Control apparatus and monitoring method

Abstract

A control apparatus and a monitoring method. The control apparatus comprises a tag unit, a management unit, a feature extraction unit and an anomaly determination unit. The tag unit is configured to obtain a plurality of pieces of segment data from a plurality of pieces of original data, and analyze each piece of segment data so as to respectively define tags for the segment data. The management unit is coupled to the tag unit, and is configured to receive the segment data from the tag unit. The management unit is configured to classify part of the segment data having the same tag into a database as a plurality of data points in the database. The feature extraction unit is coupled to the management unit, and is configured to analyze a plurality of feature values of the data points. The anomaly determination unit is coupled to the feature extraction unit, and is configured to compare the feature values with an anomaly threshold value, so as to generate notification information.

Description

Control devices and monitoring methods Technical Field

[0001] This disclosure relates to control technology, and more particularly to a control device and monitoring method. Background Technology

[0002] In automatic control systems, the controller or edge server is the core of the entire system's computation, responsible for generating control commands to cause the drivers to perform corresponding actions, such as motor control. However, under heavy computational loads, this can affect system performance and make it difficult to promptly report or respond to system anomalies. Summary of the Invention

[0003] This disclosure relates to a control device, comprising a tagging unit, a management unit, a feature extraction unit, and an anomaly detection unit. The tagging unit acquires multiple data segments from multiple raw data sources and analyzes each of these data segments to define a tag for a portion of each data segment. The management unit is coupled to the tagging unit and receives the data segments from the tagging unit. The management unit classifies portions of the data segments that share the same tag into a database as multiple data points in the database. The feature extraction unit is coupled to the management unit and analyzes multiple feature values ​​of the data points. The anomaly detection unit is coupled to the feature extraction unit and compares the feature values ​​of the data points with an anomaly threshold to generate a notification message.

[0004] This disclosure also relates to a monitoring method, comprising: acquiring multiple segments of data from multiple raw data sources via a control device; analyzing each of the segments of data to define a label for each of the segments of data; classifying a portion of the segments of data that have the same label into a database as multiple data points in the database; analyzing each of the data points to obtain multiple feature values ​​for the data points; and comparing the feature values ​​of the data points with an anomaly threshold to generate a notification message.

[0005] The control device and monitoring method disclosed herein can analyze the operating conditions (system health status) of the control system in real time, and can improve system cost, reduce computational load, reduce hardware requirements and reduce system complexity. Attached Figure Description

[0006] Figure 1 is a schematic diagram of a control system according to some embodiments of the present disclosure;

[0007] Figure 2 is a schematic diagram of a control device according to some embodiments of the present disclosure;

[0008] Figure 3 is a schematic diagram of data storage in a control system according to some embodiments of the present disclosure;

[0009] Figures 4A and 4B are schematic flowcharts of monitoring methods according to some embodiments of the present disclosure;

[0010] Figure 5A is a schematic diagram of the original data and segment data according to some embodiments of the present disclosure;

[0011] Figure 5B is a schematic diagram of segment data according to some embodiments of the present disclosure;

[0012] Figure 5C is a schematic diagram of segment data according to some embodiments of the present disclosure;

[0013] Figure 6 is a schematic diagram of data points according to some embodiments of the present disclosure;

[0014] Figure 7 is a schematic diagram of the control device determining an abnormal threshold in some embodiments according to the present disclosure.

