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 .
Status of Claims
Claims 1, 3, 8, 10, 15, and 17 have been amended.
Claims 1-20 are currently pending and addressed below.
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 as being directed to non-statutory subject matter because the claimed invention is directed to an abstract idea without reciting significantly more:
Step One: Does the Claim Fall Within a Statutory Category?
Yes. Claims 1-7 are directed towards a process (a method). Claims 8-14 are directed towards a machine (a computer program product comprising a computer-readable storage medium). Claims 15-20 are directed towards a machine (a system comprising one or more processors and a memory).
Step Two A, Prong One: Is a Judicial Exception Recited?
Yes. The claims recited are directed towards “applying one or more criteria to the flight sensor data to generate a training dataset comprising a plurality of first instances corresponding to flights of the plurality of flights; assigning, using component fault data, respective labels to the plurality of first instances”. These limitations represent abstract ideas, specifically, mental processes, that can be performed in the human mind because the claimed steps involve merely organizing and modifying data in a manner which may be performed mentally by a human, as set forth in further detail below.
The claims further recite “generating, for groups of one or more labels of the respective labels, a respective plurality of flight series, each flight series comprising a respective sequence of second instances that is based on some of the plurality of first instances, and that concludes with a second instance that is assigned a label included in the group, wherein the sequence of second instances is formed by adding noise to values of one or more respective features of the plurality of first instances such that the values are varied within a respective sensor resolution of a respective sensor of the plurality of sensors”. These limitations represent abstract ideas, specifically, mathematical concepts, that can be performed by taking existing information, manipulating the data using mathematical functions, and organizing this information into a new form (see at least MPEP 2106.04(a)(2)(I)(A)).
Step Two A, Prong Two: Is the Abstract Idea Integrated into a Practical Application?
No. As per claims 1, 8, and 15, the claims are broad enough to be performed mentally (or by pen and paper), and by using mathematical relationships. The step of applying one or more criteria to flight sensor data to generate a training dataset is broad enough to be performed by a human obtaining a set of flight data and partitioning a subset of the flight data to use as a training dataset, either mentally or with the aid of pen and paper. The step of assigning respective labels to a plurality of first instances in the training dataset is broad enough to be performed by a human labelling each of the first instances based on the respective data associated with each of the first instances, either mentally or with the aid of pen and paper.
The step of generating a plurality of flight series each comprising a respective sequence of second instances, wherein the sequence of second instances is formed by adding noise to values of one or more respective features of the plurality of first instances, is broad enough to be performed by manipulating existing data corresponding to the plurality of first instances using mathematical functions (e.g., by adding noise values which correspond to realistic sensor noise), to generate synthetic data which corresponds to the claimed second instances.
The step of receiving flight sensor data corresponding to a plurality of flights is broad enough to be performed by any well-known and conventional method for obtaining data, and amounts to insignificant, extra-solution activity in the form of routine data gathering.
As per claim 8, the structures recited in the claim include a computer-readable storage medium. As per claim 15, the structures recited in the claim include one or more processors and a memory. These structures perform functions which are insignificant, extra-solution activity in the form of mere data processing and storage and, therefore, do not amount to integration into a practical application. See MPEP 2106.05(g). These functions that the structures are performing are well-understood, routine and conventional activity and, therefore, the recitation of these structures does not amount to significantly more than the abstract idea. See Symantec, 838 F.3d at 1321, 120 USPQ2d at 1362 (utilizing an intermediary computer to forward information); TLI Communications LLC v. AV Auto. LLC, 823 F.3d 607, 610, 118 USPQ2d 1744, 1745 (Fed. Cir. 2016) (using a telephone for image transmission); OIP Techs., Inc., v. Amazon.com, Inc., 788 F.3d 1359, 1363, 115 USPQ2d 1090, 1093 (Fed. Cir. 2015) (sending messages over a network); buySAFE, Inc. v. Google, Inc., 765 F.3d 1350, 1355, 112 USPQ2d 1093, 1096 (Fed. Cir. 2014) (computer receives and sends information over a network); but see DDR Holdings, LLC v. Hotels.com, L.P., 773 F.3d 1245, 1258, 113 USPQ2d 1097, 1106 (Fed. Cir. 2014) ("Unlike the claims in Ultramercial, the claims at issue here specify how interactions with the Internet are manipulated to yield a desired result‐‐a result that overrides the routine and conventional sequence of events ordinarily triggered by the click of a hyperlink." (emphasis added)).
