DETAILED ACTION
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claims 1-4 and 7-14 are presented for examination based on the amended claims in the application filed on June 23, 2026. Claims 5-6 have been cancelled by the applicant.
Claims 1-4 and 7-14 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.
Claims 1-4 and 7-14 are rejected under 35 U.S.C. § 112(b) or 35 U.S.C. § 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. § 112, the applicant), regards as the invention.
Claims 1-4 and 7-14 rejected under 35 U.S.C. § 103 as being unpatentable over WO 2019/028269 A2 Cella et al. [herein “Cella”] in view of US 2018/0136617 A1, Xu et al. [herein, “Xu”], and in further view of Jin, Chao. "A sequential process monitoring approach using hidden Markov model for unobservable process drift." Master's thesis, University of Cincinnati, 2015 [herein “Jin”].
This action is made Final.
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 .
Response to Amendment
The amendment filed June 23, 2026 has been entered. Claims 1-4 and 7-14 remain pending in the application. Applicant’s amendments to the Claims have overcome each and every objection and 112(b) rejections previously set forth in the Non-Final Office Action mailed March 18, 2026, with the exception of the objection as included below.
Claim Objections
Claims 1-4 and 7-14 are objected to because of the following informality: recitations of elements with no previous recitations. For example, claim 1, “the time pre-adaptive learning is initiated” on Pg. 3 Ln. 2-3, is improper because there has been no previous recitation of “the time pre-adaptive learning is initiated”. For the purpose of examination, “the time pre-adaptive learning is initiated” will be interpreted as “a time pre-adaptive learning is initiated”. Claims 11 and 14, having similar limitations of claim 1, are also objected. Claims 2-4 and 7-10 and Claims 12-13 are also objected to for incorporating the deficiency of its dependent claims 1 and 11, respectively. Similarly, the following are objected under similar rationale:
Claim 1, “the plant” on Pg. 3 Ln. 27, should be “the industrial manufacturing plant”. Also see Pg. 4 Ln. 6 and Pg. 5 Ln. 3.
Claim 1, “through automatically selecting database from models database based on the industrial manufacturing plant and classify state or health of the plant” on Pg. 4 Ln. 5-6 should be “through automatically selecting database from a models database based on the industrial manufacturing plant and classify a state or health of the industrial manufacturing plant”.
Claim 1, “determining whether a percentage of the variables that are out of the ranges is above a certain threshold, wherein the variables in the first set of data that are out of the ranges of all the variables for the second set of data” on Pg. 6 Ln. 6-8 should be “determining whether a percentage of the variables in the first set of data that are out of the ranges of all the variables for the second set of data is above a certain threshold”. Further the next limitation with includes the phrase “in response to determining that the percentage of the variables are out of the ranges is above a threshold” should incorporated this change to be “in response to determining that the percentage of the variables in the first set of data are out of the ranges of all the variables for the second set of data is above the certain threshold”.
Claim 1, “in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold” on Pg. 6 Ln. 14-16 should be “and in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold”.
Claims 11 and 14, having similar limitations of claim 1, are also objected.
All claims dependent on an objected base claim are objected based on their dependency
Appropriate correction is required.
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 1-4 and 7-14 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 applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
Regarding claim 1, the original disclosure filed November 16, 2021 lacks support for the following limitation “wherein each model being mapped to a corresponding physical subsystem of the industrial manufacturing plant” on Pg. 3 Ln. 24-25. While the specification hints at receiving data from subsystems of the plant (see Para. 034, “Operations data is recorded by sensors in the plant and includes temperatures, pressures, flow rates and vibrations from processes and equipment in the units of the plant”), the specification lacks support for have a single model for each system of the plant.
Regarding claim 1, the original disclosure filed November 16, 2021 lacks support for the following limitation “automatically generating, by the MOAD, one or more control signals based on the selected at least one active model, wherein the control signals are transmitted to the industrial manufacturing plant to dynamically adjust operating parameters of industrial equipment including at least one of flow rate, temperature, pressure, or actuator state, reducing deviation in key performance parameters and improving operational efficiency of the industrial manufacturing plant” on Pg. 4 Ln. 1-8. While the specification discloses receiving parameters of flow rate, temperature, pressure (see Para. 034, “Operations data is recorded by sensors in the plant and includes temperatures, pressures, flow rates and vibrations from processes and equipment in the units of the plant”), the specification lacks support for generating and sending control signals to dynamically control these parameters, dynamically control an actuator state, reduce the key performance parameters, and improve efficiency of the plant.
Regarding claim 1, the original disclosure filed November 16, 2021 lacks support for the following limitation “automatically deploying and executing, via the one or more hardware processors, at least one model upon activating, in the industrial manufacturing plant based on the adaptive learning of the plurality of models” on Pg. 5 Ln. 17-18. While the specification discloses activating a model prior to pre-adaptive and adaptive learning (see Para. 049, “The first set of data is the data on which the active model was trained, and the second set of data is the data accumulated due to plant operation from the time of activation of the active model until the time pre-adaptive learning is initiated”) and discloses recommending the active model to be used in the plant (Para. 092, “at the last step (1222), recommending at least one model for activation in the industrial manufacturing plant based on adaptive learning of the plurality of models”), the specification lacks support for activating a model in the plant based on the adaptive learning of the model.
Regarding claim 4, the original disclosure filed November 16, 2021 lacks support for the following limitation “wherein the pre-processing is performed subject to plant-specific data acquisition constraints”. While the specification discloses receiving plant-specific data (see Para. 034, “Operations data is recorded by sensors in the plant and includes temperatures, pressures, flow rates and vibrations from processes and equipment in the units of the plant”), the specification lacks support for any specific constraints on this data during the pre-processing step.
Claims 11 and 14, having similar limitations of claim 1, are also rejected under similar rationale. All claims dependent on an rejected base claim are rejected based on their dependency.
Claim Rejections - 35 U.S.C. § 112
The following is a quotation of 35 U.S.C. § 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. § 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-4 and 7-14 are rejected under 35 U.S.C. § 112(b) or 35 U.S.C. § 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. § 112, the applicant), regards as the invention.
Claim 1 recites the terms “substantially unchanged” and “changed substantially”, which is a relative term that renders the claim indefinite. The term “substantially” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention (See MPEP § 2173.05(b)). Claims 11 and 14, having similar limitations of claim 1, are also rejected under similar rationale. All claims dependent on an rejected base claim are rejected based on their dependency.
Claim Rejections - 35 U.S.C. § 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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. § 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. § 102(b)(2)(C) for any potential 35 U.S.C. § 102(a)(2) prior art against the later invention.
Claims 1-4 and 7-14 rejected under 35 U.S.C. § 103 as being unpatentable over WO 2019/028269 A2 Cella et al. [herein “Cella”] in view of US 2018/0136617 A1, Xu et al. [herein, “Xu”], and in further view of Jin, Chao. "A sequential process monitoring approach using hidden Markov model for unobservable process drift." Master's thesis, University of Cincinnati, 2015 [herein “Jin”].
As per claim 1, Cella teaches “A processor-implemented method comprising steps of receiving, receiving, by a model-based optimization and advisory device (MOAD) via an input/output interface, a plurality of data from one or more databases of an industrial manufacturing plant at a pre-determined frequency, wherein the plurality of data comprises real-time data and non-real-time data”. (Para. 9, “Methods and systems are provided herein for data collection in industrial environments” [method of receiving data in an industrial manufacturing plant]. Para. 13, “Throughout the present disclosure, wherever a crosspoint switch, multiplexer (MUX) device, or other multiple-input multiple-output data collection or communication device is described, any multi-sensor acquisition device is also contemplated” [receiving via an input/out interface]. Para. 15, “In embodiments, the local data collection system includes multiple multiplexing units and multiple data acquisition units receiving multiple data streams from multiple machines in the industrial environment” [receiving by a model-based optimization and advisory device (MOAD) a plurality of data from one or more databases of an industrial manufacturing plant]. “In embodiments, the local data collection system includes distributed complex programmable hardware device (‘CPLD’) chips” [processor-implemented method] “each dedicated to a data bus for logic control of the multiple multiplexing units and the multiple data acquisition units that receive the multiple data streams from the multiple machines in the industrial environment.” Para. 17, “In embodiments, the local data collection system is configured to obtain long blocks of data at a single relatively high-sampling rate” [receiving data at a frequency] “as opposed to multiple sets of data taken at different sampling rates. In embodiments, the single relatively high-sampling rate corresponds to a maximum frequency of about forty kilohertz” [pre-determined frequency]. Para. 253, “Embodiments of the methods and systems disclosed herein may include data acquisition parking features. In embodiments, a data acquisition box used for route collection, real time analysis” [real-time data]. Para. 302, “In many embodiments, the RPM information can be used to mark segments of the raw waveform data over its collection history” [non-real-time data]. “Further embodiments include techniques for collecting instrument data following a prescribed route of a vibration study. The dynamic markers can enable analysis and trending software to utilize multiple segments of the collection interval indicated by the markers (e.g., two minutes) as multiple historical collection ensembles, rather than just one as done in previous systems where route collection systems would historically store data for only one RPM setting.” Further see Para. 13-17, 253, and 302. The examiner has interpreted a system and method through CPLD chips for data collection by a multi-sensor acquisition device that receives multiple data streams from multiple machines in the industrial environment to obtain long blocks of data at a single relatively high-sampling rate of 44kHz for real time analysis and collection of RPM information that can be used to mark segments of the raw waveform data over its collection history as a processor-implemented method comprising steps of receiving, receiving, by a model-based optimization and advisory device (MOAD) via an input/output interface, a plurality of data from one or more databases of an industrial manufacturing plant at a pre-determined frequency, wherein the plurality of data comprises real-time data and non-real-time data).
Cella also teaches “pre-processing, via one or more hardware processors, the received plurality of data for identification and removal of outliers, imputation of missing data, and synchronization and integration of a plurality of variables from one or more databases”. (Para. 524, “In embodiments, in response to a faulty sensor or sensor data missing from a smart band template data collection activity, one or more alternate sensors may be temporarily included in the set of sensors so as to provide data that may effectively substitute for the missing data in data processing algorithms” [imputation of missing data]. Para. 1000, “Based on continuous or periodic analysis of sensor data, as patterns/trends are identified, or outliers appear, or a group of sensor readings begin to change, etc., the expert system may modify its data collection bands intelligently” [identification and removal of outliers]. Para. 1079, “a greater number of samples from the high resolution data set may be utilized in a subsequent sensor fusion operation in response to confidence that improvements are present, narrowing of the potential result effective variables,” [integration of a plurality of variables from databases] “and/or a determination that higher resolution data is required to determine the result effective parameters.” Para. 276, “the sampling rates used during the vibration survey can be digitally synchronized to predetermined operational frequencies that can relate to pertinent parameters of the machine such as rotating or oscillating speed” [synchronization of data with variables]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via one or more hardware processors]. Further see Para. 276, 520-524, and 1079. The examiner has interpreted that including a set of sensors to provide data that may effectively substitute for the missing data, modifying its data collection bands when patterns/trends are identified or outliers appear, narrowing of the potential result effective variables for a greater number of samples, and digitally synchronizing the sample rates to predetermined operational frequencies that can relate to pertinent parameters by use of a microprocessor as pre-processing, via one or more hardware processors, the received plurality of data for identification and removal of outliers, imputation of missing data, and synchronization and integration of a plurality of variables from one or more databases).
