Dike safety monitoring and maintenance system and method based on multi-sensor fusion

Abstract

The invention discloses a dike safety monitoring and maintenance system and method based on multi-sensor fusion, and belongs to the technical field of hydraulic engineering safety. Comprising a pore water pressure sensor, a horizontal displacement sensor, a vertical displacement sensor, a seepage monitoring sensor, a data acquisition and transmission module and a data acquisition center. And the precision of the sensor is high, for example, the precision of a pore water pressure sensor is within + / -0.1 kPa. And the transmission module adopts a hybrid communication network which is mainly composed of an optical fiber and is assisted by ZigBee and LoRa wireless. The data acquisition center is provided with a high-performance processor and the like. The method comprises the steps of data acquisition, preprocessing, anomaly detection and early warning, risk assessment, maintenance strategy formulation and the like. Data are processed through joint filtering, dynamic normalization and other algorithms, the safety state is judged through clustering analysis and a composite early warning model, and risks are evaluated by integrating multiple factors. According to the method, the accuracy, scientificity and effectiveness of dike safety monitoring and maintenance are remarkably improved, and a powerful guarantee is provided for hydraulic engineering safety management.

Description

Technical Field

[0001] The present invention relates to the field of water conservancy engineering safety technology, and in particular to a levee safety monitoring and maintenance system and method based on multi-sensor fusion. Background Art

[0002] As an important engineering facility for flood control, the safety of levees is directly related to the safety of life and property of people in the surrounding areas and the social and economic stability. Traditional levee monitoring and maintenance methods have many shortcomings, such as limited accuracy of monitoring data, untimely warning, insufficient scientific and comprehensive risk assessment, and lack of targeted maintenance strategies. With the rapid development of sensor technology, communication technology and intelligent algorithms, it is necessary to develop an efficient and intelligent levee safety assurance system and method to improve the level of levee safety management. Summary of the invention

[0003] The purpose of the present invention is to provide a levee safety monitoring and maintenance system and method based on multi-sensor fusion, which solves the problems of poor accuracy of monitoring data, untimely and inaccurate anomaly detection and early warning, and incomplete risk assessment.

[0004] To achieve the above-mentioned object, the present invention provides a levee safety monitoring and maintenance system based on multi-sensor fusion, including a pore water pressure sensor installed inside the levee body, for monitoring the water pressure changes inside the levee body; Horizontal displacement sensors installed on the top and slope of the embankment are used to monitor the lateral displacement of the embankment; Vertical displacement sensors installed at key locations of the embankment foundation and embankment body are used to monitor the settlement of the embankment body; Seepage monitoring sensors installed on the embankment are used to monitor the seepage conditions of the embankment; The data acquisition and transmission module is used to transmit the data collected by the sensor to the data acquisition center.

[0005] Preferably, the pore water pressure sensor is a high-precision MEMS pore water pressure sensor, and its measurement accuracy is within ±0.1 kPa.

[0006] Preferably, the horizontal displacement sensor is a laser interferometer displacement sensor, and its resolution is not less than 0.01 mm.

[0007] Preferably, the vertical displacement sensor is a high-precision static level with an accuracy within ±0.05 mm.

[0008] Preferably, the seepage monitoring sensor is an intelligent pressure measuring tube, which has a built-in microprocessor and a water flow sensor, and can sense changes in water flow speed and direction, and make a preliminary judgment on the type of seepage anomaly.

[0009] Preferably, the data acquisition and transmission module adopts a hybrid communication network, which includes: The transmission rate of optical fiber communication as the backbone is not less than 10Gbps; As a supplementary ZigBee wireless sensor network, the operating frequency band is 2.4GHz and the transmission distance is 100 meters; As a supplementary LoRa wireless communication module, the operating frequency band is 433MHz / 868MHz / 915MHz, and the transmission distance is 15 kilometers; Ad hoc networking and adaptive routing algorithms are used to ensure reliable data transmission.

