Data extraction method, data processing device, and abnormality detection system using the same

Abstract

To provide a data extraction method, a data processing device, and an abnormality detection system using the same, that can detect data which is difficult to detect but effective for an abnormality detection and can improve an accuracy of extracting data which is significant for the abnormality detection, and can reduce an amount of data.SOLUTION: In a data extraction method for receiving a signal from a sensor and extracting significant data, the data extraction method: uses data generated by filter processing, which filters the signal from the sensor, envelope processing of the sensor signal or filtered data, or moving average processing of the signal from the sensor or envelope processed data, as data used for a significant section extraction; detects a significant section based on the data used for the significant section extraction; and performs data extraction on the detected significant section with respect to the signal from the sensor.SELECTED DRAWING: Figure 7

Description

[Technical field]

[0001] The present invention relates to a data extraction method for extracting significant data from measurement data of a sensor, a data processing device, and an anomaly detection system using the same. [Background technology]

[0002] In recent years, with the development of IoT (Internet of Things), various measurement data are collected from various sensors installed in devices or facilities, and services are being provided to monitor the status of manufacturing industries and work sites, as well as to perform management and maintenance. For example, equipment abnormality diagnosis and piping gas leakage diagnosis can be detected by analyzing vibration data and sound data in the ultrasonic band (20 kHz and above). Here, vibration data and sound data require a sampling rate of several ksps (sampling / sec) or more, and sound data in particular requires a sampling rate of about 200 ksps, which results in a huge amount of data. In order to transmit these data to a public network such as LTE-M (Long Term Evolution for machine-type-communication) or a high-level abnormality diagnosis device via low-power wireless, it is necessary to reduce the amount of data. One method of reducing the amount of data is to reduce the amount of data by making the data into features. Another method is to perform filtering on vibration and sound data to detect significant sections.

[0003] Patent Document 1 is a prior art document in this technical field. Patent Document 1 discloses a data compression method for identifying an interval including a characteristic signal change based on data detected by a sensor and compressing the data using this method, as well as a sensor device and a sensor network system that can reduce the amount of communication by transmitting data compressed using this method. Specifically, the system extracts time-series data of a specific frequency component by filtering time-series data detected by a sensor element within a predetermined period, and compares the extracted time-series data with a predetermined threshold to determine only the portion where the signal strength changes significantly as a significant interval, and wirelessly transmits only the time-series data of this significant interval, thereby reducing the amount of communication by the sensor device. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2017-122976 A Summary of the Invention [Problem to be solved by the invention]

[0005] In Patent Document 1, significant intervals are detected by performing filtering on time-series data detected by a sensor element. However, there are problems with filtering alone, such as the inability to detect significant intervals or the loss of useful data, which affects the accuracy of anomaly diagnosis.

[0006] In view of the above problems, the present invention aims to provide a data extraction method, a data processing device, and an anomaly detection system using the same, which are capable of detecting data that is difficult to detect but is effective for anomaly detection, improving the accuracy of extracting data significant for anomaly detection, and reducing the amount of data. [Means for solving the problem]

[0007] One example of the present invention is a data extraction method that receives a signal from a sensor and extracts significant data, and the data used to extract significant intervals is generated by either a filter process that filters the signal from the sensor, an envelope process of the signal from the sensor or the filtered data, or a moving average process of the signal from the sensor or the envelope processed data. Significant intervals are detected based on the data used to extract significant intervals, and data is extracted from the signal from the sensor in the detected significant intervals. Effect of the Invention

[0008] According to the present invention, it is possible to provide a data extraction method, a data processing device, and an anomaly detection system using the same, which are capable of improving the accuracy of extracting significant data and reducing the amount of data. [Brief description of the drawings]

