DETAILED ACTION
This action is responsive to the claims filed on 12/23/2025. Claims 1-5, 7-10, 13, 18-19 21-26 and 29-30 are pending for examination.
This action is Final
Response to Arguments
Applicant’s arguments have been considered but are not persuasive.
(1) Iwamura allegedly fails to teach the claimed extraction condition / sub-group / larger sub-group because its data relates to images rather than sensor-measured physical parameters. (Remarks, page 18)
The rejection does not rely on Iwamura for any “image” limitation, but rather for its generic distance-based range search technique: setting a search area within a radius R around a query, and increasing R until a predetermined number of points is included. (Iwamura, page 6, paragraph 2). Iwamura expressly characterizes the nearest-neighbor technique as applicable beyond image recognition (i.e., not limited to images). (Iwamura, page 1, paragraph 4). In the combination, the claimed “selected measured physical parameter value … associated with the current operation” corresponds to selecting one of Maruchi’s sensor-measured current values (diagnosis target data) as the query value used for retrieval. (Maruchi paragraphs [0030] and [0039]). Thus, Iwamura teaches determining an extraction condition that sets a first range including the selected value (radius around the query), extracting records (points/data sets) within that range, and modifying the extraction condition by increasing the range until at least a minimum threshold number of records is obtained. (Iwamura, page 6, paragraph 2). Maruchi provides the industrial sensor-measured context and the stored “normal” learning data, while Iwamura provides the predictable mechanism for ensuring a sufficient number of comparable data sets are extracted based on proximity to a current measured value. (Maruchi paragraph 0099; Iwamura, page 6, paragraph 2).
(2) Iwamura allegedly is incompatible with Maruchi’s intended purpose because limiting regression data points to those “closest” to the sample would make the regression line inaccurate. (Remarks, page 19)
This argument is not persuasive because the claims do not require limiting the learning information to only a single nearest neighbor or to an arbitrarily small set, nor do they require using Iwamura to “approximate the nearest neighbor” for purposes of identifying one closest point. Rather, the claims require extracting a sub-group within a first range and expanding to a larger sub-group until a minimum threshold number of data sets is achieved. (Iwamura, page 6, paragraph 2). Iwamura’s cited disclosure teaches expanding the search radius until a predetermined number of points is included, which is consistent with extracting a sufficiently sized sub-group rather than selecting a single nearest neighbor. (Iwamura, page 6, paragraph 2). Applying Iwamura’s range-expansion extraction to Maruchi’s mostly normal learning data is a predictable way to obtain a sufficiently sized set of comparable normal samples under current operating conditions. (Maruchi paragraph 0099; Iwamura, page 6, paragraph 2). Further, Maruchi itself uses a distance-based decision rule (distance/anomaly degree compared to a threshold) in its diagnosis context, so selecting an adequate comparison set is a routine optimization rather than rendering Maruchi unsatisfactory. (Maruchi paragraph 0105).
(3) Tong allegedly fails to teach “generate from extracted data sets of the larger sub-group, learning information that indicates a unit space,” because Tong selects instances to label and does not generate new data points. (Remarks, page 20)
The claim does not require generating “new data points,” but rather generating learning information from extracted data sets that indicates a unit space. Tong teaches learning/training an SVM classifier/model, and expressly describes the hypothesis/version space as lying on the surface of a hypersphere representing unit weight vectors (i.e., a “unit” space in which consistent hypotheses can lie). (Tong, page 49, Fig. 2 caption). Tong also teaches using distances/margins relative to the hyperplane in feature space as part of the SVM learning/selection framework. (Tong, page 54, paragraph 1). “Generate learning information” reasonably encompasses constructing/training the model/representation itself from training data; it does not require synthesizing new instances. (Tong Abstract; Tong, page 54, paragraph 1). Further, the claim term “unit space” is not limited to Applicant’s disclosed “Mahalanobis space,” because the claims do not recite the MT method or Mahalanobis distance as a requirement. Accordingly, Tong’s taught SVM learning framework constitutes “learning information” produced from training data within a defined unit/feature space, satisfying the recited limitation in combination with the extracted larger sub-group supplied by Maruchi/Iwamura.
(4) Maruchi allegedly fails to teach “update the source data… in response to determining that the diagnosis result corresponds to the normal state,” because model update occurs at a later step (S105) and involves maintenance-person evaluation. (Remarks, pages 20-21)
Applicant’s argument is not persuasive because the claim does not require that the update occur immediately at the same step as the initial diagnosis computation, nor does it exclude user validation as part of the overall determination flow. Maruchi teaches model update processing at a later step in the workflow after evaluation of a graph and diagnosis result, and explains that model update processing is executed in connection with that evaluation (and may be executed periodically). (Maruchi paragraphs 0154–0155). Maruchi further teaches that learning data can be continuously added and used to repeat diagnosis model regeneration/verification for the same diagnosis target system, which is consistent with updating stored data used for diagnosis based on diagnosis outcomes within the disclosed process. (Maruchi paragraphs 0162–0163). Thus, the “in response to determining” language is satisfied where the update occurs as a consequence of a normal determination within the disclosed workflow (including evaluation/confirmation), even if performed after evaluation and/or periodically.
For at least the foregoing reasons, the §103 rejection of claims 1–5, 7–10, 13, 18, 19, 21–26 and 29-30 is maintained.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or non-obviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-5, 8, 10, 18-19, 21, 23-25 and 30 are rejected under 35 U.S.C. 103 as being unpatentable over Maruchi et al. (US 20190012553 A1), hereinafter referred to as Maruchi in view of Iwamura et al., (WO 2013129580 A1), hereafter referred to as Iwamura, and in further view of Tong et al. (Tong, S., & Koller, D. (2001). Support vector machine active learning with applications to text classification. Journal of machine learning research, 2(Nov), 45-66.), hereinafter referred to as Tong.
