DETAILED ACTION
Claims 1-7 and 9-13 are amended. Claims 1-7 and 9-13 remain pending in the application.
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 .
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 02/09/2026 has been entered.
Response to Arguments
Applicant's arguments filed in response to the previous rejections have been fully considered.
In reference to Applicant’s arguments:
- Claim rejections under 35 USC 112(b).
Examiner’s response:
Applicant’s arguments have been fully considered. Rejections are withdrawn in view of amendments and applicant’s arguments.
In reference to Applicant’s arguments:
- Claim rejections under 35 USC 101.
Examiner’s response:
Applicant argues that the newly added limitation “wherein a maintenance on the device is performed based on the outputted prediction result so that downtime in operation of the device is reduced” provides an improvement in the technical field since performing pre-maintenance on a device such as a gas engine allows for its operation with a minimum required downtime (according to paragraph [0002]), however, Examiner respectfully disagrees. After careful consideration, Examiner understands that the specification explicitly sets forth an improvement but in a conclusory manner (i.e., a bare assertion of an improvement without the detail necessary to be apparent to a person of ordinary skill in the art), as the Specification fails to describe how this maintenance is being performed or implemented as an unconventional technical solution. Furthermore, this limitation is considered to incorporate new matter in the claim which does not contain support in the original disclosure, as the Specification only recites the possibility (emphasis added) of a device to be operated with minimum required downtime by performing pre-maintenance, however, the Specification fails to describe how or when this pre-maintenance is performed.
Applicant further points out to the Specification at paragraphs 0015, 0018, 0034 and 0057, however, these paragraphs simply describe producing an output with a notification of a predicted abnormality (e.g. predicting a misfire) and how reliable the prediction is, therefore, they fail to describe the performing of any maintenance in response to this predicted abnormality. The recitation of a claim limitation that attempts to cover any solution to an identified problem with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result, does not integrate a judicial exception into a practical application or provide significantly more because this type of recitation is equivalent to the words "apply it" (see MPEP 2106.05 (f)).
Applicant further asserts that the limitation “outputting a prediction result generated by the abnormality prediction to a monitor or another apparatus via electronic transmission, and performing maintenance on the device based on the outputted prediction result” improves the technical field, since performing pre-maintenance on a device allows for its operation with a minimum required downtime, however, Examiner respectfully disagrees. First, Examiner would like to point out that this limitation was amended and now recites “outputting a prediction result generated by the abnormality prediction to a monitor or another apparatus via electronic transmission”, and it is still considered under Step 2A Prong 2 as insignificant extra solution activity, mere data outputting as per MPEP 2106.05(g); and under Step 2B as well-understood, routine, and conventional activity as per MPEP 2106.05(d): “i. Receiving or transmitting data over a network”. This limitation simply transmits the result electronically, and the Specification does not explain or recite any further maintenance being performed after this transmission is made, as explained above. For these reasons, the 35 USC 101 rejection is still maintained.
In reference to Applicant’s arguments:
- Claim rejections under 35 USC 103.
Examiner’s response:
Applicant’s arguments filed in response to rejections under 35 USC 103 are mainly directed to the prior art Guo failing to teach the limitation “wherein the first prediction model is a model that predicts whether or not an abnormality occurs in the device in the future within a predetermined period based on the probability density of the operation data between a first day when the abnormality has actually occurred and a second day that goes back the predetermined period from the first day”. Applicant’s arguments have been considered but are moot in view of new grounds of rejection.
Claim Rejections - 35 USC § 112
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
Claims 1-7 and 9-13 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for pre-AIA the inventor(s), at the time the application was filed, had possession of the claimed invention.
Independent claim 1 (and analogous claims 9-10) recite the limitation (as amended) “wherein a maintenance on the device is performed based on the outputted prediction result so that downtime in operation of the device is reduced”. Said limitation is considered to incorporate new matter in the claim which does not contain support in the original disclosure. Specification at [0002] recites: “There is a possibility that a device such as a gas engine that normally operates has to be shut down for an extended period once a failure occurs, and accordingly, there is also a possibility that a large loss occurs. If an abnormality of the device can be accurately predicted, the device can be operated with the minimum required downtime by performing pre-maintenance”. This paragraph merely describes the idea of a possibility to perform pre-maintenance if an abnormality could be predicted, it does not explain an actual execution of maintenance in response to a prediction result or how this maintenance performed. Therefore, examiner cannot find support in the Specification for this limitation, and cannot conclude how exactly this maintenance is performed in view of an outputted prediction result.
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 1-7 and 9-13 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1:
In the instant case, the claims are directed to a system, method and non-transitory computer readable medium. Thus, each of the claims falls within one of the four statutory categories (i.e., process, machine, manufacture, or composition of matter).