[0009] [Figure 1] 1 is a schematic configuration diagram of an anomaly detection system according to an embodiment. [Diagram 2] FIG. 2 is a hardware configuration diagram of a data processing device according to an embodiment. [Diagram 3] FIG. 2 is another hardware configuration diagram of the data processing device according to the embodiment. [Figure 4] FIG. 3 is a functional configuration diagram of a data processing device corresponding to FIG. 2 in a case where the data used to extract significant sections is different from the data to be converted into features. [Diagram 5] FIG. 3 is a functional configuration diagram of a data processing device corresponding to FIG. 2 in a case where the data used for extracting significant sections is the same as the data to be converted into features. [Figure 6] FIG. 4 is a functional configuration diagram of a data processing device corresponding to FIG. 3. [Figure 7] 11 is a flowchart showing the first half of a data extraction process of the data processing device according to the embodiment. [Figure 8] 11 is a second half of a flowchart showing the data extraction process of the data processing device in the embodiment. [Figure 9]FIG. 5 is a functional block diagram of a data processing device corresponding to FIG. 4, showing a specific example of a case where vibration data is extracted. [Figure 10] FIG. 5 is a functional block diagram of a data processing device corresponding to FIG. 4, showing a specific example of a case where sound data is extracted. [Figure 11] FIG. 5 is a functional block diagram of a data processing device corresponding to FIG. 4, showing a specific example of a case where current data is extracted. [Figure 12] FIG. 6 is a functional block diagram of a data processing device corresponding to FIG. 5, showing a specific example of a case where vibration data is extracted. [Figure 13] FIG. 2 is a diagram illustrating the configuration of a target device for data measurement in an experiment for explaining an embodiment. [Figure 14] FIG. 13 is an explanatory diagram of the operation of a target device for data measurement in an experiment for explaining an embodiment. [Figure 15] 11 is a waveform diagram showing a data extraction procedure based on a threshold value of vibration data when the data used to extract a significant section in a data processing device and the data to be characterized are different in an experiment for explaining an embodiment. FIG. [Figure 16] 11 is a waveform diagram showing a data extraction procedure based on a threshold value of vibration data when the data used to extract a significant section in a data processing device and the data to be characterized are the same in an experiment for explaining an embodiment. FIG. [Figure 17] 11 is a waveform diagram showing a data extraction procedure based on peak values ​​of sound data when the data used to extract a significant section in a data processing device and the data to be characterized are different in an experiment for explaining an embodiment. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. EXAMPLES

[0011] 1 is a schematic diagram of an anomaly detection system according to the present embodiment. The anomaly detection system 100 shown in FIG.

[0012] The sensor 10 converts physical quantities such as vibration, sound, current, and temperature into electrical signals and outputs them. The data processing device 20 receives the electrical signals from the sensor 10, extracts significant data, and transmits them to a higher-level host system 30. The host system 30 is an information processing device such as a PC or a server, and performs analysis using the received data to perform abnormality diagnosis and predictive diagnosis.

[0013] Fig. 2 is a hardware configuration diagram of a data processing device in this embodiment. In Fig. 2, the data processing device 20 is composed of an analog circuit 210, a CPU (Central Processing Unit) (processor) 220, a memory (storage device) 221, and an output interface (I / F) 222. These are connected via a data bus 223.

[0014] The analog circuit 210 has a filter 211 that performs filtering (filter processing) to extract specific frequency components from the electrical signal from the sensor 10, a gain adjustment circuit 212 that adjusts the gain, an envelope processing circuit 213 that performs envelope processing to extract the outline of a given waveform, and an analog / digital conversion circuit (A / D) 214. Note that, as will be described later, the analog circuit 210 also has switches for turning each processing circuit ON / OFF and a control circuit for controlling the switches so that various processes of each processing circuit can be selected.

[0015] The CPU 220 takes in the digital data converted by analog / digital conversion from the analog circuit 210 into a memory 221, performs necessary data processing, and outputs the processed data to the upper system 30 via an output I / F 222. That is, the data processing device 20 shown in Fig. 2 performs mixed processing of hardware processing and software processing using analog circuits and digital circuits.

