DETAILED ACTION
1. Claims 1-19 are pending.
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 .
Priority
2. Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. KR10-2023-0041438, filed on 3/29/2023.
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Information Disclosure Statement
3. The information disclosure statement (IDS) submitted on March 28, 2024 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
4. The information disclosure statement (IDS) submitted on February 27, 2026 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
5. The information disclosure statement (IDS) submitted on June 25, 2026 is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Rejections - 35 USC § 103
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.
6. Claims 1 and 3-8, and 13-16 are rejected under 35 U.S.C. 103 as being unpatentable over Kaul et al. US 20230022733 A1 in view of Wu US 20230075005 A1.
7. With regard to claim 1, Kaul teaches:
A virtual sensor system for a digital twin application, comprising:
an edge gateway configured to collect data collected from physical sensors in the real world, apply the collected data to a virtual sensor model, and operate virtual sensors for configuring a digital twin world ([0003] Sensors, recording devices, and internet-of-things (IoT) devices connected to the physical asset collect data, often in real-time. The collected data can then be mapped to the virtual model of the digital twin. Any individual with access to the digital twin can see the real-time information about the physical asset operating in the real world without having to be physically present and viewing the physical asset while operating. Rather, users such as engineers can use the digital twin to understand not only how the physical asset is performing, but to predict how the physical asset may perform in the future, using the collected data from sensors, IoT devices and other sources of data and information being collected.); and
a virtual sensor framework configured to train the virtual sensor model using data, which is measured by the physical sensors, from the edge gateway and distribute the virtual sensor model to the edge gateway ([0017] In some embodiments, the operational conditions and/or changes to the performance and lifecycle of the manufacturing systems may be tracked and monitored using digital twin models and/or simulations. The digital twin models can be constantly updated to reflect the current operating conditions of the manufacturing systems in real time, and stream performance data, simulation data, lifecycle information, insights and recommendations that may be taken into consideration by the reinforcement learning model during model training and/or updates to the model over time when selecting which order dressing rules to apply. Sensor devices integrated into the manufacturing systems and positioned within the surrounding environment can collect manufacturing system performance data and operation metrics, monitor the changes in performance of the manufacturing systems in real time, including changes resulting in decreased performance over time. Digital twins may collect the data during actual production of products or run simulated production runs; offering insights or recommendations to the reinforcement learning model. The reinforcement learning model can analyze the digital twin data to determine how to compensate for the changes in manufacturing system performance by adjusting manufacturing characteristics, such as manufacturing equipment settings, during the translation of commercial characteristics to manufacturing characteristics, as described herein. Moreover, digital twins can also alert operators of the manufacturing systems about faulty or failing parts, settings or configuration adjustments that may modify or impact performance and suggest upgrades or replacement parts that may return the manufacturing system to previous optimal performance and/or improved operational performance.).
Although Kaul teaches of physical sensors in each device or within the surrounding environment that help digital twins measure operational performance and lifecycle of the manufacturing equipment against historical baselines. Reinforcement models dynamically adjust equipment settings for producing products to account for equipment performance degradation over time and changes in operation performance (Abstract). Additionally, Kaul teaches that communication networks can comprise, for example, copper wires, optical fibers, wireless transmission, routers, load balancers, firewalls, switches, gateway computers, edge servers, and/or other network hardware which may be part of, or connect to, nodes of the communication networks including devices, host systems, terminals or other network computer systems, but does not explicitly teach of an edge gateway.
However, in analogous art, Wu teaches:
[0042] Edge devices 161a-161n are any type of device configured to access network 110, or be accessed by other devices through network 110, such as via an edge gateway 162a-162n. According to various embodiments, edge devices 161a-161n are “IoT devices” which include any type of network-connected (e.g., Internet-connected) device. For example, in one or more embodiments, the edge devices 161a-161n include assets, sensors, actuators, processors, computers, valves, pumps, ducts, vehicle components, cameras, displays, doors, windows, security components, boilers, chillers, pumps, air handler units, HVAC components, factory equipment, and/or any other devices that are connected to the network 110 for collecting, sending, and/or receiving information.
