Time series data processing methods

The time-series data processing method addresses the challenge of decreased accuracy in abnormality detection by aligning feature data with operating states, using autoencoders to enhance detection precision in industrial equipment.

JP7593476B2Active Publication Date: 2024-12-03NEC CORP
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2023506702
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-03-19
Publication Date
2024-12-03
Estimated Expiration
2041-03-19

AI Technical Summary

Technical Problem

Existing methods for abnormality detection in industrial equipment operating in multiple states suffer from decreased accuracy due to varying operating conditions, leading to false positives.

Method used

A time-series data processing method that associates measured data with operating state information, converts it into feature data, and corrects it based on time, using autoencoders to enhance accuracy by aligning feature data with similar operating conditions.

Benefits of technology

Improves the accuracy of abnormality detection by clustering feature data based on time and operating states, reducing false positives and enhancing detection precision.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007593476000001
    Figure 0007593476000001
  • Figure 0007593476000002
    Figure 0007593476000002
  • Figure 0007593476000003
    Figure 0007593476000003
Patent Text Reader

Abstract

A time series data processing device (100) according to the present invention is provided with: a database (121) in which time series data, measured from a measurement object, and operating state information, expressing an operating state of the measurement object when the time series data was measured, are associated; a conversion unit (122) which respectively converts pieces of the time series data into feature quantity data at each prescribed time period and respectively converts pieces of the feature quantity data into corrected feature quantity data that has been corrected on the basis of the time of the corresponding time period; and an extraction unit (123) which, on the basis of the operating state information associated with the time series data, extracts the corrected feature quantity data corresponding to the time series data.
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a time series data processing method, a time series data processing device, and a program. [Background technology]

[0002] In facilities such as industrial plants that manufacture energy (electricity, gas, clean water, etc.), chemical products (crude oil, gasoline, plastics, etc.), metal products (iron, semiconductors, etc.), mechanical products (automobiles, computers, etc.), food, medicine, etc., and information processing systems, time-series data that is measurement values ​​from various sensors is analyzed, and the occurrence of an abnormal state is detected and output. For example, in Patent Document 1, operation data that is detection values ​​from sensors installed in equipment on a production line in a factory, etc. is acquired, and an anomaly detection model is learned from the operation data under normal conditions. Then, the anomaly detection model is used to calculate an anomaly score for operation data newly acquired from the equipment thereafter, and the user is notified, thereby monitoring the abnormal state. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-008997 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, measurement targets such as plants and equipment may operate in multiple operating states, making it difficult to detect anomalies while taking into account the operating states, and this may result in a problem of reduced accuracy in anomaly detection. For example, in the case of measurement targets that operate in operating states that vary depending on the time of day or season, even if data acquired during monitoring is detected as abnormal in the actual operating state, it may be determined to be normal in another operating state, resulting in a false positive. As a result, a problem occurs in which the accuracy of anomaly detection for measurement targets that operate in multiple operating states is reduced.

[0005] Therefore, an object of the present invention is to provide a time series data processing method that can solve the above-mentioned problem of a decrease in anomaly detection accuracy for measurement objects operating in multiple operating states. [Means for solving the problem]

[0006] A time series data processing method according to one aspect of the present invention includes the steps of: Associating time series data measured from a measurement object with operating state information that represents an operating state of the measurement object at the time the time series data was measured; converting the time series data into feature amount data for each predetermined period, and converting the feature amount data into corrected feature amount data that is corrected based on the time of the corresponding period; extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data; The structure is as follows.

[0007] Moreover, a time series data processing device according to one aspect of the present invention comprises: a database in which time series data measured from a measurement object is associated with operational status information representing the operational status of the measurement object when the time series data was measured; a conversion unit that converts the time series data into feature amount data for each predetermined period, and converts the feature amount data into corrected feature amount data that is corrected based on a time of the corresponding period; an extraction unit that extracts the corrected feature amount data corresponding to the time-series data based on the operational state information associated with the time-series data; Equipped with The structure is as follows.