[0016] Fig. 3 is another hardware configuration diagram of the data processing device in this embodiment. In Fig. 3, the same components as those in Fig. 2 are assigned the same reference numerals, and their description will be omitted. In Fig. 3, the data processing device 20 converts the measurement data from the sensor 10 into digital data by the A / D 214, imports the digital data into the memory 221, and performs software processing by the CPU 220, such as filtering by a digital filter, as a subsequent process. That is, in the data processing device 20 shown in Fig. 3, the CPU 220 reads out programs and information that realize each function from the memory 221, executes predetermined processing by software processing, and performs all software processing by digital signals.

[0017] Fig. 4 is a functional configuration diagram when the data used for extracting significant intervals and the data to be characterized are different in a data processing device corresponding to Fig. 2. Here, the data to be characterized is data from which significant interval data is extracted based on a significant interval detected by the data used for extracting significant intervals, and the extracted significant interval data is characterized.

[0018] In Fig. 4, the hardware processes include a filter process 21, a gain adjustment process 22, an envelope process 23, and an A / D 24 for the data to be characterized. Similarly, the hardware processes include a filter process 25, a gain adjustment process 26, an envelope process 27, and an A / D 28 for generating data to be used for extracting significant intervals. These processes are executed by the filter 211, the gain adjustment circuit 212, the envelope processing circuit 213, and the A / D 214 in Fig. 2. In addition, the software processes by the CPU 220 include a moving average process 41 for the data to be characterized, a moving average process 42 for the data to be used for extracting significant intervals, a significant interval detection process 43, a significant interval data extraction process 44, a feature extraction process 45, and a format forming process 46. Here, the moving average process is a process for smoothing so-called time series data. Note that these processes are configured so that each process can be selected from various types, as described later.

[0019] Fig. 5 is a functional configuration diagram of a data processing device corresponding to Fig. 2 in which the data used to extract significant intervals is the same as the data to be converted into features. In Fig. 5, the hardware processes include a filter process 21, a gain adjustment process 22, an envelope process 23, and an A / D 24. These processes are executed by the filter 211, the gain adjustment circuit 212, the envelope processing circuit 213, and the A / D 214 in Fig. 2. In addition, the software processes executed by the CPU 220 include a moving average process 41, a significant interval detection process 43, a significant interval data extraction process 44, a feature extraction process 45, and a format forming process 46. Note that these processes are also configured so that each process can be selected from various types, as will be described later.

[0020] Fig. 6 is a functional block diagram of a data processing device corresponding to Fig. 3. In Fig. 6, the digital data after A / D 24 is used, and software processing by CPU 220 includes filter processing 21, gain adjustment processing 22, envelope processing 23, moving average processing 41, significant section detection processing 43, significant section data extraction processing 44, feature extraction processing 45, and format forming processing 46. As will be described later, these processes are also configured so that each process can be selected from various types.

[0021] 7 and 8 are process flowcharts for extracting data in the data processing device in this embodiment. Due to space limitations, the process flowchart is divided into a first half and a second half, with the first half shown in FIG. 7 and the second half shown in FIG.

[0022] 7, first, in step S21, data measurement is started from a sensor installed in a target device or facility, etc. Then, steps S22 to S27 are carried out on the measurement data to generate data (1) to be characterized, and steps S28 to S33 are carried out to generate data (2) to be used for extracting significant intervals.

[0023] In step S22, a setting as to whether or not to perform filter processing is determined, and if filter processing is to be performed, filter processing is performed in step S23, and if filter processing is not to be performed, step S23 is skipped and the process proceeds to the next step S24. In step S24, a setting as to whether or not to perform gain adjustment is determined, and if gain adjustment is to be performed, gain adjustment is performed in step S25, and if gain adjustment is not to be performed, step S25 is skipped and the process proceeds to the next step S26. In step S26, a setting as to whether or not to perform envelope processing is determined, and if envelope processing is to be performed, envelope processing is performed in step S27, and if envelope processing is not to be performed, step S27 is skipped and the process proceeds to the next step S34. That is, as data (1) to be characterized, data generated by any of filter processing of measurement data, gain adjustment processing of measurement data or filtered data, and envelope processing of measurement data or gain adjusted data is used.