[0043] The edge gateways 162a-162n include devices for facilitating communication between the edge devices 161a-161n and the cloud 105 via network 110. For example, the edge gateways 162a-162n include one or more communication interfaces for communicating with the edge devices 161a-161n and for communicating with the cloud 105 via network 110. According to various embodiments, the communication interfaces of the edge gateways 162a-162n include one or more cellular radios, Bluetooth, WiFi, near-field communication radios, Ethernet, or other appropriate communication devices for transmitting and receiving information. According to various embodiments, multiple communication interfaces are included in each gateway 162a-162n for providing multiple forms of communication between the edge devices 161a-161n, the gateways 162a-162n, and the cloud 105 via network 110. For example, in one or more embodiments, communication are achieved with the edge devices 161a-161n and/or the network 110 through wireless communication (e.g., WiFi, radio communication, etc.) and/or a wired data connection (e.g., a universal serial bus, an onboard diagnostic system, etc.) or other communication modes, such as a local area network (LAN), wide area network (WAN) such as the Internet, a telecommunications network, a data network, or any other type of network.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul with the teachings of Wu of edge gateways. Similarly to Kaul, Wu teaches of a digital twin architecture of an IoT platform that employs a variety of modeling techniques ([0059]). Moreover, Wu teaches that IoT devices are edge devices. Edge devices are a type of device that can be accessed by other devices through a network via an edge gateway. The edge gateways facilitate communication between the edge devices and the cloud via a network. The IoT layer 205 includes one or more components for device management, data ingest, and/or command/control of the edge devices 161a-161n. The components of the IoT layer 205 enable data to be ingested into, or otherwise received at, the IoT platform 125 from a variety of sources. [...] In some embodiments, only authorized data is sent to the IoT platform 125, and the IoT platform 125 only accepts data from authorized edge gateways 162a-162n and/or edge devices 161a-161n. According to various embodiments, data is sent from the edge gateways 162a-162n to the IoT platform 125 via direct streaming and/or via batch delivery. Further, after any network or system outage, data transfer will resume once communication is re-established and any data missed during the outage will be backfilled from the source system or from a cache of the IoT platform 125. According to various embodiments, the IoT layer 205 also includes components for accessing time series, alarms and events, and transactional data via a variety of protocols. This enables multiple forms of communication between the edge devices, the gateways, and the cloud that are secure, as discussed in Wu ([0052]).
8. With regard to claim 3, Kaul further teaches:
wherein the virtual sensor model learns based on the data collected from the physical sensors that are correlated with each other and predicts time series data in an area where no physical sensor is installed ([0003] A digital twin is a virtual representation of a physical system, machine, device or other asset. The digital twin tracks changes to the physical asset across the asset's lifespan and records changes to the asset as they occur. Digital twins are a complex virtual model that is an exact counterpart to the physical asset existing in real space. Sensors, recording devices, and internet-of-things (IoT) devices connected to the physical asset collect data, often in real-time. The collected data can then be mapped to the virtual model of the digital twin. Any individual with access to the digital twin can see the real-time information about the physical asset operating in the real world without having to be physically present and viewing the physical asset while operating. Rather, users such as engineers can use the digital twin to understand not only how the physical asset is performing, but to predict how the physical asset may perform in the future, using the collected data from sensors, IoT devices and other sources of data and information being collected. Moreover, digital twins can help manufacturers and providers of physical assets with information that helps the manufacturer understand how customers continue to use the products after the purchasers have bought the physical asset.).