[0015] Explanation of icon numbers

[0016] 100: Control System

[0017] 110: Controller

[0018] 120: Driver

[0019] 200: Control device

[0020] 210: Control Module

[0021] 220: Monitoring Module

[0022] 221: Data Collection Unit

[0023] 222: Tag Unit

[0024] 223: Management Unit

[0025] 224: Feature Extraction Unit

[0026] 225: Anomaly Detection Unit

[0027] 226: Data Output Unit

[0028] 230: Servo Motor

[0029] 240: Sensor

[0030] 300: Storage Unit

[0031] 510A-510D: Raw Data

[0032] 520: Section Data

[0033] 520A-520D: Section Data

[0034] 520E-520F: Section Data

[0035] 610: Distribution range

[0036] 620: Distribution range

[0037] DB: Database

[0038] DR: Data Range

[0039] D31: Raw Data

[0040] D32: Section Data

[0041] P60: Data Points

[0042] P61: Historical Data Range

[0043] V7: Abnormal Threshold

[0044] S410: Labeling Procedure

[0045] S420: Data Management Program

[0046] S430: Feature Extraction Program

[0047] S440: Judgment Procedure

[0048] S411-S412: Steps

[0049] S421-S425: Steps

[0050] S431-S432: Steps

[0051] S441-S443: Steps Detailed Implementation

[0052] Several embodiments of the present invention will be disclosed below with reference to the accompanying drawings. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential. Furthermore, for the sake of simplicity, some conventional structures and elements will be shown in the drawings in a simple schematic manner.

[0053] In this document, when an element is referred to as a “connection” or “coupled,” it may mean an “electrical connection” or “electrical coupling.” “Connection” or “coupled” can also be used to indicate the operation or interaction between two or more elements. Furthermore, although terms such as “first,” “second,” etc., are used herein to describe different elements, these terms are merely used to distinguish elements or operations described using the same technical terminology. Unless the context clearly indicates otherwise, these terms do not specifically refer to or imply any order or sequence, nor are they intended to limit the invention.

[0054] Figure 1 is a schematic diagram of a control system 100 according to a partial embodiment of the present disclosure. The control system 100 may be an embedded system, including a controller 110 and a driver 120. The controller 110 is coupled to the driver 120 to provide a plurality of control commands. The driver 120 performs corresponding actions according to the control commands. In one embodiment, the driver 120 is disposed in a servo motor.

[0055] To monitor whether the operation of the control system 100 is abnormal, in one approach, the controller 110 or edge processor calculates the threshold value that should be used to determine "abnormal operating conditions," while simultaneously monitoring time-series data (such as drive current) or detection data (such as speed feedback values) in the control system 100 to determine whether the abnormal threshold is exceeded. However, this method significantly increases the computational load on the controller 110 (or edge processor), resulting in a decrease in system performance.

[0056] This disclosure is used to improve the signal processing method of the control system 100 to accurately determine whether an abnormality has occurred in the current operating condition (e.g., the health status of the system). The monitoring method disclosed herein can be executed by the controller 110 or the driver 120. For ease of explanation, the processing circuit executing the monitoring method is referred to herein as a "control device". In other words, the control device can be located in the controller 110 or the driver 120.

[0057] Figure 2 is a schematic diagram of a control device 200 according to a partial embodiment of the present disclosure. The control device 200 includes a control module 210 and a monitoring module 220. The control module 210 and the monitoring module 220 may each be an arithmetic circuit and may be uniformly packaged in a microprocessor or integrated circuit chip, or separately packaged in multiple different microprocessors or integrated circuit chips.

[0058] The control module 210 is used to transmit / generate control commands to drive the corresponding device. In this embodiment, the control module 210 is used to drive the servo motor 230. As shown in Figures 1 and 2, if the control device 200 is located on the controller 110, the control module 210 provides control commands to the driver 120, and the driver 120 can send back corresponding feedback signals. If the control device 200 is located on the driver 120, the control module 210 drives the servo motor 230 according to the control commands.

[0059] The monitoring module 220 is used to acquire various parameters of the control system 100 during operation, such as the aforementioned time-series data, feedback signals, or detection data, to determine whether the current operating condition is abnormal. For ease of explanation, the time-series data, feedback signals, and detection data monitored by the monitoring module 220 are collectively referred to as "raw data." In other words, the control device 200 can use the control commands (such as motor position, drive current / voltage, operating torque, and operating speed) and detection / feedback data (such as detection current value and speed feedback value) from the control system 100 as raw data for analysis.

[0060] In one embodiment, the control device 200 performs separate analyses on different categories of raw data to determine whether the current operating condition is abnormal from different perspectives. In other words, each analysis performed by the control device 200 is for the same category of raw data, such as analyzing multiple "drive currents" or multiple "speed feedback values" to determine whether the operating condition is abnormal.

