Embedded bridge cable force data abnormity monitoring device and method

Abstract

The invention discloses an embedded bridge cable force data abnormity monitoring device and method, and belongs to the technical field of bridge safety monitoring and intelligent operation and maintenance. According to the method, the difference characteristic of computing resources of an embedded platform and an upper computer is fully utilized, a bridge anomaly detection algorithm based on AT-DBSCAN and a bridge anomaly classification algorithm based on fuzzy logic reasoning are developed, and the anomaly detection and classification algorithms are reasonably deployed at an embedded end and an upper computer end. Wherein the embedded platform is responsible for real-time acquisition and anomaly detection of bridge data, and the upper computer executes refined processing and analysis operations such as bridge anomaly classification and early warning, and constructs a bridge anomaly monitoring graphical user interface (GUI), thereby realizing visual display of the whole process from data acquisition to result analysis. And comprehensive guarantee is provided for intelligent operation and maintenance and safety early warning of the bridge structure.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge safety monitoring and intelligent operation and maintenance. Background Art

[0002] With the proposal of the strategy of building a transportation power, the transportation volume in China has increased significantly. However, most bridges have been in service for a long time and are in an environment eroded by the external environment and natural disasters. At the same time, under the influence of negative factors such as the long-term cumulative fatigue effect of vehicle loads on the bridge deck, the destructive effect of accidental loads, and the increasingly frequent overweight loads, the durability and stability of bridges are challenged. There is a great risk of bridge accidents, which pose a potential threat to people's lives and property. Therefore, it is urgent to monitor bridge information.

[0003] At present, Zhang Zhengwei has carried out research on the application of wireless sensing technology in the stress structure safety monitoring of elevated bridges. By collecting on-site data of the stress structure through wireless sensing technology, the problem of a large difference between the monitoring results and the actual measurement results has been solved. (See Zhang Zhengwei, Yuan Junying. Research on the Stress Structure Safety Monitoring Method of Elevated Bridges [J]. Modern Transportation Technology, 2022, 19(05): 37-39+46.); Wei Youzhi et al. established a scientific bridge and tunnel health monitoring system through Internet of Things technology and communication infrastructure to strengthen the structural safety monitoring and management of bridges and tunnels. (See Wei Youzhi. Bridge and Tunnel Structural Health Monitoring System Based on the Internet of Things [J]. China New Telecommunications, 2023, 25(02): 42-45.); Wang Lingbo et al. summarized the future development direction based on the development of bridge health monitoring (BHM) technology in aspects such as system applicability, structural damage monitoring algorithms, and monitoring data preprocessing in recent years. (See Wang Lingbo, Wang Qiuling, Zhu Zhao, etc. Research Status and Prospect of Bridge Health Monitoring Technology [J]. China Journal of Highway and Transport, 2021, 34(12): 25-45. DOI: 10.19721 / j.cnki.1001-7372.2021.12.003.).

[0004] Current research mostly focuses on the anomaly detection of single sensors, without considering the spatio-temporal distribution relationship of sensors at different positions on the bridge, and lacks a comprehensive study of the software and hardware systems for bridge anomaly detection, making it difficult to accurately detect the overall health of the bridge and carry out intelligent operation and maintenance. At the same time, due to problems such as the sensor layout position and maintenance, it is difficult to ensure the real-time nature of early warnings, directly process a large amount of noisy data collected by sensors, and the low accuracy of mainstream algorithms will lead to problems such as missed reports and false alarms occurring continuously.

[0005] Regarding the key issues in the current field, through in-depth research on the software and hardware integration platform of the bridge monitoring system, in order to make full use of the computing resources of the embedded platform and the host computer, the present invention will construct a bridge anomaly detection algorithm based on AT-DBSCAN and fuzzy classification and reasonably deploy it on the embedded side and the host computer side. Among them, the embedded platform is responsible for real-time data collection and anomaly detection, while the host computer undertakes the visualization and analysis of bridge anomaly data. Through the graphical user interface (GUI), the whole process visualization display from data collection to result analysis is realized, providing comprehensive guarantee for the intelligent operation and maintenance and safety warning of the bridge structure. Summary of the Invention

[0006] In order to solve the problems that the existing bridge anomaly detection mainly based on single-sensor anomaly detection methods does not consider the influence of different positions of the bridge on the data collected by sensors, and the data collected by sensors contains a large amount of inevitable noise and the low accuracy of mainstream algorithms will lead to false alarms, the present invention discloses an embedded-based bridge cable force data anomaly monitoring device and method. The technical solutions adopted by the present invention are as follows:

[0007] The embedded-based bridge cable force data anomaly monitoring device in the present invention includes a sensor module, an embedded platform, a wireless transmission module, a power supply battery, and a host computer;

[0008] Among them, the sensor module is composed of several cable force sensors, and the cable force sensors are arranged at equal intervals on the bridge body and collect cable force data at the same phase interval.

[0009] The embedded platform converts the original analog signal of the cable force data collected by the cable force sensors connected to it into a digital signal, and each dimension corresponds to the cable force data collected by a cable force sensor; the embedded platform stores it in the form of data frames, and the data frames include timestamps, sensor numbers, and measured values, and the embedded platform performs real-time processing and analysis on the collected cable force data and marks the abnormal data.

[0010] The wireless transmission module is used to transmit the cable force data with abnormal data marks analyzed and processed by the single-chip microcomputer module to the host computer in a wireless transmission manner.

