DETAILED ACTION
This office action is in response to application filed on May 6, 2024.
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 .
Information Disclosure Statement
The information disclosure statements (IDS) submitted on 05/06/2024 and 11/24/2025 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Claim Objections
Claim 1 is objected to because of the following informalities:
Claim language “A method of inspecting a component, comprising:” should read “A method of inspecting a component[[,]] comprising:” in order to correct minor informalities.
Claim language “using a transducer to transmit a first signal into a component comprising a solid metallic material” should read “using a transducer to transmit a first signal into [[a]]the component comprising a solid metallic material” in order to provide appropriate antecedence basis.
Appropriate correction is required.
Claim 6 is objected to because of the following informalities:
Claim language should read “The method of claim 3, wherein the augmentation invariance pretext training of the self-supervised machine learning technique includes determining a contrastive loss between a processed .
Appropriate correction is required.
Claim 7 is objected to because of the following informalities:
Claim language should read “The method of claim 3, wherein the processed .
Appropriate correction is required.
Claim 9 is objected to because of the following informalities:
Claim language should read “The method of claim [[7]]8, wherein the training of the self-supervised machine learning technique using the masked reconstruction pretext includes determining a reconstructive loss” in order to provide appropriate dependency.
Appropriate correction is required.
Claim 10 is objected to because of the following informalities:
Claim language should read “The method of claim 1, wherein .
Appropriate correction is required.
Claim 11 is objected to because of the following informalities:
Claim language “A method of inspecting a rotor disk for a defect, comprising” should read “A method of inspecting a rotor disk for a defect, the method comprising:” in order to correct minor informalities.
Claim language “using a transducer to transmit a first signal into a rotor disk of a gas turbine engine, the rotor disk comprising a solid metallic material” should read “using a transducer to transmit a first signal into [[a]]the rotor disk of a gas turbine engine, the rotor disk comprising a solid metallic material” in order to provide appropriate antecedence basis.
Claim language “using the transducer to sense the rotor disk for a second signal produced as a result of the first signal being transmitted into the component, and produce a transducer response signal representative of the second signal” should read “using the transducer to sense the rotor disk for a second signal produced as a result of the first signal being transmitted into the rotor disk, and produce a transducer response signal representative of the second signal” in order to provide appropriate antecedence basis.
Appropriate correction is required.
Claim 12 is objected to because of the following informalities:
Claim language should read “The method of claim 11, wherein the processing .
Appropriate correction is required.
Claim 13 is objected to because of the following informalities:
Claim language should read “The method of claim 12, wherein the processing .
Appropriate correction is required.
Claim 14 is objected to because of the following informalities:
Claim language should read “The method of claim 13, wherein the processing .
Appropriate correction is required.
Claim 16 is objected to because of the following informalities:
Claim language “A component inspection system, comprising:” should read “A component inspection system[[,]] comprising:” in order to correct minor informalities.
Claim language “a controller in communication with the signal transmitter, the signal receiver, and a non-transitory memory storing instructions, which instructions when executed cause the controller to:” should read “a controller in communication with the signal transmitter, the signal receiver, and a non-transitory memory storing instructions, wherein the instructions when executed cause the controller to:” in order to correct minor informalities.
Claim language “control the signal receiver to sense the component for a second ultrasonic signal and produce a response signal representative of the second ultrasonic signal;” should read “control the signal receiver to sense the component for a second ultrasonic signal and produce a response signal representative of the second ultrasonic signal; and” in order to correct minor informalities.
Appropriate correction is required.
Claim 19 is objected to because of the following informalities:
Claim language should read “The method of claim 18, wherein the training of the self-supervised machine learning technique using the augmentation invariance pretext includes determining a contrastive loss between a processed .
Appropriate correction is required.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception without significantly more.
Regarding claim 1, the examiner submits that under Step 1 of the 2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence (see also 2019 Revised Patent Subject Matter Eligibility Guidance) for evaluating claims for eligibility under 35 U.S.C. 101, the claim is to a process, which is one of the statutory categories of invention.
Continuing with the analysis, under Step 2A - Prong One of the test:
the limitation “processing the response signal to determine a presence or an absence of a defect in the solid metallic material of the component” is a process that, under its broadest reasonable interpretation in light of the specification, covers performance of the limitation using mental processes to manipulate data and obtain a result (i.e., a presence or an absence of a defect; see specification at [0046]-[0047]). Except for the recitation of the particular technological environment or field of use (e.g., equipment inspection), the limitation in the context of the claim mainly refers to performing a mental evaluation or judgment of data for identification purposes (e.g., identify defects).
Therefore, the claim recites a judicial exception under Step 2A - Prong One of the test.
Furthermore, under Step 2A - Prong Two of the test, this judicial exception is not integrated into a practical application when considering the claim as a whole. In particular, the additional elements recited in the claim:
“A method of inspecting a component” generally links the use of the judicial exception to a particular technological environment or field of use (see specification at [0001], [0028]) (see MPEP 2106.05(h));
“using a transducer to transmit a first signal into a component comprising a solid metallic material” adds extra-solution activities (e.g., mere data transmission) using elements recited at a high level of generality (i.e., transducer, a component comprising a solid metallic material, see specification at [0028], [0036]-[0038]) (see MPEP 2106.05(g));
“using the transducer to sense the component for a second signal produced as a result of the first signal being transmitted into the component, and produce a response signal representative of the second signal” adds extra-solution activities (e.g., mere data gathering) using elements recited at a high level of generality (i.e., transducer, component, see specification at [0028], [0036]-[0038]) (see MPEP 2106.05(g)); and
“the processing using a controller configured with a self-supervised machine learning technique that is trained to be invariant to a component variability portion of the response signal” adds the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer (e.g., mere computer implementation using a self-supervised machine learning technique, see specification at [0043]-[0044]), or merely uses a computer (e.g., controller, see specification at [0039]) as a tool to perform an abstract idea (see MPEP 2106.05(f)).
