High-frequency sensor single-point gross error identification method, computer and storage medium

Abstract

The application discloses a high-frequency sensor single-point gross error identification method, a computer and a storage medium, relates to the technical field of sensor monitoring, and comprises the following steps: acquiring original time sequence data of all high-frequency sensors in a monitoring system, performing data interception on the original time sequence data of the high-frequency sensors and performing statistical processing; counting the number of out-of-limit sensors whose real-time statistical values are greater than a preset historical threshold value, performing data preprocessing on the preprocessed time sequence data according to a preset time sequence identification rule, and performing dynamic time warping calculation on the preprocessed time sequence data to obtain a current similarity distance; and judging whether the number of out-of-limit sensors is less than a preset sensor quantity threshold value, if the judgment is yes, performing abnormality discrimination processing on a single high-frequency sensor, and otherwise, performing multi-sensor abnormality discrimination processing. The application can improve the alarm accuracy of the high-frequency sensors in the monitoring system, and avoid unnecessary alarm processing and resource waste caused by false alarms.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sensor monitoring, in particular to a high-frequency sensor single-point gross error identification method, a computer and a storage medium. BACKGROUND

[0002] In the automatic monitoring system of bridge structure, various types of high-frequency (acquisition frequency greater than or equal to 1 Hz) sensors are the key components of the automatic monitoring system. The structural response index abnormal duration under the condition of bridge special events (such as vehicle / ship collision, earthquake, etc.) generally will not be less than 1 second, and the specific data waveform has the process of rising and falling, and will not suddenly start and end. High-frequency sensors can generate more data points to reflect the response process of the structure under special events, and these data are crucial for evaluating the safety and health status of the structure. However, in the actual monitoring process, due to the influence of various factors such as noise of the sensor itself, error in the data acquisition and transmission process, environmental interference, etc., single-point over-limit alarm may occur. This false alarm not only interferes with the judgment of the real state of the bridge by the monitoring personnel, but also may lead to unnecessary waste of resources and start of emergency response measures.

[0003] At present, for the abnormal detection and false alarm processing of sensor data, the common methods include:

[0004] 1. Simple judgment based on threshold, which usually only simply sets a fixed maximum value or minimum value, and triggers an alarm when the data exceeds this range. Although this method is simple, it cannot effectively distinguish between real structural abnormalities and false alarms caused by noise or errors.

[0005] 2. Abnormal detection algorithm based on machine learning, which can improve the accuracy of abnormal detection to some extent, but usually needs a large amount of labeled data for training, and for single-point gross error of this specific type of abnormality, its detection effect also has certain limitations, and the single-point gross error identification precision is limited. SUMMARY

[0006] The technical problem to be solved by the present application is to provide a high-frequency sensor single-point gross error identification method, a computer and a storage medium, which can improve the alarm accuracy and reliability of high-frequency sensors, avoid unnecessary alarm processing and resource waste caused by false alarms of high-frequency sensors, and thus improve the alarm accuracy of the bridge monitoring system.

[0007] In order to solve the above technical problems, the application provides a high-frequency sensor single-point gross error identification method, comprising the following steps: acquiring original time sequence data of all high-frequency sensors in a monitoring system, and performing interception and statistical processing on the original time sequence data of the high-frequency sensors to obtain initial time sequence data and real-time statistical values; counting the number of sensors exceeding the limit of a preset historical threshold value, and performing preprocessing on the initial time sequence data of the high-frequency sensors exceeding the limit according to a preset time sequence identification rule to obtain preprocessed time sequence data of the high-frequency sensors; performing dynamic time warping calculation on the preprocessed time sequence data of the high-frequency sensors and preset sensor gross error sequence data to obtain the current similarity distance of the high-frequency sensors; judging whether the number of sensors exceeding the limit is less than a preset sensor quantity threshold value, if the judgment is yes, judging whether the current similarity distance of the high-frequency sensors is greater than the corresponding preset distance threshold value, if the judgment is yes, performing normal alarm processing, if the judgment is no, issuing a gross error alarm information and performing review processing; if the judgment is no, performing multi-sensor abnormality discrimination processing.