[0061] To facilitate the explanation of the operation of the control device 200, the data storage method of the control system 100 will be described first with reference to Figure 3. Referring to Figures 1-3, the control system 100 includes a storage unit 300, which may be included in or coupled to the control device 200, for example, as a memory located within the controller 110 or the driver 120. The storage unit 300 is used to store the raw data received by the control device 200, and the control device 200 can classify and manage the raw data, storing it in different databases DB. Specifically, the control device 200 first obtains the judgment of the raw data D31, which is then used as segment data D32. Next, the control device 200 analyzes the segment data D32 to classify / categorize each segment data D32 into different databases DB. Therefore, the control device 200 can analyze the data in different databases DB to accurately determine the operating condition of the control system 100.

[0062] Please refer to Figures 1 to 3. Specifically, in one embodiment, the control module 210 includes a data collection unit 221, a tag unit 222, a management unit 223, a feature extraction unit 224, an anomaly detection unit 225, and a data output unit 226. The aforementioned units 221 to 226 may be separate circuits or may be integrated into the same processing chip.

[0063] The data collection unit 221 is used to acquire multiple pieces of raw data D31, such as detecting the operation of the servo motor 230 through the sensor 240, to periodically or continuously acquire multiple pieces of detection data. As mentioned above, the control device 200 analyzes different categories of raw data D31 from different perspectives; therefore, the "multiple pieces of raw data D31" mentioned here all refer to the same category, such as drive current. In some embodiments, the raw data is used to control the servo motor 230, or is generated by the servo motor 230, or is generated by an external sensor.

[0064] The tagging unit 222 is used to obtain multiple segment data D32 based on multiple raw data D31, and is also used to define / set corresponding tags according to the characteristics of the segment data D32. A "tag" is a marker used in data processing to identify different files or data, for example, a classification marker representing "acceleration time 10 milliseconds" or "acceleration time 10 milliseconds, speed 10 RPM". The tag is an encoding appended to the segment data D32 in a specific format. Since those skilled in the art can understand the meaning and definition of the tags, they will not be described further here. In one embodiment, the tagging unit 222 obtains raw data D31 according to a preset time segment to generate segment data D32. For example, the tagging unit 222 can obtain "the signal with the middle 50% time length of one cycle" or "the waveform of 6 signal cycles" from each raw data D31 as segment data D32.

[0065] Management unit 223 is coupled to tag unit 222 to receive multiple tagged segment data D32 and to classify the segment data D32, so as to classify multiple segment data D32 with the same tag into the same database DB. Each segment data D32 in the database DB can be statistically regarded as a data point. Since those skilled in the art can understand the statistical data definition, it will not be described in detail here.

[0066] Feature extraction unit 224 is coupled to management unit 223 to analyze the feature values ​​of these data segments D32 / data points. Anomaly detection unit 225 is coupled to feature extraction unit 224 to compare multiple feature values ​​generated by feature extraction unit 224 with anomaly thresholds and generate notification information based on the comparison results. For example, when the number of feature values ​​exceeding the anomaly threshold is greater than a preset value, it indicates that there is an anomaly in the current operating condition, and therefore anomaly detection unit 225 will generate notification information. The method of comparing feature values ​​with anomaly thresholds will be described in detail in subsequent paragraphs.

[0067] Data output unit 226 is coupled to anomaly detection unit 225 and is used to forward notification information to corresponding devices, such as displaying a warning on the display screen of control system 100 or transmitting notification information to the terminal device of management personnel. In one embodiment, the notification information may also include tag usage rate, database usage rate, and / or the health status of control system 100 calculated by control device 200.

[0068] Figures 4A and 4B are flowcharts of a monitoring method according to some embodiments of the present disclosure. In one embodiment, the monitoring method may be divided into a tagging procedure S410, a data management procedure S420, a feature extraction procedure S430, and a judgment procedure S440. The following description will be based on Figures 1-3 and Figures 4A-4B.

