Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Rejections - 35 USC § 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.
Claim 4 is 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.
In claim 4, the limitation of “the each context indicator” has unclear antecedent basis, and is therefore indefinite. While the preceding claims 2 and 3 recites “a context indicator” (in claim 2) and then “the context indicator” (in claim 3), the preceding claims do not use “each context indicator.” Therefore, usage of the term “each” renders it unclear as to whether “the each context indicator” is the same as the previously recited “context indicator,” especially since the preceding claims use this term to describe a singular “context indicator” whereas “each” covers the concept of a plurality of context indicators. For purposes of examination, “the each context indicator” has been interpreted as “each context indicator.”
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
1. Claims 1-3, 9, and 15-22 are rejected under 35 U.S.C. 103 as being unpatentable over Guo et al. (US 2019/0302707 A1) (“Guo”) in view of Cantrell (US 10,579,932 B1) (cited in an IDS).
As to claim 1, Guo teaches a non-transitory computer readable medium storing a computer program, wherein when the computer program is executed by one or more processors of a computing device, the computer program performs methods for processing input data, [[0025]: “a non-transitory computer readable storage medium embodied thereon a program executable by a processor for performing a method”; [0024]: “the processor is coupled with stored instructions implementing the method, wherein the instructions”] and the methods include:
obtaining input data based on sensor data obtained during manufacturing of an article using one or more manufacturing recipes in a plurality of manufacturing equipment or using a plurality of manufacturing recipes in one or more manufacturing equipment; [[0040]: “FIG. 1 is a schematic diagram illustrating components of the manufacturing anomaly detection system 100 according to some embodiments. The system 100 includes manufacturing production line 110, a training data pool 120, machine learning model 130 and anomaly detection model 140. The production line 110 uses sensors to collect data. The sensor can be digital sensors, analog sensors, and combination thereof. The collected data serve two purposes, some of data are stored in training data pool 120 and used as training data to train machine learning model 130 and some of data are used as operation time data by anomaly detection model 140 to detect anomaly.” See also [0056] (examples of sensors), [0060] (training data). That is, the input data is taught in the form of training and operation time data, and this is collected from sensors of a manufacturing equipment. The limitation of “manufacturing of an article using one more manufacturing recipes” is met by the prior art’s description of a manufacturing process for discrete items. See, e.g., [0003]: “Discrete manufacturing produces distinct items, e.g., automobiles, furniture, toys, and airplanes”; [0005]: “Discrete manufacturing includes a sequence of operations performed on work units, such as machining, soldering, assembling, etc.”; [0008]: “classes or types of the manufacturing operations can include process manufacturing and discrete manufacturing.” That is, manufacturing operations constitute a recipe, since the operations are performed to produce an item. The limitation of “in a plurality of manufacturing equipment” is met because a production line as shown in FIG. 1 includes a plurality of parts (equipment), noting that the instant claim recites no limitations on how the plurality of equipment differ from one another.]
feeding the input data and additional information for identifying one or more contexts about the input data into a neural network model loaded on a computer device, [Abstract: “An apparatus for controlling a system including a plurality of sources of signals causing a plurality of events includes an input interface to receive signals from the sources of signals, a memory to store a neural network trained to diagnose a control state of the system, a processor to submit the signals into the neural network to produce the control state of the system.” The input data is described in [0040]: “The collected data serve two purposes, some of data are stored in training data pool 120 and used as training data to train machine learning model 130 and some of data are used as operation time data by anomaly detection model 140 to detect anomaly.” The limitation of a “computer device” is met by the apparatus with the memory storing the neural network, as described above. The limitation of “additional information” is taught by the input data because there are multiple input variables from sensor data, e.g., 163, 166, and 169 as shown in FIG. 1 and described in [0042] (“Operation time data X1 163 and X2 166 are classified as normal and operation time data X3 169 is classified as anomalous.”) and the signals S1, S2, and S3 shown in FIG. 4B and signals Si0 and Si1 (1<i<3) FIG. 7 (as described in [0071]). See also [0061]: “Assume there are M data signals {Si}1M, which generate N events.” In other words, one portion of the signal data input into the neural network can be regarded as input data, while another portion can be regarded as additional information on the context, especially since the signals represent a state of the system as described in [0059]: “an event can represent abnormal status such as measured data being out of admissible operating range or normal status such as system changes from one state to another state.”] wherein the neural network model, including an encoder and a decoder, is trained with the input data and the additional information; [[0049]: “FIG. 3 shows a schematic of an autoencoder neural network used by some embodiments for unsupervised machine learning. An autoencoder neural network is a special artificial neural network to reconstruct the input data signals X 240 with the encoder 310 and the decoder 320 composed of a single or multiple hidden layers as shown in FIG. 3, where X 330 is the reconstructed data from the input data signals X 240. The reconstruction gives X=X. For the time delay autoencoder neural network, input data X 240 includes both current data and historic data.” As discussed above, the sensor data form the basis for training data. See [0060]: “Before training neural network, the training data are processed to extract events for all training data signals”; [0095]: “FIG. 10A shows a block diagram of a method used by a neural network trainer 933 to train the neural network 931 according to one embodiment.”] […]