[0008] As an improvement of the above scheme, the step of intercepting and statistically processing the original time sequence data of the high-frequency sensors comprises: intercepting the original time sequence data according to a preset interval time interception rule to obtain initial time sequence data in a current preset time period; and statistically processing the initial time sequence data to obtain required real-time statistical values.

[0009] As an improvement of the above scheme, the real-time statistical values comprise any one or more of the average value, the change value and the maximum value of the initial time sequence data.

[0010] As an improvement of the above scheme, the step of preprocessing the initial time sequence data of the high-frequency sensors exceeding the limit according to a preset time sequence identification rule to obtain preprocessed time sequence data of the high-frequency sensors comprises: determining the time of the maximum value in the intercepted initial time sequence data, and acquiring sequence data of the time N seconds before and N seconds after the time, wherein N is a positive integer; counting the maximum value in the sequence data of each second to obtain 2N monitoring maximum values; and performing normalization processing on the 2N monitoring maximum values to obtain the preprocessed time sequence data.

[0011] As an improvement of the above scheme, the step of calculating the current similarity distance of the high-frequency sensor according to the preprocessed time series data of the high-frequency sensor and the preset sensor gross error sequence data includes: constructing a distance matrix from the preprocessed time series data and the preset sensor gross error sequence data; and dynamically finding a path with the minimum sum of matrix elements from the top-left element in the distance matrix as the starting point and the bottom-right element as the ending point, and calculating the current similarity distance of the path.

[0012] As an improvement of the above scheme, the step of performing multi-sensor abnormality discrimination processing includes: determining whether the current similarity distance of at least one high-frequency sensor in the over-limit high-frequency sensor is greater than the corresponding preset distance threshold; if the determination is yes, performing normal alarm processing; if the determination is no, determining the correlation between the over-limit multiple high-frequency sensors and performing alarm classification processing on the multiple high-frequency sensors.

[0013] As an improvement of the above scheme, the step of determining the correlation between the over-limit multiple high-frequency sensors and performing alarm classification processing on the multiple high-frequency sensors includes: obtaining label data of the over-limit high-frequency sensor, wherein the label data includes category setting data of multiple categories; determining whether the category setting data of any two high-frequency sensors are different; if the determination is yes, indicating that the multiple high-frequency sensor data are in an unrelated state, and performing normal alarm processing; if the determination is no, indicating that the multiple high-frequency sensor data are in a related state, issuing a gross error alarm information, and performing review processing.

[0014] As an improvement of the above scheme, the preset sensor gross error sequence data is [0, 1, 0].

[0015] The application further provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the high-frequency sensor single-point gross error identification method when executing the computer program.

[0016] The application further provides a readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the high-frequency sensor single-point gross error identification method when executed by a processor.

[0017] The application has the following beneficial effects:

[0018] The application can identify single-point gross error of raw time series data of high-frequency sensors in a bridge monitoring system in real time, determine whether the high-frequency sensor over-limit alarm is caused by single-point gross error, and if not, perform normal sensor alarm work, thereby improving the alarm accuracy and reliability of the high-frequency sensor, avoiding unnecessary alarm processing and resource waste caused by false alarm of the high-frequency sensor, and further improving the alarm accuracy of the bridge monitoring system. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of the single-point gross error identification method of the high-frequency sensor of the application;

[0020] Figure 2 is a flowchart of the multi-sensor abnormality discrimination processing step of the application;

[0021] Figure 3 is a flowchart of the step of determining the correlation between multiple high-frequency sensors that are over-limit and performing alarm classification processing on the multiple high-frequency sensors. DETAILED DESCRIPTION

[0022] To make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below with reference to the drawings. It is hereby declared that the up, down, left, right, front, back, inner and outer directions appearing or about to appear in the text of the application are based on the drawings of the application, and are not specific limitations on the application.

[0023] As shown in Figure 1 , the embodiment of the application provides a single-point gross error identification method of a high-frequency sensor, which comprises:

[0024] S101, obtaining raw time series data of all high-frequency sensors in a monitoring system, performing interception and statistical processing on the raw time series data of the high-frequency sensors to obtain initial time series data and real-time statistical values;

[0025] It should be noted that for the high-frequency sensors in the bridge monitoring system, the raw time series data generated during operation is obtained in real time, so that the real-time limit value of the high-frequency sensor is determined according to the raw time series data, which facilitates subsequent comparison with the historical limit value of the high-frequency sensor to determine whether an over-limit alarm condition occurs.