[0008] In addition, a program according to one aspect of the present invention includes: An information processing device that can access a database in which time series data measured from a measurement object and operational status information representing the operational status of the measurement object at the time the time series data was measured are associated with each other; converting the time series data into feature amount data for each predetermined period, and converting the feature amount data into corrected feature amount data that is corrected based on the time of the corresponding period; extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data; Execute the process, The structure is as follows. Effect of the Invention

[0009] By being configured as described above, the present invention can improve the accuracy of anomaly detection for a measurement target operating in a plurality of operating states. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of a time-series data processing device according to a first embodiment of the present invention. [Diagram 2] 2 is a diagram showing how time series data is processed by the time series data processing device disclosed in FIG. 1. [Diagram 3] 2 is a diagram showing how time series data is processed by the time series data processing device disclosed in FIG. 1. [Figure 4] 2 is a diagram showing how time series data is processed by the time series data processing device disclosed in FIG. 1. [Diagram 5] 2 is a diagram showing how time series data is processed by the time series data processing device disclosed in FIG. 1. [Figure 6] 2 is a diagram showing how time series data is processed by the time series data processing device disclosed in FIG. 1. [Figure 7] 2 is a diagram showing how time series data is processed by the time series data processing device disclosed in FIG. 1. [Figure 8] 2 is a flowchart showing the operation of the time-series data processing device disclosed in FIG. 1. [Figure 9] 2 is a flowchart showing the operation of the time-series data processing device disclosed in FIG. 1. [Figure 10] FIG. 11 is a block diagram showing a hardware configuration of a time-series data processing device according to a second embodiment of the present invention. [Figure 11] FIG. 11 is a block diagram showing a configuration of a time-series data processing device according to a second embodiment of the present invention. [Figure 12] 10 is a flowchart showing the operation of a time series data processing device according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] <Embodiment 1> A first embodiment of the present invention will be described with reference to Fig. 1 to Fig. 9. Fig. 1 is a diagram for explaining the configuration of a time series data processing device, and Fig. 2 to Fig. 9 are diagrams for explaining the processing operation of the time series data processing device.

[0012] [composition] The time series data processing device 10 in the present invention is connected to a measurement target P such as a plant. The time series data processing device 10 acquires and analyzes the measured values ​​of at least one or more data items of the measurement target P, and monitors the state of the measurement target P based on the analysis result. For example, the measurement target P is a plant such as a manufacturing factory or a processing facility, and the measured values ​​of each data item are composed of values ​​of multiple types of data items such as temperature, pressure, flow rate, power consumption value, and raw material supply amount and remaining amount in the plant. Here, in this embodiment, the state of the measurement target P monitored by the time series data processing device 10 is assumed to be an abnormal state of the measurement target P. For this reason, the time series data processing device 10 converts the measured values ​​composed of multiple data items into feature quantities, and further detects the abnormal state based on the degree of abnormality calculated from the feature quantities. However, the time series data processing device 10 in this embodiment does not necessarily need to perform a process of detecting that the measurement target P is in an abnormal state, and as described later, it is sufficient to simply perform a process of converting the measured values ​​into feature quantities and extracting the feature quantities as a preprocessing for detecting that the measurement target P is in an abnormal state.

[0013] The measurement target P in the present invention is not limited to a plant, and may be any equipment such as an information processing system. For example, when the measurement target P is an information processing system, the CPU (Central Processing Unit) usage rate, memory usage rate, disk access frequency, number of input / output packets, input / output packet rate, power consumption value, etc. of each information processing device such as a terminal or server constituting the information processing system may be measured as the measurement value of each data item, and the state of the information processing system may be monitored by analyzing such measurement values.

[0014] The time series data processing device 10 is composed of one or more information processing devices each having a calculation device and a storage device. As shown in FIG. 1, the time series data processing device 10 includes an acquisition unit 11, a conversion unit 12, and an extraction unit 13. The functions of the acquisition unit 11, the conversion unit 12, and the extraction unit 13 can be realized by the calculation device executing a program for realizing each function stored in the storage device. The time series data processing device 10 also includes a measurement data storage unit 16 and a feature data storage unit 17. The measurement data storage unit 16 and the feature data storage unit 17 are each composed of a storage device. Each component will be described in detail below.

