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 statement (IDS) submitted on 12/29/2023 is in compliance
with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1, 4, 6-9, 12 are rejected under 35 U.S.C. 103 as being unpatentable over Farahani et al. (“A NOVEL METHOD FOR DESIGNING TRANSFERABLE SOFT SENSORS AND ITS APPLICATION”) in view of Angliker et al. (US 2019/0265087 A1)
Regarding claim 1, Farahani explicitly discloses:
An apparatus for recording a value of a physical variable of a first technical system, wherein the first technical system includes a first system specification, comprising: (Farahani, Pg. 5, Section 4.1: “We used an industrial process data set, which is collected from the SCADA system of a natural gas power plant. The power plant consists of five power units. Each of these units utilizes a Siemens™heavy-duty gas-turbines of class”, Pg. 6, Section 4.2: “Practically, the distribution of data collected from machines or plants of the same type are different from each other. This discrepancy is derived from different maintenance events that each unit experiences, measurement settings, mechanical behavior, and so forth. In this part, the aim is to evaluate the capability of DANN-R for TL between different machines. The source and target data are collected from the gas-turbines of different units of the power plant. Therefore, all environmental condition, like the temperature, humidity and the plant site altitude are the same.”, Pg. 6, Table 1:
PNG
media_image1.png
188
851
media_image1.png
Greyscale
) [Examiner’s note: These passages provide a strong teachings of the source and target system specifications at least being partially coincide]
wherein the generative machine learning model is configured to generate and output, in dependence on at least one value of a first physical variable of a second technical system, at least one value of a second physical variable of the second technical system, (Farahani, Pg. 4, ¶[1]: “Fig. 1 illustrates the architecture of DANN-R. This neural network consists of three major parts, feature extractor, regression model and domain discriminator. The input space is formed by an m-dimensional input data, χ = Rm, which is fed into the feature extractor, Netf (; θf ), with the model parameters θf . The feature extractor is a neural network that maps the input vector into an l-dimensional feature representation, F = Rl. Under the representation of these features, a regression model, Netr(; θr), with the model parameters θr, maps F into a 1-dimensional space Y, which represents the space of corresponding output value of the input sample, yi ∈ Y ”, Pg. 6, Section 4.2, ¶[3]: “Fig. 2 depicts the results of applying DANN-R for designing a transferable soft sensor in the case of TL between two different plants. The figure includes the plots related to the prediction of models in the both source and target domains.”) [Examiner’s note: Farahani states that the regression model maps the extracted feature representation into a one-dimensional output space representing the output value corresponding to an input sample. It further explains that the regression model trained using source data predicts the output value of target domain data.]
wherein the second technical system is characterized by a second system specification and the second system specification at least partially coincides with the first system specification, (Farahani, Pg. 5, Section 4.1: “We used an industrial process data set, which is collected from the SCADA system of a natural gas power plant. The power plant consists of five power units. Each of these units utilizes a Siemens™heavy-duty gas-turbines of class”, Pg. 6, Section 4.2: “Practically, the distribution of data collected from machines or plants of the same type are different from each other. This discrepancy is derived from different maintenance events that each unit experiences, measurement settings, mechanical behavior, and so forth. In this part, the aim is to evaluate the capability of DANN-R for TL between different machines. The source and target data are collected from the gas-turbines of different units of the power plant. Therefore, all environmental condition, like the temperature, humidity and the plant site altitude are the same.”, Pg. 6, Table 1:
PNG
media_image1.png
188
851
media_image1.png
Greyscale
) [Examiner’s note: These passages provide a strong teachings of the source and target system specifications at least being partially coincide]
a measured value generator which is configured to generate, by the generative machine learning model, a value of a second physical variable of the first technical system in dependence on the measured value of the first physical variable, (Farahani, Pg. 1, Section 1, ¶[2]: “In general, soft sensor or virtual sensor is used to make a conclusion based upon observed process variables whenever hardware measurements are not feasible [2–4]. Actually, soft sensor is a software by which several measurement’s signals are processed together in order to estimate the value of another variable of the systems.”, Pg. 3, Setion 3: “we propose a neural network structure for learning transferable regression models based on DANN, called DANNR.It is successfully employed for designing transferable soft sensors that can adapt to new plants and new working conditions”, Pg. 6, Table 1:
PNG
media_image1.png
188
851
media_image1.png
Greyscale
, Pg. 6, Section 4.2, ¶[2]: “In this part, we design soft sensors that predict the value of the active power. The input variables of the soft sensor models are introduced in Table 1. This set of variables are selected according to the performance analysis of the gas-turbines [46], [47]. Three single layer neural networks with proper dimensions are selected for the feature extractor, domain discriminator and regression model”) [Examiner’s note: “measured values of the first physical varibale” is being interpreted as the observed process varaibale or sensor measurements, th e”measured value generator” is being interpreted as the soft-sensor regression model, the “generated second physical vaariable” is being interpreted as the estimated system variable]
wherein the first physical variable of the first technical system corresponds to the first physical variable of the second technical system and the second physical variable of the first technical system corresponds to the second physical variable of the second technical system, and (Farahani, Pg. 6, Section 4.2, ¶[1]: “Practically, the distribution of data collected from machines or plants of the same type are different from each other. This discrepancy is derived from different maintenance events that each unit experiences, measurement settings, mechanical behavior, and so forth. In this part, the aim is to evaluate the capability of DANN-R for TL between different machines. The source and target data are collected from the gas-turbines of different units of the power plant. Therefore, all environmental condition, like the temperature, humidity and the plant site altitude are the same.”, Pg. 6, Table 1:
PNG
media_image1.png
188
851
media_image1.png
Greyscale
) [Examiner’s note: Farahani teaches the same input sensor types are used for source and target gas-turbine units: ambient temperature, ambient humidity, IGV angle and fuel flow. The same estimated sensor/output variable is active power for both the source and target gas-turbine units]
an output unit which is configured to output the generated value of the second physical variable of the first technical system. (Farahani, Pg. 3, Section 2, ¶[2]: “the feature space, XS = {xS1, . . . , xSn}, xSi ∈ χS is the data and P (XS) is the marginal distribution from which the source data is drawn. The corresponding ground truth of the source data is denoted by YS = {yS1, . . . , ySn}, ySi ∈ YS, where YS is the output space. The assumption is that enough amount of labeled data from the source domain is available which enables training a predictive function fˆS (x) to estimate the output yS based on the PS(y|x). Actually, fˆS (.) is an approximation of the optimal function in the source domain fS(x).”)
Farahani fails to disclose:
an interface which is configured to read in a measured value of a first physical variable of the first technical system recorded by means of a physical sensor,
a memory unit is configured to store a generative machine learning model
Angliker explicitly discloses:
an interface which is configured to read in a measured value of a first physical variable of the first technical system recorded by means of a physical sensor, (Angliker, ¶[0030]: “The evaluation unit comprises a data processing processor, a data memory and a display screen. Not only is the digital measurement signal transmitted from the electric amplifier G3 to the evaluation G4 via a plurality of interfaces but the evaluation unit G4 is also able to control the electric amplifier G3 in this manner… the digital measurement signal may be displayed on the display screen.”, ¶[0018]: “A first transmission member Gl typically is a sensor, such as a pressure sensor, an acceleration sensor, a temperature sensor, etc. Accordingly, the sensor Gl measures the measured physical variable M such as a pressure, acceleration, temperature, etc., and generates an analog measurement signal such as an electric current, an electric voltage, etc.”)
a memory unit is configured to store a generative machine learning model, (Angliker, ¶[0030]: “The digital measurement signal may be stored in the data memory.”)
It would have obvious to one of ordinary skill in the art before the effective filing date of the present application to combine Farahani and Angliker. Farahani teaches a new transfer learning (TL) based regression method, called Domain Adversarial Neural Network Regression (DANN-R), and employs it for designing transferable soft sensors. Angliker teaches method and computerized measuring system for detecting a measured physical variable. One of ordinary skill would have motivation to combine Farahani and Angliker because MPEP 2143 sets forth the Supreme Court rationales for obviousness including: (D) Applying a known technique to a known device (method, or product) ready for improvement to yield predictable results; (E): “Obvious to try” choosing from a finite number of identified, predictable solutions, with a reasonable expectation of success; (F) Known work in one field of endeavor may prompt variations of it for use in either the same field or a different one based on design incentives or other market forces if the variations are predictable to one of the ordinary skill in the art.