[0069] First, let's explain "Tag Procedure S410". In step S411, the control device 200 receives multiple raw data sets through the data collection unit 221, and the tag unit 222 performs segment filtering on the multiple raw data sets to obtain multiple segment data sets. As mentioned earlier, the raw data can be used as control commands to control the servo motor, or as detection data or feedback data generated by the servo motor / driver.

[0070] Please refer to Figure 5A, which shows a schematic diagram of multiple raw data 510A-510D and segment data 520. The raw data 510A-510D and segment data 520 can correspond to the raw data D31 and segment data D32 in Figure 3. In one embodiment, the raw data 510A-510D is a periodic signal waveform. The tag unit 222 acquires the raw data 510A-510D according to a preset time interval to generate segment data 520 respectively. For example, every 5 seconds, the waveform in four signal cycles is acquired.

[0071] In step S412, the tagging unit 222 assigns tags to at least a portion of the segment data 520. Specifically, the tagging unit 222 analyzes each segment data 520 to generate / define / set tags based on the characteristics of the segment data 520 (e.g., waveform, numerical value, etc.). In other words, similar segment data 520 will have the same tags.

[0072] Please refer to Figure 5B, which shows a schematic diagram of multiple data segments 520A to 520D. In one embodiment, data segments 520A to 520D are signal waveforms representing "rotational speed," therefore, the horizontal axis in Figure 5B represents time, and the vertical axis represents rotational speed (RPM). As shown, the rotational speed of data segments 520A / 520C is approximately 600 RPM, while the rotational speed of data segments 520B / 520D is approximately 1200 RPM, which can be considered as two different ranges. Therefore, data segments 520A / 520C and 520B / 520D will correspond to two different tags. Specifically, the tag unit 222 can determine whether the data segments belong to the same tag based on a preset interval value. For example, if the difference in the variables of the data segments is within "50 RPM," the same tag can be set. The size of the interval value can be adjusted as needed.

[0073] In the embodiment shown in Figure 5B, segment data 520A-520D are tagged based on a single variable (e.g., rotational speed). In other embodiments, tagging unit 222 analyzes multiple variables to set tags. Referring to Figure 5C, which shows a schematic diagram of multiple segment data 520E-520F, segment data 520E-520F form different data points on the coordinate axis. Tagging unit 222 analyzes "rotational speed" and "acceleration time" to determine the tags for segment data 520E-520F. As shown in Figure 5C, segment data 520E is "rotational speed increases from 0 to 3000 RPM within 67 milliseconds," and segment data 520F is "rotational speed increases from 0 to 3500 RPM within 67 milliseconds." If the distance between data points is less than a preset interval value, they can be classified into the same data range DR, that is, they can be set with the same tag (e.g., "67 milliseconds, 3000 RPM").

[0074] Next, the "Data Management Program S420" will be described. In step S421, the management unit 223 receives the segment data 520 and determines whether the tag in each segment data 520 has been defined and exists. As shown in Figure 3, the control device 200 can establish multiple databases DB, each database DB corresponding to different tags. That is, segment data with the same tag are stored in the same database DB.

[0075] If the management unit 223 determines that the tag for the currently received segment data 520 already exists, then in step S422, the management unit 223 will increase the usage rate of the corresponding tag and classify the segment data into the corresponding database as a statistical data point. In some embodiments, the management unit 223 may also simultaneously determine whether the database is stable (e.g., determine the number of all segment data in the database).

[0076] If the management unit 223 determines that the tag of the currently received segment data 520 does not exist, then in step S423, the management unit 223 will further determine whether the number of databases in the control device 200 has reached the upper limit (e.g., determine whether the number of databases has reached the preset value).

[0077] If the number of databases has not yet reached the upper limit, in step S424, the management unit 223 will add a new database corresponding to the new tag in the control device 200. If the number of databases has reached the upper limit, in step S425, the management unit 223 will first remove the database corresponding to the tag with the lowest usage rate, and then add a new database corresponding to the new tag.

