Trend extraction method and system for geological disaster monitoring crack displacement meter data

Abstract

This invention provides a method and system for trend extraction from crack displacement gauge data used in geological disaster monitoring. The method includes: collecting crack displacement gauge data and performing data cleaning; labeling the collected data into n data points according to time sequence and calculating the cost of trend classification for each data point; calculating the total cost of trend classification for all data points; determining the trend classification results for all data points with the objective of minimizing the total cost; grouping adjacent data points with the same trend classification into the same trend region, which can be an upward trend region, a downward trend region, or a stable region; and converting the upward and downward trend regions into an output format, including start time, end time, upward or downward trend, and an upward or downward threshold. This invention extracts the changing trends from crack displacement gauge data to predict crack width changes. This method is simple, efficient, and can effectively improve the timeliness of early warning.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology for geological disaster monitoring data, and in particular to a method and system for trend extraction of crack displacement gauge data for geological disaster monitoring. Background Technology

[0002] A crack gauge is an instrument used to monitor changes in the width of cracks on the surface of soil or structures, and can monitor geological hazards such as landslides and collapses. As the hazardous material changes, the crack gauge data changes accordingly, monitoring the speed and magnitude of these changes.

[0003] With a large number of devices online, checking the data of each crack gauge every day requires a lot of time and effort and is prone to human error. Therefore, an algorithm is needed to automatically filter crack gauge data that has deformed recently (e.g., within a month). This would allow for the detection of slowly deforming disaster points before an alarm is triggered, making early warnings more timely and enabling relevant departments to prepare earlier. Summary of the Invention

[0004] To address the technical problems of time-consuming and labor-intensive data inspection and processing of a large number of crack gauges in existing technologies, as well as the lack of timely early warning, this invention provides a method and system for trend extraction of crack displacement gauge data in geological disaster monitoring. By extracting the changing trend of crack gauge data, the method can predict changes in crack width. This method is simple and efficient, and can effectively improve the timeliness of early warning.

[0005] The technical solution adopted by the present invention to solve the above problems is as follows:

[0006] This invention provides a method for trend extraction of crack displacement gauge data in geological disaster monitoring, comprising the following steps:

[0007] S1. Collect crack gauge data and perform data cleaning;

[0008] S2. Mark the data collected by the crack gauge into n data points in chronological order, and calculate the cost of trend classification for each data point of the crack gauge. The trend classification includes stable, rising, and falling.

[0009] S3. Calculate the total cost of trend classification for all data points in the crack gauge;

[0010] S4. With the goal of minimizing total cost, determine the trend classification results of all data points of the crack gauge, and divide adjacent data points with the same trend classification into the same trend region. The trend region can be an upward region, a downward region, or a stable region.

[0011] S5. Convert the rising region and the falling region into an output format, wherein the output format includes start time, end time, rising or falling trend, and rising or falling threshold.

[0012] Furthermore, the data cleaning method is as follows: using an anomaly data determination method to clean the outliers in the crack gauge data, retaining the data that meets the measurement range or accuracy range.

[0013] Furthermore, the cost of calculating the trend classification for each data point of the crack gauge specifically includes the following steps:

[0014] The cost of classifying data point i as a stationary point is:

[0015]

[0016] in, This indicates that data point i-1 is divided into class j, where j=0 represents stationary, j=1 represents increasing, and j=2 represents decreasing. i This represents the monitoring data of crack gauge data point i, z i-1 Let i-1 represent the monitoring data of data point i-1, p represent the maximum error value of the crack gauge sensor, a1 is the penalty cost, which is the cost of misclassifying data outside the normal fluctuation range, and d0 is the cost of starting a new stable classification.