Step 2A Prong 1:
Claims 1, 9, and 10 recite:
“estimating a probability density of the operation data”; estimating a probability density is a mathematical calculation based on a mathematical equation, and is thus a mathematical concept;
“predicting whether or not an abnormality occurs in the device, based on an estimation result of the probability density of the operation data and a first prediction model”; making a prediction based on an estimation result is an evaluation that can be carried out by a human in the mind or with pen and paper as it is a comparison of mathematical values (comparing the calculated probability density value against a threshold) and is thus a mental process and mathematical concepts. The use of the first prediction model is discussed next at Prong 2;
“wherein the first prediction model is a model that predicts whether or not an abnormality occurs in the device in the future within a predetermined period based on the probability density of the operation data between a first day when the abnormality has actually occurred and a second data that goes back the predetermined period from the first day”; making a prediction based on a probability density based on historical data is an evaluation that can be carried out by a human in the mind or with pen and paper as it is a comparison of mathematical values (comparing the calculated probability density value against a threshold) and is thus a mental process and mathematical concepts. The use of the first prediction model is discussed next at Prong 2;
Step 2A Prong 2:
This judicial exception is not integrated into a practical application because the additional elements are as follows:
“a prediction system comprising: a device, a prediction apparatus comprising: a processor; a memory to store a program which, when executed by the processor, the processor performs processes of”; these limitations amount to nothing more than an instruction to apply the abstract idea using a generic computer as per MPEP 2106.05(f). Further, the device is merely described as the thing which the abnormality is being predicted, and this amounts to merely indicating a field of use or technological environment in which to apply a judicial exception as per MPEP 2106.05(h);
“acquiring operation data indicating an operation state of the device”; this limitation amounts to insignificant extra solution activity, mere data gathering as per MPEP 2106.05(g);
“a first prediction model”; this model is recited at a high level of generality to perform the calculations/ determinations of the probability density, interpreted as mere instructions to apply a judicial exception on a computer per MPEP 2106.05 (f);
“outputting a prediction result generated by the abnormality prediction to a monitor or another apparatus via electronic transmission”; this limitation amounts to insignificant extra solution activity, mere data outputting as per MPEP 2106.05(g);
“wherein a maintenance on the device is performed based on the outputted prediction result so that downtime in operation of the device is reduced”; the recitation of a claim limitation that attempts to cover any solution to an identified problem (such as the downtime problem in an operation) with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result, does not integrate a judicial exception into a practical application or provide significantly more because this type of recitation is equivalent to the words "apply it" (see MPEP 2106.05 (f)).
Step 2B:
The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the additional elements are as follows:
(Claim 1) “a prediction system comprising: a device, a prediction apparatus comprising: a processor; a memory to store a program which, when executed by the processor, the processor performs processes of”; (Claim 10) “a non-transitory computer-readable storage medium storing a program that causes a computer with a processor to perform”; these limitations amount to nothing more than an instruction to apply the abstract idea using a generic computer as per MPEP 2106.05(f). Further, the device is merely described as the thing which the abnormality is being predicted, and this amounts to merely indicating a field of use or technological environment in which to apply a judicial exception as per MPEP 2106.05(h);
“acquires operation data indicating an operation state of a device”; this limitation amounts to insignificant extra solution activity, mere data gathering as per MPEP 2106.05(g); furthermore, this amounts to well-understood, routine, and conventional activity as per MPEP 2106.05(d): “i. Receiving or transmitting data over a network”
“a first prediction model”; this model is recited at a high level of generality to perform the calculations/ determinations of the probability density, interpreted as mere instructions to apply a judicial exception on a computer per MPEP 2106.05 (f);
“outputting a prediction result generated by the abnormality prediction to a monitor or another apparatus via electronic transmission”; this limitation amounts to insignificant extra solution activity, mere outputting as per MPEP 2106.05(g); furthermore, this amounts to well-understood, routine, and conventional activity as per MPEP 2106.05(d): “i. Receiving or transmitting data over a network”;
“wherein a maintenance on the device is performed based on the outputted prediction result so that downtime in operation of the device is reduced”; the recitation of a claim limitation that attempts to cover any solution to an identified problem (such as the downtime problem in an operation) with no restriction on how the result is accomplished and no description of the mechanism for accomplishing the result, does not integrate a judicial exception into a practical application or provide significantly more because this type of recitation is equivalent to the words "apply it" (see MPEP 2106.05 (f)).
Dependent Claims
Claim 2 recites “wherein the processor further performs the process of estimating the probability density using a variational Bayesian method”; estimating a pdf using a variational Bayesian method is a mathematical calculation.
Claim 3 recites “wherein the processor further performs the process of estimating a probability density of operation data for each of a plurality of operation modes of the device, and predicting an occurrence of an abnormality for each operation mode of the plurality of operation modes based on an estimation result of the probability density for each operation mode of the plurality of operation modes and a second prediction model for each operation mode of the plurality of operation modes”; estimating a probability density and making a prediction based on an estimation result and a model is an evaluation that can be performed by a human in the mind or with pen and paper, and is thus a mental process and mathematical concepts.
Claim 4 recites:
- “wherein the device is a rotary machine”; this amounts to merely indicating a field of use or technological environment in which to apply a judicial exception as per MPEP 2106.05(h);
- “the processor further performs the process of determining the operation mode of the plurality of operation modes based on an output and a rotation speed of the device”; determining a mode based on data like output and speed is an evaluation that can be performed by a human in the mind or with pen and paper, and is thus a mental process.
Claim 5 recites:
“wherein the processor further performs the processes of estimating the probability density of the operation data and the probability density of the operation data for the operation mode of the plurality of operation modes”; estimating a pdf is an evaluation that can be performed by a human in the mind or with pen and paper, and is thus a mental process;
“predicting whether or not the abnormality occurs in the device, based on the estimation result of the probability density of the operation data and the first prediction model, and predicting the occurrence of the abnormality for the operation mode of the plurality of operation modes based on the estimation result of the probability density for the operation mode of the plurality of operation modes and the second prediction model”; making a prediction based on an estimation result is an evaluation that can be performed by a human in the mind or with pen and paper, and is thus a mental process.
