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
Status of Claims
Claims 1-20 are rejected under 35 U.S.C. 101.
Claims 1-9, 11, and 13 are rejected under 35 U.S.C. 102 Rejection.
Claims 10, 12, and 14-20 are rejected under 35 U.S.C. 103 Rejection.
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 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more as addressed below.
The new 2019 Revised Patent Subject Matter Eligibility Guidance published in the Federal Register (Vol. 84 No. 4, Jan 7, 2019 pp 50-57) has been applied and the claims are deemed as being patent ineligible.
The current 35 USC 101 analysis is based on the current guidance (Federal Register vol. 79, No. 241. pp. 74618-74633). The analysis follows several steps. Step 1 determines whether the claim belongs to a valid statutory class. Step 2A prong 1 identifies whether an abstract idea is claimed. Step 2A prong 2 determines whether an abstract idea is integrated into a practical application. If the abstract idea is integrated into a practical application the claim is patent eligible under 35 USC 101. Last, step 2B determines whether the claims contain something significantly more than the abstract idea. In most cases the existence of a practical application predicates the existence of an additional element that is significantly more.
Under the Step 1 of the eligibility analysis, we determine whether the claims are to a statutory category by considering whether the claimed subject matter falls within the four statutory categories of patentable subject matter identified by 35 U.S.C. 101: Process, machine, manufacture, or composition of matter. The below claim is considered to be in a statutory category (process).
Under Step 1 of the analysis, claims 1, 11 and 16 does belong to a statutory category, namely they are process, machine, or manufacture claims.
Under Step 2A Prong 1, the independent claim 1 includes abstract ideas as highlighted (using a bold font) below.
“1. A method for conditioning seismic data, the method comprising:
receiving unconditioned seismic data;
introducing transformations into the unconditioned seismic data to generate transformed seismic data;
training a neural network model using the transformed seismic data to attempt to reproduce the unconditioned seismic data; and
conditioning the unconditioned seismic data using the trained neural network model to produce conditioned seismic data”.
“11. A computing system comprising:
one or more processors; and
a memory system including one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations, the operations including:
receiving unconditioned seismic data, the unconditioned seismic data includes a 3D cube representing a subterranean formation;
introducing transformations into the unconditioned seismic data to generate transformed seismic data, the transformations include a distortion and noise, parameters of the distortion and the noise are bounded, variable, randomized, or a combination thereof, and the parameters control a nature, a strength, or both of the distortion and the noise;
training a neural network model using the transformed seismic data to attempt to reproduce the unconditioned seismic data, training the neural network model includes determining weights of the neural network model, and the weights are updated using loss functions in a data domain, a frequency domain, or both; and conditioning the unconditioned seismic data using the trained neural network model to produce conditioned seismic data, conditioning the unconditioned seismic data includes correcting the distortion and the noise, removing the distortion and noise, or both.”
“16. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor of a computing system, cause the computing system to perform operations, the operations comprising:
receiving unconditioned seismic data, wherein the unconditioned seismic data includes a 3D cube representing a subterranean formation;
introducing transformations into the unconditioned seismic data to generate transformed seismic data, the transformations include a distortion and noise, parameters of the distortion and the noise are bounded, variable, and randomized, the parameters control a nature and a strength of the distortion and the noise, the distortion and the noise are introduced before or after reducing a dynamic range of the unconditioned seismic data, the transformed seismic data, or both, and the distortion and the noise are configured to be introduced in either order or simultaneously;
training a neural network model using the transformed seismic data to attempt to reproduce the unconditioned seismic data, training the neural network model includes determining weights of the neural network model, the weights are determined in a single or a multi-objective fashion, the weights are determined simultaneously or in a piecemeal iterative fashion, the weights are determined using single or multiple loss functions, the weights are determined with varying relative proportions of the loss functions, and the weights are updated using the loss functions in a data domain, a frequency domain, or both;
conditioning the unconditioned seismic data using the trained neural network model to produce conditioned seismic data, conditioning the unconditioned seismic data includes correcting and removing the distortion and the noise, and conditioning the unconditioned seismic data enhances geological features; and displaying the conditioned seismic data.”