[0017] The cost of dividing data point i into ascending categories is:

[0018]

[0019] Where a0 is the penalty cost, which is the cost of misclassifying data within the normal fluctuation range, and d1 is the cost of initiating a new ascending classification;

[0020] The cost of dividing data point i into decreasing segments is set as follows:

[0021]

[0022] Where d2 is the cost of initiating a new descent classification;

[0023] The cost of trend classification for the current data point is determined by the difference between the current data point and the previous data point, as well as the trend classification result of the previous data point.

[0024] Furthermore, the calculation of the cost of trend classification for each data point in the crack gauge also includes the step of setting constraints: for any data point i, i∈(0~n-1), there is only one trend classification. This indicates that data point i is divided into class j, where j=0 represents stationary, j=1 represents rising, and j=2 represents falling.

[0025] Furthermore, a0:a1:d0:d1:d2 = 1:3:24:9:9.

[0026] Furthermore, the calculation of the total cost of trend classification for all data points specifically includes the following steps:

[0027] The trend classification cost for starting data point 0 is:

[0028]

[0029] The trend classification cost for the first i+1 data points is:

[0030]

[0031] The cost of trend classification for n data points is:

[0032]

[0033] in, Let represent the optimal total cost of the first i+1 data points when data point i is classified into class j. This indicates that data point i is classified into class j. d0 represents the cost of classifying data point i into class j, d1 represents the cost of classifying a new stationary class, d2 represents the cost of classifying a new ascending class, and d2 represents the cost of classifying a new descending class.

[0034] Furthermore, the method for dividing adjacent data points with the same trend classification into the same trend region is as follows:

[0035] If the trend classification of data point i is the same as that of data point i-1, then they are in the same trend region. j represents the nature of the trend region, where j=0 represents stability, j=1 represents rising, and j=2 represents falling.

[0036] If the trend classification of data point i is different from that of data point i-1, then they are in two different trend regions.

[0037] This invention also provides a trend extraction system for crack displacement gauge data in geological disaster monitoring, comprising:

[0038] A database, which stores data collected by all projects and their crack gauges;

[0039] An algorithm library is provided to store algorithms for calculating the cost of trend classification for each data point of the crack gauge, as well as algorithms for calculating the total cost of trend classification for all data points of the crack gauge.

[0040] The data input module is used for project names or keywords;

[0041] The data acquisition module connects to the database and is used to find all crack gauge data for the input project and send the data to the data cleaning module in a fixed format.

[0042] The data cleaning module is used for data cleaning and marking the data collected by each crack gauge in chronological order of collection time.

[0043] The trend extraction module is connected to the algorithm library and is used to call the algorithms stored in the algorithm library to calculate the cost of trend classification for each data point and the total cost of trend classification for all data points of the crack meter. With the goal of minimizing the total cost, the trend classification results of all data points of the crack meter are determined, and adjacent data points with the same trend classification are divided into the same trend region to obtain several rising regions, falling regions, and stable regions.

[0044] The data output module is used to convert the rising and falling regions into an output format, which includes the start time, end time, rising or falling trend, and rising or falling threshold, and then output the data.

[0045] The beneficial effects of this invention are:

[0046] This invention provides a method and system for trend extraction of crack displacement gauge data for geological disaster monitoring. By extracting the changing trend of crack gauge data, the method can predict changes in crack width. This method is simple and efficient, and can effectively improve the timeliness of early warning. It solves the technical problems of time-consuming and labor-intensive data inspection and processing of large numbers of crack gauges and untimely early warning.

[0047] The trend extraction method and system for crack displacement gauge data in geological disaster monitoring provided by this invention can automatically filter crack displacement gauge data that have recently undergone deformation, detect disaster points with slow deformation before an alarm occurs, make early warning more timely, and allow relevant departments to prepare earlier. When an alarm occurs from crack displacement gauge data, it can provide a basis for operation and maintenance personnel.

[0048] The trend classification cost calculation method provided by this invention has a time complexity that can be represented by a polynomial. That is, as the length of the calculation data increases, the calculation time increases slowly, resulting in high processing efficiency.