[0010] Preferably, the data collection center includes: High-performance multi-core processors and data processing servers; Real-time data processing software based on Linux operating system; The Hadoop distributed file system is used to store data, and multiple copies of data blocks are set; MySQL database, used to record data collection time, sensor location and equipment status.

[0011] The present invention also provides a dike safety monitoring and maintenance method based on multi-sensor fusion, comprising the following steps: S1. Use the data acquisition and transmission module to obtain data collected by the water pressure sensor, horizontal displacement sensor, vertical displacement sensor and seepage monitoring sensor; S2, using data preprocessing algorithms to process the collected raw data, including filtering and normalization processing; S3, using anomaly detection and early warning algorithms to detect anomalies in the processed data and issue early warning information; S4. Use risk assessment algorithms to assess levee risks; S5. Develop and implement maintenance strategies based on the assessment results.

[0012] Preferably, in step S2, the data preprocessing algorithm includes: a joint filtering algorithm, which first uses a wavelet filtering algorithm to remove high-frequency noise, and then uses a Kalman filtering algorithm for state prediction and correction; a dynamic normalization algorithm, which normalizes data of different physical dimensions to the [0,1] interval according to the real-time change range; a data missing value processing algorithm, which uses an interpolation algorithm based on support vector regression to fill in the missing values ​​when data is missing.

[0013] Preferably, in step S3, the anomaly detection and early warning algorithm includes: performing cluster analysis in combination with the DBSCAN density clustering algorithm and the K-Means clustering algorithm to determine the distribution range of normal data; a composite early warning model based on the long short-term memory network LSTM and the multi-layer perceptron MLP, taking the preprocessed sensor data as input, and outputting the safety status information of the levee.

[0014] Therefore, the present invention adopts a multi-sensor fusion-based levee safety monitoring and maintenance system and method with the above structure, which has the following beneficial effects: The present invention adopts high-precision sensors, hybrid communication networks, advanced data processing and intelligent algorithms, and refined maintenance strategies, which significantly improve the accuracy of levee monitoring data, enhance the stability and flexibility of data transmission, optimize data processing and analysis capabilities, improve anomaly detection and early warning effects, achieve more scientific and comprehensive risk assessment, and improve the pertinence and effectiveness of maintenance strategies, forming a dynamically optimized integrated system for levee safety monitoring, early warning and maintenance, which effectively guarantees levee safety and provides a more complete and efficient solution for water conservancy project safety management. The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a system and method for monitoring and maintaining levee safety based on multi-sensor fusion according to the present invention; Figure 2 It is a flow chart of a levee safety monitoring and maintenance system and method based on multi-sensor fusion according to the present invention. DETAILED DESCRIPTION

[0016] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0017] Unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example like Figure 1 As shown, the present invention provides a levee safety monitoring and maintenance system based on multi-sensor fusion, which is specifically shown as follows: Pore ​​water pressure sensor inside the embankment: High-precision MEMS pore water pressure sensor is used, and its measurement accuracy can reach ±0.1kPa, which can keenly capture the tiny water pressure changes inside the embankment. It is arranged in layers every 10 meters in the longitudinal direction and every 3 meters in the transverse direction of the embankment to form a three-dimensional monitoring network to ensure that there are no blind spots in monitoring, and to comprehensively and accurately obtain water pressure data at different positions of the embankment, providing a detailed basis for analyzing the seepage status inside the embankment.

[0019] Displacement sensor: The horizontal displacement sensor uses a laser interferometer displacement sensor with a resolution of up to 0.01mm. It is installed on the preset concrete base of the embankment top and embankment slope, and monitors the lateral displacement of the embankment by accurately measuring the phase change of the reflected laser. The vertical displacement sensor uses a high-precision static level with an accuracy of ±0.05mm. It is installed at the key structural nodes of the embankment foundation and embankment body, and uses the connecting pipe principle to measure the height difference between each measuring point, thereby achieving high-precision monitoring of the embankment settlement and timely discovering the slight deformation trend of the embankment body.

