DETAILED ACTION
This office action is responsive to the Request for Reconsideration-After Non-Final filed 6/9/2026. The application contains claims 1, 3-11, 13-21, 23-24, all examined and rejected.
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 .
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, 3-11, 13-21, 23-24 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter. Claim 1 is rejected under 35 USC 101 because the claimed inventions are directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more.
While independent claims 1, 10 and 11 are each directed to a statutory category, it recites a series of steps pertaining to analyze received data to identify features that are used to predict machine failure, which appears to be directed to an abstract idea (mental process, mathematical concept).
Claims 1, 3-11, 13-21, 23-24 are rejected under 35 U.S.C. § 101 because the instant application is directed to non-patentable subject matter. Specifically, the claims are directed toward at least one judicial exception without reciting additional elements that amount to significantly more than the judicial exception. The rationale for this determination is in accordance with the guidelines of USPTO, applies to all statutory categories, and is explained in detail below.
When considering subject matter eligibility under 35 U.S.C. 101, (1) it must be determined whether the claim is directed to one of the four statutory categories of invention, i.e., process, machine, manufacture, or composition of matter. If the claim does fall within one of the statutory categories, (2a) it must then be determined whether the claim is directed to a judicial exception (i.e., law of nature, natural phenomenon, and abstract idea), and if so (2b), it must additionally be determined whether the claim is a patent-eligible application of the exception. If an abstract idea is present in the claim, any element or combination of elements in the claim must be sufficient to ensure that the claim amounts to significantly more than the abstract idea itself. Examples of abstract ideas include certain methods of organizing human activities; a mental processes; and mathematical concepts, (2019 PEG)
STEP 1.
Per Step 1, the claims are determined to include process, manufacture, and machine as in independent Claim 1, 10, and 11, and in the therefrom dependent claims. Therefore, the claims are directed to a statutory eligibility category.
At step 2A, prong 1, The invention is directed to identifying features within received data that could be an indication of the probability of occurrence of a machine failure based on analyzed historic data which is akin to Mental Process (see Alice), As such, the claims include an abstract idea. When considering the limitations individually and as a whole the limitations directed to the abstract idea are:
“generating a plurality of data features based on at least a portion of the sensor data”, “selecting, from the plurality of data features, at least one indicative data feature for a machine failure detection”, “detect machine failure indicators based on the selected at least one indicative data feature”, “determining, selected at least one indicative data feature that is associated with the new sensor data, whether at least one machine failure indicator was detected in the new sensor data”, “tagging the at least one machine failure indicator upon determination that the at least one machine failure indicator was detected, wherein upon determination that no machine failure indicators were detected, continuously searches for machine failure indicators (Mental process, observation, evaluation and judgment) “selecting, from the plurality of data features, at least one indicative data feature for machine failure prediction; applying to the selected at least one indicative data feature a supervised machine failure prediction process; wherein the supervised machine failure prediction process is configured to predict machine failures based on the selected at least one indicative data feature; and updating the supervised machine failure prediction process with the tagged at least one machine failure indicator, such that the supervised machine failure prediction process is continuously and automatically updated and improved” (Mental process, observation, evaluation and judgment).
The claim recites additional elements as
“online computer server”, “online machine learning based method for detection and prediction of industrial machine failures”, “non-transitory computer readable medium having stored thereon instructions for Causing a processing circuitry to perform a process”, “a processing circuitry; and a memory, the memory containing instructions that, when executed by the processing circuitry” (“Using a computer as a tool to perform a mental process”, MPEP 2106.04(a)(2)(III)(C));
receiving sensor data related to at least one industrial machine (insignificant extra-solution activity, MPEP 2106.05(g));
“wherein the selected at least one indicative data feature for machine failure detection is different from the selected at least one indicative data feature for machine failure prediction” (data description, which is directed to generally linking the use of a judicial exception to a particular technological environment or type or source of data or field of use MPEP 2106.05(h));
“applying an unsupervised machine failure”, “unsupervised machine failure detection process is configured”, “applying the unsupervised machine failure detection process to the selected at least one indicative data feature”, “unsupervised machine failure detection process “ (merely indicates a field of use or technological environment in which the judicial exception is performed and fails to add an inventive concept to the claims. See MPEP 2106.05(h)).
This judicial exception is not integrated into a practical application. The elements are recited at a high level of generality, i.e. a generic computing system performing generic functions including generic processing of data. Accordingly the additional elements do not integrate the abstract into a practical application because it does not impose any meaningful limits on practicing the abstract idea. Therefore the claims are directed to an abstract idea. (2019 Revised Patent Subject Matter Eligibility Guidance ("2019 PEG"). Thus, under Step 2A of the Mayo framework, the Examiner holds that the claims are directed to concepts identified as abstract.
STEP 2B.
Because the claims include one or more abstract ideas, the examiner now proceeds to Step 2B of the analysis, in which the examiner considers if the claims include individually or as an ordered combination limitations that are "significantly more" than the abstract idea itself. This includes analysis as to whether there is an improvement to either the "computer itself," "another technology," the "technical field," or significantly more than what is "well-understood, routine, or conventional" (WURC) in the related arts.
The instant application includes in Claim 1 additional steps to those deemed to be abstract idea(s).
When taken the steps individually, these steps are:
“online computer server”, “online machine learning based method for detection and prediction of industrial machine failures”, “non-transitory computer readable medium having stored thereon instructions for Causing a processing circuitry to perform a process”, “a processing circuitry; and a memory, the memory containing instructions that, when executed by the processing circuitry” (“Using a computer as a tool to perform a mental process”, MPEP 2106.05(f)(2));
receiving sensor data related to at least one industrial machine (WELL-UNDERSTOOD, ROUTINE, CONVENTIONAL ACTIVITY, sending, receiving, displaying and processing data are common and basic functions in computer technology, MPEP 2106.05(d)(II)(i));
“wherein the selected at least one indicative data feature for machine failure detection is different from the selected at least one indicative data feature for machine failure prediction” (data description, which is directed to generally linking the use of a judicial exception to a particular technological environment or type or source of data or field of use MPEP 2106.05(h));
“applying an unsupervised machine failure”, “unsupervised machine failure detection process is configured”, “applying the unsupervised machine failure detection process to the selected at least one indicative data feature”, “unsupervised machine failure detection process “ (merely indicates a field of use or technological environment in which the judicial exception is performed and fails to add an inventive concept to the claims. See MPEP 2106.05(h) and mere instructions to “apply” the abstract ideas, which cannot provide an inventive concept. See MPEP 2106.05(f));
In the instant case, Claim 1 is directed to above mentioned abstract idea. Technical functions such as receiving, and extracting are common and basic functions in computer technology. The individual limitations are recited at a high level and do not provide any specific technology or techniques to perform the functions claimed.
In addition, when the claims are taken as a whole, as an ordered combination, the combination of steps does not add "significantly more" by virtue of considering the steps as a whole, as an ordered combination. The instant application, therefore, still appears only to implement the abstract idea to the particular technological environments using what is well-understood, routine, and conventional in the related arts. The steps are still a combination made to the abstract idea. The additional steps only add to those abstract ideas using well understood and conventional functions, and the claims do not show improved ways of, for example, an unconventional non-routine functions for analyzing model operations or updating the model that could then be pointed to as being "significantly more" than the abstract ideas themselves.
Moreover, Examiner was not able to identify any "unconventional" steps, which, when considered in the ordered combination with the other steps, could have transformed the nature of the abstract idea previously identified. The instant application, therefore, still appears to only implement the abstract ideas to the particular technological environments using what is well-understood, routine, and conventional (WURC) in the related arts.
Further, note that the limitations, in the instant claims, are done by the generically
recited computing devices. The limitations are merely instructions to implement the abstract idea on a computing device that is recited in an abstract level and require no more than a generic computing devices to perform generic functions.