[0078] Next, the "feature extraction procedure S430" will be described. The feature extraction unit 224 is used to analyze each segment of data / data point to obtain the feature values ​​of each segment of data / data point. In step S431, the feature extraction unit 224 is used to calculate the average of multiple data points in the same database (e.g., the average of speed feedback values, the average of current) and / or the covariance matrix, and then generate feature values ​​accordingly.

[0079] In step S432, the feature extraction unit 224 calculates the Mahalanobis distance for each data point based on the covariance matrix. The Mahalanobis distance can then be used as the aforementioned "feature value of segment data / data point". Mahalanobis distance is a statistical method used to calculate the difference or similarity between "data point" and "data distribution". Since those skilled in the art can understand how Mahalanobis distance is calculated, it will not be described in detail here.

[0080] Figure 6 shows a schematic diagram of multiple data points P60 classified under the same label. The position of data point P60 on the coordinate axis is determined by two different variables (variable one and variable two in Figure 6, which can be time, rotational speed, current, torque, etc.). In one embodiment, when calculating the Mahalanobis distance of a data point, the feature extraction unit 224 first selects a subset of data points from these data points P60 that were generated earliest as multiple historical data points. Then, using these historical data points as a reference, it calculates the Mahalanobis distance of other data points P60. For example, it calculates the Mahalanobis distance of other data points P60 with the center of these historical data points as the origin. As shown in Figure 6, the multiple data points included in the historical data range P61 shown in the figure are the historical data points, and the number of data points with the "earliest generation time" (e.g., the earliest 10) or the time segment (the first 10% of the generation time) can be adjusted according to requirements.

[0081] The advantage of using Mahalanobis distance for analysis is that it considers the correlation between multiple variables and is not affected by the data scale between different variables, thus providing a more comprehensive analysis and more accurate detection of outliers. As shown in Figure 6, the distribution range of 610 is a reasonable range for analyzing the similarity between data points P60 using Mahalanobis distance; the distribution range of 620 is a reasonable range for analyzing the similarity between data points P60 using Euclidean distance. It is clear from the figure that the distribution range of 610 more accurately reflects the distribution trend of data points P60 and identifies outlier data points.

[0082] The following describes the "judgment procedure S440". In step S441, the anomaly judgment unit 225 compares the obtained "feature values ​​of data points / segments" with the set anomaly threshold and calculates the number of feature values ​​(e.g., Mahalanobis distance) of multiple data points that exceed the anomaly threshold.

[0083] In step S442, the anomaly judgment unit 225 determines whether the number of "feature values ​​exceeding the anomaly threshold (i.e., abnormal feature values)" exceeds a warning threshold. If so, a notification message is generated in step S443. Figure 7 shows a schematic diagram of determining whether abnormal feature values ​​exceed the warning threshold in the same database (corresponding to the same label). The horizontal axis represents multiple feature values, and the vertical axis represents the numerical value of the feature values. As shown in the figure, among the 22 feature values, three feature values ​​exceed the anomaly threshold V7. If the warning threshold set in the control device 200 is "2", it means that the analysis result of the data points in this database is "abnormal operating condition".

[0084] This disclosure segments the raw data and identifies the characteristics of each segment to categorize it into different databases. Based on this, the control device can make judgments about the overall behavior of the system, rather than being limited to a single abnormal state. In other words, the control device can analyze different databases in the same way to detect a variety of different anomalies. As mentioned earlier, different databases correspond to different segments of data (raw data), such as motor position, motor current, motor voltage, vibration level, torque, etc.

[0085] On the other hand, since the control device 200 is directly located in the controller 110 or the driver 120, there is no need to transmit large amounts of raw data frequently, nor is there a need to set up an additional edge processor. Accordingly, the cost, hardware requirements, and system complexity of the control system 100 can be improved.