[0026] Specifically, the step of intercepting and statistically processing the raw time series data of the high-frequency sensor comprises:

[0027] Step 1, intercepting the raw time series data according to a preset interval time interception rule to obtain initial time series data in a current preset time period;

[0028] It should be noted that the original time series data continuously acquired is time period intercepted according to a preset interval time interception rule to obtain initial time series data in a preset time period in succession. Preferably, the preset interval time interception rule is to intercept the original time series data acquired in real time at an interval of 1 minute to obtain the original time series data in the corresponding 1 minute. Preferably, the interval time is not limited to 1 minute, and can be adjusted according to actual needs.

[0029] Step 2, statistics of the initial time series data are obtained to obtain the required real-time statistical value.

[0030] It should be noted that the original time series data in the current preset time period is statistically processed, such as calculating any one or more of the average value, the change value and the maximum value of the initial time series data, that is, the real-time statistical value includes any one or more of the average value, the change value and the maximum value of the sequence data.

[0031] S102, the number of sensors exceeding the limit of the real-time statistical value greater than the preset historical threshold is counted, and the initial time series data of the high-frequency sensor exceeding the limit is preprocessed according to a preset time sequence identification rule to obtain the preprocessed time series data of the high-frequency sensor;

[0032] It should be noted that the preset historical threshold is a threshold data determined according to the historical original time series data of the high-frequency sensor, which can be not affected by human experience factors, thereby improving the data processing accuracy. The threshold data includes any one or more of the average value, the change value and the maximum value determined according to the historical original time series data; by comparing the real-time statistical value with the preset historical threshold, it can be quickly determined whether the current monitoring data of the high-frequency sensor appears an alarm exceeding the limit, thereby facilitating the subsequent further analysis and processing of the alarm exceeding the limit to determine whether the alarm exceeding the limit is caused by a single-point gross error. In addition, the number of sensors exceeding the limit of the real-time statistical value greater than the preset historical threshold is counted, so as to facilitate alarm classification processing for the monitoring system of the high-frequency sensor with different number of exceeding limits, avoid unnecessary alarm processing and resource waste caused by false alarm, and further improve the alarm accuracy of the bridge monitoring system.

[0033] The step of preprocessing the initial time series data of the high-frequency sensor exceeding the limit according to the preset time sequence identification rule to obtain the preprocessed time series data of the high-frequency sensor includes:

[0034] Step 1, the time of the maximum value in the intercepted initial time series data is determined, and the sequence data of the time N seconds before and after the time is obtained, wherein N is a positive integer;

[0035] Step 2, the maximum value in the sequence data of each second is counted to obtain 2N monitoring maximum values;

[0036] Step 3, the 2N monitoring maximum values are normalized to obtain the preprocessed time sequence data.

[0037] It should be noted that when the real-time statistical value is greater than the preset historical threshold, the time of the maximum value in the intercepted initial time sequence data is determined, and the sequence data of the previous N seconds and the subsequent N seconds of the time are obtained, and the sequence data of each second includes a plurality of monitoring data values. Wherein, if there are a plurality of same maximum values in the sequence data, the sequence data of the previous N seconds and the subsequent N seconds of the time when the maximum value first appears is taken. Preferably, N is preferably 30 seconds, but is not limited thereto.

[0038] In the 2N second sequence data, the maximum value in the sequence data of each second is counted to obtain 2N monitoring maximum values. In order to eliminate the dimensional influence of data and reduce the gap between feature data, the data range needs to be unified, so that the data is linearly mapped between the maximum value and the minimum value of the data, thereby mapping the above maximum value sequence data to the [0, 1] interval, and obtaining the required preprocessed time sequence data , is the i-th mapping data, i=[1, 60] and i is a positive integer, so that the above normalization processing can eliminate the dimensional influence of data, reduce the gap between feature data, and improve the convergence speed of the algorithm, thereby improving the accuracy of data processing.