Claim 1: Maruchi teaches the following:
A system for fault diagnosis of a power-operated machine comprising: a storage device configured to store source data including a plurality of data sets, wherein the plurality of data sets include stored physical parameter values representing a normal state of the operation of the power operated machine; (Maruchi paragraph 99, “It is supposed that sample data acquired from a system of social infrastructure such as power generation station, plant, or railway mostly belongs to normal data and includes almost no anomalous data. Thus, measured data is used intact as learning data to generate a model representing the normal state of a diagnosis target system by unsupervised learning, and the model is used as an anomaly detection model.”,
Maruchi, paragraph 30, “The diagnostic device 100 performs anomaly diagnosis of the diagnosis target system 500 by using a diagnosis model generated by machine learning in advance based on measured data 1 (for example, data measured by a plurality of sensors) of the diagnosis target system 500.”, measured data used as input is almost entirely normal data, indicating a normal representation of the system. Furthermore, the claim’s “power-operated machine” corresponds to Maruchi’s diagnosis target system in industrial infrastructure (e.g., a power generation station/plant/railway system) that necessarily operates using power. The claim’s “physical parameter values” correspond to Maruchi’s sensor values measured by one or more sensors (physical measurements such as temperature/pressure/vibration/etc. depending on the sensor). Therefore, Maruchi’s “measured data … measured by … sensors” teaches the claimed measured/stored physical parameter values for a power-operated diagnosis target system.)
one or more sensors configured to obtain measured physical parameter values of the power-operated machine; (Maruchi, paragraph 39, “The measured data 1 includes data measured by one or a plurality of sensors. Measured data includes a measurement time and a sensor value.”)
circuitry configured to acquire diagnosis target data including the measured physical parameter values from the one or more sensors, wherein the measured physical parameter values represent a current operation of the power-operated machine; (Maruchi, paragraph 30, “The diagnostic device 100 performs anomaly diagnosis of the diagnosis target system 500 by using a diagnosis model generated by machine learning in advance based on measured data 1 (for example, data measured by a plurality of sensors) of the diagnosis target system 500”, in this fault diagnosis system, measured data, including parameters, of a system is being read by sensors and is used as input for classifying fault using machine learning algorithms.)
Iwamura, in the same field of machine learning data extraction, teaches the following limitations which the above fails to teach:
determine an extraction condition for a selected measured physical parameter value among the measured physical parameter values of the diagnosis target data, associated with the current operation of the power-operated machine, wherein the extraction condition sets a first physical parameter range that includes the selected measured physical parameter value; extract a sub-group of the plurality of data sets of the source data, for which the stored physical parameter value are within the first physical parameter range of the extraction condition; (Iwamura, page 6, paragraph 2, “Alternatively, the search range determination unit sets an area having a representative point within the range of the search radius R around the query as the search area, and searches until the number of points included in the search area reaches a predetermined number The radius R may be gradually increased.”, Iwamura further indicates the nearest neighbor search technique is not limited to image recognition and extends to other technical fields such as statistical classification, recommendation systems, etc. Iwamura’s “search range … within the range of the search radius R around the query” teaches “determin[ing] an extraction condition” that “sets a first physical parameter range that includes the selected measured physical parameter value” because the range is defined around (and includes) the selected/query value. Iwamura’s “sets an area … as the search area” and “searches until the number of points included … reaches a predetermined number” teaches extracting (retrieving) a sub-group of stored data sets whose stored values are within the first range, and doing so in a manner that is tied to the current (query) measured value and ensures at least a threshold count of data sets is obtained.)
modify the extraction condition, in response to determining that the sub-group contains fewer data sets than a minimum threshold number of data sets, wherein the modified extraction condition sets a second physical parameter range, that includes the selected measured physical parameter value and that is increased relative to the first physical parameter range; extract a larger sub-group of the plurality of data sets of the source data, for which the stored physical parameter values are within the second physical parameter range of the modified extraction condition, wherein the larger sub-group contains at least the minimum threshold number of data sets; (Iwamura, page 6, paragraph 2, “Alternatively, the search range determination unit sets an area having a representative point within the range of the search radius R around the query as the search area, and searches until the number of points included in the search area reaches a predetermined number The radius R may be gradually increased.”, The claim’s selected measured physical parameter value corresponds to Iwamura’s query (the current value used to retrieve comparable stored values). Iwamura’s “search radius R around the query” is a first physical parameter range that includes the query value (i.e., centered on it). Iwamura then explicitly teaches continuing the search “until … a predetermined number” of points is included and that “radius R may be gradually increased,” which matches modifying the extraction condition when the sub-group has fewer than a minimum threshold number of data sets, by increasing the range to extract a larger sub-group meeting the minimum count.)
Both Maruchi and Iwamura utilize data extraction techniques to learn patterns in data. Maruchi teaches a system that uses learning data from a certain period of time without any indication of an order or use of a sorting algorithm. Iwamura teaches a system that gradually extracts source data based on a dynamic search radius. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the teachings disclosed by Maruchi with the teachings disclosed by Iwamura. A motivation for the combination is to have a method for a predictable way of ensuring enough comparable “normal” samples are for stable diagnosis.
Tong, in the same field of machine learning data extraction, teaches the following:
generate from extracted data sets of the larger sub-group, learning information that indicates a unit space; (Tong, page 54, first paragraph, “For each unlabeled instance x, the shortest distance between its hyperplane in W and the vector wi is simply the distance between the feature vector Φ(x) and the hyperplane wi in F—which is easily computed by |wi · Φ(x)|. This results in the natural rule: learn an SVM on the existing labeled data and choose as the next instance to query the instance that comes closest to the hyperplane in F.”, the active learning algorithm selects the next data point by measuring the distance (via the SVM margin) between each candidates feature representation and a decision boundary.)
Maruchi teaches a system that uses learning data from a certain period of time without any indication of an order or use of a sorting algorithm. Tong teaches a system that places extraction conditions on its data sets. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the teachings disclosed by Maruchi and Iwamura with the teachings disclosed by Tong. A motivation for the combination is to have produce high quality accuracies from a limited set of data examples, (Tong, page 52, paragraph 4, “Thus, by shrinking the size ofthe version space as much as possible with each query, we are reducing as fast as possible the space in which w∗ can lie. Hence, the SVM that we learn from our limited number of queries will lie close to w∗.”, by selecting queries based on distance (e.g. choosing examples closest to the current SVM decision boundary), the algorithm rapidly narrows down the version space).