Accordingly, these additional elements, when considered individually and in combination, do not integrate the judicial exception into a practical application because they do not impose any meaningful limits on practicing the abstract idea when considering the claim as a whole. The claim is directed to a judicial exception under Step 2A of the test.
Additionally, under Step 2B of the test, the claim, when considered as a whole, does not include additional elements that, when considered individually and in combination, are sufficient to amount to significantly more than the judicial exception because the additional elements:
generally link the use of the judicial exception to a particular technological environment or field of use (i.e., component inspection, see specification at [0001], [0028]), which as indicated in the MPEP: “As explained by the Supreme Court, a claim directed to a judicial exception cannot be made eligible “simply by having the applicant acquiesce to limiting the reach of the patent for the formula to a particular technological use.” Diamond v. Diehr, 450 U.S. 175, 192 n.14, 209 USPQ 1, 10 n. 14 (1981). Thus, limitations that amount to merely indicating a field of use or technological environment in which to apply a judicial exception do not amount to significantly more than the exception itself, and cannot integrate a judicial exception into a practical application” (see MPEP 2106.05(h));
recite extra-solution activities (i.e., mere data gathering/transmission) using elements (i.e., transducer, component, see specification at [0028], [0036]-[0038]) specified at a high level of generality, which as indicated in the MPEP: “Another consideration when determining whether a claim integrates the judicial exception into a practical application in Step 2A Prong Two or recites significantly more in Step 2B is whether the additional elements add more than insignificant extra-solution activity to the judicial exception. The term “extra-solution activity” can be understood as activities incidental to the primary process or product that are merely a nominal or tangential addition to the claim. Extra-solution activity includes both pre-solution and post-solution activity. An example of pre-solution activity is a step of gathering data for use in a claimed process” (see MPEP 2106.05(g)), and “Use of a machine that contributes only nominally or insignificantly to the execution of the claimed method (e.g., in a data gathering step or in a field-of-use limitation) would not provide significantly more” (see MPEP 2106.05(b)); and
append generic computer components (i.e., controller, see specification at [0039]) used to facilitate the application of the abstract idea (e.g., mere computer implementation using machine learning, see specification at [0043]-[0044]), which as indicated in the MPEP: “Use of a computer or other machinery in its ordinary capacity for economic or other tasks (e.g., to receive, store, or transmit data) or simply adding a general purpose computer or computer components after the fact to an abstract idea (e.g., a fundamental economic practice or mathematical equation) does not provide significantly more” (see MPEP 2106.05(f), item 2).
The claim, when considered as a whole, does not provide significantly more under Step 2B of the test.
Based on the analysis, the claim is not patent eligible.
Similarly, independent claims 11 and 16 are directed to a judicial exception (abstract idea, Step 2A – Prong One) without integrating the judicial exception into a practical application (Step 2A – Prong Two) and/or without providing significantly more (Step 2B) when considering the claimed invention as a whole, as explained above with regards to claim 1.
With regards to the dependent claims they are also directed to the non-statutory subject matter because:
they just extend the abstract idea of the independent claims by additional limitations (Claims 6-7, 9, 12-14 and 19-20), that under the broadest reasonable interpretation in light of the specification, cover performance of the limitations using mental processes (e.g., evaluations, judgments, opinions), and
the additional elements recited in the dependent claims, when considered individually and in combination, refer to extra-solution activities (e.g., mere data gathering/transmission using a data type or source), generic computer components/implementation and/or field of use (Claims 2-5, 7-8, 10, 12, 15, 17-18 and 20), which as indicated in the Office’s guidance does not integrate the judicial exception into a practical application (Step 2A – Prong Two) and/or does not provide significantly more (Step 2B) when considering the claimed invention as a whole.
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
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 1-2, 10 and 16-17 are rejected under 35 U.S.C. 103 as being unpatentable over Loic (US 20230359869 A1), hereinafter ‘Loic’, in view of Ha (US 20250334550 A1), hereinafter ‘Ha’.
Regarding claim 1.
Loic discloses:
A method of inspecting a component (Fig. 1, item 901 – “monitored target”; [0028], [0033]: an equipment anomaly detection method for detecting anomalies using signals of a monitored target is presented (see also [0001]-[0002])), comprising:
using the transducer (Fig. 1, item 111-1 – “sensor”) to sense the component for a second signal (Fig. 1, item 902 – “vibration signals”), and produce a response signal (Fig. 1, item 112 – “sensor signals”) representative of the second signal ([0034]: a sensor system includes sensors for detecting vibration signals of the monitored target and generating a stream of sensor signals representing the vibration signals); and
processing the response signal to determine a presence or an absence of a defect of the component, the processing using a controller configured with a self-supervised machine learning technique that is trained ([0034]: the sensor signals are converted into input event streams (Fig. 1, item 190; see also [0040]), which are used to generate reproduced event streams (Fig. 1, item 149) using self-supervised machine learning (i.e., embedding network and decoding network; see Fig. 1, items 130 and 140, [0011], [0052], [0082]) by an equipment anomaly detection device (see Fig. 1, item 101, [0037]-[0038], [0119]), with both input event streams and reproduced event streams being compared to determine anomalies in the monitored target (see also [0001]-[0002], [0086], [0124])).
Loic does not explicitly disclose (see italic text):
using a transducer to transmit a first signal into a component comprising a solid metallic material;
a second signal produced as a result of the first signal being transmitted into the component;
processing the response signal to determine a presence or an absence of a defect in the solid metallic material of the component; and
a self-supervised machine learning technique that is trained to be invariant to a component variability portion of the response signal.