[0039] S103, according to the preprocessed time sequence data of the high-frequency sensor and the preset sensor gross error sequence data, dynamic time warping calculation is performed to obtain the current similarity distance of the high-frequency sensor;

[0040] It should be noted that in order to determine whether the preprocessed time sequence data has a gross error anomaly, the present application specially sets the corresponding preset sensor gross error sequence data , wherein, is the j-th sensor gross error data, j=[1, 3] and j is a positive integer; the dynamic time warping algorithm is used to measure the similarity between the preprocessed time sequence data and the preset sensor gross error sequence data , that is, the current similarity distance d1 between the two is calculated.

[0041] Specifically, the step of calculating the current similarity distance of the high-frequency sensor according to the pre-processed time series data of the high-frequency sensor and the preset sensor gross error sequence data includes:

[0042] Step 1, constructing a distance matrix from the pre-processed time series data and the preset sensor gross error sequence data;

[0043] It should be noted that the pre-processed time series data and the preset sensor gross error sequence data A 60X3 distance matrix (i,j) can be constructed, and the elements in the distance matrix (i,j) are the distances of the corresponding points of the two sequences .

[0044] Step 2, starting from the upper left element in the distance matrix as the starting point and the lower right element as the ending point, dynamically finding a path with the smallest sum of matrix elements, and calculating the current similarity distance of the path.

[0045] It should be noted that starting from the upper left element w 1,1 as the starting point and the lower right element w 60,3 as the ending point, a path with the smallest sum of matrix elements is dynamically found, so as to obtain the required optimal path, and the sum of all elements on the optimal path is the current similarity distance d1 of the path.

[0046] S104, judging whether the number of the ultra-limit sensors is less than a preset sensor number threshold,

[0047] If the judgment is yes, it is judged whether the current similarity distance of the high-frequency sensor is greater than the corresponding preset distance threshold, if the judgment is yes, normal alarm processing is performed, if the judgment is no, a gross error alarm information is sent and review processing is performed;

[0048] If the judgment is no, multi-sensor abnormality discrimination processing is performed.

[0049] Preferably, the preset sensor number threshold is 2.

[0050] When the number of the over-limit high-frequency sensors in the high-frequency sensors of the monitoring system is less than the preset sensor number threshold, it indicates that there is only one over-limit high-frequency sensor in the current monitoring system. At this time, the current similar distance d1 of the high-frequency sensor is compared with the preset distance threshold d0. When the current similar distance d1 is greater than the preset distance threshold d0, it indicates that the monitoring data of the high-frequency sensor is normal over-limit alarm data, and normal alarm processing can be performed to accurately inform the relevant staff to go to the scene for troubleshooting. When the current similar distance d1 is less than or equal to the preset distance threshold d0, it indicates that the current over-limit alarm data is caused by a single-point gross error factor, and a gross error alarm information is sent for review processing. The review processing specifically includes: eliminating the maximum value (i.e. single-point limit data) in the initial time sequence data in the current preset time period, recalculating the real-time statistical value according to the remaining sequence data and comparing it with the preset historical threshold. When the new real-time statistical value is greater than the preset historical threshold, it indicates that the current monitoring data of the high-frequency sensor is normal over-limit alarm data, and normal alarm processing can be performed to accurately inform the relevant staff to go to the scene for troubleshooting. When the new real-time statistical value is less than or equal to the preset historical threshold, it indicates that this alarm is an abnormal alarm caused by a single-point gross error, and no structural abnormal event occurs, so the alarm can be canceled.

[0051] When the number of the over-limit high-frequency sensors in the high-frequency sensors of the monitoring system is greater than or equal to the preset sensor number threshold, it indicates that there are at least two or more over-limit high-frequency sensors in the current monitoring system. At this time, multi-sensor abnormality discrimination processing needs to be performed to determine whether the monitoring data of the multiple over-limit high-frequency sensors is normal over-limit alarm or over-limit alarm caused by a single-point gross error factor, so as to avoid unnecessary alarm processing and resource waste caused by false alarms of the high-frequency sensors in the monitoring system, and further improve the alarm accuracy of the bridge monitoring system.

