TWI768606B - System and method for monitoring sensor - Google Patents

Abstract

The present disclosure relates to a system for monitoring sensor. The system includes an acquiring module, a clustering module, and a determining module. The acquiring module is configure to acquire sensor data of a plurality of sensors. The clustering module is configure to cluster the plurality of sensors according to the sensor data. The determining module is configure to determine whether the cluster of the plurality of sensors changes in comparison with a standard cluster. The present disclosure also relates to a method for monitoring sensor.

Description

Sensor monitoring system and method

The present invention relates to a sensor monitoring system and method, and more specifically relates to an automatic monitoring system and method for a sensor.

Wireless Sensor Network (WSN) technology can be applied in various fields, using multiple sensors to work together to monitor physical or environmental conditions in different locations, in the Internet of Things (IoT) and automated production lines Development plays a key role.

When an abnormal signal occurs in the sensor, manual interpretation is usually required to identify the abnormal signal as an environmental variable or a misjudgment. In addition, in order to improve the reliability of the sensors, it is usually necessary to manually check each sensor on a regular basis, which is not only time-consuming and labor-intensive, but also difficult to calibrate or replace the sensors in real time. Therefore, how to reduce the cost of manual interpretation and detection of the sensor and ensure the reliability of the sensor is the problem to be solved by the present invention.

An embodiment of the present disclosure relates to a sensor monitoring system, including a capture module, a clustering module, and a determination module. The capture module is configured to capture the plurality of sensor data of the plurality of sensors. The clustering module is configured to classify the sensor clusters according to the sensor data. The determination module is configured to determine whether the cluster classification of the sensors has a cluster change compared to a standard cluster classification.

An embodiment of the present disclosure relates to a sensor monitoring method, including capturing a plurality of sensor data of a plurality of sensors, clustering the sensors according to the sensor data, and determining Whether the cluster classification of the sensors has a cluster change compared to a standard cluster classification.

1: Sensor monitoring system

10: Capture module

11: Swarm Module

12: Judgment Module

13: Detection module

2: Sensor monitoring method

20: Steps

21: Steps

22: Steps

23: Steps

24: Steps

25: Steps

26: Steps

27: Steps

28: Steps

7: Field

70: Location

71: Location

A-Z: Sensors

a-h: sensor

Various aspects of the embodiments are discussed below with reference to the drawings, which are not drawn to scale, are illustrative only, and do not limit the scope of the invention. The reference numerals used in the drawings and the description are only examples, and do not limit the scope of the present invention. The same or similar components are represented by the same reference numerals, wherein: FIG. 1 shows a schematic diagram of a sensor monitoring system according to some embodiments of the present disclosure; FIG. 2 shows a sensing system according to some embodiments of the present disclosure. Figure 3 shows a schematic diagram of one or more steps in a sensor monitoring method according to some embodiments of the present disclosure; Figure 4 shows a sensing method according to some embodiments of the present disclosure Figure 5 shows a standard cluster classification diagram according to some embodiments of the present disclosure; Figure 6 shows a cluster classification diagram according to some embodiments of the present disclosure; FIG. 7 is a schematic diagram showing the position of the sensor according to some embodiments of the present disclosure; and FIG. 8 is the average value, the average value of the change amount, and the median value of the sensor data according to some embodiments of the present disclosure. , kurtosis value, deviation value, and standard deviation.

Referring to FIG. 1 , FIG. 1 is a schematic diagram of a sensor monitoring system 1 according to some embodiments of the present disclosure. The sensor monitoring system 1 may include a capture module 10 , a clustering module 11 , a determination module 12 , and a detection module 13 .

In some embodiments, the capture module 10 , the clustering module 11 , the determination module 12 , and the detection module 13 can be integrated to realize all their functions by a single electronic device. In some embodiments, the capturing module 10 , the clustering module 11 , the determination module 12 , and the detection module 13 may be dispersed in several devices, and all functions are implemented by a plurality of devices.

Terms used in this disclosure, such as "system," "module," and "device," refer to hardware, software, firmware, or any combination of these, associated with a computer or server. For example, the sensor monitoring system 1 may include software executing on a computer, a computer executing the software, or both software and a computer. For example, the capture module 10, the clustering module 11, the determination module 12, and the detection module 13 may include software executing on a computer, a computer executing the software, or both software and a computer. In some embodiments, computers and servers may have one or more processors configured to execute computer-executable instructions stored in memory, and processors may include electronic integrated circuits that apply electronic signals to perform logical operations circuit.