[0086] Furthermore, in one embodiment, during the execution of the monitoring method, the control device 200 may simultaneously calculate the tag usage rate, tag recognition rate, database usage rate, and / or the health status of the control system 100. The control device 200 may transmit the aforementioned statistical data (i.e., tag usage rate, tag recognition rate, database usage rate, and health status) to the display screen of the control system 100 or the terminal device of the administrator. In one embodiment, the anomaly judgment unit 225 is used to calculate the tag recognition rate, tag usage rate, database usage rate, and / or health status, and may generate notification information based on one or more of them. The definitions of the aforementioned statistical data are explained below:

[0087] Tag recognition rate: defined as "the proportion of tags with diagnostic capabilities". As mentioned earlier, segment data of the same tag are stored in the same database DB, and the anomaly detection unit 225 analyzes the data in the same database DB (i.e., the feature values ​​of data points or segment data) to determine whether an anomaly has occurred. During monitoring, when any database is determined to be abnormal by the anomaly detection unit 225 (e.g., the number of feature values ​​exceeding the anomaly threshold is too high), the tags corresponding to this part of the database DB are defined as "with diagnostic capabilities" or diagnostic tags, meaning that the database corresponding to this type of tag can identify whether it is abnormal. Therefore, if M out of N tags have diagnostic capabilities, the tag recognition rate is "M / N". The calculation method is the ratio of the number of diagnostic tags to the total number of tags.

[0088] Label usage rate: Defined as "the proportion of labels used". That is, the ratio of the number of times each label is set to the total number of data segments. For example: if in W analyses, label unit 222 defines / sets three different labels in W data segments 520, and the number of times these three labels are set / used are Z1, Z2, and Z3 respectively (that is, Z1 to Z3 can be any value not greater than W), then the usage rates of these three labels are "Z1 / W", "Z2 / W", and "Z3 / W".

[0089] Database utilization rate: defined as "the proportion of data recorded in the database". For example, if a database can store a maximum of X data segments, and the number of data segments currently stored is Y, then the database utilization rate is "Y / X".

[0090] Health level: Defined as "the similarity between newly added data and the database". That is, if a database contains K data segments, the similarity between these K data segments and newly added data is the health level of the new data. Another definition of health level is "the proportion exceeding the anomaly threshold". As shown in Figure 7 above, data points corresponding to feature values ​​exceeding the anomaly threshold are called "abnormal data points", and data points corresponding to feature values ​​not exceeding the anomaly threshold are called "normal data points" (i.e., data points that do not meet the anomaly threshold). Health level can be calculated by dividing the number of normal data points in the same database by the total number of all data points (data segments).

[0091] Overall, this disclosure not only optimizes monitoring methods but also reduces system cost and complexity. Furthermore, through tag and database categorization management, users can easily understand the current status and calculation results of each database using metrics such as database usage rate and tag usage rate, thereby improving management convenience. Users can also adjust various control parameters according to actual needs, such as the aforementioned preset values, interval values, time segments, anomaly thresholds, and warning thresholds, providing flexible control flexibility.

[0092] The elements, method steps, or technical features in the foregoing embodiments can be combined with each other, and are not limited to the order of textual description or the order of presentation of the drawings in this disclosure.

[0093] Although the present disclosure has been described above with reference to embodiments, it is not intended to limit the present disclosure. Any person skilled in the art may make various modifications and refinements without departing from the spirit and scope of the present disclosure. Therefore, the scope of protection of the present disclosure shall be determined by the scope of the appended claims.

Claims

1. A control device, comprising: A tag unit is used to obtain multiple segments of data from multiple raw data, and to analyze each of the multiple segments of data to define tags for portions of the multiple segments of data respectively. A management unit, coupled to the tag unit, is used to receive the plurality of segment data from the tag unit, wherein the management unit is used to classify the portions of the plurality of segment data that have the same tag into a database, so as to serve as multiple data points in the database; A feature extraction unit, coupled to the management unit, is used to analyze multiple feature values ​​of the multiple data points; as well as An anomaly detection unit, coupled to the feature extraction unit, is used to compare the multiple feature values ​​of the multiple data points with an anomaly threshold to generate notification information.

2. The control device according to claim 1, wherein the feature extraction unit is used to calculate the average and covariance matrix of the plurality of data points to generate the plurality of feature values.