Cella also teaches “obtaining, via the one or more hardware processors, simulated data based on the pre-processed data and using at least one soft sensor, wherein the at least one soft-sensor comprises a physics-based soft sensor and a data-driven soft sensor, wherein the simulated data is integrated with preprocessed data to obtain integrated data”. (Para. 327, “In embodiments, the local cognitive input selection system 4004 may organize fusion of data” [integrated data] “for various onboard sensors, external sensors (such as in the local environment) and other input sources 116 to the local collection system 102 into one or more fused data streams, such as using the multiplexer 4002 to create various signals that represent combinations, permutations, mixes, layers, abstractions, data-metadata combinations, and the like” [obtaining simulated data using soft sensors] “of the source analog and/or digital data that is handled by the data collection system 102.” Para. 344, “Storage may be optimized by configuring what data types are used (e.g., byte-like structures, structures representing fused data from multiple sensors, structures representing statistics or measures calculated by applying mathematical functions on data, and the like)” [combing simulated data with preprocessed date to obtain integrated data]. Para. 328, “In embodiments, the analytic system 4018 may apply to any of a wide range of analytic techniques, including statistical and econometric techniques (such as linear regression analysis, use similarity matrices, heat map based techniques, and the like), reasoning techniques (such as Bayesian reasoning, rule-based reasoning, inductive reasoning, and the like), iterative techniques (such as feedback, recursion, feed-forward and other techniques), signal processing techniques (such as Fourier and other transforms), pattern recognition techniques (such as Kalman and other filtering techniques), search techniques, probabilistic techniques (such as random walks, random forest algorithms, and the like), simulation techniques (such as random walks, random forest algorithms, linear optimization and the like), and others. This may include computation of various statistics or measures” [data-driven and physical-based]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 327-329, 344, and 520. The examiner has interpreted that using the multiplexer 4002 to create various signals that represent combinations, permutations, mixes, layers, abstractions, and data-metadata combinations to fuse data from multiple sensors representing statistics or measures calculated by applying mathematical functions on data through statistical and measurable computations by use of a microprocessor as obtaining, via the one or more hardware processors, simulated data based on the pre-processed data and using at least one soft sensor, wherein the at least one soft-sensor comprises a physics-based soft sensor and a data-driven soft sensor, wherein the simulated data is integrated with preprocessed data to obtain integrated data).
Cella also teaches “determining, via the one or more hardware processors, one or more predicted values of each of a plurality of response variables to detect and diagnose process, equipment anomalies, classifying a state of the process of an equipment, and estimating remaining useful life time to time failure using the integrated data and a plurality of models, wherein the plurality of models comprising at least one active model, wherein the plurality of response variables include one or more key performance parameters of the industrial manufacturing plant”. (Para. 650, “Correlation of trends and values for different types of data” [integrated data] “may be analyzed to identify those parameters whose short-term analysis might provide the best prediction regarding expected performance” [predicted values of each of the response variables (i.e., key performance parameters) of the industrial plant using integrated data]. “This information may be transmitted back to the monitoring device to update types of data collected and analyzed locally or to influence the design of future monitoring devices.” Para. 356, “Embodiments include using a cloud-based platform to identify patterns in data across a plurality of data pools that contain data published from industrial sensors. Embodiments include training a model to identify preferred sensor sets to diagnose a condition of an industrial environment” [to detect and diagnose process]. Para. 656, “An example system for data collection, processing, and utilization of signals in an industrial environment includes a plurality of monitoring devices, each monitoring device comprising a data acquisition circuit structured to interpret a plurality of detection values from a plurality of input sensors communicatively coupled to the data acquisition circuit, each of the plurality of detection values corresponding to at least one of the input sensors; a signal evaluation circuit structured to obtain at least one of vibration amplitude, vibration frequency and a vibration phase location corresponding to at least one of the input sensors in response to the corresponding at least one of the plurality of detection values; a data storage facility for storing a subset of the plurality of detection values; a communication circuit structured to communicate at least one selected detection value to a remote server; and a monitoring application on the remote server structured to: receive the at least one selected detection value; jointly analyze a subset of the detection values received from the plurality of monitoring devices; and recommend an action”. Para. 657, “where the subset of detection values is selected based on data associated with a detection value including at least one: common type of component, common type of equipment, and common operating conditions and further selected based on one of anticipated life of a component associated with detection values, type of the equipment associated with detection values, and operational conditions under which detection values were measured; and/or where the analysis of the subset of detection values includes feeding a neural net with the subset of detection values and supplemental information to learn to recognize various operating states, health states, life expectancies and fault states utilizing deep learning techniques, wherein the supplemental information comprises one of component specification, component performance, equipment specification, equipment performance, maintenance records, repair records and an anticipated state model” [classifying a state of the process of an equipment, and estimating remaining useful life time to time failure]. Para. 318, “The platform 100 may, therefore, learn from and make decisions on a set of data, by making data-driven predictions and adapting according to the set of data” [predict values using an active model]. “In embodiments, machine learning may involve performing a plurality of machine learning tasks by machine learning systems, such as supervised learning, unsupervised learning, and reinforcement learning” [using a plurality of models]. Para. 318, “the tasks may also be classified as machine learning problems such as classification, regression, clustering, density estimation, dimensionality reduction, anomaly detection, and the like” [equipment anomalies]. Para. 319, “data streams from vibration, pressure, temperature, accelerometer, magnetic, electrical field, and other analog sensors may be multiplexed or otherwise fused, relayed over a network, and fed into a cloud-based machine learning facility, which may employ one or more models relating to an operating characteristic of an industrial machine, an industrial process, or a component or element thereof” [using a plurality of models]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 318-319, 356, 520, 650, and 656-657. The examiner has interpreted that providing the best prediction regarding the expected performance based on the correlation of trends and values for different types of data as well as using data-driven predictions and adapting to the set of data for performing a plurality of machine learning tasks by machine learning systems, such as supervised learning, unsupervised learning, and reinforcement learning employing one or more models relating to an operating characteristic of an industrial machine to identify preferred sensor sets to diagnose a condition of an industrial environment, to interpret a plurality of detection values from a plurality of input sensors such as operational conditions under which detection values were measured, to learn to recognize various operating states, health states, life expectancies and fault states utilizing deep learning technique, and anomaly detection by use of a microprocessor as determining, via the one or more hardware processors, one or more predicted values of each of a plurality of response variables to detect and diagnose process, equipment anomalies, classifying a state of the process of an equipment, and estimating remaining useful life time to time failure using the integrated data and a plurality of models, wherein the plurality of models comprising at least one active model, wherein the plurality of response variables include one or more key performance parameters of the industrial manufacturing plant).
Cella also teaches “estimating, via the one or more hardware processors, predictions of response variables using a combination of a first set of data and a second set of data, wherein the active model trained on the first set of data, and the second set of data is accumulated due to plant operation from a time of activation of the active model until the time pre-adaptive learning is initiated, wherein the response variables include key process parameters in process plants including productivity, yield, cycle time, energy consumption, waste generation, emission, quality parameters, condition of equipment, availability, mean time between failures, number of unplanned shutdowns, cost of operation, cost of maintenance, or a weighted combination of all to indicate a condition of the industrial manufacturing plant, process and equipment”. (Para. 354, “Embodiments include making a stream of continuous ultrasonic monitoring data from an industrial environment available as a service from a data marketplace. Embodiments include feeding a stream of continuous ultrasonic monitoring data into a self-organizing data pool. Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream,” [wherein the active model trained on the first set of data on the first set of data] “and is improved based on data collected on performance in an industrial environment” [the second set of data is accumulated due to plant operation from a time of activation of the active model until the time pre-adaptive learning is initiated]. Para. 650, “Correlation of trends and values for different types of data may be analyzed to identify those parameters whose short-term analysis might provide the best prediction regarding expected performance” [estimating predictions of response variables (i.e., key performance parameters) using a combination of a first set of data and a second set of data, wherein the response variables include key process parameters in process plants including productivity]. Para. 337, “a platform is provided having training AI models based on industry-specific feedback. In embodiments, the various embodiments of cognitive systems disclosed herein may take inputs and feedback from industry-specific and domain-specific sources 116 (such as relating to optimization of specific machines, devices, components, processes, and the like). Thus, learning and adaptation of storage organization, network usage, combination of sensor and input data, data pooling, data packaging, data pricing, and other features ( such as for a marketplace 4102 or for other purposes of the host processing system 112) may be configured by learning on the domain-specific feedback measures of a given environment or application, such as an application involving IoT devices (such as an industrial environment). This may include optimization of efficiency (such as in electrical, electromechanical, magnetic, physical, thermodynamic, chemical and other processes and systems),” [key process parameters in process plants] “optimization of outputs (such as for production of energy, materials, products, services and other outputs), prediction, avoidance and mitigation of faults (such as in the aforementioned systems and processes), optimization of performance measures (such as returns on investment, yields, profits, margins, revenues and the like)” [yield] “reduction of costs (including labor costs, bandwidth costs, data costs, material input costs, licensing costs, and many others),” [cost of operation, cost of maintenance] “optimization of benefits (such as relating to safety, satisfaction, health), optimization of work flows (such as optimizing time and resource allocation to processes), and others”. Para. 356, “Embodiments include using a cloud-based platform to identify patterns in data across a plurality of data pools that contain data published from industrial sensors. Embodiments include training a model to identify preferred sensor sets to diagnose a condition of an industrial environment” [to indicate a condition of the process]. Para. 656, “An example system for data collection, processing, and utilization of signals in an industrial environment includes a plurality of monitoring devices, each monitoring device comprising a data acquisition circuit structured to interpret a plurality of detection values from a plurality of input sensors communicatively coupled to the data acquisition circuit, each of the plurality of detection values corresponding to at least one of the input sensors; a signal evaluation circuit structured to obtain at least one of vibration amplitude, vibration frequency and a vibration phase location corresponding to at least one of the input sensors in response to the corresponding at least one of the plurality of detection values; a data storage facility for storing a subset of the plurality of detection values; a communication circuit structured to communicate at least one selected detection value to a remote server; and a monitoring application on the remote server structured to: receive the at least one selected detection value; jointly analyze a subset of the detection values received from the plurality of monitoring devices; and recommend an action”. Para. 657, “where the subset of detection values is selected based on data associated with a detection value including at least one: common type of component, common type of equipment, and common operating conditions and further selected based on one of anticipated life of a component associated with detection values, type of the equipment associated with detection values, and operational conditions under which detection values were measured; and/or where the analysis of the subset of detection values includes feeding a neural net with the subset of detection values and supplemental information to learn to recognize various operating states, health states, life expectancies and fault states utilizing deep learning techniques, wherein the supplemental information comprises one of component specification, component performance, equipment specification, equipment performance, maintenance records, repair records and an anticipated state model” [to indicate a condition of the industrial manufacturing plant and equipment]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 337, 354-356, 520, 650, and 656-657. The examiner has interpreted that providing the best prediction regarding the expected performance based on the correlation of trends and values for different types of data first from a training set and then improved based on data collected on the performance in an industrial environment for measures such as optimization of efficiency, optimization of outputs, optimization of performance measures and yield, and reduction of labor and material cost to diagnose a condition of an industrial environment, to interpret a plurality of detection values from a plurality of input sensors such as operational conditions under which detection values were measured, to learn to recognize various operating states, health states, life expectancies and fault states utilizing deep learning technique, and anomaly detection by use of a microprocessor as estimating, via the one or more hardware processors, predictions of response variables using a combination of a first set of data and a second set of data, wherein the active model trained on the first set of data, and the second set of data is accumulated due to plant operation from a time of activation of the active model until the time pre-adaptive learning is initiated, wherein the response variables include key process parameters in process plants including productivity, yield, cost of operation, or cost of maintenance to indicate a condition of the industrial manufacturing plant, process and equipment.)