According to some embodiments of the present disclosure, the sensor monitoring system 1 can be applied in various fields to monitor sensors therein. The aforementioned fields may include, for example, but not limited to, factories, laboratories, and other indoor spaces. However, the aforementioned field is not limited to a confined space, but includes a virtual space defined by the detection range of the sensor. For example, a plurality of sensors work together to monitor physical or environmental conditions at different locations, and the detection range of the plurality of sensors is the applicable field of the sensor monitoring system 1 . The aforementioned sensors may include particulate or particle sensors such as, but not limited to, falling dust sensors, odor particle sensors, temperature sensors, humidity sensors, sound sensors, or Sensors with spatially overlapping correlation characteristics of other signals.

In some embodiments, the capture module 10 may be configured to capture sensor data for the sensors. Clustering module 11 may be configured to classify sensor clusters according to sensor data. The determination module 12 may be configured to determine whether the cluster classification of the sensor has a cluster change compared to the standard cluster classification. The detection module 13 may be configured to detect whether the sensor data exceeds a standard value.

Embodiments of the sensor monitoring system 1 and the sensor monitoring system 2 will be described below with reference to FIGS. 2 to 8 .

FIG. 2 is a flowchart of a sensor monitoring method 2 according to some embodiments of the present disclosure. The sensor monitoring method 2 can be implemented by the sensor monitoring system 1 .

In step 20, a capture module (eg, capture module 10 of FIG. 1) may capture sensor data of the sensor.

In some embodiments, the sensor data may include the relationship of the signal generated by the sensor to time. Since the signal values (or data points) can be ordered in time, sensor data can also be referred to as time series. In some embodiments, sensor data may include, for example, but not limited to, number of particles detected per unit time, odor (eg, concentration), temperature, humidity, or sound (eg, decibels).

In step 21, step 22, and step 23, the clustering module (eg, the clustering module 11 of FIG. 1) can calculate the similarity of the captured sensor data, and cluster the sensors according to the similarity Sort until the number of clusters drops to the number of needs.

In some embodiments, the similarity of the sensor data may represent the similarity of the change trend of the sensor data in a unit time. The smaller the time difference between sensor data changes, the higher the similarity. Specifically, after an event (such as dust) is transmitted through the air, the propagation path will pass through two adjacent sensors in sequence, so the signals of the two sensors will have similar performance, but The time series are different, so through such a comparison concept, it can be known which sensors are adjacent and which sensors are in different clusters.

In some embodiments, the similarity of sensor data may be calculated using Dynamic Time Warping. For example, as shown in FIG. 3, the two sets of sensor data (ie, the two time series) are normalized to align their time. The respective initial data points i and j and the next data points i-1 and j-1 of the two sets of sensor data can be obtained. Then, the distance between the data points of the two sets of sensor data is calculated, and the shortest distance is found. For example, calculate the distance between data point i and data point j-1 (denoted as D(i,j-1)), D(i-1,j), and D(i-1,j-1), where If D(i, j-1) is the shortest distance, the next data points of i and j-1 (ie, i-1 and j-2), respectively, are regarded as starting data points. By repeating the above steps until the end of the data point, a plurality of shortest distances can be obtained, and the sum of the plurality of shortest distances is the distance between the two sets of sensor data. If the distance is smaller (or the value is smaller), it means that the time difference between the data points is smaller, and it can be determined that the similarity of the two sets of sensor data is higher.

In some embodiments, calculating the distances of any two sets of sensor data can draw a distance matrix, and can use Hierarchical Agglomerative Clustering and Average-linkage Agglomerative Algorithm according to the distance matrix Classify sensor clusters. For example, as shown in FIG. 4 , P1 , P2 , P3 , and P4 are four sensors (grouped in four groups), respectively. First, the P1 and P2 with the highest similarity (that is, the minimum distance "2") are merged into the first-order cluster, and the similarity of the first-order cluster is "2" (that is, the distance between P1 and P2 is "2") . P1 and P2 in the first-order cluster are connected by a node. Then, the P3 with the highest similarity with the first-order cluster (that is, the minimum distances "4" and "3") is aggregated with the first-order cluster to form the second-order cluster, and the similarity of the second-order cluster is " 3.5" (that is, the average "3.5" of the distance "4" between P3 and P1 and the distance "3" between P3 and P2). P3 of the second-order cluster is connected with P1 (or P2) of the first-order cluster by two nodes. Then will be clustered with the second order The P4 with the highest similarity (that is, the minimum distances "10", "13", "22") aggregates with the second-order group and becomes the third-order cluster, and the similarity of the third-order cluster is "15" (ie, The average of the distance "10" between P4 and P1, the distance "13" between P4 and P2, and the distance "22" between P4 and P3). P4 in the third-order cluster is connected with P1 (or P2) in the first-order cluster by three nodes.