Regarding claim 4, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker further discloses:
wherein the value of the second physical variable of the first technical system cannot be measured directly or cannot be measured sufficiently by a physical sensor. (Farahani, Pg. 8, Section 5: “Providing an approach for designing transferable sensors, this paper reveals that TL can dramatically enhance the performance of models in these problems. By using our transferable soft sensor, it is possible to predict the value of sensors that are defective or not installed in a power plants via the knowledge transferred from other gas turbine fleets. For instance, Lower Heating Value (LHV) sensor [48], which the hardware sensor is hard to be maintained and expensive to operate. Furthermore, there are some sensors that are installed in system only during limited periods of time, for example during Performance Guarantee Test (Commonly known as PG Test) whose values are very useful for condition monitoring purposes. A model trained with data gathered during such a limited period of time might not be able to make accurate predictions in all working conditions. Again in such cases, our transferable soft sensor can be useful.”)
Regarding claim 6, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker further discloses:
wherein the generative machine learning model is configured based on measured values of the second technical system, (Farahani, Pg. 4, ¶[1]: “Fig. 1 illustrates the architecture of DANN-R. This neural network consists of three major parts, feature extractor, regression model and domain discriminator. The input space is formed by an m-dimensional input data, χ = Rm, which is fed into the feature extractor, Netf (; θf ), with the model parameters θf . The feature extractor is a neural network that maps the input vector into an l-dimensional feature representation, F = Rl. Under the representation of these features, a regression model, Netr(; θr), with the model parameters θr, maps F into a 1-dimensional space Y, which represents the space of corresponding output value of the input sample, yi ∈ Y ”, Pg. 6, Section 4.2, ¶[3]: “Fig. 2 depicts the results of applying DANN-R for designing a transferable soft sensor in the case of TL between two different plants. The figure includes the plots related to the prediction of models in the both source and target domains.”) [Examiner’s note: Farahani states that the regression model maps the extracted feature representation into a one-dimensional output space representing the output value corresponding to an input sample. It further explains that the regression model rained using source data predicts the output value of target domain data.]
wherein the measured values comprise at least one value of the first physical variable and at least one value of the second physical variable of the second technical system. (Farahani, Pg. 6, Section 4.2, ¶[1]: “Practically, the distribution of data collected from machines or plants of the same type are different from each other. This discrepancy is derived from different maintenance events that each unit experiences, measurement settings, mechanical behavior, and so forth. In this part, the aim is to evaluate the capability of DANN-R for TL between different machines. The source and target data are collected from the gas-turbines of different units of the power plant. Therefore, all environmental condition, like the temperature, humidity and the plant site altitude are the same.”, Pg. 6, Table 1:
PNG
media_image1.png
188
851
media_image1.png
Greyscale
) [Examiner’s note: Farahani teaches the same input sensor types are used for source and target gas-turbine units: ambient temperature, ambient humidity, IGV angle and fuel flow. The same estimated sensor/output variable is active power for both the source and target gas-turbine units]
Regarding claim 7, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker further discloses:
wherein the apparatus comprises a computing unit for artificial intelligence. (Angliker, ¶[0030]: “The evaluation unit comprises a data processing processor, a data memory and a display screen.”)
Regarding claim 8, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker further discloses:
wherein the apparatus is realized as a virtual sensor.(Farahani, Pg. 3, ¶[4]: “In our problems, χ is the space of gas-turbine’s sensors, Y is the space of estimated variable and fˆ(.) is the soft sensor model.”)
Regarding claim 9, this claim is rejected under the same rationale as independent claim 1 as they are analogous claims.
Claim(s) 2, 3, 5, 10-11 are rejected under 35 U.S.C. 103 as being unpatentable over Farahani et al. (“A NOVEL METHOD FOR DESIGNING TRANSFERABLE SOFT SENSORS AND ITS APPLICATION”) in view of Angliker et al. (US 2019/0265087 A1) and further in view of Jiang & Farimani (“Deep Learning Convective Flow Using Conditional Generative Adversarial Networks”)
Regarding claim 2, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker fails to disclose:
wherein the generative machine learning model is configured by generative adversarial networks
However, Jiang explicitly discloses:
wherein the generative machine learning model is configured by generative adversarial networks (Jiang, Pg. 1, Col. 2, ¶[1]: “By using the conditional Generative Adversarial Network (cGAN) [11, 12, 13], a generator G, which trained adversarially with a discriminator D, could capture the distribution from multi-physics training data and make prediction G(x; z) directly using condition x and random noise z.”)