9. With regard to claim 4, Kaul further teaches:
wherein the edge gateway includes:
a database configured to collect the data from the physical sensors ([0073] Examples of monitoring devices 231, 235 may include sensor devices, recording devices such as cameras and other types of audio or video recorders, as well as IoT devices. In some embodiments, the data metrics outputted by the monitoring devices 231, 235 may be transmitted to the metrics module 217, where the collected data may be stored or processed for storage in one or more data structures, such as a database.);
a data preprocessing module configured to preprocess the data stored in the database ([0073] In some embodiments, the data metrics outputted by the monitoring devices 231, 235 may be transmitted to the metrics module 217, where the collected data may be stored or processed for storage in one or more data structures, such as a database.);
a virtual sensor database configured to store the virtual sensor model ([0075] Embodiments of the creation engine 225 may perform tasks or functions associated with creating a digital twin models 223 reflecting a current state of a manufacturing asset 207. In some embodiments, initial versions of the digital twin models 223 depicting the base form of a manufacturing asset 207 provided by the manufacturer at the time of purchase, may be provided to the digital twin module 221 and/or stored as a digital twin model 223, or as part of one or more digital twin files maintained in a repository.); and
a virtual sensor operation module configured to operate the virtual sensors through the virtual sensor model using the data preprocessed in the data preprocessing module ([0003] Any individual with access to the digital twin can see the real-time information about the physical asset operating in the real world without having to be physically present and viewing the physical asset while operating; [0081] Embodiments of the digital twin module 221 may further use the collected data to aid in the performance of one or more simulations that may simulate manufacturing asset 207 performance within the digital twin model 223 and/or provide simulations using various scenarios of the digital twin models 223 to predict results implementing one or more changes to the configuration of manufacturing asset 207.).
10. With regard to claim 5, Kaul further teaches:
wherein the edge gateway further includes an abnormal signal detection module configured to, when it is determined that the characteristics of the data preprocessed in the data preprocessing module have changed, request update of the virtual sensor model from the virtual sensor framework according to a distribution of the changed data ([0017] Sensor devices integrated into the manufacturing systems and positioned within the surrounding environment can collect manufacturing system performance data and operation metrics, monitor the changes in performance of the manufacturing systems in real time, including changes resulting in decreased performance over time; [0077] Changes to the manufacturing asset 207 that may result in the creation of a new digital twin model 223 may be automatically created in response to changes in the configuration or operating conditions of a manufacturing asset 207. For example, a manufacturing asset 207 may receive repairs, maintenance, reconfigure asset settings 233 and/or install or remove components on the manufacturing asset 207. In response to the changes, the creation engine 225 may create a new version of the digital twin model 223 to reflect the changes and store the new digital twin model 223 within a repository and/or as part of the timeline tracking the evolution of the manufacturing asset 207. ).
11. With regard to claim 6, Kaul further teaches:
wherein, when the data preprocessed in the data preprocessing module is an abnormal signal, the abnormal signal detection module transmits an alarm to an administrator so that the administrator is able to check whether there is an abnormality in the physical sensors ([0078] For example, changes in performance may indicate the degradation of existing parts, new or replacement parts or components, modified configurations and asset settings 233, software or firmware update, damage, repairs, etc. Embodiments of the digital twin module 221 may analyze the performance changes based on the changes in the performance data collected and reflect the changes to the manufacturing asset 207 as a new digital twin model 223 in some embodiments. In other embodiments, the digital twin module 221 may report the detected changes in performance data to the metrics module 217, along with additional performance or lifecycle insights, and recommendations for compensating for such changes to operating conditions of the manufacturing assets 207, as reflected by the digital twin models 223.).