[0015] The acquisition unit 11 acquires the measurement values ​​of each data item measured by various sensors installed on the measurement target P at a predetermined time interval as time series data, and stores the data in the measurement data storage unit 16. At this time, since there are multiple types of data items to be measured, the acquisition unit 11 acquires a time series data set D, which is a collection of time series data of multiple data items as shown in the line graph of FIG. 2. In the line graph of FIG. 2, the horizontal axis represents time, and each line represents the measurement value of each data item. The acquisition unit 11 constantly acquires and stores the time series data set D, and the acquired time series data set D is used when accumulating feature amount data when the measurement target P is operating in a predetermined operation mode and when monitoring the state of the measurement target P, as described later.

[0016] As described later, the acquisition unit 11 associates the operation mode (operation state information) representing the operation state of the measurement object P when the time series data set D is measured with the model (type information) representing the type of the measurement object P in the measurement data storage unit 16 (database) for the time series data set D used in the process of accumulating feature amount data when the measurement object P is operating in a predetermined operation mode. For example, the operation mode is information representing the time period (morning, afternoon, evening, etc.), month (January, February, etc.), season (spring, summer, autumn, winter, etc.), and operation state after startup (startup, normal operation, etc.). In addition, the model represents the type of the device (model A, B, C, etc.) when there are multiple types of devices that are the measurement object P. The operation mode and type may be, for example, information on the operation mode and model input by the operator when the measurement object P is in operation, acquired by the acquisition unit 11 and stored in association with the time series dataset D, or information on the operation mode and model already set in the measurement object P that is in operation may be acquired by the acquisition unit 11 and stored in association with the time series dataset D.

[0017] The converter 12 converts the time series data set stored in the measurement data storage unit 16 into feature amount data consisting of information representing the features of the time series data set D. At this time, as shown by symbols D1 and D2 in Fig. 2, the converter 12 divides the time series data set D into predetermined time intervals to generate partial time series data sets D1 and D2 consisting of predetermined periods, and converts each of the partial time series data sets D1 and D2 into feature amount data. Note that the multiple partial time series data sets D1 and D2 may have periods that are continuous with no gaps, may overlap each other, or may be separated from each other.

[0018] Specifically, as shown in FIG. 3, the conversion unit 12 converts the partial time series data set into a feature using an autoencoder by unsupervised learning that uses the partial time series data set as input data and aims to output output data that matches the input data. At this time, the autoencoder compresses the dimension of the input data in an intermediate layer located between an input layer that inputs input data and an output layer that outputs output data, and the value of the intermediate layer becomes the input data, that is, the feature data of the partial time series data set. The example of FIG. 3 shows a case where the input layer is composed of four nodes and the intermediate layer is composed of two nodes. For example, as shown in FIG. 4, in the autoencoder, when the partial time series data set Di for the period corresponding to the time "10:40" is composed of four values ​​(a1, a2, a3, a4), it is converted into feature data composed of two values ​​such as (0.82.0.15). Note that the conversion unit 12 uses, for example, an autoencoder generated by machine learning a partial time series data set measured from a measurement target P during normal operation in the past. In other words, the autoencoder uses the partial time series data set measured from the measurement target P during normal operation as input data, and has learned the weighting between each node so as to output matching output data. However, the conversion unit 12 is not necessarily limited to converting the partial time series data set into feature data using the autoencoder, and may convert the partial time series data set into feature data using other methods.

[0019] The converter 12 also has a function of converting the feature data obtained by converting the partial time series data set as described above into corrected feature data, and stores the converted corrected feature data in the feature data storage unit 17. At this time, the converter 12 corrects the feature data based on the time when the partial time series data set before being converted into the feature data was measured, and converts the feature data to the corrected feature data. Specifically, the converter 12 adds time data, which is a value based on the time when the partial time series data set corresponding to the feature data was measured, to the feature data, and converts the feature data to which the time data has been added into the corrected feature data using an autoencoder. At this time, the time data to be added to the feature data is data whose value is closer to the closer time. As an example, as shown in the lower left of FIG. 5, the times when the partial time series data set was measured are arranged on a circumference in order so that the closer the time, the closer the time, and two values ​​such as (sinθ, cosθ) which are trigonometric ratios according to the angle of the time are used. In the example of FIG. 5, for example, the trigonometric ratio value (0.8, 0.2) corresponding to the position on the circumference where the time "10:40" is located is used as the time data Ti. Then, the conversion unit 12 inputs four values ​​(0.82, 0, 15, 0.8, 0.2) obtained by adding the time data Ti (0.8, 0.2) to the feature data (0.82, 0.15) obtained by converting the partial time series data set of the time "10:40" as input data to the autoencoder. In this way, the conversion unit 12 can obtain, as corrected feature data, values ​​of the intermediate layer of the autoencoder that has been machine-learned to output output data that matches the input data. Note that, as the time, values ​​that differ on a yearly basis, such as "month and day", may be used. In other words, the closer the "month and day" is, the closer the time data in this case is, the more approximate the value becomes.