Regarding “using a transducer to transmit a first signal into a component; and a second signal produced as a result of the first signal being transmitted into the component”, Ha teaches:
“The present disclosure relates to a non-destructive inspection method, and particularly, to a non-destructive inspection method and system based on self-supervised
learning, which detect the inside of an inspection object in a non-destructive way by using ultrasonic waves and also predict the depth of a defect through self-supervised learning” ([0001]: a non-destructive inspection method based on self-supervised learning includes emitting ultrasonic waves into an inspection object, and collecting returning signals from the inspection objection (see [0009], [0046]); examiner notes that signal emission implies the use of a transmitter or transducer).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to use a transducer to transmit a first signal into a component, and to obtain a second signal produced as a result of the first signal being transmitted into the component, in order to implement commonly non-destructive inspection techniques to detect defects in an object, as discussed by Ha ([0002], [0056]).
Regarding “a component comprising a solid metallic material; and processing the response signal to determine a presence or an absence of a defect in the solid metallic material of the component”, Loic teaches:
“A monitored target system 901 may include one or more entities, such as one or more industrial plants, motors, devices, facilities, or systems” ([0033]: monitored target includes industrial plants, motors, devices; examiner notes that equipment in industrial plants and motors may include solid metallic material (see also [0002])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to incorporate a component comprising a solid metallic material, and to process the response signal to determine a presence or an absence of a defect in the solid metallic material of the component, in order to identify common defects (e.g., corrosion, cracks, wear) in typical components used in industrial or residential machines that may rapidly affect the operation of the component and reduce its lifetime.
Regarding “a self-supervised machine learning technique that is trained to be invariant to a component variability portion of the response signal”, Loic further teaches:
“The embedding network 130 are trained through unsupervised learning. Training data are a small subset of the full data set, that supposedly do not contain any anomalies. The embodiment of the disclosure allows to learn an alternative representation of normal input signals by an encoder, that is the embedding network 130, and to accurately reconstruct normal input signals from such representation by a decoder, that is the decoding network 140. The basic principle behind is that the auto-encoder never learns to reconstruct abnormal signals. Thus, the reconstruction loss is small during normal conditions, and this loss is large in presence of abnormal signals since the auto encoder never learned to reconstruct abnormal signals” ([0052]: during training, the embedding network learns from a subset of data not containing anomalies, with the reconstruction loss being small during normal conditions; examiner notes that by allowing a small reconstruction loss, variability of signal during normal operations is taken into account without affecting model performance (see also [0048] regarding different normal sub-states)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to train a self-supervised machine learning technique to be invariant to a component variability portion of the response signal, in order to provide a more robust detection technique that takes into account typical signal variations during normal operation without triggering false anomalies.
Regarding claim 10.
Loic in view of Ha discloses all the features of claim 1 as described above.
Loic does not explicitly disclose:
the step of processing the response signal to determine the presence or the absence of the defect in the solid metallic material of the component includes processing an entirety of the response signal produced by the transducer.
However, Loic further teaches:
“Each of the sensors 111-1 to 111-N detects the monitored target and generates a stream of sensor signals representing vibration signals of the monitored target. The vibration signals 902 are detected by the sensor system 110 as sensor signals 112, and converted to trains of asynchronous input event streams 190 by a signal conversion module 120. The sensor signals 112 may be transmitted through wirelines or wireless connections. An embedding network 130 generates and outputs embedded encoding streams 139 to a decoding network 140 based on the input event streams 190. The embedding network 130 and the decoding network 140 form a cascade of neural networks and learn to reproduce reconstructed event streams 149 as a network reconstructed version of the input event streams 190. The detection module 150 compares the reproduced event streams 149 against the input event streams 190, and computes difference between the reproduced event streams 149 and the input event streams 190. The difference below a threshold represents no anomaly is detected, and the detection module 150 accordingly does not raise a trigger signal 180. If the difference is above a threshold, which represents anomaly is detected, then the detection module 150 accordingly raises a trigger signal 180” ([0034]: the sensor signals are converted into input event streams (Fig. 1, item 190; see also [0040]), which are used to generate reproduced event streams (Fig. 1, item 149) using self-supervised machine learning (i.e., embedding network and decoding network; see Fig. 1, items 130 and 140, [0011], [0052], [0082]) by an equipment anomaly detection device (see Fig. 1, item 101, [0037]-[0038], [0119]), with both input event streams and reproduced event streams being compared to determine anomalies in the monitored target (see also [0001]-[0002], [0086], [0124]); examiner notes that entirety of signals are used during analysis).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to incorporate the processing of the response signal to determine the presence or the absence of the defect in the solid metallic material of the component including processing an entirety of the response signal produced by the transducer, in order to provide a continuous monitoring technique for anomaly detection.
Regarding claim 16.
Loic discloses:
A component inspection system (Fig. 1, item 100 – “equipment anomaly detection system”; [0033]: an equipment anomaly detection system for detecting anomalies using signals of a monitored target is presented (see also [0001]-[0002])), comprising:
a signal receiver (Fig. 1, item 111-1 – “sensor”; [0034]: a sensor system (Fig. 1, item 110) includes sensors for detecting the monitored target); and
a controller (Fig. 1, item 101 – “equipment anomaly detection device”; Fig. 9, item 101a) in communication with the signal receiver, and a non-transitory memory (Fig. 14, items 104, 105 – ‘memory’ and “storage device”) storing instructions, which instructions when executed cause the controller ([0037]-[0038], [0119]: an equipment anomaly detection device includes modules to perform self-supervised machine learning (i.e., embedding network and decoding network; see Fig. 1, items 130 and 140, [0011], [0052] and [0082]) based on the sensor signals) to:
control the signal receiver to sense the component for a second signal (Fig. 1, item 902 – “vibration signals”) and produce a response signal (Fig. 1, item 112 – “sensor signals”) representative of the second signal ([0034]: the sensor system includes sensors for detecting vibration signals of the monitored target system and generating a stream of sensor signals representing the vibration signals);
use the response signal and a self-supervised machine learning technique to determine a presence or an absence of a defect of the component ([0034]: the sensor signals are converted into input event streams (Fig. 1, item 190; see also [0040]), which are used to generate reproduced event streams (Fig. 1, item 149) using self-supervised machine learning (i.e., embedding network and decoding network; see Fig. 1, items 130 and 140, [0011], [0052], [0082]) by the equipment anomaly detection device ([0037]-[0038], [0119]), with both input event streams and reproduced event streams being compared to determine anomalies in the monitored target (see also [0001]-[0002], [0086], [0124])).