12. With regard to claim 7, Wu further teaches:
when an error between time series data measured by the physical sensors and time series data restored by a deep learning model for detecting an abnormal signal is greater than a preset threshold, the abnormal signal detection module determines that the data is an abnormal signal, and when the error is smaller than the threshold, the abnormal signal detection module determines that the data is a normal signal ([0092] Thus, method 600 provides an example embodiment of determining an anomaly score. Returning to FIG. 4, the method 400 proceeds to decision point 408. At decision point 408, a determination of whether the anomaly score is indicative of a potential fault of the asset is made. For example, the anomaly score is compared to one or more anomaly score thresholds. If the one or more anomaly score thresholds are satisfied, then it is determined that the anomaly score is not indicative of a potential fault of the asset. In contrast, it is determined that the anomaly score is indicative of a potential fault of the asset if an anomaly score threshold is not satisfied.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul with the teachings of Wu when an error between time series data measured by the physical sensors and time series data restored by a deep learning model for detecting an abnormal signal is greater than a preset threshold, the abnormal signal detection module determines that the data is an abnormal signal, and when the error is smaller than the threshold, the abnormal signal detection module determines that the data is a normal signal. Similarly to Kaul, Wu teaches of training an ML model using raw sensor data in a digital twin environment. Moreover, Wu teaches of an anomaly score that determines if there is a potential fault of the asset. When a fault is determined, fault data indicative of the potential fault of the asset is generated. In various embodiments, the fault data provides additional and contextual information regarding the potential fault indicated by the anomaly score. For example, the fault data generated includes specific sensors or sensor signals 702 determined to be anomalous (e.g., large temperature delta associated with a sensor signal 702 from the average temperature), an amount of time spanned by anomalous operation periods 804, and/or the like. As a further example, the fault data includes a fault identifier (e.g., predicted via a fault classification model) and/or one or more fault messages describing the type of fault (e.g., excess temperatures, low temperatures, incorrect temperature relationships between sensors). The method 400 then includes block 412 that causes presentation of the fault data and an indication of the one or more features considered in the determination of the anomaly score via the user interface. Accordingly, a user can visually examine the fault data and reach an understanding of why the potential fault may occur, how the potential fault may occur, where within the asset the potential fault may occur, and/or the like, as discussed in Wu ([0094]). Having a threshold to compare the anomaly score to enables a user to determine where in the asset the fault may occur, and allows the user to address the potential issues accordingly.
13. With regard to claim 8, Kaul further teaches:
wherein the virtual sensor framework includes:
a database configured to store preprocessed data from the edge gateway ([0073] In some embodiments, the data metrics outputted by the monitoring devices 231, 235 may be transmitted to the metrics module 217, where the collected data may be stored or processed for storage in one or more data structures, such as a database.);
a virtual sensor learning model module configured to train the virtual sensor model with the data stored in the database and generate the virtual sensors ([0003] Any individual with access to the digital twin can see the real-time information about the physical asset operating in the real world without having to be physically present and viewing the physical asset while operating; [0081] Embodiments of the digital twin module 221 may further use the collected data to aid in the performance of one or more simulations that may simulate manufacturing asset 207 performance within the digital twin model 223 and/or provide simulations using various scenarios of the digital twin models 223 to predict results implementing one or more changes to the configuration of manufacturing asset 207.);
a signal error analysis module configured to compare the data stored in the database with data of the virtual sensors input from the virtual sensor learning model module and correct errors of the virtual sensors ([0080] Embodiments of the asset monitoring devices 231 integrated into the manufacturing asset 207 can also provide error or diagnostic codes, which may further assist with identifying potential performance issues and changes in operating conditions. Through the use of the collected datasets, organized, analyzed and/or formatted by the data collection engine 227, the digital twin module 221 may analyze each of the manufacturing assets' 207 performance, identify failing parts, provide resolutions to cure errors or diagnostic codes and recommend optimal actions to improve or optimize performance of the manufacturing asset 207 based on one or more simulations.); and
a data and error monitoring visualization engine configured to monitor the data stored in the database and the errors received from the signal error analysis module, determine whether the virtual sensor model needs to be updated, and request data collection for learning from the edge gateway according to a result of the determination ([0081] Embodiments of the digital twin module 221 may further use the collected data to aid in the performance of one or more simulations that may simulate manufacturing asset 207 performance within the digital twin model 223 and/or provide simulations using various scenarios of the digital twin models 223 to predict results implementing one or more changes to the configuration of manufacturing asset 207. For example, simulations predicting effects of replacing particular parts, changing timings, adjusting asset settings 233 for manufacturing products, modifying onboard electrical or computing components or even replacing potentially defective asset monitoring devices 231 and/or environment monitoring devices 235. Data sets collected by the data collection engine 227 may contribute to building one or more simulation models that may be used by a simulation engine. In some embodiments, manufacturers and/or users of the manufacturing asset 207 may share the collected datasets amongst owners of the same manufacturing asset 207 to improve modeling that uses the data, increasing the overall amounts of data available amongst the community of owners, thus improving the prediction models, performance insights, simulation results and recommendations; [0082] Embodiments of the digital twin module 221 may comprise a reporting module 229. The reporting module 229 may perform functions, tasks and/or processes of the digital twin module 221 which may be directed toward reporting simulation results for digital twin models 223 of manufacturing assets 207, as well as provide to the order dressing module 211 with additional digital twin data that may be considered by the reinforcement learning model 215, including performance and lifecycle insights about the manufacturing assets 207 and recommendations for alleviating changes in operating conditions or performance changes. In some embodiments of the reporting module 229, the reporting module 229 may save and archive the simulation results to one or more files of a repository or database, which may be accessed by the order dressing module 211.).