The combination of Farahani, Angliker and Jiang are analogous art because they are in the same field of training time series data. Therefore, it would have been obvious to one of ordinary skill in the art before the effective filling date of the claimed invention, having the teachings of Farahani, Angliker and Jiang before them, to modify the teachings of Farahani and Angliker to include the teachings of Jiang using generative adversarial networks as the generative machine learning model because by using the conditional Generative Adversarial Network (cGAN), a generator G, which trained adversarially with a discriminator D, could capture the distribution from multi-physics training data and make prediction G(x; z) directly using condition x and random noise z.
Regarding claim 3, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker fails to disclose:
wherein the generative machine learning model is configured to generate, in dependence on at least one value of a first physical variable of the second technical system at a first time, at least one value of a second physical variable of this technical system at a second time, wherein the second time is later than the first time.
However, Jiang explicitly discloses:
wherein the generative machine learning model is configured to generate, in dependence on at least one value of a first physical variable of the second technical system at a first time, at least one value of a second physical variable of this technical system at a second time, wherein the second time is later than the first time. (Jiang , Pg. 3, Fig. 2: “Time-dependent convective flow prediction This figure shows some good prediction results in the time-dependent dataset of u, v, p, T respectively (in second). For each category, the first row is the prediction of FluidGAN model and the second row is the ground truth.”
PNG
media_image2.png
764
742
media_image2.png
Greyscale
) [Examiner’s note: Fig 2 teaches receiving a velocity value or field at the initial time t=0, and generate a pressure value or field at a subsequent time t =2…t=30]
Regarding claim 5, the combination of Farahani and Angliker discloses all the limitations in claim 1 (as shown in the rejections above).
Farahani in view of Angliker fails to disclose:
wherein the generative machine learning model is configured based on simulation data of a computer-aided simulation of the second technical system,
However, Jiang explicitly discloses:
wherein the generative machine learning model is configured based on simulation data of a computer-aided simulation of the second technical system, (Jiang, Pg. 5, Fig 4:
PNG
media_image3.png
581
339
media_image3.png
Greyscale
“Discriminator performance for ”fake time” inputs. The first two rows show the output of discriminator for outputs and targets respectively, suggesting that the discriminator could tell the difference between outputs and targets. The last two rows show the output of discriminator for ”fake time” inputs, in which we add t0 to the time channel. When t0 is getting bigger (i.e. from 0s to 4s), the discriminator could gradually classify this time channel dis-match.”, Pg. 5, Col. 2: “In order to fool our model, we add some positive value t0 to the time channel of targets, namely combining u, v, p, T values of time t with time value t + t0. As Fig. 4 suggests, when we increase t0, the discriminator is gradually able to tell the mismatch between fake time channel and actual time channel. Therefore, the discriminator we get from the training could be used to identify non-physical prediction, which may be further used to identify the quality of a numerical simulation”)
wherein the simulation data comprises at least one value of the first physical variable and at least one value of the second physical variable of the second technical system. (Jiang, Pg. 5, Col. 1, ¶[2]: “This result suggests that the FluidGAN discriminator is very robust in identifying nonphysical prediction. We further explore this finding by constructing some ”fake” predictions. In the time-dependent prediction, we know that if we feed predictions into the discriminator, it will have low discriminator score. In the meantime, if we input targets to the discriminator it should predict high discriminator score, which should be close to 1. In order to fool our model, we add some positive value t0 to the time channel of targets, namely combining u, v, p, T values of time t with time value t + t0. As Fig. 4 suggests, when we increase t0, the discriminator is gradually able to tell the mismatch between fake time channel and actual time channel. Therefore, the discriminator we get from the training could be used to identify non-physical prediction, which may be further used to identify the quality of a numerical simulation.”)