Cella also teaches “wherein the pre-adaptive learning comprises of: identifying, via the one or more hardware processors, a subset of each of the plurality of models based on input raw materials, condition of the process, health of the equipment and environmental conditions”. (Para. 319, “embodiments, methods and systems are disclosed herein for cloud-based, machine pattern recognition based on fusion of remote, analog industrial sensors. For example, data streams from vibration, pressure, temperature, accelerometer, magnetic, electrical field, and other analog sensors may be multiplexed or otherwise fused, relayed over a network, and fed into a cloud-based machine learning facility, which may employ one or more models relating to an operating characteristic of an industrial machine, an industrial process, or a component or element thereof)” [wherein the pre-adaptive learning comprises of: identifying a subset of each of the plurality of models based on environmental conditions]. Para. 311, “platform 100 may include the local data collection system 102 deployed in the environment 104 to monitor signals from sensors that may provide signals for fault detection including excessive vibration, incorrect material, incorrect material properties,” [input raw materials]. Para. 321, “The cognitive input selection systems 4004, 4014 may use intelligence and machine learning capabilities described elsewhere in this disclosure, such as using detected conditions (such as conditions informed by the input sources 116 or sensors), state information (including state information determined by a machine state recognition system 4020 that may determine a state), such as relating to an operational state, an environmental state” [environmental conditions], “a state within a known process or workflow” [condition of the process], “a state involving a fault or diagnostic condition” [health of the equipment]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 311, 319, 321, and 520. The examiner has interpreted that that by use of a microprocessor and machine pattern recognition from data stream of different engineering areas to employ a model relating to the operating characteristics of the industrial machine, an industrial process, or a component and detection of material properties, a state within a known process or workflow, state involving a fault or diagnostic condition, and environmental state as wherein the pre-adaptive learning comprises of: identifying, via the one or more hardware processors, a subset of each of the plurality of models based on input raw materials, condition of the process, health of the equipment and environmental conditions.)
Cella also teaches “computing, via the one or more hardware processors, a model quality index (MQI) for each subset of the plurality of models, where the MQI is different for each output, then MQI is customized by an user to have at least one threshold; and activating, via the one or more hardware processors, an at least one model with a highest MQI from the computed MQI for each subset of the plurality of models for the execution.” (Para. 211, “Machine learning may be used to improve the foregoing, such as by adjusting one or more weights, structures, rules, or the like (such as changing a function within a model) based on feedback (such as regarding the success of a model in a given situation) or based on iteration (such as in a recursive process).” Para. 211, “For example, the system may learn to recognize faults,” [adjusting weights based on feedback or iteration to recognize faults, i.e. computing and improving the MQI of the model] “to recognize patterns, to develop models or functions, to develop rules, to optimize performance, to minimize failure rates, to optimize profits, to optimize resource utilization, to optimize flow (such as flow of traffic), or to optimize many other parameters that may be relevant to successful outcomes (such as outcomes in a wide range of environments)” [choosing the best model outcome, i.e., activating a model with the highest MQI]. “Machine learning may use genetic programming techniques,” [activating per-adaptive learning] “such as promoting or demoting one or more input sources, structures, data types, objects, weights, nodes, links, or other factors based on feedback (such that successful elements emerge over a series of generations). For example, alternative available sensor inputs for a data collection system 102 may be arranged in alternative configurations and permutations, such that the system may, using generic programming techniques over a series of data collection events,” [for each subset of the models] “determine what permutations provide successful outcomes based on various conditions” [computing MQI for each subset of the plurality of models, where the MQI is different for each output]. Para 319, “The machine learning facility may take feedback, such as one or more inputs or measures of success, such that it may train, or improve, its initial model (such as improvements by adjusting weights, rules, parameters, or the like, based on the feedback).” [e.g., then MQI is customized by an user to have at least one threshold]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [hardware processor]. Further see Para. 211, 319, and 520. The examiner has interpreted that by use of a microprocessor recognizing faults and patterns in the development of models by adjusting weights based on feedback, inputs, and measures of success to improve the model and optimize parameters for successful outcomes, performance, failure rates, resource utilization, and many other parameters that may be relevant to successful outcomes using generic programming techniques over a series of data collection events as computing, via the one or more hardware processors, a model quality index (MQI) for each subset of the plurality of models, where the MQI is different for each output, then MQI is customized by an user to have at least one threshold; and activating, via the one or more hardware processors, an at least one model with a highest MQI from the computed MQI for each subset of the plurality of models for the execution).
Cella also teaches “displaying an output of a service through a user interface of MOAD, by performing predictions, wherein the predictions are obtained using a selected at least one active prediction, detection, classification, diagnosis or prognosis model including one or more data driven models, wherein the data driven models include reduced order models for the plurality of pre-processed real time and non-real time data, wherein each model being mapped to a corresponding physical subsystem of the industrial manufacturing plant, wherein prediction from various models aid a plant operator or an engineer to take informed decisions concerning the operation of the plant to keep check on root cause of possible anomalies”. (Para. 318, “The platform 100 may, therefore, learn from and make decisions on a set of data, by making data-driven predictions and adapting according to the set of data” [by performing predictions, wherein the predictions are obtained using a selected at least one active model and pre-processed data]. Para. 904, “References to a neural net throughout this disclosure should be understood to encompass a wide range of different types of neural networks, machine learning systems, artificial intelligence systems, and the like, such as model-based systems (including ones based on physical models, statistical models, flow-based models, biological models, biomimetic models, and the like)” [active model including one or more data driven models]. Para. 253, “Embodiments of the methods and systems disclosed herein may include data acquisition parking features. In embodiments, a data acquisition box used for route collection, real time analysis” [for the plurality of pre-processed real-time data]. Para. 302, “In many embodiments, the RPM information can be used to mark segments of the raw waveform data over its collection history” [non-real-time data]. “Further embodiments include techniques for collecting instrument data following a prescribed route of a vibration study. The dynamic markers can enable analysis and trending software to utilize multiple segments of the collection interval indicated by the markers (e.g., two minutes) as multiple historical collection ensembles, rather than just one as done in previous systems where route collection systems would historically store data for only one RPM setting.” Para. 926, “Therefore, the auto encoders may operate as an unsupervised learning model. An auto encoder may be used, for example, for unsupervised learning of efficient codings, such as for dimensionality reduction, for learning generative models of data, and the like. In embodiments, an auto-encoding neural network may be used to self-learn an efficient network coding for transmission of analog sensor data from an industrial machine over one or more networks” [wherein the data driven models include reduced order models]. Para. 656, “An example system for data collection, processing, and utilization of signals in an industrial environment includes a plurality of monitoring devices, each monitoring device comprising a data acquisition circuit structured to interpret a plurality of detection values from a plurality of input sensors communicatively coupled to the data acquisition circuit, each of the plurality of detection values corresponding to at least one of the input sensors; a signal evaluation circuit structured to obtain at least one of vibration amplitude, vibration frequency and a vibration phase location corresponding to at least one of the input sensors in response to the corresponding at least one of the plurality of detection values; a data storage facility for storing a subset of the plurality of detection values; a communication circuit structured to communicate at least one selected detection value to a remote server; and a monitoring application on the remote server structured to: receive the at least one selected detection value; jointly analyze a subset of the detection values received from the plurality of monitoring devices; and recommend an action”. Para. 657, “where the subset of detection values is selected based on data associated with a detection value including at least one: common type of component, common type of equipment, and common operating conditions and further selected based on one of anticipated life of a component associated with detection values, type of the equipment associated with detection values, and operational conditions under which detection values were measured; and/or where the analysis of the subset of detection values includes feeding a neural net with the subset of detection values and supplemental information to learn to recognize various operating states, health states, life expectancies and fault states utilizing deep learning techniques, wherein the supplemental information comprises one of component specification, component performance, equipment specification, equipment performance, maintenance records, repair records and an anticipated state model” [to take informed decisions concerning the operation of the plant and classify the state or health of the plant]. Para. 314, “the platform 100 may include the local data collection system 102 deployed in the environment 104 to monitor signals from additional large machines such as turbines, windmills, industrial vehicles, robots, and the like. These large mechanical machines include multiple components and elements providing multiple subsystems on each machine” and Para. 318, “The platform 100 may, therefore, learn from and make decisions on a set of data, by making data-driven predictions and adapting according to the set of data” [predict values using an active model]. “In embodiments, machine learning may involve performing a plurality of machine learning tasks by machine learning systems, such as supervised learning, unsupervised learning, and reinforcement learning” [wherein prediction from various models, e.g., wherein each model being mapped to a corresponding physical subsystem of the industrial manufacturing plant]. Para. 318, “the tasks may also be classified as machine learning problems such as classification, regression, clustering, density estimation, dimensionality reduction, anomaly detection, and the like” [possible anomalies]. Para. 522, “Each set, which may be a logical set of sensors, may be selected to provide information about elements in an industrial environment that may provide insight into potential problems, root causes of problems, and the like. Each set may be associated with a condition that may be monitored for compliance with an acceptable range of values. The set of sensors may be based on a machine architecture, hierarchy of components, or a hierarchy of data that contributes to a finding about a machine that may usefully be applied to maintaining or improving performance in the industrial environment.” [e.g., displaying an output of a service to aid a plant operator or an engineer to take informed decisions concerning the operation of the plant to keep check on root cause of possible anomalies]. Para. 1166, “a system for data collection in an industrial environment may include an expert system graphical user interface in which a user may, by interacting with a graphical user interface element, select a component of an industrial machine displayed in the graphical user interface for data collection, view a set of sensors that are available to provide data about the industrial machine, and select a subset of sensors for data collection” [through a user interface of MOAD]. Para. 319, “embodiments, methods and systems are disclosed herein for cloud-based, machine pattern recognition based on fusion of remote, analog industrial sensors. For example, data streams from vibration, pressure, temperature, accelerometer, magnetic, electrical field, and other analog sensors may be multiplexed or otherwise fused, relayed over a network, and fed into a cloud-based machine learning facility, which may employ one or more models relating to an operating characteristic of an industrial machine, an industrial process, or a component or element thereof)” [e.g., through automatically selected database from the models database based on the plant]. Further see Para. 253, 318-319, 522, 656-657, 904, 926 and 1166. The examiner has interpreted that a neural network consisting of model-based systems such as a physical model to make date driving prediction and adapting according to the set of data for real-time analysis and data over its collection history through the use of an auto encoder in unsupervised learning for dimensionality reduction of learning generative models, and analyze detection data values to anticipate life of a components and equipment, recognize various operating, health and life states to recommend an action into and provide information about potential problems, root causes of problems, and anomaly detection for the monitoring of compliance with an acceptable range of values through an expert system graphical user interface and the employing of models relating to the operating characteristic of an industrial machine, an industrial process, or a component of the multiple components and elements providing multiple subsystems on each machine as displaying an output of a service through a user interface of MOAD, by performing predictions, wherein the predictions are obtained using a selected at least one active prediction, detection, classification, diagnosis or prognosis model including one or more data driven models, wherein the data driven models include reduced order models for the plurality of pre-processed real time and non-real time data, wherein prediction from various models aid a plant operator or an engineer to take informed decisions concerning the operation of the plant to keep check on root cause of possible anomalies.)