[0052] Specifically, as shown in Figure 2 the multi-sensor abnormality discrimination processing includes:

[0053] S201, it is judged whether the current similar distance of at least one high-frequency sensor in the over-limit high-frequency sensors is greater than the corresponding preset distance threshold. If the judgment is yes, normal alarm processing is performed. If the judgment is no, the correlation between the multiple over-limit high-frequency sensors is determined, and alarm classification processing is performed on the multiple high-frequency sensors.

[0054] It should be noted that among multiple high-frequency sensors that exceed limits, if the current similarity distance of at least one high-frequency sensor is greater than its corresponding preset distance threshold, the current monitoring data is considered to be normal over-limit alarm data, and normal alarm processing is performed to accurately notify relevant personnel to conduct on-site investigation and processing; otherwise, it is considered that there is still a possibility of over-limit alarm caused by single-point gross error factors. In this case, it is necessary to determine the correlation between the multiple high-frequency sensors that exceed limits and to classify the alarms of multiple high-frequency sensors.

[0055] Among them, such as Figure 3 As shown, the steps of determining the correlation between multiple high-frequency sensors that exceed the limits and performing alarm classification processing on the multiple high-frequency sensors include:

[0056] S301. Obtain the tag data of the high-frequency sensor that exceeds the limit, wherein the tag data is preset and includes category setting data for multiple categories;

[0057] It should be noted that, in this embodiment of the invention, the label data includes at least category setting data for the first category and the second category.

[0058] The category setting data for the first category includes the category setting data for the acquisition device category. For example, if multiple high-frequency sensors are connected to the same acquisition device, the category setting data for the high-frequency sensors will be the same; otherwise, the category setting data for the multiple high-frequency sensors will have different values.

[0059] The second category of category setting data includes category setting data for power supply equipment. For example, high-frequency sensors powered by the same power supply equipment have the same category setting data, while high-frequency sensors powered by different power supply equipment have different category setting data values.

[0060] S302. Determine whether there are any two high-frequency sensors whose category setting data are different. If the determination is yes, it means that the data of multiple high-frequency sensors are uncorrelated, and normal alarm processing is performed. If the determination is no, it means that the data of multiple high-frequency sensors are correlated, a gross error alarm message is issued, and a review process is performed.

[0061] It should be noted that, in the multiple over-limit high-frequency sensors, if the category setting data of any two of the high-frequency sensors are not the same, it indicates that the multiple high-frequency sensor data are in an unrelated state, and the over-limit sensor is unrelated, which indicates that the over-limit sensors are from different acquisition and power supply equipment, so it is judged that the probability of data abnormality caused by the same acquisition and power supply equipment is very low, so normal alarm processing is performed to timely inform the staff to check the bridge site condition; otherwise, it is considered that the multiple high-frequency sensors are in a related state, at this time, the current over-limit alarm situation may be caused by a single-point gross error factor, and a gross error alarm information is sent and a review processing is performed.

[0062] The above-mentioned manner can improve the alarm accuracy and reliability of the high-frequency sensor in the bridge monitoring system, avoid unnecessary alarm processing and resource waste caused by false alarm of the high-frequency sensor (such as false alarm caused by a single-point gross error), and further improve the alarm accuracy of the bridge monitoring system.

[0063] The preset distance threshold is obtained by dynamic time warping processing of historical normal sequence data of the high-frequency sensor and time sequence data containing a gross error reviewed by an artificial, and the above-mentioned manner can be not affected by artificial experience factors, so as to improve the data processing accuracy.

[0064] The application further provides a computer device including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above-mentioned high-frequency sensor single-point gross error identification method when executing the computer program.

[0065] The application further provides a readable storage medium having a computer program stored thereon, and the computer program implements the steps of the above-mentioned high-frequency sensor single-point gross error identification method when executed by a processor.

[0066] In summary, the application can perform single-point gross error identification processing on the original time sequence data of the high-frequency sensor in the bridge monitoring system in real time to determine whether the high-frequency sensor over-limit alarm is caused by a single-point gross error, and perform normal sensor alarm work if it is not caused by a single-point gross error, so as to improve the alarm accuracy and reliability of the high-frequency sensor, avoid unnecessary alarm processing and resource waste caused by false alarm of the high-frequency sensor, and further improve the alarm accuracy of the bridge monitoring system.

