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 .
Information Disclosure Statement
1. The information disclosure statements (IDS) submitted on 7/10/2023 and is in compliance with the provisions of 37 CFR 1.97. According, the information disclosure statement is being considered by the Examiner.
Election/Restrictions
2. Applicant elected Group I with claims 1-6 and 16-20 as readable on the elected invention, in the reply filed on 12/19/2025 is acknowledged. Applicant elected without an indication of “traverse” or “without traverse” and because applicant did not distinctly and specifically point out the supposed errors in the restriction requirement, the election has been treated as an election without traverse (MPEP § 818.03(a)). Claims 7-15 are withdrawn from further consideration pursuant to 37 CFR 1.142(b) as being drawn to a nonelected claim invention. Claims 1-6 and 16-20 are presented for examination.
Claim Rejections - 35 USC § 101
3. The following is a quotation from 35 U.S.C. 101:
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.
PNG
media_image1.png
298
532
media_image1.png
Greyscale
4. Claims 1-6 and 16-20 are rejected under 35 U.S.C. 101 as being directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
Regarding claim 1, the claim recites a method for monitoring voltage drift, the method comprising: measuring a voltage across a diode of a power device; providing the measured voltage as an input to a controller, the controller being configured to run a transformer-based model, wherein the transformer-based model comprises a temporal fusion transformer with a temporal convolutional neural network, the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm; and forecasting a range of expected future values of the voltage across the diode of the power device with the transformer-based model.
Step
Analysis
1: Statutory Category?
Yes. The claim 1 recites a method for monitoring voltage drift which falls within one of the four statutory categories (i.e. process, machine, manufacture, or composition of matter).
2A - Prong 1: Judicial Exception Recited?
Yes. The claim recites the step of “providing the measured voltage as an input to a controller, the controller being configured to run a transformer-based model, wherein the transformer-based model comprises a temporal fusion transformer with a temporal convolutional neural network, the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm”. This limitation, as drafted, under its broadest reasonable interpretation, recites a mathematical calculation. The grouping of “mathematical concepts” includes “mathematical calculations” as an exemplar of an abstract idea. The claim recites the step of “forecasting a range of expected future values of the voltage across the diode of the power device with the transformer-based model”. Inputting the measured data to the transformer model in order to provide a forecast of futures values of the voltage across the diode is directed to a mathematical calculation. Therefore, mathematical operation is an abstract idea.
2A - Prong 2: Integrated into a Practical Application?
No. The claim recites additional elements “measuring a voltage across a diode of a power device” is a generic measurement which is a mere indication of the field of use, which does not integrate the judicial exception into a practical application because the additional elements do not impose any meaningful limits on practicing the abstract ideas.The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is directed to the abstract idea.
2B: Claim provides an Inventive Concept?
No. When considering claim 1 as a whole, The claim does not amount to significantly more than judicial exception, it does not provide enough to be an inventive concept because it merely amounts to applying a judicial exception to a well-known industry.
Claims 2-6 are also rejected as they depended on the rejected base claim 1.
Regarding claim 2, the claim recites “the adversarial compensation model is configured to compensate for noise with Jacobian regularization” is a mathematical operation which is an abstract idea.
Regarding claim 3, the claim recites “training the transformer-based model with a power cycling test” is a mathematical operation using the transformer model. The mathematical operation is an abstract idea.
Regarding claim 4, the claim recites “forecasting when the power device will reach an end of its operational lifetime based on the range of expected future values of the voltage across the diode of the power device” is a mathematical operation using the transformer model. The mathematical operation is an abstract idea.
Regarding claim 5, the claim recites “ the temporal fusion transformer comprises a multi-head attention block” is a mathematical operation using the transformer model. The mathematical operation is an abstract idea.
Regarding claim 6, the claim recites “the backpropagation algorithm computes a gradient of a loss function, the loss function being a Mean Square Error (MSE) loss function” is a mathematical operation using the transformer model. The mathematical operation is an abstract idea.
Regarding claim 16, the claim recites a system for monitoring voltage drift, the system comprising: a power device; a non-transitory memory comprising a program; and a microprocessor coupled to the non-transitory memory and the power device, the microprocessor being configured to execute the program, the program comprising a transformer-based model, wherein the transformer-based model comprises a temporal fusion transformer with a temporal convolutional neural network, the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm, and based on the transformer-based model monitoring the voltage drift of the power device and predicting future voltage drift of the power device.
Step
Analysis
1: Statutory Category?
Yes. The claim 16 recites a system for monitoring voltage drift which falls within one of the four statutory categories (i.e. process, machine, manufacture, or composition of matter).
2A - Prong 1: Judicial Exception Recited?
