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 .
Response to Amendment
The amendment filed 06/19/2026 has been entered. As directed, claims 1, 9, 10, 18 and 19
have been amended, no claim is added and canceled. Thus claims 1-6, 8-15 and 17-19 are remain pending in the application.
Response to Arguments
With respect to the Applicant’s argued rejection under 35 U.S.C 101 in “Applicant
Arguments/Remarks Made in an Amendment,”
Applicant argues:
III. Rejection of Claims 1-6, 8-15 and 17-19 under 35 U.S.C. § 101
…
Here, the Office Action alleges that Claims 1, 10 and 19 fail to recite details of how the a
solution to a problem is accomplished and, therefore, the claims fail to provide a practical application of an abstract idea. Applicant respectfully disagrees. Pending Claims 1, 10, and 19 are specific in how improved steering of a drill bit is performed in that each of these claims recite specific steps to improve steering of a drill bit - the recited steps being:
accessing historical run information (stored in a memory of a controller),
determining drilling measurements based on the historical run information, training at least one initial model with a machine learning method using the determined drilling measurements,
utilizing the trained at least one initial model to automatically determine a mud motor drilling mode for a mud motor,
actuating the mud motor or a rotary table operable to provide rotation to a conveyance based on the automatically determined mud motor drilling mode, and
steering the drill bit based on the actuation of the mud motor or rotation of the rotary table.
That is, pending Claims 1, 10, and 19 recite that in order to improve steering of a drill bit (automatically without manual monitoring and adjustment), a mud motor drilling mode is automatically determined based on drilling measurements from historical run information used to train a model with machine learning and then the automatically determined mud motor drilling mode is used to either actuate a mud motor (if the automatically determined mud motor drilling mode is a sliding mode) or actuate a rotary table (if the automatically determined mud motor drilling mode is a rotation mode) to steer the drill bit. Thus, pending Claims 1, 10, and 19 recite a practical application that is demonstrated by meaningful additional elements such as how a direction of a drill bit is controlled since pending Claims 1, 10, and 19 recite details of how improved steering of a drill bits is accomplished.
The Manual of Patent Examining Procedure (MPEP) states:
If the additional elements in the claim integrate the recited exception into a practical application of the exception, then the claim is not directed to the judicial exception (Step 2A: NO) and thus is eligible at Pathway B. This concludes the
eligibility analysis.1
Since, as established above, pending Claims 1, 10, and 19 recite a practical application of the claimed automatically determined mud motor drilling mode, per the above-cited portion of the MPEP, Claims 1, 10, and 19 are eligible at Step 2A of the Alice/Mayo test. Claims 5-6 and 8-9 and Claims 11-15 and 17-18, at least by their dependence on Claims 1 and 10, respectively, are also eligible at Step 2A of the Alice/Mayo test.
For at least this reason, the § 101 rejection of pending Claims 1-6, 8-15, and 17-19 should be overturned and set to issue. Accordingly, Applicant respectfully requests the Office to withdraw the § 101 rejection of pending Claims 1-6, 8-15, and 17-19 and allow issuance thereof.
(see Response filed 06/19/2026 [pages 8-10]).
With respect to applicant’s argument, it is persuasive because the recited additional limitations, when considered individually and in combination, integrated the judicial exception into practical application. In particular, the automatically determined mud motor drilling model is used to actuate the mud motor or rotary table, and the resulting actuation or rotation is used to steer the drill bit. This claimed relationship allows the drilling system to account for the operating model of the mud motor when controlling the direction of drilling, thereby improving directional well drilling and borehole trajectory control. Accordingly, claims 1, 10 and 19, and the claims dependent therefrom, are eligible under 35 U.S.C. § 101.