generating an output by processing the input data using the neural network model based on the additional information about the input data, wherein the neural network model processes the input data for different contexts of the input data identified by the additional information; [Abstract: “submit the signals into the neural network to produce the control state of the system.” [0076]: “The neural network 931 is trained to diagnose a control state of the system.” That is, the control state corresponds to an output. Alternatively, or additionally, the computations of the neural network layers as shown in FIG. 3, for example, (e.g., layers 330 and 340) can also correspond to outputs of the neural network.] and
detecting an anomaly for a plurality of normal states corresponding to the input data based on the output of the neural network model. [[0042]: “For example, using normal data patterns 155 and 158, the trained machine learning model 150 can classify operation time data into normal data 170 and abnormal data 180.” [0074]: “For the test error threshold 810=0.018, FIG. 8A shows that the fully connected autoencoder neural network detected two anomalies, one corresponding to test error 820 and other corresponding to test error 830. The anomaly corresponding to test error 820 is a false alarm and the anomaly corresponding to test error 830 is the true anomaly. On the other hand, the structured autoencoder neural network only detected true anomaly corresponding to test error 840.”]
Guo does not explicitly teach “further feeding a context characteristic indicator that associates the input data with at least one of a characteristic of one manufacturing recipe of the one or more manufacturing recipes or a characteristic of one manufacturing equipment of the one or more manufacturing equipment, into the neural network model by matching with the input data.”
Cantrell teaches “further feeding a context characteristic indicator that associates the input data with at least one of a characteristic of one manufacturing recipe of the one or more manufacturing recipes or a characteristic of one manufacturing equipment of the one or more manufacturing equipment, into the neural network model by matching with the input data.” [FIG. 6 and col. 29 lines 32 to 41: “the data analytics platform may be configured to carry out a separate instance of phases 604-606 for each different mode that may be detected in the received stream of multivariate data points originating from data source 601, and data acquisition phase 602 may then include a mode detection operation 603 that serves to (1) detect the mode to which each data point in the received stream of multivariate data points belongs and then (2) direct each data point to the instance of phases 604-606 that is specific to the detected mode.” That is, this part teaches that the input data and the detected mode (context characteristic indicator) are fed into the model wherein the input data are matched with modes and the asset. See also col. 29 lines 19-24: “continuing with the example above where data source 601 takes the form of an asset that is capable of outputting data belonging to a first mode when the asset is operating in an idle state and outputting data belonging to a second mode when the asset is operating in an heavy usage state,” which teaches the mode associates the input data with an asset (equipment); col. 20 lines 33-35 and col. 20 lines 51-60, which teaches that the model can be an artificial neural network. In regards to the limitation of “of one manufacturing recipe…or… of one manufacturing equipment,” the context of a recipe and equipment is already taught in the base reference Guo, as discussed above, and Cantrell is cited for its general technique which is applicable to manufacturing equipment and processes. See Cantrell, col. 7, lines 28-35: “types of assets that may be monitored by asset data platform 102 may include… manufacturing equipment (e.g., robotics devices, conveyor systems, and/or other assembly-line machines).” Therefore, the instant limitation of the characteristic being of a recipe or equipment is suggested when Cantrell’s general techniques are applied to the context of Guo.]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Guo with the teachings of Cantrell by implementing the technique of detecting and using a mode as taught in Cantrell so as to arrive at the limitations of the instant claim. The motivation would have been to account for different modes during anomaly detection, which is beneficial for eliminating inaccuracies associated, as suggested by Cantrell (Col. 28, line 63 to col. 29, 13: “if the data analytics platform attempts to create and/or run an anomaly detection model based on data points originating from data source 601 that belong to different modes, this may result in inaccuracies while attempting to detect anomalies in the stream of data points originating from data source 601. For example, if the data analytics platform attempts to create an anomaly detection model based on data points belonging to different modes, this may result in what effectively amounts to a “mixed-mode” anomaly detection model that does not accurately score multivariate data points belonging to any particular mode.”).