14. Regarding claim 13, it is rejected under the same reasoning as claims 4 and 5 above. Therefore, it is rejected under the same rationale.
15. Regarding claim 14, it is rejected under the same reasoning as claim 7 above. Therefore, it is rejected under the same rationale.
16. Regarding claim 15, it is rejected under the same reasoning as claim 8 above. Therefore, it is rejected under the same rationale.
17. Regarding claim 16, it is rejected under the same reasoning as claim 8 above. Therefore, it is rejected under the same rationale.
18. Claim 2 is rejected under 35 U.S.C. 103 as being unpatentable over Kaul et al. US 20230022733 A1, as applied in claim 1, in further view of Berti et al. US 20220114310 A1.
19. With regard to claim 2, Kaul teaches:
wherein the virtual sensor model predicts time series data on the basis of correlation characteristics of the data collected from the physical sensors and the data collected from the physical sensors at a current moment in order to construct a digital twin world in the field of bridges ([0003] A digital twin is a virtual representation of a physical system, machine, device or other asset. The digital twin tracks changes to the physical asset across the asset's lifespan and records changes to the asset as they occur. Digital twins are a complex virtual model that is an exact counterpart to the physical asset existing in real space. Sensors, recording devices, and internet-of-things (IoT) devices connected to the physical asset collect data, often in real-time. The collected data can then be mapped to the virtual model of the digital twin. Any individual with access to the digital twin can see the real-time information about the physical asset operating in the real world without having to be physically present and viewing the physical asset while operating. Rather, users such as engineers can use the digital twin to understand not only how the physical asset is performing, but to predict how the physical asset may perform in the future, using the collected data from sensors, IoT devices and other sources of data and information being collected. Moreover, digital twins can help manufacturers and providers of physical assets with information that helps the manufacturer understand how customers continue to use the products after the purchasers have bought the physical asset.).
Although Kaul teaches that the virtual sensor model predicts time series data based on the characteristics of the data collected from the physical sensors at the current moment in order to construct a digital twin world, Kaul fails to explicitly teach that it is in the field of bridges.
However, in analogous art, Berti teaches:
[0001] The present invention relates generally to the field of digital twins, which are a type of machine readable data sets that can be used in managing various kinds of physical assets, like vehicles, bridges, buildings, utility poles, fire hydrants, flood walls, emergency phones, tended gardens, lawns, pumps, large scale electrical components (for example, transformers, power towers), HVAC (heating, ventilation and air conditioning) components (for example, ductwork), land, facilities, infrastructure, major manufacturing equipment, and so on.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul with the teachings of Berti where the digital twins can manage various kinds of physical assets like bridges. Kaul teaches that engineers use digital twins to see how the physical asset is currently operating and how it will operate in the future. Similarly, Berti teaches that the physical assets can be bridges. Therefore, it would have been obvious that the digital twins taught in Kaul can be used in a bridge environment like in Berti.
20. Claims 9 and 17 are rejected under 35 U.S.C. 103 as being unpatentable over Kaul et al. US 20230022733 A1, as applied in claim 8, in further view of Brikis et al. US 20210109973 A1.
21. With regard to claim 9, Kaul teaches the virtual sensor system of claim 8 but fails to explicitly teach wherein the virtual sensor learning model module includes an encoder configured to receive time series data of the physical sensors as an input and a decoder configured to receive the time series data of the physical sensors as an input, and the virtual sensor learning model module sets the time series data of the physical sensors as an input of the encoder, sets the time series data of the physical sensors as an input of the decoder, sets a result of applying teacher forcing to an input sequence of the decoder as a target sequence of the decoder, and performs training of the virtual sensor model by inputting a last internal state of the encoder as an initial state of the decoder.