With respect to the Applicant’s argued rejection under 35 U.S.C 103 in “Applicant Arguments/Remarks Made in an Amendment,”
Applicant argues:
IV. Rejection of Claims 1-2, 4, 8-11, 13, and 17-19 under 35 U.S.C. § 103
…
While Yu may only be relied upon to teach a trained model to automatically determine a mud motor drilling mode and the cited portions of Dursun are relied upon to teach other elements of pending Claims 1, 10, and 19, the cited portions of Yu only describe using sensor data received during drilling to allegedly automatically determine a mud motor drilling mode. There is no disclosure in the cited portions of Yu that Yu uses drilling measurements based on historical run information to train a machine learning model to automatically determine a mud motor drilling mode as recited in Claims 1, 10, and 19. Thus, Yu's principle of operation is use drilling measurements during drilling, rather than from historical run information, to allegedly automatically determine a mud motor drilling mode.
While the Office Action alleges the "combination of the prior art would not change the principle of operation of the prior art invention being modified, since the trained agent as a neural network (Yu) can be modified to incorporate with Dursun's trained model by utilizing historical data with a filtering process to train and generate predictive model with machine learning method, while still using neural network and sensor data for determining drilling mode" and alleges the "modification of Yu in view of Dursun merely enhances the data used for training and improving the trained model, rather than replacing or altering Yu's use of neural network and sensor data," the Office Action cites no portion of Yu that supports these allegations, i.e., that Yu can be modified to utilize historical data rather than from sensor information during drilling. These allegations are merely conclusory.
As such, Yu's principle of operation is to only use sensor information gathered while drilling to determine the mud motor drilling mode (and not historical run information). And modifying Yu, absent any evidence that Yu can be modified to use historical data to automatically determine a mud motor drilling mode, to use historical data to determine the mud motor drilling mode with the cited portions of Dursun (or any other art for that matter), would clearly change Yu's principle of operation.
The Manual of Patent Examining Procedure (MPEP) states:
If the proposed modification or combination of the prior art would change the principle of operation of the prior art invention being modified, then the teachings of the references are not sufficient to render the claims prima facie obvious. In re Ratti, 270 F.2d 810, 813, 123 USPQ 349, 352 (CCPA 1959)3
Since, as established above, the proposed modification of Yu with Dursun changes the principle of operation of Yu (the art being modified), the references, per the above-cited portion of the MPEP, are not sufficient to render pending Claims 1, 10, and 19 and claims that depend thereon prima facie obvious.
Thus, the applied combination of the cited portions of Yu and Dursun does not provide a prima facie case of obviousness for pending Claims 1, 10, and 19 and claims that depend thereon. For at least this reason, the § 103 rejection of Claims 1-2, 4, 8-11, 13, and 17-19 should be overturned and the claims set to issue. Accordingly, Applicant respectfully requests the Office to withdraw the § 103 rejection of Claims 1-2, 4, 8-11, 13, and 17-19 and allow issuance thereof.
V. Rejection of Claims 3, 5-6, 12, and 14-15 under 35 U.S.C. § 103
Claims 3, 5-6, 12, and 14-15 stand rejected under 35 U.S.C. § 103 as allegedly being unpatentable over Yu and Dursun in view of:
…
As established above, the applied combination of the cited portions of Yu and Dursun does not provide a prima facie case of obviousness for pending Claims 1 and 10. Gunawardena and Lu and Kanevsky have not been cited to cure the above-noted deficiencies of the applied combination of the cited portions of Yu and Dursun. Instead, Gunawardena and Lu and Kanevsky have been cited to teach the features of the above-mentioned dependent claims.4 As such, the cited portions of Yu and Dursun in combination with either of the cited portions of Gunawardena or Lu and Kanevsky, as applied by the Office Action, do not provide a prima facie case of obviousness for pending Claims 1 and 10 and claims that depend thereon.
For at least this reason, the § 103 rejections of Claims 3, 5-6, 12, and 14-15 should be overturned and the claims set to issue. Accordingly, Applicant respectfully requests the Office to withdraw the § 103 rejections of Claims 3, 5-6, 12, and 14-15 and allow issuance thereof.