[0024] Steps S28 to S33 for generating data (2) used to extract significant intervals are performed in the same manner as steps S22 to S27. Although filtering is considered essential for data (2) used to extract significant intervals, filtering can also be selected, and the settings for whether or not to perform filtering, gain adjustment, and envelope processing are appropriately set according to the target device or facility.

[0025] Next, in step S34, the data (1) to be characterized using steps S22 to S27 and the data (2) to be used to extract significant sections using steps S28 to S33 are temporarily stored for the measurement data. Then, in step S35, it is determined whether the time required for data measurement has elapsed, and if the time has not yet elapsed, the process returns to step S21 to continue the sensor data measurement and repeats steps S22 to S35. If the time required for data measurement has elapsed in step S35, the process proceeds to step S36, and the data (1) to be characterized and the data (2) to be used to extract significant sections stored in step S34 are read out.

[0026] Next, in step S37, it is determined whether the data used to extract the significant intervals is the same as the data to be characterized, and if they are the same, steps S38 to S43 are carried out to detect significant intervals using the data to be characterized (1) as the data to be used to extract the significant intervals, whereas if they are different, steps S44 to S49 are carried out to detect significant intervals using the data to be used to extract the significant intervals (2).

[0027] That is, in step S38, a setting is determined as to whether or not to perform moving average processing, and if moving average processing is performed, moving average processing of data (1) is performed in step S39, and if moving average processing is not performed, step S39 is skipped and the process proceeds to the next step S40. In step S40, a setting is determined as to whether or not to perform significant interval detection using a threshold, and if significant interval detection using a threshold is performed, a threshold value for data (1) is calculated in step S41, and the process proceeds to step S43. If significant interval detection using a threshold is not performed in step S40, a peak value of data (1) is calculated in step S42 to detect significant intervals using a peak value, and the process proceeds to step S43. In step S43, significant interval detection using data (1) is performed based on a threshold value or a peak value. That is, in the case of significant interval detection using a threshold, detection is performed by comparing the data used to extract significant intervals with the threshold value. Also, in the case of significant interval detection using a peak value, detection is performed by using a peak value as a base point and a predetermined period before and after the peak value as a significant interval. Here, detection of significant intervals using a threshold value can be applied when the operating time is relatively long, such as a few seconds, while detection of significant intervals using a peak value is applied when the operating time is short or when it is difficult to detect significant intervals using a threshold value. The threshold value is automatically determined from the data (2) used to extract significant intervals. For example, the threshold value is set to 1 / 2 the maximum value of the data used to extract significant intervals. The time when the threshold value is exceeded and the significant interval from the base point of the peak value can be expanded or reduced, assuming that the operating time differs for each device or facility.

[0028] When the data used to extract the significant intervals is different from the data to be characterized, steps S44 to S49 are the same as steps S38 to S43, but the target data is the data (2) used to extract the significant intervals.

[0029] In addition, the settings of whether the data used to extract significant intervals is the same as the data to be characterized, whether moving average processing is performed, and whether significant intervals are detected using a threshold value are each set appropriately according to the target device or facility.

[0030] 8, it is determined in step S50 whether a significant interval has been detected, and if not, the process returns to START and the data measurement is repeated. If a significant interval has been detected, the process proceeds to step S51, where significant interval data is extracted from the data (1) to be characterized, feature quantities are extracted from the significant interval data in step S52, the data is formatted in step S53, and the data is transmitted to the upper system in step S54.

[0031] 7 and 8, for example, when performing filtering, envelope processing, and moving average processing, data generated by filtering measurement data, envelope processing of the measurement data or the filtered data, and moving average processing of the measurement data or the envelope processed data are used as the data (1) to be characterized and the data (2) to be used to extract significant sections. The order of processing is filtering, envelope processing, and moving average processing.

[0032] 9 to 11 are specific examples of functional configuration diagrams of a data processing device in the case where the data used to extract the significant interval corresponding to FIG. 4 is different from the data to be characterized.