Claim 10 recites a system comprising “non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry” configured to perform the same method as set forth in claim 1, the added element of “non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry to perform a process” do not transform the judicial exception into a practical application because they are amount to a mere instruction to apply the judicial exception to a generic computer. The additional elements are also not sufficient to amount to significantly more than the judicial exception because the action of implementing the method on a general purpose computer with non-transitory computer readable medium having stored thereon instructions for causing a processing circuitry is mere instruction to apply the judicial exception to a computer.
Claim 10 is therefore rejected according to the same findings and rationale as provided above.
Claim 11 recites a system comprising processing circuitry; and a memory, the memory containing instructions that, when executed by the processing circuitry” configured to perform the same method as set forth in claim 1, the added element of “processing circuitry; and a memory, the memory containing instructions that, when executed by the processing circuitry” do not transform the judicial exception into a practical application because they are amount to a mere instruction to apply the judicial exception to a generic computer. The additional elements are also not sufficient to amount to significantly more than the judicial exception because the action of implementing the method on a general purpose computer with processing circuitry; and a memory, the memory containing instructions that, when executed by the processing circuitry is mere instruction to apply the judicial exception to a computer.
Claim 11 is therefore rejected according to the same findings and rationale as provided above.
Independent claims 10 and 11 are the same analogy and rejected using similar analysis as claim 1.
CONCLUSION
It is therefore determined that the instant application not only represents an abstract idea identified as such based on criteria defined by the Courts and on USPTO examination guidelines, but also lacks the capability to bring about "Improvements to another technology or technical field" (Alice), bring about "Improvements to the functioning of the computer itself" (Alice), "Apply the judicial exception with, or by use of, a particular machine" (Bilski), "Effect a transformation or reduction of a particular article to a different state or thing" (Diehr), "Add a specific limitation other than what is well-understood, routine and conventional in the field" (Mayo), "Add unconventional steps that confine the claim to a particular useful application" (Mayo), or contain "Other meaningful limitations beyond generally linking the use of the judicial exception to a particular technological environment" (Alice), transformed a traditionally subjective process performed by humans into a mathematically automated process executed on computers (McRO), or limitations directed to improvements in computer related technology, including claims directed to software (Enfish).
The dependent claims, when considered individually and as a whole, likewise do not provide "significantly more" than the abstract idea for similar reasons as the independent claim.
claims 3 disclose “wherein the plurality of data features represents a behavior of at least a component of the at least one industrial machine” data description , which is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 4 disclose “wherein the plurality of data features is generated based on at least one statistical method.” data description , which is directed to generally linking the use of a judicial exception to a particular technological environment or field of use. It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 5 disclose “wherein the at least one indicative data feature is selected from the plurality of data features based on a probability to detect machine failures”, (mental and mathematical concept), It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 6 disclose “wherein the at least one indicative data feature is selected from the plurality of data features based on a probability to predict machine failures”(mental and mathematical concept), It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 7 disclose “selecting a plurality of indicative data features from the plurality of data features based on at least a distribution of the plurality of indicative data features, wherein the at least a distribution indicates at least an association between the plurality of data features towards a machine failure” (mental and mathematical concept), It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 8 disclose “wherein at least a portion of the sensor data is previously tagged with at least one machine failure indicator.” It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea. Claim 9 disclose “wherein determining whether at least one machine failure indicator were detected in the new sensor data is based on semi- supervised machine learning”, It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea. Claim 20 disclose “wherein the at least one of the selected indicative data feature for machine failure prediction is selected when it is determined that a portion of the selected indicative data feature for machine failure prediction has a better probability to contribute more to predicting a machine failure with respect to others of the plurality of data features by identifying in the portion an increasing change in a distribution of the selected indicative data feature prior to a machine failure with respect to a normal state of the at least one industrial machine” (mental and mathematical concept), It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 21 disclose “wherein the selected at least one indicative data feature for machine failure prediction comprises at least two indicative data feature for machine failure prediction and wherein the at least two indicative data features for machine failure prediction are selected such that abnormal parameters of at least two of the at least two indicative data features for machine failure prediction demonstrate an association, wherein such association is indicative of a forthcoming machine failure.” (mental and mathematical concept), It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 23 disclose “the tagging is performed after the determining by applying the unsupervised machine failure detection process to the selected at least one indicative data feature associated with the new sensor data that the at least one machine failure indicator was detected” (data description, which is directed to generally linking the use of a judicial exception to a particular technological environment or type or source of data or field of use MPEP 2106.05(h)); It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea; claim 24 disclose “the sensor data is received in real-time” (data description, which is directed to generally linking the use of a judicial exception to a particular technological environment or type or source of data or field of use MPEP 2106.05(h)); It does not integrate the abstract idea into a practical application and did not add significantly more to the abstract idea.
The dependent claims which impose additional limitations also fail to claim patent eligible subject matter because the limitations cannot be considered statutory. The dependent claim(s) have been examined individually and in combination with the preceding claims, however they do not cure the deficiencies of claim 1 ; where all claims are directed to the same abstract idea, "addressing each claim of the asserted patents [is] unnecessary." Content Extraction &. Transmission LLC v, Wells Fargo Bank, Natl Ass'n, 776 F.3d 1343, 1348 (Fed. Cir. 2014). If applicant believes the dependent claims are directed towards patent eligible subject matter, they are invited to point out the specific limitations in the claim that are directed towards patent eligible subject matter. Claims for the other statutory classes are similarly analyzed.
For at least these reasons, the claimed inventions of each of dependent claims 3-9, 13-21, 23-24 are directed or indirect to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more and are rejected under 35 USC 101.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 3-8, 10-11, 13-18, 20-21, and 23-24 are rejected under 35 U.S.C. 103 as being unpatentable over Bates et al. [US 2017/0083830 A1, hereinafter D1] further in view of “Degradation Feature Selection for Remaining Useful Life Prediction of Rolling Element Bearings” Published 2015 [hereinafter D2].