[0020] Here, the above-mentioned time data is a value of (sinθ, cosθ) which is a trigonometric ratio corresponding to the position of the time arranged in order on the circumference, so the closer the time, the more similar the value becomes. Therefore, time data with a value that is similar to each feature data of the partial time series data sets for periods close to each other is added, so the input data input to the autoencoder also have values ​​that are similar to each other. In particular, the partial time series data set measured from the measurement target P during normal operation is expected to have less significant changes as the time approaches, so the feature data is also expected to be similar, and it can be expected that the input data itself to which the time data is added will have a similar value. As a result, the closer the measurement time, the more similar the corrected feature data, which is the value of the intermediate layer of the autoencoder, can be. Note that the conversion unit 12 uses, for example, an autoencoder generated by machine learning using a partial time series data set measured from the measurement target P during normal operation in the past, and a value obtained by adding time data to the feature data converted in the same manner as described above as input data. However, the conversion unit 12 is not necessarily limited to converting the feature data to which time data has been added into corrected feature data using an autoencoder, and may convert into corrected feature data by any method. For example, the above-mentioned time data is just an example, and other data based on measured time may be used, and the feature data may be corrected into corrected feature data based on time by any method.

[0021] Note that, when the measurement target P is monitored, the converter 12 only performs a process of converting a time series data set (second time series data) newly measured from the measurement target P into feature data (second feature data) in the same manner as described above. That is, when the measurement target P is monitored, the converter 12 divides the time series data set D into partial time series data sets D1 and D2 in the same manner as described above, and inputs each of the partial time series data sets D1 and D2 into an autoencoder to convert them into feature data. At this time, the converter 12 does not convert the feature data into corrected feature data by further correcting the feature data.

[0022] The extraction unit 13 extracts and outputs the corrected feature amount data stored in the feature amount data storage unit 17 based on the operation mode and the model associated with the time series data set that is the source of the corrected feature amount data. For example, when the operation mode and the model are specified by an instruction from an operator, the extraction unit 13 refers to the data stored in the measurement data storage unit 16, extracts only the corrected feature amount data obtained by converting the partial time series data set associated with the specified operation mode and the model, and outputs the data to be displayed. For example, the left side of FIG. 6 shows a case where the corrected feature amount data corresponding to each operation mode (summer, winter) and each model (A, B, C) is displayed, but when "summer" is specified as the operation mode, as shown in the right side of FIG. 6, only the corrected feature amount data corresponding to the operation mode "summer" is extracted and displayed. Note that when the operation mode and the model are specified, the extraction unit 13 extracts the corrected feature amount data corresponding to the specified operation mode and the model, and when only the model is specified, the extraction unit 13 extracts the corrected feature amount data corresponding to the specified model.

[0023] The extraction unit 13 also has a function of performing a process of comparing feature data (second feature data) of time series data (second time series data) newly measured from the measurement target P with corrected feature data extracted by specification by an operator during monitoring of the measurement target P. For example, the left side of FIG. 7 shows a case where all corrected feature data and feature data f obtained by converting newly measured time series data are displayed on the same screen for comparison. In this case, the degree of abnormality of the feature data f is calculated as a value based on the distance d1 from the corrected feature data of the closest model "A" and operation mode "winter". However, if the newly measured time series data is measured when the model "A" is operating in the operation mode "summer", it is desirable to calculate the degree of abnormality of the feature data f as a value based on the distance d2 from the corrected feature data of the model "A" and operation mode "summer". Therefore, when the operator designates "summer" as the operation mode, the extraction unit 13 extracts and displays only the corrected feature data corresponding to that operation mode, as shown on the right side of Fig. 7, and can calculate the degree of abnormality of the feature data f based on the distance d2 from the corrected feature data of the model "A" and the operation mode "summer". Note that the extraction unit 13 may automatically calculate and compare the degree of abnormality of new feature data f with the extracted corrected feature data, or may simply display the extracted corrected feature data and feature data f so that they can be compared, as shown on the right side of Fig. 7.