As to claim 2, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, as set forth above.
Cantrell further teaches “wherein the methods further include: further feeding a context indicator that associates the input data with at least one of one manufacturing recipe of the one or more manufacturing recipes or one manufacturing equipment of the one or more manufacturing equipment, into the neural network model by matching with the input data.” [FIG. 6 and col. 29 lines 32 to 41: “the data analytics platform may be configured to carry out a separate instance of phases 604-606 for each different mode that may be detected in the received stream of multivariate data points originating from data source 601, and data acquisition phase 602 may then include a mode detection operation 603 that serves to (1) detect the mode to which each data point in the received stream of multivariate data points belongs and then (2) direct each data point to the instance of phases 604-606 that is specific to the detected mode.” That is, this part teaches that the input data and the detected mode (context indicator) are fed into the model wherein the input data are matched with modes and the asset. Col. 29 lines 19-24: “continuing with the example above where data source 601 takes the form of an asset that is capable of outputting data belonging to a first mode when the asset is operating in an idle state and outputting data belonging to a second mode when the asset is operating in an heavy usage state.” This part teaches that the mode associates the input data with an asset (equipment). Col. 20 lines 33-35 and Col. 20 lines 51-60 teaches that the model can be an artificial neural network).]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Guo with the teachings of Cantrell to have further arrived at the limitations of the instant dependent claim. The motivation for doing so is the same as the one given for the teachings of Cantrell in the rejection of the parent claim.
As to claim 3, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 2, wherein the neural network model is configured to process each input data differently [This limitation is met by the fact that the neural network model in Guo takes inputs in a specific manner. See, e.g., Guo, [0049]: “An autoencoder neural network is a special artificial neural network to reconstruct the input data signals X 240 with the encoder 310 and the decoder 320 composed of a single or multiple hidden layers as shown in FIG. 3.” Guo, [0071]: “It can be seen that the number of nodes at the input layer is six, i.e., number of data signals multiplied by number of time delay steps, and the number of nodes at the first hidden layer is three, i.e., number of data signals.” That is, different data signals are processed differently with respect to the different nodes at the neural network’s input layers. Alternatively, input data that is input at different times are considered to be processed differently.]
Cantrell further teaches “based on the context indicator matched with the each input data.” [FIG. 6 and col. 29 lines 32 to 41: “the data analytics platform may be configured to carry out a separate instance of phases 604-606 for each different mode that may be detected in the received stream of multivariate data points originating from data source 601, and data acquisition phase 602 may then include a mode detection operation 603 that serves to (1) detect the mode to which each data point in the received stream of multivariate data points belongs and then (2) direct each data point to the instance of phases 604-606 that is specific to the detected mode.” That is, this part teaches that the model processes the input data differently based on the mode detected (for example, some data can be processed in deployment of the model while other data points can be processed in the training of the model) to detect anomaly occurrences (generate output); Col. 20 lines 33-35 and Col. 20 lines 51-60 teaches the model can be an artificial neural network).]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Guo with the teachings of Cantrell to have further arrived at the limitations of the instant dependent claim. The motivation for doing so is the same as the one given for the teachings of Cantrell in the rejection of the parent claim.
As to claim 9, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein the neural network model is configured to process each input data differently [This limitation is met by the fact that the neural network model in Guo takes inputs in a specific manner. See, e.g., Guo, [0049]: “An autoencoder neural network is a special artificial neural network to reconstruct the input data signals X 240 with the encoder 310 and the decoder 320 composed of a single or multiple hidden layers as shown in FIG. 3.” Guo, [0071]: “It can be seen that the number of nodes at the input layer is six, i.e., number of data signals multiplied by number of time delay steps, and the number of nodes at the first hidden layer is three, i.e., number of data signals.” That is, different data signals are processed differently with respect to the different nodes at the neural network’s input layers. Alternatively, input data that is input at different times are considered to be processed differently.]