Claim 6 recites:
“wherein the processor further performs the processes of calculating a reliability of a prediction of the processor based on the prediction and an actual result of whether or not the abnormality has occurred for the prediction”; making a calculation based on a prediction and result is a mathematical concept. Further, the use of the processor amounts to nothing more than an instruction to apply the abstract idea using a generic computer as per MPEP 2106.05(f);
“wherein the processor calculates the reliability for each combination of predicted values based on each of the first prediction model and the second prediction model”; making a calculation based on a prediction and result is a mathematical concept. Further, the use of the processor amounts to nothing more than an instruction to apply the abstract idea using a generic computer as per MPEP 2106.05(f).
Claim 7 recites “wherein the processor further performs the process of creating a prediction model that predicts whether or not the abnormality occurs in the device, based on learning data in which the estimation result of the probability density estimated from the operation data in a predetermined period is associated with information indicating whether or not the abnormality has occurred in the device from which the operation data has been acquired in the predetermined period”; creating a prediction model based on data including mathematical values describing period of time and historical values to perform the calculations/ determinations of the probability density is interpreted as mere instructions to apply a judicial exception on a computer per MPEP 2106.05 (f).
Claim 11 recites:
“wherein the processor further performs the process of estimating a probability density of operation data for each of a plurality of operation modes of the device”; estimating a probability density is mathematical calculation, therefore, mathematical concepts. Further, the use of the processor amounts to nothing more than an instruction to apply the abstract idea using a generic computer as per MPEP 2106.05(f);
- “predicting an occurrence of an abnormality for each operation mode of the plurality of operation modes based on an estimation result of the probability density for each operation mode of the plurality of operation modes and a second prediction model for each operation mode of the plurality of operation modes”; making a prediction based on an estimation result (mathematical values based on mathematical calculations) and a model is an evaluation that can be performed by a human with pen and paper, and is a mental process and mathematical concepts.
Claim 12 recites:
“wherein the processor further performs the process of estimating the probability density of the operation data and the probability density of the operation data for the operation mode of the plurality of operation modes”; estimating a probability density is mathematical calculation, therefore, mathematical concepts. Further, the use of the processor amounts to nothing more than an instruction to apply the abstract idea using a generic computer as per MPEP 2106.05(f);
“predicting whether or not the abnormality occurs in the device, based on the estimation result of the probability density of the operation data and the first prediction model, and predicting the occurrence of the abnormality for the operation mode of the plurality of operation modes based on the estimation result of the probability density for the operation mode of the plurality of operation modes and the second prediction model”; estimating a probability density and making a prediction based on an estimation result and a model is an evaluation that can be performed by a human in the mind or with pen and paper, and is thus a mental process and mathematical concepts.
Claim 13 recites “wherein the processor further performs the process of creating a prediction model that predicts whether or not the abnormality occurs in the device, based on learning data in which the estimation result of the probability density estimated from the operation data in a predetermined period is associated with information indicating whether or not the abnormality has occurred in the device from which the operation data has been acquired in the predetermined period”; creating a prediction model based on data including mathematical values describing period of time and historical values to perform the calculations/ determinations of the probability density is interpreted as mere instructions to apply a judicial exception on a computer per MPEP 2106.05 (f).
Claim Rejections - 35 USC § 103
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, 7, 9-10, and 13 are rejected under 35 U.S.C. 103 as being unpatentable over San Martin (NPL: "Deep variational auto-encoders: A promising tool for dimensionality reduction and ball bearing elements fault diagnosis"; hereinafter "San Martin") in view of Anderson et al. (“US 2013/0232094”; hereinafter “Anderson”).
As per Claim 1, San Martin teaches a prediction system comprising: a device, a prediction apparatus comprising: processor; a memory to store a program which, when executed by the processor, the processor performs processes of (San Martin, Page 1093: left column “One approach to develop a methodology for the early detection of faults is to assume that the operational data have information about the health state of the system at the moment of its measure and, from there, use some technique to extract such information in a clear and clever way”. Therefore, this system is interpreted as the ‘device’. Further San Martin at Page 1107: “All the results shown in the subsequent sections were obtained using the following hardware and software configuration: i7 7700k processor, 32 GB of DDR4 RAM, and a Nvidia Titan X GPU with Tensorflow 1.2.0 and CUDNN 8.0”) comprising:
acquiring operation data indicating an operation state of the device (San Martin, Page 1093: left column “One approach to develop a methodology for the early detection of faults is to assume that the operational data have information about the health state of the system at the moment of its measure and, from there, use some technique to extract such information in a clear and clever way”. Therefore, this system is interpreted as the ‘device’. Further, San Martin, Page 1099 Left Column: “Step 1: Acquire vibration signals from the system under analysis.”)
estimating a probability density of the operation data (San Martin, Page 1099 Left Column: Step 2: From the original vibration signals, generate the dataset X that will be used to train the VAE … Step 3: Perform unsupervised training of the VAE model.”
San Martin, Page 1094 “Background”, “Variational Inference”: “One of the main challenges in probabilistic inference in general, and in fault diagnosis and prognosis in particular, is the approximation of difficult to compute probabilities densities. Currently, a powerful and flexible approach to approximate these probabilities densities is variational inference.”