[0011] The host computer further classifies these abnormal points by using fuzzy logic reasoning to determine their specific abnormal types.

[0012] Preferably, the embedded platform uses two-point or three-point differential signal processing to enhance signal characteristics.

[0013]

[0014]

[0015] wherein is the two-point difference of the i-th point, is the three-point difference of the i-th point.

[0016] Preferably, the embedded platform uses the density-based spatial clustering algorithm adaptive time series data AT-DBSCAN clustering method to perform real-time processing and analysis on the collected cable force data, and mark the abnormal data. The specific steps are as follows:

[0017] Step 1: Data preprocessing and formatting

[0018] (1) Input data processing: Input the original data of the cable force sensor with a time interval of from a certain moment, as well as the two-point and three-point difference data, expressed as , where n is the total number of data, represents the original value and the two-point and three-point difference values of the i-th data;

[0019] (2) Data weight calculation: w(x j ) is the weight of the sample x j , which is determined by the data acquisition time. The calculation formula is as follows:

[0020]

[0021] where represents the time interval between the sample x j and the current moment, and T is the forgetting threshold. The sample points exceeding this time have a weight of 0.5;

[0022] (3) Calculate clustering parameters:

[0023] Determine the clustering parameters and calculate the total weight of the data as:

[0024]

[0025] Define the minimum number of points to include and the scanning radius ;

[0026] (4) Apply the clustering algorithm:

[0027] According to the data within the time interval of the current moment, execute the AT-DBSCAN step process:

[0028] Utilize the determined scanning radius eps, forgetting threshold T, and minimum number of points to include minPts;

[0029] For the current point, construct data points based on the cable force value and differential characteristics, and find all nearby points within the scanning radius eps from it; assign weights to each point , when the total weight within the scanning radius eps of the nearby points ≥ the minimum inclusion points minPts, it is considered a normal point; when the number of points within the scanning radius eps of the nearby points < the minimum inclusion points minPts, this point is marked as an abnormal point.

[0030] Preferably, the host computer further classifies these abnormal points using fuzzy logic reasoning, and divides them into single-point abnormality, multi-point abnormality, data gain, data drift, and comprehensive multi-abnormality situations. Among them, single-point abnormality means that a single data point deviates from the normal range, usually caused by accidental measurement errors or sudden environmental factors; multi-point abnormality means that there are multiple abnormal points in a time series, but these abnormal points do not change the overall trend, usually caused by sensor failures; data gain abnormality means that the data as a whole is significantly larger than the normal range, usually caused by sensor gain errors or system setting problems; data drift abnormality means that the data gradually deviates from the normal range or trend, showing systematic and trend-based deviations; comprehensive multi-abnormality situation means that multiple types of abnormalities exist simultaneously, usually requiring a combination of multiple methods for comprehensive analysis; the specific steps of the fuzzy logic reasoning classification are as follows:

[0031] (1)Define input variables

[0032] Input the data obtained from the cable force sensor and the two-point difference and three-point difference, and measure the following indicators:

[0033] Rate of change: Used to measure the change amplitude of the data over time, reflecting the fluctuation characteristics of the data. According to the change of the data within one hour, it is divided into three levels: high, medium, and low. More than 10% is a high rate of change, less than 2% is a low rate of change, and between 2% - 10% is a medium rate of change;

[0034] Offset: Represents the degree to which the data deviates from the historical mean. According to the deviation between the current data and the historical mean, it is divided into three levels: high, medium, and low. Greater than 15% is a high offset, less than 5% is a low offset, and the deviation between 5% - 15% is a medium offset;

[0035] Time span: Measures the duration of abnormal data in the time series. According to the duration of the abnormality, it is divided into three levels: long, medium, and short. Continuing for more than 10 hours is a long time span, less than 3 hours is a short time span, and continuing for 3 - 10 hours is a medium time span;

[0036] (2)Fuzzify the membership degree of the input data

[0037] Convert the rate of change, offset, and time span into membership degree values through the membership degree function. The membership degree function is expressed as:

[0038]

[0039]

[0040]

[0041] where R represents the change rate, offset, or time span of the input; is a predefined interval parameter representing the interval boundaries of different membership sets, where represent 2%, 5%, and 3 hours respectively, represent 10%, 15%, and 10 hours respectively; represent the membership values of the low, medium, and high levels respectively;

[0042] (3) Construct fuzzy rules

[0043] Construct fuzzy rules for each input variable and output variable:

[0044] Rule 1: IF the change rate IS high AND the offset IS small AND the time span IS short THEN the anomaly type IS single-point anomaly (linear relationship)

[0045] Rule 2: IF the change rate IS medium AND the offset IS medium AND the time span IS medium THEN the anomaly type IS multi-point anomaly (linear relationship)

[0046] Rule 3: IF the change rate IS low AND the offset IS large AND the time span IS long THEN the anomaly type IS data gain anomaly (linear relationship)

[0047] Rule 4: IF the change rate IS low AND the offset IS medium AND the time span IS long THEN the anomaly type IS data drift anomaly (linear relationship)

[0048] Rule 5: IF the change rate IS high AND the offset IS large AND the time span IS medium OR long THEN the anomaly type IS comprehensive multi-anomaly situation (linear relationship)