[0024] [Operation] Next, the operation of the above-mentioned time series data processing device 10 will be described mainly with reference to the flowcharts of Figures 8 and 9. First, with reference to the flowchart of Figure 8, the operation when accumulating feature amount data in a case where the measurement object P is operating normally in a predetermined operation mode will be described.

[0025] First, the time series data processing device 10 acquires the measurement values ​​of each data item measured by various sensors installed on the measurement object P at a predetermined time interval as a time series data set, and stores it in the measurement data storage unit 16. At this time, the time series data processing device 10 associates the acquired time series data set with information on the operation mode indicating the operating state of the measurement object P and the model indicating the type of the measurement object P, and stores it in the measurement data storage unit 16 (step S2).

[0026] Next, the time series data processing device 10 converts the time series data set stored in the measurement data storage unit 16 into feature data consisting of information representing the features of the feature data set (step S2). At this time, as shown by symbols D1 and D2 in FIG. 2, the time series data processing device 10 divides the time series data set D into partial time series data sets D1 and D2 consisting of predetermined periods divided by predetermined times, and converts each partial time series data set D1 and D2 into feature data. For example, as shown in FIG. 3, the time series data processing device 10 converts the partial time series data set into feature data using an autoencoder by unsupervised learning that aims to output output data that matches the input data. That is, the time series data processing device 10 inputs the partial time series data set as input data for the autoencoder, and acquires the value of an intermediate layer in which the dimension of the input data is compressed as feature data.

[0027] Next, the time series data processing device 10 converts the feature data obtained by converting the partial time series data set into corrected feature data, and stores the corrected feature data in the feature data storage unit 17 (step S3). At this time, the time series data processing device 10 corrects the feature data based on the time when the partial time series data set before being converted into the feature data was measured, and converts the feature data into corrected feature data. Specifically, the time series data processing device 10 adds time data, whose value is based on the time when the partial time series data set corresponding to the feature data was measured, to the feature data, and converts the feature data to which the time data has been added into corrected feature data using an autoencoder. In particular, as the time data to be added to the feature data, data whose value becomes closer the closer the time is is used. As a result, the feature data of partial time series data sets whose times are close to each other are converted into corrected feature data whose values ​​are closer to each other. The time series data processing device 10 receives feature data with added time data as input data, and acquires the intermediate layer values ​​as corrected feature data using an autoencoder that uses unsupervised learning to output output data that matches the input data.

[0028] Then, the time series data processing device 10 outputs the corrected feature amount data stored in the feature amount data storage unit 17 so as to be extracted and displayed based on the operation mode and model associated with the time series data set on which the corrected feature amount data is based (step S4). For example, when "summer" is specified as the operation mode, the time series data processing device 10 extracts and displays only the corrected feature amount data corresponding to the operation mode, as shown on the right side of Fig. 6. Note that when an operation mode and a model are specified, the time series data processing device 10 extracts the corrected feature amount data corresponding to the specified operation mode and model, and when only a model is specified, the time series data processing device 10 extracts the corrected feature amount data corresponding to the specified model.

[0029] As described above, according to the time series data processing device 10 of the present invention, the feature data of the partial time series data set is further corrected based on time, so that the corrected feature data with close time has a closer value. As a result, the corrected feature data when the operating conditions such as the operation mode and the model are the same can be clustered with higher accuracy.

[0030] Next, the operation when monitoring the operating state of the measurement object P will be described with reference to the flowchart of FIG. 9. First, the time series data processing device 10 acquires the measurement values ​​of each data item measured by various sensors installed in the measurement object P whose operating state is being monitored as a time series data set at a predetermined time interval (step S11). Then, the time series data processing device 10 sequentially divides the time series data set into partial time series data sets D1 and D2 at predetermined time intervals as shown in FIG. 2, and converts each of the partial time series data sets into feature data (step S12). For example, the time series data processing device 10 inputs the partial time series data set to an autoencoder as shown in FIG. 3, as described above, and acquires the value of the intermediate layer as feature data.