With regard to Claim 1,
D1 teach an online machine learning based method performed by an online computer server for detection and prediction of industrial machine failures (Fig. 1, ¶32), comprising:
receiving, by the computer server, sensor data related to at least one industrial machine (¶12, “processor is configured by computer code to receive sensor data relating to the unit of equipment”, ¶¶33-35, ¶38);
Generating, by the computer server, a plurality of data features based on at least a portion of the sensor data (¶36, ¶54, “Importing the sensor data leading up to and including a failure condition allows the failure signature recognition system to identify what leads up to the failure condition”, ¶68, “input would be a vector of length 24 * 10=240 for each time step, since the input would contain current data as well as prior data”);
selecting, by the computer server, from the plurality of data features, at least one indicative data feature for a machine failure detection (¶47, “failure identification module 330 provides a screen … that allows a user to identify failures from maintenance work order history … which work orders represent failures”, ¶54, “signature of a failure is a characteristic pattern of sensor readings, oscillations, some changing variable, etc. … Importing the sensor data leading up to and including a failure condition allows the failure signature recognition system to identify what leads up to the failure condition, not just the failure condition”, ¶40);
applying, by the computer server, to the selected at least one indicative data feature an unsupervised machine failure detection process, wherein the unsupervised machine failure detection process is configured to detect machine failure indicators based on the selected at least one indicative data feature (¶54, ¶87, “anomaly detection component 220 utilizes a Kohonen self organizing map (SOM) to perform the analysis”, ¶89, “Kohonen Self-Organizing Map (SOM) methodology essentially clusters tag data for each time step into an output, which can be thought of as an operating state”);
receiving, by the computer server, new sensor data related to the at least one industrial machine (¶91, “Agent feeds the new data into the trained SOM model, which classifies it into one of the known operating states”, ¶76, “failure signature recognition component 210 receives, via the plant data interface 240, current trend data from plant historians related to the plant data sources”);
determining, by the computer server, by applying the unsupervised machine failure detection process to the selected at least one indicative data feature that is associated with the new sensor data, whether at least one machine failure indicator was detected in the new sensor data (¶92, “Anomaly Detection works is, it compares the error E of the current classification to the maximum error detected on the Training DataSet, E′. If E exceeds E′ by a factor T, known as the Anomaly Threshold, then an Anomaly Alert is generated”, ¶95, “probability (P, returned by f(x)) is compared to the minimum baseline probability calculated from the Training DataSet (P′). If P is smaller than P′ by a factor T, known as the Anomaly Threshold … an Anomaly Alert is generated”); and
tagging, by the computer server, the at least one machine failure indicator upon determination that the at least one machine failure indicator were detected, wherein upon determination that no machine failure indicators was detected, the unsupervised machine failure detection process continuously searches for machine failure indicators (¶¶84-85, ¶91, “anomaly agent is activated as a live profile for monitoring. The anomaly agents can monitor the new sensor data during the process 1100 in the same way that the failure agents monitor the new sensor data. The Agent feeds the new data into the trained SOM model, which classifies it into one of the known operating states”, ¶92, “Anomaly Detection works is, it compares the error E of the current classification to the maximum error detected on the Training DataSet, E′. If E exceeds E′ by a factor T, known as the Anomaly Threshold, then an Anomaly Alert is generated”, ¶95); and
selecting, by the computer server, from the plurality of data features, at least one indicative data feature for machine failure prediction (¶40, ¶44, “user can select from a list of tags listed in a tag data store shown in the screen 410. Each tag corresponds to a sensor associated with the pump selected with the screen 405 in this example. A sensor could be associated with an operating parameter of the pump such as pressure or temperature. For each tag in the screen”, ¶47);
applying, by the computer server, to the selected at least one indicative data feature a supervised machine failure prediction process (¶54, “At stage 1025, the learning agent training module 340 analyzes the sensor data at times leading up to and during the identified failures ... By identifying when a failure occurs for a given asset, the sensor data leading up to the failure and during the failure can be identified”, ¶57, “training at stage 1025 involves creating a failure agent that takes in the sensor data in the training set and, using machine learning, parameters of the failure agent are adjusted such that the failure agent successfully predicts the identified failures before the failures occur”, ¶55, “Machine learning techniques such as Resilient Back Propagation (RPROP), Logistic Regression (LR), and Support Vector machines (SVM) can all be used at stage 1025”);
wherein the supervised machine failure prediction process is configured to predict machine failures based on the selected at least one indicative data feature (¶¶54-55, ¶57, “training at stage 1025 involves creating a failure agent that takes in the sensor data in the training set and, using machine learning, parameters of the failure agent are adjusted such that the failure agent successfully predicts the identified failures before the failures occur”, ¶58); and
updating the supervised machine failure prediction process with the tagged at least one machine failure indicator, such that the supervised machine failure prediction process of the computer server is continuously and automatically updated and improved (Fig. 11, ¶79, ¶84, “Due to the retraining at stages 1125 and 1135, the process 1100 allows a failure agent to adapt itself over time, becoming more and more fine-tuned for the equipment it is monitoring”).
D1 does not explicitly the selected at least one indicative data feature for machine failure detection is different from the selected at least one indicative data feature for machine failure prediction.
D2 teach generating, by the computer server, a plurality of data features based on at least a portion of the sensor data (P. 2, 2. Degradation feature selection “condition monitoring data are sampled in the data acquisition process, and some pre-conditioning operations are made. Then candidate degradation features are generated from condition monitoring data using signal processing techniques”, P. 3, 2.1. Feature generation, “Candidate prognostic features can be generated by processing time domain, frequency domain and time-frequency domain of original condition monitoring signals”, P. 5, 3.2, Results and analysis, “10 common statistical features and 16 WPNE features are generated for each horizontal or vertical vibration signals, making a total of 52 features for each testing bearing” );
selecting, by the computer server, from the plurality of data features, at least one indicative data feature for a machine failure detection (P. 2, “Features with larger interclass and smaller intraclass distances are selected in diagnostic feature evaluation”, 2.1. Feature generation, “Statistical indices that are effective in fault diagnosis, such as RMS and wavelet packet node energy(WPNE)”);
selecting, by the computer server, from the plurality of data features, at least one indicative data feature for machine failure prediction (P. 2, “features retained for prognostics should have better predictabilities with trend, robustness and so on”, “candidate deteriorative features are evaluated and optimal ones are selected in the feature selection step“, P. 3, 2.2, “good prognostic features should be well correlated with item performance degradation progressing, monotonically increasing or decreasing, robust to outliers and common across individual item and soon. Thus correlation, monotonicity and robustness based on trend and residual are proposed here for more relevant degradation features election”, 2.3, “features of high J score should be retained for effective and efficient RUL estimation”); wherein selected at least one indicative data feature for machine failure detection (P. 2, ¶2, “Statistical indices that are effective infault diagnosis, such as RMS and wavelet packet node energy (WPNE), have been considered here”, P. 3, Table 1, P. 5, ¶3.2, “both horizontal and vertical vibration signals are pre-processed to extract candidate degradation features”) is different from the selected at least one indicative data feature for machine failure prediction (P. 2, ¶3, “Unlike static point clustering in diagnostic feature evaluation, a sequence of consecutive realizations should be considered in prognostic feature evaluation because degradation is a continuous stochastic process “, “Features with larger interclass and smaller intraclass distances are selected in diagnostic feature evaluation, while features retained for prognostics should have better predictabilities with trend, robustness and so on“, “features retained for prognostics should have better predictabilities with trend, robustness and so on”, P. 2, 2.2, “good prognostic features should be well correlated with item performance degradation progressing, monotonically increasing or decreasing, robust to outliers and common across individual item and soon. Thus correlation, monotonicity and robustness based on trend and residual are proposed here for more relevant degradation feature selection”, D2 teaches that diagnostic feature selection criteria differ from prognostic feature selection criteria which results in different retained or selected features).
D1 and D2 are analogous art to the claimed invention because they are from a similar field of endeavor of machine prognostics and health management for industrial equipment. Thus, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify D1 resulting in resolutions as disclosed by D2 with a reasonable expectation of success.
One of ordinary skill in the art would be motivated to modify D1 as described above to improve prognostic accuracy by selecting features associated with degradation (D2, P. 2, 2.2, “good prognostic features should be well correlated with item performance degradation progressing, monotonically increasing or decreasing, robust to outliers and common across individual item and soon. Thus correlation, monotonicity and robustness based on trend and residual are proposed here for more relevant degradation feature selection”) which is Simple substitution of one known element for another to obtain predictable results; usage of known technique to improve similar devices (methods, or products) in the same way; and Combining prior art elements according to known methods to yield predictable results (MPEP 2143).