VI. Comments
All of Applicant's arguments are without prejudice or disclaimer. Applicant reserves the right to discuss the distinctions between the cited references and the claims in a later response or on appeal, if appropriate. The arguments offered by Applicant are sufficient to overcome the Office Action rejections and Applicant does not acquiesce to additional statements in the Office Action even if not specifically addressed in the response.
Applicant respectfully requests appropriate evidentiary support should a rejection based on any of the above asserted rejections be maintained. Additionally, if a rejection relies upon "common knowledge" or "well known" principles or "Official Notice" or other information within the personal knowledge of the Examiner, Applicant respectfully requests that the Examiner cite a reference or provide an affidavit as documentary evidence in support of this position in accordance with M.P.E.P. § 2144.03 and 37 C.F.R. 1.104(d)(2).
(see Response filed 06/19/2026 [pages 11-14]).
With respect to Applicant’s arguments with respect to claims 1, 10, and 19 have been considered but are not persuasive. Applicant argues that “the proposed modification of Yu with Dursun changes the principle of operation of Yu (the art being modified), the references, per the above-cited portion of the MPEP, are not sufficient to render pending Claims 1, 10, and 19 and claims that depend thereon prima facie obvious. Thus, the applied combination of the cited portions of Yu and Dursun does not provide a prima facie case of obviousness for pending Claims 1, 10, and 19 and claims that depend thereon.”
In response to applicant’s argument that Yu does not disclose using historical drilling information to train its neural-network agent and that modifying Yu in view of Dursun would change Yu’s principle of operation, the test for obviousness is not whether the features of a secondary reference may be bodily incorporated into the structure of the primary reference; nor is it that the claimed invention must be expressly suggested in any one or all of the references. Rather, the test is what the combined teachings of the references would have suggested to those of ordinary skill in the art. See In re Keller, 642 F.2d 413, 208 USPQ 871 (CCPA 1981).
In this case, Yu teaches training an agent component to generate a trained agent component by selecting a drilling mode, simulating drilling using the selected drilling mode, generating a reward based on the resulting state and a planned borehole trajectory, and using the reward to train the agent component. See Yu at cited paragraph [0412]. Yu further teaches subsequently using the trained neural network and sensor data received during drilling to determine a drilling mode and issue a control instruction for drilling using the determined drilling mode. See Yu at cited paragraph [0415]. Although Yu describes an exemplary simulation and reward training process, Yu does not limit its agent to a single fixed training event or preclude further training using drilling derived information. Yu clearly suggests or teaches learning may improve the agent using information acquired during drilling and that further learning may be performed after reaching the target so that the improved agent is used for drilling another borehole. See Yu at paragraph [0406]. Information acquired during a completed drilling operation and subsequently used to improve the agent for another borehole constitutes information from a prior drilling run. Thus, Yu provides evidence that its neural network agent is not limited to a single simulation and reward training process, but may also be trained and improved using drilling information obtained from prior drilling operations.
As set forth in the rejection, Dursun is relied upon to teach collecting and storing drilling information for later retrieval, including drilling parameters generated during drilling operations; retrieving and segregating drilling datasets; preprocessing the data to remove noisy, corrupted, or missing data and to form training datasets; and training a machine-learning model using the processed training datasets. See Dursun at cited paragraphs [0022], [0025]-[0027], [0029]-[0030] and [0036]-[0037].
Accordingly, the proposed combination does not require bodily incorporating Dursun’s entire predictive model system into Yu or replacing Yu’s disclosed drilling mode determination and control framework. Rather, consistent with the rationale previously set forth, Dursun provides the particular technique for collecting, processing, and using prior run drilling information to train and improve a machine learning model. While Yu permits its agent to be further improved using information obtained from completed drilling operations. Yu continues to use the trained neural network and sensor data during a drilling operation to determine the drilling mode and issue the corresponding control instruction.
