a mathematical relationship/mental process. The limitations include:
generating a domain adaptation model based on a source data associated with a source area that includes a well and a target area that does not include a well, and predicting a P-impedance value of the target area; and generating, using the P-impedance value generated by the domain adaptation model, a P-impedance low frequency model configured to predict a final P-impedance value of the target area by performing an inversion.
These limitations describe mathematical concepts, including mathematical modeling, statistical learning, and inversion calculations. Such operations fall within the abstract idea grouping of mathematical relationships., mathematical formulas, and mathematical calculations (see MPEP § 2106.04 (a)(2). The steps amount to receiving data, applying mathematical algorithms to analyze the data, and outputting predicted values, which constitute abstract data processing. Accordingly, the claim recites a judicial exception.
Next, Step 2A, Prong Two evaluates whether additional elements of the claim “integrate the abstract idea into a practical application” in a manner that imposes a meaningful limit on the judicial exception, such that the claim is more than a drafting effort designed to monopolize the exception. The claim does not recite additional elements that integrate the judicial exception into a practical application.
Although the claim recites “seismic data”, “source data”, “source area that includes a well”, and “target area that does not include ”, these limitations merely apply the abstract mathematical processes to a particular technological field (geophysical analysis), which constitutes a field-of-use limitation. The claim does not recite any specific operation performed on the well, any control of drilling equipment , any modification of a wellbore or geological formation, or any improvement to seismic acquisition technology. Rather, the well is only referenced as characteristic of the dataset used to train or apply the machine learning model. Thus, the recited well limitation merely limits the abstract idea to a particular field of use (geophysical analysis involving well-related data) and does not impose a meaningful limit on the judicial exception. The remaining steps are performed generic computing technology to execute mathematical models on seismic data.
Accordingly, the additional elements do not integrate the judicial exception into a practical application.
At Step 2B, consideration is given to additional elements that may make the abstract idea significantly more. Under Step 2B, there are no additional elements that make the claim significantly more than the abstract idea.
The additional elements of “machine learning model”, “domain adaptation”, and “inversion process” are described at a high level of generality and are implemented using conventional data processing techniques. There is no recitation of a specific unconventional algorithmic structure that improves computer functionality itself, nor any specialized hardware. The claimed method merely applies known mathematical and machine learning techniques to seismic data to obtain predicted impedance values.
Accordingly, the limitations have been considered individually and as a whole and do not amount to significantly more than the abstract idea itself and therefore claim 1 is not patent eligible under 35 U.S.C. § 101.
Dependent claims 2-8 do not add anything which would render the claimed invention a patent eligible application of the abstract idea. The additional limitations recited in claims 2-8, including extracting feature information from seismic data, generating first and second loss function values retaining a feature extraction algorithm to minimize a combined loss, performing smoothing and filtering operations to generate low-frequency models, extending P-impedance values to S-impedance and density values using relational expressions derived from well log data, and performing simultaneous inversion, constitute further mathematical concepts and data processing operations. These limitations merely add additional mathematical modeling, statistical learning techniques, and signal processing steps to the abstract idea identified in claim 1. The claims do not recite any specific improvement to computer functionality, any specialized or unconventional hardware, any transformation of matter, or any integration of mathematical concepts into a practical application beyond the field of seismic data analysis. Rather, the additional elements refine the abstract algorithm and limit it to particular mathematical techniques and data relationships, which does not amount to significantly more than the judicial exception.
Accordingly, claims 2-8 are not patent eligible under 35 U.S.C. § 101.
Claim 9 is rejected 35 USC § 101 for the same rationale as in claim
1.
The additional elements of “a processor” and “storage unit” are recited at a high level of generality and are recited as performing generic computer functions routinely used in computer applications. Generic computer components recited as performing generic computer functions that are well-understood, routine and conventional activities amount to no more than implementing the abstract idea with a computerized system (Alice Corp. Pty. Ltd. v. CLS Bank Int’l 573 U.S. __, 134 S. Ct. 2347, 110 U.S.P.Q.2d 1976 (2014)).