With regard to Claim 3,
D1-D2 disclose the method of claim 1, wherein the plurality of data features represents a behavior of at least a component of the at least one industrial machine (D1, ¶57, “failures for equipment where a false negative can be catastrophic such as an oil rig”, ¶59, ¶61). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 4,
D1-D2 disclose the method of claim 1, wherein the plurality of data features is generated based on at least one statistical method (D1, ¶93, “Gaussian algorithm fits a probability distribution to each tag (variable) in the Training DataSet, estimating the mean u and standard deviation σ from the data. With these parameters estimated, the Gaussian probability function is used for each tag”). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 5,
D1-D2 disclose the method of claim 1, wherein the at least one indicative data feature is selected from the plurality of data features based on a probability to detect machine failures (D1, ¶¶53-54, ¶93, “Gaussian algorithm fits a probability distribution to each tag (variable) in the Training DataSet, estimating the mean u and standard deviation σ from the data. With these parameters estimated, the Gaussian probability function is used for each tag”, ¶94, “For a given time step, the value for each tag Xi is fed into the Gaussian function for that tag (with the associated mean and standard deviation), and the probability is calculated”, ¶95, “After the probability is calculated for each tag for a given time step, these probabilities are multiplied together to get the overall probability (based on assumption of independence of the random variables for each tag). The probability (P, returned by f(x)) is compared to the minimum baseline probability calculated from the Training DataSet (P′). If P is smaller than P′ by a factor T, known as the Anomaly Threshold, then the new tag data is considered to be an anomaly, and an Anomaly Alert is generated”). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 6,
D1-D2 disclose the method of claim 1, wherein the at least one indicative data feature is selected from the plurality of data features based on a probability to predict machine failures (D1, ¶¶53-54, ¶93, “Gaussian algorithm fits a probability distribution to each tag (variable) in the Training DataSet, estimating the mean u and standard deviation σ from the data. With these parameters estimated, the Gaussian probability function is used for each tag”, ¶94, “For a given time step, the value for each tag Xi is fed into the Gaussian function for that tag (with the associated mean and standard deviation), and the probability is calculated”, ¶95, “After the probability is calculated for each tag for a given time step, these probabilities are multiplied together to get the overall probability (based on assumption of independence of the random variables for each tag). The probability (P, returned by f(x)) is compared to the minimum baseline probability calculated from the Training DataSet (P′). If P is smaller than P′ by a factor T, known as the Anomaly Threshold, then the new tag data is considered to be an anomaly, and an Anomaly Alert is generated”). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 7,
D1-D2 disclose the method of claim 1, further comprising:
selecting a plurality of indicative data features from the plurality of data features based on at least a distribution of the plurality of indicative data features (D1, ¶47, ¶87, “anomaly detection component 220 analyzes sensor data at times where conditions are normal in order to determine baseline or normal operating conditions. In one aspect, the anomaly detection component 220 utilizes a Kohonen self organizing map (SOM) to perform the analysis at stage”, ¶88, “analysis at stage 1225 can comprise BIC (Bayesian Information Criteria) to determine the number of regions (e.g., the groups 910 and 920). Gausian probability can be used to determine the odds that sensor A (temperature) is one value and sensor B (pressure) is one value and this can detect the anomaly”, wherein the at least a distribution indicates at least an association between the plurality of data features towards a machine failure (D1, ¶40, “failure signature recognition component 210 uses pattern recognition techniques to learn when failures are about to occur. The failure signature recognition component identifies fault conditions in the work order histories of the CM system 110, takes the sensor data from the plant data sources and learns failure signatures based on the sensor data”, ¶42, ¶88, “An anomaly agent is trained to detect an anomaly when the current operating state of a piece of equipment is outside of the first group 910 and the second group 920. The analysis at stage 1225 can comprise BIC (Bayesian Information Criteria) to determine the number of regions (e.g., the groups 910 and 920). Gausian probability can be used to determine the odds that sensor A (temperature) is one value and sensor B (pressure) is one value and this can detect the anomaly”). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 8,
D1-D2 disclose the method of claim 1, wherein at least a portion of the sensor data is previously tagged with at least one machine failure indicator (D1, ¶49, “training data set importer module 320 retrieves a set of training data comprising sensor data corresponding to all the tags identified at stage 1005 that exhibit changes during the identified failures for the selected asset”, ¶50, ¶52, “imported data is stored with metadata to flag which intervals are failure intervals versus normal intervals”). The same motivation to combine for claim 1 equally applies for current claim.
With regards to claim 10,
Claim 10 is similar in scope to claim 1; therefore it is rejected under similar rationale. Further D1 teach non-transitory computer readable medium having stored thereon instructions for Causing a processing circuitry to perform a process (¶¶9-10).
With regards to claim 11,
Claim 11 is similar in scope to claim 1; therefore it is rejected under similar rationale. Further D1 teach a processing circuitry; and a memory, the memory containing instructions that, when executed by the processing circuitry (¶¶9-10).
With regards to claim 13,
Claim 13 is similar in scope to claim 3; therefore it is rejected under similar rationale.
With regards to claim 14,
Claim 14 is similar in scope to claim 4; therefore it is rejected under similar rationale.
With regards to claim 15,
Claim 15 is similar in scope to claim 5; therefore it is rejected under similar rationale. With regards to claim 16,
Claim 16 is similar in scope to claim 6; therefore it is rejected under similar rationale. With regards to claim 17,
Claim 17 is similar in scope to claim 7; therefore it is rejected under similar rationale. With regards to claim 18,
Claim 18 is similar in scope to claim 8; therefore it is rejected under similar rationale.
With regard to Claim 20,
D1-D2 disclose the method of claim 1, wherein the at least one of the selected (D1, ¶44, “user can select from a list of tags listed in a tag data store shown in the screen 410. Each tag corresponds to a sensor associated with the pump selected with the screen 405 in this example. A sensor could be associated with an operating parameter of the pump such as pressure or temperature”, ¶45, “ user interface 270 renders a user interface screen 420 shown in FIG. 4E. The screen 420 is used to create a sensor template for the chosen asset (the pump)”) indicative data feature for machine failure prediction is selected when it is determined that a portion of the selected indicative data feature for machine failure prediction has a better probability to contribute more to predicting a machine failure with respect to others of the plurality of data features (D1, ¶51, “After the user selects to execute the import of the training data with the screen 440, the training data set importer module 320 displays a screen 445, as shown in FIG. 4J, that shows sensor data for normal conditions both before and after a portion 446 of training data that includes the identified failure”, ¶55, “Machine learning techniques such as Resilient Back Propagation (RPROP), Logistic Regression (LR), and Support Vector machines (SVM) can all be used at stage 1025. RPROP can be used for certain non-linear patterns, LR enables ranking of tag prediction rank, and SVM enables confidence intervals for prediction”, ¶63, “learning agent training module 340 uses different spans of time to identify the optimal time interval using Receiver Operating Characteristic methodology and Area Under Curve (AUC) methodology”) by identifying in the portion an increasing change in a distribution of the selected indicative data feature prior to a machine failure (D1, ¶54, “The signature of a failure is a characteristic pattern of sensor readings, oscillations, some changing variable, etc. By identifying when a failure occurs for a given asset, the sensor data leading up to the failure and during the failure can be identified. Importing the sensor data leading up to and including a failure condition allows the failure signature recognition system to identify what leads up to the failure condition, not just the failure condition”, ¶66, “When analyzing a memory process or non-Markov process, one looks at the past readings for a period of time to sense the signature. Historyless (memoryless) processes, in contrast, are analyzed at each time period independently and the analysis tries to learn what is different in the failure period compared to the normal periods. As described below, one can vary the memory settings to get the optimum prediction interval”, ¶67, “output of the Agent depends on previous time steps in addition to the current time step”, ¶70, “FIG. 8 shows a graph 800 including a first trace 810 and a second trace 820 from two different sensors. In this example of a failure signature with memory, the amplitude is about the same before and after the failure, but the frequency changes”) with respect to a normal state of the at least one industrial machine (¶51, “After the user selects to execute the import of the training data with the screen 440, the training data set importer module 320 displays a screen 445, as shown in FIG. 4J, that shows sensor data for normal conditions both before and after a portion 446 of training data that includes the identified failure”, ¶87, “ the anomaly detection component 220 analyzes sensor data at times where conditions are normal in order to determine baseline or normal operating conditions”). Examiner notes that a full mapping has been provided for compact persecution. However, “machine failure prediction is selected when it is determined that a portion of the selected indicative data feature” the “when” clause is a contingent clause that is non-limiting in scope See MPEP 2111.04. The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 21,