The highlighted steps indicated as abstract idea limitations are considered to be equivalent to mathematical steps and fundamental aspect of mathematics or directed to mental processes performed in the human mind (including observation, evaluation and opinion).
The limitation of “transformations into the unconditioned seismic data to generate transformed seismic data …” is generally converting raw measurements into structured, interpretable dataset, which involve fundamental mathematical steps. Similarly, the “conditioning” step is a mathematical step performed on the unconditioned seismic data to convert it to conditioned seismic data.
Regarding the step of “training a neural network”, this is seen as closely analogous to the neural network training step in Example 47, claim 2, of the guidance on 35 USC 101. The later claims explicitly recite particular mathematical concepts involving determining weights and using functions and claim 1 implicitly includes such concepts within its scope. (To the extent that any of these neural network training steps was not deemed to be a mathematical concept, it would alternatively be treated as an additional element which is a generic computer processing step and does not integrate the other abstract idea limitations into a particular practical application.) In contrast, the neural network training step recited here is not closely analogous to the neural network training step in Example 39 of the guidance on 35 USC 101, where the claim recites constructing a neural network capable of performing a specific facial recognition task. The training here does not produce a neural network which is an end-result capable of performing a specific task; like claim 2 in Example 47, these claims only recite the training in a broad sense as an intermediate step in the analysis being performed in the claim.
Next, under Step 2A, Prong Two, we consider whether the claim that recites a judicial exception is integrated into a practical application. In this step, we evaluate whether the claim recites additional elements that integrate the exception into a practical application of that exception.
This judicial exception is not integrated into a practical application because there is no improvement to another technology or technical field; improvements to the functioning of the computer itself; a particular machine; or effecting a transformation or reduction of a particular article to a different state or thing.
Examiner notes that since the claimed methods and computing system just comprise a field of use but are not tied to a particular machine or apparatus, they do not represent an improvement to another technology or technical field.
The steps of “receiving unconditioned seismic data “, “receiving unconditioned seismic data, the unconditioned seismic data includes a 3D cube representing a subterranean formation”, “receiving unconditioned seismic data, wherein the unconditioned seismic data includes a 3D cube representing a subterranean formation” is just data gathering, which is insignificant extra solution activity.
The step of “displaying the conditioned seismic data” just data outputting, which is insignificant extra solution activity.
The step of “training a neural network” could be considered either a mathematical process (including such steps as determining weights and updating weights using loss functions as recited in claim 8, for instance) as done above, or the act of training the neural network could alternately be taken as an additional element which would be considered to amount to the generic use of computer processing components, namely the construction of a neural network computer module. In the latter case, the mere training of the neural network, which as presently recited is not recited as being intended to perform a sufficiently specific task but would cover a broad range of tasks in the field of seismic data analysis, would not be enough to integrate the abstract idea into a particular practical application.
Under step 2B Claims 1, 11 and 16 do not comprise any additional elements which would make the claim significantly more than the abstract idea, for the same reasons as discussed at Prong Two.
Claims 11 and 16 comprising the “processors” and “a memory system including one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors” and “a non-transitory computer-readable medium storing instructions that, when executed by at least one processor of a computing system” are just general parts of the computer and software running on the computer. The computer is a general computer, which is not significantly more.
The additional limitations in relation to the computer, computer product, or computer system does not offer a meaningful limitation beyond generally linking the use of the method to a computer (see ALICE CORP. v. CLS BANK INT’L 573 U. S. 208 (2014)). The claim does not recite a particular machine applying or being used by the abstract idea.
The dependent claims 2, 3, 4, 5, 6, 7, 17 and 19 merely additionally describe the type of data.
The dependent claims 8, 9, 13, 14, 18 and 20 are merely extending the details of the abstract idea of mathematical concepts, more particularly mathematical calculations or mental steps as recited.
Claim 10 additionally performs the “wellsite action”, which is common action in this field, therefore is directed to insignificant additional steps.
Claim 12 additionally performs obtaining data with more details.