[0031] Thereafter, the time series data processing device 10 compares the corrected feature data stored in the feature data storage unit 17 with the feature data obtained by converting the newly measured partial time series data as described above (step S13). For example, the time series data processing device 10 extracts and displays only the corrected feature data corresponding to the current operation mode of the monitored measurement object P or the same model, and also displays the feature data obtained by converting the newly measured partial time series data. This makes it possible to compare the feature data of the monitored measurement object P with the corrected feature data obtained in a normal state obtained in the same operating state. As a result, the feature data obtained by converting the measurement data during monitoring of the measurement object P can be compared with the feature data clustered with high accuracy in the operating state the same as the operating state of the measurement object P, thereby making it possible to suppress erroneous detection of an abnormal state of the measurement object P and improve the accuracy of abnormality detection.

[0032] <Embodiment 2> Next, a second embodiment of the present invention will be described with reference to Fig. 10 to Fig. 12. Fig. 10 to Fig. 11 are block diagrams showing the configuration of a time series data processing device in the second embodiment, and Fig. 12 is a flowchart showing the operation of the time series data processing device. Note that this embodiment shows an outline of the configuration of the time series data processing device and the time series data processing method described in the above-mentioned embodiments.

[0033] First, the hardware configuration of the time series data processing device 100 in this embodiment will be described with reference to Fig. 10. The time series data processing device 100 is configured as a general information processing device, and is equipped with the following hardware configuration, as an example. ·CPU(Central Processing Unit)101(Arithmetic unit) ROM (Read Only Memory) 102 (storage device) RAM (Random Access Memory) 103 (storage device) · Programs 104 loaded into RAM 103 A storage device 105 for storing a program group 104 A drive device 106 that reads and writes data from a storage medium 110 external to the information processing device A communication interface 107 for connecting to a communication network 111 outside the information processing device Input / output interface 108 for inputting and outputting data Bus 109 connecting each component

[0034] The time series data processing device 100 can be equipped with the database 121, the conversion unit 122, and the extraction unit 123 shown in Fig. 11 by the CPU 101 acquiring the program group 104 and executing it. The program group 104 is stored in the storage device 105 or the ROM 102 in advance, for example, and the CPU 101 loads the program group 104 into the RAM 103 and executes it as necessary. The program group 104 may be supplied to the CPU 101 via the communication network 111, or may be stored in the storage medium 110 in advance, and the drive device 106 may read out the program and supply it to the CPU 101. However, the database 121, the conversion unit 122, and the extraction unit 123 described above may be constructed of dedicated electronic circuits for realizing such functions.

[0035] 10 shows an example of the hardware configuration of the information processing device which is the time-series data processing device 100, and the hardware configuration of the information processing device is not limited to the above-mentioned case. For example, the information processing device may be configured with a part of the above-mentioned configuration, such as not having the drive device 106.

[0036] The time series data processing device 100 executes the time series data processing method shown in the flowchart of FIG. 12 by the functions of the database 121, the conversion unit 122, and the extraction unit 123 constructed by the program as described above.

[0037] As shown in FIG. 12, the time series data processing device 100 Associating time series data measured from a measurement object with operating state information representing the operating state of the measurement object when the time series data was measured (step S101); The time series data is converted into feature amount data for each predetermined period, and the feature amount data is converted into corrected feature amount data that is corrected based on the time of the corresponding period (step S102); Extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data (step S103); The following process is executed.

[0038] The above-mentioned program may be stored in various types of non-transitory computer readable media and provided to a computer. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer readable media include magnetic recording media, Media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memories (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)). The program may also be provided to the computer by various types of transitory computer readable media. Examples of the computer-readable medium include an electric signal, an optical signal, and an electromagnetic wave. The temporary computer-readable medium can provide the program to the computer via a wired communication path such as an electric wire or an optical fiber, or a wireless communication path.

[0039] Although the present invention has been described above with reference to the above-mentioned embodiment, the present invention is not limited to the above-mentioned embodiment. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. In addition, at least one or more functions of the above-mentioned database 121, conversion unit 122, and extraction unit 123 may be executed by an information processing device installed and connected anywhere on a network, that is, they may be executed by so-called cloud computing.