The technological components are recited at such a high level of generality to amount to a generic computer structure for performing generic computer functions which is not enough to integrate the abstract idea into a practical application or be significantly more than the abstract idea itself. Although technological elements are included for performing the claimed steps, a mental process performed in a computer environment is still considered a mental process (see at least MPEP 2106.04(a)(2)(III)(C)).
Step Two B: Does the Claim Provide an Inventive Concept
No. There are no additional elements recited in the independent claims that amount to significantly more than performing the abstract idea. The technological components recite the well-understood, routine, and conventional computing functions of mere data processing and storage. See MPEP 2106.05(d)(II). The claims do not require any particular technological elements besides generic computer components performing functions which are well-known by one of ordinary skill in the art.
It is noted that the claimed limitations set forth generating a training dataset, assigning labels to first instances in the training dataset, and generating flight series comprising a sequence of second instances based on the first instances, which, as noted above, can be done mentally and with the use of mathematical relationships, but none of the claimed limitations actively set forth wherein the generated flight series is subsequently applied in a manner which could be considered an inventive concept or integration in a practical application.
Dependent Claims
The dependent claims are merely further defining the abstract idea by providing additional limitations on how the abstract idea is performed, and are not adding anything to the abstract idea set forth in the independent claims such that the invention will amount to significantly more than the abstract idea.
As per claims 2, 9, and 16, generating a target number of copies of information is a well-known and conventional process which does not amount to integration into a practical application or an inventive concept.
As per claims 3, 10, and 17, adding noise values to data is a well-known and conventional mathematical transformation which does not amount to integration into a practical application or an inventive concept.
As per claims 4, 11, and 18, dropping (e.g., erasing, removing, and/or deleting) data is a well-known and conventional process which does not amount to integration into a practical application or an inventive concept.
As per claims 5-6, 12-13, and 19-20, the claimed RUL function amounts to an abstract idea in the form of a mathematical concept, and does not amount to integration into a practical application or an inventive concept.
As per claims 7 and 14, the generation of one or more cross-flight features may be performed by any well-known and conventional mathematical process to determine features from a dataset. For example, as recited in paragraph 0083 of the specification of the instant application, the cross-flight features may correspond to a median, a standard deviation, a moving average, a linear regression, or any other conventional statistical method that is well-known in the art.
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, 5-8, 12-15, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Chen et al. (US 2024/0391608), hereinafter referred to as Chen, in view of Nemani et al. (US 2024/0210913), hereinafter referred to as Nemani, and Donderici (US 2023/0382407), hereinafter referred to as Donderici. Chen, Nemani, and Donderici are considered analogous to the claimed invention because they are in the same field of generating training data sets.