Claim 15 is just additionally displaying data, which is merely output of the data.
Therefore claims 2-10, 12-15 and 17-20 are similarly rejected under 35 U.S.C. 101.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 1-9, 11, and 13 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Liuqing Yang et al., “Unsupervised 3-D Random Noise Attenuation Using Deep Skip Autoencoder”, hereinafter Yang.
Regarding Claim 1, Yang discloses a method for conditioning seismic data, the method comprising:
receiving unconditioned seismic data (Fig. 2(a); Page 4, right column, A, 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; e.g., the clean seismic data is the "unconditioned seismic data");
introducing transformations into the unconditioned seismic data to generate transformed seismic data (Figs. 1, 2(a)-(b); Page 4, right column, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)";
a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data", see Page 2, lines 16-18, where UL algorithm for separating useful signals from noisy data; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data", as per FIG. 1, see Fig. 4, left col, lines 7-10, where input noisy patches seismic data P and output patches in DDUL; see B. Patching and Unpatching Process);
training a neural network model using the transformed seismic data to attempt to reproduce the unconditioned seismic data (Fig. 1, when 2D slices of the unconditioned seismic data, when provided with, as input 2D slices of the transformed seismic data; abstract; "In this article, we propose a deep-denoising unsupervised learning (DDUL) network to attenuate random noise in 2-D/3-D seismic data... We use the fully symmetrical structure of the autoencoder to construct the network";
page 2, right-hand-column, lines 6-7, and 15-24: “denoising CNN to attenuate desert seismic random noise”, lines 15-24; "In seismic random noise attenuation, autoencoder (AE) [49]-[51] is a classic UL algorithm for separating useful signals from noisy data. The traditional AE is mainly divided into two structures i.e., encoder and decoder. The encoder learns the feature representation of the useful signal in the input noisy data and then maps the feature representation to an abstract feature space. The decoder reconstructs the abstract features and then maps these abstract features back to the original seismic data-space to achieve random noise attenuation [52]";
page 4, left-hand-column, lines 6-9, "To obtain the best denoising performance, the mean-squared error (MSE) loss function is used to minimize the error between input noisy patched seismic data P and output patches in DDUL";
since the error function to update the and autoencoder (AE) aims to minimize the error between the AEs output and input (shown in FIG. 1), the AE is trained to minimize the difference between it's input and output, and in doing so, attenuates random noise in the input; therefore, the AE attempts to reconstruct a noise-free version of it's input (which is the "clean data" or "the unconditioned seismic data");
conditioning the unconditioned seismic data using the trained neural network model to produce conditioned seismic data (FIG. 1; the process of the DDUL AE shown conditions unconditioned seismic data (that is, produces "denoising seismic data" from "input noisy seismic data"), e.g., conditioning, producing , as output, by the trained neural network model, 2D slices of conditioned seismic data, when provided with as input 2D slices of unconditioned seismic data).
Regarding Claim 2, Yang discloses the method of claim 1, wherein the unconditioned seismic data includes a 3D cube representing a subterranean formation with regards to claim 2, D1 discloses wherein the unconditioned seismic data includes a 3D cube representing a subterranean formation (Figs.1, and 9(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; the clean seismic data is the "unconditioned seismic data"; while the above passage only explicitly describes 2-D seismic data; as per FIG. 1, the process of the DDUL may be applied to 2-D or 3-D seismic data (such as the 3-D seismic data shown in FIG. 9), and it is therefore implicit that the "unconditioned seismic data" may be either 2-D or 3-D (e.g., the clean data of FIG. 9(a)).
Regarding Claim 3, Yang discloses the method of claim 1, wherein the transformations include a distortion, noise, or both (FIGS. 1, 2(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data"; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data, as per Fig. 1).
Regarding Claim 4, Yang discloses the method of claim 3, wherein parameters of the distortion, the noise, or both are bounded (FIGS. 1, 2(a)-(b); page 4, A. 2-D Seismic Random Noise Attenuation, III. Example; the distortion is bounded by parameter "P" which defines the size of each patch ("PxP"); implicit that the Gaussian noise is variable (i.e., the strength and bounds of the noise can be modified) and bounded (e.g., bounded by the domain over which the Gaussian noise is defined); "DDUL is used to attenuate random noise in 2-D and 3-D seismic data").