The examiner notes that the element “training module configured to generate a domain adaptation model for predicting a P-impedance value of a target area that does not include a well by using source data obtained from a source area that includes a well and target data obtained from the target area” is considered performing mathematical calculation which falls within the “mathematical concept” grouping of abstract ideas (see Example 47, in the 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence).
This judicial exception is not integrated into a practical application because the remaining elements amount to no more than general purpose computer components programmed to perform the abstract ideas. As set forth in the 2019 Eligibility Guidance, 84 Fed. Reg. at 55 “merely include[ing] instructions to implement an abstract idea on a computer” is an example of when an abstract idea has not been integrated into a practical application.
Accordingly, the limitations have been considered individually and as a whole and do not amount to significantly more than the abstract idea itself and therefore claim 9 is not patent eligible under 35 U.S.C. § 101.
Dependent claims 10-15 do not add anything which would render the claimed invention a patent eligible application of the abstract idea. The additional limitations recited in claims 10-15, including extracting feature information from seismic data, generating first and second loss function values retaining a feature extraction algorithm to minimize a combined loss, performing smoothing and filtering operations to generate low-frequency models, extending P-impedance values to S-impedance and density values using relational expressions derived from well log data, and performing simultaneous inversion, constitute further mathematical concepts and data processing operations. These limitations merely add additional mathematical modeling, statistical learning techniques, and signal processing steps to the abstract idea identified in claim 9. The claims do not recite any specific improvement to computer functionality, any specialized or unconventional hardware, any transformation of matter, or any integration of mathematical concepts into a practical application beyond the field of seismic data analysis. Rather, the additional elements refine the abstract algorithm and limit it to particular mathematical techniques and data relationships, which does not amount to significantly more than the judicial exception.
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, 3-6, 9, and 11-13 are rejected under 35 U.S.C. 103 as being
unpatentable over Meek (Pub. No. US 2019/0293818) in view of Yoo et al. (NPL: “Domain adaptation-based acoustic impedance estimation with seismic data constraint”, Geoscience (15 November 2021) (hereinafter Yoo).
As per claims 1 and 9, Meek teaches generating P-impedance (i.e., acoustic impedance) value, a P-impedance low frequency model configured to predict a final P-impedance value of the target area by performing an inversion and “generating acoustic impedance volumes” (which in seismic terminology corresponds to P-impedance) (see ¶¶ [0047]-[0053] and [0057]-[0061], where Meek discloses a machine learning-assisted seismic inversion workflow including generating acoustic impedance (i.e., P-impedance, P-wave impedance)) volumes and creating low-frequency background models used in performing inversion).
However, Meek fails to explicitly teach generating a domain adaptation model based on a source data associated with a source area that includes a well and a target area that does not include a well and predicting a P-impedance value of the target area.
Yoo, however, explicitly teaches Domain adaptation is a ML method that can be applied not only to the source domain (domain with labeled data) but also to the target domain (domain without labeled data). Therefore, we have applied the domain adaptation technique to predict the acoustic impedance more accurately in the area not only around the well but also away from the well. Furthermore, to generalize the ML model, the reconstruction process of seismic data was added as a constraint and a pseudo labeling strategy was applied” and that the “predicts the acoustic impedance of areas around the well and areas far from the wells, using domain adaptation” (see Abstract). Thus, Yoo teaches generating impedance predictions for a target area without labeled well data using a source domain that includes labeled well log data. It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to incorporate the domain adaptation impedance prediction of Yoo into the inversion workflow of Meek because doing so would enable estimation of P-impedance in well target areas, thereby permitting construction of low-frequency model for use in the inversion framework of Meek.