Loic does not explicitly disclose (see italic text):
a signal transmitter;
a controller in communication with the signal transmitter;
control the signal transmitter to transmit a first ultrasonic signal into a component comprising a solid metallic material;
the second signal is a second ultrasonic signal; and
use the response signal to determine a presence or an absence of a defect in the solid metallic material of the component.
Regarding “a signal transmitter; a controller in communication with the signal transmitter; control the signal transmitter to transmit a first ultrasonic signal into a component; and the second signal is a second ultrasonic signal”, Ha teaches:
“The present disclosure relates to a non-destructive inspection method, and particularly, to a non-destructive inspection method and system based on self-supervised learning, which detect the inside of an inspection object in a non-destructive way by using ultrasonic waves and also predict the depth of a defect through self-supervised learning” ([0001]: a non-destructive inspection method based on self-supervised learning includes emitting ultrasonic waves into an inspection object, and collecting returning signals from the inspection objection (see [0009], [0046]); examiner notes that signal emission implies the use of a transmitter or transducer).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to incorporate a signal transmitter; to incorporate a controller in communication with the signal transmitter; to control the signal transmitter to transmit a first ultrasonic signal into a component; and to incorporate the second signal as a second ultrasonic signal, in order to implement commonly non-destructive inspection techniques to detect defects in an object, as discussed by Ha ([0002], [0056]).
Regarding “a component comprising a solid metallic material; and use the response signal to determine a presence or an absence of a defect in the solid metallic material of the component”, Loic teaches:
“A monitored target system 901 may include one or more entities, such as one or more industrial plants, motors, devices, facilities, or systems” ([0033]: monitored target includes industrial plants, motors and devices (see also [0002] and [0033]-[0034]); examiner notes that industrial plants and motors may include solid metallic material).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to incorporate a component comprising a solid metallic material, and to use the response signal to determine a presence or an absence of a defect in the solid metallic material of the component, in order to identify common defects (e.g., corrosion, cracks, wear) in typical components used in industrial or residential machines that may rapidly affect the operation of the component and reduce its lifetime.
Regarding claim 17.
Loic in view of Ha discloses all the features of claim 16 as described above.
Loic does not explicitly disclose:
the self-supervised machine learning technique is trained to be invariant to a variability of the component.
However, Loic further teaches:
“The embedding network 130 are trained through unsupervised learning. Training data are a small subset of the full data set, that supposedly do not contain any anomalies. The embodiment of the disclosure allows to learn an alternative representation of normal input signals by an encoder, that is the embedding network 130, and to accurately reconstruct normal input signals from such representation by a decoder, that is the decoding network 140. The basic principle behind is that the auto-encoder never learns to reconstruct abnormal signals. Thus, the reconstruction loss is small during normal conditions, and this loss is large in presence of abnormal signals since the auto encoder never learned to reconstruct abnormal signals” ([0052]: during training, the embedding network learns from a subset of data not containing anomalies, with the reconstruction loss being small during normal conditions; examiner notes that by allowing a small reconstruction loss, variability of signal during normal operations is taken into account without affecting model performance (see also [0048] regarding different normal sub-states)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha to train the self-supervised machine learning technique to be invariant to a variability of the component, in order to provide a more robust detection technique that takes into account typical signal variations during normal operation while accurately identifying anomalies.
Claims 2-9 and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Loic, in view of Ha, and in further view of Simumba (US 20240394547 A1), hereinafter ‘Simumba’.
Regarding claim 2.
Loic in view of Ha discloses all the features of claim 1 as described above.
Loic does not disclose:
the self-supervised machine learning technique is trained with a pretext task.
Simumba teaches:
“Discussing elements displayed in FIG. 2 in further detail, self-supervised learning module 210 represents computer software (and, in various embodiments, associated computer hardware), for accessing images available from image database 280 and performing various modifications, transformations, and comparisons upon the accessed images (as discussed herein) to provide for self-supervised learning and training of one or more machine learning models. In various embodiments of the invention, self-supervised learning module 210 includes one or more of image database access module 212, image masking module 215, autoencoder 218, comparison module 221, data augmentation module 224, loss calculation module 226, and machine learning improvement module 228” ([0034]: self-supervised learning and training includes pretext tasks such as masking and data augmentation to remove noise and improve model accuracy (see also [0036]-[0040])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha, and in further view of Simumba, to train the self-supervised machine learning technique with a pretext task, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Regarding claim 3.
Loic in view of Ha and Simumba discloses all the features of claim 2 as described above.
Loic does not disclose:
the pretext task is an augmentation invariance pretext that is used to train the self-supervised machine learning technique based on an augmented response signal.