Regarding Claim 5, Yang discloses the method of claim 3, wherein parameters of the distortion, the noise, or both are variable(FIGS. 1, 2(a)-(b); page 4, A. 2-D Seismic Random Noise Attenuation, III. Example; the distortion is bounded by parameter "P" which defines the size of each patch ("PxP"); implicit that the Gaussian noise is variable (i.e., the strength and bounds of the noise can be modified) and bounded (e.g., bounded by the domain over which the Gaussian noise is defined); "DDUL is used to attenuate random noise in 2-D and 3-D seismic data").
Regarding Claim 6, Yang discloses the method of claim 3, wherein parameters of the distortion, the noise, or both are randomized (FIGS. 1, 2(a)-(b); page 4, A. 2-D Seismic Random Noise Attenuation, III. Example; the distortion is bounded by parameter "P" which defines the size of each patch ("PxP"); implicit that the Gaussian noise is variable (i.e., the strength and bounds of the noise can be modified) and bounded (e.g., bounded by the domain over which the Gaussian noise is defined); "DDUL is used to attenuate random noise in 2-D and 3-D seismic data").
Regarding Claim 7, Yang discloses the method of claim 3, wherein parameters of the distortion, the noise, or both control a nature and a strength of the distortion, the noise, or both (page 2, left col, lines 9-12, where Rank-reduction-based algorithms assume that the useful signal in the seismic data has a low-rank structure and then attenuates the random noise by reducing the rank of the noisy data; page 3, right col. lines 15-19, where encoder layer can be regarded as a seismic data dimension-reduction structure. In the process of dimension-reduction. the features of the useful seismic signal c(t) are extracted layer by layer from the noisy data x(t), and then used to attenuate random noise and see Page 4, right Col. lines 9-10, where to extract the waveform feature of noisy data and reconstruct denoised data).
Regarding Claim 8, Yang discloses the method of claim 1, wherein training the neural network model includes determining weights of the neural network model, and wherein the weights are updated using loss functions in a data domain, a frequency domain, or both (FIG. 1; page 4, left-hand-column; An Adam optimizer with root-mean-square propagation to minimize a loss-function based on a single objective (to minimize the error between input noisy patched seismic data P and output patches in DDUL) is used to update the weights of the DDUL).
Regarding Claim 9, Yang discloses the method of claim 1, wherein conditioning the unconditioned seismic data includes: correcting the distortion, the noise, or both; removing the distortion and noise, or both; or both FIG. 1; "Unpatching" removes the "Patching" (or distortion) from the input; the decoder part of the autoencoder (AE) shown in FIG. 1, by attempting to reconstruct the input as output, denoises, or
removes the Gaussian noise, from the input; the output 2D slices of the AE shown in FIG. with noise and distortion removed, are 2D slices of conditioned seismic data)..