As per claim 3, the combination of Meek and of Yoo teaches the system as stated
Above. Meek further teaches preprocessing operation by generating a P-impedance low frequency model an inversion operation by predicting the final P-impedance value using the P-impedance low frequency model and the seismic data (see ¶¶ [0009]-[0012], describing creation of a low-frequency background model and application of pre-stack inversion using the model and seismic records).
However, Meek fails to explicitly disclose generating the P-impedance values used to construct the low-frequency model using machine-learning domain-adaptation model trained using source and target seismic data.
Yoo, however, teaches predicting acoustic impedance using a machine domain-adaptation model that uses labeled source data associated with wells and unlabeled target seismic data from areas without wells, explaining that domain adaptation enables prediction of acoustic impedance “in areas far from the wells” where labeled well-log data are unavailable (see Abstract). . It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to incorporate the domain adaptation impedance prediction of Yoo into the inversion workflow of Meek because Yoo identifies the problem that impedance prediction models trained using well-log data cannot reliably predict impedance in regions lacking wells, and Meek teaches constructing low-frequency impedance models as inputs to seismic inversion. Combining the techniques would allow the predicted impedance values generated by Yoo to be used to construct the low-frequency background model used in the inversion workflow of Meek, thereby enabling seismic inversion to be performed in regions lacking well control while utilizing the standard low-frequency model-based inversion framework taught by Meek.
As per claims 4 and 11, the combination of Meek and of Yoo teaches the system
as stated above. Meek further teaches performing a smoothing operation by obtaining a smoothed P-impedance value by smoothing the P-impedance value and performing a filtering operation by obtaining the P-impedance low frequency model by applying a filter to the smoothed P-impedance value (see ¶¶ [0009]-[0012]). In particular, Meek describes creating a low-frequency background model by applying bandpass filtering to impedance or well-log data and using the resulting model as input to seismic inversion. Thus, Meek teaches performing a preprocessing operation including smoothing a P-impedance value and applying a filter to obtain a P-impedance low-frequency model.
As per claims 5, 6, 12 and 13, the combination of Meek and of Yoo teaches the
system as stated above. Meek further teaches performing an extension operation by obtaining a smoothed S-impedance value using the smoothed P-impedance value, wherein the filtering operation further includes obtaining an S-impedance low frequency model by applying a filter to the smoothed S-impedance value (see ¶ [0010], describing volumes including acoustic impedance, shear impedance, density, P-wave velocity, and S-wave velocity). Such inversion workflows involve generating low-frequency models for the relevant parameters and performing simultaneous inversion to estimate these parameters from seismic data. Thus, Meek teaches generating an S-impedance model corresponding to acoustic impedance and performing inversion to predict both P-impedance and S-impedance values using seismic data and low-frequency models.
Yoo teaches predicting impedance values in regions without wells using domain adaptation (see Abstract). It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to use the predicted impedance values of Yoo as input impedance values used in the inversion workflow of Meek because Yoo addresses prediction of impedance in areas lacking well logs, while Meek teaches simultaneous inversion of multiple impedance parameters using low-frequency models, thereby enabling estimation of both P-impedance and S-impedance values for target regions lacking well control.
The examiner notes that the P-impedance (i.e., acoustic impedance), P-wave velocity, and density (of the rock) are directly related : ZP = ρ VP. Therefore, claims 6 and 13 is also included in this rejection.
Claims 2 and 10 are rejected under 35 U.S.C. 103 as being unpatentable over
Meek in view of Yoo and further in view of Ganin et al. (NPL: “Domain-Adversarial Training of Neural Networks” (Journal of Machine Learning Research 17 (01 June 2016)) (hereinafter Ganin).