Regarding claim 1, Chen teaches:
A method of generating a balanced training dataset for a machine learning model ("Training data is extended via generative GPR model with averaging techniques to produce realistic emulated data for training more accurate RUL prediction models." – see at least Chen: paragraph 0024);
the method comprising: receiving flight sensor data from a plurality of sensors corresponding to a plurality of flights ("In one embodiment, the method further includes obtaining sensing data from sensors of the aircraft engine" – see at least Chen: paragraph 0036) (The examiner notes that Chen teaches that sensor readings span over a life cycle of an aircraft engine ("On the one hand, in the training trajectories of the training dataset, historical run-to-failure sensor measurements of entire engine may be available along with entire life cycle until the engine totally fails. On the other hand, the testing trajectories of the testing dataset may contain sensor measurements that are truncated at certain time cycle before the engine failure, so that the RUL may be predicted at an earlier time based on given limited sensor measurements." – see at least Chen: paragraph 0047). As such, one of ordinary skill in the art would recognize that the life cycle of an engine is typically expected to span over a plurality of flights, and that the flight sensor data as taught by Chen would therefore include data from a plurality of flights);
applying one or more criteria to the flight sensor data to generate a training dataset comprising a plurality of first instances corresponding to flights of the plurality of flights ("In S100, observation data is partitioned into training data, validation data, and testing data." – see at least Chen: paragraph 0028);
assigning, using component fault data, respective labels to the plurality of first instances ("For each segment, the RUL label of the last data sample in the segment may be considered as the label of the segment. After preprocessing the dataset, the segment of sensor readings and corresponding labels may be fed into the deep learning models for training." – see at least Chen: paragraph 0063) (The examiner notes that RUL as taught by Chen refers to Remaining Useful Life, which is based on aircraft component health status and failure thresholds, which corresponds to the claimed component fault data ("In one embodiment, RUL is calculated as a first passage time when a health status value of the aircraft engine exceeds a predefined failure threshold." – see at least Chen: paragraph 0040));
and generating, for groups of one or more labels of the respective labels, a respective plurality of flight series ("Readings from monitoring sensors of the aircraft engine may be highly correlated to the health condition of the aircraft engine. Assume that N sensors are employed for monitoring the aircraft engine. The time series sensor readings at jth time cycle are denoted as Xj={X1j, X2j, . . . Xij=1, 2, . . . , N}, where i denotes the index of a sensor, j denotes the time cycle when the sensor data is recorded. The health status at jth time cycle is denoted as Yj." – see at least Chen: paragraph 0044),
each flight series comprising a respective sequence of second instances that is based on some of the plurality of first instances, and that concludes with a second instance that is assigned a label included in the group ("The learned GPR model may serve as a synthetic data generator, which can generate data samples that behave similarly to original training data according to learned patterns from the GPR model training." – see at least Chen: paragraph 0042) (The examiner notes that the generated synthetic data based on the original training data as taught by Chen corresponds to the claimed sequence of second instances. One of ordinary skill in the art would recognize that since the generated synthetic data samples of Chen behave similarly to the corresponding original training data, it is expected that the labels of the generated synthetic data samples would include the same set of labels as the original training data (i.e., RUL labels as set forth above)).
Chen does not explicitly disclose, but Nemani teaches:
wherein the sequence of second instances is formed by adding noise to values of one or more respective features of the plurality of first instances ("When very little data is available and is noisy, the LSTM forecast can almost be flat especially near the onset of bearing degradation. By using data augmentation of duplicating the training data with added Gaussian noise, we observe the forecast to be much more intuitive and stable. To this end, for the XJTU-SY bearing dataset, we add Gaussian noise to V0.2ω-sf/2RMS as a simple data augmentation technique similar to Refs. [71,72]." – see at least Nemani: paragraph 0130).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with these above aforementioned teachings from Nemani such that the sequence of second instances is formed by adding noise to values of one or more respective features of the plurality of first instances. At the time of the effective filing date of the claimed invention, one of ordinary skill in the art would have been motivated to incorporate Nemani’s method of duplicating training data with added Gaussian noise with Chen’s method of using deep learning to predict the RUL of an aircraft engine in order to augment datasets with additional data, in particular for situations in which data associated with certain labels are limited or underrepresented (“Particularly in the bearing prognostic scenario, we find the following challenges: (1) very noisy feature data, (2) limited training data, and (3) most of the training data is in the domain pertaining to a healthy bearing suppressing learning from the bearing degradation domain. Although the third scenario can be tackled by considering only the bearing degradation data for training the LSTM network, this further accentuates the second problem of limited data. The use of data augmentation is particularly useful to address this aspect for a stable forecast. To demonstrate this, we use a simple toy example of linear degradation with noise to train and test an LSTM network as shown in FIG. 9B.” – see at least Nemani: paragraph 0130). Doing so would provide the benefit of utilizing data augmentation to improve the accuracy and robustness of the RUL prediction model (“A simple data augmentation technique is used during the training phase of the LSTM networks to improve the accuracy and robustness of RUL prediction.” – see at least Nemani: paragraph 0069).