However, in analogous art, Brikis teaches:
wherein the virtual sensor learning model module includes an encoder configured to receive time series data of the physical sensors as an input and a decoder configured to receive the time series data of the physical sensors as an input, and the virtual sensor learning model module sets the time series data of the physical sensors as an input of the encoder, sets the time series data of the physical sensors as an input of the decoder, sets a result of applying teacher forcing to an input sequence of the decoder as a target sequence of the decoder, and performs training of the virtual sensor model by inputting a last internal state of the encoder as an initial state of the decoder ([0010] The training procedure may be described as follows: Sampled sensor observations from the training data are fed as input to a sensor encoder of the graph generative model (e.g. Long-short-term-memory (LSTMs) may be used to get encodings of multi-variate sensor observations). A graph decoder network takes the encoded sensor observations and maps them to a graph structure (e.g. Graph Convolutional Policy Network (GCPN) may be used for this). A Graph Discriminator network estimates how close the generated graph is to the original example from the training data. The estimation is used as reward for the graph generative model. Once the GCPN pulls a stop action (or after a maximum number of steps), a corresponding final generated graph is fed into a queue of “fake” generated examples. The “fake” example queue is used to train a graph discriminator network in parallel to the graph generative model—e.g. a generative adversarial network (GAN) training model; [0038] Training step 2 T2: A graph decoder network takes the encoded sensor observations and maps the observations to a graph structure (e.g. Graph Convolutional Policy Network (GCPN) may be used for this). A Graph Discriminator network estimates how close the generated graph is to the original example from the training data. This estimation is used as reward for the graph generative model.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul with the teachings of Brikis wherein the virtual sensor learning model module includes an encoder configured to receive time series data of the physical sensors as an input and a decoder configured to receive the time series data of the physical sensors as an input, and the virtual sensor learning model module sets the time series data of the physical sensors as an input of the encoder, sets the time series data of the physical sensors as an input of the decoder, sets a result of applying teacher forcing to an input sequence of the decoder as a target sequence of the decoder, and performs training of the virtual sensor model by inputting a last internal state of the encoder as an initial state of the decoder. Similarly to Kaul, Brikis teaches of collecting training data from physical sensors in order to train a digital twin model. Moreover, Brikis teaches of using an encoder-decoder deep neural network model. This dynamic continuous improvement method allows the model to learn and evaluate the generate better outputs in order to maximize its reward, as discussed in Brikis ([0010]; [0038]).
22. Regarding claim 17, it is rejected under the same reasoning as claim 9 above. Therefore, it is rejected under the same rationale.
23. Claims 10 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Kaul et al. US 20230022733 A1, as applied in claim 8, in further view of Serackis et al. US 20230067081 A1.
24. With regard to claim 10, Kaul teaches the virtual sensor system of claim 8 but fails to explicitly teach wherein the virtual sensor model includes a recurrent neural network (RNN)-based encoder and an RNN-based decoder, the encoder and the decoder are composed of a cell RNN with a stacked structure, and the cell RNN is implemented with at least one of a SimpleRNN, a long short-term memory (LSTM), and a gated recurrent unit (GRU).
However, in analogous art Serackis teaches:
wherein the virtual sensor model includes a recurrent neural network (RNN)-based encoder and an RNN-based decoder, the encoder and the decoder are composed of a cell RNN with a stacked structure, and the cell RNN is implemented with at least one of a SimpleRNN, a long short-term memory (LSTM), and a gated recurrent unit (GRU) ([0032] The digital twin unit 130 uses a separate ML model, model No. 4, trained according to the method in FIG. 4, blocks 407, 408, and 414. The data processing and ML model No. 4 execution diagram is shown in FIG. 3. Before ML model No. 4 is executed in bloc 304, the input data is prepared (302), the 3D marker cloud, obtained from unit 120, is combined with a previously estimated output of the model, obtained from block 304. In one embodiment, ML model No. 4 may have feed-forward structure-based neural network architectures. In another embodiment, ML model No. 4 can be based on recurrent neural network architecture, with RNN, LSTM, GRU, or other neural network cells with recurrent (feedback) connections between one or several neurons.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul with the teachings of Serackis wherein the virtual sensor model includes a recurrent neural network (RNN)-based encoder and an RNN-based decoder, the encoder and the decoder are composed of a cell RNN with a stacked structure, and the cell RNN is implemented with at least one of a SimpleRNN, a long short-term memory (LSTM), and a gated recurrent unit (GRU). Similarly to Kaul, Serackis teaches of a digital twin using a ML model. Specifically, Serackis teaches that the ML model can be based on recurrent neural network architecture, with RNN, LSTM, GRU, or other neural network cells with recurrent (feedback) ([0032]). Therefore, it would have been obvious to one of ordinary skill in the art that Kaul’s digital twin ML model can include a RNN, LSTM, and GRU as shown in Serackis’s ML model.