Cella teaches “automatically generating, by the MOAD, one or more control signals based on the selected at least one active model, wherein the control signals are transmitted to the industrial manufacturing plant to dynamically adjust operating parameters of industrial equipment including at least one of flow rate, temperature, pressure, or actuator state, through automatically selected database from models database based on the industrial manufacturing plant and classify state or health of the plant, reducing deviation in key performance parameters and improving operational efficiency of the industrial manufacturing plant”. (Para. 319 “The model may be refined by feeding in data from sensors disposed in the environment of a machine, in the machine, and the like, as well as data indicating actual fuel consumption, so that the machine can provide increasingly accurate, sensor-based, estimates of fuel consumption and can also provide output that indicate what changes can be made to increase fuel consumption (such as changing operation parameters of the machine or changing other elements of the environment, such as the ambient temperature, the operation of a nearby machine, or the like)” [i.e. automatically generating, by the MOAD, one or more control signals based on the selected at least one active model, wherein the control signals are transmitted to the industrial manufacturing plant to dynamically adjust operating parameters of industrial equipment including at least one of flow rate]. Para. 319, “embodiments, methods and systems are disclosed herein for cloud-based, machine pattern recognition based on fusion of remote, analog industrial sensors. For example, data streams from vibration, pressure, temperature, accelerometer, magnetic, electrical field, and other analog sensors may be multiplexed or otherwise fused, relayed over a network, and fed into a cloud-based machine learning facility, which may employ one or more models relating to an operating characteristic of an industrial machine, an industrial process, or a component or element thereof)” [e.g., through automatically selected database from the models database based on the industrial manufacturing plant]. Para. 656, “An example system for data collection, processing, and utilization of signals in an industrial environment includes a plurality of monitoring devices, each monitoring device comprising a data acquisition circuit structured to interpret a plurality of detection values from a plurality of input sensors communicatively coupled to the data acquisition circuit, each of the plurality of detection values corresponding to at least one of the input sensors; a signal evaluation circuit structured to obtain at least one of vibration amplitude, vibration frequency and a vibration phase location corresponding to at least one of the input sensors in response to the corresponding at least one of the plurality of detection values; a data storage facility for storing a subset of the plurality of detection values; a communication circuit structured to communicate at least one selected detection value to a remote server; and a monitoring application on the remote server structured to: receive the at least one selected detection value; jointly analyze a subset of the detection values received from the plurality of monitoring devices; and recommend an action”. Para. 657, “where the subset of detection values is selected based on data associated with a detection value including at least one: common type of component, common type of equipment, and common operating conditions and further selected based on one of anticipated life of a component associated with detection values, type of the equipment associated with detection values, and operational conditions under which detection values were measured; and/or where the analysis of the subset of detection values includes feeding a neural net with the subset of detection values and supplemental information to learn to recognize various operating states, health states, life expectancies and fault states utilizing deep learning techniques, wherein the supplemental information comprises one of component specification, component performance, equipment specification, equipment performance, maintenance records, repair records and an anticipated state model” [classify the state or health of the plant, reducing deviation in key performance parameters and improving operational efficiency of the industrial manufacturing plant]. Further see Para. 319 655-657. The examiner has interpreted that having the model output changes can be made to increase fuel consumption to change the ambient temperature and operation of a machine through the employing of models relating to the operating characteristic of an industrial machine, an industrial process, or a component of the multiple components and analyze detection data values to anticipate life of a components and equipment, recognize various operating, health and life states to recommend an action into and provide information about potential problems, root causes of problems, and anomaly detection for the monitoring of compliance with an acceptable range of values as automatically generating, by the MOAD, one or more control signals based on the selected at least one active model, wherein the control signals are transmitted to the industrial manufacturing plant to dynamically adjust operating parameters of industrial equipment including at least one of flow rate, temperature, pressure, or actuator state, through automatically selected database from models database based on the industrial manufacturing plant and classify state or health of the plant, reducing deviation in key performance parameters and improving operational efficiency of the industrial manufacturing plant.)
Cella also teaches “computing, via the one or more hardware processors, the MQI for each of the plurality of models by comparing the determined one or more predicted values and one or more actual values of each of the plurality of response variables”. (Para. 519, “While a sensed value of a condition may be sufficient to trigger a smart bands data collection template activity, data may need to be collected and processed over time from a plurality of sensors to generate a data value that may be compared to a set of data collection band parameters for conditionally triggering the data collection activity” [computing the MQI for each of the plurality of models by comparing the determined one or more predicted values and one or more actual values of each of the plurality of response variables]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 519-520. The examiner has interpreted that generating a data value from the processed data and comparing it to the set of data collection band parameters by use of a microprocessor as computing, via the one or more hardware processors, the MQI for each of the plurality of models by comparing the determined one or more predicted values and one or more actual values of each of the plurality of response variables).
Cella also teaches “determining, via the one or more hardware processors, a drift in performance of each of the plurality of models based on one or more predefined thresholds of MQI, wherein the computed MQI of each of the plurality of models is compared with the predefined thresholds of MQI for each of the plurality of models”. (Para. 321, “Thus, an automatically adapting, multi-sensor data collection system is provided, where cognitive input selection is used (with feedback) to improve the effectiveness, efficiency, or other performance parameters of the data collection system within its particular environment. Performance parameters may relate to overall system metrics (such as financial yields, process optimization results, energy production or usage, and the like), analytic metrics (such as success in recognizing patterns, making predictions, classifying data, or the like), and local system metrics (such as bandwidth utilization, storage utilization, power consumption, and the like)” [improving the effectiveness of the model by relating the performance parameters of the model to system metrics, i.e., comparing computed MQI to predefined threshold of MQI]. Para. 522, “Each set, which may be a logical set of sensors, may be selected to provide information about elements in an industrial environment that may provide insight into potential problems, root causes of problems, and the like” [determining a drift for each model]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 321 and 520-522. The examiner has interpreted improving the effectiveness, efficiency, or other performance parameters of the data collection system within its particular environment which relate to system and analytical metrics to provide insight into potential problems by use of a microprocessor as determining, via the one or more hardware processors, a drift in performance of each of the plurality of models based on one or more predefined thresholds of MQI, wherein the computed MQI of each of the plurality of models is compared with the predefined thresholds of MQI for each of the plurality of models).
Cella also teaches “identifying, via the one or more hardware processors, at least one cause of the determined drift in the performance of the plurality of models using one or more key performance parameters of the industrial manufacturing plant”. (Para. 321, “Thus, an automatically adapting, multi-sensor data collection system is provided, where cognitive input selection is used (with feedback) to improve the effectiveness, efficiency, or other performance parameters of the data collection system within its particular environment. Performance parameters may relate to overall system metrics (such as financial yields, process optimization results, energy production or usage, and the like), analytic metrics (such as success in recognizing patterns, making predictions, classifying data, or the like), and local system metrics (such as bandwidth utilization, storage utilization, power consumption, and the like).” Para. 318, “The platform 100 may, therefore, learn from and make decisions on a set of data, by making data-driven predictions and adapting according to the set of data.” [identifying a cause from the performance drift of the models]. “In embodiments, machine learning may involve performing a plurality of machine learning tasks by machine learning systems, such as supervised learning, unsupervised learning, and reinforcement learning.” Para. 522, “Each set, which may be a logical set of sensors, may be selected to provide information about elements in an industrial environment that may provide insight into potential problems, root causes of problems, and the like” [identifying drift for each model]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 318-321 and 520-522. The examiner has interpreted that collecting information about operation, quality, and performance to provide insight into the root cause of problems of the model by use of a microprocessor in determining, via the one or more hardware processors, a root cause of potential problem by use of a microprocessor as identifying at least one cause of the determined drift in the performance of the plurality of models using one or more key performance parameters of the industrial manufacturing plant).