The examiner notes that the claimed methods of generating training data (e.g., by adding noise to copies of original data) are well-known to one of ordinary skill in the art and may be readily applied to a variety of machine learning applications. See the Conclusion section of the Non-Final Rejection filed 12 January 2026 for further detail on pertinent prior art references which provide additional evidence that the claimed methods of generating training data are well known by one of ordinary skill in the art.
Chen does not explicitly disclose, but Donderici teaches:
such that the values are varied within a respective sensor resolution of a respective sensor of the plurality of sensors (“For instance, the simulated onboard sensor suites may generate simulated sensor data of the virtual objects, e.g., images, audio, depth information, location information, and so on. In some embodiments, the simulation module 340 determines a fidelity for a sensor in the virtual onboard sensor suite. The fidelity of a sensor indicates a quality of the sensor, such as accuracy, precision, amount of noise, resolution, etc.” – see at least Donderici: paragraph 0052).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with these above aforementioned teachings from Donderici such that the values are varied within a respective sensor resolution of a respective sensor of the plurality of sensors. At the time of the effective filing date of the claimed invention, one of ordinary skill in the art would have been motivated to incorporate Donderici’s method of generating simulated sensor data with Chen’s method of using deep learning to predict the RUL of an aircraft engine in order to generate simulated data that is similar in precision and accuracy to data obtained by an actual sensor (“The simulation module 340 may also generate a simulated AV, which is a virtual AV that simulates an AV 110... An accuracy of the sensor may be a difference between an actual value and a value captured by the sensor. For a sensor with a higher fidelity, the sensor is more accurate, and the difference between the actual value and the captured value is less. The simulation module 340 may select a fidelity for a sensor from pre-determined fidelity options.” – see at least Donderici: paragraph 0052). Doing so would provide the benefit of allowing the simulated data to be used to train models related to vehicle performance (“As another example, the simulator datastore 330 stores data collected from simulations run by the simulation module 340, such data to be used for training AV control models, data to be used to evaluate AV performances, and so on.” – see at least Donderici: paragraph 0047).
The examiner acknowledges that Donderici is directed toward a different application of machine learning than the instant application. However, the same benefits of generating additional training data (e.g., simulated sensor data which is similar to data collected by a real sensor) are realized regardless of the particular task for which machine learning is being applied. These methods are well-known to one of ordinary skill in the art and may be readily applied to a variety of machine learning applications. See the Response to Arguments section of this Office Action for further detail on pertinent prior art references which provide additional evidence that the claimed methods of generating training data are well known by one of ordinary skill in the art.
Regarding claim 5, Chen in view of Nemani and Donderici teaches all of the elements of the current invention as stated above. Chen further teaches:
wherein assigning respective labels to the plurality of first instances is according to a remaining useful life (RUL) function for an aircraft component ("In order to mitigate such issue, the RUL may be assigned based on a suitable degradation model such as a piece-wise linear degradation model. As shown in FIG. 6, the piece-wise linear degradation model assumes that the RUL target function limits the maximum value of the RUL until a threshold is reached, and after the threshold, the RUL value may decrease linearly with usage. This is based on the fact that the aircraft engine typically works normally and the health status has negligible degradation in early age. After a certain degree of wear or usage, the aircraft engine may start to degrade, which is consistent with a typical degradation model." – see at least Chen: paragraph 0063).