D1-D2disclose the method of claim 1, wherein the selected at least one indicative data feature for machine failure prediction comprises at least two indicative data feature for machine failure prediction (D1, ¶68, “If there is data from 10 tags in the current training data set, then, with no memory, the input to the machine learning agent would be a vector of length 10 for each time step”, ¶40, “failure signature recognition component 210 uses pattern recognition techniques to learn when failures are about to occur. The failure signature recognition component identifies fault conditions in the work order histories of the CM system 110, takes the sensor data from the plant data sources and learns failure signatures based on the sensor data”) and wherein the at least two indicative data feature for machine failure prediction are selected such that abnormal parameters of at least two of the at least two indicative data feature for machine failure prediction demonstrate an association (D1, ¶54, “The signature of a failure is a characteristic pattern of sensor readings, oscillations, some changing variable, etc. By identifying when a failure occurs for a given asset, the sensor data leading up to the failure and during the failure can be identified. Importing the sensor data leading up to and including a failure condition allows the failure signature recognition system to identify what leads up to the failure condition, not just the failure condition”, ¶70, “FIG. 8 shows a graph 800 including a first trace 810 and a second trace 820 from two different sensors. In this example of a failure signature with memory, the amplitude is about the same before and after the failure, but the frequency changes”, ¶41, “ the anomaly detections component 220 can look at temperature and pressure time histories and identify abnormal measurements based on trained learning agents”), wherein such association is indicative of a forthcoming machine failure (D1, ¶41, “ learning agents of the anomaly detection component are trained to identify an anomaly in the sensor data before a failure occurs. If an anomaly is detected, the affected equipment can be shut down and inspected to identify what may be causing the anomaly before a catastrophic failure occurs”, system interprets associated sensor behavior (across multiple features) as an indication of a forthcoming machine failure, ¶54, “ learning agent training module 340 analyzes the sensor data at times leading up to and during the identified failures … allows the failure signature recognition system to identify what leads up to the failure condition”). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 23,
D1-D2 disclose the method of claim 1, wherein the tagging is performed after the determining by applying the unsupervised machine failure detection process to the selected at least one indicative data feature associated with the new sensor data that the at least one machine failure indicator was detected (¶87, “the anomaly detection component 220 utilizes a Kohonen self organizing map (SOM) to perform the analysis at stage 1225“, ¶91, “The Agent feeds the new data into the trained SOM model, which classifies it into one of the known operating states, and returns the output state along with the classification error E“, ¶92, “it compares the error E of the current classification to the maximum error detected on the Training DataSet, E′. If E exceeds E′ by a factor T, known as the Anomaly Threshold, then an Anomaly Alert is generated. Whenever an Anomaly is detected and determined to be a valid predictor of a fault, a supervised learning profile (Failure Signature Recognition) agent is created to learn the specifics of the new signature, and flagged with extra metadata about the specifics of the fault and remedy. In this format, the system goes from anomalies to failure signatures (with improved recommended corrective action)”). The same motivation to combine for claim 1 equally applies for current claim.
With regard to Claim 24,
D1-D2 disclose the method of claim 1, wherein the sensor data is received in real-time (D1, ¶4, “control system data is real-time data measured in terms of seconds”, ¶5, “Manufacturers are drowning in a flood of real-time and non-real time data “).The same motivation to combine for claim 1 equally applies for current claim.
Claims 9 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Bates et al. [US 2017/0083830 A1, hereinafter D1] further in view of “Degradation Feature Selection for Remaining Useful Life Prediction of Rolling Element Bearings” Published 2015 [hereinafter D2] in view of Gundel et al . [US 2021/0373063 A1, hereinafter Gundel].
With regard to Claim 9,
D1-D2 teach the method of claim 1, wherein determining whether at least one machine failure indicator were detected in the new sensor data (D1, ¶54, “At stage 1025, the learning agent training module 340 analyzes the sensor data at times leading up to and during the identified failures ... By identifying when a failure occurs for a given asset, the sensor data leading up to the failure and during the failure can be identified”, ¶57, “training at stage 1025 involves creating a failure agent that takes in the sensor data in the training set and, using machine learning, parameters of the failure agent are adjusted such that the failure agent successfully predicts the identified failures before the failures occur”, ¶55, “Machine learning techniques such as Resilient Back Propagation (RPROP), Logistic Regression (LR), and Support Vector machines (SVM) can all be used at stage 1025”). The same motivation to combine for claim 1 equally applies for current claim.
D1-D2 does not explicitly teach
Gundel teach determining whether at least one machine failure indicator were detected in the new sensor data is based on semi-supervised machine learning (¶¶73-74, ¶77, “Example machine learning techniques that may be employed to generate models 74C can include various learning styles, such as supervised learning, unsupervised learning, and semi-supervised learning”).
D1-D2 and Gundel are analogous art to the claimed invention because they are from a similar field of endeavor of predicting machine failure. Thus, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify D1-D2 resulting in resolutions as disclosed by Gundel with a reasonable expectation of success.
One of ordinary skill in the art would be motivated to modify D1-D2 as described above to reduce labeling cost that can be time-consuming and expensive process specially in the availability of limited amount of labeled data as this is Simple substitution of one known element for another to obtain predictable results; usage of known technique to improve similar devices (methods, or products) in the same way; and Combining prior art elements according to known methods to yield predictable results (MPEP 2143).
With regard to Claim 19,
Claim 19 is similar in scope to claim 9; therefore it is rejected under similar rationale.
Response to Arguments
Applicant argue that the Office Action to mischaracterize the claim as data collection and analysis and abstracting away the particular computer-server implementation and the ordered interaction of the recited machine-learning processes. As the independent claims do not merely analyze machine data and report a result. Rather, they recite a computer server architecture in which a supervised machine failure prediction process is continuously and automatically updated using tagged machine failure indicators detected by a separate unsupervised machine failure detection process. Thus, the claim is directed to innovatively improving the operation of the machine failure prediction process itself, rather than merely generating information or results from received machine sensor data.
Examiner respectfully disagrees; machine failure process prediction is broad term that could be completely manual as the user could have a process to monitor failure and update the monitoring process based on previous experience. This fall under mental process as a user is able to monitor machines and predict failure (e.g. if machine temperature increase to a specific degree and it is getting close to a specific threshold, then a failure is expected, in addition the threshold could be updated based on previous failure that occurred previously at temperature less that the threshold). In addition the current claims does not provide a special or new way of training machine learning model for prediction task that could be considered that it is not WURC. Using and training supervised and unsupervised machine learning model merely indicates a field of use or technological environment in which the judicial exception is performed and fails to add an inventive concept to the claims. See MPEP 2106.05(h). Applicant is welcomed to provide specific citation from the specification that reflects limitations in the claims that show the difference between the state of the art of the time of the invention and the current inventions and how this difference is considered an improvement to the technology. Regarding the continuous update of the machine learning models, examiner notes that training and tuning of machine learning model is WURC activity and cannot overcome the abstract 35 USC 101 rejection unless the specification that reflect limitations in the claims disclose a novel or unique specific for of tuning/training/updating of machine learning model.
Applicant argue that paragraph 19 of the specification as filed support that the claims provide an improvement in the functioning of a computer and to other technology so as to integrate any alleged abstract ideas into a practical application. In particular, not only is the computer operation improved by updating the supervised machine failure prediction process but the whole machine monitoring system itself is improved. Note that, essentially, the computer is a self-programming system that improves its operation over time. Thus, the arrangement as claimed is not a computer that merely executes a fixed algorithm on new data but rather is a computer that changes how it will operate in the future in an improved manner.
Examiner respectfully disagrees, paragraph 19 of the specification as filed disclose a known process of fine tuning models. Retraining models based on received feedback and data for improvement cannot be considered as an improvement to the technology unless the invention is disclosing a new form of retraining that differentiate the disclosed retraining or tuning of the model from normal well known retraining methods as clarified in the specification.
Applicant argue that the claims are clearly not directed merely to manually monitoring a machine or updating a threshold based on experience. Rather, the claims recite a particular online computer server implementation in which sensor data is processed to generate data features, a selected indicative data feature is applied to an unsupervised machine failure detection process, a detected machine failure indicator is tagged, a different selected indicative data feature is applied to a supervised machine failure prediction process, and the supervised machine failure prediction process is continuously and automatically updated and improved with the tagged machine failure indicator. This detailed ordered machine-learning architecture specifically for implementation by a computer is not reasonably characterized as a manual mental process.