[0067] The above-mentioned only is the preferred embodiment of the application, and of course cannot limit the scope of the application, so the equivalent changes made according to the claims of the application still belong to the scope covered by the application.

Claims

1. A method for identifying single-point gross errors using a high-frequency sensor, characterized in that: include: The raw time series data of all high-frequency sensors in the monitoring system are acquired, and the raw time series data of the high-frequency sensors are extracted and statistically processed to obtain initial time series data and real-time statistical values. The number of over-limit sensors whose real-time statistical values ​​are greater than a preset historical threshold is counted, and the initial time series data of the over-limit high-frequency sensors are preprocessed according to a preset time series identification rule to obtain the preprocessed time series data of the high-frequency sensors. Dynamic time warping is performed based on the preprocessed time series data of the high-frequency sensor and the preset sensor gross error sequence data to obtain the current similarity distance of the high-frequency sensor. Determine whether the number of over-limit sensors is less than a preset sensor number threshold. If the determination is yes, then determine whether the current similarity distance of the high-frequency sensor is greater than its corresponding preset distance threshold. If the determination is yes, then perform normal alarm processing. If the determination is no, then issue gross error alarm information and perform verification processing. If the determination is negative, then multi-sensor anomaly detection processing is performed; The step of preprocessing the initial time series data of the high-frequency sensor that exceeds the limit according to a preset time series identification rule to obtain the preprocessed time series data of the high-frequency sensor includes: Determine the moment when the maximum value is found in the initial time series data after truncation, and obtain the sequence data of the time N seconds before and N seconds after that moment, where N is a positive integer; Count the maximum value in the sequence data for each second to obtain 2N monitored maximum values; The 2N monitored maximum values ​​are normalized to obtain the preprocessed time series data; The step of performing dynamic time warping calculations based on the preprocessed time series data of the high-frequency sensor and the preset sensor gross error sequence data to obtain the current similarity distance of the high-frequency sensor includes: The preprocessed time series data and the preset sensor gross error sequence data are used to construct a distance matrix; Starting from the top-left element of the distance matrix and ending at the bottom-right element, dynamically find a path with the minimum sum of matrix elements and calculate the current similarity distance of that path.

2. The high-frequency sensor single-point gross error identification method according to claim 1, characterized in that, The step of extracting and statistically processing the raw time series data from the high-frequency sensor to obtain initial time series data and real-time statistical values ​​includes: The original time series data is truncated according to the preset interval truncation rules to obtain the initial time series data within the current preset time period; The initial time series data is statistically analyzed to obtain the required real-time statistical values.

3. The high-frequency sensor single-point gross error identification method according to claim 1 or 2, characterized in that, The real-time statistical values ​​include any one or more of the average, change, and maximum values ​​of the initial time series data.

4. The high-frequency sensor single-point gross error identification method according to claim 1, characterized in that, The steps for performing multi-sensor anomaly detection processing include: If it is determined whether at least one of the high-frequency sensors that exceeds the limit has a current similarity distance greater than its corresponding preset distance threshold, then normal alarm processing is performed if the determination is yes; otherwise, the correlation between the multiple high-frequency sensors that exceed the limit is determined and alarm classification processing is performed on the multiple high-frequency sensors.

5. The high-frequency sensor single-point gross error identification method according to claim 4, characterized in that, The steps of determining the correlation between multiple high-frequency sensors that exceed the limits and classifying the multiple high-frequency sensors for alarm processing include: Acquire the tag data of the high-frequency sensor that exceeds the limit, wherein the tag data includes category setting data for multiple categories; If the system determines whether any two high-frequency sensors have different category settings, and if so, it indicates that the data from multiple high-frequency sensors are uncorrelated, and a normal alarm is triggered. If not, it indicates that the data from multiple high-frequency sensors are correlated, and a gross error alarm is issued, followed by a review process.

6. The high-frequency sensor single-point gross error identification method according to claim 1, characterized in that, The preset sensor gross error sequence data is [0,1,0].

7. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the high-frequency sensor single-point gross error identification method according to any one of claims 1 to 6.

8. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the high-frequency sensor single-point gross error identification method according to any one of claims 1 to 6.

Images