Regarding claim 6, Chen in view of Nemani and Donderici teaches all of the elements of the current invention as stated above. Chen further teaches:
wherein assigning respective labels to the plurality of first instances comprises: applying a clipped linear function to the RUL function, such that values of the RUL function between an upper threshold and a lower threshold are assigned linearly interpolated values as the respective labels ("Since the physics-based degradation model is difficult to be obtained, the piecewise linear model may be the most common choice for the RUL target function. In the C-MAPSS dataset (e.g., model), the maximum RUL may be configured as 125 for the dataset." – see at least Chen: paragraph 0061) (The examiner notes that Fig. 6 of Chen as shown below illustrates a linear degradation model which corresponds to the claimed clipped linear function, wherein Fig. 6 of Chen illustrates a linear function with an upper threshold value of 125 corresponding to a maximum RUL value, and a lower threshold value of 0 corresponding to the RUL value when the health status of the engine surpasses a predefined failure threshold).
PNG
media_image1.png
467
582
media_image1.png
Greyscale
Regarding claim 7, Chen in view of Nemani and Donderici teaches all of the elements of the current invention as stated above. Chen further teaches:
further comprising: generating one or more cross-flight features for the plurality of flight series ("For RUL prediction, the DL techniques including DNN, CNN, LSTM, and BNN have been widely used to model the sensor features. For instance, CNN is applied to extract the features of the sensor data for RUL prediction. The multivariate time series of the sensor data may be convolved with multiple convolutional kernels which may be further learned during the training process. The architecture of the CNN for RUL prediction is shown in FIG. 5B. The predicted RUL may be obtained from the output of the trained CNN model. LSTM may be used to build long-term time dependencies for modeling the sensor data features." – see at least Chen: paragraph 0055).
Regarding claim 8, this claim is substantially similar to claim 1 and is, therefore, rejected in the same manner as claim 1 as has been set forth above.
Regarding claim 12, this claim is substantially similar to claim 5 and is, therefore, rejected in the same manner as claim 5 as has been set forth above.
Regarding claim 13, this claim is substantially similar to claim 6 and is, therefore, rejected in the same manner as claim 6 as has been set forth above.
Regarding claim 14, this claim is substantially similar to claim 7 and is, therefore, rejected in the same manner as claim 7 as has been set forth above.
Regarding claim 15, this claim is substantially similar to claim 1 and is, therefore, rejected in the same manner as claim 1 as has been set forth above.
Regarding claim 19, this claim is substantially similar to claim 5 and is, therefore, rejected in the same manner as claim 5 as has been set forth above.
Regarding claim 20, this claim is substantially similar to claim 6 and is, therefore, rejected in the same manner as claim 6 as has been set forth above.
Claims 2-3, 9-10, and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Chen in view of Nemani and Donderici, further in view of Kung (US Patent 12,361,138), hereinafter referred to as Kung. Kung is considered analogous to the claimed invention because they are in the same field of generating training data sets.
Regarding claim 2, Chen in view of Nemani and Donderici teaches all of the elements of the current invention as stated above. Chen does not explicitly disclose, but Kung teaches:
wherein generating the respective plurality of flight series comprises, for a first group of the groups: determining a count of those first instances, of the plurality of first instances, that have a label included in the first group; determining a scale factor based on a quotient of a target number of flight series and the count of the first instances; and generating a scale factor number of copies of each of the first instances having a label included in the first group ("For example, each supported product may have a predetermined target number of labeled sample data for training. For each supported product, the labeled sample data may be reduced if the number of labeled sample data exceeds the target number, or augmented if the number of labeled training data is below the target number. For example, assuming a target number of 200, 200 application names may be randomly selected from labeled sample data greater than 200 samples. For labeled sample data less than 200 samples, additional application names may be added by creating synthetic names" – see at least Kung: Column 11 lines 56-66) (The examiner notes that the factor by which labeled sample data is augmented to reach a target number of labeled samples as taught by Kung corresponds to the claimed scale factor).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with these above aforementioned teachings from Kung such that generating the respective plurality of flight series comprises, for a first group of the groups: determining a count of those first instances, of the plurality of first instances, that have a label included in the first group; determining a scale factor based on a quotient of a target number of flight series and the count of the first instances; and generating a scale factor number of copies of each of the first instances having a label included in the first group. At the time of the effective filing date of the claimed invention, one of ordinary skill in the art would have been motivated to incorporate Kung’s method of augmenting a data set to include a target number of labeled sample data with Chen’s method of using deep learning to predict the RUL of an aircraft engine in order to balance a training dataset with a plurality of different labels (“In step 304, a training dataset is selected from a corresponding labeled sample data. The goal of training dataset selection is to have a similar number of training data for each supported product to avoid imbalance. For example, each supported product may have a predetermined target number of labeled sample data for training.” – see at least Kung: Column 11 lines 53-58). Doing so would provide the benefit of preventing overfitting to labels which are overrepresented in the dataset (“That is, instead of repeating exactly the same application name, using a version number that is consistent with, but different from existing versions, will help reduce the risk of overfitting.” – see at least Kung: Column 12 lines 2-5).