Cella also teaches “selecting, via the one or more hardware processors, the first set of data and the second set of data out of the plurality of data of the industrial manufacturing plant, wherein the first set of data is used for training of the plurality of models and the second set of data is stored since activation of the plurality of models”. (Para. 354, “Embodiments include making a stream of continuous ultrasonic monitoring data from an industrial environment available as a service from a data marketplace. Embodiments include feeding a stream of continuous ultrasonic monitoring data into a self-organizing data pool” [selecting data]. “Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream,” [first set of data] “and is improved based on data collected on performance in an industrial environment” [second set of data]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further see Para. 354 and 520-522. The examiner has interpreted feeding a stream of continuous ultrasonic monitoring data where the model is based on a training set created from human analysis of such a data stream and is improved based on data collected on performance in an industrial environment by use of a microprocessor as selecting, via the one or more hardware processors, the first set of data and the second set of data out of the plurality of data of the industrial manufacturing plant, wherein the first set of data is used for training of the plurality of models and the second set of data is stored since activation of the plurality of models).
Cella also teaches “activating, via the one or more hardware processors, a pre-adaptive learning to compute MQI for each subset of the plurality of models based on the selected first set of data and second set of data and the identified cause of the drift in the performance of the plurality of models”. (Para. 211, “Machine learning may be used to improve the foregoing, such as by adjusting one or more weights, structures, rules, or the like (such as changing a function within a model) based on feedback (such as regarding the success of a model in a given situation) or based on iteration (such as in a recursive process).” Para. 211, “For example, the system may learn to recognize faults,” [adjusting weights based on feedback or iteration to recognize faults, i.e. computing and improving the MQI of the model based on the performance] “to recognize patterns, to develop models or functions, to develop rules, to optimize performance, to minimize failure rates, to optimize profits, to optimize resource utilization, to optimize flow (such as flow of traffic), or to optimize many other parameters that may be relevant to successful outcomes (such as outcomes in a wide range of environments). Machine learning may use genetic programming techniques,” [activating per-adaptive learning] “such as promoting or demoting one or more input sources, structures, data types, objects, weights, nodes, links, or other factors based on feedback (such that successful elements emerge over a series of generations)” [based on the first and second set of data]. “For example, alternative available sensor inputs for a data collection system 102 may be arranged in alternative configurations and permutations, such that the system may, using generic programming techniques over a series of data collection events,” [for each subset of model based on the first and second set of data] “determine what permutations provide successful outcomes based on various conditions” [i.e., based on identified causes of performance drift]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [via the one or more hardware processors]. Further Para. 211 and 520-522. The examiner has interpreted that using a machine learning programming techniques that are arranged in alternative configurations over a series of data collection events and successful outcomes based on various conditions to recognize faults by adjusting wights based on feedback by use of a microprocessor as activating, via the one or more hardware processors, a pre-adaptive learning to compute MQI for each of subset of the plurality of models based on the selected the first set of data and the second set of data based on, and the identified cause of the drift in the performance of the plurality of models).
Cella teaches “selecting, using a two-stage adaptive learning pipeline, the first set of data corresponding to stable plant operating regions and the second set of data corresponding to drift conditions identified in the plant”. (Para. 354, “Embodiments include making a stream of continuous ultrasonic monitoring data from an industrial environment available as a service from a data marketplace. Embodiments include feeding a stream of continuous ultrasonic monitoring data into a self-organizing data pool. Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream” and Para. 362, “Embodiments include training a model to present the most valuable data in a data marketplace, where training is based on industry-specific measures of success” [selecting the first set of data corresponding to stable plant operating regions, i.e., a first stage of adaptive learning]. Para. 354, “where the model is based on a training set created from human analysis of such a data stream, and is improved based on data collected on performance in an industrial environment” and Para. 522, “Each set, which may be a logical set of sensors, may be selected to provide information about elements in an industrial environment that may provide insight into potential problems, root causes of problems, and the like” [the second set of data corresponding to drift conditions identified in the plant, i.e., a second stage of adaptive learning]. Further see Para. 354-362 and 522. The examiner has interpreted that training a model with industry-specific measures of success and then improved based on data collected on the performance in an industrial environment that provides information on elements to provide insight into potential problems as selecting, using a two-stage adaptive learning pipeline, the first set of data corresponding to stable plant operating regions and the second set of data corresponding to drift conditions identified in the plant.)
Cella also teaches “only triggering, via the one or more hardware processors, an adaptive learning based on the computed MQI of each subset of the plurality of models when the computed MQI is below the one or more predefined thresholds of MQI, wherein the adaptive learning of the plurality of models includes model performance diagnosis, model re-tuning, model re-building, and model-recreating on the selected first set of data and the second set of data, wherein when active models are a physics-based model then the adaptive learning involves steps of data selection and model re-tuning, wherein when the active models are hybrid models, then data-driven components of the plurality of models are subjected to the adaptive learning via a data-driven route and physics-based components of the active models are subjected to the adaptive learning via a physics-based route, wherein after the adaptive learning, both the adaptively learnt physics-based components and data-driven components are placed back together and the hybrid models are tested for performance”. (Para. 211, “Machine learning may use genetic programming techniques such as promoting or demoting one or more input sources, structures, data types, objects, weights, nodes, links, or other factors based on feedback (such that successful elements emerge over a series of generations)” [triggering an adaptive learning, data-driven components of the models are subjected to the adaptive learning via a data-driven route]. Para. 211, “Machine learning may be used to improve the foregoing, such as by adjusting one or more weights,” [based on the computed MQI of each subset of the plurality of models when the computed MQI is below the one or more predefined thresholds of MQI, e.g., model re-tuning] “structures, rules, or the like (such as changing a function within a model)” [model rebuilding] “based on feedback (such as regarding the success of a model in a given situation) or based on iteration (such as in a recursive process)” [only triggering the adaptive learning based on the computed MQI of each subset of the plurality of models when the computed MQI is below the one or more predefined thresholds of MQI]. Para. 211, “For example, the system may learn to recognize faults to recognize patterns, to develop models or functions, to develop rules, to optimize performance,” [model performance diagnosis] “to minimize failure rates, to optimize profits, to optimize resource utilization, to optimize flow (such as flow of traffic), or to optimize many other parameters that may be relevant to successful outcomes (such as outcomes in a wide range of environments)” [model recreating]. Para. 692, “In embodiments, the monitoring application 8776 may feed a neural net with the selected subset to learn to recognize various operating state, health states (e.g., lifetime predictions) and fault states utilizing deep learning techniques” [triggering adaptive learning for each subset of the plurality of models]. “In embodiments, a hybrid of the two techniques (model-based learning and deep learning) may be used.” Para. 354, “Embodiments include making a stream of continuous ultrasonic monitoring data from an industrial environment available as a service from a data marketplace. Embodiments include feeding a stream of continuous ultrasonic monitoring data into a self-organizing data pool” [selecting data]. “Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream,” [creating model based on first set of data] “and is improved based on data collected on performance in an industrial environment” [updating model based on second set of data]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [hardware processor]. Para. 904, “References to a neural net throughout this disclosure should be understood to encompass a wide range of different types of neural networks, machine learning systems, artificial intelligence systems, and the like” “and/or holographic associative memory neural networks, or hybrids or combinations of the foregoing, or combinations with other expert systems, such as rule-based systems, model-based systems (including ones based on physical models, statistical models, flow-based models, biological models, biomimetic models, and the like), model-based systems (including ones based on physical models, statistical models, flow-based models, biological models, biomimetic models, and the like)” [wherein when active models are a physics-based model, data-drive model, and hybrid model]. Para. 321, “a model of fuel consumption by an industrial machine may include physical model parameters that characterize weights, motion, resistance, momentum, inertia, acceleration, and other factors that indicate consumption, and chemical model parameters (such as those that predict energy produced and/or consumed e.g., such as through combustion, through chemical reactions in battery charging and discharging, and the like). The model may be refined by feeding in data from sensors disposed in the environment of a machine, in the machine, and the like, as well as data indicating actual fuel consumption, so that the machine can provide increasingly accurate, sensor-based, estimates of fuel consumption and can also provide output that indicate what changes can be made to increase fuel consumption” [when active models are a physics-based model then the adaptive learning involves steps of data selection and model re-tuning and physics-based components of the active models are subjected to the adaptive learning via a physics-based route]. Para. 339, “Parameters that may be managed, varied, selected and adapted by cognitive, machine learning may include storage parameters (location, type, duration, amount, structure and the like across the swarm 4202), network parameters (such as how the swarm 4202 is organized, such as in mesh, peer-to-peer, ring, serial, hierarchical and other network configurations as well as bandwidth utilization, data routing, network protocol selection, network coding type, and other networking parameters), security parameters (such as settings for various security applications and services), location and positioning parameters (such as routing movement of mobile data collectors 102 to locations, positioning and orienting collectors 102 and the like relative to points of data acquisition, relative to each other, and relative to locations where network availability may be favorable, among others), input selection parameters (such as input selection among sensors, input sources 116 and the like for each collector 102 and for the aggregate collection), data combination parameters (such as those for sensor fusion, input combination, multiplexing mixing, layering, convolution, and other combinations)” [wherein after the adaptive learning, both the adaptively learnt physics-based components and data-driven components are placed back together]. Para. 348, “cognitive haptic system may be provided, where selection of inputs or triggers for haptic feedback, selection of outputs, timing, intensity levels, durations, and other parameters ( or weights applied to them) may be varied in a process of variation, promotion, and selection (such as using genetic programming) with feedback based on real world responses to feedback in actual situations” [and the hybrid models are tested for performance]. Further see Para. 211, 321, 339, 349-354, 520-522, 692, and 904. The examiner has interpreted that using machine learning programming techniques on alternative configurations and permutations when data types are altered based on feedback to adjust weights based on feedback or iterations, structures and rules, and successful outcomes to optimized parameters which are based on the training set created from human analysis of a data stream on physical model parameters that are refined based on the data obtained from the senor and the model is select after a real world response is generated from the feedback weights that were applies and the data collected on performance in an industrial environment by use of a microprocessor as only triggering, via the one or more hardware processors, an adaptive learning based on the computed MQI of each subset of the plurality of models when the computed MQI is below the one or more predefined thresholds of MQI, wherein the adaptive learning of the plurality of models includes model performance diagnosis, model re-tuning, model re-building, and model-recreating on the selected first set of data and the second set of data, wherein when active models are a physics-based model then the adaptive learning involves steps of data selection and model re-tuning, wherein when the active models are hybrid models, then data-driven components of the plurality of models are subjected to the adaptive learning via a data-driven route and physics-based components of the active models are subjected to the adaptive learning via a physics-based route, wherein after the adaptive learning, both the adaptively learnt physics-based components and data-driven components are placed back together and the hybrid models are tested for performance.)