25. Regarding claim 18, it is rejected under the same reasoning as claims 10 and 11 above. Therefore, it is rejected under the same rationale.
26. Claims 11 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Kaul et al. US 20230022733 A1 and Serackis et al. US 20230067081 A1, as applied in claim 10, in further view of Cheon et al. US 20230259112 A1.
27. With regard to claim 11, Kaul and Serackis teach the virtual sensor system of claim 10 but fail to explicitly teach wherein the virtual sensor model stores a final hidden state of the encoder in a state input representation layer (state), then inputs the final hidden state as the initial state of the decoder, generates a target sequence by shifting an input sequence of the decoder by one time step, and then sets an output signal that passes through a time-distributed fully connected layer as a target signal to perform learning, and the time-distributed fully connected layer performs learning so that the decoder is able to know what a next target signal is at each time step.
However, in analogous art, Cheon teaches:
wherein the virtual sensor model stores a final hidden state of the encoder in a state input representation layer (state), then inputs the final hidden state as the initial state of the decoder, generates a target sequence by shifting an input sequence of the decoder by one time step, and then sets an output signal that passes through a time-distributed fully connected layer as a target signal to perform learning, and the time-distributed fully connected layer performs learning so that the decoder is able to know what a next target signal is at each time step ([0222] Model 700 includes a first portion 720 (e.g., an encoder) and a second portion 740 (e.g., decoder). In some embodiments, the model is one or more of an autoencoder, a neural network model, a convolutional neural network model, a deep belief network, a feedforward neural network, a multilayer neural network, etc. The first portion 720 dimensionally reduces the input data 710 (e.g., metrology data) to a compressed form (e.g., compressed data 730). The input layer of the machine learning model may be separated from the compressed data by a number of hidden layers (two are depicted in FIG. 7, but any number of hidden layers may be used). During training of the machine learning model 700, the first portion 720 may generate one or more functions to fit input data 710 to a lower dimensional representation without guidance from a user. The reducing (e.g., compressing, encoding) may take place over several stages (i.e. convert input data 710 to partially compressed data first, then further to compressed data 730), or reducing (e.g., compressing, encoding) may be done in a single stage; [0223] Second portion 740 takes as input compressed data 730 and produces output data 750. During training, model 700 is trained to minimize the difference between input data 710 and output data 750, where output 750 is a reconstruction of input data 710 from compressed data 730. The minimization function used to train model 700 may also enforce penalties on the dimensionality of compressed data 730, to avoid returning a function with insufficient compression (e.g., the identity function, which perfectly recreates input data 710 but does not compress the data to a reduced dimensionality). Model 700 may be trained such that the output data 750 of model 700 approximately matches input data 710.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul and Serackis with the teachings of Cheon wherein the virtual sensor model stores a final hidden state of the encoder in a state input representation layer (state), then inputs the final hidden state as the initial state of the decoder, generates a target sequence by shifting an input sequence of the decoder by one time step, and then sets an output signal that passes through a time-distributed fully connected layer as a target signal to perform learning, and the time-distributed fully connected layer performs learning so that the decoder is able to know what a next target signal is at each time step. Cheon discusses an encoder-decoder model that utilizes a hidden layer in order to train the model. Because the hidden state is compressed data, it ensures that data is simplified and reduced, as discussed in Cheon ([0222]). Therefore, ensuring efficient storage, analysis, and faster processing.