Regarding Claim 11, Yang discloses a computing system comprising: one or more processors; and a memory system including one or more non-transitory computer-readable media storing instructions that, when executed by at least one of the one or more processors, cause the computing system to perform operations (FIG. 1; implicit that the functioning shown is performed by a computing device which comprises a processor, memory and instructions), the operations including:
receiving unconditioned seismic data, the unconditioned seismic data includes a 3D cube representing a subterranean formation (Figs. 1, 9(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; the clean seismic data is the "unconditioned seismic data"; while the above passage only explicitly describes 2-D seismic data; as per FIG. 1, the process of the DDUL may be applied to 2-D or 3-D seismic data (such as the 3-D seismic data shown in FIG. 9), and it is therefore implicit that the "unconditioned seismic data" may be either 2-D or 3-D (e.g., the clean data of FIG. 9(a)));
introducing transformations into the unconditioned seismic data to generate transformed seismic data, the transformations include a distortion and noise (Figs. 1, 2(a)-(b); Page 4, A. 2-1) Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data"; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data, as per FIG.1),
parameters of the distortion and the noise are bounded, variable, randomized, or a combination thereof, and the parameters control a nature, a strength, or both of the distortion and the noise(FIGS. 1, 2(a)-(b)); Page 4, A. 2-D Seismic Random Noise Attenuation; the distortion is bounded by parameter "P" which defines the size of each patch ("PxP"); implicit that the Gaussian noise is variable (i.c., the strength and bounds of the noise can be modified) and bounded (e.g., bounded by the domain over which the Gaussian noise is defined));
training a neural network model using the transformed seismic data to attempt to reproduce the unconditioned seismic data (Fig. 1, when 2D slices of the unconditioned seismic data, when provided with, as input 2D slices of the transformed seismic data; abstract; "In this article, we propose a deep-denoising unsupervised learning (DDUL) network to attenuate random noise in 2-D/3-D seismic data... We use the fully symmetrical structure of the autoencoder to construct the network";
Page 2, right-hand-column lines 6-7, and 15-24: “denoising CNN to attenuate desert seismic random noise”, lines 15-24; "In seismic random noise attenuation, autoencoder (AE) [49]-[51] is a classic UL algorithm for separating useful signals from noisy data. The traditional AE is mainly divided into two structures i.e., encoder and decoder. The encoder learns the feature representation of the useful signal in the input noisy data and then maps the feature representation to an abstract feature space. The decoder reconstructs the abstract features and then maps these abstract features back to the original seismic data-space to achieve random noise attenuation [52]";
page 4, left-hand-column, lines 6-9, "To obtain the best denoising performance, the mean-squared error (MSE) loss function is used to minimize the error between input noisy patched seismic data P and output patches in DDUL";
since the error function to update the and autoencoder (AE) aims to minimize the error between the AEs output and input (shown in FIG. 1), the AE is trained to minimize the difference between its input and output, and in doing so, attenuates random noise in the input; therefore, the AE attempts to reconstruct a noise-free version of its input (which is the "clean data" or "the unconditioned seismic data"), training the neural network model includes determining weights of the neural network model, and the weights are updated using loss functions in a data domain, a frequency domain, or both (page 4, left-hand-column; "Adam optimizer [59] is used to update the [theta] = {W, b} during the network training stage"; {W, b} clearly refer to the weights and biases of the autoencoder network); and
conditioning the unconditioned seismic data using the trained neural network model to produce conditioned seismic data, conditioning the unconditioned seismic data includes correcting the distortion and the noise, removing the distortion and noise, or both (FIG. 1; "Unpatching" removes the "Patching" (or distortion) from the input; the decoder part of the autoencoder (AE) shown in FIG. 1, by attempting to reconstruct the input as output, denoises, or removes the Gaussian noise, from the input).
Regarding Claim 13, Yang discloses the computing system of claim 11, wherein the weights are determined in a single or a multi-objective fashion, and wherein the weights are determined simultaneously or in a piecemeal iterative fashion (Fig. 1; page 4, left-hand-column; The weights of the DDUL are updated using an Adam optimizer with root-mean-square propagation that minimizes a loss-function based on a single objective (the error between input noisy patched seismic data P and output patches in DDUL)).
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.
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over in Yang in view of Aldred et al., (US Pat.8,838,426 B2), hereinafter Aldred.
Regarding Claim 10, Yang discloses the method of claim 1, but does not disclose comprising performing a wellsite action based at least partially upon the conditioned seismic data.
Aldred discloses: “driller uses these measurements to predict whether the desired target is likely to be intersected and may take corrective actions to parameters such as weight-on-bit and drilling-rotational-speed to cause the drilling trajectory to change in the direction of the target if necessary” (Col. 10, lines 23-29) and see “that data and changes in the operating environment that the data reflects may have significant effect on how an operator of the drilling of the hydrocarbon well or operation of the hydrocarbon related procedure would set parameters for optimal process performance or where the such data”(Page 15, lines 35-41).
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to performing a wellsite action as taught by Aldred based on conditioned seismic data of the Yang in order to more optimally perform the drilling operation.