After clustering and classifying all the sensors, a cluster classification map can be generated, as shown in the cluster classification map on the right side of Figure 4. The standard cluster classification map of FIG. 5 and the cluster classification map of FIG. 6 can also be generated in the manner described above. However, the present disclosure is not limited thereto, and in some embodiments, other feasible algorithms can also be used to calculate the similarity of sensor data. In addition, other feasible algorithms can also be used to classify the sensor clusters.

The sensor monitoring system 1 and the sensor monitoring system 2 generate a standard cluster classification map (such as the standard cluster classification map in FIG. 5 ) through the above method, and determine the sensor based on the standard cluster classification map (such as the standard cluster classification map in FIG. 6 ). Whether the sensor in the cluster classification diagram has a cluster change, and then determine whether the sensor is abnormal. Since the sensors may be added or removed, replaced, changed positions, etc., the standard cluster classification map can be corrected in real time by feedback of the results of each determination.

At step 24 , a determination module (eg, determination module 12 of FIG. 1 ) may determine a sensor (eg, the cluster classification map of FIG. 6 ) based on a standard cluster classification map (eg, the standard cluster classification map of FIG. 5 ) sensor) has a cluster change. If there is no cluster change, the determination module (eg, the determination module 12 in FIG. 1 ) can determine that the sensor is not abnormal (eg, step 28 ). If there is a cluster change, the determination module (eg, the determination module 12 in FIG. 1 ) can further determine whether the cluster change exceeds a preset value (eg, step 25 ).

The case of no cluster change includes, for example, the sensor H is connected to the sensor I with one node in both FIG. 5 and FIG. 6 and the similarity is within the preset value (belonging to the same cluster), so it is determined that The module 12 can determine that the sensor H has no cluster change. In addition, sensor H and sensor I are in 5 and 6 are located in the first-order cluster, and are connected to each other by a node (belonging to the same cluster). , so the determination module 12 can determine that the sensor H and the sensor I have no cluster change. Further, sensor H and sensor I are located in the first-order cluster in the standard clustering classification diagram of FIG. 5, and are connected to each other by a node (belonging to the same cluster), which can represent sensors H and The similarity of the change trend of the signal of sensor 1 in unit time is higher than that of other sensors. If there is no abnormality in the two sensors, the signals of the two sensors will also belong to the same in Figure 6. crowd gathering. Conversely, if the signals of the two sensors also belong to the same cluster in FIG. 6 , it can be determined that neither of the two sensors is abnormal (eg, step 28 ).

When there is a cluster change, in step 25 , the determination module (eg, the determination module 12 in FIG. 1 ) can further determine whether the cluster change exceeds the preset value. If the predetermined value is not exceeded, the determination module (eg, the determination module 12 in FIG. 1 ) may determine that the sensor is not abnormal (eg, step 28 ). If it exceeds the preset value, the detection module (eg, the detection module 13 in FIG. 1 ) can further determine whether the sensor data exceeds the standard value (eg, step 26 ).

In some embodiments, whether the predetermined value is exceeded may be determined through similarity. For example, sensor U and sensor V are located in the first order cluster in the standard cluster classification diagram of FIG. 5 and are connected to each other by a node (belonging to the same cluster). Sensor U and Sensor V are connected to each other by two nodes (belonging to different clusters) in FIG. 6, from the first order cluster to the second order cluster, the clusters of sensors U and sensor V The cluster changes in FIGS. 5 and 6 , so the determination module 12 can determine that the sensor U and the sensor V have cluster changes. However, the similarity of the second-order clustering of the sensor U and the sensor V in FIG. 6 is "5". Assuming that the default value is the similarity of "10", the sensor U and the sensor The similarity of V is within the preset value, so the determination module 12 can determine that the cluster change of the sensor U and the sensor V does not exceed the preset value. Therefore, the determination module 12 can determine that neither of the two sensors is abnormal (eg, step 28 ).