[0033] First, Fig. 9 is a functional configuration diagram showing a specific example of vibration data extraction. In Fig. 9, the data processing device takes in vibration data as measurement data from the sensor, and the data (1) to be converted into a feature quantity is the data after A / D 24 with settings that do not perform filter processing, gain adjustment, or envelope processing. Then, the data (2) to be used for extracting the significant interval is the data after envelope processing 27 and A / D 28 with settings that do not perform filter processing or gain adjustment, and is the data to be processed for moving average processing 42. Then, a significant interval detection process 43 is performed, and the data (1) to be converted into a feature quantity is subjected to significant interval data extraction 44 of the detected significant interval, feature extraction process 45, and format forming process 46.

[0034] Fig. 10 is a functional configuration diagram showing a specific example of sound data extraction. In Fig. 10, the data processing device takes in sound data as data (1) to be characterized as measurement data from a sensor, and sets the data to be subjected to envelope processing 23 and A / D 24 without filter processing or gain adjustment. The data processing device then takes in vibration data as data (2) to be used for extracting significant intervals, and sets the data to be subjected to filter processing 25, gain adjustment processing 26, envelope processing 27, and A / D 28, followed by moving average processing 42. Then, significant interval detection processing 43 is performed, and significant interval data extraction 44 of the detected significant interval is performed on the data (1) to be characterized, feature extraction processing 45 is performed, and format forming processing 46 is performed.

[0035] Similarly, Fig. 11 is a functional configuration diagram showing a specific example when current data is extracted. In Fig. 11, the data processing device takes in current data as measurement data from the sensor, and data (1) to be converted into a feature quantity is data after A / D 24 with settings that do not perform filter processing, gain adjustment, or envelope processing. Data (2) to be used for extracting significant intervals is data after filter processing 25, gain adjustment processing 26, envelope processing 27, and A / D 28, and then moving average processing 42. Then, significant interval detection processing 43 is performed, and significant interval data extraction 44 of the detected significant interval is performed on the data (1) to be converted into a feature quantity, feature extraction processing 45 is performed, and format forming processing 46 is performed.

[0036] Fig. 12 is a specific example of the case of extracting vibration data in the functional configuration diagram of a data processing device in the case where the data used to extract significant intervals and the data to be characterized are the same, corresponding to Fig. 5. In Fig. 12, the data processing device takes in vibration data as measurement data from the sensor, assuming that the data to be characterized (1) is also used as the data to be used to extract significant intervals, and sets it to data after envelope processing 23 and A / D 24 without filter processing or gain adjustment. Then, it performs significant interval detection processing 43 on the data, performs significant interval data extraction 44 of the detected significant interval, performs feature extraction processing 45, and performs format forming processing 46.

[0037] Note that, although a specific example of a functional configuration diagram corresponding to Figure 6 in which the measurement data from the sensor is A / D converted to digital data and processed entirely by software is omitted, the processing program can control the ON / OFF of each function, etc.

[0038] FIG. 13 is a diagram illustrating the configuration of a target device for data measurement in an experiment to explain this embodiment. In FIG. 13, the target device for data measurement in the experiment is a device in which a stage 51 moves along a linear guide 52, and the linear guide 52 is supported by bearings 53 and 54. A rotor 55 is driven by a motor 56, so that the linear guide 52 has a ball screw structure, and the rotary motion of the rotor 55 is converted into linear motion, and the stage 51 moves linearly along the linear guide 52. As sensors, the bearing 53 has an acceleration sensor 57, which is a vibration sensor, a microphone 58, which is a sound sensor, and an ammeter 59, which is a current sensor. And, as an abnormal state for the experiment, for example, the bearing is corroded by applying salt water to cause bearing damage, the sliding surface of the linear guide is wiped off to cause insufficient lubrication, and a weight is attached to the rotor to cause imbalance.

[0039] Fig. 14 is an explanatory diagram of the operation of the target device for data measurement in an experiment to explain this embodiment. As shown in Fig. 14, the experiment involves movement 1 of moving the stage 51 to the right in the figure, followed by movement 2 of moving it to the left, and movement 3 of moving it further to the right. The data processing device acquires measurement data of vibration, sound, and current from each sensor during these movements.