[0020] Seepage monitoring sensor: The intelligent pressure measuring tube has a built-in microprocessor and water flow sensor, which can not only monitor the water pressure in real time, but also sense the changes in water flow speed and direction through the water flow sensor. When the water flow speed changes suddenly or there is an abnormal flow direction, the microprocessor uses the built-in gradient algorithm to preliminarily determine whether there is a concentrated seepage channel, and classifies the type of seepage anomaly according to the water pressure change pattern, such as piping, leakage, etc., to provide key information for subsequent maintenance measures.

[0021] Soil moisture sensor: A capacitive soil moisture sensor with a measurement accuracy of ±2% is used. It is layered at different depths (0.5m, 1m, 1.5m) in the soil around the embankment, with a horizontal interval of 5 meters to form a soil moisture monitoring profile. The soil moisture is accurately monitored by measuring the change in the soil dielectric constant. Combined with the soil mechanics model, it provides multi-dimensional data support for assessing the stability of the soil around the embankment, and provides early warning of risks such as landslides caused by changes in soil moisture content.

[0022] Hybrid communication architecture: Build a hybrid communication network with optical fiber as the backbone and wireless as the branch. Optical fiber uses single-mode optical fiber with a transmission rate of up to 10Gbps, ensuring high-speed and stable data transmission to the data collection center. In areas where optical fiber is difficult to lay, such as complex terrain on embankments or temporary monitoring points, ZigBee wireless sensor networks (working frequency band 2.4GHz, transmission distance up to 100 meters) and LoRa wireless communication modules (working frequency band 433MHz / 868MHz / 915MHz, transmission distance up to 15 kilometers) are deployed as supplements. Through self-organizing networks and adaptive routing algorithms, reliable transmission of sensor data is achieved to ensure the integrity of data collection.

[0023] Adaptive acquisition frequency mechanism: The system has a built-in intelligent data monitoring module to analyze the trend of sensor data changes in real time. For displacement data, when the change rate of data collected for three consecutive times exceeds 0.05mm / h, the acquisition frequency will be automatically increased from once per hour to once every 30 minutes; for pore water pressure data, if the pressure change exceeds 0.5kPa or the fluctuation frequency is greater than 0.1Hz within 15 minutes, the acquisition frequency will be increased from once every 10-15 minutes to once every 5 minutes. Conversely, the acquisition frequency will be appropriately reduced after the data is stable for 2 hours to optimize system resource utilization and data processing efficiency.

[0024] Data Collection Center: Data processing servers equipped with high-performance multi-core processors and large-capacity high-speed memory run real-time data processing software based on the Linux operating system. The Hadoop distributed file system is used for data storage, and data is divided into blocks and stored on multiple storage nodes. Three copies are set for each data block to improve the redundancy and reliability of data storage. At the same time, the MySQL database is used to record key information such as data collection time, sensor location, equipment status, etc., and data indexes are established to facilitate quick query and retrieval.

[0025] like Figure 2 As shown, the present invention also provides a dike safety monitoring and maintenance method based on multi-sensor fusion, which is specifically shown as follows: Joint filtering algorithm: First, the wavelet filter uses the Daubechies4 wavelet basis to decompose the original data into three layers, removes high-frequency noise through threshold quantization processing, and retains the main features of the signal. Then, based on the linear system state space model, the Kalman filter uses the state estimate value of the previous moment and the measured value of the current moment to make the optimal estimate, continuously correct the system state, further reduce noise interference, improve the accuracy and stability of the data, and provide a reliable data basis for subsequent analysis.

[0026] Dynamic normalization method: Design a dynamic normalization algorithm to monitor the changes in the maximum and minimum values ​​of the data in real time. Update the normalization parameters every hour, normalize the data of different physical dimensions such as displacement and pore water pressure to the [0,1] interval, ensure that the data is processed under the same dimension, avoid algorithm deviations caused by dimensional differences, and improve the applicability and accuracy of the intelligent algorithm.