28. Regarding claim 18, it is rejected under the same reasoning as claims 10 and 11 above. Therefore, it is rejected under the same rationale.
29. Claims 12 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Kaul et al. US 20230022733 A1 and Serackis et al. US 20230067081 A1, as applied in claim 10, in further view of Suzani et al. US 20220188694 A1.
30. With regard to claim 12, Kaul and Serackis teach the virtual sensor system of claim 10 but fail to explicitly teach wherein the signal error analysis module obtains an absolute value of an error between the data of the virtual sensors and data of the physical sensors, obtains an absolute error average during a set period, and then monitors the errors of the virtual sensors by performing an exponential moving average on the absolute error average.
However, in analogous art, Suzani teaches:
wherein the signal error analysis module obtains an absolute value of an error between the data of the virtual sensors and data of the physical sensors, obtains an absolute error average during a set period, and then monitors the errors of the virtual sensors by performing an exponential moving average on the absolute error average ([0069] Thus in an embodiment, while raw anomaly scores may have negative and positive values, the squared anomaly score instead isolates the magnitude of the anomaly score as an absolute value. A squared anomaly score may be used as follows; [0100] Step 403 adjusts the adaptive anomaly threshold based on an exponential moving average of anomaly scores that is μ.sub.t in the above adaptive anomaly threshold formula. Mathematics of adjusting the exponential moving average are presented later herein; [0106] Exponential moving average is a calculation that is safe and efficient with various arithmetic embodiments. In one embodiment, a counter tallies how many items were processed so far, and a delta measures an arithmetic difference of a previous moving average less an anomaly score of a current item. The delta is then scaled down according to a progressively diminishing weight such as 2/counter. That weighted delta is then added to the previous moving average to derive the next moving average; [0107] Various embodiments have various formulae for calculating an exponential moving average. In a most efficient embodiment that lacks a counter, an exponential moving average may instead be calculated according to the following exponential moving average formula; [0108] The above exponential moving average formula is based on the following numeric terms; [0109] x.sub.t is the normalized anomaly score of the current item; [0110] μ.sub.t-1 is the moving average of anomaly scores before adjustment based on the current normalized anomaly score [0111] α is a sensitivity constant that indicates how important or unimportant are past anomaly scores such that increasing α to almost one maximizes apparent volatility and decreasing α to almost zero instead maximizes smoothing; [0123] In step 501, the ML model calculates an anomaly score for an internet of things (IoT) telemetry item. For example, the stream of items that provides input for the ML model may contain items such as measurement records or operation log entries from remote devices such as sensors. For example, the anomaly score may reflect that a remote device may be hijacked, spoofed, or malfunctioning.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Kaul and Serackis with the teachings of Suzani wherein the signal error analysis module obtains an absolute value of an error between the data of the virtual sensors and data of the physical sensors, obtains an absolute error average during a set period, and then monitors the errors of the virtual sensors by performing an exponential moving average on the absolute error average. A statistical moving average, such as an exponential moving average, are computationally efficient and therefore are suitable for large scale problems and/or embedded deployments that may have high data rates, low computational resource availability, and/or low latency deadlines such as for live real-time stream processing. These moving statistics are robust in dealing with abrupt changes in the input data stemming from the natural distribution of data compared to gradual changes from concept drift as explained herein. That is, these measurements can distinguish trends versus spikes, as discussed in Suzani ([0021]). This is beneficial for a digital twin environment where monitoring can be real-time and allows the virtual model to react to systematic changes in the physical asset.
31. Regarding claim 19, it is rejected under the same reasoning as claim 12 above. Therefore, it is rejected under the same rationale.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to AN-AN N NGUYEN whose telephone number is (571)272-6147. The examiner can normally be reached Monday-Friday 8:00-5:00 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, AIMEE LI can be reached at (571) 272-4169. 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.
/AN-AN NGOC NGUYEN/Examiner, Art Unit 2195 /KEVIN L YOUNG/Supervisory Patent Examiner, Art Unit 2194
Prosecution Insights