Maruchi further teaches:
after generating the learning information from the extracted data sets of the larger sub-group, determine a diagnosis result that indicates whether the diagnosis target data associated with the power-operated machine is faulty based on a comparison of a distance from the diagnosis target data to the learning information within the unit space, with a diagnosis threshold. (Maruchi, paragraph 105, “This regression model is generated from measured data in the normal state, and the distance D between sample data 911 and a regression line is defined to be an anomaly degree. It can be estimated that anomaly occurs to the diagnosis target system when the anomaly degree exceeds a threshold.”, Maruchi expressly uses distance D as an anomaly degree and teaches comparing it to a threshold to determine anomaly/fault. Therefore Maruchi teaches the “comparison … with a diagnosis threshold” part explicitly, and supports the “determine a diagnosis result” outcome (faulty/anomalous vs normal) based on that comparison.)
determine that the diagnosis result corresponds to the normal state of operation of the power-operated machine when the distance is less than the diagnosis threshold (Maruchi, paragraph 122, “The anomaly degree is an index indicating the degree of deviation of sample data determined to be anomalous from normal data. In the above-described exemplary linear regression model illustrated in FIG. 11, the distance from the straight line corresponds to the anomaly degree. The threshold is a value compared with the anomaly degree to perform anomaly determination. For example, it is determined that sample data is normal data (a diagnosis target is normal) when the anomaly degree of the sample data is equal to or lower than the threshold, or it is determined that the sample data is anomalous data (the diagnosis target is anomalous) when the anomaly degree is higher than the threshold.”, Maruchi teaches a distance that is compared to a threshold to detect anomaly. If anomaly occurs when distance exceeds the threshold, then the complementary condition distance less than the threshold corresponds to non-anomalous (normal) operation. Thus, the claimed determination that the diagnosis result corresponds to the normal state of operation when the distance is less than the diagnosis threshold is taught by Maruchi’s explicit “distance-vs-threshold” decision rule.)
and update the source data by storing the measured physical parameter values in a new data set of the source data in the storage device, in response to determining that the diagnosis result corresponds to the normal state of operation of the power-operated machine (Maruchi, paragraph 155, “The following describes, in detail, the model update processing at step S105 illustrated in FIG. 16. The model update processing is executed at each evaluation of a graph and a diagnosis result by the maintenance person in the process illustrated in FIG. 16, but does not necessarily need to be executed at each evaluation. The execution frequency of the model update processing can be determined with taken into account cost of a time, a calculation resource, and the like taken for the model update processing”, The claim requires “update the source data … in response to determining that the diagnosis result corresponds to the normal state.” Maruchi teaches a workflow where the diagnosis result is evaluated (graph/diagnosis evaluation), and then “model update processing” is executed in connection with that evaluation. In this arrangement, when the diagnosis is confirmed as corresponding to a normal state during evaluation, the system proceeds to update the stored learning information/learning data used for diagnosis (i.e., the stored source data) via the model update processing. Thus, Maruchi teaches updating stored data used for diagnosis in response to a normal determination within its disclosed diagnosis-and-evaluation process.)
and a display device configured to display the diagnosis result (Maruchi, paragraph 30, “The diagnostic device 100 generates, from the measured data 1, a graph referred to by a user such as a maintenance person or an operator (hereinafter collectively referred to as a maintenance person) to determine the validity of an anomaly diagnosis result. The diagnostic device 100 displays the created graph together with the anomaly diagnosis result on the screen display device 900.”)
Claim 3: Maruchi, Iwamura, and Tong teaches claim 2. Maruchi further teaches the following:
wherein the source data is updated with the normal data sets at a timing when a maintenance of the power-operated machine is completed. (Maruchi paragraph 77, “When the diagnoser 120 determines that anomaly is detected and the maintenance person determines that the determination is correct based on the content of a graph and an anomaly factor is correct, an approval operation may be performed without a correction operation.”, when an anomaly is detected a maintenance person analyzes the graphical output (information regarding the validity of the anomaly) and if the anomalous activity is deemed appropriately classified then it can undergo an approval operation without a correction operation.
Paragraph 73, “Upon the approval operation, the diagnosis result register 160 registers the diagnosis result (whether anomaly is detected and the factor classification result) approved by the maintenance person to the diagnosis history database 240. The following describes an exemplary specific case of the diagnosis result correction.”, when the approval operation is conducted (indicating maintenance is completed) then the measured data used to classify the anomalous activity is recorded in the historical database as learning data.)
Claim 4: Maruchi, Iwamura, and Tong teaches claim 2. Maruchi further teaches the following:
wherein the source data is updated with the normal data sets, after a predetermined time period has elapsed since a previous update of the source data. (Maruchi paragraph 155, “The model update processing may be executed at a timing determined by a method different from that in the present process. For example, the model update processing may be performed in a period such as once per week. The model update processing may be executed each time when evaluation of a graph and a diagnosis result by the maintenance person is performed a certain number of times. The model update processing may be performed at a timing other than those described above.”, extracted data is updated (updated data in the historical databases) after a certain time period has elapsed, in this case it is 1 week.)
Claim 5: Maruchi, Iwamura, and Tong teaches claim 2. Maruchi further teaches the following:
wherein the diagnosis target data includes a plurality of data pieces acquired at different times within a given time period, (Maruchi, paragraph 105, “This regression model is generated from measured data in the normal state, and the distance D between sample data 911 and a regression line is defined to be an anomaly degree. It can be estimated that anomaly occurs to the diagnosis target system when the anomaly degree exceeds a threshold.”, a fault level score here in being interpreted as the distance D.)
wherein the processor is further configured to calculate respective distances associated with the data pieces of the diagnosis target data based on the learning information (Maruchi, paragraph 105, “This regression model is generated from measured data in the normal state, and the distance D between sample data 911 and a regression line is defined to be an anomaly degree. It can be estimated that anomaly occurs to the diagnosis target system when the anomaly degree exceeds a threshold.”
Maruchi, paragraph 122, “An anomaly degree, a threshold, and an anomaly probability are displayed in addition to the anomaly determination result. The anomaly degree is an index indicating the degree of deviation of sample data determined to be anomalous from normal data.”, an anomaly degree is analogous to the fault level scores of the diagnosis data based on the learning information.)
and wherein the source data is updated with the normal data sets when each of the respective distances is greater than the diagnosis threshold for a predetermined number of data pieces acquired consecutively. (Maruchi, paragraph 46, “The sample data may be measured data at the measurement time of the diagnosis or measured data in a certain period including a plurality of measurement times. The sample data may be defined by a method other than the above-described method, such as a method of performing anomaly diagnosis.”
Maruchi, paragraph 105, “...the distance D... is defined to be an anomaly degree... anomaly occurs... when the anomaly degree exceeds a threshold”;
Maruchi, paragraph 155, “The following describes, in detail, the model update processing at step S105 illustrated in FIG. 16. The model update processing is executed at each evaluation of a graph and a diagnosis result by the maintenance person in the process illustrated in FIG. 16, but does not necessarily need to be executed at each evaluation. The execution frequency of the model update processing can be determined with taken into account cost of a time, a calculation resource, and the like taken for the model update processing”,
Maruchi, paragraph 157, “...additional learning data is prepared... learning data is newly added by reflecting... a diagnosis result...”, Maruchi teaches computing a distance/anomaly degree for sample data and comparing it to a threshold (distance greater than threshold indicates anomaly). Maruchi also teaches that sample data may include multiple data pieces acquired at a plurality of measurement times within a period (i.e., consecutive/series data pieces). Further, Maruchi teaches performing model update processing after evaluation occurs a “certain number of times,” which corresponds to a predetermined number of successive determinations on consecutively acquired data pieces. Maruchi further teaches that, during this update processing, additional learning data is newly added by reflecting recorded diagnosis results (i.e., updating the stored data sets used for learning/diagnosis based on the diagnosis outcomes over the series of measurements).)