Therefore, the proposed modification does not change Yu’s principle of operation. Yu remains operable to use a trained neural network and drilling sensor data to determine a drilling mode and control the drilling operation, while Dursun provides a known technique for preparing and using drilling information to train and improve the machine-learning model. Accordingly, Applicant’s arguments are not persuasive, and the rejection of claims 1, 10, and 19, and claims depending therefrom, under 35 U.S.C. § 103 is maintained.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and
103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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(s) 1, 2, 4, 8-11, 13 and 17-19 are rejected under 35 U.S.C. 103 as being unpatentable over Yu
US20200370409A1 in view of Dursun US20180025269 A1.
Claim 1, Yu teaches A method for identifying a mud motor drilling mode (abstract, “…determining a drilling mode from a plurality of drilling modes using a trained neural network…”; [0441], “determining a drilling mode from a plurality of drilling modes using a trained neural network and at least a portion of the sensor data; … the plurality of drilling modes can include a rotary drilling mode and a sliding drilling mode.” [0198], “a mud motor can be capable of delivering a desired well curvature via operations that can include switching between rotating and sliding modes (e.g., rotate mode and slide mode).”), comprising:
training at least one initial model with a machine learning method ([0412], “… a train block 2940 for, using the reward, training the agent component to generate a trained agent component …” [0425], “a trained agent component can include a trained value-based network as a trained neural network.”)
utilizing the trained at least one initial model to automatically determine the mud motor drilling mode for a mud motor without manual monitoring and adjustment ([0160], “Through use of a mud motor, a directional drilling operation can alternate between rotating and sliding modes of drilling.” [0198], “a mud motor can be capable of delivering a desired well curvature via operations that can include switching between rotating and sliding modes (e.g., rotate mode and slide mode).” [0415], “FIG. 30 shows an example of a method 3000 … a determination block 3020 for determining a drilling mode from a plurality of drilling modes using a trained neural network and at least a portion of the sensor data; and an issuance block 3030 for issuing a control instruction for drilling an additional portion of the borehole using the determined drilling mode.” [0425], “… a trained agent component can include a trained value-based network as a trained neural network …” [0452], “As an example, a controller can include an agent component that selects a drilling mode using sensor data. In such an example, the drilling mode can be selected from a plurality of drilling modes, which may include one or more of a sliding mode (e.g., sliding up, sliding down, etc.), a rotary mode, a survey interval, etc.” Examiner note: the reference teaches automated control framework in which a trained neural network (determination block 3020) automatically determines a drilling mode based on sensor data and an issuance block issues control instructions for drilling using the determined mode, without requiring user intervention ([0415]), and a controller including an agent component, where the agent is a trained neural network ([0425]) that selects the drilling mode ([0452]). A POSITA would understand that the agent selection and control instruction framework constitutes automated determination and execution of drilling modes to perform the determining step without manual monitoring and adjustment.);
actuating the mud motor or a rotary table operable to provide rotation to a conveyance based, at least in part, on the automatically determined mud motor drilling mode, wherein the mud motor is coupled to the conveyance (Fig.2., drillstring 225, rotary table 220; motor 260, drill bit 226; [0054], “… the drillstring 225 pass through an opening in the rotary table 220.” [0055], “The kelly 218 can be used to transmit rotary motion from the rotary table 220 via the kelly drive bushing 219 to the drillstring 225, while allowing the drillstring 225 to be lowered or raised during rotation.” [0056], “As to a top drive example, the top drive 240 can provide functions performed by a kelly and a rotary table. The top drive 240 can turn the drillstring 225 …”; [0058], “The mud can then flow via a passage (e.g., or passages) in the drillstring 225 (examiner note: i.e. conveyance) and out of ports located on the drill bit 226 ...”; [0158], “A mud motor can include a bend in a motor bearing housing that provides for steering a bit toward a desired target.” [0194], “… rotation from the surface rig (e.g., table or top drive) may be stopped such that circulation of mud (e.g., drilling fluid) acts to drive the mud motor to rotate the bit downhole. As mentioned, in some instances, there can be a combination of surface rotation and downhole rotation. In general, where surface rotation is not provided, the drill string is in a sliding mode as it slides downward as drilling ahead occurs via rotation of the bit via operation of the mud motor. Such an operation can be referred to as a sliding operation (e.g., sliding mode). Another mode can be for holding the borehole direction tangent where surface equipment rotates the drillstring such that the motor bend also rotates with drillstring. In such a mode, the BHA does not have a particular drill-ahead direction. Such an operation can be referred to as a rotating operation (e.g., a rotating mode or rotary mode).” [0441], “determining a drilling mode from a plurality of drilling modes using a trained neural network … issuing a control instruction for drilling an additional portion of the borehole using the determined drilling mode. In such an example, the plurality of drilling modes can include a rotary drilling mode and a sliding drilling mode” See also [0415], [0425] and [0452]); and