In some embodiments, whether the predetermined value is exceeded may be determined by other sensors. For example, sensor Q and sensor S are connected to each other by three nodes (belonging to different clusters) in the standard cluster classification diagram of FIG. 5, and the similarity is within "10". Sensors Q to Y with a similarity within "10" in FIG. 5 can be regarded as the same cluster. In FIG. 6 , the sensor Q and the sensor W are connected to each other as a node (belonging to the same cluster), and the similarity is within “10”. The sensor Q is connected to the same cluster of sensors as a node in both Figure 5 and Figure 6, so it can be determined that the cluster change of the sensor Q does not exceed the preset value, and the sensor Q is not abnormal ( For example, step 28).

In some embodiments, a group of sensors can be preset as the same cluster, and when any two in the same cluster are connected to each other by a node or the similarity of the cluster is within a preset value, it is determined that the two sensors are The clustering changes of those did not exceed the default value. As shown in FIG. 7 , in field 7, sensor H and sensor 1 are located at position 70, and sensor H and sensor 1 can be set to the same cluster. The sensor Q, the sensor S, the sensor T, the sensor U, the sensor V, the sensor W, and the sensor c are located at the position 71, and can also be set to the same cluster. In FIG. 6 , the sensor S and the sensor c are connected to each other as a node (belonging to the same cluster), and the similarity is within “10”. Referring to the field of FIG. 7 , since the sensor S and the sensor c are at similar positions, if similar signal responses occur, it should be possible to determine that the sensors are not abnormal. In FIG. 6 , the sensor S is connected with the sensors of the same cluster as a node, so it can be determined that the cluster change of the sensor S does not exceed the preset value, and the sensor S is not abnormal (for example, step 28 ) .

The case where the clustering change exceeds the preset value includes, for example, that the sensor T is clustered with different sensors in FIG. 5 and FIG. 6 , the clustering order is changed, and thus has a clustering change. And the sensor T is not connected to the same cluster of sensors as a node in FIG. 6, so it can be determined that the cluster change of the sensor T exceeds the preset value, and the sensor data of the sensor T is abnormal .

When the cluster change exceeds the preset value, in step 26, the detection module (for example, the The detection module 13) can further determine whether the sensor data exceeds the standard value. For example, it is determined whether the average value, the average value of the change amount, the median value, the kurtosis value, the deviation value, and the standard deviation of the sensor data of the detection sensor T exceed the standard value (as shown in FIG. 8 ). In some embodiments, historical sensor data for sensor T may be used for the standard value. In some embodiments, the standard value can also use sensors at similar positions (for example, the sensor Q, the sensor S, the sensor U, the sensor V, the sensor W, the sensor in FIG. 7 , the sensor W, the sensor c) Historical sensor data. If it is found that the sensor data does not exceed the standard value after detection, it can be determined that the sensor is not abnormal (eg, step 28 ). If it is found that the sensor data exceeds the standard value after detection, it can be determined that the sensor is abnormal (eg, step 27 ). In some embodiments, a clustering module (eg, clustering module 11 of FIG. 1 ) may be configured to modify the standard clustering classification based on the above determinations.

The sensor monitoring system 1 and the sensor monitoring system 2 are used to monitor the sensors, which greatly reduces the cost of manual periodic interpretation and detection of the sensors. When an abnormal signal occurs in the sensor, the signal of the sensor with high similarity is used to determine whether the abnormal signal is an environmental variable or a misjudgment (also known as a spatial abnormality). If it is a spatial abnormality, the sensor is judged. Whether the data exceeds the standard value of historical sensor data (also known as time anomaly). When there are sensors judged to be both spatial anomalies and temporal anomalies, the sensor monitoring system 1 can be set to issue an alarm to remind the sensors to be calibrated or replaced, so as to ensure the reliability of the sensors.

It will be appreciated that the embodiments of the systems and methods discussed herein are not limited to the details of construction and/or configurations described herein or depicted in the drawings, but may be practiced or carried out in various ways. The specific embodiments herein are illustrative only and are not intended to limit the invention.

