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 .
Status of the Claims
Claims 1-8 and 10-15 are pending for examination.
Claims 1, 8 and 15 are independent Claims.
Claims 1-8 and 10-15 are rejected under 35 U.S.C. §103.
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.
Claims 1, 4-7 and 15 are rejected under 35 U.S.C. § 103 as being obvious over Poornaki et al (US 20200210824, hereinafter “Poornaki”) in view of Balan et al (US 20160132786, hereinafter “Balan”) in further view of Martin et al. (U.S. 2021/0118248 hereinafter Martin).
Regarding claim 1, Poornaki discloses [a] method, comprising: for receipt of input data from one or more assets: (Abstract; “An example method utilizing different pipelines of a prediction system, comprises receiving failure data, and asset data from SCADA system(s), receiving and dividing historical sensor data from sensors of components of wind turbines into different classes of different lead times, training a set of models to predict faults for each component using the historical sensor data and lead times with a deep neural network, evaluating each model of a set using standardized metrics, comparing evaluations of each model of a set to select a model with preferred lead time and accuracy, receive current sensor data from the sensors of the components, apply the selected model(s) to the current sensor data to generate a component failure prediction, compare the component failure prediction to a threshold, and generate an alert and report based on the comparison to the threshold.”, which discloses a method to perform as well as a method to receive input data from one or more assets(for example the SCADA systems)).
identifying and separating different event contexts from the input data (Abstract; “receiving failure data, and asset data from SCADA system(s), receiving and dividing historical sensor data from sensors of components of wind turbines into different classes of different lead times,” which discloses receiving and dividing historical sensor data (interpreted to be the input data) into different event contexts (interpreted to be the different classes of different lead times) and Figure. 8A; which discloses identified and separated event log examples in a table.)
training a plurality of machine learning models for each of the different event contexts; (Abstract; “training a set of models to predict faults for each component using the historical sensor data and lead times with a deep neural network,” which discloses training multiple models based on the different event contexts (the different lead times; and [0003]; “train failure prediction models for a first component using different lead times to create multi-class classifications, training a first set of failure prediction models using a deep neural network, the first historical sensor data and different lead times, the deep neural network including layers of a fully connected neural network, convolutional neural network, and a recurrent neural network to create a first set of failure prediction models”)
selecting, for each of the different event contexts, a best performing model from the plurality of machine learning models, ([0186]; “In step 2010, the model evaluation module 514 may compare any number of the model evaluations of failure prediction models of a set of failure prediction models to any of the other set of model evaluations to select a preferred model of the set of models. It will be appreciated that each failure prediction model of a set may be compared using similar metrics and/or different metrics as described above. Based on the two different failure prediction models in this example, the model evaluation module 514 or authorized entity may select the failure prediction model with the longer lead time, higher AUC, train sensitivity, train precision, and train specificity even though the lookback time is larger”, wherein the selected model is best performing based on a longer lead time, train precision, etc; and Abstract; “evaluating each model of a set using standardized metrics, comparing evaluations of each model of a set to select a model with preferred lead time and accuracy” which discloses selecting a best performing model based on performance and accuracy to form a compound model (interpreted to be the best performing model).)
deploying the compound model for the selected subset. ([0187]; “The model application module 516 may apply the selected failure prediction model to the current sensor data to generate a prediction”; and Abstract; “apply the selected model(s) to the current sensor data to generate a component failure prediction, compare the component failure prediction to a threshold, and generate an alert and report based on the comparison to the threshold.”; and [0114]; “Feature extraction may be used to provide a more manageable representative subset of input variables. It will be appreciated that feature extraction may extract features for the data as well as create new features from the initial set of data.” Which discloses feature extraction methods of choosing a subset of input data based on a metric and deploying it to apply it on the selected models).
Poornaki fails to explicitly disclose but Balan discloses selecting a best performing subset of the input data for the compound model based on maximizing a metric; and (Par [0036]; “In some implementations, a subset-specific machine-learning classifier may be trained on each learned subset created by the partitioning of the set of training data. In other implementations, subset-specific machine-learning classifiers may be trained for selected subsets. For example, subset-specific machine-learning classifiers may be trained only on selected subsets of training data that indicate confusion of the global machine-learning classifier (e.g., the number of data objects in the subset is greater than a confusion threshold).”, which discloses a subset selection process based on a performance driven metric (interpreted to be the confusion threshold) to select a best performing subset of input data on a maximized metric (confusion threshold)).
Poornaki and Balan are analogous art because both are concerned optimizing or enhancing machine learning model performance through data driven processing techniques. Before the effective filing date of the claimed invention, it would have been obvious to one skilled in the art of machine learning and data driven model optimization to combine the subset selection based on confusion thresholds from Balan and the event-based machine learning system of Poornaki to yield to the predictable result of selecting a best performing subset of the input data for the compound model based on maximizing a metric. The motivation for doing so would be to train a classifier based only on selected subsets of training data that indicate confusion of the global machine-learning classifier (e.g., the number of data objects in the subset is greater than a confusion threshold).”, Balan; Par [0036]).