[0049] This invention is applicable to various types of single time-series data trend processing to obtain trend development or to provide early warnings. Attached Figure Description

[0050] The accompanying drawings, which form part of this specification, are provided to further illustrate embodiments of the invention and, together with the textual description, explain the principles of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0051] Figure 1 This is a schematic diagram illustrating the principle of calculating the total cost of trend classification for the first i data points, provided for a specific embodiment of the present invention.

[0052] Figure 2 A block diagram of a trend extraction system for crack displacement gauge data in geological disaster monitoring, provided as a specific embodiment of the present invention. Detailed Implementation

[0053] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that the terminology used herein is for describing specific embodiments only and is not intended to limit the exemplary embodiments according to this application. Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of components and steps set forth in these embodiments do not limit the scope of the present invention.

[0054] As one aspect of the present invention, a method for trend extraction of crack displacement gauge data for geological disaster monitoring is provided, comprising the following steps:

[0055] S1. Collect crack gauge data and perform data cleaning.

[0056] Preferably, outliers in the crack gauge data are cleaned using anomaly detection methods to ensure that only normal data is pushed to the trend extraction module. Anomaly detection methods include: Isolation Forest algorithm, 3σ detection, etc.

[0057] It should be noted that, in this invention, abnormal data refers to data points that deviate from the normal measurement range or data segments whose overall accuracy is not within the normal accuracy range.

[0058] S2. Mark the data collected by the crack gauge into n data points in chronological order, and calculate the cost of trend classification for each data point of the crack gauge. The trend classification includes stable, rising, and falling.

[0059] To facilitate subsequent processing, the data collected by each crack gauge is marked as n data points in the order of collection time. Data point i represents the i-th data point, which corresponds to the i-th monitoring data of the crack gauge.

[0060] This invention utilizes dynamic programming to extract trends from monitoring data of n data points of a crack gauge.

[0061] First, the n data points are classified into three trends: upward, downward, and stable. The trend extraction problem of the crack gauge data is divided into n stages. In each stage, the trend classification problem of a single data point is considered, and the ultimate goal is to minimize the trend classification cost.

[0062] Set z i This represents the monitoring data at data point i of the crack gauge, where i ∈ (0 to n-1), and n is the total number of data points collected by the crack gauge. p represents the maximum error value of the crack gauge sensor, which is typically 0.5. This indicates that data point i is divided into class j, where j=0 represents stationary, j=1 represents increasing, and j=2 represents decreasing. This represents the class into which data point i is classified as j. This represents the class that data point i is not classified into, i.e., j. This represents the cost of classifying data point i into class j. This represents the optimal total cost for the first i+1 data points when data point i is classified into class j.

[0063] Set constraints such that for any data point i, there can only be one classification.

[0064]

[0065] for The rules for setting are divided into stable costs. Rising costs Decreasing costs The design cost rules are as follows:

[0066] (1) Setting The rule is: the data point preceding data point i is classified as stationary, i.e. In the case where the difference between the monitoring data of data point i and the monitoring data of data point i-1 is within the normal fluctuation range (the difference does not exceed the maximum error value of the crack gauge sensor), the cost of classifying data point i as a stationary point is... The cost of classifying data point i as a stationary point is zero; when the difference between the monitoring data of data point i and the monitoring data of data point i-1 is outside the normal fluctuation range (the difference exceeds the maximum error value of the crack gauge sensor), the cost of classifying data point i as a stationary point is zero. Let the value be a1. The data point preceding data point i was not classified as stationary, i.e. In the case of data point i being classified as a stationary point, the cost It is d0.

[0067]

[0068] Where a1 is the penalty cost, which is the cost of misclassifying data outside the normal fluctuation range, and d0 is the cost of starting a new stable classification.