[0049] (4) Apply the TS model for inference

[0050] According to the membership degree of the input and the "IF-THEN" part of the rule, calculate the membership degree value and output value of each rule. The formula is as follows:

[0051]

[0052] Among them is the output of the i-th rule; is the membership value of the rate of change, is the membership value of the offset, is the membership value of the time span;

[0053] (5)Output the classification result

[0054] According to the result of fuzzy inference, output the determination of the abnormal data type, and the formula is as follows:

[0055] 。

[0056] The present invention also provides a monitoring method for an abnormal monitoring device of cable force data of a bridge based on an embedded system. The steps of the method are as follows:

[0057] Step 1: Arrange a number of cable force sensors at equal intervals on the bridge body, and collect cable force data at the same time phase interval;

[0058] Step 2: The embedded platform converts the original analog signal of the cable force data collected by the cable force sensors connected to it into a digital signal. Each dimension corresponds to the cable force data collected by a cable force sensor; the embedded platform stores it in the form of data frames. The data frames include timestamps, sensor numbers, and measured values, and the embedded platform performs real-time processing and analysis on the collected cable force data and marks the abnormal data;

[0059] Step 3: The wireless transmission module is used to transmit the cable force data with abnormal data marks analyzed and processed by the single-chip microcomputer module to the upper computer in a wireless transmission manner;

[0060] Step 4: The upper computer further classifies these abnormal points by using fuzzy logic inference to determine their specific abnormal types.

[0061] Preferably, in Step 2, the embedded platform converts the cable force data into 2-point or 3-point differential signals,

[0062]

[0063]

[0064] Among them is the 2-point difference of the i-th point, is the 3-point difference of the i-th point, is the cable force data of the i-th cable force sensor.

[0065] Preferably, the method for the embedded platform to perform real-time processing and analysis on the collected cable force data and mark the abnormal data in Step 2 is:

[0066] Step 1: Data Preprocessing and Formatting

[0067] (1) Input Data Processing: The input is the original data of the cable force sensor with a time interval of from a certain moment and the differential data between two points and three points, denoted as , where n is the total number of data, represents the original value of the i-th data and the differential value between two points and three points;

[0068] (2) Data Weight Calculation: w(x j ) is the weight of sample x j , which is determined by the data acquisition time, and the calculation formula is as follows:

[0069]

[0070] where represents the time interval between sample x j and the current moment, T is the forgetting threshold, and the sample points exceeding this time have a weight of 0.5;

[0071] (3) Calculate Clustering Parameters:

[0072] Determine the clustering parameters and calculate the total weight of the data as:

[0073]

[0074] Define the minimum number of included points and the scanning radius ;

[0075] (4) Apply the Clustering Algorithm:

[0076] According to the data within the time interval of the current moment, execute the AT-DBSCAN step process:

[0077] Utilize the determined scanning radius eps, forgetting threshold T, and minimum number of included points minPts;

[0078] For the current point, construct a data point based on the cable force value and differential characteristics, and find all nearby points within the scanning radius eps of it; assign a weight to each point. When the total weight within the scanning radius eps of the nearby points ≥ the minimum number of included points minPts, it is considered a normal point; when the number of nearby points within the scanning radius eps < the minimum number of included points minPts, this point is marked as an abnormal point.

[0079] Preferably, in step four, the host computer further classifies these abnormal points by using fuzzy logic reasoning, and divides them into single-point abnormality, multi-point abnormality, data gain, data drift, and comprehensive multi-abnormality situations. Among them, the single-point abnormality refers to a single data point deviating from the normal range, usually caused by accidental measurement errors or sudden environmental factors; the multi-point abnormality refers to multiple abnormal points in a time series, but these abnormal points do not change the overall trend, usually caused by sensor failures; the data gain abnormality refers to the overall data being significantly larger than the normal range, usually caused by sensor gain errors or system setting problems; the data drift abnormality refers to the data gradually deviating from the normal range or trend, manifested as systematic and trend-based deviations; the comprehensive multi-abnormality situation refers to the coexistence of multiple types of abnormalities, usually requiring comprehensive analysis by combining multiple methods. The specific steps of the fuzzy logic reasoning classification are as follows:

[0080] (1) Define input variables

[0081] Input the data obtained from the cable force sensor and the two-point difference and three-point difference to measure the following indicators:

[0082] Rate of change: Used to measure the change amplitude of the data over time, reflecting the fluctuation characteristics of the data. According to the change of the data within one hour, it is divided into three levels: high, medium, and low. More than 10% is a high rate of change, less than 2% is a low rate of change, and between 2% - 10% is a medium rate of change;

[0083] Offset: Represents the degree of deviation of the data from the historical mean. According to the deviation between the current data and the historical mean, it is divided into three levels: high, medium, and low. Greater than 15% is a high offset, less than 5% is a low offset, and the deviation between 5% - 15% is a medium offset;

[0084] Time span: Measures the duration of abnormal data in the time series. According to the duration of the abnormality, it is divided into three levels: long, medium, and short. Continuing for more than 10 hours is a long time span, less than 3 hours is a short time span, and continuing for 3 - 10 hours is a medium time span;

[0085] (2) Fuzzify the input data

[0086] Convert the rate of change, offset, and time span into membership values through the membership function. The membership function is expressed as:

[0087]

[0088]

[0089]