steering a drill bit coupled to the conveyance based on the actuation of the mud motor or rotation of the rotary table ([0158], “A mud motor can include a bend in a motor bearing housing that provides for steering a bit toward a desired target.” [0163], “When transitioning from the rotating mode to the sliding mode, … orienting a bit to drill, … to steer the bit as appropriate to keep the trajectory on course. [0194], “… the bend can be pointed to a desired orientation while rotation from the surface rig (e.g., table or top drive) may be stopped such that circulation of mud (e.g., drilling fluid) acts to drive the mud motor to rotate the bit downhole.” Examiner note: for the limitation of “actuating the mud motor .. steering a drilling bit …”, Yu teaches multiple alternative actuation mechanisms, including rotation from the surface rig via a rotary table or top drive, and actuation of a mud motor via circulation of drilling fluid, the mud motor can provide directional drilling capability, including a bend in the motor housing for steering the drill bit toward a desire target, and issuing a control instruction for drilling using the determined drill mode. A POSITA would understand that provides different way to achieve rotation of drillstring (i.e., conveyance) by actuating the mud motor or rotary table, and to steer a drill bit coupled to the conveyance based on the actuation of the mud motor or rotation of the rotary table based on the control instruction using the determined drilling mode).
However, Yu fails to teach accessing historical run information stored in a memory of a controller; determining drilling measurements based on the historical run information; training at least one initial model with a machine learning method using the determined drilling measurements, wherein the at least one initial model comprises one or more inputs selected from a group consisting of revolutions per minute, tool-face, torque, flowrate, weight on bit, rate of penetration, differential pressure, a derivative thereof, and any combination thereof.
Dursun teaches accessing historical run information stored in a memory of a controller ([0023] “… the drilling system 100 may comprise a control unit 160 … The control unit 160 may comprise an information handling system …” [0010], “…The information handling system may include random access memory (RAM), one or more processing resources such as a central processing unit (CPU) or hardware or software control logic, ROM, and/or other types of nonvolatile memory.” [0022], “… The output of the sensors may be collected at the surface and stored, for example, in a database or data warehouse to be retrieved later.” [0025], “… the dataset comprises dynamic data 250 and static data 250. The dynamic data 250 may comprise drilling parameters, … including, … WOB, rotary speed, drill bit RPM, hook load, surface torque and torque on bit, downhole mud flow rate, return mud flow rate, SPP, and ROP. [0027], “ … an information handling system may include software executable by a processor … including accessing or otherwise receiving raw data from a remote data storage facility through a data network, manipulating the raw data, generating one or more predictive models, …” [0026], “These datasets may be retrieved and segregated according to the types of drilling operations and formations from which they were produced …”);
determining drilling measurements based on the historical run information ([0029], “Step 302 may comprise pre-processing steps to eliminate noisy, corrupted, or missing data from the received data 301. For example, the pre-processing steps may include the application of one or more thresholds, data filters and noise reduction algorithms, to alter or remove specific data entries or entire data sets.” [0030], “Step 303 comprises a feature extraction step that may be used to reduce the dimensionality of the training data sets T1-Tn before they are used to generate predictive models.”);