San Martin, Page 1095 “The VAE model for dimensionality reduction”: “VAEs, originally proposed by Kingma and Welling, are generative models that combine neural networks, variational inference, and unsupervised learning to address the problem of finding an approximation to the posterior probability distribution p(zjx).”
Examiner notes that here, San Martin discloses taking the operation data (“original vibration signals”) and using it to train a VAE, which estimates a probability density (“approximate these probabilities densities”) of the operation data.)
predicting whether or not an abnormality occurs in the device, based on an estimation result of the probability density of the operation data and a first prediction model (San Martin, Page 1099 Left Column: “Step 4: Use the encoder of the trained VAE to transform the vectors of the original dataset, which has dimensionality equal to c, to the latent space of dimension k.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
Examiner notes that here, San Martin predicts whether or not an abnormality occurs (“fault diagnosis”), based on an estimation result of the probability density (“VAE-based latent representation”, which is based on the probability density) and a prediction model (“neural network classifier for fault diagnosis”)).
However, San Martin does not teach:
outputting a prediction result generated by the abnormality prediction to a monitor or another apparatus via electronic transmission;
wherein the first prediction model is a model that predicts whether or not an abnormality occurs in the device in the future within a predetermined period based on the probability density of the operation data between a first day when the abnormality has actually occurred and a second day that goes back the predetermined period from the first day;
wherein a maintenance on the device is performed based on the outputted prediction result so that downtime in operation of the device is reduced.
Anderson teaches, in an analogous system,
outputting a prediction result generated by the abnormality prediction to a monitor or another apparatus via electronic transmission (see Anderson at [0011-0012]: “The evaluation engine can further include an outage derived database to store outage derived data sets (ODDS) that capture dynamic precursor to fail data representative of at least one of the like components. In one embodiment, dynamic precursor to fail data is obtained from a time-shifted time domain ending at the time of, or just before, the failure and beginning at a pre-selected time prior to the failure”. Further, Anderson at [0014]: “evaluating the collection of propensity to failure metrics in an evaluation engine to detect and remove non-complying metrics from the collection of propensity to failure metrics and to provide the collection of filtered propensity to failure metrics; and (f) ranking the collection of filtered propensity to failure metrics obtained from the evaluation engine and displaying the ranking on a decision support application”. Further at [0101]: “A number of business management applications can use the data in the Output Data Repository to highlight areas of risk through graphical displays and map overlays”. Anderson is directed to predict preventive maintenance by using precursor data as training data in order to display propensity to failure metrics on a display, which is interpreted as ‘outputting a prediction result’);
wherein the first prediction model is a model that predicts whether or not an abnormality occurs in the device in the future within a predetermined period based on the probability density of the operation data between a first day when the abnormality has actually occurred and a second day that goes back the predetermined period from the first day (see Anderson at [0011-0012]: “The evaluation engine can further include an outage derived database to store outage derived data sets (ODDS) that capture dynamic precursor to fail data representative of at least one of the like components. In one embodiment, dynamic precursor to fail data is obtained from a time-shifted time domain ending at the time of, or just before, the failure and beginning at a pre-selected time prior to the failure”. Further, Anderson at [0022]: “FIG. 7 illustrates an example illustrating the training and test time windows in an Outtage Derived Data Set (ODDS). In this example, the current time is taken as Aug. 13, 2008 and the failure data for training is derived from Jul. 30, 2007 to Aug. 27, 2007 and Jul. 30, 2008--Aug. 13, 2008”. Therefore, this precursor to fail data used as training data has operation data from a time that a failure happened (interpreted as the first day) till a pre-selected time prior to the failure (interpreted as the second day that goes back the predetermined period from the first day). This data is used by Anderson for predicting failure and predicting maintenance, being analogous to the instant application);
wherein a maintenance on the device is performed based on the outputted prediction result so that downtime in operation of the device is reduced (see Anderson at [0008]: “The presently disclosed subject matter provides methods and systems for proactive predictive maintenance programs for electrical grid reliability”. Further, Anderson at [0048]: “Proactive Maintenance Tasks: Power companies are beginning to switch from reactive maintenance plans (fix when something goes wrong) to proactive maintenance plans (fix potential problems before they happen). There are advantages to this: reactive plans, which allow failures to happen, can lead to dangerous situations, for instance fires and cascading failures, and costly emergency repairs”. Further, Anderson at [0052]: “This added load elevates the risk of failure for the remaining feeders and transformers, and past a certain point, the network will experience a cascading failure, where the remaining distribution assets are unable to carry the network's load, and the entire network must be shut down until the system can be repaired”. Therefore, Anderson provides advantages for this proactive maintenance which include avoiding dangerous situations, avoiding cascading failures and avoiding shutting down systems, therefore, Anderson’s predictions will minimize downtime of an operation).
Anderson is analogous art because it is in the field of endeavor of predictive maintenance. It would have been obvious before the effective filing date of the claimed invention to combine the predictive maintenance system of San Martin with the proactive predictive maintenance system using precursor to failure data of Anderson. One of ordinary skill in the art would have been motivated to do so in order to be able to predict maintenance needed using precursor to fail data to develop proactive predictive maintenance programs, as suggested by Anderson at [0008]: “For example, the methods and systems of the present application, via machine learning, provide for proactive predictive maintenance of secondary components in electrical grid based on improved machine learning techniques and making use of data, obtained in the ordinary course of grid management, which was not designed for predictive purposes”).