[0069] (2) Setting The rule is: the data point preceding data point i is classified as ascending, i.e. In the case where the monitored data of data point i is larger than the monitored data of data point i-1, and the difference is not within the normal fluctuation range (indicating that this fluctuation is not caused by systematic error, but is most likely a fluctuation of the true value), the cost of classifying data point i as an upward point is determined. The cost of classifying data point i as an upward point is zero; when the difference between the monitoring data of data point i and data point i-1 is within the normal fluctuation range (indicating that this difference is most likely caused by system error), the cost of classifying data point i as an upward point is zero. for When the monitoring data of data point i is smaller than that of data point i-1, and the difference is not within the normal fluctuation range, data point i is classified as a cost of rising points. Let the value be a1. The data point preceding data point i was not classified as ascending, i.e. In the case of classifying data point i as an ascending point, the cost Let's call it d1.

[0070]

[0071] Where a0 is the penalty cost, which is the cost of misclassifying data within the normal fluctuation range, and d1 is the cost of initiating a new ascending classification. Preferably, the ratios of a0, a1 and d0, d1 are as follows:

[0072] a0:a1:d0:d1=1:3:24:9

[0073] Where a1:d1 = 1:3 indicates that three consecutive rising points can be classified as an upward range. a0:d0 = 1:24, under normal circumstances, the crack gauge measures 24 sets of data per day. When all the data in a day are normal fluctuation data, it will definitely be classified as normal fluctuation data. a0:a1 = 1:3 indicates that the cost of the falling point in the upward segment is 3 times that of the stable point.

[0074] The ratios of a0, a1 and d0, d1 are uncertain and can be adjusted based on crack displacement gauge data. Preferably, the cost 'a' of a point should be less than the cost 'd' of the region. This setting better ensures the continuity of the region and provides some tolerance for points that do not conform to the region's characteristics.

[0075] (3) Setting The rule is: the data point preceding data point i is classified as a descending point, i.e. In the case where the monitoring data of data point i is smaller than that of data point i-1, and the difference is not within the normal fluctuation range, data point i is classified as a cost of a decreasing point. The cost is zero when the difference between the monitoring data of data point i and data point i-1 is within the normal fluctuation range; data point i is classified as a cost of a decreasing point. Let a0 be the cost of data point i when the monitoring data of data point i is greater than that of data point i-1, and the difference is not within the normal fluctuation range. Let the value be a1. The data point preceding data point i was not classified as decreasing, i.e. In the case of data point i being classified as a descent point, the cost It is d2.

[0076]

[0077] Where d2 is the cost of initiating a new descent classification. Preferably, the ratios of a0, a1 and d0, d1, d3 are as follows:

[0078] a0:a1:d0:d1:d2=1:3:24:9:9

[0079] S3. Calculate the total cost of trend classification for all data points in the crack gauge.

[0080] For the initial data point 0, there are three possibilities: the data point is rising, falling, or stationary. The costs for these three scenarios are as follows:

[0081]

[0082] For the i-th data point (i > 0), the cost of classifying it as stationary is the minimum cost when the previous data point is classified as stationary, rising, or falling. The same logic applies when classifying the i-th data point as rising or falling, and the formula is as follows:

[0083]

[0084] And so on, such as Figure 1 As shown, the cost of the last data point n-1 can be obtained as follows:

[0085]

[0086] in, Let represent the optimal total cost of the first i+1 data points when data point i is classified into class j. This indicates that data point i is classified into class j. d0 represents the cost of classifying data point i into class j, d1 represents the cost of classifying a new stationary class, d2 represents the cost of classifying a new ascending class, and d2 represents the cost of classifying a new descending class.

[0087] S4. With the goal of minimizing total cost, determine the trend classification results of all data points in the crack gauge, and divide adjacent data points with the same trend classification into the same trend region, which can be an upward region, a downward region, or a stable region.

[0088] In this case, the cost of the optimal trend classification method is as follows:

[0089]

[0090] The final trend classification cost-optimal result is obtained, and the optimal choice calculated from the trend classification cost-optimal result constitutes the trend classification result.