Claim 8: Maruchi, Iwamura, and Tong teaches claim 1 Maruchi further teaches the following:
wherein the plurality of data sets of the source data includes old data sets having old physical parameter values that represent a previous normal state of operation of the power-operated machine, (Maruchi, paragraph 98, “When learning data used to generate an anomaly detection model does not include or hardly includes anomalous data and thus a large number of pieces of sample data are assumed to be normal data, unsupervised learning may be performed. The unsupervised learning is performed to generate a model indicating the normal state of a diagnosis target system, and anomaly detection is performed by using the model.”, the learning data used to generate anomaly detection aligns with the normal data sets of the claim language because these learning datasets predominantly consist of normal operational data)
and new data sets having new physical parameter values that represent an updated normal state of operation of the power-operated machine, (Maruchi, paragraph 134, “At step S101, the diagnosis model generator 110 generates a diagnosis model (anomaly detection model or factor classification model) by machine learning using the measurement database 210. The diagnosis model generator 110 stores the generated diagnosis model in the diagnosis model database 220. Labels such as a label indicating the normal or anomalous state and a label indicating an anomaly factor are applied to measured data in the measurement database 210 in advance in accordance with a learning method for an anomaly detection model or a factor classification model.”, the measurement data indicative of a normal state (as labeled ‘normal’ or ‘anomalous’) is saved in a measurement database.
Maruchi, paragraph 162, “Learning data can be continuously added to repeat diagnosis model regeneration and verification by repeatedly executing the model update processing at steps S501 to S503 on a diagnosis model for an identical diagnosis target system. [0163] An improved learning data stored in the measurement database 210 and a diagnosis model stored in the diagnosis model database 220 can be transmitted from the diagnostic device 100 to another diagnostic device.”, updated learning data can be continually added to the measurement database, it is interpreted by the examiner that this continuous updating of the learning data is analogous to ‘new data sets… that represent an updated normal state’ as disclosed in the claim language.)
and wherein the larger sub-group is extracted from among the new data sets that represent the updated normal state of the power-operated machine, to exclude the old data sets from the extracted larger sub-group. (Maruchi, paragraph 61, “For example, when anomaly of a diagnosis target system is detected based on certain sample data, large deviation of a trend graph in the sample data, which corresponds to a case in which Sensor B is on, from normal sample data (refer to FIG. 3 or FIG. 8 to be described later) can be information effective for determining this sample data to be anomalous data.”, learning data is indicated by deviations of a trend graph of the measured data. This indicates that learning data extracted to historical databases occurs after a period of normal activity.)
Claim 18-19 recites limitations substantially similar to claim 1, therefore the rejection of claim 1 similarly applies to claim 18-19.
Claim 19 teaches the following additional limitations for consideration:
A non-transitory computer readable storage device storing processor-executable instructions to a cause a processor to (Maruchi, claim 12, “A non-transitory computer readable medium having a computer program stored therein which causes a computer to perform processes”)
Claim 2: Maruchi, Iwamura, and Tong teaches claim 19. Maruchi further teaches the following:
wherein the processor is further configured to acquire a plurality of normal data sets including the new data set, that represents the normal state of operation of the power-operated machine (Maruchi, paragraph 98, “When learning data used to generate an anomaly detection model does not include or hardly includes anomalous data and thus a large number of pieces of sample data are assumed to be normal data, unsupervised learning may be performed. The unsupervised learning is performed to generate a model indicating the normal state of a diagnosis target system, and anomaly detection is performed by using the model.”, the learning data used to generate anomaly detection aligns with the normal data sets of the claim language because these learning datasets predominantly consist of normal operational data.)
wherein the source data is updated in the storage device, by adding the normal data sets to the source data. (Maruchi paragraph 99, “It is supposed that sample data acquired from a system of social infrastructure such as power generation station, plant, or railway mostly belongs to normal data and includes almost no anomalous data. Thus, measured data is used intact as learning data to generate a model representing the normal state of a diagnosis target system by unsupervised learning, and the model is used as an anomaly detection model.”, measured data used as input is almost entirely normal data, indicating a normal representation of the system.)
Claim 10: Maruchi, Iwamura, and Tong teaches claim 19 Maruchi further teaches the following:
wherein the extraction condition is additionally determined based on condition setting information that corresponds to a time period when the diagnosis target data is acquired. (Maruchi, paragraph 155, “The model update processing may be executed at a timing determined by a method different from that in the present process. For example, the model update processing may be performed in a period such as once per week. The model update processing may be executed each time when evaluation of a graph and a diagnosis result by the maintenance person is performed a certain number of times. The model update processing may be performed at a timing other than those described above.”,
Maruchi, paragraph 39, “Measured data includes a measurement time and a sensor value.”;
Maruchi, paragraph 46, “The sample data may be measured data at the measurement time of the diagnosis or measured data in a certain period including a plurality of measurement times.”, Maruchi expressly teaches that measured data includes “measurement time,” and further teaches that the diagnosis/sample data may be taken at the diagnosis time or over a “certain period” including multiple measurement times. Therefore, the “condition setting information” corresponding to a time period when diagnosis target data is acquired is taught by Maruchi’s measurement-time/time-period teaching, and the extraction condition may be set (additionally) based on that time period (e.g., selecting data associated with a relevant time interval).)
Claim 21: Maruchi, Iwamura, and Tong teaches claim 19. Maruchi further teaches the following:
The computer-readable storage device according to claim 19, wherein the plurality of data sets of the source data includes old data sets having old physical parameter values that represent a previous normal state of operation of the power-operated machine, (Maruchi paragraph 99, “It is supposed that sample data acquired from a system of social infrastructure such as power generation station, plant, or railway mostly belongs to normal data and includes almost no anomalous data. Thus, measured data is used intact as learning data to generate a model representing the normal state of a diagnosis target system by unsupervised learning, and the model is used as an anomaly detection model.”, measured data used as input is almost entirely normal data, indicating a normal representation of the system.)
and new data sets having new physical parameter values that represent an updated normal state of operation of the power-operated machine, (Maruchi, paragraph 134, “At step S101, the diagnosis model generator 110 generates a diagnosis model (anomaly detection model or factor classification model) by machine learning using the measurement database 210. The diagnosis model generator 110 stores the generated diagnosis model in the diagnosis model database 220. Labels such as a label indicating the normal or anomalous state and a label indicating an anomaly factor are applied to measured data in the measurement database 210 in advance in accordance with a learning method for an anomaly detection model or a factor classification model.”, the measurement data indicative of a normal state (as labeled ‘normal’ or ‘anomalous’) is saved in a measurement database.