As per Claim 2, the combination of San Martin, Anderson teaches the prediction system according to claim 1. San Martin teaches wherein the processor further performs the process of estimating the probability density using a variational Bayesian method. (San Martin, Page 1094 Left Column: “VAEs, originally proposed by Kingma and Welling,19 are generative models that combine neural networks, variational inference, and unsupervised learning to address the problem of finding an approximation to the posterior probability distribution p(zjx).” Examiner notes that the cited Kingma and Welling paper is called “Auto-encoding variational Bayes”).
As per Claim 7, the combination of San Martin and Anderson teaches the prediction system according to claim 1. San Martin teaches wherein the processor further performs the process of creating a prediction model that predicts whether or not the abnormality occurs in the device, based on learning data in which the estimation result of the probability density estimated from the operation data in a predetermined period is associated with information indicating whether or not the abnormality has occurred in the device from which the operation data has been acquired in the predetermined period. (San Martin, Page 1099 Left Column: “Step 4: Use the encoder of the trained VAE to transform the vectors of the original dataset, which has dimensionality equal to c, to the latent space of dimension k.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
San Martin, Page 1107 Left Column: “Train the chosen NN fault classifier with only the portion of labeled data extracted from the transformed dataset.”
Examiner notes that here, San Martin discloses a prediction model (“neural network classifier for fault diagnosis”) that predicts whether or not an abnormality occurs in the device (“fault diagnosis”). This model is created based on learning data (“series of experiments”) in which an estimation result of the probability density estimated from the operation data (“VAE-based latent representation”) in a predetermined period (“time scale”: San Martin Page 1095: “In the STFT, the color intensity of the image shows the energy level of that point in frequency and time, the horizontal axis is the time scale and the vertical axis is the frequency scale. In this article, we use STFT to obtain spectrograms from vibration signals”) is associated with information indicating whether or not the abnormality has occurred in the device from which the operation data has been acquired in the predetermined period (“true system’s health states are known”, “labeled data”)).
In addition, Anderson teaches at [0011-0012]: “The evaluation engine can further include an outage derived database to store outage derived data sets (ODDS) that capture dynamic precursor to fail data representative of at least one of the like components. In one embodiment, dynamic precursor to fail data is obtained from a time-shifted time domain ending at the time of, or just before, the failure and beginning at a pre-selected time prior to the failure”. Further, Anderson at [0022]: “FIG. 7 illustrates an example illustrating the training and test time windows in an Outtage Derived Data Set (ODDS). In this example, the current time is taken as Aug. 13, 2008 and the failure data for training is derived from Jul. 30, 2007 to Aug. 27, 2007 and Jul. 30, 2008--Aug. 13, 2008”. Therefore, this precursor to fail data used as training data used by Anderson for predicting failure and predicting maintenance is also analogous to the claimed “learning data”).
Anderson is analogous art because it is in the field of endeavor of predictive maintenance. It would have been obvious before the effective filing date of the claimed invention to combine the predictive maintenance system of San Martin with the proactive predictive maintenance system using precursor to failure data of Anderson. One of ordinary skill in the art would have been motivated to do so in order to be able to predict maintenance needed using precursor to fail data as learning data to develop proactive predictive maintenance programs, as suggested by Anderson at [0008]: “For example, the methods and systems of the present application, via machine learning, provide for proactive predictive maintenance of secondary components in electrical grid based on improved machine learning techniques and making use of data, obtained in the ordinary course of grid management, which was not designed for predictive purposes”).
As per Claim 9, this is a method claim corresponding to the system of Claim 1, and is rejected for similar reasons, mutatis mutandis.
As per Claim 10, this is a non-transitory computer readable storage medium claim corresponding to the system of Claim 1, and is rejected for similar reasons, mutatis mutandis.
As per Claim 13, the combination of San Martin and Anderson teaches the prediction apparatus according to claim 2. San Martin teaches wherein the processor further performs the process of creating a prediction model that predicts whether or not the abnormality occurs in the device, based on learning data in which the estimation result of the probability density estimated from the operation data in a predetermined period is associated with information indicating whether or not the abnormality has occurred in the device from which the operation data has been acquired in the predetermined period. (San Martin, Page 1099 Left Column: “Step 4: Use the encoder of the trained VAE to transform the vectors of the original dataset, which has dimensionality equal to c, to the latent space of dimension k.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
San Martin, Page 1107 Left Column: “Train the chosen NN fault classifier with only the portion of labeled data extracted from the transformed dataset.”
Examiner notes that here, San Martin discloses a prediction model (“neural network classifier for fault diagnosis”) that predicts whether or not an abnormality occurs in the device (“fault diagnosis”). This model is created based on learning data (“series of experiments”) in which an estimation result of the probability density estimated from the operation data (“VAE-based latent representation”) in a predetermined period (the period of the “series of experiments”) is associated with information indicating whether or not the abnormality has occurred in the device from which the operation data has been acquired in the predetermined period (“true system’s health states are known”, “labeled data”)).