As per claim 2, the combination of Meek and of Yoo teaches the system as stated
above specifically, Meek teaches generating acoustic impedance values from seismic data using machine-learning assisted inversion workflow (see ¶ [0053], describing generating of acoustic impedance volumes derived from seismic data). Yoo further teaches applying domain adaptation to seismic impedance inversion so that a model trained using labeled well-log data in a source domain containing wells can predict acoustic impedance in a target domain without wells. In particular, Yoo explains that domain adaptation allows a model trained using labeled data in a source domain to be applied to a target domain without labels and used to predict acoustic impedance in areas far from wells. Thus, the combination of Meek and of Yoo teaches extracting information from seismic data associated with a source area containing wells and seismic data associated with a target area lacking wells and predicting impedance values for the target area.
However, Meek and of Yoo do not explicitly disclose determining whether extracted feature information corresponds to the source or target domain, generating a first loss function based on that determination, and retraining a feature extraction algorithm based on both that loss and a label-prediction loss as recited in claim 2.
Ganin, however, teaches a neural network architecture including a feature extractor, a label predictor, and a domain classifier, where the domain classifier predicts where extracted features originate from a source or target domain and where the feature extractor is trained to minimize label prediction loss while maximizing domain classification loss (i.e., the training process involves optimizing both a label prediction loss and a domain classification loss). Ganin explains that the feature extractor produces feature representations that are optimized using both losses during training so that the model learns features useful for prediction while accounting for domain differences (see Section 1, Introduction, page 2, last paragraph and page 3, paragraphs 2 and 4, describing a feature extractor, label predictor, domain classifier, and associated loss functions) and Ganin further explains that the network parameters are trained by optimizing a loss function during iterative training until the loss converges (see Section 4.1-4.2, pages 8-11, equations (5) and (10), which inherently involves stopping training when the loss reaches a minimum or convergence threshold (claim 10). It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to incorporate the feature-extraction and dual-loss training framework described in Ganin into the seismic domain-adaptation impedance inversion method taught by Yoo and implemented in the seismic inversion workflow of Meek because Yoo explicitly identifies the problem that machine-learning impedance inversion models trained using well-log labels in a source region cannot reliably predict impedance in regions far from wells lacking labels, and Ganin teaches a known machine-learning technique for addressing this exact type of domain-shift problem by training a feature extractor using both label-prediction loss and domain-classification loss so that the learned features become invariant across source and target domains, thereby enabling models trained on labeled source data to generalize to unlabeled target data such as seismic regions without wells.
Claims 7 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over
Meek in view of Yoo and further in view of Castagna et al. (NPL: “Relationships between compressional-wave and shear-wave velocities in clastic silicate rocks” (Geophysics (1985)) (hereinafter Castagna).
As per claim 7, the combination of Meek and of Yoo teaches the system as stated
above. Meek further teaches generating rock-property volumes including acoustic impedance and shear impedance using seismic inversion models constructed from well-log data and seismic attributes (see ¶¶ [0008]-[0010]). Yoo teaches predicting impedance values for a target region lacking well control using a machine-learning domain-adaptation model trained using source data associated with wells and target seismic data without well (see Abstract).
However, the combination of Meek and of Yoo fails to explicitly disclose converting a smoothed P-impedance value into a smoothed S-impedance value using a relational expression derived from a relationship between a P-impedance value and an S-impedance value included in well log data of the source data.
Castagna, however, teaches that empirical-relationships between P-wave velocity (VP) and S-wave velocity (VS) can be derived from well-log measurements of clastic rocks, commonly referred as the “mudrocks”, and that such relationships are used to estimate S-wave velocity when only P-wave measurements are available (see page 572). Because acoustic impedance and shear impedance are functions of density and the respective wave velocities, these VP & VS relationships derived from well logs enable estimation of S-impedance from P-impedance (i.e., acoustic impedance) values. It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to apply such relationship described by Castagna to convert P-impedance values into S-impedance values in the inversion workflow of Meek using the impedance values predicted by Yoo because Castagna teaches that empirical relationships between P-wave and S-wave properties derived from well-log data are routinely used to estimate S-wave parameters when direct measurements are unavailable, thereby enabling estimation of S-impedance values for regions lacking well-log measurements.