Examiner respectfully disagrees; examiner did not characterize the whole claim as a mental process. Analyzing, tagging and identifying data is mental process. However the other claimed limitations as collecting data is well understood and conventional activity, and the usage of machine learning merely indicates a field of use or technological environment in which the judicial exception is performed and fails to add an inventive concept to the claims. See MPEP 2106.05(h).
Applicant argue that the claims, like those in Enfish, are directed to a specific implementation of a solution in the software arts, namely deficiencies in the functioning of computers used for monitoring behavior of a machine in order to detect anomalies. In particular, the claims are directed to development of an improved machine monitoring system. Thus, Applicants' claims do not recite an abstract idea for the same reasons that the claims in Enfish did not recite an abstract idea.
Examiner respectfully disagrees, the claim disclose abstract idea See at least “generating a plurality of data features based on at least a portion of the sensor data”, “selecting, from the plurality of data features, at least one indicative data feature for a machine failure detection”, “detect machine failure indicators based on the selected at least one indicative data feature”, “determining, selected at least one indicative data feature that is associated with the new sensor data, whether at least one machine failure indicator was detected in the new sensor data”, “tagging the at least one machine failure indicator upon determination that the at least one machine failure indicator was detected, wherein upon determination that no machine failure indicators were detected, continuously searches for machine failure indicators (Mental process, observation, evaluation and judgment) “selecting, from the plurality of data features, at least one indicative data feature for machine failure prediction; applying to the selected at least one indicative data feature a supervised machine failure prediction process; wherein the supervised machine failure prediction process is configured to predict machine failures based on the selected at least one indicative data feature; and updating the supervised machine failure prediction process with the tagged at least one machine failure indicator, such that the supervised machine failure prediction process is continuously and automatically updated and improved” (Mental process, observation, evaluation and judgment). Further examiner notes that to consider that the claims are directed to development of an improved machine monitoring system, the specification as filed as reflect in the claims should disclose the state of art and how the new method or machine monitoring system differentiate from the previous system and how that solve or improve the technology.
Applicant argue that the computer is not recited in high level of generality as the claims recite supervised and unsupervised machine learning models in a specific way.
Examiner respectfully disagrees; the claims does not use the model in any specific form that differentiate using it from any other form.
Applicant argue that the rejection untethers the claim from its actual claim limitations and treats the claim as if it merely covers observing data and predicting failure. It seems that the Office Action feels at liberty to redefine what the claim actually calls for, which is clearly improper. Rather, the claim must be evaluated as a whole and in view of all recited limitations, including the ordered relationship between the unsupervised machine learning
detection process, tagging, supervised machine failure prediction, and updating of the
supervised machine failure prediction.
Examiner respectfully disagrees; the Office Action provided a clear specific rejection for every limitation in the claims.
Applicant argue that the Response to Arguments further improperly collapses: what is being called for in the claim, i.e., the specific application of two types of machine learning and how it is done, i.e., the specific functional loop. Note that the claims recite a specific technical mechanism by which a plurality of data features is generated and then at least one of the plurality is selected for machine failure detection using unsupervised machine learning and at least one of the plurality is selected for machine failure prediction using supervised machine learning. The claim then requires updating the supervised machine failure prediction process with the tagged at least one machine failure indicator, such that the supervised machine failure prediction process is continuously and automatically updated and improved. In other words, the supervised prediction process is updated using knowledge of the tagged failure indicators from the unsupervised machine leaning process.
Examiner respectfully disagrees, the response to the arguments address the core of the argument, using supervised and unsupervised machine leaning models is not an improvement to the technology and cannot be considered extra elements that overcome the abstract rejection. As previously clarified by the examiner the arguments need to connect the claims to the specification that disclose the state of the art upon the filing of the invention and how the current invention improve what is well known to solve a problem in the field.
Applicant argue that the provided architecture is not WURC as shown in regards to the rejection based on prior art. In addition the provided architecture enable continuous improvement.
Examiner respectfully disagrees. While § 101 subject matter eligibility is a threshold test that typically precedes the novelty or obviousness inquiry (Bilski v. Kappos, 561 U.S. 593, 602 (2010)), it is a requirement separate from those other patentability inquiries. Return Mail, Inc. v. USPS, 123 USPQ2d 1813, 1827 (Fed. Cir. 2017); see also Mayo Collaborative Servs v. Prometheus Labs, Inc., 566 U.S. 66, 90 (2012). Even assuming that a particular claimed feature is novel does not avoid the problem of abstractness. Affinity Labs of Tex., LLC v. DirecTV, LLC, 120 USPQ2d 1201, 1206 (Fed. Cir. 2016). In addition, tuning or retraining model is not an improvement to the technology.
Applicant argue that it is not possible for a human mind to do the disclosed claims without a computer and that the usage of the computer goes far beyond what is acceptable as a mental process, which is always described as a person in their mind or using a pen and paper. The computer server is not merely displaying information for a person to evaluate. It performs the claimed feature generation, feature selection, unsupervised detection, tagging, supervised prediction, and updating operations. None of this would be done by a person, certainly not in their head with pen and paper. Moreover, even if the computer displayed the sensor data to a person, i.e., as a set of numbers or a graph, a person would have no practical way to understand them or know what to do with them.
Examiner respectfully disagrees, applicant is mixing mental steps with computer usage. Human is able to analyze data and to select features. Using a machine learning model is limitations that link the use of a judicial exception to a particular technological environment or field of use. This does not integrate the abstract idea into a practical application. See MPEP 2106.05(h). In addition training a machine learning model is a high-generic computer software process of training data. This limitation does not amount to significantly more than the judicial exception, see MPEP 2106.05 (f). Regarding the argument provided related to the ability of human to handle and use data. It is unclear why a human would have no practical way to understand them or know what to do with them. For example if a temperature reach a specific degree, then there is a probability of failure and machine must be turned off.
Applicant argue that a person cannot practically process such real-time sensor data to perform the requisite machine monitoring. That is why, fundamentally, these systems are needed.
Examiner respectfully disagrees,
It is unclear why a person cannot do that mentally. User can observe, analyze data and identify if the data is out of specific range. Using a computer as a tool to display data is considered a “Using a computer as a tool to perform a mental process”, MPEP 2106.04(a)(2)(III)(C) and sending and receiving data is WURC activity See at least MPEP 2106.05(d)(II)(i) sending, receiving, displaying and processing data are common and basic functions in computer technology. Using supervised and unsupervised machine learning model merely indicates a field of use or technological environment in which the judicial exception is performed and fails to add an inventive concept to the claims. See MPEP 2106.05(h).
Applicant argue that user looking at displayed data does not meet the claimed requirements that include the usage of supervised and unsupervised machine learning models.
Examiner respectfully disagrees; the rejection did not map all the claimed limitation to the human looking at displayed data. Therefore the argument is unrelated to the actual rejection provided.
Applicant argue that user that the human mind cannot analyze or otherwise process the sensory inputs as called for in the claim in real time, and certainly not in the manner and according to the specific arrangement called for in the claims.
Examiner respectfully disagrees; human has the ability to monitor systems and analyze received data in real time using a computer as a tool.
Applicant argue that the examiner misunderstood the provided arguments and the arguments did not argue two separate selection steps then the applicant argue that D1 does not disclose two selection steps. Applicant further argue that D1 does not disclose 1) a first selection for selecting data feature an unsupervised process used for detecting machine failure indicators and 2) a second, separate selection of at least one indicative data feature for a supervised machine failure prediction process used for predicting machine failures based on a different selecting step. Instead, D1 treats its features as a shared input space used by its agents across functions. In other words, D1 simply does not teach or suggest two selection steps. Moreover, even if the data features in each selecting step of the claim overlap, each act of selection is distinct and purposive, i.e., having or serving a purpose, or acting with intention. By contrast, D1 does not disclose feature selection keyed to the type of machine learning to which they will be applied, i.e., to unsupervised or supervised learning.