In addition, Anderson teaches at [0011-0012]: “The evaluation engine can further include an outage derived database to store outage derived data sets (ODDS) that capture dynamic precursor to fail data representative of at least one of the like components. In one embodiment, dynamic precursor to fail data is obtained from a time-shifted time domain ending at the time of, or just before, the failure and beginning at a pre-selected time prior to the failure”. Further, Anderson at [0022]: “FIG. 7 illustrates an example illustrating the training and test time windows in an Outtage Derived Data Set (ODDS). In this example, the current time is taken as Aug. 13, 2008 and the failure data for training is derived from Jul. 30, 2007 to Aug. 27, 2007 and Jul. 30, 2008--Aug. 13, 2008”. Therefore, this precursor to fail data used as training data used by Anderson for predicting failure and predicting maintenance is also analogous to the claimed “learning data”).
Anderson is analogous art because it is in the field of endeavor of predictive maintenance. It would have been obvious before the effective filing date of the claimed invention to combine the predictive maintenance system of San Martin with the proactive predictive maintenance system using precursor to failure data of Anderson. One of ordinary skill in the art would have been motivated to do so in order to be able to predict maintenance needed using precursor to fail data as learning data to develop proactive predictive maintenance programs, as suggested by Anderson at [0008]: “For example, the methods and systems of the present application, via machine learning, provide for proactive predictive maintenance of secondary components in electrical grid based on improved machine learning techniques and making use of data, obtained in the ordinary course of grid management, which was not designed for predictive purposes”).
Claims 3, 5, and 11 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of San Martin, Anderson, further in view of Jiang et al. (“Multimode Process Monitoring Using Variational Bayesian Inference and Canonical Correlation Analysis”; hereinafter “Jiang”).
As per Claim 3, the combination of San Martin and Anderson teaches the prediction system according to claim 1. San Martin teaches wherein the processor further performs the processes of estimating a probability density of operation data [for each of a plurality of operation modes of the device], and predicting an occurrence of an abnormality [for each operation mode of the plurality of operation modes] based on an estimation result of the probability density [for each operation mode of the plurality of operation modes] and a [second] prediction model [for each operation mode of the plurality of operation modes] (San Martin, Page 1099 Left Column: Step 2: From the original vibration signals, generate the dataset X that will be used to train the VAE … Step 3: Perform unsupervised training of the VAE model.”
San Martin, Page 1094 “Background”, “Variational Inference”: “One of the main challenges in probabilistic inference in general, and in fault diagnosis and prognosis in particular, is the approximation of difficult to compute probabilities densities. Currently, a powerful and flexible approach to approximate these probabilities densities is variational inference.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
Examiner notes this is also explained above in the rejection to Claim 1, that San Martin estimates probability density for the operation data with a VAE, and then predicts abnormality using the VAE and a prediction model. However, San Martin does not teach for each of a plurality of operation modes of the device and a second prediction model for the operation mode of the plurality of operation modes.
Jiang teaches for each of a plurality of operation modes of the device and a second prediction model for the operation mode of the plurality of operation modes. (Jiang, Page 1817 Left Column: “The CCA monitoring model effectively handles linearly correlated process data from a single operation mode. However, a modern process can be characterized by multiple operation modes ... Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y. The local operation modes must be identified, and local monitoring models must be constructed.”
Here, San Martin has established, as a first model, what Jiang calls a “global model”. Furthermore, Jiang discloses a second model, called a “local model” which is created for each operation mode of the device.
Jiang is analogous art because it is in the field of endeavor of machine learning and condition monitoring. It would have been obvious before the effective filing date of the claimed invention to combine the fault detection of San Martin and Anderson with the local models for each operation mode of Jiang. One of ordinary skill in the art would have been motivated to do so in order to better monitor the behavior under the conditions of specific modes, because global level data can be insufficient for proper monitoring (Jiang Page 1817 Left Column: “Thus, local CCA models must be established to explore local linear correlations among variables. For example, a process is operated in three different modes during a period, and the process data are distributed, as shown in Fig. 1. Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y.”)
As per Claim 5, the combination of San Martin, Anderson and Jiang teaches the prediction system according to claim 3. San Martin teaches wherein the processor further performs the processes of estimating the probability density of the operation data and the probability density of the operation data [for the operation mode of the plurality of operation modes], and predicting whether or not the abnormality occurs in the device, based on the estimation result of the probability density of the operation data and the first prediction model (San Martin, Page 1099 Left Column: Step 2: From the original vibration signals, generate the dataset X that will be used to train the VAE … Step 3: Perform unsupervised training of the VAE model.”
San Martin, Page 1094 “Background”, “Variational Inference”: “One of the main challenges in probabilistic inference in general, and in fault diagnosis and prognosis in particular, is the approximation of difficult to compute probabilities densities. Currently, a powerful and flexible approach to approximate these probabilities densities is variational inference.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
Examiner notes this is also explained above in the rejection to Claim 1, that San Martin estimates probability density for the operation data with a VAE, and then predicts abnormality using the VAE and a prediction model. However, San Martin does not teach for the operation mode of the plurality of operation modes; and predicting the occurrence of the abnormality for the operation mode of the plurality of operation modes based on the estimation result of the probability density for the operation mode of the plurality of operation modes and the second prediction model.
Jiang teaches for the operation mode of the plurality of operation modes; and predicting the occurrence of the abnormality for the operation mode of the plurality of operation modes based on the estimation result of the probability density for the operation mode of the plurality of operation modes and the second prediction model. (Jiang, Page 1817 Left Column: “The CCA monitoring model effectively handles linearly correlated process data from a single operation mode. However, a modern process can be characterized by multiple operation modes ... Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y. The local operation modes must be identified, and local monitoring models must be constructed.”