Cella also teaches “and automatically deploying and executing, via the one or more hardware processors, at least one model upon activating, in the industrial manufacturing plant based on the adaptive learning of the plurality of models, wherein the at least one model includes a re-tuned model, a re-built model, and a re-created model, [wherein the model re-building is invoked when the MQI of the re-tuned models are lower than the predefined thresholds of MQL, wherein the model re-creating is invoked when the MQI of the re-built models are lower than the predefined thresholds of MQI or after model diagnosis] wherein in a data-driven model and data-driven components of the hybrid models, a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data.” (Para. 350 “Embodiments include training a model to determine what policies should be deployed in an industrial data collection system. Embodiments include a system for data collection in an industrial environment with a policy engine for deploying policy within the system and, optionally, self-organizing network coding for data transport, wherein in certain embodiments, a policy applies to how data will be presented in a multi-sensory interface, a heat map visual interface, or in an interface that operates with self-organized tuning of the interface layer” [automatically deploying and executing at least one model upon activating, in the industrial manufacturing plant based on the adaptive learning of the plurality of models]. Para 1005, “The expert system may (optionally using a neural net, machine learning system, deep learning system, or the like, which may occur under supervision by one or more supervisors (human or automated)) intelligently manage bands aligned with different goals and assign weights, parameter modifications, or recommendations based on a factor, such as a bias towards one goal or a compromise to allow better alignment with all goals being tracked, for example” [based on the adaptive learning of the models]. Para. 211, “Machine learning may be used to improve the foregoing, such as by adjusting one or more weights,” [model retuning] “structures, rules, or the like (such as changing a function within a model)” [model rebuilding] “based on feedback (such as regarding the success of a model in a given situation) or based on iteration (such as in a recursive process).” Para. 211, “For example, the system may learn to recognize faults to recognize patterns, to develop models or functions, to develop rules, to optimize performance,” [model performance diagnosis] “to minimize failure rates, to optimize profits, to optimize resource utilization, to optimize flow (such as flow of traffic), or to optimize many other parameters that may be relevant to successful outcomes (such as outcomes in a wide range of environments)” [model recreating]. Para. 354, “Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream and is improved based on data collected on performance in an industrial environment” [modifying the model based on the first and second set of data, e.g., a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data]. Para. 339, “Parameters that may be managed, varied, selected and adapted by cognitive, machine learning may include storage parameters (location, type, duration, amount, structure and the like across the swarm 4202), network parameters (such as how the swarm 4202 is organized, such as in mesh, peer-to-peer, ring, serial, hierarchical and other network configurations as well as bandwidth utilization, data routing, network protocol selection, network coding type, and other networking parameters), security parameters (such as settings for various security applications and services), location and positioning parameters (such as routing movement of mobile data collectors 102 to locations, positioning and orienting collectors 102 and the like relative to points of data acquisition, relative to each other, and relative to locations where network availability may be favorable, among others), input selection parameters (such as input selection among sensors, input sources 116 and the like for each collector 102 and for the aggregate collection), data combination parameters (such as those for sensor fusion, input combination, multiplexing mixing, layering, convolution, and other combinations)” [wherein in a data-driven model and data-driven components of the hybrid models]. Para. 520, “a programmable logic component that interfaces with a sensor and processes data from the sensor; by use of a computer processor, such as a microprocessor” [hardware processor]. Further see Para. 211, 339, 350, 520, 929, 1005, and 1179. The examiner had interpreted that training a model to determine what policies should be deployed in an industrial data collection system for data collection in an industrial environment with a policy engine for deploying policy within the system where a model that is adjusted with weights, structures or rules, and parameters to meet certain outcomes and combining input data parameters using sensor and hybrid combinations where the model is based on a training set created from human analysis of a data stream and is improved based on data collected on performance in an industrial environment by use of a microprocessor as automatically deploying and executing, via the one or more hardware processors, at least one model upon activating, in the industrial manufacturing plant based on the adaptive learning of the plurality of models, wherein the at least one model includes a re-tuned model, a re-built model, and a re-created model, a data-driven model and data-driven components of the hybrid models, a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data).
Cella teaches “wherein the model diagnosis performed on data-driven models or the data-driven components of the hybrid models includes: detecting variables crossing ranges of variables in the first set of data for training the active model; computing ranges of all variables for the second set of data; comparing the ranges of all the variables for the second set of data with the ranges of the variables from the first set of data; determining whether a percentage of the variables that are out of the ranges is above a certain threshold, wherein the variables in the first set of data that are out of the ranges of all the variables for the second set of data; performing model re-tuning with data of variables already in the active models, in response to determining that the percentage of the variables are out of the ranges is above a threshold”. (Para. 319, “The machine learning facility may take feedback, such as one or more inputs or measures of success, such that it may train, or improve, its initial model (such as improvements by adjusting weights, rules, parameters, or the like, based on the feedback).” Para. 333, “the cognitive data packaging system 4110 can automatically vary packaging, such as using different combinations, permutations, mixes, and the like, and varying weights applied to given input sources, sensors, data pools and the like, using learning feedback 4012 to promote favorable packages and de-emphasize less favorable packages” [re-tuning]. Para. 339, “Feedback may be based on any of the kinds of feedback described herein, such that over time the swarm may adapt to its current and anticipated situation to achieve a wide range of desired objectives” [comparing ranges of variables]. Para. 1003, “expert system may determine that the system should either keep or modify operational parameters, equipment or a weighting of a neural net or other model given a constraint of operation (e.g., meeting a required endpoint (e.g., delivery date, amount, cost, coordination with another system), operating with a limited resource (e.g., power, fuel, battery), storage (e.g., data storage), bandwidth (e.g., local network, p2p, WAN, internet bandwidth, availability, or input/output capacity), authorization (e.g., role-based)), a warranty limitation, a manufacturer's guideline, a maintenance guideline)…As the expert system iterates and receives the off-nominal data, it may predict that the refinery will not achieve a specified goal and will recommend an action”[wherein the model diagnosis performed on data-driven models or the data-driven components of the hybrid models includes: detecting variables crossing ranges of variables in the first set of data for training the active model; computing ranges of all variables for the second set of data; comparing the ranges of all the variables for the second set of data with the ranges of the variables from the first set of data; determining whether a percentage of the variables that are out of the ranges is above a certain threshold, wherein the variables in the first set of data that are out of the ranges of all the variables for the second set of data]… “the expert system may determine that the system should adjust the weights/biases of a model used by the expert system” [performing model re-tuning with data of variables already in the active models, in response to determining that the percentage of the variables are out of the ranges is above a threshold]. Further see Para. 319, 339, and 1003. The examiner has interpreted that determining that the system should modify operational parameters and adjust weighting of the neural network model and parameters when it no longer meeting a requirement when the off-nominal data received as wherein the model diagnosis performed on data-driven models or the data-driven components of the hybrid models includes: detecting variables crossing ranges of variables in the first set of data for training the active model; computing ranges of all variables for the second set of data; comparing the ranges of all the variables for the second set of data with the ranges of the variables from the first set of data; determining whether a percentage of the variables that are out of the ranges is above a certain threshold, wherein the variables in the first set of data that are out of the ranges of all the variables for the second set of data; performing model re-tuning with data of variables already in the active models, in response to determining that the percentage of the variables are out of the ranges is above a threshold.)
Cell also teaches “computing [T2 metric from] principal component analysis [or Mahalanobis distance (MD)] for the second set of data”. (Para. 318, “The platform 100 may, therefore, learn from and make decisions on a set of data, by making data-driven predictions and adapting according to the set of data… In examples, machine learning may include a plurality of other tasks based on an output of the machine learning system. In examples the tasks may also be classified as machine learning problems such as classification, regression, clustering, density estimation, dimensionality reduction, anomaly detection, and the like”. Para. 1450, “dimensionality reduction algorithm can be one or more of a Decision Tree, Random Forest, Principal Component Analysis” [e.g., computing principal component analysis for the second set of data]. Further see Para. 318, 1003, and 1450. The examiner has interpreted that using a principal component analysis in dimensionality reduction algorithm based on the output of the system and its data predictions and adapting to the data as computing principal component analysis for the second set of data.)
Cella does not specifically teach “wherein the model re-building is invoked when the MQI of the re-tuned models are lower than the predefined thresholds of MQL, wherein the model re-creating is invoked when the MQI of the re-built models are lower than the predefined thresholds of MQI or after model diagnosis, wherein in a data-driven model and data-driven components of the hybrid models, a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data.”
However, in the same field of endeavor namely optimizing the performance of industrial manufacture monitoring models, Xu teaches “wherein the model re-building is invoked when the MQI of the re-tuned models are lower than the predefined thresholds of MQL, wherein the model re-creating is invoked when the MQI of the re-built models are lower than the predefined thresholds of MQI or after model diagnosis, wherein in a data-driven model and data-driven components of the hybrid models, a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data.” (Para. 47, “The predicted performance is evaluated/processed with new monitored data samples received, step 122, from the industrial asset. This stream of monitored data samples can be combined with accurate, observed (i.e., “ground truth”) data, with subsequent filtering to be used by the continuous learning block to update/create models for addition to the model ensemble” [a pre-processed first set of data and second set of data together are used for the model re-tuning]. Para. 48, “If the error difference is determined at step 130 to be greater than the predetermined threshold,” [when the MQI of the re-tuned models are lower than the predefined thresholds of MQL] “a new model is created, step 133. This new model is then added to the model ensemble” [wherein the model re-building is invoked]. Para. 46, “In accordance with embodiments the model ensemble can be a collection of models, where each model implements a different modeling approach” [created models use different learning techniques, i.e., wherein the model re-creating is invoked]. Para. 47, “If the error difference is less than or equal to a predetermined threshold, ERA algorithm returns to the model ensemble, where each individual model is updated 135 and its corresponding weight is adjusted based on its performance 140” [wherein the model re-creating is invoked when the MQI of the re-built models are lower than the predefined thresholds of MQI or after model diagnosis]. Para. 53-54, “the model ensemble can be associated with a variable, named Life, which counts the total number of online evaluations the model has seen so far. Thus, Life is initialized as 0 for each new model. The mean square error (MSE) of the model on the data points that it is evaluated on (with upper threshold≤ws) is denoted as a variable, mse, which is also initially set as 0. The voting strategy of the ensemble is weighted voting, and the weight of the first model is 1. In the online learning block, the ensemble generates the prediction custom-character for a new input point xt, based on weighted voting from all of its components,” [when the weights reach a certain limit, new input values are incorporated in the model ensembles, i.e., models are recreating using new variables]. Fig. 1 shows these steps to be a successive process. Further see Para. 46-48 and 53-54. The examiner has interpreted that updating models with new data combined with the observed data, then creating model that that is added to the ensemble when the error difference is determined at step 130 to be greater than the predetermined threshold, and finally generating models based on new input when the error difference is determined if the error difference is less than the threshold to adjust the weight on it performance of all of its components as wherein the model re-building is invoked when the MQI of the re-tuned models are lower than the predefined thresholds of MQL, wherein the model re-creating is invoked when the MQI of the re-built models are lower than the predefined thresholds of MQI or after model diagnosis, a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “wherein the model re-building is invoked when the MQI of the re-tuned models are lower than the predefined thresholds of MQL, wherein the model re-creating is invoked when the MQI of the re-built models are lower than the predefined thresholds of MQI or after model diagnosis” and “wherein in a data-driven model and data-driven components of the hybrid models, a pre-processed first set of data and second set of data together are used for the model re-tuning and the model re-tuning entails building the models again using new data” as conceptually seen from the teaching of Xu, into that of Cella because this modification of modifying the model in a successive manner for the advantageous purpose of providing an efficient continuous learning approach when modifying models to save on computing resources (Xu, Para. 19 & 22-23). Further motivation to combine be that Cella and Xu are analogous art to the current claim and are directed to optimizing the performance of industrial manufacture monitoring models.