training at least one initial model with a machine learning method using the determined drilling measurements, wherein the at least one initial model comprises one or more inputs selected from a group consisting of revolutions per minute, tool-face, torque, flowrate, weight on bit, rate of penetration, differential pressure, a derivative thereof, and any combination thereof ([0036], “Step 305 comprises a training step, in which at least one learning algorithm 305 a with associated parameters 305b may be trained with the training data sets T1-Tn to produce one or more context-specific predictive models M1-Mn. For instance, a learning algorithm may receive as an input training data set T1 and determine a relationship between the drilling parameters and operational conditions within training data set T1 and the ROP values within training data set T1 that result from the associated drilling parameters and operational conditions.” [0037], “… the learning algorithm 305a may comprise supervised and unsupervised learning algorithms and may include a decision tree, a Bayesian belief networks, a genetic algorithms, an artificial neural network, and/or a support vector machines. Each of the above learning algorithms may “learn” by generating and refining an internal model based on the training data set. This internal model may be the context-specific predictive model corresponding to the training data set.” [0029], “Each of the training data sets T1-Tn may be associated with one or more different static variables, identified either through the binarized variables in the received data 301, or through nominal values 350 received at the pre-processing step 302. For instance, one of the training data sets T1-Tn may comprise all of the pre-processed data entries from the received data 301 …” [0030], “… a combination of drilling parameters and operating conditions are used as an input to the model …” Examiner note: the reference teaches that the training data set (T1-Tn) are formed from pre-processed data entries (i.e., data subjected to filtering and preprocessing in Step 302), which correspond to the determined drilling measures, and the training step uses these training data sets as input; therefore, the training is performed using data derived from the determined drilling measurement. The reference further teaches the drilling parameters , including RPM, torque, flow rate, WOB, and ROP, are included in the training data sets and are used as input to the predictive model, and correspond to the group of inputs. Therefore, the reference teaches that the at least one model comprises one or more inputs selected from the recited group).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Yu to incorporate the teachings of Dursun, and apply accessing stored historical drilling data, preprocessing the data to generate processed drilling measurements, and training predictive models using the processed data as input in order to enable model training based on historical drilling measurements and improve the data used for training the model. In this cases, Yu teaches training a neural network model, and using the trained model to determine drilling mode. Dursun teaches accessing historical drilling datasets, preprocessing the datasets including filtering and noise reduction to generate processed drilling parameters, segregating the processed data into training data sets, and training machine learning models using the training data sets that include drilling parameters as inputs. The combination of teachings would predictably provide the benefit of improving the accuracy and robustness of the trained model by incorporating training based on processed historical drilling datasets, thereby enabling more reliable and data driven drilling operation decisions.
Claim 2, Yu fails to teach processing the historical run information to remove ancillary data and determine statistical information associated with the determined drilling measurements.
Dursun teaches processing the historical run information to remove ancillary data ([0043], “Step 402 comprises a filtering step in which data entries may be removed based on a qualitative assessment of the data entry (note: i.e., remove ancillary data). For example, certain raw data sets include ROP values that were measured during the drilling operation, but calculated after the fact, and data entries or entries raw data sets containing these ROP values may be removed from the received data.”) and determine statistical information associated with the determined drilling measurements ([0032], “… principal component analysis may comprise a statistical algorithm in which a set of observations of possibly correlated variables, e.g., the dynamic variables and the ROP for a drilling operation, are converted using an orthogonal transformation into a set of values of linearly uncorrelated variables referred to as principal components … the variables with high variance with the ROP may be determined and selected … The processor may then determine a linear regression model that identifies the covariance structures” ).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Yu to incorporate the teachings of Dursun, and apply preprocessing including filtering to remove noisy or invalid data entries and determining statistical information from drilling measurements using statistical algorithms to identify relationships and covariance structures among drilling parameters and operational conditions in order to improve the quality and reliability of input data and extract statistical relationships for use in generating a predictive model, and the combination of teachings would predictably provide the benefit of improving predictive accuracy and robustness of the predictive model by reducing noise and leveraging statistical features derived from the drilling data.