Here, San Martin has established, as a first model, what Jiang calls a “global model”. Furthermore, Jiang discloses a second model, called a “local model” which is created for each operation mode of the device.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Jiang with San Martin and Anderson for at least the reasons recited in the rejection to Claim 3.
As per Claim 11, the combination of San Martin and Anderson teaches the prediction system according to claim 2. San Martin teaches wherein the processor further performs the processes of estimating a probability density of operation data [for each of a plurality of operation modes of the device], and predicting an occurrence of an abnormality [for each operation mode of the plurality of operation modes] based on an estimation result of the probability density for [each operation mode of the plurality of operation modes] and a [second] prediction model for [each operation mode of the plurality of operation modes] (San Martin, Page 1099 Left Column: Step 2: From the original vibration signals, generate the dataset X that will be used to train the VAE … Step 3: Perform unsupervised training of the VAE model.”
San Martin, Page 1094 “Background”, “Variational Inference”: “One of the main challenges in probabilistic inference in general, and in fault diagnosis and prognosis in particular, is the approximation of difficult to compute probabilities densities. Currently, a powerful and flexible approach to approximate these probabilities densities is variational inference.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
Examiner notes this is also explained above in the rejection to Claim 1, that San Martin estimates probability density for the operation data with a VAE, and then predicts abnormality using the VAE and a prediction model. However, San Martin does not teach for each of a plurality of operation modes of the device and a second prediction model for each operation mode of the plurality of operation modes.
Jiang teaches for each of a plurality of operation modes of the device and a second prediction model for each operation mode of the plurality of operation modes. (Jiang, Page 1817 Left Column: “The CCA monitoring model effectively handles linearly correlated process data from a single operation mode. However, a modern process can be characterized by multiple operation modes ... Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y. The local operation modes must be identified, and local monitoring models must be constructed.”
Here, San Martin has established, as a first model, what Jiang calls a “global model”. Furthermore, Jiang discloses a second model, called a “local model” which is created for each operation mode of the device.
Jiang is analogous art because it is in the field of endeavor of machine learning and condition monitoring. It would have been obvious before the effective filing date of the claimed invention to combine the fault detection of San Martin and Anderson with the local models for each operation mode of Jiang. One of ordinary skill in the art would have been motivated to do so in order to better monitor the behavior under the conditions of specific modes, because global level data can be insufficient for proper monitoring (Jiang Page 1817 Left Column: “Thus, local CCA models must be established to explore local linear correlations among variables. For example, a process is operated in three different modes during a period, and the process data are distributed, as shown in Fig. 1. Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y.”)
Claims 4 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of San Martin, Anderson, and Jiang, further in view of Timusk et al. (“Automated Operating Mode Classification for Online Monitoring Systems”; hereinafter “Timusk”).
As per Claim 4, the combination of San Martin, Anderson and Jiang teaches the prediction apparatus according to claim 3 as well as operation mode (see Jiang in rejection to Claim 3).
San Martin teaches wherein the device is a rotary machine (San Martin, Page 1101 Left Column: “In the CWR dataset, a Reliance electric motor with 2 horsepower was used with ball bearings in experiments for the acquisition of vibration data on both the drive end and fan end bearings … Also, 0 to 3 horsepower motor loads were used in the experiments, with motor speeds ranging from 1720 to 1797 r/min.”)
However, combination does not teach the processor further performs the process of determining the operation mode of the plurality of operation modes based on an output and a rotation speed of the device.
Timusk teaches the processor further performs the process of determining the operation mode of the plurality of operation modes based on an output and a rotation speed of the device. (Timusk, Page 4 “3.4 Choice of Signals” discloses: “Vibration and speed are chosen as the parameters on which the mode classification would be based.”)
Timusk is analogous art because it is in the field of endeavor of machine monitoring. It would have been obvious before the effective filing date of the claimed invention to combine the multi-mode fault prediction process of San Martin, Anderson and Jiang with the identification of the operating mode using vibration and rotational speed of Timusk. One of ordinary skill in the art would have been motivated to do so because these features are common to several types of systems and are thus broadly applicable (Timusk, Page 4 “3.4 Choice of Signals”: “Using these common diagnostic parameters allows the same basic program structure to be applied directly to other applications such as hydraulic swing machinery or even the planetary final drives on a haul truck. The common denominator between applications is the vibration signals coupled with an accurate synchronized speed measurement.”)
As per Claim 12, the combination of San Martin, Anderson, Jiang, and Timusk teaches the prediction system according to claim 4. San Martin teaches wherein the processor further performs the processes of estimating the probability density of the operation data and the probability density of the operation data [for the operation mode of the plurality of operation modes], and predicting whether or not the abnormality occurs in the device, based on the estimation result of the probability density of the operation data and the first prediction model (San Martin, Page 1099 Left Column: Step 2: From the original vibration signals, generate the dataset X that will be used to train the VAE … Step 3: Perform unsupervised training of the VAE model.”
San Martin, Page 1094 “Background”, “Variational Inference”: “One of the main challenges in probabilistic inference in general, and in fault diagnosis and prognosis in particular, is the approximation of difficult to compute probabilities densities. Currently, a powerful and flexible approach to approximate these probabilities densities is variational inference.”