Claim 12 is rejected under 35 U.S.C. 103 as being unpatentable over in Yang.
Regarding Claim 12, Yang discloses the computing system of claim 11, but does not disclose wherein the distortion and the noise are introduced before or after reducing a dynamic range of the unconditioned seismic data, the transformed seismic data, or both, and the distortion and the noise are configured to be introduced in either order or simultaneously.
The dynamic range compression was well-known technics in the relevant art for improving image interpretation.
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to apply dynamic range compression to the unconditioned seismic data of the Yang before or after distortion and noise are introduced in order for improving image interpretation and visual clarity.
Claim 14 is rejected under 35 U.S.C. 103 as being unpatentable over in Yang in view of Li et al., (US Pub.20190050734A1), hereinafter Li.
Regarding Claim 14, Yang discloses the computing system of claim 11, but does not disclose wherein the weights are determined or updated with varying relative proportions of the loss functions.
Li discloses the weights are determined or updated with varying relative proportions of the loss functions (para [0104], where equations mean updating the weight matrix by subtracting the product of learning rate and gradient of the loss function from the weight matrix).
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to adjust the "learning rate" of these loss functions, thereby adjusting the proportion of the single or multiple loss functions with which weights as taught by Li into Yang in order to improve the balance contributions, stabilizes training, speeds convergence, and improves overall model performance.
Claims 15-19 are rejected under 35 U.S.C. 103 as being unpatentable over in Yang in view of Yarus et al., (WO2018045255A1), hereinafter Yarus.
Regarding Claim 15, Yang discloses the computing system of claim 11, but does not disclose wherein the operations include displaying the conditioned seismic data.
Yarus discloses the operations include displaying the conditioned seismic data (para [0046], where additional conditioning data for a visual representation of the PV model to be displayed).
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to displaying the conditioned seismic data as taught by Yarus into Yang in order to improves the visual quality of the subsurface image and directly enhances the reliability of quantitative analysis.
Regarding Claim 16, Yang discloses a non-transitory computer-readable medium storing instructions that, when executed by at least one processor of a computing system, cause the computing system to perform operations, the operations comprising:
receiving unconditioned seismic data, wherein the unconditioned seismic data includes a 3D cube representing a subterranean formation(Figs. 1, 9(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; the clean seismic data is the "unconditioned seismic data"; while the above passage only explicitly describes 2-D seismic data; as per FIG. 1, the process of the DDUL may be applied to 2-D or 3-D seismic data (such as the 3-D seismic data shown in FIG. 9), and it is therefore implicit that the "unconditioned seismic data" may be either 2-D or 3-D (e.g., the clean data of FIG. 9(a)));
introducing transformations into the unconditioned seismic data to generate transformed seismic data, the transformations include a distortion and noise (FIGS. 1, 2(a)-(b); Page 4, A. 2-1) Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data"; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data, as per FIG. 1), parameters of the distortion and the noise are bounded, variable, and randomized, the parameters control a nature and a strength of the distortion and the noise (FIGS. 1, 2(a)-(b)); Page 4, A. 2-D Seismic Random Noise Attenuation; the distortion is bounded by parameter "P" which defines the size of each patch ("PxP"); implicit that the Gaussian noise is variable (i.e., the strength and bounds of the noise can be modified) and bounded (e.g., bounded by the domain over which the Gaussian noise is defined)),
the distortion and the noise are configured to be introduced in either order or simultaneously (FIGS. 1, 2(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4msand contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data"; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data", as per FIG. 1; while in the embodiment described, the distortion is introduced after the noise; there appears to be no barrier for a person skilled in the art to change the order of the applied transformations, SO that noise is introduced after the distortion; at least in-part because of this, it is considered that the distortion and noise are "configured" to be introduced in either order, e.g, distortion and noise introduced in order);
training a neural network model using the transformed seismic data to attempt to reproduce the unconditioned seismic data (Fig.1, when 2D slices of the unconditioned seismic data, when provided with, as input 2D slices of the transformed seismic data; abstract; "In this article, we propose a deep-denoising unsupervised learning (DDUL) network to attenuate random noise in 2-D/3-D seismic data... We use the fully symmetrical structure of the autoencoder to construct the network";