Cantrell further teaches “based on each context characteristic indicator matched with each input data.” [FIG. 6 and col. 29 lines 32 to 41: “the data analytics platform may be configured to carry out a separate instance of phases 604-606 for each different mode that may be detected in the received stream of multivariate data points originating from data source 601, and data acquisition phase 602 may then include a mode detection operation 603 that serves to (1) detect the mode to which each data point in the received stream of multivariate data points belongs and then (2) direct each data point to the instance of phases 604-606 that is specific to the detected mode.” That is, this part teaches that the input data and the detected mode (context characteristic indicator) are fed into the model wherein the input data are matched with modes and the asset. See also col. 29 lines 19-24: “continuing with the example above where data source 601 takes the form of an asset that is capable of outputting data belonging to a first mode when the asset is operating in an idle state and outputting data belonging to a second mode when the asset is operating in an heavy usage state,” which teaches the mode associates the input data with an asset (equipment); col. 20 lines 33-35 and col. 20 lines 51-60, which teaches that the model can be an artificial neural network.]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Guo and Cantrell to have also arrived at the claimed invention of the instant dependent claim. The motivation for doing so is the same as the one given for Cantrell in the rejection of the parent independent claim.
As to claim 15, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein the neural network model is a neural network model capable of processing all or one of encoding and decoding of the input data. [[0049]: “FIG. 3 shows a schematic of an autoencoder neural network used by some embodiments for unsupervised machine learning. An autoencoder neural network is a special artificial neural network to reconstruct the input data signals X 240 with the encoder 310 and the decoder 320 composed of a single or multiple hidden layers as shown in FIG. 3, where X 330 is the reconstructed data from the input data signals X 240. The reconstruction gives X=X. For the time delay autoencoder neural network, input data X 240 includes both current data and historic data.”]
As to claim 16, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein the anomaly includes all or one of an article anomaly of the article and manufacturing equipment anomaly of the one or more manufacturing equipment. [Guo, [0011]: “To that end, some embodiments apply neural network methods for anomaly detection in manufacturing systems. Using neural networks, additional anomalies that are not obvious from domain knowledge can be detected.” Guo, [0014]: “neural network trained to detect anomalies in the complex manufacturing systems.” Guo, [0005]: “Discrete manufacturing includes a sequence of operations performed on work units, such as machining, soldering, assembling, etc. Anomalies can include incorrect execution of one or more of tasks, or an incorrect order of the tasks.”]
As to claim 17, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein the anomaly includes a manufacturing anomaly detected by a sensor data when the article is produced in the one or more manufacturing equipment. [Guo, [0011]: “To that end, some embodiments apply neural network methods for anomaly detection in manufacturing systems. Using neural networks, additional anomalies that are not obvious from domain knowledge can be detected.” Guo, [0014]: “neural network trained to detect anomalies in the complex manufacturing systems.” Guo, [0005]: “Discrete manufacturing includes a sequence of operations performed on work units, such as machining, soldering, assembling, etc. Anomalies can include incorrect execution of one or more of tasks, or an incorrect order of the tasks.” As noted above, discrete manufacturing refers to the manufacturing of an article (Guo, [0003]).]
As to claim 18, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein the neural network model includes a neural network function selected from the group consisting of an AutoEncoder (AE), a Denoising AutoEncoder (DAE), or a Variational AutoEncoder (VAE). [Guo, [0046]: “For anomaly detection in manufacturing systems, some embodiments apply the time delay feedforward neural network and some embodiments apply the time delay autoencoder neural network.” Guo, [0049]: “FIG. 3 shows a schematic of an autoencoder neural network used by some embodiments for unsupervised machine learning.]
As to claim 19, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein the one or more manufacturing recipes includes an operating parameter of the manufacturing equipment for producing the article that is loaded on the one or more manufacturing equipment. [Guo, [0004]: “This method is common in process manufacturing industries, for example oil refining, where there is usually a good understanding of permissible ranges for physical variables, and quality metrics for the product quality are often defined directly in terms of these variables.” Guo, [0005]: “Discrete manufacturing includes a sequence of operations performed on work units, such as machining, soldering, assembling, etc. Anomalies can include incorrect execution of one or more of tasks, or an incorrect order of the tasks. …physical variables, such as temperature or pressure are out of range.” Guo, [0056]: “the data measurement collected from a sensor that monitors a specific property of the manufacturing system is called as a data signal, e.g., a voltage sensor measures voltage signal.” That is, in the sequence of operations (manufacturing recipe), there are physical parameters (e.g., voltage) for producing the item. The limitation of loaded on equipment is understood from the description of discrete and process manufacturing, where it is understood that the manufacturing system includes equipment for manufacturing the product. This is also illustrated as item 110 in FIG. 1, which shows physical manufacturing equipment.]