[0091] The formula for calculating the properties of a trend region is as follows:

[0092]

[0093] If the trend classification of data point i is the same as that of data point i-1, they are considered to be in the same trend region, and j represents the nature of this trend region. In this embodiment, the trend region is divided into an upward region, a downward region, and a stable region. Using the formula for calculating the nature of the trend region, the data collected by the crack gauge is divided into several upward, downward, and stable regions in chronological order.

[0094] S5. Convert the rising and falling regions into a standard output format, including start time, end time, rising / falling time, and rising / falling threshold, and then output them.

[0095] By outputting rising and falling regions, changes in crack gauge data can be predicted in advance, thereby realizing an early warning function and enabling relevant departments to prepare earlier.

[0096] As one aspect of the present invention, a trend extraction system for crack displacement gauge data in geological disaster monitoring is provided, including a data input module, a data acquisition module, a data cleaning module, a trend extraction module, an output module, a database, and an algorithm library, such as... Figure 2 As shown.

[0097] 1. Data Input Module

[0098] The data input module is the interaction module between operations and maintenance personnel and the trend extraction system. Operations and maintenance personnel input the name or keywords of the project they need to investigate into the program.

[0099] It should be noted that the project name or project keywords directly correspond to several crack meters associated with the project. By entering the project name or keywords, maintenance personnel can quickly find the crack meters that need to be checked.

[0100] 2. Data Acquisition Module

[0101] The data acquisition module connects to the database, searches for all crack gauge data in the project that needs to be investigated, and sends the data to the data cleaning module in a fixed format.

[0102] In this invention, the database is used to store all projects and the data collected by the crack gauge. Through data preprocessing, all the data collected by the crack gauge is standardized to obtain crack gauge data in a fixed format.

[0103] 3. Data Cleaning Module

[0104] The data cleaning module cleans the acquired crack gauge data and marks the data collected by each crack gauge in chronological order of acquisition time.

[0105] 4. Trend Extraction Module

[0106] The trend extraction module connects to the algorithm library. By calling the algorithms stored in the algorithm library, the trend extraction module calculates the cost of trend classification for each data point and the total cost of trend classification for all data points in the crack meter. With the goal of minimizing the total cost, it determines the trend classification results for all data points in the crack meter. Adjacent data points with the same trend classification are divided into the same trend region, resulting in several rising regions, falling regions, and stable regions.

[0107] 5. Data Output Module

[0108] The data output module is used to convert the rising and falling regions into a standard output format, including start time, end time, rise / fall, and rise / fall threshold, before outputting them.

[0109] 6. Database

[0110] The database is used to store all projects and the data collected by their crack gauges.

[0111] 7. Algorithm Library

[0112] The algorithm library stores algorithms for calculating the cost of trend classification for each data point in the crack gauge, as well as algorithms for calculating the total cost of trend classification for all data points in the crack gauge.

[0113] This invention provides a method and system for trend extraction of crack displacement gauge data for geological disaster monitoring. By extracting the changing trend of crack gauge data, the method can predict changes in crack width. This method is simple and efficient, and can effectively improve the timeliness of early warning. It solves the technical problems of time-consuming and labor-intensive data inspection and processing of large numbers of crack gauges and untimely early warning.

[0114] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0115] The parts of this invention not described in detail are techniques known to those skilled in the art.

Claims

1. A method for trend extraction from crack displacement gauge data in geological disaster monitoring, characterized in that, Includes the following steps: S1. Collect crack gauge data and perform data cleaning; S2. Mark the data collected by the crack gauge into n data points in chronological order, and calculate the cost of trend classification for each data point of the crack gauge. The trend classification includes stable, rising, and falling. S3. Calculate the total cost of trend classification for all data points in the crack gauge; S4. With the goal of minimizing total cost, determine the trend classification results of all data points of the crack gauge, and divide adjacent data points with the same trend classification into the same trend region. The trend region can be an upward region, a downward region, or a stable region. S5. Convert the rising region and the falling region into an output format, wherein the output format includes start time, end time, rising or falling trend, and rising or falling threshold.