[0090] Among them, R represents the input rate of change, offset, or time span; is a predefined interval parameter representing the interval boundaries of different membership sets, where represent 2%, 5%, and 3 hours respectively, represent 10%, 15%, and 10 hours respectively; represent the membership values of the low, medium, and high levels respectively;

[0091] (3)Construct fuzzy rules

[0092] Construct fuzzy rules for each input variable and output variable:

[0093] Rule 1: IF the rate of change IS high AND the offset IS small AND the time span IS short THEN the type of anomaly IS single-point anomaly (linear relationship)

[0094] Rule 2: IF the rate of change IS medium AND the offset IS medium AND the time span IS medium THEN the type of anomaly IS multi-point anomaly (linear relationship)

[0095] Rule 3: IF the rate of change IS low AND the offset IS large AND the time span IS long THEN the type of anomaly IS data gain anomaly (linear relationship)

[0096] Rule 4: IF the rate of change IS low AND the offset IS medium AND the time span IS long THEN the type of anomaly IS data drift anomaly (linear relationship)

[0097] Rule 5: IF the rate of change IS high AND the offset IS large AND the time span IS medium OR long THEN the type of anomaly IS comprehensive multi-anomaly situation (linear relationship)

[0098] (4)Apply the TS model for inference

[0099] According to the input membership degrees and the "IF-THEN" part of the rules, calculate the membership degree values and output values of each rule. The formula is as follows:

[0100]

[0101] where is the output of the i-th rule; is the membership degree value of the rate of change, is the membership degree value of the offset, is the membership degree value of the time span;

[0102] (5)Output the classification result

[0103] Based on the result of fuzzy inference, the determination of the abnormal data type is output, and the formula is as follows:

[0104] 。

[0105] Preferably, the method further includes: Step Five, realizing graphical visualization of data abnormal points at the host computer end:

[0106] Visualize the detection results in the form of charts, which enables users to observe the health status of the bridge structure in real time; save the detection results of abnormal data for subsequent analysis. The system supports displaying the complete monitoring process of sensor data through a graphical user interface (GUI), including function modules such as data reading, abnormal detection, repair results, and file saving.

[0107] Advantages of the present invention:

[0108] The present invention makes full use of the difference characteristics of the computing resources of the embedded platform and the host computer, utilizes the complementary advantages of real-time acquisition and processing at the embedded end and the rich computing resources of the host computer, develops a bridge abnormal detection algorithm based on AT-DBSCAN and a bridge abnormal classification algorithm based on fuzzy logic inference, decouples and divides the abnormal detection method and effectively deploys it to the embedded platform and the host computer. Among them, the embedded platform is responsible for real-time acquisition and abnormal detection of bridge data, and the host computer performs refined processing and analysis operations such as bridge abnormal classification and early warning, and constructs a graphical user interface (GUI) for bridge abnormal monitoring, realizing the whole-process visualization display from data acquisition to result analysis, providing comprehensive guarantee for the intelligent operation and maintenance and safety early warning of the bridge structure. Description of the Drawings

[0109] Figure 1 is the overall flowchart of the method used in the present invention;

[0110] Figure 2 is the annotation effect diagram of abnormal values by the density-based abnormal detection algorithm used in the present invention;

[0111] Figure 3 is the effect diagram of the GUI visualization platform involved in the present invention;

[0112] Figure 4 is the schematic diagram of the cable force sensor involved in the present invention;

[0113] Figure 5 is the schematic diagram of the single-chip microcomputer system in the present invention. Detailed Embodiments

[0114] The technical solution of the present invention will be further explained and illustrated below in the form of specific embodiments. In this experiment, the dataset uses the real data of the cable forces of the Heilongjiang Bridge, with a total of 48 measurement points upstream and downstream, and the monitoring duration is eight months. The dataset contains a total of 103,639 records. We divide the dataset into a training set, a validation set, and a test set according to a ratio of 7:2:1.

[0115] Step 1: Data preprocessing

[0116] Extract the original data of the cable force sensors of the Heilongjiang Bridge and first perform preprocessing on the original data, which specifically includes the following steps:

[0117] (a) Collect data frames

[0118] Convert the sensor analog signal into a digital signal and store it in the form of a data frame, including a timestamp, a sensor number, and a measured value. The embedded single-chip microcomputer reads the data frames obtained by extracting sensor data through a timing sampling window (for example, 500 milliseconds), and uses 2-point or 3-point differential signal processing to enhance the signal characteristics

[0119]

[0120]

[0121] where is the 2-point difference of the i-th point, is the 3-point difference of the i-th point.

[0122] Then send the obtained data to the next module.

[0123] (b) Adjust the output format

[0124] Organize the output format of the cable force data measured in a group of signals into , where n is the total number of data, represents the original value of the i-th data and the 2-point and 3-point difference values;

[0125] (c) Data weight calculation: w(x j ) is the weight of sample x j , which is determined by the data collection time, and the calculation formula is as follows:

[0126]

[0127] where represents the time interval between sample x j and the current moment, T is the forgetting threshold, and the sample points exceeding this time have a weight of 0.5;

[0128] (d) Calculate clustering parameters:

[0129] Determine the clustering parameters and calculate the total weight of the data as follows:

[0130]

[0131] Define the minimum number of points to be included and the scanning radius ;

[0132] (e) Apply the clustering algorithm:

[0133] Based on the data within the time interval at the current moment range, execute the AT-DBSCAN step process:

[0134] Utilize the determined scanning radius eps, forgetting threshold T, and minimum number of points to be included minPts;

[0135] For the current point, construct data points based on the cable force value and differential features, and find all nearby points within the scanning radius eps from it; assign weights to each point , when the total weight within the scanning radius eps of the nearby points ≥ the minimum number of points to be included minPts, it is considered a normal point; when the number of nearby points within the scanning radius eps < the minimum number of points to be included minPts, this point is marked as an abnormal point.