Cella nor Xu specifically teach “computing T2 metric from principal component analysis or Mahalanobis distance (MD) for the second set of data, in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold, an operating regime of the industrial manufacturing plant changed substantially, and data comprising all available process variables is used for a model re-creating step, wherein, in response to determining that the percentage of data points exceeding the upper limit of the statistical metric is less than the predefined threshold, the operating regime of the industrial manufacturing plant is determined to be substantially unchanged, and data comprising only variables present in one or more active models is used for a model re-tuning step”.
However, in the same field of endeavor namely optimizing the performance of industrial manufacture monitoring models, Jin teaches “computing T2 metric from principal component analysis or Mahalanobis distance (MD) for the second set of data, in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold, an operating regime of the industrial manufacturing plant changed substantially, and data comprising all available process variables is used for a model re-creating step, wherein, in response to determining that the percentage of data points exceeding the upper limit of the statistical metric is less than the predefined threshold, the operating regime of the industrial manufacturing plant is determined to be substantially unchanged, and data comprising only variables present in one or more active models is used for a model re-tuning step”. (Pg. 5, “Solutions to lack of robustness in fault detection models include methods such as using local models for the encountered regimes, introducing a correction factor in model parameters” [model re-tuning step], “or adaptively re-training of models when process drift is detected” [data comprising all available process variables is used for a model re-creating step]. Pg. 19 “Recursive PCA updates the correlation matrix by adding a new sample to its current value, and yet moving-window PCA updates the correlation matrix by getting rid of the first sample from the correlation matrix, and then adding a new sample to the matrix. However, the conventional moving-window PCA consists of two steps for modeling: elimination of old sample and inclusion of new sample. An improved moving-window PCA method takes advantage of both of the aforementioned methods, and derives a set of equations to directly calculate from the original matrix to the updated matrix which eliminates the oldest sample and includes the new sample” [data comprising only variables present in one or more active models is used for a model re-tuning step]. Pg. 16 “When applying PCA for fault detection, statistics such as Hotelling’s T2 and squared prediction error (SPE) are adopted as the health index to monitor whether the data set is deviating from the normal condition” [computing T2 metric from principal component analysis for the second set of data]. “The health limit of T2 and SPE statistics are based on the distribution of each of them, and usually a 99% or 95% confidence limit is calculated as the threshold” and Fig. 4-4 shows that a deviation via a detected fault occurs only above the limit otherwise, a deviation does not occur [in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold, an operating regime of the industrial manufacturing plant changed substantially and wherein, in response to determining that the percentage of data points exceeding the upper limit of the statistical metric is less than the predefined threshold, the operating regime of the industrial manufacturing plant is determined to be substantially unchanged]. Further see Chap. 1-2 and Chap. 4. The examiner has interpreted that applying PCA for fault detection using Hotelling’s T2 to monitor whether the data set is deviating from the normal condition and determine the limit of T2 is above a certain threshold or not to be determined as a deviation via a detected fault or not for adding samples to current values of the original matrix to introduce a correction factor to the model parameters or adaptively re-train the model when a process drift is detected as computing T2 metric from principal component analysis or Mahalanobis distance (MD) for the second set of data, in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold, an operating regime of the industrial manufacturing plant changed substantially, and data comprising all available process variables is used for a model re-creating step, wherein, in response to determining that the percentage of data points exceeding the upper limit of the statistical metric is less than the predefined threshold, the operating regime of the industrial manufacturing plant is determined to be substantially unchanged, and data comprising only variables present in one or more active models is used for a model re-tuning step.)
Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the claimed invention to add “computing T2 metric from principal component analysis or Mahalanobis distance (MD) for the second set of data, in response to determining a percentage of data points of the second set of data exceeding an upper limit of a statistical metric is greater than a predefined threshold, an operating regime of the industrial manufacturing plant changed substantially, and data comprising all available process variables is used for a model re-creating step, wherein, in response to determining that the percentage of data points exceeding the upper limit of the statistical metric is less than the predefined threshold, the operating regime of the industrial manufacturing plant is determined to be substantially unchanged, and data comprising only variables present in one or more active models is used for a model re-tuning step” as conceptually seen from the teaching of Jin, into that of Cella and Xu because this modification of determining the deviation from second data set for the advantageous purpose of providing information on the variables that deviate the models data and are no longer within the acceptable range (Jin, Pg. 2-3). Further motivation to combine be that Cella, Xu, and Jin are analogous art to the current claim and are directed to optimizing the performance of industrial manufacture monitoring models.
As per claim 2, Cella teaches “wherein the plurality of models includes one or more prediction models, one or more detection models, one or more classification models, one or more diagnostic models and one or more prognostic models.” (Para. 317, “In the cloud platform, optionally using massively parallel computational capability, a plurality of different neural networks of several types (including modular forms, structure-adaptive forms, hybrids, and the like) may be used to undertake prediction,” [prediction models] “classification,” [classification models] “control functions, and provide other outputs as described in connection with expert systems disclosed throughout this disclosure.” Para. 318, “In examples, the tasks may also be classified as machine learning problems such as classification, regression, clustering, density estimation, dimensionality reduction, anomaly detection, and the like” [detection models]. Para 321, “The cognitive input selection systems 4004, 4014 may use intelligence and machine learning capabilities described elsewhere in this disclosure, such as using detected conditions (such as conditions informed by the input sources 116 or sensors), state information (including state information determined by a machine state recognition system 4020 that may determine a state), such as relating to an operational state, an environmental state, a state within a known process or workflow, a state involving a fault or diagnostic condition, or many others” [diagnostic models]. Para. 600, “A monitoring device may be structured to interpret detection values received from a plurality of sensors measuring a variety of characteristics associated with the gearbox such as temperature, vibration, and the like. The monitoring device may process the detection values to identify gear and gearbox health and anticipated life” [prognostic models]. Further see Para. 317-321 and 600. The examiner has interpreted that models that can predict values, detect anomalies, classify equipment states, and determine equipment health and lifespan as one or more prediction models, one or more detection models, one or more classification models, one or more diagnostic models and one or more prognostic models).
As per claim 3, Cella teaches “wherein each of the plurality of models is either the physics-based model or the data-driven model or a hybrid physics plus data-driven model.” (Para. 904, “References to a neural net throughout this disclosure should be understood to encompass a wide range of different types of neural networks, machine learning systems, artificial intelligence systems, and the like, such as model-based systems (including ones based on physical models, statistical models, flow-based models, biological models, biomimetic models, and the like).” Further see Para. 904. The examiner has interpreted that a neural network consisting of mode-based systems like physical model and statistical models as a physics-based model or a data-driven model).
As per claim 4, Cella teaches “wherein the preprocessing of the received plurality of data is also for verification of availability of the received plurality of data, removal of redundant data, and unification of sampling frequency, and wherein the pre-processing is performed subject to plant-specific data acquisition constraints.” (Para. 208, “Figure 5 depicts a programmatic data marketplace 70, which may be a self-organizing marketplace, such as for making available data that is collected in industrial environments, such as from data collectors,” [verification of availability of received data]. Para. 522, “Gathering and processing data from sets of sensors may facilitate determining which sensors contribute meaningful data to the set, and those sensors that do not contribute can be removed from the set” [removing unmeaningful data from data set which does not contribute, e.g., removal of redundant data]. Para. 36, “The stream of sensed data includes a range of frequencies that exceeds the frequency range of the set of sensed data, the processing comprising executing an algorithm on a portion of the stream of sensed data that is constrained to the frequency range of the set of sensed data, the algorithm configured to process the set of sensed data” [unification of sampling frequency, and wherein the pre-processing is performed subject to plant-specific data acquisition constraints]. Further see Para. 36, 208 and 522. The examiner has interpreted making available data that is collected in industrial environments, removing unmeaningful data from data set, and constraining data to frequency range of the set of the sensed data from the sensors as verification of availability of the received plurality of data, removal of redundant data, and unification of sampling frequency, and wherein the pre-processing is performed subject to plant-specific data acquisition constraints).
As per claim 7, Cella teaches “wherein the model retuning of each of the plurality of models is carried out based on the combination of the selected first set of data and second set of data of the industrial manufacturing plant without changing the input variables and the learning techniques used in the plurality of models.” (Para. 211, “Machine learning may be used to improve the foregoing, such as by adjusting one or more weights,” [model retuning]. Para. 354, “Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream and is improved based on data collected on performance in an industrial environment” [modifying the model based on the first and second set of data]. Para. 692, “In embodiments, the monitoring application 8776 may feed a neural net with the selected subset to learn to recognize various operating state, health states (e.g., lifetime predictions) and fault states utilizing deep learning techniques” [modifying each model]. Further see Para. 211, 354, and 692. The examiner has interpreted that using machine learning programming techniques to improve the model by adjusting the weights where the model is based on a training set created from human analysis of a data stream and is improved based on data collected on performance in an industrial environment for the selected subset as wherein the model retuning of each of the plurality of models is carried out based on the combination of the selected first set of data and second set of data of the industrial manufacturing plant. The examiner has also interpreted that the retuning of the model does not change the input variables and learning techniques used in the models as these steps are conducted in model recreating and model rebuilding, respectively, as detailed below).