As to claim 20, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, wherein one input data comprises sensor data obtained during manufacturing of an article by using one manufacturing recipe of the one or more manufacturing recipes in one manufacturing equipment of the one or more manufacturing equipment. [Guo, [0056]: “the data measurement collected from a sensor that monitors a specific property of the manufacturing system is called as a data signal, e.g., a voltage sensor measures voltage signal.” The manufacturing of an article is disclosed in Guo, [0003]: “Discrete manufacturing produces distinct items, e.g., automobiles, furniture, toys, and airplanes.” [0005]: “Discrete manufacturing includes a sequence of operations performed on work units, such as machining, soldering, assembling, etc. Anomalies can include incorrect execution of one or more of tasks, or an incorrect order of the tasks. …physical variables, such as temperature or pressure are out of range.” That is, in the sequence of operations (manufacturing recipe), there is sensor data (e.g., voltage) for producing the item. The limitation of manufacturing equipment is understood from the description of discrete manufacturing, where it is understood that the manufacturing system includes equipment for manufacturing the product. This is also illustrated as item 110 in FIG. 1, which shows physical manufacturing equipment.]
As to claim 21, this claim is directed to a method comprising the same or substantially the same operations as those of claim 1. Therefore, the rejection made to claim 1 is applied to claim 21.
As to claim 22, this claim is directed to a computing device for performing operations that are the same or substantially the same operations as those of claim 1. Therefore, the rejection made to claim 1 is applied to claim 21.
Furthermore, Guo teaches “A computing device for processing input data, comprising: one or more processors; and a memory for storing computer programs executable on the one or more processors; wherein the one or more processors are configured to” [[0025]: “a non-transitory computer readable storage medium embodied thereon a program executable by a processor for performing a method”; [0024]: “the processor is coupled with stored instructions implementing the method, wherein the instructions”]
2. Claims 4, 7-8, 10, and 13-14 are rejected under 35 U.S.C. 103 as being unpatentable over Guo in view of Cantrell, and further in view of Zhao et al., “Learning to Monitor Machine Health with Convolutional Bi-Directional LSTM Networks,” Sensors 2017, 17, 273; doi:10.3390/s17020273 (“Zhao”) (cited in an IDS).
As to claim 4, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 3, wherein the neural network model processes the each input data differently,” [This limitation is met by the fact that the neural network model in Guo takes inputs in a specific manner. See, e.g., Guo, [0049]: “An autoencoder neural network is a special artificial neural network to reconstruct the input data signals X 240 with the encoder 310 and the decoder 320 composed of a single or multiple hidden layers as shown in FIG. 3.” Guo, [0071]: “It can be seen that the number of nodes at the input layer is six, i.e., number of data signals multiplied by number of time delay steps, and the number of nodes at the first hidden layer is three, i.e., number of data signals.” That is, different data signals are processed differently with respect to the different nodes at the neural network’s input layers. Alternatively, input data that is input at different times are considered to be processed differently.] but does not teach the further limitations of “by specifying one or all of one manufacturing equipment of the one or more manufacturing equipment and one manufacturing recipe of the one or more manufacturing recipes, based on the each context indicator matched with the each input data.”
Zhao teaches “by specifying one or all of one manufacturing equipment of the one or more manufacturing equipment and one manufacturing recipe of the one or more manufacturing recipes, based on the each context indicator matched with the each input data.” [Page 10, first paragraph: “three tool life tests named C1, C4 and C6 were selected as our dataset. Each test contains 315 data samples, while each data sample has a corresponding flank wear. For training/testing splitting, a three-fold setting is adopted such that two tests are used as the training domain and the other one is used as the testing domain. For example, when C4 and C6 are used as the training datasets, C1 will be adopted as the testing dataset. This splitting is denoted as c1. The details about training/testing splitting are shown in Table 1. Our task is defined as the prediction of tool wear depth based on the sensory input.” That is, this part teaches that the reference’s neural network, called CBLSTM, processes each dataset differently wherein each dataset represents a different tool life test; for example, C4 and C6 will be processed as training datasets while C1 will be processed as testing datasets; Table 1 teaches symbols (context indicators) matched with input datasets used to indicate the processing of datasets by the neural network).]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Zhao by implementing the neural network model so that it processes the each input data differently “by specifying one or all of one manufacturing equipment of the one or more manufacturing equipment and one manufacturing recipe of the one or more manufacturing recipes, based on the each context indicator matched with the each input data,” so as to arrive at the limitations of the instant claim. One of ordinary skill in the arts would have been motivated to make this modification in order to utilize a data set for both training and testing purposes, as suggested by Zhao (see parts cited above).