Claim 4, Yu further teaches The method of claim 1, wherein the at least one initial model is selected from a group consisting of a neural network model ([0425], “a trained agent component can include a trained value-based network as a trained neural network.”), Random Forest, Decision Tree, K-nearest neighbors, Naive Bayes Classifier, and any combination thereof.
Claim 8, Yu further teaches The method of claim 1,wherein the automatically determined mud motor drilling mode is determined as one selected from a group consisting of rotating, sliding, sliding without pipe rocking, sliding with pipe rocking, a derivative thereof, and any combination thereof ([0452], “As an example, a controller can include an agent component that selects a drilling mode … the drilling mode can be selected from a plurality of drilling modes, which may include one or more of a sliding mode (e.g., sliding up, sliding down, etc.), a rotary mode, a survey interval, etc.”.
Claim 9, Yu further teaches The method of claim 1 further comprising selecting one of one or more trained initial models for utilization ([0005], “… determining a drilling mode from a plurality of drilling modes using a trained neural network … issuing a control instruction for drilling an additional portion of the borehole using the determined drilling mode.”).
The elements of claims 10-11, 13 and 17-19 are substantially the same as those of claims 1-2, 4 and 8-9. Therefore, the elements of claims 10-11, 13 and 17-19 are rejected due to the same reasons as outlined above for claims 1-2, 4 and 8-9. Further, the additional limitations of claims 10 and 19, “A system for determining a mud motor historical drilling mode, comprising: … and a processor operable to:” and “A non-transitory computer-readable medium comprising instructions that are configured, when executed by a processor, to:” (See Yu; [0455]).
Claim(s) 3 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Yu and Dursun
as applied to claims 1 and 10 above, and further in view of Gunawardena US20200110943A1.
Claim 3, Yu teaches The method of claim 1, further comprising: (see Yu; i.e., trained neural network); and (see Yu; i.e., trained neural network).
However, Yu and Dursun fail to teach determining hyper-parameters of model; and re-training the model using the determined hyper-parameters.
Gunawardena teaches determining hyper-parameters of model; and re-training the model using the determined hyper-parameters. ([0051], “a trained neural network and related machine assisted technologies for each discipline and sub-discipline. The system and method further include using a specific training dataset for each discipline, refining the neural network by tuning hyper-parameters, and retraining the neural network based on new data.”).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Yu and Dursun to incorporate the teachings of Gunawardena, and apply method include using a specific training dataset for each discipline, refining the neural network by tuning hyper-parameters, and retraining the neural network based on new data in order to minimize the error rate to obtain a desired accuracy [0105]. In this case, the hyper-parameters would optimize the neural network model performance with the new data by re-training, therefore, it can significantly impact its accuracy on different data distributions.
The elements of claim 12 is substantially the same as those of claim 3. Therefore, the elements of claim 12 is rejected due to the same reasons as outlined above for claim 3.
Claim(s) 5-6 and 14-15 are rejected under 35 U.S.C. 103 as being unpatentable over Yu
and Dursun as applied to claims 1 and 10 above, and further in view of Lu US20200234471A1 and Kanevsky US20160180214A1.