The examiner acknowledges that Kung is directed toward a different application of machine learning than the instant application. However, the same benefits of augmenting training data to generate a target number of data samples (e.g., balancing datasets and avoiding overfitting) are realized regardless of the particular task for which machine learning is being applied. These methods are well-known to one of ordinary skill in the art and may be readily applied to a variety of machine learning applications. See the Conclusion section of the Non-Final Rejection filed 12 January 2026 for further detail on pertinent prior art references which provide additional evidence that the claimed methods of generating training data are well known by one of ordinary skill in the art.
Regarding claim 3, Chen in view of in view of Nemani, Donderici, and Kung teaches all of the elements of the current invention as stated above. Chen does not explicitly disclose, but Nemani teaches:
wherein generating the respective plurality of flight series further comprises: forming the sequence of second instances, wherein forming the sequence of second instances comprises: adding the noise to the values of the one or more respective features of the scale factor number of copies of each of the first instances ("When very little data is available and is noisy, the LSTM forecast can almost be flat especially near the onset of bearing degradation. By using data augmentation of duplicating the training data with added Gaussian noise, we observe the forecast to be much more intuitive and stable. To this end, for the XJTU-SY bearing dataset, we add Gaussian noise to V0.2ω-sf/2RMS as a simple data augmentation technique similar to Refs. [71,72]." – see at least Nemani: paragraph 0130).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with these above aforementioned teachings from Nemani such that generating the respective plurality of flight series further comprises: forming the sequence of second instances, wherein forming the sequence of second instances comprises: adding noise to values of one or more respective features of the scale factor number of copies of each of the first instances. At the time of the effective filing date of the claimed invention, one of ordinary skill in the art would have been motivated to incorporate Nemani’s method of duplicating training data with added Gaussian noise with Chen’s method of using deep learning to predict the RUL of an aircraft engine in order to augment datasets with additional data, in particular for situations in which data associated with certain labels are limited or underrepresented (“Particularly in the bearing prognostic scenario, we find the following challenges: (1) very noisy feature data, (2) limited training data, and (3) most of the training data is in the domain pertaining to a healthy bearing suppressing learning from the bearing degradation domain. Although the third scenario can be tackled by considering only the bearing degradation data for training the LSTM network, this further accentuates the second problem of limited data. The use of data augmentation is particularly useful to address this aspect for a stable forecast. To demonstrate this, we use a simple toy example of linear degradation with noise to train and test an LSTM network as shown in FIG. 9B.” – see at least Nemani: paragraph 0130). Doing so would provide the benefit of utilizing data augmentation to improve the accuracy and robustness of the RUL prediction model (“A simple data augmentation technique is used during the training phase of the LSTM networks to improve the accuracy and robustness of RUL prediction.” – see at least Nemani: paragraph 0069).
Regarding claim 9, this claim is substantially similar to claim 2 and is, therefore, rejected in the same manner as claim 2 as has been set forth above.
Regarding claim 10, this claim is substantially similar to claim 3 and is, therefore, rejected in the same manner as claim 3 as has been set forth above.
Regarding claim 16, this claim is substantially similar to claim 2 and is, therefore, rejected in the same manner as claim 2 as has been set forth above.