Simumba further teaches:
“Data augmentation module 224 represents software and/or associated hardware for augmentation of images made available from image database 280, according to one or more data augmentation policies … As would be understood by one of skill in the art, since images may contain “noise” which would increase the likeliness of bad interpretations of future data made by a machine learning model, by augmenting the images, potential “noise” may be removed from images, which allows as further discussed herein, more accurate predictions to be made by machine learning models. In various embodiments of the invention, data augmentation module 224 augments partially masked images and/or original images available from the image database 280 to generate augmented partially masked images or augmented original images. The augmented partially masked images, or augmented original images are used, as more fully discussed herein. In an embodiment of the invention, the augmented partially masked images or augmented original images are input into autoencoder 218 to obtain an augmented model output” ([0039]: autoencoder (Fig. 2, item 218) of the self-supervised learning module (Fig. 2, item 210) uses data augmentation in original images to produce augmented original images (analogous to augmented response signal) as augmentation invariance pretext (see also [0040], [0045], [0047]; see current application at [0021]) for training (see [0031])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to implement the pretext task as an augmentation invariance pretext that is used to train the self-supervised machine learning technique based on an augmented response signal, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040], [0045], [0047]).
Regarding claim 4.
Loic in view of Ha and Simumba discloses all the features of claim 3 as described above.
Loic does not disclose:
the augmented response signal is based on empirical data.
Simumba further teaches:
“Data augmentation module 224 may, in “augmenting” data cause channels of
images to be removed or altered, certain colors to be altered, boundaries of images or portions of images to be resized, color(s) changed, brightness changed, pixels to be jointed” ([0039]: images from measurements (see [0032], empirical data) are used by the data augmentation module).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to implement the augmented response signal being based on empirical data, in order to utilize measurements representing actual component conditions for improving analysis and results.
Regarding claim 5.
Loic in view of Ha and Simumba discloses all the features of claim 4 as described above.
Loic does not explicitly disclose:
the component is a rotor disk for a gas turbine engine, and the empirical data is collected from a plurality of control rotor disks free of defects.
However, Loic further teaches:
“A monitored target system 901 may include one or more entities, such as one or more industrial plants, motors, devices, facilities, or systems” ([0033]: monitored target includes industrial plants, motors and devices (see also [0002] and [0033]-[0034]); examiner notes that industrial plants may include gas turbine engines having motors, which may include rotor discs); and
“Training data are a small subset of the full data set, that supposedly do not contain any anomalies. The embodiment of the disclosure allows to learn an alternative representation of normal input signals by an encoder, that is the embedding network 130, and to accurately reconstruct normal input signals from such representation by a decoder, that is the decoding network 140. The basic principle behind is that the auto-encoder never learns to reconstruct abnormal signals. Thus, the reconstruction loss is small during normal conditions, and this loss is large in presence of abnormal signals since the auto encoder never learned to reconstruct abnormal signals. Such a training procedure allows detecting anomalies without defining what an anomaly is supposed to look like” ([0052]: signals used for training do not contain anomalies).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to incorporate the component as a rotor disk for a gas turbine engine, and the empirical data as collected from a plurality of control rotor disks free of defects, in order to identify common defects (e.g., corrosion, cracks, wear) in typical components used in industrial or residential machines that may rapidly affect the operation of the component and reduce its lifetime, while also easily detecting anomalies by training model based on normal conditions.
Regarding claim 6.
Loic in view of Ha and Simumba discloses all the features of claim 3 as described above.
Loic does not disclose:
the augmentation invariance pretext training of the self-supervised machine learning technique includes determining a contrastive loss between a processed said response signal and a processed said augmented response signal.
Simumba further teaches:
“In an embodiment of the invention, loss calculation module 226 determines total loss by comparing the unaugmented model output (generated by comparison module 221, as discussed herein), with the augmented model output (generated by data augmentation module 224). In another embodiment of the invention, loss calculation module 226 determines a total loss by comparing the unaugmented model output generated by comparison module 221 with one or more augmented images generated by data augmentation module 224 in determining total loss. Total loss may be used, in various embodiments, to improve “weights” in an encoding portion of autoencoder 218, for more accurate interpretations, as would be understood by one of skill in the invention” ([0040]: loss calculation module computes total loss (contrastive loss) between unaugmented model output (processed said response signal) and augmented model output (processed said augmented response signal) to improve model (see also [0045], [0047]))
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to implement the augmentation invariance pretext training of the self-supervised machine learning technique including determining a contrastive loss between a processed said response signal and a processed said augmented response signal, in order to remove noise and improve model accuracy, as discussed by Simumba ([0039]-[0040], [0045], [0047]).
Regarding claim 7.
Loic in view of Ha and Simumba discloses all the features of claim 3 as described above.
Loic does not disclose:
the processed said response signal is processed using a first autoencoder, and the processed said augmented response signal is processed using a second autoencoder.
Simumba further teaches:
“Continuing with regard to FIG. 3, data augmentation module 224 augments partially masked image 315 to generate an augmented partially masked image 340. The augmented partially masked image 340 is also input into autoencoder 218 to obtain an augmented model output 360. The augmented model output 360 and unaugmented model output 330 are used by loss calculation module 226 to determine total loss 390. Total loss 390 is used by machine learning improvement module 228 to further improve autoencoder 218 (or provide other downstream machine learning functions), as further discussed herein” ([0045]: unaugmented model output (see Fig. 3, item 330) (processed said response signal) is obtained by a first implementation of autoencoder 218 (left autoencoder in Fig. 3) (analogous to first autoencoder), while augmented model output (see Fig. 3, item 360) (processed said augmented response signal) is obtained by a second implementation of autoencoder 218 (right autoencoder in Fig. 3) (analogous to second autoencoder)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to process the processed said response signal using a first autoencoder, and to process the processed said augmented response signal using a second autoencoder, in order to train unique models based on corresponding signals (augmented or unaugmented) while avoiding noise or inaccurate responses.
Regarding claim 8.
Loic in view of Ha discloses all the features of claim 1 as described above.