As to claim 7, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, as set forth above.
Cantrell further teaches “wherein the methods further include: feeding a context indicator that associates the input data with at least one of a characteristic of one manufacturing recipe of the one or more manufacturing recipes or a characteristic of one manufacturing equipment of the one or more manufacturing equipment, into a first preprocessing neural network model; processing the context indicator using the first preprocessing neural network model” [Fig. 6 and Col. 29 lines 32 to 41: “the data analytics platform may be configured to carry out a separate instance of phases 604-606 for each different mode that may be detected in the received stream of multivariate data points originating from data source 601, and data acquisition phase 602 may then include a mode detection operation 603 that serves to (1) detect the mode to which each data point in the received stream of multivariate data points belongs and then (2) direct each data point to the instance of phases 604-606 that is specific to the detected mode.” That is, Cantrell teaches feeding input stream of multivariate data points (corresponds to input data) and detected mode of the input data points (corresponds to additional information identifying context about the input data) into a model (training, creation, or deployment phase of the model) for anomaly detection. See also col. 29 lines 19-24: “continuing with the example above where data source 601 takes the form of an asset that is capable of outputting data belonging to a first mode when the asset is operating in an idle state and outputting data belonging to a second mode when the asset is operating in an heavy usage state,” which teaches the mode (context indicator) associates the input data with a characteristic of an asset (equipment) such as “idle state” or “heavy usage state.” Col. 20 lines 33-35 and Col. 20 lines 51-60 teaches the model can be an artificial neural network, which corresponds to a first preprocessing neural network model.]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Guo with the teachings of Cantrell to have further arrived at the above limitations of the instant dependent claim. The motivation for doing so is the same as the one given for the teachings of Cantrell in the rejection of the parent claim.
The combination of references thus far does not teach the limitations of “further feeding a preprocessed context indicator which is an output of the first preprocessing neural network model, into the neural network model, wherein the preprocessed context indicator is a dense representation of the context indicator.”
Zhao teaches “further feeding a preprocessed context indicator which is an output of the first preprocessing neural network model, into the neural network model, wherein the preprocessed context indicator is a dense representation of the context indicator” [Page 2, last full paragraph: “In this paper, we combine CNN with bi-directional LSTM to propose a novel machine health monitoring system named Convolutional Bi-directional LSTM networks (CBLSTMs). In our proposed CBLSTMs, CNN can extract local robust features, and bi-directional LSTMs, which are built on CNN, are able to encode the temporal information and learn representations”; page 7, second full paragraph: “However, in machine health monitoring systems, the sequential sensory data have strong temporal dependencies. It is meaningful to consider the future context. Therefore, the bi-directional LSTM is applied here.” That is, Zhao teaches teach that the bi-directional LSTMs (analogous to a second preprocessing neural network model) output temporal information and representations (corresponding to a context indicator as dense representations), which are then fed into a neural network with two Fully-connected Layers and Linear Regression Layer (see Figure 2). See also page 9, first full paragraph: “In this part, the output representation of the temporal encoder, a two-layer bi-directional LSTMs, is fed into multiple hidden layers to seek a higher-level representation. Two fully-connected dense layers are stacked together, in which the output of one layer is used as the input into the next layer.”]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of references combined thus far with the teachings of Zhao by modifying Guo, as already modified thus far, to include the technique of using a model that outputs temporal information and representations that are then fed into another neural network, so as to arrive at the limitations of the instant dependent claim. One of ordinary skill in the arts would have been motivated to make this modification because of the following: “an advanced recurrent model, bi-directional LSTMs are able to capture long-term dependencies in forward and backward ways. Additionally, the stacked LSTM layers can enable our module to learn more abstract and deep features” (Zhao et al. pg. 15 first full paragraph).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of references combined thus far with the teachings of Zhao by modifying Guo, as modified thus far, to include the technique of using a model that outputs temporal information and representations that are then fed into another neural network, so as to arrive at the limitations of the instant dependent claim. One of ordinary skill in the arts would have been motivated to make this modification because of the following: “an advanced recurrent model, bi-directional LSTMs are able to capture long-term dependencies in forward and backward ways. Additionally, the stacked LSTM layers can enable our module to learn more abstract and deep features” (Zhao et al. pg. 15 first full paragraph).