[0040] <Additional Notes> A part or all of the above-described embodiments can be described as follows: The configurations of the time series data processing method, the time series data processing device, and the program according to the present invention are outlined below. However, the present invention is not limited to the following configurations. (Appendix 1) Associating time series data measured from a measurement object with operating state information that represents an operating state of the measurement object at the time the time series data was measured; converting the time series data into feature amount data for each predetermined period, and converting the feature amount data into corrected feature amount data that is corrected based on the time of the corresponding period; extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data; Time series data processing methods. (Appendix 2) 2. The time series data processing method according to claim 1, converting the feature amount data for each predetermined period into corrected feature amount data having a value corresponding to the closeness of time between the corrected feature amount data and another period; Time series data processing methods. (Appendix 3) 3. The time series data processing method according to claim 1, further comprising: converting the feature amount data for each predetermined period into corrected feature amount data such that the closer the time of the feature amount data is to another period, the closer the value of the corrected feature amount data becomes to the corrected feature amount data corresponding to the other period; Time series data processing methods. (Appendix 4) A time series data processing method according to any one of claims 1 to 3, adding time data, the value of which is based on the time of the corresponding period, to the feature amount data for each predetermined period, and converting the feature amount data to which the time data has been added into the corrected feature amount data; Time series data processing methods. (Appendix 5) 5. The time series data processing method according to claim 4, adding the time data, the value of which becomes closer the closer the time of the corresponding period is, to the feature amount data for each predetermined period, and converting the feature amount data to which the time data has been added into the corrected feature amount data; Time series data processing methods. (Appendix 6) 6. The time series data processing method according to claim 4, further comprising: converting the time series data into the feature data using an autoencoder for each predetermined period, and converting the feature data to which the time data has been added into the corrected feature data using an autoencoder. Time series data processing methods. (Appendix 7) 7. A time series data processing method according to any one of claims 1 to 6, comprising: converting second time series data measured from the measurement object into second feature amount data; performing a process of comparing the corrected feature amount data extracted based on the operating state information with the second feature amount data; Time series data processing methods. (Appendix 8) 8. The time series data processing method according to claim 7, extracting the corrected feature amount data corresponding to the time-series data associated with the selected operational state information; Time series data processing methods. (Appendix 9) A time series data processing method according to any one of claims 1 to 8, The operating state information includes information representing an operating mode when the measurement target is operating. Time series data processing methods. (Appendix 10) A time series data processing method according to any one of claims 1 to 9, Associating the time-series data with type information that indicates a type of the measurement target; extracting the corrected feature amount data corresponding to the time series data based on the type information associated with the time series data; Time series data processing methods. (Appendix 11) a database in which time series data measured from a measurement object is associated with operational status information representing the operational status of the measurement object when the time series data was measured; a conversion unit that converts the time series data into feature amount data for each predetermined period, and converts the feature amount data into corrected feature amount data that is corrected based on a time of the corresponding period; an extraction unit that extracts the corrected feature amount data corresponding to the time-series data based on the operational state information associated with the time-series data; A time series data processing device comprising: (Appendix 12) 12. The time series data processing device according to claim 11, the conversion unit converts the feature amount data for each predetermined period into corrected feature amount data having a value corresponding to a time closeness of the other period to the corrected feature amount data corresponding to the other period, Time series data processing device. (Appendix 13) 13. The time-series data processing device according to claim 11, the conversion unit converts the feature amount data for each predetermined period into the corrected feature amount data such that the closer the time of the feature amount data is to another period, the closer the value of the corrected feature amount data becomes to the corrected feature amount data corresponding to the other period. Time series data processing device. (Appendix 14) 14. A time series data processing device according to any one of claims 11 to 13, the conversion unit adds time data, the value of which is based on the time of a corresponding period, to the feature amount data for each predetermined period, and converts the feature amount data to which the time data has been added into the corrected feature amount data. Time series data processing device. (Appendix 15) 15. The time series data processing device according to claim 14, the conversion unit converts the feature amount data added with the time data, the value of which becomes closer the closer the time of the corresponding period is to the feature amount data for each predetermined period, into the corrected feature amount data. Time series data processing device. (Appendix 16) 16. The time series data processing device according to claim 14, the conversion unit converts the time series data into the feature data by using an autoencoder for each predetermined period, and converts the feature data to which the time data has been added into the corrected feature data by using an autoencoder. Time series data processing device. (Appendix 17) 17. A time series data processing device according to any one of claims 11 to 16, The conversion unit converts second time series data measured from the measurement object into second feature amount data; the extraction unit performs a process of comparing the corrected feature amount data extracted based on the operation state information with the second feature amount data. Time series data processing device. (Appendix 18) 18. The time series data processing device according to claim 17, The extraction unit extracts the corrected feature amount data corresponding to the time-series data associated with the selected operational status information. Time series data processing device. (Appendix 19) A time series data processing device according to any one of claims 11 to 8, the database stores data in which the time-series data is associated with type information that indicates a type of the measurement target, the extraction unit extracts the corrected feature amount data corresponding to the time series data based on the type information associated with the time series data. Time series data processing device. (Appendix 20) An information processing device that can access a database in which time series data measured from a measurement object and operational status information representing the operational status of the measurement object at the time the time series data was measured are associated with each other; converting the time series data into feature amount data for each predetermined period, and converting the feature amount data into corrected feature amount data that is corrected based on the time of the corresponding period; extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data; A program for executing a process. [Explanation of symbols]