Loic does not disclose:
the self-supervised machine learning technique is trained with an augmentation invariance pretext task and a masked reconstruction pretext.
Simumba teaches:
“Discussing elements displayed in FIG. 2 in further detail, self-supervised learning module 210 represents computer software (and, in various embodiments, associated computer hardware), for accessing images available from image database 280 and performing various modifications, transformations, and comparisons upon the accessed images (as discussed herein) to provide for self-supervised learning and training of one or more machine learning models. In various embodiments of the invention, self-supervised learning module 210 includes one or more of image database access module 212, image masking module 215, autoencoder 218, comparison module 221, data augmentation module 224, loss calculation module 226, and machine learning improvement module 228” ([0034]: self-supervised learning and training includes masking and data augmentation to remove noise and improve model accuracy (see also [0036]-[0040]; see current application at [0021])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha, and in further view of Simumba, to train the self-supervised machine learning technique with an augmentation invariance pretext task and a masked reconstruction pretext, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Regarding claim 9.
Loic in view of Ha and Simumba discloses all the features of claim 8 as described above.
Loic does not disclose:
the training of the self-supervised machine learning technique using the masked reconstruction pretext includes determining a reconstructive loss.
Simumba further teaches:
“Autoencoder 218 in embodiments serves to encode partially masked and/or unmasked images from image database 280 into reduced dimension encodings or other representations of the images, as well as decode the encodings/representations of the images back to representations of their original states … “Loss” from actions of autoencoder 218 encoding/decoding is used, as further discussed herein. In embodiments of the invention, as autoencoder 218 encodes/decodes masked/unmasked images, “noise” may be removed from the images which could lead to more accurate interpretations and predictions associated with future data input into autoencoder 218, as discussed further herein” ([0037]: autoencoder encodes/decodes masked images to remove noise (see also [0044] and [0046] regarding reconstruction loss)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to train the self-supervised machine learning technique using the masked reconstruction pretext including determining a reconstructive loss, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Regarding claim 18.
Loic in view of Ha discloses all the features of claim 17 as described above.
Loic does not explicitly disclose:
the self-supervised machine learning technique is trained using an augmentation invariance pretext and an augmented response signal, wherein the augmented response signal is configured to mimic variability associated with a plurality of control rotor disks that are free of defects.
Regarding “the self-supervised machine learning technique is trained using an augmentation invariance pretext and an augmented response signal”, Simumba teaches:
“Data augmentation module 224 represents software and/or associated hardware for augmentation of images made available from image database 280, according to one or more data augmentation policies … As would be understood by one of skill in the art, since images may contain “noise” which would increase the likeliness of bad interpretations of future data made by a machine learning model, by augmenting the images, potential “noise” may be removed from images, which allows as further discussed herein, more accurate predictions to be made by machine learning models. In various embodiments of the invention, data augmentation module 224 augments partially masked images and/or original images available from the image database 280 to generate augmented partially masked images or augmented original images. The augmented partially masked images, or augmented original images are used, as more fully discussed herein. In an embodiment of the invention, the augmented partially masked images or augmented original images are input into autoencoder 218 to obtain an augmented model output” ([0039]: autoencoder of the self-supervised learning module (see Fig. 2) uses data augmentation in original images (analogous to response signal) to produce augmented original images (analogous to augmented response signal) as augmentation invariance pretext (see also [0040], [0045], [0047]; see current application at [0021]) for training (see [0031])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha, and in further view of Simumba, to train the self-supervised machine learning technique using an augmentation invariance pretext and an augmented response signal, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040], [0045], [0047]).
Regarding “the augmented response signal is configured to mimic variability associated with a plurality of control rotor disks that are free of defects’, Loic further teaches:
“A monitored target system 901 may include one or more entities, such as one or more industrial plants, motors, devices, facilities, or systems” ([0033]: monitored target includes industrial plants, motors and devices (see also [0002] and [0033]-[0034]); examiner notes that industrial plants may include gas turbine engines having motors, which may include rotor discs); and
“Training data are a small subset of the full data set, that supposedly do not contain any anomalies. The embodiment of the disclosure allows to learn an alternative representation of normal input signals by an encoder, that is the embedding network 130, and to accurately reconstruct normal input signals from such representation by a decoder, that is the decoding network 140. The basic principle behind is that the auto-encoder never learns to reconstruct abnormal signals. Thus, the reconstruction loss is small during normal conditions, and this loss is large in presence of abnormal signals since the auto encoder never learned to reconstruct abnormal signals. Such a training procedure allows detecting anomalies without defining what an anomaly is supposed to look like” ([0052]: signals used for training do not contain anomalies).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to configure the augmented response signal to mimic variability associated with a plurality of control components (i.e., rotor disks) that are free of defects, in order to identify common defects (e.g., corrosion, cracks, wear) in typical components used in industrial or residential machines that may rapidly affect the operation of the component and reduce its lifetime while also allowing detecting anomalies without defining what an anomaly looks like, as discussed by Loic ([0052]).
Regarding claim 19.
Loic in view of Ha and Simumba discloses all the features of claim 18 as described above.
Loic does not disclose:
the training of the self-supervised machine learning technique using the augmentation invariance pretext includes determining a contrastive loss between a processed said response signal and a processed said augmented response signal.
Simumba further teaches:
“In an embodiment of the invention, loss calculation module 226 determines total loss by comparing the unaugmented model output (generated by comparison module 221, as discussed herein), with the augmented model output (generated by data augmentation module 224). In another embodiment of the invention, loss calculation module 226 determines a total loss by comparing the unaugmented model output generated by comparison module 221 with one or more augmented images generated by data augmentation module 224 in determining total loss. Total loss may be used, in various embodiments, to improve “weights” in an encoding portion of autoencoder 218, for more accurate interpretations, as would be understood by one of skill in the invention” ([0040]: loss calculation module computes total loss (contrastive loss) between unaugmented model output (processed said response signal) and augmented model output (processed said augmented response signal) (see also [0045], [0047])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to implement the training of the self-supervised machine learning technique using the augmentation invariance pretext including determining a contrastive loss between a processed said response signal and a processed said augmented response signal, in order to remove noise and improve model accuracy, as discussed by Simumba ([0039]-[0040], [0045], [0047]).