Also, the phraseology and terminology used herein is exemplary only and not intended to limit the invention. The singular or plural form is exemplary only and is not intended to limit the system or method of the present invention, its elements, components, or steps. "includes", "includes", "includes" Has," "contains," "involves," and other similar terms cover the items listed thereafter, equivalents, and additional items. "Or" and other similar terms may be taken to indicate any of the described items.

1: Sensor monitoring system

10: Capture module

11: Swarm Module

12: Judgment Module

13: Detection module

Claims (16)

A sensor monitoring system, comprising: a capture module configured to capture a plurality of sensor data from a plurality of sensors; a cluster module configured to capture data from the plurality of sensors classifying the plurality of sensor clusters; and a determination module configured to determine whether a clustering classification of the plurality of sensors has a clustering change compared to a standard clustering classification; wherein the The plurality of sensors includes a first sensor and a second sensor; and wherein the determination module is further configured to: if the clustering of the first sensor in the standard clustering classification is not equal to If the first sensor is clustered in the clustering classification, it is determined that the first sensor has a clustering change; and if the first sensor and the second sensor are in the standard clustering classification If the order difference is not equal to the order difference between the first sensor and the second sensor in the cluster classification, it is determined that the first sensor and the second sensor have a cluster change . The sensor monitoring system of claim 1, wherein the plurality of sensor data includes the relationship between the signals generated by the plurality of sensors and time. The sensor monitoring system of claim 2, wherein the clustering module is configured to classify the plurality of sensor clusters according to the relationship between the signals generated by the plurality of sensors and time. The sensor monitoring system of claim 3, wherein the clustering module is configured to calculate signal similarity based on signals generated by the plurality of sensors. The sensor monitoring system of claim 2, wherein the signal includes the number of particles, odor, temperature, humidity or sound detected by the plurality of sensors per unit time. The sensor monitoring system of claim 1, wherein the cluster classification of the plurality of sensors includes M-order clustering, the standard cluster classification includes N-order clustering, and M and N are positive Integer. The sensor monitoring system of claim 6, wherein the clustering module is configured to calculate the signal similarity according to the signals generated by the plurality of sensors, and assign the first group of sensors with the highest signal similarity The sensors are combined into a first-order cluster, and the second group of sensors with the next highest signal similarity are aggregated with the first-order cluster to form a second-order cluster. The sensor monitoring system of claim 7, wherein the clustering module is configured to make the clustering order of the plurality of sensors less than or equal to M by means of cluster merging. The sensor monitoring system of claim 1, wherein the determination module is further configured to determine the first sensor if the cluster change of the first sensor or the second sensor is greater than a predetermined value The sensor or the second sensor has an abnormality. The sensor monitoring system of claim 1, wherein the clustering module is configured to be based on the The standard taxonomic cluster is corrected by determining the determination result of the module. The sensor monitoring system according to claim 1, further comprising: a detection module for detecting the average value, the average value of the change amount, the median value, the kurtosis value, the deviation value, and the standard value of the sensor data Whether at least one of the differences exceeds a standard value. A sensor monitoring method, comprising: acquiring a plurality of sensor data of a plurality of sensors; clustering and classifying the plurality of sensors according to the plurality of sensor data; and determining the plurality of sensors whether a clustering classification of the sensors has a clustering change compared to a standard clustering classification; wherein the plurality of sensors includes a first sensor and a second sensor; and wherein the sensor The sensor monitoring method further includes: if the clustering of the first sensor in the standard clustering classification is not equal to the clustering of the first sensor in the clustering classification, determining that the first sensor has cluster change; and if the cluster change of the first sensor or the second sensor is greater than a predetermined value, determining that the first sensor or the second sensor has an abnormality. The sensor monitoring method of claim 12, wherein the plurality of sensor data includes the relationship between the signals generated by the plurality of sensors and time. The sensor monitoring method of claim 13, wherein the signal includes the number of particles, odor, temperature, humidity or sound detected by the plurality of sensors per unit time. The sensor monitoring method of claim 12, wherein the clustering classification of the plurality of sensors includes M-order clustering, the standard clustering classification includes N-ordering clustering, and M and N are positive integers . The sensor monitoring method of claim 12, further comprising: detecting an average value, a change average value, a median value, a kurtosis value, a deviation value, and a standard deviation of the plurality of sensor data Whether at least one exceeds a standard value.

Images