[0041] 10 Time series data processing device 11 Acquisition Department 12 Conversion section 13 Extraction part 16 Measurement data storage unit 17 Feature data storage unit 100 Time series data processing device 101 CPU 102 ROM 103 RAM 104 Programs 105 Storage device 106 Drive device 107 Communication Interface 108 Input / Output Interface 109 Bus 110 Storage medium 111 Communication Network 121 Database 122 Conversion unit 123 Extraction part

Claims

1. An information processing device comprising: Associating time series data measured from a measurement object with operating state information that represents an operating state of the measurement object at the time the time series data was measured; converting the time series data into feature data for each predetermined period, adding time data to the feature data, the time data being a value based on the time at which the time series data corresponding to the feature data was measured, and converting the feature data with the added time data into corrected feature data, the time series data being converted into the feature data using an autoencoder, and the feature data with the added time data being converted into the corrected feature data using an autoencoder; extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data; Time series data processing methods.

2. 2. The time series data processing method according to claim 1, The information processing device, adding, to the feature amount data for each predetermined period, the time data which becomes closer in value to the closer the time at which the time-series data corresponding to the feature amount data is measured, and converting the feature amount data to which the time data has been added into the corrected feature amount data; Time series data processing methods.

3. 3. The time series data processing method according to claim 1, further comprising: The information processing device, converting second time series data measured from the measurement object into second feature amount data; performing a process of comparing the corrected feature amount data extracted based on the operating state information with the second feature amount data; Time series data processing methods.

4. 4. The time series data processing method according to claim 3, The information processing device, extracting the corrected feature amount data corresponding to the time-series data associated with the selected operational state information; Time series data processing methods.

5. a database in which time series data measured from a measurement object is associated with operational status information representing the operational status of the measurement object when the time series data was measured; a conversion unit that converts each of the time series data into feature data for each predetermined period, adds time data to the feature data, the time data being a value based on the time at which the time series data corresponding to the feature data was measured, and converts the feature data to which the time data has been added into corrected feature data, converting the time series data into the feature data using an autoencoder and converting the feature data to which the time data has been added into the corrected feature data using an autoencoder; an extraction unit that extracts the corrected feature amount data corresponding to the time-series data based on the operational state information associated with the time-series data; A time series data processing device comprising:

6. An information processing device that can access a database in which time series data measured from a measurement object and operational status information representing the operational status of the measurement object at the time the time series data was measured are associated with each other, converting the time series data into feature data for each predetermined period, adding time data to the feature data, the time data being a value based on the time at which the time series data corresponding to the feature data was measured, and converting the feature data with the added time data into corrected feature data, the time series data being converted into the feature data using an autoencoder, and the feature data with the added time data being converted into the corrected feature data using an autoencoder; extracting the corrected feature amount data corresponding to the time series data based on the operational state information associated with the time series data; A program for executing a process.

Citation Information

Patent Citations

  • Manufacture of mats utilizing waste tire

    JP1981084941A

  • Boiler tube leak detection device, boiler tube leak detection method, data monitoring center using the device and method, information providing service and boiler plant

    JP2015007509A

  • Abnormality detection system

    JP2020008997A

  • Feature amount generation device, feature amount generation method, and program

    WO2018142694A1