Regarding claim 17, this claim is substantially similar to claim 3 and is, therefore, rejected in the same manner as claim 3 as has been set forth above.
Claims 4, 11, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Chen in view of Nemani, Donderici, and Kung, further in view of Shalaby et al. (US 2022/0187819), hereinafter referred to as Shalaby. Shalaby is considered analogous to the claimed invention because they are in the same field of generating training data sets for predicting remaining useful life.
Regarding claim 4, Chen in view of Nemani, Donderici, and Kung teaches all of the elements of the current invention as stated above. Chen does not explicitly disclose, but Shalaby teaches:
wherein forming the sequence of second instances further comprises: dropping one or more second instances from an initial sequence of second instances ("To this end, example implementations involve various techniques for augmenting the data with synthetic samples from the available samples using: 1) dropout of events/subsequences within the sequence" – see at least Shalaby: paragraph 0027).
It would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to have modified Chen with these above aforementioned teachings from Shalaby such that forming the sequence of second instances further comprises: dropping one or more second instances from an initial sequence of second instances. At the time of the effective filing date of the claimed invention, one of ordinary skill in the art would have been motivated to incorporate Shalaby’s method of dropping out events within a sequence of events with Chen’s method of using deep learning to predict the RUL of an aircraft engine in order to create a more diverse set of training data (“Example implementations involve techniques for data augmentation to increase the diversity of data available for training and to improve the machine learning model generalization.” – see at least Shalaby: paragraph 0027). Doing so would provide the benefit of improving a RUL estimation model by augmenting scarce training data (“Example implementations described herein involve a methodology for failure prediction and remaining useful life (RUL) estimation on event-based sequential data. The example implementations include: 1) Techniques for data augmentation to handle scarcity of event-based failure data” – see at least Shalaby: paragraph 0009).
The examiner notes that the claimed methods of generating training data (e.g., by dropping a subset of copies of original data) are well-known to one of ordinary skill in the art and may be readily applied to a variety of machine learning applications. See the Conclusion section of the Non-Final Rejection filed 12 January 2026 for further detail on pertinent prior art references which provide additional evidence that the claimed methods of generating training data are well known by one of ordinary skill in the art.
Regarding claim 11, this claim is substantially similar to claim 4 and is, therefore, rejected in the same manner as claim 4 as has been set forth above.
Regarding claim 18, this claim is substantially similar to claim 4 and is, therefore, rejected in the same manner as claim 4 as has been set forth above.
Response to Arguments
Applicant’s arguments filed 13 April 2026 with respect to claims 1-20 have been considered but are moot in view of the new grounds of rejection based on the teachings of the newly relied upon reference by Donderici, which has been introduced to address the amended claims.
In particular, the amended claims add limitations which recite wherein the sequence of second instances is formed by adding noise to values of one or more respective features of the plurality of first instances such that the values are varied within a respective sensor resolution of a respective sensor of the plurality of sensors. The examiner acknowledges that while the previously applied references by Chen, Nemani, Kung and Shalaby teach relevant aspects of the claimed invention, as set forth in further detail in the Non-Final Rejection filed 12 January 2026, these references do not explicitly teach these limitations of the amended independent claims. At best, Nemani teaches adding Gaussian noise to duplicated training data in order to augment the training data, but Nemani does not explicitly teach that the added noise is varied within a respective sensor resolution of a respective sensor, as required by the amended claims.
To address the Applicant’s Response, the examiner has introduced the reference by Donderici which teaches methods of generating simulated data for training models related to vehicle performance, wherein the simulated data is generated using virtual sensors with variable levels of fidelity, and wherein the level of fidelity of the virtual sensor may be selected such that the characteristics of the virtual sensor (e.g., accuracy, precision, amount of noise, and resolution) are similar to the characteristics of the actual sensors of a vehicle (see at least Donderici: paragraph 0052). As set forth in further detail above in the section for Claim Rejections under 35 U.S.C. 103, these teachings from Donderici are considered to cure the aforementioned deficiencies of the previously applied prior art with regard to the amended claims.