As to claim 8, the combination of Guo, Cantrell, and Zhao teaches the non-transitory computer readable medium according to claim 7, as set forth above.
Zhao further teaches “wherein the further feeding a preprocessed context indicator, which is an output of the first preprocessing neural network model, into the neural network model includes: feeding the preprocessed context indicator into an input layer or an intermediate layer of the neural network model.” [Page 2, last full paragraph: “In this paper, we combine CNN with bi-directional LSTM to propose a novel machine health monitoring system named Convolutional Bi-directional LSTM networks (CBLSTMs). In our proposed CBLSTMs, CNN can extract local robust features, and bi-directional LSTMs, which are built on CNN, are able to encode the temporal information and learn representations”; page 7, second full paragraph: “However, in machine health monitoring systems, the sequential sensory data have strong temporal dependencies. It is meaningful to consider the future context. Therefore, the bi-directional LSTM is applied here.” That is, Zhao teaches teach that the bi-directional LSTMs (analogous to a second preprocessing neural network model) output temporal information and representations (correspond to context indicator as dense representations), which are then fed into a neural network with two Fully-connected Layers and Linear Regression Layer (see Figure 2). See also page 9, first full paragraph: “In this part, the output representation of the temporal encoder, a two-layer bi-directional LSTMs, is fed into multiple hidden layers to seek a higher-level representation. Two fully-connected dense layers are stacked together, in which the output of one layer is used as the input into the next layer.”]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have further combined the teachings of Zhao and the other references combined thus far to arrive at the above limitations of the instant dependent claim. The motivation for doing so is the same as the one given for the teachings of Zhao in the rejection of the parent claim.
As to claim 10, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 9, wherein the neural network model processes the each input data differently [This limitation is met by the fact that the neural network model in Guo takes inputs in a specific manner. See, e.g., Guo, [0049]: “An autoencoder neural network is a special artificial neural network to reconstruct the input data signals X 240 with the encoder 310 and the decoder 320 composed of a single or multiple hidden layers as shown in FIG. 3.” Guo, [0071]: “It can be seen that the number of nodes at the input layer is six, i.e., number of data signals multiplied by number of time delay steps, and the number of nodes at the first hidden layer is three, i.e., number of data signals.” That is, different data signals are processed differently with respect to the different nodes at the neural network’s input layers. Alternatively, input data that is input at different times are considered to be processed differently.]
Zhao teaches “based on material characteristic information of the article that is obtained based on the each context characteristic indicator matched with the each input data.” [Page 10 first paragraph: “three tool life tests named C1, C4 and C6 were selected as our dataset. Each test contains 315 data samples, while each data sample has a corresponding flank wear. For training/testing splitting, a three-fold setting is adopted such that two tests are used as the training domain and the other one is used as the testing domain. For example, when C4 and C6 are used as the training datasets, C1 will be adopted as the testing dataset. This splitting is denoted as c1. The details about training/testing splitting are shown in Table 1. Our task is defined as the prediction of tool wear depth based on the sensory input” teaches the reference’s neural network, called CBLSTM, processes each dataset differently wherein each dataset represents a different tool life test (material characteristic information); for example, C4 and C6 will be processed as training datasets while C1 will be processed as testing datasets; Table 1 teaches symbols (context characteristic indicator) matched with input datasets used to indicate the processing of datasets by the neural network).]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of the references combined thus far with the teachings of Zhao by implementing the neural network model to process the each input data differently “based on material characteristic information of the article that is obtained based on the each context characteristic indicator matched with the each input data.” One of ordinary skill in the arts would have been motivated to make this modification in order to utilize a data set for both training and testing purposes, as suggested by Zhao (see parts cited above).
As to claim 13, the combination of Guo and Cantrell teaches the non-transitory computer readable medium according to claim 1, as set forth above.