Yes. The claim recites the step of “the program comprising a transformer-based model, wherein the transformer-based model comprises a temporal fusion transformer with a temporal convolutional neural network, the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm, and based on the transformer-based model monitoring the voltage drift of the power device and predicting future voltage drift of the power device” is directed to a mathematical operation by using a software transformer model to calculate and forecast future voltage drift of the power device. Therefore, this limitation, as drafted, under its broadest reasonable interpretation, recites a mathematical calculation. The grouping of “mathematical concepts” includes “mathematical calculations” as an exemplar of an abstract idea.
2A - Prong 2: Integrated into a Practical Application?
No. The claim does not include additional elements that are sufficient to amount to significantly more than the judicial exception. The claim is directed to the abstract idea.
2B: Claim provides an Inventive Concept?
No. When considering claim 16 as a whole, The claim does not amount to significantly more than judicial exception, it does not provide enough to be an inventive concept because it merely amounts to applying a judicial exception to a well-known industry.
Claims 17-20 are also rejected as they depended on the rejected base claim 16.
Regarding claim 17, the claim recites “the non-transitory memory is firmware” is a generic element which is a mere indication of the field of use, which does not integrate the judicial exception into a practical application because the additional elements do not impose any meaningful limits on practicing the abstract ideas.
Regarding claim 18, the claim recites “the power device is a SiC power device” is a generic device which is a mere indication of the field of use, which does not integrate the judicial exception into a practical application because the additional elements do not impose any meaningful limits on practicing the abstract ideas.
Regarding claim 19, the claim recites “the power device is coupled to a cooling system of a traction drive” is a generic operation which is a mere indication of the field of use, which does not integrate the judicial exception into a practical application because the additional elements do not impose any meaningful limits on practicing the abstract ideas.
Regarding claim 20, the claim recites “ the program further comprises instructions to produce an alert message in response to determining that degradation of the performance of the power device is predicted to occur within twenty weeks of additional use” is a mathematical operation using the transformer model. The mathematical operation is an abstract idea.
Examiner Notes
5. Examiner cites particular paragraphs, columns and line numbers in the references as applied to the claims below for the convenience of the applicant. Although the specified citations are representative of the teachings in the art and are applied to the specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested that, in preparing responses, the applicant fully consider the references in entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner.
Claim Rejections - 35 USC § 102
6. The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention.
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
7. Claims 16-17 and 19 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Li et al. (CN-115600502, with attached English translation; hereinafter “Li”).
Regarding claim 16, Li discloses a system for monitoring voltage drift (a system for predicting the health of IGBT module in a power semiconductor device based on obtaining the on-state voltage drop of the IGBT module, see abstract and page 3 lines 28-31), the system comprising: a power device (a power semiconductor device includes IGBT and freewheel diode, see page 5); a non-transitory memory comprising a program; and a microprocessor coupled to the non-transitory memory and the power device, the microprocessor being configured to execute the program (a readable storage medium, the readable storage medium is stored with computer program instructions, the computer program instructions are read and run by a processor, executing the construction method of the IGBT module life prediction model based on GRU neural network, see page 11), the program comprising a transformer-based model (a life prediction model based on GRU neural network, see abstract and Fig. 6), wherein the transformer-based model comprises a temporal fusion transformer with a temporal convolutional neural network (a GRU circulation neural network for the life prediction model, comprising a sequence input layer, a hidden layer, a full connection layer and a regression output layer, the sequence input layer inputs the aging characteristic quantity data input sequence of the IGBT module into the network, the hidden layer through input weight, recursion weight and offset weight to learn advanced characteristic of the IGBT module circulating aging test time sequence data, the hidden layer comprises multiple groups of sub-hidden layers, any group of sub-hidden layers comprises a plurality of GRU layers with configurable hidden layer node number and a Dropout layer set with the discarding probability of the initial value, then it is a full connection layer and a regression output layer, the predicted value of the aging characteristic quantity of the IGBT module predicted by the GRU network is output, wherein the sub-hidden layer specific number, the node specific number of each GRU layer is obtained by subsequent model construction…. See pages 8-9 and Fig. 6), the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm (“In order to weaken the adverse effect of noise to the neural network, improving the life prediction precision, the invention uses S-G filtering method to reduce the data fluctuation, enhancing data periodicity. The algorithm of the S-G filtering method is to perform weighted smoothing to the data in the sliding window of the specified width, and the weighting is obtained according to the given high order polynomial using the least squares fitting…”, see at least in page 8), the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm (“In order to weaken the adverse effect of noise to the neural network, improving the life prediction precision, the invention uses S-G filtering method to reduce the data fluctuation, enhancing data periodicity. The algorithm of the S-G filtering method is to perform weighted smoothing to the data in the sliding window of the specified width, and the weighting is obtained according to the given high order polynomial using the least squares fitting…”, see at least in page 7), and based on the transformer-based model monitoring the voltage drift of the power device and predicting future voltage drift of the power device (“using the IGBT module life prediction model based on GRU neural network Specifically, based on the GRU neural network the IGBT module life prediction method comprises: obtaining the on-state voltage drop of the IGBT module, introducing the IGBT module life prediction model based on GRU neural network obtaining the aging degree of the IGBT module, which also represents the remaining service life of the IGBT module. based on the verification can be known, using the life prediction model can accurately obtain the aging degree of the IGBT module”, see page 11).