Poornaki in view of Balan may not explicitly disclose:
and combining the selected best performing models for the different event contexts to form a compound model;
wherein the identifying and the separating the different event contexts comprises defining the different event contexts based on event type, impact of the event, and asset attribute from which event is captured.
Martin teaches:
and combining the selected best performing models (Martin (¶0048 line 1-4), “method 200 involves dynamically developing multiple RUL models 130 (a compound model) (block 251) and selecting RUL model 131 from multiple RUL models 130 (block 252).”) for the different event contexts to form a compound model (Martin (¶0048 line 5-6), “Each selection of sensors and subgroups of vehicles results in a different RUL model (best performing models).”);
wherein the identifying and the separating the different event contexts (Martin (¶0083 line 2-6), “wherein each of multiple RUL models 130 corresponds to a different one of multiple RUL intervals and a different one of multiple ENS intervals, and wherein each of multiple RUL models 130 has a corresponding one of precision values”) comprises defining the different event contexts based on event type, impact of the event, and asset attribute from which event is captured (Martin (¶0090 line 2-4), “wherein ENS 126 comprises one or more of end-of-life time of component 192 (impact of the event), end-of-life type (event type), or a confidence level of RUL 127 (asset attribute).”).
Poornaki and Balan disclose a system/method for optimizing or enhancing machine learning model performance through data driven processing techniques. Martin discloses a system/method to generate a group of best performance models for each context. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify models of Poornaki in view of Balan instead be a generated best model taught by Martin, with a reasonable expectation of success. The motivation would be to “provide direct metrics of accuracy for specific forecast measures and objective comparison of all developed RUL models, while selecting one for the actual use for determining the component RUL.” (Martin (¶0018 line 6-10)).
Regarding claim 15, is an apparatus claim corresponding to the steps of claim 1, and is rejected for the same reasons as claim 1.
Regarding claims 4, the rejection of claims 1and 8 are incorporated and Poornaki further discloses wherein the selecting the best performing model from the plurality of machine learning models to form the compound model is based on comparison of the plurality of models to a ground truth. (Par [0139]; “In various embodiments, the model evaluation module 514 compares the predictions of each failure prediction model of a set of failure prediction models using historical sensor data to compare the results against ground truth (e.g., known failures and known periods of time that the component did not fail).” Which discloses comparing the prediction models to a ground truth, this comparison resulting in the compound model.)
Regarding claims 5, the rejection of claims 1 and 8 are incorporated and Poornaki further discloses wherein the selecting the best performing model from the plurality of machine learning models to form the compound model is based on an average metric. (Summary Section, Par [0003]; “evaluating each of the first set of failure prediction models using at least a confusion matrix including metrics for true positives, false positives, true negatives, and false negatives as well as a positive prediction value, comparing by the model training and testing pipeline, the confusion matrix and the positive prediction value of each of the first set of failure prediction models,” and Par [0148] and [0157], “Examples of the metrics may include the following:…Accuracy (ACC), ACC=(TP+TN)/(P+N)=(TP+TN)/(TP+TN+FP+FN)” which discloses use of an average metric(for example accuracy or the confusion metric) to select a best performing model(compound model). )
Regarding claims 6, the rejection of claims 1and 8 are incorporated and Poornaki further discloses wherein the compound model is configured to output event prognostics based on another input data from a client. (Par [0170]; “In various embodiments, the model application module 516 may compare new sensor data to classified and/or categorized states identified by the selected model to identify when sensor data indicates a failure state or a state associated with potential failure is reached… In another example, the model application module 516 may compare the fit of sensor data to a failure state or state associate with potential failure that has been identified by the model of the model application module 516 in order to trigger or not trigger an alert or report.” Which discloses training results of the compound model to output event prognostics (failure state or state associate with potential failure that has been identified by the model).)
Regarding claims 7, the rejection of claims 1and 8 are incorporated and Poornaki further discloses wherein the input data from the one or more assets is indicative of sequential events obtained from the one or more assets. (Par [0207];” In step 2112, the data extraction module 504 and/or the data preparation module 506 may mine and discover patterns among the event and alarm data in the longitudinal history (e.g., patterns may be as simple as unique event code counts in a past time period such as a month, advanced time sequence patterns such as A->B->C, or complicated encoded event sequence vectors).” and Par [0210];” The data extraction module 504 and/or the data preparation module 506 may then count a number of each event code that occurred during the period of time and sequence events” which discloses the different types of sequential data types/ event contexts collected from one or more assets.)
Claims 2, 3, 8 and 10-14 are rejected under 35 U.S.C. § 103 as being obvious over Poornaki and Balan in view of Martin further in view of Nguyen et al (US 20200327225, hereinafter “Nguyen”).