[0136] This step is to detect abnormal data points in bridge monitoring data in real time through a density-based spatial clustering algorithm. The dataset is the cable force data collected from sensors, and these data may be affected by noise or environmental changes. Therefore, appropriate algorithms must be used to accurately identify outliers. Applying the AT-DBSCAN density clustering method can identify those abnormal data points or noise points.

[0137] By calculating the density offset of adjacent data intervals, potential abnormal changes can be identified. The calculation of density offset can help determine the clustering structure between data points. When the offset between adjacent intervals is greater than a preset threshold, it can be determined that the data in this interval may be abnormal. Therefore, by dynamically adjusting the threshold, the clustering algorithm can flexibly handle different data characteristics and change trends.

[0138] In this step, a clustering algorithm is used to perform density analysis on the data. By clustering the data points and setting appropriate clustering parameters (such as the radius threshold ε and the minimum number of neighbors minPts), the algorithm can effectively distinguish normal data from abnormal data, and thus identify potential bridge structure problems. The detection results of outliers will be further used for bridge health assessment to achieve accurate structural anomaly early warning. The type of anomaly is judged through fuzzy logic reasoning.

[0139] Step 3: The wireless transmission module is used to transmit the cable force data with abnormal data marks obtained by the single-chip microcomputer module to the upper computer in a wireless transmission manner;

[0140] Step 4: The upper computer further classifies these abnormal points by using fuzzy logic reasoning to determine their specific abnormal types.

[0141] The upper computer further classifies these abnormal points by using fuzzy logic reasoning, and divides them into single-point abnormality, multi-point abnormality, data gain, data drift, and comprehensive multi-abnormality situations. Among them, the single-point abnormality refers to that a single data point deviates from the normal range, usually caused by accidental measurement errors or sudden environmental factors; the multi-point abnormality refers to that there are multiple abnormal points in a time series, but these abnormal points do not change the overall trend, usually caused by sensor failures; the data gain abnormality refers to that the data as a whole is significantly increased compared with the normal range, usually caused by sensor gain errors or system setting problems; the data drift abnormality refers to that the data gradually deviates from the normal range or trend, showing a systematic and trend-based deviation; the comprehensive multi-abnormality situation refers to the coexistence of multiple types of abnormalities, usually requiring a combination of multiple methods for comprehensive analysis. The specific steps of the fuzzy logic reasoning classification are as follows:

[0142] (1) Define input variables

[0143] Input the data obtained by the cable force sensor and the two-point difference and three-point difference, and measure the following indicators:

[0144] Change rate: Used to measure the change amplitude of the data over time, reflecting the fluctuation characteristics of the data. According to the change of the data within one hour, it is divided into three levels: high, medium, and low. More than 10% is a high change rate, less than 2% is a low change rate, and between 2% and 10% is a medium change rate;

[0145] Offset: Represents the degree to which the data deviates from the historical mean. According to the deviation between the current data and the historical mean, it is divided into three levels: high, medium, and low. Greater than 15% is a high offset, less than 5% is a low offset, and the deviation between 5% and 15% is a medium offset;

[0146] Time span: Measures the duration of abnormal data in the time series. According to the duration of the abnormality, it is divided into three levels: long, medium, and short. Continuing for more than 10 hours is a long time span, less than 3 hours is a short time span, and continuing for 3 - 10 hours is a medium time span;

[0147] (2) Fuzzify the input data

[0148] Convert the change rate, offset, and time span into membership values through the membership function. The membership function is expressed as:

[0149]

[0150]

[0151]

[0152] Where R represents the change rate, offset or time span of the input; is a pre-defined interval parameter, representing the interval boundaries of different membership sets, where represent 2%, 5%, and 3 hours respectively, represent 10%, 15%, and 10 hours respectively; represent the membership values of low, medium, and high levels respectively;

[0153] (3)Construct fuzzy rules

[0154] Construct fuzzy rules for each input variable and output variable:

[0155] Rule 1: IF the change rate IS high AND the offset IS small AND the time span IS short THEN the anomaly type IS single-point anomaly (linear relationship)

[0156] Rule 2: IF the change rate IS medium AND the offset IS medium AND the time span IS medium THEN the anomaly type IS multi-point anomaly (linear relationship)

[0157] Rule 3: IF the change rate IS low AND the offset IS large AND the time span IS long THEN the anomaly type IS data gain anomaly (linear relationship)

[0158] Rule 4: IF the change rate IS low AND the offset IS medium AND the time span IS long THEN the anomaly type IS data drift anomaly (linear relationship)

[0159] Rule 5: IF the change rate IS high AND the offset IS large AND the time span IS medium OR long THEN the anomaly type IS comprehensive multi-anomaly situation (linear relationship)

[0160] (4)Apply the TS model for inference

[0161] According to the membership degree of the input and the "IF-THEN" part of the rule, calculate the membership degree value and output value of each rule. The formula is as follows:

[0162]

[0163] Where is the output of the i-th rule; is the membership value of the rate of change, is the membership value of the offset, is the membership value of the time span;

[0164] (5)Output the classification result

[0165] According to the result of fuzzy inference, output the determination of the type of data anomaly, and the formula is as follows:

[0166] .