Maruchi, paragraph 162, “Learning data can be continuously added to repeat diagnosis model regeneration and verification by repeatedly executing the model update processing at steps S501 to S503 on a diagnosis model for an identical diagnosis target system. [0163] An improved learning data stored in the measurement database 210 and a diagnosis model stored in the diagnosis model database 220 can be transmitted from the diagnostic device 100 to another diagnostic device.”, updated learning data can be continually added to the measurement database, it is interpreted by the examiner that this continuous updating of the learning data is analogous to ‘new data sets… that represent an updated normal state’ as disclosed in the claim language.)
and wherein the larger sub-group is extracted from among the new data sets that represent the updated normal state of the power-operated machine, to exclude the old data sets from the extracted larger sub-group. (Maruchi, paragraph 97, “The diagnosis model generator 110 creates an anomaly detection model and a factor classification model when the diagnostic device 100 starts or when a new system or device is added as a diagnosis target. In addition, a model specialized for a case in which the existing system is placed under a particular condition may be generated. The particular condition means an installation environment, an operation mode, a load situation, and the like of a system or device. To generate a model specialized for a case in which the existing system is placed under a particular condition, only measured data when the existing system is placed under the particular condition may be extracted from among measured data acquired from the existing system and stored in the measurement database 210, and may be used as learning data”, in cases where specialized conditions are present, the model may only use new learning data obtained from the devices in that case rather than from older learning data. It is interpreted that it would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to extract the larger sub-group from the Tong prior art reference using only this new data as similarly disclosed in Maruchi.)
Claim 23: Maruchi, Iwamura, and Tong teaches claim 19. Maruchi further teaches the following:
The computer-readable storage device according to claim 19 wherein the distance for determining the diagnosis result, is taken from a point converted from the diagnosis target data, within the unit space, (Maruchi, paragraph 69, “In a case of a trend graph or a scatter diagram, the visual determination depends on whether determination of whether sample data as a diagnosis target is an outlier is possible based on a calculated distance of the graph from a graph or a point of other sample data. The distance may be an existing distance scale such as Manhattan distance, Euclidean distance, or DTW.”)
Claim 24: Maruchi, Iwamura, and Tong teaches claim 18. Maruchi further teaches the following:
The fault diagnosis method according to claim 18 wherein the diagnosis target data includes a plurality of data pieces acquired at different times within a given time period, (Maruchi, paragraph 46, “The sample data may be measured data at the measurement time of the diagnosis or measured data in a certain period including a plurality of measurement times.”)
wherein the method further comprises: acquiring a plurality of normal data sets including the new data set, that represent the normal state of operation of the power-operated machine, (Maruchi, paragraph 98, “When learning data used to generate an anomaly detection model does not include or hardly includes anomalous data and thus a large number of pieces of sample data are assumed to be normal data, unsupervised learning may be performed. The unsupervised learning is performed to generate a model indicating the normal state of a diagnosis target system, and anomaly detection is performed by using the model.”, the learning data used to generate anomaly detection aligns with the normal data sets of the claim language because these learning datasets predominantly consist of normal operational data.)
calculating respective distances associated with the data pieces of the diagnosis target data based on the learning information, (Maruchi, paragraph 122, “An anomaly degree, a threshold, and an anomaly probability are displayed in addition to the anomaly determination result. The anomaly degree is an index indicating the degree of deviation of sample data determined to be anomalous from normal data.”, an anomaly degree is analogous to the fault level scores of the diagnosis data based on the learning information.)
and wherein the source data is updated in the storage device, by adding the normal data sets to the source data, when the each of the respective distances is greater than a diagnosis threshold for a predetermined number of data pieces acquired consecutively. (Maruchi, paragraph 155, “The following describes, in detail, the model update processing at step S105 illustrated in FIG. 16. The model update processing is executed at each evaluation of a graph and a diagnosis result by the maintenance person in the process illustrated in FIG. 16, but does not necessarily need to be executed at each evaluation. The execution frequency of the model update processing can be determined with taken into account cost of a time, a calculation resource, and the like taken for the model update processing”, model update processing (which is interpreted to include updating the learning data of normal data sets) is done so at each step of the flow chart of the entire diagnosis process of figure 16. One step, S103, includes diagnosis result of being anomalous (or normal) and it is interpreted by the examiner that this updating step upon reaching an anomalous diagnosis at step S103, is analogous to updating the normal data sets to the source data (databases) when the fault level scores (the anomaly degree as disclosed above) is greater than a diagnosis threshold (detection of an anomaly). )
Claim 25: Maruchi, Iwamura, and Tong teaches claim 18, Maruchi further teaches the following:
The fault diagnosis method according to claim 18 further comprising: wherein the plurality of data sets of the source data includes old data sets having old physical parameter values that represent a previous normal state of operation of the power-operated machine, (Maruchi, paragraph 98, “When learning data used to generate an anomaly detection model does not include or hardly includes anomalous data and thus a large number of pieces of sample data are assumed to be normal data, unsupervised learning may be performed. The unsupervised learning is performed to generate a model indicating the normal state of a diagnosis target system, and anomaly detection is performed by using the model.”, the learning data used to generate anomaly detection aligns with the normal data sets of the claim language because these learning datasets predominantly consist of normal operational data)
and new data sets having new physical parameter values that represent an updated normal state of operation of the power-operated machine, (Maruchi, paragraph 134, “At step S101, the diagnosis model generator 110 generates a diagnosis model (anomaly detection model or factor classification model) by machine learning using the measurement database 210. The diagnosis model generator 110 stores the generated diagnosis model in the diagnosis model database 220. Labels such as a label indicating the normal or anomalous state and a label indicating an anomaly factor are applied to measured data in the measurement database 210 in advance in accordance with a learning method for an anomaly detection model or a factor classification model.”, the measurement data indicative of a normal state (as labeled ‘normal’ or ‘anomalous’) is saved in a measurement database.