Regarding claims 2 and 8, the rejection of claims are incorporated but Poornaki and Balan in view of Martin fail to explicitly disclose but Nguyen discloses separating different events in the input data into one-step incremented event subsets; and (Par [0234]; “At 810, the representation subsystem 246 can determine a second indicator 722(j) at least partly by applying a second representation vector 718(j) ("R. Vee.") of the representation vectors 718 to the trained classifier 704. In some examples, the second indicator 722(j) is associated with a second event-data value 714(j) that immediately follows the first event-data value 714(i) in the ordered sequence 712 of event-data values 714, e.g., j=i+l for sequences using incrementing indices, or j=i-1 for sequences using decrementing indices.” Which discloses incrementing event data values at a time (j = i + 1), effectively separating events into subsets based on incremental indices.)
forming each of the different event contexts from subsets of the one-step incremented event subsets (Par [0234]; “At 810, the representation subsystem 246 can determine a second indicator 722(j) at least partly by applying a second representation vector 718(j) ("R. Vee.") of the representation vectors 718 to the trained classifier 704. In some examples, the second indicator 722(j) is associated with a second event-data value 714(j) that immediately follows the first event-data value 714(i) in the ordered sequence 712 of event-data values 714, e.g., j=i+l for sequences using incrementing indices, or j=i-1 for sequences using decrementing indices.” And Par [0235]; “At 812, the representation subsystem 246 can determine a token 814 in the ordered sequence 712 of event-data values 714 based at least in part on the first indicator 722(i) and the second indicator 722(j) satisfying the tokenization criterion 808.” Which discloses that the incremented indices are used to form tokens in an ordered sequence event data. This process creates subsets of one-step incremented event data that are used to form distinct event contexts.)
Poornaki, Balan, Martin and Nguyen are analogous art because all are concerned event-driven data processing and machine learning model training. Before the effective filing date of the claimed invention, it would have been obvious to one skilled in the art of machine learning and event based analysis to combine the one-step incremented event subset processing from Nguyen and the even based machine learning system of Poornaki to yield to the predictable result separating different events in the input data into one-step incremented event subsets; and forming each of the different event contexts from subsets of the one-step incremented event subsets. The motivation for doing so would be to predict, based on a character of training command-line text 606, an immediately following character in the training command-line text 606, within a predetermined accuracy. (Nguyen; Par [0215]).
Regarding claims 3 and 10, the rejection of claims 1,2 8 and 10 are incorporated but Poornaki and Balan in view of Martin fail to explicitly disclose but Nguyen discloses wherein the training the plurality of machine learning models for each of the different event contexts comprises training machine learning models for each of the subsets of the one-step incremented event subsets. (Par [0240]; “At 902, the training module 228 can determine a trained representation mapping 904 at least partly by adjusting parameters of a first model structure 906 (which can represent first model structure 604) so that the trained representation mapping 904 predicts, based on a character of training command-line text 908 ( or other value in event-data values 312), the immediately following character in the training command-line text 908 (or other value of training event-data values 326), within a predetermined accuracy.” Which discloses a training process of a model (adjusted based on set parameters) forming a different model structure with a next data point (following character of the input data/ in a one-step increment).)
Poornaki, Balan, Martin and Nguyen are analogous art because both are concerned event-driven data processing and machine learning model training. Before the effective filing date of the claimed invention, it would have been obvious to one skilled in the art of machine learning and event based analysis to combine the one-step incremented event subset processing from Nguyen and the even based machine learning system of Poornaki to yield to the predictable result of training the plurality of machine learning models for each of the different event contexts comprises training machine learning models for each of the subsets of the one-step incremented event subsets. The motivation for doing so would be to predicts, based on a character of training command-line text 606, an immediately following character in the training command-line text 606, within a predetermined accuracy. (Nguyen; Par [0215]).
As Claim 11-14, the Claims are rejected for the same reasons as Claims 4-7, respectively.
Conclusion
1. Claim Rejections under 35 U.S.C. §101
Applicant argues that the Claims implement a practical application because the Claims are directed to event based compound model architecture and one-step incremented sequential event data division (these improvements are disclosed in paragraph 0003, 0005, 0007 and 0009 of applicant’s original disclosure) (third paragraph of page 8 in the remarks).
Applicant’s arguments are persuasive. Therefore, 35 U.S.C. §101 rejection(s) are respectfully withdrawn.
2. Claim Rejections under 35 U.S.C. §103
As Claim 1 and 15, Applicant argues that current amendment(s) overcomes the cited reference Poornaki and Balan (last paragraph of page 9).
Applicant’s arguments are moot because new reference Martin discloses the amended limitations.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Upadhyay et al. (U.S. 2023/0068432) discloses method for estimating RUL of asset based on vehicle class and sub-class.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to NHAT HUY T NGUYEN whose telephone number is (571)270-7333. The examiner can normally be reached M-F: 12:00-8:00 EST.
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, Viker Lamardo can be reached at 571-270-5871. 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.
/NHAT HUY T NGUYEN/Primary Examiner, Art Unit 2147
Prosecution Insights