Regarding claim 20.
Loic in view of Ha and Simumba discloses all the features of claim 19 as described above.
Loic does not disclose:
the self-supervised machine learning technique is trained using a masked reconstruction pretext, and the training of the self-supervised machine learning technique using the masked reconstruction pretext includes determining a reconstructive loss between the response signal and a reconstructed portion of the response signal.
Simumba further teaches:
“Autoencoder 218 in embodiments serves to encode partially masked and/or unmasked images from image database 280 into reduced dimension encodings or other representations of the images, as well as decode the encodings/representations of the images back to representations of their original states … “Loss” from actions of autoencoder 218 encoding/decoding is used, as further discussed herein. In embodiments of the invention, as autoencoder 218 encodes/decodes masked/unmasked images, “noise” may be removed from the images which could lead to more accurate interpretations and predictions associated with future data input into autoencoder 218, as discussed further herein” ([0037]: autoencoder encodes/decodes masked images to remove noise (see also [0044] and [0046] regarding reconstruction loss)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Ha and Simumba to train the self-supervised machine learning technique using a masked reconstruction pretext, and the training of the self-supervised machine learning technique using the masked reconstruction pretext includes determining a reconstructive loss between the response signal and a reconstructed portion of the response signal, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Claims 11-15 are rejected under 35 U.S.C. 103 as being unpatentable over Loic, in view of Simumba, and in further view of Ha.
Regarding claim 11.
Loic discloses:
A method of inspecting a component (Fig. 1, item 901 – “monitored target”) for a defect ([0028], [0033]: an equipment anomaly detection method for detecting anomalies using signals of a monitored target is presented (see also [0001]-[0002])), comprising:
providing a controller (Fig. 1, item 101 – “equipment anomaly detection device”) that is configured with stored instructions that cause the controller to perform a self-supervised machine learning technique ([0037]-[0038], [0119]: an equipment anomaly detection device includes modules to perform self-supervised machine learning (i.e., embedding network and decoding network; see Fig. 1, items 130 and 140, [0011], [0052] and [0082]));
using the transducer (Fig. 1, item 111-1 – “sensor”) to sense the component for a second signal (Fig. 1, item 902 – “vibration signals”), and produce a transducer response signal (Fig. 1, item 112 – “sensor signals”) representative of the second signal ([0034]: a sensor system includes sensors for detecting vibration signals of the monitored target and generating a stream of sensor signals representing the vibration signals); and
processing the transducer response signal to determine a presence or an absence of a defect of the component ([0034]: the sensor signals are converted into input event streams (Fig. 1, item 190; see also [0040]), which are used to generate reproduced event streams (Fig. 1, item 149) using self-supervised machine learning (i.e., embedding network and decoding network; see Fig. 1, items 130 and 140, [0011], [0052], [0082]) by the equipment anomaly detection device ([0037]-[0038], [0119]), with both input event streams and reproduced event streams being compared to determine anomalies in the monitored target (see also [0001]-[0002], [0086], [0124])).
Loic does not explicitly disclose (see italic text):
the component is a rotor disk;
a self-supervised machine learning technique that includes a trained masked reconstruction pretext;
using a transducer to transmit a first signal into a rotor disk of a gas turbine engine, the rotor disk comprising a solid metallic material;
a second signal produced as a result of the first signal being transmitted into the component;
processing the transducer response signal, including masking a portion of the transducer response signal and using the trained masked reconstruction pretext to determine a presence or an absence of a defect in the solid metallic material of the rotor disk.
Regarding “the component is a rotor disk of a gas turbine engine; the rotor disk comprising a solid metallic material; and processing the transducer response signal to determine a presence or an absence of a defect in the solid metallic material of the rotor disk”, Loic teaches:
“A monitored target system 901 may include one or more entities, such as one or more industrial plants, motors, devices, facilities, or systems” ([0033]: monitored target includes industrial plants, motors and devices (see also [0002] and [0033]-[0034]); examiner notes that industrial plants may include gas turbine engines having motors, which may include rotor discs made of solid metallic material).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic to incorporate the component as a rotor disk of a gas turbine engine, the rotor disk comprising a solid metallic material, and to process the transducer response signal to determine a presence or an absence of a defect in the solid metallic material of the rotor disk, in order to identify common defects (e.g., corrosion, cracks, wear) in typical components used in industrial or residential machines that may rapidly affect the operation of the component and reduce its lifetime.