Maruchi, paragraph 162, “Learning data can be continuously added to repeat diagnosis model regeneration and verification by repeatedly executing the model update processing at steps S501 to S503 on a diagnosis model for an identical diagnosis target system. [0163] An improved learning data stored in the measurement database 210 and a diagnosis model stored in the diagnosis model database 220 can be transmitted from the diagnostic device 100 to another diagnostic device.”, updated learning data can be continually added to the measurement database, it is interpreted by the examiner that this continuous updating of the learning data is analogous to ‘new data sets… that represent an updated normal state’ as disclosed in the claim language.)
and wherein the larger sub-group is extracted from among the new data sets that represent the updated normal state of the power-operated machine, to exclude the old data sets from the extracted larger sub-group. (Maruchi, paragraph 97, “The diagnosis model generator 110 creates an anomaly detection model and a factor classification model when the diagnostic device 100 starts or when a new system or device is added as a diagnosis target. In addition, a model specialized for a case in which the existing system is placed under a particular condition may be generated. The particular condition means an installation environment, an operation mode, a load situation, and the like of a system or device. To generate a model specialized for a case in which the existing system is placed under a particular condition, only measured data when the existing system is placed under the particular condition may be extracted from among measured data acquired from the existing system and stored in the measurement database 210, and may be used as learning data”, in cases where specialized conditions are present, the model may only use new learning data obtained from the devices in that case rather than from older learning data.)
Claim 30: Maruchi, Iwamura, and Tong teaches claim 19. Maruchi further teaches:
The computer-readable storage device according to claim 19, wherein the selected measured physical parameter value corresponds to a temperature value associated with the current operation of the power-operated machine. (Maruchi, paragraph 39, “Measured data includes a measurement time and a sensor value. FIG. 2 illustrates exemplary measured data. The exemplary data illustrated in FIG. 2 is measured by four sensors. Examples of the kind of measured data include a temperature”, The measured physical parameters gathered from the sensors comprise temperature associated with the current operation of the power-operated machine.)
Claims 7, and 22 are rejected under 35 U.S.C. 103 as being unpatentable over Maruchi in view of Iwamura, Tong and Gupta et al. (US 20190294524 A1), hereinafter referred to as Gupta.
Claim 7: Maruchi, Iwamura, and Tong teaches claim 1, Iwamura further teaches:
wherein the sub-group that is extracted corresponds to first data sets among the plurality of data sets, that satisfy the extraction condition (Iwamura, page 6, paragraph 2, “the search range determination unit sets an area having a representative point within the range of the search radius R around the query as the search area, and searches until the number of points included in the search area reaches a predetermined number”, Iwamura’s “search area” defined by the radius around the query is the claimed “extraction condition.” The “points included in the search area” correspond to the first/extracted data sets that satisfy that extraction condition (i.e., those whose stored values fall within the defined range around the selected/query value).)
Gupta, in the same field of anomaly detection/inference, further teaches the following which Maruchi, Iwamura, and Tong fail to teach:
wherein the first data sets are selected based on a sorting criterion of the plurality of data sets, that indicates an order of temporal proximity to a time when the diagnosis target data is acquired. (Gupta, paragraph 22, “The anomaly detector reads in time-series data (i.e., measurements of application related metrics) collected and/or generated from monitoring agents. The anomaly detector accumulates the time-series data across a series of time instants to form a multivariate time-series data slice or multivariate data slice.”, learning data measured in Gupta is accumulated time-series data, time-series data is inherently sorted based on time and is considered a sorting criterion.)
Maruchi teaches a system that uses learning data from a certain period of time without any indication of an order or use of a sorting algorithm. Gupta does teach sorting its learning data by utilizing multivariate time-series data, indicating that the data is sorted from the time it was received by the system. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the teachings disclosed by Maruchi, Iwamura, and Tong with the teachings disclosed by Gupta (i.e. sorting the learning data temporally). A motivation for the combination is to have a method for continuously scanning and updating using the newest data or having time any similar time based criteria, (Gupta, paragraph 58, “The scanner can apply various unknown anomaly labeling criteria, such as a time period threshold criterion, an immediate change criterion, and a new cluster criteria... The time period threshold is likely a multiple of the slice width. Applying the immediate change criterion is similar to applying the time period threshold criterion, except that the scanner immediately labels the current point as an unknown anomaly if the scanner has not detected any canonical behaviors or known anomalies without considering preceding multivariate data slice derived points.”).
Claim 22: Maruchi, Iwamura, and Tong teaches claim 19. Maruchi further teaches the following:
The computer-readable storage device according to claim 19, wherein the processor-executable instructions are further configured to cause the processor to: evaluate the learning information as reliable when a ratio of a number of the extracted data of the larger sub-group relative to a number of the plurality of data sets in the source data, is equal to or greater than a lower limit value, (Gupta, paragraph 92, “If the current path is not a mature path, the path determines whether the current path meets a set of reliability criteria (1270). The set of reliability criteria are a set of one or more criterion that determines whether a path is sufficiently reliable during a path analysis. The reliability criteria can include criterion such as: that a minimum number of sets of points are used to generate the current path… If the current path does not meet the reliability criteria, the scanner does not change the status of the current path. Otherwise, the scanner establishes the current path as a mature path in the path library (1274). Once the current path is established as a mature path, the newly-mature path can be used during a path analysis.”, if a current path formed by sample points is not reliable (does not exceed a certain minimum amount of points) then is continually updated until it is mature (ready for analysis by the learning system).)
wherein the diagnosis result is determined in response to evaluating that the learning information is reliable. (Gupta, paragraph 92, “Once the current path is established as a mature path, the newly-mature path can be used during a path analysis.”, once a reliable (mature) path is formed (e.g. a path that has a minimum amount of points to generate the path) it is placed under path analysis for a diagnosis result.)
Claim 9 is rejected under 35 U.S.C. 103 as being unpatentable over Maruchi in view of Iwamura, Tong and Rockwell, (Rockwell Automation. (2010, March). FactoryTalk Historian ME VERSION 2.1 USER’S GUIDE.), hereafter referred to as Rockwell.