Regarding claim 17, Li discloses the system of claim 16, wherein the non-transitory memory is firmware (see claim 10).
Regarding claim 19, Li discloses the system of claim 16, wherein the power device is coupled to a cooling system of a traction drive (see page 6).
Claim Rejections - 35 USC § 103
8. 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 of this title, 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.
9. Claims 1 and 3-6 are rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Sjoroos et al. (US. Pub. 2016/0377671; hereinafter “Sjoroos”).
Regarding claim 1, Li discloses a method for monitoring voltage drift (a method of predicting the health of IGBT module in a power semiconductor device based on obtaining the on-state voltage drop of the IGBT module, see abstract and page 3 lines 28-31), the method comprising: measuring a voltage across a The GRU neural network IGBT module life prediction method comprises: obtaining the on-state voltage drop of the IGBT module”; see at least in page 3 lines 28-31); providing the measured voltage as an input to a controller (“introducing the GRU neural network the optimal parameter of the IGBT module life prediction model”, see at least in page 3 lines 28-31), the controller being configured to run a transformer-based model (a life prediction model based on GRU neural network, see abstract and Fig. 6), wherein the transformer-based model comprises a temporal fusion transformer with a temporal convolutional neural network (a GRU circulation neural network for the life prediction model, comprising a sequence input layer, a hidden layer, a full connection layer and a regression output layer, the sequence input layer inputs the aging characteristic quantity data input sequence of the IGBT module into the network, the hidden layer through input weight, recursion weight and offset weight to learn advanced characteristic of the IGBT module circulating aging test time sequence data, the hidden layer comprises multiple groups of sub-hidden layers, any group of sub-hidden layers comprises a plurality of GRU layers with configurable hidden layer node number and a Dropout layer set with the discarding probability of the initial value, then it is a full connection layer and a regression output layer, the predicted value of the aging characteristic quantity of the IGBT module predicted by the GRU network is output, wherein the sub-hidden layer specific number, the node specific number of each GRU layer is obtained by subsequent model construction…. See pages 8-9 and Fig. 6), the transformer-based model further comprising an adversarial compensation model with a backpropagation algorithm (“In order to weaken the adverse effect of noise to the neural network, improving the life prediction precision, the invention uses S-G filtering method to reduce the data fluctuation, enhancing data periodicity. The algorithm of the S-G filtering method is to perform weighted smoothing to the data in the sliding window of the specified width, and the weighting is obtained according to the given high order polynomial using the least squares fitting…”, see page 7); and forecasting a range of expected future values of the voltage across the diode of the power device with the transformer-based model (“using the IGBT module life prediction model based on GRU neural network Specifically, based on the GRU neural network the IGBT module life prediction method comprises: obtaining the on-state voltage drop of the IGBT module, introducing the IGBT module life prediction model based on GRU neural network obtaining the aging degree of the IGBT module, which also represents the remaining service life of the IGBT module. based on the verification can be known, using the life prediction model can accurately obtain the aging degree of the IGBT module”, see page 11).
Li does not explicitly specify the step of measuring a voltage across a diode of a power device.
Sjoroos discloses, in Figs. 1-2, a method of monitoring aging or health of a power semiconductor device (see abstract), comprising the step of measuring a voltage across a diode of a power device (“the power semiconductor device includes a diode, and determining the value includes determining a voltage drop via sensing leads electrically connected to an anode and a cathode of the diode”, see claim 8 and [0020]).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to employ the monitoring system of Li by measuring a voltage across a diode of a power semiconductor device as taught by Sjoroos for purpose of monitoring aging of a power semiconductor device e.g. insulated gate bipolar transistor and freewheeling diode in an electric drive system.
Regarding claim 3, Li and Sjoroos disclose the method of claim 1, Li further teaches comprising training the transformer-based model with a power cycling test (“step one, performing power cycle aging test to the IGBT module, obtaining the IGBT module aging characteristic quantity data, …” in page 2, “setting the cycle training time epoch as 30, the hidden layer neural node number is substituted, selecting mean square error (MSE) as loss function, selecting simple cross validation method, observing the change of trend loss and verification loss, if training loss is continuously reduced, the verification loss is continuously reduced,…” in page 10, ).
Regarding claim 4, Li and Sjoroos disclose the method of claim 1, Li further teaches comprising forecasting when the power device will reach an end of its operational lifetime based on the range of expected future values of the voltage across the diode of the power device (see at least in claims 1-6).