Regarding “a self-supervised machine learning technique that includes a trained masked reconstruction pretext; and processing the transducer response signal, including masking a portion of the transducer response signal and using the trained masked reconstruction pretext to determine a presence or an absence of a defect”, Simumba teaches:
“Discussing elements displayed in FIG. 2 in further detail, self-supervised learning module 210 represents computer software (and, in various embodiments, associated computer hardware), for accessing images available from image database 280 and performing various modifications, transformations, and comparisons upon the accessed images (as discussed herein) to provide for self-supervised learning and training of one or more machine learning models. In various embodiments of the invention, self-supervised learning module 210 includes one or more of image database access module 212, image masking module 215, autoencoder 218, comparison module 221, data augmentation module 224, loss calculation module 226, and machine learning improvement module 228” ([0034]: self-supervised learning and training of images (analogous to transducer response signal) includes masking portion of the images to remove noise and improve model accuracy (see also [0037], [0044], [0046])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Simumba to incorporate a self-supervised machine learning technique that includes a trained masked reconstruction pretext; and to process the transducer response signal, including masking a portion of the transducer response signal and using the trained masked reconstruction pretext to determine a presence or an absence of a defect, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Regarding “using a transducer to transmit a first signal into a rotor disk of a gas turbine engine; and a second signal produced as a result of the first signal being transmitted into the component”, Ha teaches:
“The present disclosure relates to a non-destructive inspection method, and particularly, to a non-destructive inspection method and system based on self-supervised
learning, which detect the inside of an inspection object in a non-destructive way by using ultrasonic waves and also predict the depth of a defect through self-supervised learning” ([0001]: a non-destructive inspection method based on self-supervised learning includes emitting ultrasonic waves into an inspection object, and collecting returning signals from the inspection objection (see [0009], [0046]); examiner notes that signal emission implies the use of a transmitter or transducer).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Simumba, and in further view of Ha, to use a transducer to transmit a first signal into a component (i.e., rotor disk of a gas turbine engine); and to produce a second signal as a result of the first signal being transmitted into the component, in order to implement commonly non-destructive inspection techniques to detect defects in an object, as discussed by Ha ([0002], [0056]).
Regarding claim 12.
Loic in view of Simumba and Ha discloses all the features of claim 11 as described above.
Loic does not disclose:
the processing step includes using the trained masked reconstruction pretext to produce a reconstructed portion of the transducer response signal.
Simumba further teaches:
“Image masking module 215 represents software and/or hardware for masking sections of images accessed from image database access module 212. … As discussed further herein, by “filling-in,” or altering in some other way images from image database, 280, machine learning models may learn to encode partially-masked and/or unmasked images from the image database 280 into reduced dimension encodings or other representations of the images … Autoencoder 218 in embodiments serves to encode partially masked and/or unmasked images from image database 280 into reduced dimension encodings or other representations of the images, as well as decode the encodings/representations of the images back to representations of their original states” ([0036]-[0037]: self-supervised machine learning models (see [0035]) learn to encode/decode masked images to generate representations of original states (analogous to a reconstructed portion of the transducer response signal) (see also [0044], [0046])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Simumba and Ha to incorporate the processing step including using the trained masked reconstruction pretext to produce a reconstructed portion of the transducer response signal, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Regarding claim 13.
Loic in view of Simumba and Ha discloses all the features of claim 12 as described above.
Loic does not disclose:
the processing step includes evaluating the reconstructed portion relative to the transducer response signal to determine a reconstructive loss.
Simumba further teaches:
“The decoded encoding 320 and original image 305 are compared by comparison module 221 to generate an unaugmented model output 330 (the unaugmented model output 330 including, in various embodiments, a “reconstruction loss” indicating differences between the original image 305 and the decoded encoding 320)” ([0044]: comparison between decoded encoding 320 (resulting from applying masking, see Fig. 3, item 215; analogous to reconstructed portion) and original image 305 (analogous to transducer response signal) generates an unaugmented model output including a reconstruction loss indicating differences (see also [0046])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Simumba and Ha to incorporate the processing step including evaluating the reconstructed portion relative to the transducer response signal to determine a reconstructive loss, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Regarding claim 14.
Loic in view of Simumba and Ha discloses all the features of claim 13 as described above.
Loic does not explicitly disclose:
the processing step includes evaluating the reconstructive loss to determine the presence or the absence of the defect in the solid metallic material of the rotor disk.
However, Loic further teaches:
“The basic principle behind is that the auto-encoder never learns to reconstruct abnormal signals. Thus, the reconstruction loss is small during normal conditions, and this loss is large in presence of abnormal signals since the auto encoder never learned to reconstruct abnormal signals” ([0052]: reconstructive loss is used to determine presence of abnormal signals in a monitored target (analogous to solid metallic material of the rotor disk)).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Simumba and Ha to incorporate the processing step including evaluating the reconstructive loss to determine the presence or the absence of the defect in the solid metallic material of the rotor disk, in order to facilitate the detection of anomalies based on the value of reconstructive loss.
Regarding claim 15.
Loic in view of Simumba and Ha discloses all the features of claim 14 as described above.
Loic does not disclose:
the self-supervised machine learning technique further includes a trained augmentation invariance pretext task.
Simumba further teaches:
“Discussing elements displayed in FIG. 2 in further detail, self-supervised learning module 210 represents computer software (and, in various embodiments, associated computer hardware), for accessing images available from image database 280 and performing various modifications, transformations, and comparisons upon the accessed images (as discussed herein) to provide for self-supervised learning and training of one or more machine learning models. In various embodiments of the invention, self-supervised learning module 210 includes one or more of image database access module 212, image masking module 215, autoencoder 218, comparison module 221, data augmentation module 224, loss calculation module 226, and machine learning improvement module 228” ([0034]: self-supervised learning and training includes pretext tasks such as masking and data augmentation to remove noise and improve model accuracy (see also [0036]-[0040], [0045], [0047]; see current application at [0021])).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Loic in view of Simumba and Ha to incorporate the self-supervised machine learning technique further including a trained augmentation invariance pretext task, in order to remove noise and improve model accuracy, as discussed by Simumba ([0036]-[0040]).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Abreu Calfa; Bruno et al., US 20220404317 A1, AUTOMATED SCAN DATA QUALITY ASSESSMENT IN ULTRASONIC TESTING
Reference discloses ultrasonic testing using a transducer and machine learning to determine defects.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LINA CORDERO whose telephone number is (571)272-9969. The examiner can normally be reached 9:30 am - 6:00 pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, ANDREW SCHECHTER can be reached at 571-272-2302. 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.
/LINA CORDERO/Primary Examiner, Art Unit 2857
Prosecution Insights