Claim 9: Maruchi, Iwamura, and Tong teaches claim 2. Maruchi further teaches the following:
wherein the stored physical parameter values are associated with a plurality of items of the power-operated machine, (Maruchi, Paragraph 39, “The measured data 1 includes data measured by one or a plurality of sensors. Measured data includes a measurement time and a sensor value. FIG. 2 illustrates exemplary measured data. The exemplary data illustrated in FIG. 2 is measured by four sensors. Examples of the kind of measured data include a temperature, a flow rate, a current, a voltage, a pressure, an operation command by a person, the position of an object in operation, a surrounding environment, and any other kind.”, kinds of measured data include parameter values of the system including temperature. Paragraph 22 of this application’s specification states that, “Examples of such parameter values include temperatures around the machine system”)
Rockwell, in the same field of data extraction techniques, teaches the following limitations which the above fails to teach:
and wherein the extraction condition is determined additionally based on condition setting information that identifies one or more items of the plurality of items, (Rockwell, page 81, paragraph 3, “Instrument Tag - indicates the controller from which the data is coming. If you replace a controller with a different one that measures the same process value, it is usually best to continue using the same point. Edit the point as required so that it will collect the new data. The format of the instrument tag name is: [<controllerprojectname>_1_<controllerslotnumber>][<controllertagname>]”, The item identifier (InstrumentTag) uniquely identifies the parameter (“item”).)
and that indicates respective physical parameter ranges of the stored physical parameter values relating to the one or more items, wherein the extraced sets are associated with the one or more items of the power-operated system and satisfy the respective physical parameter ranges of the stored physical parameter values. (Rockwell, page 78, paragraph 2, “Point attributes tell FactoryTalk Historian ME how and when the server should collect data from a particular data source. Point attributes specify information such as the data source location, how often the servers should get new values from the data source, which values the server can ignore, and which represent valid data.”;
Rockwell, page 80, “The General section contains the following fields:… Name - provides a label for the point. Follow these conventions when naming a point:… The name must be unique.”;
Rockwell, page 103, “The deadband when exceeded causes an exception. This is configured for each point in either the ExcDev or ExcDevPercent attribute.”;
Rockwell, page 215, “Compression Deviation (CompDev attribute) Specifies in engineering units how much a value may differ from the previous value before it is considered to be a significant value…For typical flows, pressures, and levels, a typical deviation specification of 1% or 2% of Span is used. For temperatures, the typical deviation is usually 1 or 2 degrees.”, Rockwell teaches that data collection is performed for specific “points/tags,” where each point has a unique name (corresponding to a unique item/parameter of the system) and has configured attributes that control which values are accepted/archived. Rockwell further teaches that deviation attributes (e.g., ExcDev/CompDev) are configured per point and define numeric bands in engineering units (i.e., physical parameter ranges) that determine whether values are treated as significant/kept. Therefore, the extracted/kept data sets are “associated with” the particular item (because they belong to the tag/point for that item) and satisfy the respective configured ranges for that item.)
Maruchi teaches a system that uses learning data from sensors at a certain period of time without any indication of an acceptable range to extract. Rockwell on the other hand teaches data extraction techniques that define numeric ranges for each sensor for purposes of extracting within those ranges. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the teachings disclosed by Maruchi, Iwamura, and Tong with the teachings disclosed by Rockwell (i.e. user-defined per-item ranges for sensor data extraction). A motivation for the combination is to have per-sensor range based extraction so that only user-defined data deemed as significant is passed through for each point (sensor), (Rockwell, page 85, paragraph 1, “COMPRESSION FILTERING POINT PROPERTIES When a new Snapshot arrives, the previous one is evaluated according to the compression specifications to see if it is a significant event. If so, it is sent to the Event Queue. If not, it is discarded. The result is that only significant data is written to the archive. This process is called compression.”).
Claim 13 is rejected under 35 U.S.C. 103 as being unpatentable over Maruchi in view of Iwamura, Tong and Rockwell, (Rockwell Automation. (2012, June). FactoryTalk Historian DataLink User Guide.), hereafter referred to as FactoryTalk.
Claim 13: Maruchi, Iwamura, and Tong teaches claim 1. FactoryTalk, in the same field of data extraction, further teaches the following:
evaluate the learning information as reliable when a ratio of a number of extracted data sets of the larger sub-group relative to a number of the plurality of data sets in the source data, is equal to or greater than a lower limit value, wherein the diagnosis result is determined in response to evaluating that the learning information is reliable; and generate an alert when the learning information is evaluated as not reliable. (FactoryTalk, page 46, paragraph 1, “Minimum Percent Good: Specify the minimum percentage of good data (page 25) required in each time range to calculate and return a value. Insufficient good data is substituted as a placeholder when a value is not returned… Percent Good: […] The percentage of good values helps in assessing the reliability of calculations built on FactoryTalk Historian point values, particularly if calculated values are to be used in further calculations.”, This is a ratio-based reliability gate (Percent Good / Minimum Percent Good). FactoryTalk expressly teaches setting a “Minimum Percent Good” required in each time range to calculate/return a value, and teaches that when there is “Insufficient good data,” a placeholder is substituted when a value is not returned. Therefore, when Percent Good is below the minimum, the system explicitly indicates insufficient/poor data quality (i.e., an alert/flag of unreliability) and the value is not produced/returned for downstream use, which corresponds to evaluating the learning information as not reliable and generating an alert/indication of unreliability.)
FactoryTalk discloses methods that assess data reliability within a time range interval and determines a ratio (percentage) of acceptable data to intake from the set. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the teachings disclosed by Maruchi, Iwamura, and Tong with the teachings disclosed by FactoryTalk (i.e. data reliability method). A motivation for the combination is to obtain higher quality data through a threshold reliability measure.
Claims 26 are rejected under 35 U.S.C. 103 as being unpatentable over Maruchi in view of Iwamura, Tong, Matboua et al., (Matboua, F., Alahyarib, H., & Namvarc, F. (2016). Anti-icing system at gas turbine compressor bell-mouth and Inlet Guide Vane (IGV). International Academic Journal of Science and Engineering, 3(6), 74-82.), hereafter referred to as Matboua, and Rockwell.
Claim 26: Maruchi, Iwamura, and Tong teaches claim 19.
Matboua, in the same field of data extraction, further teaches the following which the above fails to teach:
The computer-readable storage device according to claim 19 wherein the power-operated machine is selected from the group consisting of: a gas turbine, an aero-engine, and a vacuum furnace (Matboua, page 178, paragraph 1, “First Referring to Figure 2 in MGT 70 (V94.2) gas turbine in base load (IGV is fully open) Ice can build up in IGV surface if ambient air temperature is below 2.98 °C, but corresponding relative humidity for ice formation varies with respect to ambient air temperature . This variation is very slightly meaning that for wide range of ambient air temperature it can be consider constant. For this reason in base load operation and sea level for upper limit of anti-icing system it can be selected +3 °C for temperature and 80% for relative humidity.”, Matboua expressly discusses a gas turbine (“MGT 70 … gas turbine”), which is one of the listed options. Therefore, the limitation is met when the selected power-operated machine is a gas turbine.)
and wherein the selected measured physical parameter value is selected to correspond to the intake-air temperature identified by the item identifier acquired. (Matboua, page 178, paragraph 1, “First Referring to Figure 2 in MGT 70 (V94.2) gas turbine in base load (IGV is fully open) Ice can build up in IGV surface if ambient air temperature is below 2.98 °C, but corresponding relative humidity for ice formation varies with respect to ambient air temperature . This variation is very slightly meaning that for wide range of ambient air temperature it can be consider constant.”, In the gas-turbine operating context, Matboua uses ambient air temperature as the relevant inlet condition for icing. That inlet/ambient temperature corresponds to the intake air temperature condition of the gas turbine, i.e., the temperature of air entering the compressor/inlet region.)