San Martin, Page 1106 Left Column: “To validate the proposed approach, a series of experiments are performed, where the VAE-based latent representation is used to train a neural network classifier for fault diagnosis, in which the true system’s health states are known.”
Examiner notes this is also explained above in the rejection to Claim 1, that San Martin estimates probability density for the operation data with a VAE, and then predicts abnormality using the VAE and a prediction model. However, San Martin does not teach for the operation mode of the plurality of operation modes; and predicting the occurrence of the abnormality for the operation mode of the plurality of operation modes based on the estimation result of the probability density for the operation mode of the plurality of operation modes and the second prediction model.
Jiang teaches for the operation mode of the plurality of operation modes; and predicting the occurrence of the abnormality for the operation mode of the plurality of operation modes based on the estimation result of the probability density for the operation mode of the plurality of operation modes and the second prediction model. (Jiang, Page 1817 Left Column: “The CCA monitoring model effectively handles linearly correlated process data from a single operation mode. However, a modern process can be characterized by multiple operation modes ... Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y. The local operation modes must be identified, and local monitoring models must be constructed.”
Here, San Martin has established, as a first model, what Jiang calls a “global model”. Furthermore, Jiang discloses a second model, called a “local model” which is created for each operation mode of the device.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Jiang with San Martin and Anderson for at least the reasons recited in the rejection to Claim 3.
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over the combination of San Martin, Anderson, and Jiang further in view of Huber et al. (US 2020/0257943 A1; hereinafter “Huber”).
As per Claim 6, the combination of San Martin, Anderson, and Jiang teaches the prediction system according to claim 5. San Martin teaches wherein the processor further performs the processes of calculating a reliability of a prediction of the abnormality prediction unit based on the prediction and an actual result of whether or not the abnormality has occurred for the prediction (San Martin, Page 1107 Left Column, discloses: “Use the test dataset to evaluate the MLP fault classifier performance” San Martin, Page 1101 Bottom Left, discloses: “For this article, the fault locations and fault sizes are included as individual classes, thus resulting in 12 health states (classes) in this dataset.” Here, San Martin discloses the use of labeled data if abnormality occurs (“health states”) to test the reliability (“performance”) of the prediction model (“fault classifier”)).
However, San Martin does not explicitly teach wherein the processor calculates the reliability for a combination of predicted values based on each of the first prediction model and the second prediction model
Jiang teaches first prediction model and the second prediction model (Jiang, Page 1817 Left Column: “The CCA monitoring model effectively handles linearly correlated process data from a single operation mode. However, a modern process can be characterized by multiple operation modes ... Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y. The local operation modes must be identified, and local monitoring models must be constructed.”
Here, San Martin has established, as a first model, what Jiang calls a “global model”. Furthermore, Jiang discloses a second model, called a “local model” which is created for each operation mode of the device.
combination of predicted values (Jiang, Page 1818 Right Column: “The proposed VBGMM-CCA integrates the results from all local models into the BIP index.”)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Jiang with San Martin and Anderson for at least the reasons recited in the rejection to Claim 3.
However, Examiner notes this amounts to a combination of multiple “second models” (“all local models”), and therefore does not explicitly teach reliability for a combination of predicted values based on each of the first prediction model and the second prediction model. While it may appear obvious that one could also combine results of a “global model” and a “local model”, Examiner will bring in another reference for clarity.
Huber teaches wherein the processor calculates the reliability for a combination of predicted values based on each of the first prediction model and the second prediction model (Huber, Para [0149], discloses: “Some aspects of embodiments of the present invention relate to combining generalized time series models (to forecast events) with specialized forecasting models for specific topic domains.” Huber, Para [0171], discloses: “In still other embodiments of the present invention, algorithms that weight the elements of the ensemble based on their confidence … are used to generate an output forecast.”
Here, Huber discloses calculating reliability (“confidence”) for each combination of predicted values based on a first prediction model (“generalized time series models”) and a second prediction model (“specialized forecasting models”). These “generalized” and “specialized” models are analogous to the “global” and “local” models disclosed by Jiang, and thus the combination with Huber discloses calculating a reliability for a combination of them.
Huber is analogous art because it is in the field of endeavor of machine learning. It would have been obvious before the effective filing date of the claimed invention to combine the global and mode-specific probabilistic abnormality prediction method of San Martin and Jiang with the combining via confidence of Huber. One of ordinary skill in the art would have been motivated to do so in order to take into account operation-specific behavior (Jiang, Page 1817: “Modeling the process with a global model ignores the local process behavior, particularly in exploring the correlation between input u and output y”) while also guarding against incorrect mode identification (Jiang, Page 1818: “This feature takes the advantages of probabilistic monitoring in avoiding the effects of incorrect mode identification, thereby increasing monitoring robustness.”) Also, ensembles can lead to a more accurate result (Huber, [0114]: “In some embodiments, updated models are added to an ensemble of models (discussed in more detail below) and may contribute to the final forecast computed in operation 321, where the weight of the user's modified model in the final output is determined by factors such as their historical accuracy and the historical accuracy of the models they produce.”)
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to LUIS A SITIRICHE whose telephone number is (571)270-1316. The examiner can normally be reached M-F 9am-6pm.
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.
/LUIS A SITIRICHE/Primary Examiner, Art Unit 2126
Prosecution Insights