Examiner respectfully disagrees; the asserted distinction appears to impose an exclusivity that is not required by the claims. The fact that the claim recite an unsupervised ML process for detecting a machine failure indicator and a supervised ML process for machine failure prediction does not mean that the unsupervised process must be incapable of prediction or that the supervised process must be incapable of detection. In other words, the system could employ unsupervised model to perform both prediction and detection and supervised model to perform both prediction and detection. This will satisfy the claim limitations, as the supervised process perform the required prediction function and the unsupervised process perform the required prediction function. Moreover, D1 do not merely treat the features as a shared space. As discussed above, D1 disclose selection of features for the respective detection and prediction functions. In particular, D2 expressly distinguishes diagnostic features evaluation for prognostic features evaluation, teaching that “Features with larger interclass and smaller intraclass distances are selected in diagnostic feature evaluation, while features retained for prognostics should have better predictabilities with trend, robustness and so on”. Thus, the teaching disclose feature selection directed to the respective detection and prediction purposes.
Also, D1 expressly disclose that the ML techniques used to generate model 74C may include supervised learning, unsupervised model, and semi-supervised learning (¶77). In addition ¶¶70-71 describe the analytics framework as both detecting anomalies and predicting failure events.
Applicant argue that Bates treats its features as a shared input space used by its agents across functions and thus Bates simply does not teach or suggest two selection steps in sequence as explained hereinabove. Thus, Applicants were correct in pointing out that Bates is using the same data from a single selection for both failure detection and anomaly detection. As such, unlike the allegation in the Response to Arguments regarding Bates, the claim actually requires two separate selecting steps, i.e., that two selections be made, and that at least one feature be selected for each stated purpose in
each of the respective selecting steps. Moreover, with regard to amended claims, Bates does not teach or suggest selecting a detection feature that is different from the prediction feature. The Office Action appears to continue to rely on a common feature set in Bates to satisfy both the detection-feature selection limitation and the prediction-feature selection limitation. In this regard, the Response to Arguments appears to be saying that if Bates selects feature X and uses feature for both anomaly detection and prediction, that still reads on the claim. However, even before the amendment, broadest reasonable
interpretation cannot make the arrangement of Bates according to the Office Action into
two separate selection steps, i.e., two separate operations, as called for in the claim. In
other words, the claim requires two separate operations, not two uses of data. In particular, doing so fails to permit that different items be selected in the different steps,
which is clearly something that the claim even prior to amendment permitted. Moreover,
with the amendment of the claim to include therein claim 22, the selected features must
be different, which cannot be achieved in Bates as explained by the Office Action. Thus,
in view of the claim amendment, Bates does not merely need different downstream models. Bates must disclose selecting a first indicative data feature for machine failure
detection and selecting a different indicative data feature for machine failure prediction.
Examiner respectfully disagrees,
applicant previously indicated that the prior arguments had been misunderstood, but again argue that D1 uses the same data from a single selection for both detection and prediction function and therefore does not teach two separate selection steps. applicant assertion that, even prior to the amendment, the claim require two separate selections operation is not supported by the claim language. The claim merely recited selecting at least one indicative data feature for failure detection and selecting at least one indicative data feature for failure prediction, it did not require the selection to be done independently, or at different times. Thus a common selection operation satisfy both limitations where the selected features are selected for both stated purposes. The limitation for different features was introduced in the amendments and is separately addressed by D2.
Regarding the arguments related to the new amendments that require the detection feature selected to be different that the prediction feature selected. This is taught by D2 that disclose generating, by the computer server, a plurality of data features based on at least a portion of the sensor data (P. 2, 2. Degradation feature selection “condition monitoring data are sampled in the data acquisition process, and some pre-conditioning operations are made. Then candidate degradation features are generated from condition monitoring data using signal processing techniques”, P. 3, 2.1. Feature generation, “Candidate prognostic features can be generated by processing time domain, frequency domain and time-frequency domain of original condition monitoring signals”, P. 5, 3.2, Results and analysis, “10 common statistical features and 16 WPNE features are generated for each horizontal or vertical vibration signals, making a total of 52 features for each testing bearing” ); selecting, by the computer server, from the plurality of data features, at least one indicative data feature for a machine failure detection (P. 2, “Features with larger interclass and smaller intraclass distances are selected in diagnostic feature evaluation”, 2.1. Feature generation, “Statistical indices that are effective in fault diagnosis, such as RMS and wavelet packet node energy(WPNE)”); selecting, by the computer server, from the plurality of data features, at least one indicative data feature for machine failure prediction (P. 2, “features retained for prognostics should have better predictabilities with trend, robustness and so on”, P. 3, 2.2, “good prognostic features should be well correlated with item performance degradation progressing, monotonically increasing or decreasing, robust to outliers and common across individual item and soon. Thus correlation, monotonicity and robustness based on trend and residual are proposed here for more relevant degradation features election”, 2.3, “features of high J score should be retained for effective and efficient RUL estimation”); wherein selected at least one indicative data feature for machine failure detection (P. 2, ¶2, “Statistical indices that are effective infault diagnosis, such as RMS and wavelet packet node energy (WPNE), have been considered here”, P. 3, Table 1) is different from the selected at least one indicative data feature for machine failure prediction (P. 2, ¶3, “Unlike static point clustering in diagnostic feature evaluation, a sequence of consecutive realizations should be considered in prognostic feature evaluation because degradation is a continuous stochastic process “, “Features with larger interclass and smaller intraclass distances are selected in diagnostic feature evaluation, while features retained for prognostics should have better predictabilities with trend, robustness and so on“, “features retained for prognostics should have better predictabilities with trend, robustness and so on”, P. 2, 2.2, “good prognostic features should be well correlated with item performance degradation progressing, monotonically increasing or decreasing, robust to outliers and common across individual item and soon. Thus correlation, monotonicity and robustness based on trend and residual are proposed here for more relevant degradation feature selection”, D2 teaches that diagnostic feature selection criteria differ from prognostic feature selection criteria which results in different retained or selected features). D1 and D2 are analogous art to the claimed invention because they are from a similar field of endeavor of machine prognostics and health management for industrial equipment. Thus, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to modify D1 resulting in resolutions as disclosed by D2 with a reasonable expectation of success. One of ordinary skill in the art would be motivated to modify D1 as described above to improve prognostic accuracy by selecting features associated with degradation (D2, P. 2, 2.2, “good prognostic features should be well correlated with item performance degradation progressing, monotonically increasing or decreasing, robust to outliers and common across individual item and soon. Thus correlation, monotonicity and robustness based on trend and residual are proposed here for more relevant degradation feature selection”) which is Simple substitution of one known element for another to obtain predictable results; usage of known technique to improve similar devices (methods, or products) in the same way; and Combining prior art elements according to known methods to yield predictable results (MPEP 2143).
Applicant argue that Zhang teaches diagnostic feature selection and prognostic feature selection may use different evaluation criteria Zhang fails to teach or suggest that the diagnostic feature selection is for unsupervised machine failure detection and prognostic feature selection is for supervised machine failure prediction. Thus, Zhang fails to teach or suggest applying its different feature selections in the specific manner called for in the claim.
Examiner respectfully disagrees, D2 is not relied on to teach the argued limitations therefore the applicant arguments are not related to the presented rejection.
Applicant argue that Zhang is merely teaching a method that determines types of features that are considered desirable to use for either diagnostic feature evaluation or prognostic feature evaluation. Zhang does not disclose selecting separate feature sets for separate machine-learning processes for an actual operating machine based on the machine's sensor data, much less for an unsupervised machine failure detection process and a supervised machine failure prediction process as claimed. In other words, Zhang is directed to determining which types of features to use for particular arrangements for
determining remaining useful life.