3. The control device according to claim 2, wherein the feature extraction unit is used to calculate multiple Mahalanobis distances of the multiple data points based on the covariance matrix, as the multiple feature values.

4. The control device according to claim 3, wherein the anomaly determination unit is used to determine the number of Mahalanobis distances exceeding the anomaly threshold.

5. The control device according to claim 3, wherein the feature extraction unit is used to select the portion of the plurality of data points that was generated earliest as a plurality of historical data points, so as to calculate the plurality of Mahalanobis distances based on the plurality of historical data points.

6. The control device according to claim 1, wherein the plurality of raw data is used to control a servo motor, generated by the servo motor, or generated by an external sensor.

7. The control device according to claim 1, wherein the tag unit defines multiple tags for the multiple data segments, and the anomaly judgment unit is used to calculate the tag usage rate, wherein the tag usage rate is the ratio of the number of each of the multiple tags to the number of the multiple data segments.

8. The control device according to claim 1, wherein the anomaly judgment unit is used to calculate the health level, the health level being the number of multiple normal data points that do not meet the anomaly threshold divided by the number of the multiple data points.

9. The control device according to claim 1, wherein the tag unit is configured to acquire the plurality of segment data from the plurality of raw data according to time segments.

10. The control device of claim 1, wherein the tag unit is configured to analyze the waveform of the plurality of data segments to define a tag for each of the plurality of data segments.

11. The control device according to claim 1, wherein the tag unit defines multiple tags for the plurality of data segments, the plurality of data segments are classified into multiple databases, and portions of the plurality of databases judged to be abnormal correspond to multiple diagnostic tags among the plurality of tags; and The anomaly detection unit is used to calculate the tag recognition rate, which is the ratio of the number of the plurality of diagnostic tags to the plurality of tags.

12. A monitoring method, comprising: Multiple data segments are obtained from multiple raw data sources through a control device; Analyze each of the multiple data segments to define a label for each of the multiple data segments respectively; The portions of the multiple data segments that have the same label are classified into a database to serve as multiple data points in the database; Analyze each of the plurality of data points to obtain multiple feature values ​​for the plurality of data points; as well as The multiple feature values ​​of the multiple data points are compared with anomaly thresholds to generate notification information.

13. The monitoring method of claim 12, wherein the method of analyzing each of the plurality of data points to obtain the plurality of feature values ​​of the plurality of data points comprises: Calculate the mean and covariance matrix of the multiple data points to generate the multiple eigenvalues.

14. The monitoring method of claim 13, wherein the method of analyzing each of the plurality of data points to obtain the plurality of feature values ​​of the plurality of data points further comprises: The Mahalanobis distances of the multiple data points are calculated based on the covariance matrix and used as the multiple eigenvalues.

15. The monitoring method according to claim 14, wherein the method of comparing the plurality of feature values ​​of the plurality of data points with the anomaly threshold to generate the notification information comprises: Determine the number of anomaly thresholds among the plurality of Mahalanobis distances; and The notification message is generated when the number exceeds the warning threshold.

16. The monitoring method of claim 14, wherein the method of analyzing each of the plurality of data points to obtain the plurality of feature values ​​of the plurality of data points further comprises: Select the earliest generated portion of the plurality of data points as the plurality of historical data points; and The multiple Mahalanobis distances are calculated based on the multiple historical data points.

17. The monitoring method according to claim 12, wherein the plurality of raw data is used to control a servo motor, generated by the servo motor, or generated by an external sensor.

18. The monitoring method according to claim 17, further comprising: Multiple control commands are obtained through the control device within the controller, wherein the multiple control commands are used to control the servo motor; and The multiple control commands are used as the multiple raw data.

19. The monitoring method according to claim 12, wherein the method for obtaining the plurality of segment data from the plurality of raw data comprises: Based on the time period, the multiple segment data are obtained from the multiple original data.

20. The monitoring method of claim 12, wherein the method of analyzing each of the plurality of segment data to define the tag for each of the plurality of segment data respectively comprises: Analyze the waveforms of the multiple data segments to define the label for each of the multiple data segments.

Images