Regarding claim 10, the combination of Farahani and Angliker discloses all the limitations in claim 9 (as shown in the rejections above).
Farahani in view of Angliker further discloses:
A computer-implemented method for providing a generative
machine learning model for use in the method as claimed in claim 9, with the method steps:
reading in training data of a technical system, wherein the training data comprises at least one value of a first physical variable and at least one value of a second physical variable of the technical system, (Farahani, Pg. 6, Section 4.2, ¶[1]: “Practically, the distribution of data collected from machines or plants of the same type are different from each other. This discrepancy is derived from different maintenance events that each unit experiences, measurement settings, mechanical behavior, and so forth. In this part, the aim is to evaluate the capability of DANN-R for TL between different machines. The source and target data are collected from the gas-turbines of different units of the power plant. Therefore, all environmental condition, like the temperature, humidity and the plant site altitude are the same.”, Pg. 6, Table 1:
PNG
media_image1.png
188
851
media_image1.png
Greyscale
) [Examiner’s note: Farahani teaches the same input sensor types are used for source and target gas-turbine units: ambient temperature, ambient humidity, IGV angle and fuel flow. The same estimated sensor/output variable is active power for both the source and target gas-turbine units]
training the generative machine learning model by means of the training data and a discriminator network such that the generative machine learning model generates and outputs the value of the second physical variable in dependence on the value of the first physical variable, and (Farahani, Pg. 4, ¶[1]: “Fig. 1 illustrates the architecture of DANN-R. This neural network consists of three major parts, feature extractor, regression model and domain discriminator… Under the representation of these that maps the input vector into an l-dimensional feature representation, F = R . features, a regression model, Netr(; θr), with the model parameters θr, maps F into a 1-dimensional space Y, which represents the space of corresponding output value of the input sample, yi ∈ Y . Moreover, F is also introduced to the domain discriminator, Netd(; θd) with the model parameters θd. Netd(; θd) is a classifier that maps features F into a 1-dimensional binary space D, which represents the space of domain label of the input sample, di ∈ D. In other words, the domain discriminator tries to detect whether input instances are from the source domain or the target domain. When an instance is from the source domain or target domain, the output of domain discriminators is expected to be 0 or 1, respectively. ”)
outputting the trained generative machine learning model for recording a value of a physical variable of a technical system. (Farahani, Pg. 4, ¶[1]: “Fig. 1 illustrates the architecture of DANN-R. This neural network consists of three major parts, feature extractor, regression model and domain discriminator. The input space is formed by an m-dimensional input data, χ = Rm, which is fed into the feature extractor, Netf (; θf ), with the model parameters θf . The feature extractor is a neural network that maps the input vector into an l-dimensional feature representation, F = Rl. Under the representation of these features, a regression model, Netr(; θr), with the model parameters θr, maps F into a 1-dimensional space Y, which represents the space of corresponding output value of the input sample, yi ∈ Y ”, Pg. 6, Section 4.2, ¶[3]: “Fig. 2 depicts the results of applying DANN-R for designing a transferable soft sensor in the case of TL between two different plants. The figure includes the plots related to the prediction of models in the both source and target domains.”) [Examiner’s note: Farahani states that the regression model maps the extracted feature representation into a one-dimensional output space representing the output value corresponding to an input sample. It further explains that the regression model rained using source data predicts the output value of target domain data.]
Farahani in view of Angliker fails to disclose:
reading in a generative machine learning model,
However, Jiang explicitly discloses:
reading in a generative machine learning model, (Jiang, Pg. 1, Col. 2, ¶[1]: “By using the conditional Generative Adversarial Network (cGAN) [11, 12, 13], a generator G, which trained adversarially with a discriminator D, could capture the distribution from multi-physics training data and make prediction G(x; z) directly using condition x and random noise z.”)
Regarding claim 11, this claim is rejected under the same rationale with claim 5 as they are analogous claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMY TRAN whose telephone number is (571)270-0693. The examiner can normally be reached Monday - Friday 7:30 am - 5:00 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, David Yi can be reached at (571) 270-7519. 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.
/AMY TRAN/Examiner, Art Unit 2126
/DAVID YI/Supervisory Patent Examiner, Art Unit 2126
Prosecution Insights