Claims 8 and 15 are rejected under 35 U.S.C. 103 as being unpatentable over
Meek in view of Yoo and further in view of Gardner et al. (NPL: “formation Velocity and Density –The diagnostic Basics for Stratigraphic Traps” (Geophysics (1974)) (hereinafter Gardner).
As per claim 8, the combination of Meek and of Yoo teaches the system as stated
above. Meek further teaches generating rock-property volumes including acoustic impedance and shear impedance using seismic inversion models constructed from well-log data and seismic attributes (see ¶¶ [0008]-[0010]). Yoo teaches predicting impedance values for a target region lacking well control using a machine-learning domain-adaptation model trained using source data associated with wells and target seismic data without well (see Abstract).
However, the combination of Meek and of Yoo fails to explicitly disclose converting the smoothed P-impedance value into the smoothed density value using a relational expression derived from a relationship between a P-impedance value and a density value included in well log data of the source data.
Gardner teaches that empirical-relationships between P-wave velocity and rock density can be derived from well-log measurements and used to estimate density from seismic velocity data (see page 779, equation (7)). Because acoustic impedance is the product of density and P-wave velocity (ZP = ρ VP), the velocity-density relationship derived from well-log data allows density to be estimated from acoustic impedance values. It would have been obvious to one having ordinary skill in the art before the effective filling date of the claimed invention to apply the velocity-density relationship described by Gardner to convert P-impedance values into density values in inversion workflow of Meek using the impedance values predicted by Yoo because Gardner teaches that well-log-derived relationships between velocity and density are routinely used to estimate density when direct density measurements are unavailable, thereby enabling estimation of density values for regions lacking well-log measurements.
Prior art
The prior art made record and not relied upon is considered pertinent to applicant’s
disclosure:
DI et al. [‘857] discloses a method for seismic processing includes extracting, using a first machine learning model, one or more seismic features from seismic data representing a subsurface domain, receiving one or more well logs representing one or more subsurface properties in the subsurface domain, and predicting, using a second machine learning model, the one or more subsurface properties in the subsurface domain at a location that does not correspond to an existing well based on the seismic data, the one or more well logs, and the one or more seismic features that were extracted from the seismic data.
Xia [‘119] discloses a computer-implemented method, and system implementing the method, for computing a final model of elastic properties, using nonlinear direct prestack seismic inversion for Poisson impedance. User inputs and earth-model data is obtained over points of incidence of a survey region, at various angles of incidence. Various models are then computed that serve for lithology identification and fluid discrimination and take part in preliminary seismic exploration and reservoir characterization. Therefore, further refinement of these models is required due to changes in burial depths, compaction and overburden pressure, as they provide limitations for reservoirs on porous media. The further refinement using nonlinear direct prestack seismic model is performed on a system computer, which produces a final model of elastic properties. This model can then be applied for lithology prediction and fluid detection to identify potential targets of oil and gas exploration and estimating spots in unconventional shale gas applications.
Wang. et al. [‘013] discloses a method for property estimation including receiving a seismic dataset representative of a subsurface volume of interest and a well log from a well location within the subsurface volume of interest; identifying seismic traces in the seismic dataset that correspond to the well location to obtain a subset of seismic traces; windowing the subset of seismic traces and the well log to generate windowed seismic traces and a windowed well log; multiplying the windowed seismic traces and the windowed well log by a random matrix to generate a plurality of training datasets; and training a neural network using the plurality of training datasets. The method may be executed by a computer system.
Contact information
Any inquiry concerning this communication or earlier communications from the
examiner should be directed to MOHAMED CHARIOUI whose telephone number is (571)272-2213. The examiner can normally be reached Monday through Friday, from 9 am to 6 pm.
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.
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.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Mohamed Charioui
/MOHAMED CHARIOUI/Primary Examiner, Art Unit 2857