Page 2, right-hand-column lines 6-7, and 15-24: “denoising CNN to attenuate desert seismic random noise”, lines 15-24; "In seismic random noise attenuation, autoencoder (AE) [49]-[51] is a classic UL algorithm for separating useful signals from noisy data. The traditional AE is mainly divided into two structures i.e., encoder and decoder. The encoder learns the feature representation of the useful signal in the input noisy data and then maps the feature representation to an abstract feature space. The decoder reconstructs the abstract features and then maps these abstract features back to the original seismic data-space to achieve random noise attenuation [52]";
page 4, left-hand-column, lines 6-9, "To obtain the best denoising performance, the mean-squared error (MSE) loss function is used to minimize the error between input noisy patched seismic data P and output patches in DDUL";
since the error function to update the and autoencoder (AE) aims to minimize the error between the AEs output and input (shown in FIG. 1), the AE is trained to minimize the difference between its input and output, and in doing so, attenuates random noise in the input; therefore, the AE attempts to reconstruct a noise-free version of its input (which is the "clean data" or "the unconditioned seismic data"),
training the neural network model includes determining weights of the neural network model, the weights are determined in a single or a multi-objective fashion, the weights are determined simultaneously or in a piecemeal iterative fashion, the weights are determined using single or multiple loss functions, the weights are determined with varying relative proportions of the loss functions, and the weights are updated using the loss functions in a data domain, a frequency domain, or both (FIG. 1; page 4, left-hand-column; An Adam optimizer with root-mean-square propagation to minimize a loss-function based on a single objective (to minimize the error between input noisy patched seismic data P and output patches in DDUL) is used to update the weights of the DDUL);
conditioning the unconditioned seismic data using the trained neural network model to produce conditioned seismic data, conditioning the unconditioned seismic data includes correcting and removing the distortion and the noise, and conditioning the unconditioned seismic data enhances geological features (FIGS. 1, 2(a)-(b); abstract; page 2, right-hand-column; page 4, left-hand-column; removing random noise from the input seismic data, as performed by the DDUL autoencoder of FIG. 1, "enhances", or improves visibility of, "geological features" in the seismic data, in the instance wherein the removed noise was initially present at the location of "geological features" in the seismic data).
Yang does not disclose the distortion and the noise are introduced before or after reducing a dynamic range of the unconditioned seismic data, the transformed seismic data, or both;
displaying the conditioned seismic data.
The dynamic range compression was well-known technics in the relevant art for improving image interpretation.
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to apply dynamic range compression to the unconditioned seismic data of the Yang before or after distortion and noise are introduced in order for improving image interpretation and visual clarity.
Yarus discloses displaying the conditioned seismic data (para [0046], where additional conditioning data for a visual representation of the PV model to be displayed).
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to displaying the conditioned seismic data as taught by Yarus into Yang in order to improves the visual quality of the subsurface image and directly enhances the reliability of quantitative analysis.
Regarding Claim 17, Yang and Yarus disclose the non-transitory computer-readable medium of claim 16, further Yang discloses wherein introducing the noise includes performing a noise task, the noise task includes adding a noise sample into the unconditioned seismic data, the transformed seismic data, or both, the noise sample includes a multi-dimensional probability distribution, and the noise sample is employed in an additive or multiplicative fashion (FIGS. 1, 2(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data" (the unconditioned seismic data); a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data" (the transformed seismic data), as per FIG. 1; the noise sample is the "Gaussian noise", wherein, as per FIGS. 1, 2(a)-(b), 9(a)-(b), the "Gaussian noise" can be applied to 2-D or 3-D data, and therefore, the "Gaussian noise" implicitly comprises at least 2 or 3 dimensions; implicit that the "Gaussian noise" is applied through addition or multiplication, since these are the standard ways for introducing noise into signals). With regards to claim 18, D1 discloses the appended features at page 4, left-hand-column. With regards to claim 19, D1 discloses wherein the distortion and the noise are intentionally introduced in the transformed seismic data (FIGS. 1, 2(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in FIG. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data"; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data", as per Fig. 1).