Examiner respectfully disagrees,
Applicant’s mischaracterize D2 as merely identifying desirable types of feature. D2 expressly teaches that candidate features are “candidate degradation features are generated from condition monitoring data using signal processing techniques”, “candidate deteriorative features are evaluated and optimal ones are selected in the feature selection step” (P. 2, 2 Degradation feature selection) and experimentally selects optimal features from generated bearing vibration features for RUL prediction. Therefore, D2 teaches actual feature selection from machine sensor data, not merely predetermined feature types. Regarding the usage of supervisor and unsupervised process, D2 is not relied on to teach the argued process therefore the applicant arguments are not related to the presented rejection.
Applicant argue that Zhang merely recognizes that diagnostic feature evaluation and prognostic feature evaluation may employ different evaluation criteria. However, Zhang treats diagnostics and prognostics as separate analytical tasks and does not teach or suggest selecting a first indicative feature for a machine failure detection process and
selecting a different indicative feature for a machine failure prediction process within a common architecture. Moreover, Zhang does not disclose any interaction between diagnostic and prognostic processes, much less the claimed arrangement in which detected machine failure indicators are tagged and used to update a supervised machine failure prediction process. As such, not only does Bates not anticipate the invention as claimed even prior to the amendment, but even with the amendment the proposed combination fails to teach or suggest the invention as claimed.
Examiner respectfully disagrees,
applicant’s argument improperly require D2 to independently teach the entire claimed architecture. The rejection is an obvious rejection that depend on a combination of D1 that teach the argued limitations and D2 to modify D1 to include the ability to distinguish features selected for diagnostic evaluation from features retained for prognostics. Thus, as clarified above, D1 in view of D2 teaches the requirements as claimed.
Applicant argue that Bates does not teach the claimed ordered relationship. The claim requires tagging the machine failure indicator upon determining that the machine failure indicator was detected by applying the unsupervised machine failure detection process to the selected indicative data feature associated with new sensor data, and then updating the supervised machine failure prediction process with the tagged machine failure indicator. The cited metadata, anomaly alerts, work requests, and sensor tags in Bates are not used in this manner, are not the kind of data to be tagged according to the claims, and do not disclose this claimed feedback path. Bates requires sensor data and known failure information relating to equipment failures. There is no teaching or suggestion in Bates that such failure information is generated internally by the system via unsupervised learning. More specifically, Bates defines the source of failure information as human-generated maintenance records in paragraph 33.
Examiner respectfully disagrees; the argument incorrectly asserts that the system rely solely on human CM work orders and therefore does not internally generate machine failure indicators through unsupervised learning. While ¶33 explain that the system receive failure ratification from CM system as labeling information, D1 separately discloses an unsupervised anomaly detection process that is trained on normal sensor data (¶87) and then applied to new sensor during live monitoring (¶91). Paragraph 92 then determine whether the detected anomaly is valid predictor of a fault and associates related metadata with the resulting failure signature information. Thus, D1 internally detects and determines fault related information from new sensor data through its unsupervised process; this disclosure is not limited to the externally supplied CM records (¶33).
Applicant argue that the Response to Arguments equates Bates' anomaly alert with the claimed machine failure indicator. However, Bates expressly distinguishes between the detection of an anomaly and the subsequent determination that the anomaly is a valid predictor of a fault. Thus, Bates does not disclose that every anomaly alert constitutes a machine failure indicator as claimed. Second, the Response to Arguments conflates the creation of a supervised learning agent and the application of metadata to that agent with the claimed tagging of a machine failure indicator. Even assuming Bates discloses metadata associated with a supervised learning agent, the Office Action does not identify where Bates discloses tagging the machine failure indicator itself, as expressly required by the claim. In other words, the Office Action is pointing to tagging the wrong thing with respect to what is called for in the claim and furthermore, associating metadata with an agent is not what is tagging as the word is clearly used in the claim and especially in view of the specification. Lastly, even assuming, arguendo, which Applicants do not admit, that Bates discloses an unsupervised anomaly detection process and creation of a supervised learning agent, the Office Action does not identify where Bates discloses updating a supervised machine failure prediction process with a tagged machine failure indicator. The cited portions of Bates describe anomaly detection and creation of supervised agents, but do not disclose the claimed use of a tagged machine failure indicator as the input to the update operation.
Examiner respectfully disagrees;
First, applicant improperly read the claim as requiring every anomaly alert to constitute a machine failure indicator. The claim only requires detecting at least one machine failure indicator. D1 expressly teaches that an anomaly may subsequently be “determined to be a valid predictor of a fault” which satisfy the claimed limitation.
Second, applicant reads additional limitations into “tagging”. The claim does not require a particular tagging format, separate data object, or storage location. D1 expressly associates the determined fault related condition with metadata concerning the fault and remedy, thereby providing identifying information for the determined fault related condition.
Third, applicant further narrow “updating a supervised machine failure prediction process” to require the tagged indicator to be supplied as a model input. The claim recite updating a supervised prediction process, not any specific model training operation or update. D1 uses the determined fault related anomaly to create a supervised failure signature agent that learn the new signature and further teach retraining failure agents as new failure occurs.
Applicant argue that the disclosure of Bates does not support that these disclosures of Bates are equivalent to the claim elements for which they are cited. As such, it can be seen that the Office Action is actually relying on an improper reconstruction of the claim from disparate disclosures that do not actually match the claim elements.
Examiner respectfully disagrees;
applicant improper reconstruction of the claim is unpersuasive because D1 itself links the relevant steps. ¶92 expressly connect anomaly detection, determination that the detected anomaly is valid failure predictor, creation of a supervised failure signature agent, and the fault/remedy metadata. Therefore, the mapping does not rely on unrelated disclosure to combine.
Applicant argue that Bates selects or identify sensor tags and only then analyze
probabilities. By contrast, the claims require selecting indicative features based on a
distribution showing association toward failure. These are not the same thing and the
former does not anticipate the latter.
Examiner respectfully disagrees;
applicant reads the “based on at least a distribution” limitation narrowly. D1 determines normal operation distributions from sensor data using SOM/BIC and Gaussian probability (¶¶87-89), while D1 teaches that the sensor data are used to learn failure signature (¶40) and identifies the tags contributing most to the resulting anomaly (¶85). Thus, the indicative features are identified based at least on distribution based behavior indicating an association toward the failure; the claim merely require that the distribution “indicates at least an association” between data features and machine failure and does not require the distribution to be the only selection criteria or require a particular degree of association. D1 distribution based identification meets he broadly recited relationship.
As to the remaining dependent claims, applicant argue that they are allowable due to their respective direct and indirect dependencies upon one of the aforementioned Independent claims. The examiner respectfully disagrees; Independent claims were not allowable as stated in the paragraph above in this “Response to Arguments” section in this office action.
Conclusion
The prior art made of record and not relied upon is considered pertinent to the applicant’s disclosure.
US Patent Application Publication No. 20220301903 filed by Liao et al. that disclose the ability to use sensor historical data to predict machine failure See at least Abstract
Examiner has pointed out particular references contained in the prior arts of record in the body of this action for the convenience of the applicant. Although the specified citations are representative of the teachings in the art and are applied to the specific limitations within the individual claim, other passages and Figures may apply as well. It is respectfully requested from the applicant, in preparing the response, to consider fully the entire references as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior arts or disclosed by the examiner. It is noted that any citation to specific pages, columns, figures, or lines in the prior art references any interpretation of the references should not be considered to be limiting in any way. A reference is relevant for all it contains and may be relied upon for all that it would have reasonably suggested to one having ordinary skill in the art. In re Heck, 699 F.2d 1331-33, 216 USPQ 1038-39 (Fed. Cir. 1983) (quoting In re Lemelson, 397 F.2d 1006, 1009, 158 USPQ 275, 277 (CCPA 1968)).
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MOHAMED ABOU EL SEOUD whose telephone number is (303)297-4285. The examiner can normally be reached Monday-Thursday 9:00am-6:00pm MT.
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, Michelle Bechtold can be reached at (571) 431-0762. 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.
/MOHAMED ABOU EL SEOUD/Primary Examiner, Art Unit 2148
Prosecution Insights