Regarding claim 5, Li and Sjoroos disclose the method of claim 1, Li further teaches wherein the temporal fusion transformer comprises a multi-head attention block (see Fig. 6).
Regarding claim 6, Li and Sjoroos disclose the method of claim 1, wherein the backpropagation algorithm computes a gradient of a loss function, the loss function being a Mean Square Error (MSE) loss function (FIG. 7 is a root-mean-square error diagram of the number of different hidden layers of the GRU neural network in FIG. 6. “using the root mean square error function as the prediction index of the IGBT service life prediction problem, using different hidden layer number and different hidden layer node number, respectively iterating the same times, comparing different layer number root mean square error average value, determining the optimal hidden layer number;” in page 3 and 9.).
10. Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Sjoroos and further in view of Liu et al. (NPL: “Jacobian Norm with Selective Input Gradient Regularization for Improved and Interpretable Adversarial Defense”; hereinafter “Liu”).
Regarding claim 2, Li and Sjoroos disclose the method of claim 1, except for specifying that wherein the adversarial compensation model is configured to compensate for noise with Jacobian regularization.
Liu discloses an AI model being trained using a joint loss function (page 3, section a) in second column: Adversarial training) comprising the adversarial compensation model is configured to compensate for noise with Jacobian regularization (page 9, second column: A set of noise injection methods, i.e., RSE [46], Adv-BNN [47] and PNI [15], combine the adversarial training and noise injection into the inputs/weights of the network. However, these methods, except for the PNI [15] manually set the noise configurations, making it very ad-hoc, and thus not generalizable to different datasets. PNI [15] exploits the min-max optimization with trade-off on clean-and perturbed data by injecting trainable Gaussian noise on various locations of the network to generate adversarial examples. Whilst PNI [15] improves the accuracy of both clean and perturbed data, noise injection is not related to the robustness response of the network. In contrast, our proposed method regularizes the Jacobian norm and the input gradients, such that the network parameters can be dynamically trained to perform better adversarial defense. In addition, the Jacobian norm regularization explicitly suggests the robustness of the classification model in response to imperceptible data perturbation).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to employ the monitoring system of Li and Sjoroos by having the adversarial compensation model is configured to compensate for noise with Jacobian regularization, as taught by Liu for purpose of providing adversarial training (AT) is often adopted to improve robustness through training a mixture of corrupted and clean data (see abstract of Liu).
11. Claim 18 is rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Lichtenwalner et al. (US. Pub. 2021/0270886; hereinafter “Lichtenwalner”).
Regarding claim 18, Li discloses the system of claim 16, except for specifying wherein the power device is a SiC power device.
Lichtenwalner discloses a power semiconductor device (10 in Fig. 5) comprising the power device is a SiC power device (see [0070]).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to employ the monitoring system of Li and Sjoroos by having the power device is a SiC power device, as taught by Lichtenwalner in order to meet the system design and specification requirement.
12. Claim 20 is rejected under 35 U.S.C. 103 as being unpatentable over Li in view of Kindu et al. (US. Pub. 2025/0052709; hereinafter “Kundu”).
Regarding claim 20, Li discloses the system of claim 16, except for explicitly specifying that wherein the program further comprises instructions to produce an alert message in response to determining that degradation of the performance of the power device is predicted to occur within twenty weeks of additional use.
Kundu discloses, in Fig. 3, a method for state-of-health monitoring of a powertrain component in an electric vehicle system includes: determining an equivalent circuit model of the powertrain component; modeling heat losses in the powertrain component considering both transient and steady-state conditions, wherein the program further comprises instructions to produce an alert message in response to determining that degradation of the performance of the power device is predicted to occur within twenty weeks of additional use (see [106, 115]).
It would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to employ the monitoring system of Li by producing an alert message in response to determining that degradation of the performance of the power device is predicted to occur within twenty weeks of additional use, as taught by Kundu for purpose of the fault diagnosis and prognosis method is used to identify solder fatigue and bond-wire liftoff fault, the voltage-current (VI) characteristics are defined with multiple points for improved accuracy conducting experiment in continuous current conduction mode.
Prior Art of Record
13. The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Zhang (U.S Pub. 20230208281) discloses a method for detecting early degradation within the inverter module (see specification for more details). Degrenne (U.S Pub. 20200256912) discloses a method to establish a degradation state of electrical connections in a power semiconductor device (see specification for more details).
Conclusion
14. Any inquiry concerning this communication or earlier communications from the examiner should be directed to THANG LE whose telephone number is (571)272-9349. The examiner can normally be reached on Monday thru Friday 7:30AM-5:00PM EST.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Huy Phan can be reached on (571) 272-7924. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see http://pair-direct.uspto.gov. Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/THANG X LE/Primary Examiner, Art Unit 2858
4/2/2026
Prosecution Insights