2. The method as described in claim 1, characterized in that, The data cleaning method is as follows: abnormal values ​​in the crack gauge data are cleaned by using an abnormal data judgment method, and data that meets the measurement range or accuracy range are retained.

3. The method as described in claim 1, characterized in that, The cost of calculating the trend classification for each data point of the crack gauge specifically includes the following steps: The cost of classifying data point i as a stationary point is: in, This indicates that data point i-1 is divided into class j, where j=0 represents stationary, j=1 represents increasing, and j=2 represents decreasing. i This represents the monitoring data of crack gauge data point i, z i-1 Let i-1 represent the monitoring data of data point i-1, p represent the maximum error value of the crack gauge sensor, a1 is the penalty cost, which is the cost of misclassifying data outside the normal fluctuation range, and d0 is the cost of starting a new stable classification. The cost of dividing data point i into ascending categories is: Where a0 is the penalty cost, which is the cost of misclassifying data within the normal fluctuation range, and d1 is the cost of initiating a new ascending classification; The cost of dividing data point i into decreasing segments is set as follows: Where d2 is the cost of initiating a new descent classification; The cost of trend classification for the current data point is determined by the difference between the current data point and the previous data point, as well as the trend classification result of the previous data point.

4. The method as described in claim 3, characterized in that, The calculation of the cost of trend classification for each data point of the crack gauge also includes the step of setting constraints: for any data point i, i∈(0~n-1), there is only one trend classification. This indicates that data point i is divided into class j, where j=0 represents stationary, j=1 represents rising, and j=2 represents falling.

5. The method as described in claim 3, characterized in that, a0∶a1∶d0∶d1∶d2=1∶3∶24∶9∶9.

6. The method as described in claim 3, characterized in that, The calculation of the total cost of trend classification for all data points specifically includes the following steps: The trend classification cost for starting data point 0 is: The trend classification cost for the first i+1 data points is: The cost of trend classification for n data points is: in, Let represent the optimal total cost of the first i+1 data points when data point i is classified into class j. This indicates that data point i is classified into class j. d0 represents the cost of classifying data point i into class j, d1 represents the cost of classifying a new stationary class, d2 represents the cost of classifying a new ascending class, and d2 represents the cost of classifying a new descending class.

7. The method as described in claim 3, characterized in that, The method for dividing adjacent data points with the same trend classification into the same trend region is as follows: If the trend classification of data point i is the same as that of data point i-1, then they are in the same trend region. j represents the nature of the trend region, where j=0 represents stability, j=1 represents rising, and j=2 represents falling. If the trend classification of data point i is different from that of data point i-1, then they are in two different trend regions.

8. A trend extraction system for crack displacement gauge data in geological disaster monitoring, characterized in that, include: A database, which stores data collected by all projects and their crack gauges; An algorithm library is provided to store algorithms for calculating the cost of trend classification for each data point of the crack gauge, as well as algorithms for calculating the total cost of trend classification for all data points of the crack gauge. The data input module is used for project names or keywords; The data acquisition module connects to the database and is used to find all crack gauge data for the input project and send the data to the data cleaning module in a fixed format. The data cleaning module is used for data cleaning and marking the data collected by each crack gauge in chronological order of collection time. The trend extraction module is connected to the algorithm library and is used to call the algorithms stored in the algorithm library to calculate the cost of trend classification for each data point and the total cost of trend classification for all data points of the crack meter. With the goal of minimizing the total cost, the trend classification results of all data points of the crack meter are determined, and adjacent data points with the same trend classification are divided into the same trend region to obtain several rising regions, falling regions, and stable regions. The data output module is used to convert the rising and falling regions into an output format, which includes the start time, end time, rising or falling trend, and rising or falling threshold, and then output the data.

Images