[0167] Step Five: Bridge structure anomaly detection and visualization

[0168] Through data processing, mark potential abnormal data points in the preprocessed image, and classify the type of anomaly according to Step Four. The observed data of the sensor is used as the feature input of the node to construct a Naive Bayes classifier for classifying the state of each node.

[0169]

[0170] Among them, represents the state of node i, and E is the observed data.

[0171] By calculating the posterior probability of each node, judge its health status. If the condition is met: , it is determined as an abnormal node, where T2 is the abnormal determination threshold. Combining the determination results of each node, comprehensively evaluate the safety of the entire bridge structure. If multiple key nodes are determined to be abnormal, the system gives a warning that there are structural safety hazards in the bridge. Finally, output a safety assessment report of the bridge structure, including the determined abnormal parts, abnormal levels, and recommended maintenance measures.

[0172] Extract the above data and transmit it to the visualization system for real-time visualization display.

[0173] Visualize the detection results in the form of charts for users to observe the health status of the bridge structure in real time; save the detection results of abnormal data for subsequent analysis; the system supports displaying the complete monitoring process of sensor data through a graphical user interface (GUI), including function modules such as data reading, anomaly detection, repair results, and file saving.

[0174] Experimental results:

[0175] In this instance experiment, by comparing the designed AT-DBSCAN algorithm with other anomaly detection methods (3-Sigma, Support Vector Machine (SVM), K-means, and Isolation Forest), experiments on execution time, precision, false alarm rate, and recall rate were conducted to evaluate its efficiency and real-time performance in practical applications. The experiments were carried out under the same dataset and processing conditions to ensure the comparability of the results.

[0176] The specific results are shown in Table 1, which details the execution time (in milliseconds) of each algorithm when processing the same dataset, as well as the values of precision, false alarm rate, and recall rate. By comparing the execution times and metrics of these algorithms, the performance differences of different algorithms in the anomaly detection task can be clarified, providing a basis for selecting the appropriate algorithm. The experimental results are shown in Table 1. The experiments show that the proposed method has performance advantages over other methods in terms of execution time precision and false alarm rate, and has a high recall rate, meeting the requirements of bridge anomaly detection and classification for accuracy and efficiency.

[0177] Table

[0178]

[0179] Facing the urgent needs in the field of bridge safety monitoring and intelligent operation and maintenance in China, this paper aims to construct an efficient and automated bridge anomaly detection system. Therefore, a lightweight clustering anomaly data detection algorithm, AT-DBSCAN, is proposed to detect bridge sensor anomaly data in real time and efficiently. The aim is to implement an integrated hardware and software prototype system for anomaly detection, decouple and effectively deploy the anomaly detection method to the embedded device and the host computer, and realize full-process visualization in combination with the GUI interface. This system can provide decision-making support for the intelligent operation and maintenance of bridges and give timely safety warnings when the structure shows anomalies, thus ensuring the long-term safe operation of bridges.

Claims

1. An embedded bridge cable force data abnormal monitoring device, characterized in that, The device includes a sensor module, an embedded platform, a wireless transmission module, a power supply battery, and a host computer; Among them, the sensor module consists of several cable force sensors, which are arranged at equal intervals on the bridge body, and collect cable force data at the same phase interval; The embedded platform converts the original analog signal of the cable force data collected by the cable force sensors connected to it into a digital signal, and each dimension corresponds to the cable force data collected by a cable force sensor; the embedded platform stores in the form of data frames, and the data frames include timestamps, sensor numbers, and measured values, and the embedded platform performs real-time processing and analysis on the collected cable force data, and marks the abnormal data; The wireless transmission module is used to transmit the cable force data with abnormal data marks analyzed and processed by the single-chip microcomputer module to the host computer in a wireless transmission manner; The host computer further classifies these abnormal points by using fuzzy logic reasoning to determine their specific abnormal types.

2. The abnormal monitoring device for cable force data of a bridge based on an embedded system according to claim 1, characterized in that, The embedded platform uses 2-point or 3-point differential signal processing to enhance signal characteristics wherein is the two-point difference of the i-th point, is the three-point difference of the i-th point.

3. The abnormal monitoring device for bridge cable force data based on embeddedness according to claim 2, wherein, The embedded platform uses the density-based spatial clustering algorithm for adaptive time series data AT-DBSCAN clustering method to perform real-time processing and analysis on the collected cable force data, and mark the abnormal data. The specific steps are as follows: Step 1: Data preprocessing and formatting (1)Input data processing: The original data of the cable force sensor with a time interval of from a certain moment and the differential data between point 2 and point 3 are represented as , where n is the total number of data, represents the original value of the i-th data and the differential value between point 2 and point 3; (2) Data weight calculation: w(x j ) is the weight of sample x j , which is determined by the data collection time, and the calculation formula is as follows: wherein represents the time interval of the sample x j with the current moment, T being the forgetting threshold, and the sample points beyond this time have a weight of 0.5; (3) Calculate clustering parameters: Determine the clustering parameters, and calculate the total weight of the data as: Define the minimum number of included points and the scanning radius ; (4) Apply the clustering algorithm: Execute the AT-DBSCAN step process for data within the time interval of the current moment : Use the determined scanning radius eps, forgetting threshold T, and minimum inclusion points minPts; For the current point, construct data points based on the cable force value and differential features, and find all nearby points within a distance of the scanning radius eps; assign weights to each point , when the total weight within the scanning radius eps of the nearby points ≥ the minimum number of included points minPts, it is considered a normal point; when the number of nearby points within the scanning radius eps < the minimum number of included points minPts, the point is marked as an abnormal point.