As per claim 8, Cella teaches “wherein the model rebuilding of the plurality of models is carried out based on the combination of the first set of data and the second set of data of the industrial manufacturing plant using a plurality of learning techniques without changing the input variables used in the plurality of models.” (Para. 211, “Machine learning may be used to improve the foregoing, such as by adjusting one or more weights, structures, rules, or the like (such as changing a function within a model)” [based on using different learning techniques]. Para. 318, “In embodiments, machine learning may involve performing a plurality of machine learning tasks” [plurality of learning techniques] “by machine learning systems, such as supervised learning, unsupervised learning, and reinforcement learning.” Para. 354, “Embodiments include training a machine learning model to monitor a continuous ultrasonic monitoring data stream where the model is based on a training set created from human analysis of such a data stream and is improved based on data collected on performance in an industrial environment” [modifying the model based on the first and second set of data]. Further see Para. 211, 318, and 354. The examiner has interpreted that using machine learning programming techniques improve the model by adjusting the functions with the model and performing a plurality of machine learning tasks where the model is based on a training set created from human analysis of such a data stream and is improved based on data collected on performance in an industrial environment as wherein the model rebuilding of the plurality of models is carried out based on the combination of the first set of data and the second set of data of the industrial manufacturing plant using a plurality of learning techniques. The examiner has also interpreted that the rebuilding of the model does not change the input variables used in the models as this step is conducted in model recreating, as detailed below).
As per claim 9, Cella teaches “wherein the model recreating of the plurality of models is carried out based on the combination of the first set of data and the second set of data of the industrial manufacturing plant using a plurality of learning techniques and new variables identified through at least one feature selection technique.” (Para. 211, “For example, the system may learn to recognize faults to recognize patterns, to develop models or functions, to develop rules, to optimize performance, to minimize failure rates, to optimize profits, to optimize resource utilization, to optimize flow (such as flow of traffic), or to optimize many other parameters that may be relevant to successful outcomes (such as outcomes in a wide range of environments)” [model recreating based on new variables identified through feature selection technique]. Para. 211, “For example, alternative available sensor inputs for a data collection system 102 may be arranged in alternative configurations and permutations, such that the system may, using generic programming techniques over a series of data collection events,” [based on the first and second set of data] “determine what permutations provide successful outcomes based on various conditions.” Para. 318, “In embodiments, machine learning may involve performing a plurality of machine learning tasks” [plurality of learning techniques] “by machine learning systems, such as supervised learning, unsupervised learning, and reinforcement learning.” Further see 211 and 318. The examiner has interpreted that using optimizing the performance of the model using other parameters found in other outcomes and implementing alternative configurations, permutations, and machine learning tasks over a series of data collection events as wherein the model recreating of the plurality of models is carried out based on the combination of the first set of data and the second set of data of the industrial manufacturing plant using a plurality of learning techniques and new variables identified through at least one feature selection technique.)
As per claim 10, Cella teaches “wherein the combination of the first set of data and the second set of data for model re-tuning, model rebuilding or model re-creating is chosen such that MQI of the re-tuned model, rebuilt model, or re-created model is maximized.” (Para. 319, “By continuously adjusting parameters to cause outputs to match actual conditions, the machine learning facility may self-organize to provide a highly accurate model of the conditions of an environment” [retuned, rebuilt, and recreated model performance is maximized based on the combination of the first and second data sets]. Para. 321, “For example, if a data stream consisting of a particular combination of sensors and inputs yields positive results in a given set of conditions (such as providing improved pattern recognition, improved prediction, improved diagnosis, improved yield, improved return on investment, improved efficiency, or the like), then metrics relating to such results from the analytic system 4018 can be provided via the learning feedback system 4012 to the cognitive input selection systems 4004, 4014 to help configure future data collection to select that combination in those conditions” [model is chosen based on the best MQI]. Further see 319-321. The examiner has interpreted that adjusting parameters to cause outputs to match actual conditions to provide a highly accurate model of the conditions of an environment that yields positive results in a given set of conditions for metrics that are provided to configure future data collection to select that combination in those conditions as the combination of the first set of data and the second set of data for model re-tuning, model rebuilding or model re-creating is chosen such that MQI of the re-tuned model, rebuilt model, or re-created model is maximized).
Re Claim 11, it is a machine claim, having similar limitations of claim 1. Thus claim 11 is also rejected under the similar rationale as cited in the rejection of claim 1.
Furthermore, regarding claim 11, Cella teaches “at least one memory storing a plurality of instructions; and one or more hardware processors communicatively coupled with the at least one memory, wherein the one or more hardware processors are configured to execute the plurality of instructions stored in the at least one memory”. (Para. 234, “intense signal processing activities including resampling, weighting, filtering, and spectrum processing may be performed by dedicated processors such as field-programmable gate array ("FPGAs"), digital signal processor ("DSP"), microprocessors, micro-controllers, or a combination thereof” [and one or more hardware processors]. Para. 344, “Self-organizing storage may allocate storage” [memory] “based on application of machine learning, which may improve storage configuration based on feedback measure over time. Storage may be optimized by configuring what data types are used ( e.g., byte-like structures, structures representing fused data from multiple sensors, structures representing statistics or measures calculated by applying mathematical functions on data, and the like), by configuring compression, by configuring data storage duration, by configuring write strategies (such as by striping data across multiple storage devices, using protocols where one device stores instructions for other devices in a chain, and the like),” [e.g., one memory storing a plurality of instructions; and one or more hardware processors communicatively coupled with the at least one memory, wherein the one or more hardware processors are configured to execute the plurality of instructions stored in the at least one memory]. Further see Para. 234 and 344. The examiner has interpreted performing signal processing activities on processors that stores instructions by a storage device as at least one memory storing a plurality of instructions; and one or more hardware processors communicatively coupled with the at least one memory, wherein the one or more hardware processors are configured to execute the plurality of instructions stored in the at least one memory.)
As per claim 12, Cella teaches “wherein the one or more databases include operations database, laboratory database, maintenance database and an environment database.” (Para. 210, “The platform 100 may include one or more local autonomous systems, such as for enabling autonomous behavior, such as reflecting artificial, or machine-based intelligence or such as enabling automated action based on the applications of a set of rules or models upon input data from the local data collection system 102 or from one or more input sources 116, which may comprise information feeds and inputs from a wide array of sources, including those in the local environment 104, in a network 110, in the host system 112, or in one or more external systems, databases, or the like.” Para. 231, In embodiments, the system may provide data to enable extended statistics capabilities for continuous monitoring as well as ambient local vibration for analysis that combines ambient temperature and local temperature and vibration levels changes for identifying machinery issues [database which has operations data, i.e., operations database]. Para. 582, “Production lines may also include one or more pumps for moving a variety of material including acidic or corrosive materials, flammable materials, minerals, fluids comprising particulates of varying sizes, high viscosity fluids, variable viscosity fluids, or high-density fluids” [model that analyzes chemical properties from various databases, i.e., laboratory database]. Para. 229, “In embodiments, the Jennie™ board also has the ability to store calibration data and system maintenance repair history data in an on-board card set” [database which has maintenance history, i.e., maintenance database]. Para. 254, “In embodiments, ambient environmental temperature and pressure,” [database which has environmental data, i.e., environmental database] “sensed temperature and pressure may be combined with long/medium term vibration analysis for prediction of any of a range of conditions or characteristics.” The examiner has interpreted the data being inputted from external databases containing ambient temperature, local temperature, and vibration levels changes for identifying machinery issues; fluids comprising particulates of varying sizes, high viscosity fluids, variable viscosity fluids, or high-density fluid; calibration data and system maintenance; and sensed ambient environmental temperature and pressure as one or more databases include operations database, laboratory database, maintenance database and an environment database).
Re Claim 13, it is a system claim, having similar limitations of claim 2 and claim 3. Thus claim 13 is also rejected under the similar rationale as cited in the rejection of claim 2 and claim 3.
Re Claim 14, it is an articles of manufacture claim, having similar limitations of claim 1. Thus claim 14 is also rejected under the similar rationale as cited in the rejection of claim 1.
Furthermore, regarding claim 14, Cella teaches “A non-transitory computer readable medium storing one or more instructions which when executed by a processor on a system, cause the processor to perform a method”. (Para. 234, “intense signal processing activities including resampling, weighting, filtering, and spectrum processing may be performed by dedicated processors such as field-programmable gate array ("FPGAs"), digital signal processor ("DSP"), microprocessors, micro-controllers, or a combination thereof” [processor on a system]. Para. 344, “Self-organizing storage may allocate storage” [computer readable medium] “based on application of machine learning, which may improve storage configuration based on feedback measure over time. Storage may be optimized by configuring what data types are used ( e.g., byte-like structures, structures representing fused data from multiple sensors, structures representing statistics or measures calculated by applying mathematical functions on data, and the like), by configuring compression, by configuring data storage duration, by configuring write strategies (such as by striping data across multiple storage devices, using protocols where one device stores instructions for other devices in a chain, and the like),” [e.g., storing one or more instructions which when executed by a processor]. Para. 1896, “The processor may access a non-transitory storage medium through an interface that may store methods, codes, and instructions as described herein and elsewhere” [A non-transitory computer readable medium storing one or more instructions]. Further see Para. 234 and 344. The examiner has interpreted performing signal processing activities on processors that stores instructions by a storage device accessed by a non-transitory storage medium through an interface that may store instructions as a non-transitory computer readable medium storing one or more instructions which when executed by a processor on a system, cause the processor to perform a method.)
Response to Arguments
Applicant’s arguments, see Pg. 19-30, filed June 23, 2026, with respect to the rejection(s) of the claims under 35 U.S.C. § 101 have been fully considered and are persuasive with regards to the amended independent claims which integrate the claimed invention into a practical application. Therefore, the rejection has been withdrawn.
Applicant’s arguments, see Pg. 31-33, filed June 23, 2026, with respect to the rejection(s) of claims 1 and 11 under 35 U.S.C. 103 have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of the amended claims and new cited art, necessitated by the applicant’s amendment, as detailed above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Liao, Linxia. “An adaptive modeling for robust prognostics on a reconfigurable platform.” PhD diss., University of Cincinnati, 2010 teaches a method for determining failure prediction in industrial machines using reconfigurable models.
Liu, Zongchang. "Cyber-physical system augmented prognostics and health management for fleet-based systems." PhD diss., University of Cincinnati, 2018 teaches improving industrial models with collection samples to drift in performance using T2 metric of PCA
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.
Examiner’s Note: The examiner has cited particular columns and line numbers in the reference that applied to the claims above for the convenience of the applicant. Although the specified citations are representative of the art and are applied to specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested from the applicant, to fully consider the references in their entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner. In the case of amending the claimed invention, the applicant is respectfully requested to indicate the portion(s) of the specification which dictate(s) the structure relied on for the proper interpretation and also to verify and ascertain the metes and bound of the claimed invention.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Simeon P Drapeau whose telephone number is (571)-272-1173. The examiner can normally be reached Monday - Friday, 8 a.m. - 5 p.m. ET.
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, Ryan Pitaro can be reached on (571) 272-4071. 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.
/SIMEON P DRAPEAU/Examiner, Art Unit 2188
/RYAN F PITARO/Supervisory Patent Examiner, Art Unit 2188
Prosecution Insights