[0040] 15 to 17 are experimental waveform diagrams obtained when predetermined processing is performed on measurement data from a sensor generated by the operation shown in FIG. 14 in the target device shown in FIG.

[0041] Specifically, FIG. 15 is an experimental waveform diagram showing a data extraction procedure by a threshold value of vibration data when the data used for extracting the significant section of the data processing device is different from the data to be characterized. That is, it is an experimental waveform diagram corresponding to the functional configuration diagram of FIG. 9. In FIG. 15, (A) is the vibration measurement data from the acceleration sensor 57, the horizontal axis indicates the time required for the motions 1, 2, and 3, and the vertical axis indicates the vibration intensity. (B) is the data after envelope processing, which is the data used for extracting the significant section, and is the data obtained by performing envelope processing on the vibration measurement data of (A). (C) is the data after moving average processing, which is the data used for extracting the significant section, and is the data obtained by performing moving average processing on the envelope processed data of (B). Then, in (C), a threshold value is set, and the significant section is detected by comparing the data used for extracting the significant section with the threshold value. (D) is the data to be characterized, and data is extracted in the significant section detected in (C) from the vibration measurement data, which is the data to be characterized in (A).

[0042] Fig. 16 is an experimental waveform diagram showing a data extraction procedure using a threshold value of vibration data when the data used to extract significant sections in the data processing device and the data to be characterized are the same. That is, it is an experimental waveform diagram corresponding to the functional configuration diagram of Fig. 12. In Fig. 16, (A) is the same vibration measurement data from acceleration sensor 57 as in Fig. 15. (B) is data after envelope processing, which is data used to extract significant sections and data to be characterized, and is data obtained by performing envelope processing on the vibration measurement data of (A). Then, in (B), a threshold value is set and significant sections are detected using the threshold value, and (C) is data to be characterized, and data extraction is performed in the detected significant sections from the envelope-processed data of (B).

[0043] FIG. 17 is an experimental waveform diagram showing a data extraction procedure based on peak values ​​of sound data when the data used to extract significant sections in the data processing device is different from the data to be made into feature quantities.

[0044] That is, it is an experimental waveform diagram corresponding to the functional configuration diagram of FIG. 10. In FIG. 17, (A) is an experimental waveform diagram after envelope processing of the sound measurement data from the microphone 58, with the horizontal axis representing the time required for the actions 1, 2, and 3, and the vertical axis representing the sound intensity. (B) is an experimental waveform diagram after filter processing and envelope processing of the vibration data from the acceleration sensor 57 as data used for extracting the significant section. (C) is data after moving average processing of the envelope-processed data of (B). Then, in (C), a peak value is detected, and a predetermined period before and after the peak value is detected as a base point as a significant section. Then, (D) is data to be made into a feature quantity, and data extraction is performed in the significant section detected in (C) for the envelope-processed data of the sound measurement data of (A).

[0045] As described above, according to this embodiment, by determining a significant interval by combining filter processing with envelope processing and moving average processing, it is possible to provide a data extraction method, a data processing device, and an anomaly detection system using the same, which make it possible to detect data that is difficult to detect but effective for anomaly detection, improve the accuracy of extracting data significant for anomaly detection, and reduce the amount of data.

[0046] The above describes an embodiment of the present invention. The present invention can provide a data processing device that can reduce the amount of data, thereby reducing carbon emissions and preventing global warming, and contributing to achieving the Sustainable Development Goals (SDGs), particularly item 7, energy.