Claim 5, Yu and Dursun fail to teach, but Lu teaches The method of claim 1, further comprising using a scaled conjugate gradient algorithm with cross entropy as a performance function for evaluating a performance of the at least one initial model, wherein a cost function is calculated as a sum of cross-entropy loss ([0088], “an error is calculated (e.g., using a loss function or a cost function) to represent a measure of the difference (e.g., a distance measure) between the RTE-method generated data (i.e., reference scatter profile, or ground truth) and the output data of the 3D CNN as applied in a current iteration of the 3D CNN. The error can be calculated using any known cost function or distance measure between the image data, including those cost functions described above. Further, in certain implementations the error/loss function can be calculated using one or more of a hinge loss and a cross-entropy loss.” [0085], “the optimization method used in training the 3D CNN can use a form of gradient descent incorporating backpropagation to compute the actual gradients. This is done by taking the derivative of the cost function with respect to the network parameters and then changing those parameters in a gradient-related direction. The backpropagation training algorithm can be: a steepest descent method (e.g., with variable learning rate, with variable learning rate and momentum, and resilient backpropagation), a quasi-Newton method (e.g., Broyden-Fletcher-Goldfarb-Shanno, one step secant, and Levenberg-Marquardt), or a conjugate gradient method (e.g., Fletcher-Reeves update, Polak-Ribiére update, Powell-Beale restart, and scaled conjugate gradient).”).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Yu and Dursun to incorporate the teachings of Lu, and apply optimization method used in training the 3D CNN can use a form of gradient descent incorporating backpropagation to compute the actual gradients and the error/loss function can be calculated using one or more of a hinge loss and a cross-entropy loss in order to minimizes the cost criterion (i.e., the error value calculated using the cost function) [0084] and provides an optimization method for training the 3D CNN [0088].
However, Yu and Dursun and Lu fail to teach a cost function is calculated as a sum of cross-entropy loss.
Kanevsky teaches a cost function is calculated as a sum of cross-entropy loss ([0005], “… calculating the gradient for the neural network by applying a sharp discrepancy output layer objective function to the output layer may comprise calculating the gradient of a cross-entropy function.” [0026,] “… The sharp discrepancy objective function is obtained from a typical objective function, for example log-likelihood or cross entropy, comprising a sum of terms for … data in a training dataset.” Equation (7).)
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Yu and Dursun and Lu to incorporate the teachings of Kanevsky, and apply a cross-entropy objective function as a summation over training data in order to improve the training objective of the neural network by aggregating perdition errors across multiple training samples for optimization, thereby improving model training accuracy and convergence performance.
Claim 6, Yu and Dursun fail to teach, but Lu teaches The method of claim 5, wherein the cost function comprises a regularization term to prevent overfitting or an over-complicated model ([0089], “the loss function can be combined with a regularization approach to avoid overfitting the network to the particular instances represented in the training data.”).
It would have been obvious for a person of ordinary skill in the art before the effective filing date of the claimed invention to have modified Yu and Dursun to incorporate the teachings of Lu, and apply the combining a loss function with a regularization term to prevent overfitting of the model (see, e.g., [0089]) in order to reduce overfitting during training and improve generalization of the predictive model to new data, and the combination would predicably provide the benefit of improving model robustness and accuracy by preventing the model from fitting noise or specific instance in the training data.
The elements of claims 14-15 are substantially the same as those of claims 5-6. Therefore, the elements of claims 14-15 are rejected due to the same reasons as outlined above for claims 5-6.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Chen US20210230998A1 teaches method can include receiving data for a borehole trajectory … and a bottom hole assembly that includes a mud motor … generating a sequence for operation of the mud motor using a model of at least the bit, where the sequence includes a sliding mode and a rotary mode for drilling the borehole in the formation … ([0004]).
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37
CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to whose telephone number is (571)270-1303. The examiner can normally be reached Monday - Friday.
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, Emerson Puente can be reached at (571)272-3652. 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.
/YI . HAO/
Examiner, Art Unit 2187
/EMERSON C PUENTE/Supervisory Patent Examiner, Art Unit 2187
Prosecution Insights