It would have been obvious to apply Rockwell’s per-tag identifier and per-tag delta configuration to the gas turbine inlet temperature parameter of Matboua and to the prior art combination of Maruchi, Iwamura, and Tong, to reliably extract/compare records for diagnosis when too few records match a narrower range.
Rockwell, in the same field of data extraction techniques, teaches the following limitations which the above fails to teach:
wherein the processor is further caused to acquire an item identifier that uniquely identifies an… temperature of the power-operated machine, (
Rockwell, page 80, “Name... The name must be unique.”;
Rockwell, page 81, paragraph 3, “Instrument Tag - indicates the controller from which the data is coming. If you replace a controller with a different one that measures the same process value, it is usually best to continue using the same point. Edit the point as required so that it will collect the new data. The format of the instrument tag name is: [<controllerprojectname>_1_<controllerslotnumber>][<controllertagname>]”;
Rockwell, page 215, “For temperatures, the typical deviation is usually 1 or 2 degrees.”,
Rockwell teaches that each point/tag has a unique name (unique identifier) and an Instrument Tag that identifies the source/controller/tag for the process value. Because Rockwell’s points/tags represent specific measured process variables, configuring a point/tag for a temperature measurement (e.g., intake-air temperature) results in the acquired point identifier (unique point name/instrument tag) uniquely identifying that temperature parameter. Rockwell further confirms temperature is a supported parameter type by expressly discussing typical deviation settings for temperatures.)
and wherein the first delta value and the second delta value are set based on the item identifier acquired that is associated with the selected measured physical parameter value. (Rockwell, page 215, paragraph 3, “Compression Deviation (CompDev attribute) Specifies in engineering units how much a value may differ from the previous value before it is considered to be a significant value. As a rule of thumb, set CompDev to the precision of the data source or hardware (instrument). Set it a little "loose" to err on the side of collecting, rather than losing data. After collecting data for a while, go back and check the data for your most important tags and adjust CompDev if necessary.”,
Rockwell, page 104, last paragraph, “The Exception Deviation (ExcDev) attribute specifies in engineering units how much a value may differ from the previous value before it is considered to be a significant value. The exception deviation should be less than the compression deviation by at least a factor of 2.”, The claim requires delta values that are set “based on the item identifier,” i.e., parameter/tag-specific settings. Rockwell teaches that deviation parameters (e.g., ExcDev and CompDev) are configured and tuned per tag (“your most important tags”), which makes the delta values depend on the identifier of the parameter. Rockwell also explicitly teaches a relationship between two deviation values (ExcDev < CompDev), supporting a smaller delta and a larger delta tied to the same tag/identifier.)
Iwamura further teaches:
wherein the first physical parameter range has a minimum value corresponding to the selected measured physical parameter value subtracted by a first delta value, and a maximum value corresponding to the selected measured physical parameter value increased by the first delta value, (Iwamura, page 6, paragraph 2, “the search range determination unit sets an area having a representative point within the range of the search radius R around the query as the search area…”, A symmetric radius R around the selected value is the ±δ first range.)
wherein the second physical parameter range has a minimum value corresponding to the selected measured physical parameter value subtracted by a second delta value that is greater than the first delta value, and a maximum value corresponding to the selected measured physical parameter value increased by the second delta value, (Iwamura, page 6, paragraph 2, “…and searches until the number of points included in the search area reaches a predetermined number. The radius R may be gradually increased.”, Iwamura increases the radius (δ₂>δ₁) to widen the extraction window.)
Claims 29 are rejected under 35 U.S.C. 103 as being unpatentable over Maruchi in view of Iwamura, Tong, and Matboua.
Claim 29: Maruchi, Iwamura, and Tong teaches claim 1.
Matboua, in the same field of data extraction, further teaches the following which the above fails to teach:
The system according to claim 1, wherein the power-operated machine is a thermomechanical machine, (Matboua, page 178, paragraph 1, “First Referring to Figure 2 in MGT 70 (V94.2) gas turbine in base load (IGV is fully open) Ice can build up in IGV surface if ambient air temperature is below 2.98 °C, but corresponding relative humidity for ice formation varies with respect to ambient air temperature . This variation is very slightly meaning that for wide range of ambient air temperature it can be consider constant. For this reason in base load operation and sea level for upper limit of anti-icing system it can be selected +3 °C for temperature and 80% for relative humidity.”, A gas turbine is a thermomechanical machine (thermal energy conversion to mechanical shaft work).)
and wherein the selected measured physical parameter value corresponds to a temperature value associated with the current operation of the thermomechanical machine. (Matboua, page 178, paragraph 1, “First Referring to Figure 2 in MGT 70 (V94.2) gas turbine in base load (IGV is fully open) Ice can build up in IGV surface if ambient air temperature is below 2.98 °C, but corresponding relative humidity for ice formation varies with respect to ambient air temperature . This variation is very slightly meaning that for wide range of ambient air temperature it can be consider constant.”, Matboua uses temperature (ambient/inlet air temperature) as an operational condition used during operation (anti-icing logic tied to operating state). This teaches selecting a measured parameter value that corresponds to a temperature value associated with current operation.)
Matboua discloses methods that have tested real-world optimal temperature threshold values for their data extraction systems. It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to have incorporated the teachings disclosed by Maruchi, Iwamura, and Tong with the teachings disclosed by Matboua (i.e. a gas turbine with specific temperature thresholds). A motivation for the combination is to gain knowledge of real-world parameter-specific temperature values that are optimal for gas turbines.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
US20120316835A1 — “Anomaly detection method and anomaly detection system”
US7043401B2 — “Multivariate data analysis method and uses thereof”
US20120041575A1 — “Anomaly Detection Method and Anomaly Detection System”
US20130024166A1 — “Monitoring System Using Kernel Regression Modeling with Pattern Sequences”
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HYUNGJUN B YI whose telephone number is (703)756-4799. The examiner can normally be reached M-F 9-5.
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, Usmaan Saeed can be reached on (571) 272-4046. 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.
/H.B.Y./Examiner, Art Unit 2124
/USMAAN SAEED/Supervisory Patent Examiner, Art Unit 2146
Prosecution Insights