[0047] Furthermore, the present invention is not limited to the above-described embodiment, and includes various modified examples. For example, the above-described embodiment has been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to the embodiment having all of the described configurations. [Explanation of symbols]

[0048] 10: sensor, 20: data processing device, 21, 25: filter processing, 22, 26: gain adjustment processing, 23, 27: envelope processing, 24, 28: A / D, 30: upper system, 41, 42: moving average processing, 43: significant section detection processing, 44: significant section data extraction processing, 45: feature extraction processing, 46: format forming processing, 51: stage, 52: linear guide, 53, 54: bearing, 55: rotating body, 56: motor, 57: acceleration sensor, 58: microphone, 59: ammeter, 100: anomaly detection system, 211: filter, 212: gain adjustment circuit, 213: envelope processing circuit, 214: analog / digital conversion circuit (A / D), 220: CPU (processor), 221: memory (storage device), 222: output interface (I / F)

Claims

1. 1. A data extraction method for receiving a signal from a sensor and extracting meaningful data, comprising: As data used for extracting the significant section, data generated by any one of a filtering process for filtering the signal from the sensor, an envelope process for the signal from the sensor or the filtered data, and a moving average process for the signal from the sensor or the envelope processed data is used; Detecting a significant section based on the data used for extracting the significant section; A data extraction method comprising the steps of: extracting data from the detected significant section of the signal from the sensor.

2. 2. The data extraction method according to claim 1, A data extraction method, comprising the steps of: detecting the significant interval by comparing the data used for extracting the significant interval with a threshold value.

3. 2. The data extraction method according to claim 1, A data extraction method, comprising the steps of: detecting the significant section by calculating a peak value of data used for extracting the significant section.

4. 3. The data extraction method according to claim 2, A data extraction method characterized in that the threshold value is automatically determined from the data used to extract the significant interval.

5. 2. The data extraction method according to claim 1, A data extraction method comprising the steps of: adjusting a gain on the data used to extract the significant section or on the signal from the sensor.

6. 2. The data extraction method according to claim 1, A data extraction method comprising the steps of: extracting data in the detected significant section from data used for extracting the significant section.

7. A data processing device for receiving signals from the sensors and extracting meaningful data, As data used for extracting the significant section, data generated by any one of a filtering process for filtering the signal from the sensor, an envelope process for the signal from the sensor or the filtered data, and a moving average process for the signal from the sensor or the envelope processed data is used; a significant interval detection process for detecting a significant interval based on the data used for extracting the significant interval; a significant section data extraction process for extracting data in the detected significant section from the signal from the sensor; a feature extraction process for extracting features of the significant section data extracted by the significant section data extraction process; The data processing device further comprises: a data processing unit configured to transmit the feature extracted by the feature extraction process to an outside device.

8. 8. The data processing device according to claim 7, A data processing device comprising: a data processor for extracting data in the detected significant section from data used for extracting the significant section.

9. 8. The data processing device according to claim 7, a filter for performing the filtering process; an envelope processing circuit for performing the envelope processing; an A / D converter that converts the output of the envelope processing circuit into digital data; a memory for storing digital data converted by the A / D converter; A CPU is included. a CPU executes the moving average process, the significant section detection process, the significant section data extraction process, and the feature extraction process through software processing; a data processing device further comprising an output interface for transmitting the feature quantity extracted by the feature quantity extraction process to an external device.

10. 8. The data processing device according to claim 7, an A / D converter for converting a signal from the sensor into digital data; a memory for storing digital data converted by the A / D converter; A CPU is included. a CPU executes the filtering process, the envelope process, the moving average process, the significant section detection process, the significant section data extraction process, and the feature extraction process through software processing; a data processing device further comprising an output interface for transmitting the feature quantity extracted by the feature quantity extraction process to an external device.

11. An anomaly detection system including a sensor, a data processing device, and a host system, The data processing device includes: As data used for extracting the significant section, data generated by any one of a filtering process for filtering the signal from the sensor, an envelope process for the signal from the sensor or the filtered data, and a moving average process for the signal from the sensor or the envelope processed data is used; a significant interval detection process for detecting a significant interval based on the data used for extracting the significant interval; a significant section data extraction process for extracting data in the detected significant section from the signal from the sensor; a feature extraction process for extracting features of the significant section data extracted by the significant section data extraction process; Transmitting the feature amount extracted by the feature amount extraction process to the host system; The anomaly detection system is characterized in that the host system performs analysis using the received feature amount and performs anomaly diagnosis.

Images