As set forth in further detail above, the instant claims are considered to be an obvious variation of the teachings of the primary reference by Chen, in that the teachings of Chen may be readily modified using data augmentation techniques that are well-known in the art (e.g., those taught by Nemani, Donderici, Kung, Shalaby, and the additional references noted in the Conclusion section of the Non-Final Rejection filed 12 January 2026) in order to reach the claimed invention. Additional newly cited references which provide further teachings regarding relevant data augmentation techniques include:
Nakanoya (US 2024/0087295), which teaches in at least paragraph 0049 a method of generating pseudo data which is similar to actual observation data, wherein the pseudo data includes added characteristics such as fluctuation, noise, and resolution similar to a real sensor.
White et al. (US 2020/0302241) teaches in at least paragraph 0027 using sensors that are simulated to have similar characteristics to real-world sensors, including the sensor’s field of view, resolution, RGB image noise, and error in depth measurements, which provides the benefit of making the data generated by the simulated sensors realistic and variable enough to train a ML model in a way that improves performance in a real-world environment.
Hassan-Shafique et al. (US 2025/0306594) teaches in at least paragraphs 0112-0114 generating simulated sensor data which applies sensor models that account for the characteristics and limitations of each sensor type, such as noise, resolution, and detection range, such that the simulated sensor data may closely resemble the actual data that would be obtained by the physical sensors in real-world conditions, and wherein the collected simulated sensor data is used to train and refine machine learning models.
Donderici, as well as each of the aforementioned additional newly cited references, teach aspects of generating simulated training data by adding noise values which are varied in a manner which is based on the actual performance of the sensor that is being simulated, in order to generate additional training data which takes into account the characteristics of the particular type of sensor being simulated (“In some embodiments, the simulation module 340 determines a fidelity for a sensor in the virtual onboard sensor suite. The fidelity of a sensor indicates a quality of the sensor, such as accuracy, precision, amount of noise, resolution, etc. An accuracy of the sensor may be a difference between an actual value and a value captured by the sensor. For a sensor with a higher fidelity, the sensor is more accurate, and the difference between the actual value and the captured value is less.” – see at least Donderici: paragraph 0052). These teachings from the aforementioned prior art are considered to provide substantially the same benefits as presented in the Applicant’s Remarks (e.g., “For example, features related to data from one type of sensor may have a different scale of noise added compared to features related to data from a different type of sensor. Sensors with relatively high resolution may have smaller or more incremental amounts of noise added”). Therefore, claims 1-20 as currently presented are rejected under 35 U.S.C. 103.
As per the pending rejections under 35 U.S.C. 101, the claims as amended are not considered to sufficiently integrate the abstract idea into a practical application or an inventive concept. The amended claims add the step of "adding noise to values… such that the values are varied within a respective sensor resolution". This feature is considered to merely further define the abstract idea by reciting a particular mathematical relationship for taking existing information, manipulating the data using mathematical functions, and organizing this information into a new form (see at least MPEP 2106.04(a)(2)(I)(A)).
The Applicant asserts that this step recites an improvement in the current technology of generating training data for machine learning applications. However, as set forth in further detail above, this step is considered to be a form of manipulating data which is well-known in the prior art for generating training data that is similar to data collected by real sensors (see at least Donderici: paragraph 0052). As such, the claims are not considered to recite a particular unconventional technical solution nor a technological improvement over the prior art (see at least MPEP 2106.05(a)). Therefore, claims 1-20 remain rejected under 35 U.S.C. 101.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DOMINICK ANTHONY MULDER whose telephone number is (571)272-3610. The examiner can normally be reached Monday - Friday 9:00am - 5:00pm.
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, RAMYA P BURGESS can be reached at (571)272-6011. 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.
/D.M./Examiner, Art Unit 3667
/TUAN C TO/Primary Examiner, Art Unit 3661
Prosecution Insights