Regarding Claim 18, Yang and Yarus disclose the non-transitory computer-readable medium of claim 16, further Yang discloses wherein the training includes regularizing the neural network model, and regularizing the neural network model includes early stopping (Page 4, left col. lines 14-20, where network training, an optimization strategy is used to save the best interaction result and avoid network overfitting, i.e., early stopping mechanism).
Regarding Claim 19, Yang and Yarus disclose the non-transitory computer-readable medium of claim 16, further Yang discloses wherein the distortion and the noise are implicit in the unconditioned seismic data (FIG. 1; page 4, A. 2-1) Seismic Random Noise Attenuation; page 6, right-hand-column; it is implicit that field data that is input to the DDUL autoencoder of FIG. 1 comprises implicit noise (unwanted signal components) and distortion (similarly unwanted signal components, such as additive or multiplicative pixel information that does not correspond to a ground truth of a desired signal, and the distortion and the noise are intentionally introduced in the transformed seismic data (FIGS. 1, 2(a)-(b); Page 4, A. 2-D Seismic Random Noise Attenuation; "the 2-D synthetic data are generated by a Ricker wavelet with a sampling rate of 4ms and contains 76 traces. The clean data and noisy data with Gaussian noise are shown in Figs. 2(a) and (b)"; a first transformation is a noise transformation, which occurs in generating the "noisy data" from the "clean data"; a second transformation is a distortion transformation, which occurs when "Patching" the "Input noisy seismic data", as per FIG. 1).
Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over in Yang in view of Yarus, as applied above and further in view of Zhang et al., (CN112882101-A), hereinafter Zhang.
Regarding Claim 20, Yang and Yarus disclose the non-transitory computer-readable medium of claim 16, but do not disclose wherein the operations further include performing geological interpretation of the subterranean formation based upon the conditioned seismic data.
Zhang discloses the operations further include performing geological interpretation of the subterranean formation based upon the conditioned seismic data.
(page 2, lines 24-25, where steps of seismic data processing is attenuation noise, highlighting the effective signal, providing reliable data for the next step of geological interpretation).
Therefore, it would have been obvious to one of ordinary skill in the art at the time the applicants' invention was made to performing geological interpretation, as taught by Zhang in combination of Yang and Yarus in order to provide cleaner data to build accurate and ensures seismic models are grounded in real geological reality, leading to safer, more cost-effective, and more successful exploration and production outcomes.
improves the visual quality of the subsurface image and directly enhances the reliability of quantitative analysis.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
1. Omar M. Saad et al., 'Deep Denoising Autoencoder for Seismic Random Noise Attenuation', Geophysics, Volume 85, Issue 4,
published 20 April 2020, DOI: 10.1190/geo2019-0468.1, Available at:
https://www.researchgate.net/publication/340793215_Dep_Denoising_Autoencoder_for_Seismic_Random_Noise_Attenuation
2. Fanlei Meng et al., 'Self-Supervised Learning for Seismic Data Reconstruction and Denoising', IEEE Geoscience and Remote Sensing Letters, Volume 19, published 2 April 2021, DOI: 10.1109/LGRS.2021.3068132, Available at: https://iccexplore.iccc.org/document/9394868
3. Claire Birnic et al., 'The Potential of self-supervised networks for random noise suppression in seismic data', Artificial
Intelligence in Geosciences, Volume 2, Pages 47-59, published 15 Sep 2021, Available at:
https://doi.org/10.48550/arXiv.2109.07344
D5: US 10985777 B2 (WILLIAM MARSH RICE UNIVERSITY) 20 April 2021.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KALERIA KNOX whose telephone number is (571)270-5971. The examiner can normally be reached M-F 8am-5pm.
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 at (571)2722302. 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.
/KALERIA KNOX/
Examiner, Art Unit 2857
/ANDREW SCHECHTER/Supervisory Patent Examiner, Art Unit 2857
Prosecution Insights