4. The abnormal monitoring device for cable force data of a bridge based on an embedded system according to claim 1, characterized in that, The host computer further classifies these abnormal points by using fuzzy logic reasoning, and divides them into single-point anomalies, multi-point anomalies, data gain, data drift, and comprehensive multi-anomaly situations. Among them, the single-point anomaly refers to a single data point deviating from the normal range, usually caused by accidental measurement errors or sudden environmental factors; multi-point anomalies refer to multiple abnormal points in a time series, but these abnormal points do not change the overall trend, usually caused by sensor failures; data gain anomalies refer to the data as a whole being significantly larger than the normal range, usually caused by sensor gain errors or system setting problems; data drift anomalies refer to the data gradually deviating from the normal range or trend, showing systematic and trend-based deviations; comprehensive multi-anomaly situations refer to the simultaneous existence of multiple types of anomalies, usually requiring a combination of multiple methods for comprehensive analysis; the specific steps of the fuzzy logic reasoning classification are as follows: (1) Define input variables Input the data obtained from the cable force sensors and 2-point difference and 3-point difference to measure the following indicators: Change rate: Used to measure the change range of data over time, reflecting the fluctuation characteristics of the data. According to the change of data within one hour, it is divided into three levels: high, medium, and low. More than 10% is a high change rate, less than 2% is a low change rate, and between 2% and 10% is a medium change rate; Offset: Represents the degree of deviation of the data from the historical mean. According to the deviation between the current data and the historical mean, it is divided into three levels: high, medium, and low. Greater than 15% is a high offset, less than 5% is a low offset, and the deviation is between 5% and 15% is a medium offset; Time span: Measures the duration of abnormal data in a time series. According to the duration of the anomaly, it is divided into three categories: long, medium, and short. If the duration exceeds 10 hours, it is a long time span; if it is less than 3 hours, it is a short time span; if it lasts from 3 to 10 hours, it is a medium time span. (2)Input data fuzzification membership degree Convert the rate of change, offset, and time span into membership degree values through the membership degree function. The membership degree function is expressed as: where R represents the rate of change, offset, or time span of the input; is a pre-defined interval parameter representing the interval boundaries of different membership sets, where represent 2%, 5%, and 3 hours respectively, represent 10%, 15%, and 10 hours respectively; represent the membership degree values for the low, medium, and high levels respectively; (3)Construct fuzzy rules Construct fuzzy rules for each input variable and output variable: Rule 1: IF the rate of change IS high AND the offset IS small AND the time span IS short THEN the anomaly type IS single-point anomaly (linear relationship) Rule 2: IF the rate of change IS medium AND the offset IS medium AND the time span IS medium THEN the anomaly type IS multi-point anomaly (linear relationship) Rule 3: IF the rate of change IS low AND the offset IS large AND the time span IS long THEN the anomaly type IS data gain anomaly (linear relationship) Rule 4: IF the rate of change IS low AND the offset IS medium AND the time span IS long THEN the anomaly type IS data drift anomaly (linear relationship) Rule 5: IF the rate of change IS high AND the offset IS large AND the time span IS medium OR long THEN the anomaly type IS comprehensive multi-anomaly situation (linear relationship) (4)Apply the TS model for inference According to the input membership degree and the "IF-THEN" part of the rule, calculate the membership degree value and output value of each rule. The formula is as follows: where is the output of the i-th rule; is the membership value of the rate of change, is the membership value of the offset, is the membership value of the time span; (5)Output classification result According to the result of fuzzy inference, output the determination of the data anomaly type. The formula is as follows: 。 5. A monitoring method for the abnormal monitoring device of cable force data of a bridge based on an embedded system as described in claim 1, characterized in that, The steps of this method are as follows: Step 1: Arrange a number of cable force sensors at equal intervals on the bridge body, and collect cable force data at the same time phase interval. Step 2: The embedded platform converts the original analog signal of the cable force data collected by the cable force sensors connected to it into a digital signal. Each dimension corresponds to the cable force data collected by a cable force sensor. The embedded platform stores in the form of data frames. The data frames include timestamps, sensor numbers, and measurement values. The embedded platform performs real-time processing and analysis on the collected cable force data and marks the abnormal data. Step 3: The wireless transmission module is used to transmit the cable force data with abnormal data marks analyzed and processed by the single-chip microcomputer module to the upper computer in a wireless transmission manner. Step 4: The upper computer further classifies these abnormal points using fuzzy logic inference to determine their specific anomaly types.

6. The monitoring method of the embedded-based abnormal monitoring device for cable force data of bridges according to claim 5, characterized in that, In Step 2, the embedded platform converts the cable force data into 2-point or 3-point differential signals. wherein is the two-point difference of the i-th point, is the three-point difference of the i-th point, is the cable force data of the i-th cable force sensor.