Cantrell further teaches “wherein the methods further include: feeding a context characteristic indicator that associates the input data with at least one of a characteristic of one manufacturing recipe of the one or more manufacturing recipes or a characteristic of one manufacturing equipment of the one or more manufacturing equipment, into a second preprocessing neural network model; processing the context characteristic indicator using the second preprocessing neural network model;” [FIG. 6 and col. 29 lines 32 to 41: “the data analytics platform may be configured to carry out a separate instance of phases 604-606 for each different mode that may be detected in the received stream of multivariate data points originating from data source 601, and data acquisition phase 602 may then include a mode detection operation 603 that serves to (1) detect the mode to which each data point in the received stream of multivariate data points belongs and then (2) direct each data point to the instance of phases 604-606 that is specific to the detected mode.” That is, this part teaches feeding input stream of multivariate data points (corresponds to input data) and detected mode of the input data points (corresponds to additional information identifying context about the input data) into a model (training, creation, or deployment phase of the model) for anomaly detection. Col. 29 lines 19-24: “continuing with the example above where data source 601 takes the form of an asset that is capable of outputting data belonging to a first mode when the asset is operating in an idle state and outputting data belonging to a second mode when the asset is operating in an heavy usage state.” That is, this part teaches that the mode (context indicator) associates the input data with a characteristic of an asset (equipment) such as “idle state” or “heavy usage state.” Col. 20 lines 33-35 and Col. 20 lines 51-60 teaches that the model can be an artificial neural network), which corresponds to a second preprocessing neural network model.]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of Guo with the teachings of Cantrell to have further arrived at the above limitations of the instant dependent claim. The motivation for doing so is the same as the one given for the teachings of Cantrell in the rejection of the parent claim.
The combination of references thus far does not teach the limitations of “further feeding a preprocessed context characteristic indicator which is an output of the second preprocessing neural network model, into the neural network model, wherein the preprocessed context characteristic indicator is a dense representation of the context characteristic indicator.”
Zhao teaches “further feeding a preprocessed context characteristic indicator which is an output of the second preprocessing neural network model, into the neural network model, wherein the preprocessed context characteristic indicator is a dense representation of the context characteristic indicator” [Page 2, last full paragraph: “In this paper, we combine CNN with bi-directional LSTM to propose a novel machine health monitoring system named Convolutional Bi-directional LSTM networks (CBLSTMs). In our proposed CBLSTMs, CNN can extract local robust features, and bi-directional LSTMs, which are built on CNN, are able to encode the temporal information and learn representations”; page 7, second full paragraph: “However, in machine health monitoring systems, the sequential sensory data have strong temporal dependencies. It is meaningful to consider the future context. Therefore, the bi-directional LSTM is applied here.” That is, Zhao teaches teach that the bi-directional LSTMs (analogous to a second preprocessing neural network model) output temporal information and representations (correspond to context indicator as dense representations), which are then fed into a neural network with two Fully-connected Layers and Linear Regression Layer (see Figure 2). See also page 9, first full paragraph: “In this part, the output representation of the temporal encoder, a two-layer bi-directional LSTMs, is fed into multiple hidden layers to seek a higher-level representation. Two fully-connected dense layers are stacked together, in which the output of one layer is used as the input into the next layer.”]
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have combined the teachings of references combined thus far with the teachings of Zhao by modifying Guo, as modified thus far, to include the technique of using a model that outputs temporal information and representations that are then fed into another neural network, so as to arrive at the limitations of the instant dependent claim. One of ordinary skill in the arts would have been motivated to make this modification because of the following: “an advanced recurrent model, bi-directional LSTMs are able to capture long-term dependencies in forward and backward ways. Additionally, the stacked LSTM layers can enable our module to learn more abstract and deep features” (Zhao et al. pg. 15 first full paragraph).
As to claim 14, the combination of Guo, Cantrell, and Zhao teaches the non-transitory computer readable medium according to claim 13, as set forth above.
Zhao further teaches “wherein the further feeding a preprocessed context characteristic indicator, which is an output of the second preprocessing neural network model, into the neural network model includes: feeding the preprocessed context characteristic indicator into an input layer or an intermediate layer of the neural network model.” [Page 2, last full paragraph: “In this paper, we combine CNN with bi-directional LSTM to propose a novel machine health monitoring system named Convolutional Bi-directional LSTM networks (CBLSTMs). In our proposed CBLSTMs, CNN can extract local robust features, and bi-directional LSTMs, which are built on CNN, are able