7. The monitoring method of the embedded-based abnormal cable force data monitoring device for bridges according to claim 5, characterized in that, In Step 2, the method for the embedded platform to perform real-time processing and analysis on the collected cable force data and mark the abnormal data is as follows: Step 1: Data preprocessing and formatting (1)Input data processing: Input the original data of the cable force sensor with a time interval of from a certain moment, as well as the differential data between point 2 and point 3, expressed as , where n is the total number of data, represents the original value of the i-th data and the differential value between point 2 and point 3; (2) Data weight calculation: w(x j ) is the weight of sample x j , which is determined by the data collection time. The calculation formula is as follows: wherein represents the time interval of the sample x j with the current moment, T being the forgetting threshold, and the sample points beyond this time have a weight of 0.5; (3)Calculate clustering parameters: Determine the clustering parameters and calculate the total weight of the data as: Define the minimum number of included points and the scanning radius ; (4): Apply the clustering algorithm Execute the AT-DBSCAN step process for data within the time interval of the current moment: within the range: Using the determined scanning radius eps, forgetting threshold T, and minimum number of included points minPts; For the current point, construct data points based on the cable force value and differential features, and find all nearby points within a distance of the scanning radius eps; assign weights to each point , when the total weight within the scanning radius eps of the nearby points ≥ the minimum number of included points minPts, it is considered a normal point; when the number of nearby points within the scanning radius eps < the minimum number of included points minPts, the point is marked as an abnormal point.

8. The monitoring method of the embedded-based abnormal monitoring device for bridge cable force data according to claim 5, characterized in that, In step 4, the host computer further classifies these abnormal points through fuzzy logic reasoning, dividing them into single-point anomalies, multi-point anomalies, data gain, data drift, and comprehensive multi-anomaly situations. Among them, single-point anomalies refer to individual data points deviating from the normal range, usually caused by accidental measurement errors or sudden environmental factors; multi-point anomalies refer to multiple abnormal points in a time series, but these abnormal points do not change the overall trend, usually caused by sensor failures; data gain anomalies refer to the overall data being significantly larger than the normal range, usually caused by sensor gain errors or system setting problems; data drift anomalies refer to the data gradually deviating from the normal range or trend, manifested as systematic and trend-based deviations; comprehensive multi-anomaly situations refer to the simultaneous existence of multiple types of anomalies, usually requiring a combination of multiple methods for comprehensive analysis. The specific steps of the fuzzy logic reasoning classification are as follows: (1) Define input variables Input the data obtained from the cable force sensor, as well as two-point differences and three-point differences, and measure the following indicators: Rate of change: Used to measure the change amplitude of data over time, reflecting the fluctuation characteristics of the data. According to the change of data within one hour, it is divided into three levels: high, medium, and low. A change exceeding 10% is a high rate of change, a change less than 2% is a low rate of change, and a change between 2% - 10% is a medium rate of change; Offset: Represents the degree to which the data deviates from the historical mean. According to the deviation between the current data and the historical mean, it is divided into three levels: high, medium, and low. A deviation greater than 15% is a high offset, a deviation less than 5% is a low offset, and a deviation between 5% - 15% is a medium offset; Time span: Measures the duration of abnormal data in the time series. According to the duration of the anomaly, it is divided into three levels: long, medium, and short. A duration exceeding 10 hours is a long time span, a duration less than 3 hours is a short time span, and a duration between 3 - 10 hours is a medium time span; (2) Fuzzify the input data Convert the rate of change, offset, and time span into membership values through the membership function. The membership function is expressed as: where R represents the rate of change, offset or time span of the input; are predefined interval parameters, representing the interval boundaries of different membership sets, where represent 2%, 5%, and 3 hours respectively, represent 10%, 15%, and 10 hours respectively; represent the membership degree values of the low, medium, and high levels respectively; (3) Construct fuzzy rules Construct fuzzy rules for each input variable and output variable: Rule 1: IF the rate of change IS high AND the offset IS small AND the time span IS short THEN the anomaly type IS single-point anomaly (linear relationship) Rule 2: IF the rate of change IS medium AND the offset IS medium AND the time span IS medium THEN the anomaly type IS multi-point anomaly (linear relationship) Rule 3: IF the rate of change IS low AND the offset IS large AND the time span IS long THEN the anomaly type IS data gain anomaly (linear relationship) Rule 4: IF the rate of change IS low AND the offset IS medium AND the time span IS long THEN the anomaly type IS data drift anomaly (linear relationship) Rule 5: IF the rate of change IS high AND the offset IS large AND the time span IS medium OR long THEN the anomaly type IS a comprehensive multi-anomaly situation (linear relationship) (4) Apply the TS model for inference Based on the input membership degrees and the "IF-THEN" part of the rules, calculate the membership degree values and output values of each rule. The formulas are as follows: wherein is the output of the i-th rule; is the membership value of the rate of change, is the membership value of the offset, is the membership value of the time span; (5) Output the classification result Based on the results of fuzzy inference, output the determination of the data anomaly type. The formula is as follows: 。 9. The monitoring method of the embedded-based abnormal cable force data monitoring device for bridges according to claim 5, characterized in that, This method further includes: Step Five, implement graphical visualization of data anomaly points at the host computer side: Visualize the detection results in the form of charts, enabling users to observe the health status of the bridge structure in real time; save the detection results of abnormal data for subsequent analysis.

Images