[0027] Data missing value processing: When sensor data is missing, an interpolation algorithm based on support vector regression (SVR) is used. The training sample set is constructed using historical data and adjacent sensor data, and the missing values ​​are predicted by training the SVR model. The radial basis function (RBF) is selected as the kernel function, and the penalty parameter C and kernel function parameter γ are adjusted to make the model achieve the minimum mean square error on the training set, ensure the accuracy of filling the missing values, and maintain the continuity and integrity of the data.

[0028] Cluster analysis: The DBSCAN density clustering algorithm sets the neighborhood radius to 0.2 (determined based on data distribution characteristics and experience) and the minimum number of sample points to 5. The data is preliminarily clustered to identify core points, boundary points, and noise points. The core point data is input into the K-Means clustering algorithm, and the number of cluster centers is set to 3 (determined based on the distribution pattern of normal data and the clustering results of historical data). The distance from the sample to the cluster center is iteratively calculated, the cluster center is updated, and the distribution range of normal data is accurately determined. When the distance between the new data point and the nearest cluster center is greater than 1.5 times the average cluster radius, it is determined to be a possible anomaly and the early warning mechanism is triggered.

[0029] Composite early warning model: A composite early warning model based on long short-term memory network (LSTM) and multi-layer perceptron (MLP) was constructed. The LSTM network has a memory function and can learn the time series characteristics of sensor data. Its hidden layer is set with 128 neurons and uses the tanh activation function. The preprocessed time series data is input into the LSTM network, and after processing for multiple time steps, a feature vector is output. The feature vector is used as the input of the MLP. The MLP has two hidden layers, containing 64 and 32 neurons respectively. The ReLU activation function is used to finally output the safety status of the levee (safe, careful, dangerous). Through the training of a large amount of historical data (including different working conditions and abnormal conditions), the model continuously adjusts the weights and biases to improve the accuracy of the judgment of the safety status of the levee and achieve timely and effective early warning.

[0030] Risk Assessment Algorithm: In the evaluation index system, in addition to the stability of the levee and the safety of seepage, new environmental factor indicators are added, such as river flow changes (obtained through hydrological monitoring station data), stability of surrounding geological structures (based on geological survey reports and earthquake monitoring data), meteorological conditions (rainfall, wind speed, etc.), etc. For each indicator, the membership is determined by a trapezoidal membership function based on sensor monitoring data, statistical analysis of historical data and expert experience. For example, for the levee displacement indicator, when the displacement is less than 5mm, the membership is 1 (safe); when the displacement is between 5-10mm, the membership decreases linearly; when the displacement is greater than 10mm, the membership is 0 (dangerous). By constructing a multi-level fuzzy relationship matrix, the fuzzy relationship between each indicator and the levee risk is fully reflected.

[0031] Determine weights using the analytic hierarchy process: Use the analytic hierarchy process to construct a judgment matrix, and invite 10 experts in the field of water conservancy projects to compare and score the relative importance of each indicator. Use the 1-9 scale method to test the consistency of the judgment matrix, calculate the eigenvector, and determine the weights of indicators such as embankment stability, seepage safety, river flow changes, surrounding geological structure stability, and meteorological conditions as 0.3, 0.25, 0.15, 0.15, and 0.15, respectively. Through fuzzy synthesis operations, such as the weighted average method, the comprehensive risk level of the embankment (low risk, medium risk, high risk) is obtained, providing a scientific basis for the formulation of maintenance strategies.

[0032] Protection strategy formulation Low-risk maintenance: For low-risk situations where individual sensor data is slightly abnormal, professional inspectors with water conservancy project inspection qualifications and rich experience are arranged, equipped with high-precision handheld measuring instruments (such as portable levels, total stations, etc.), to conduct inspections every 30 minutes according to the predetermined route. At the same time, multi-rotor drones equipped with 4K high-definition cameras and infrared thermal imagers are used to conduct all-round aerial inspections at a height of 50 meters from the levee. The drone automatically collects images and thermal imaging data according to the preset flight trajectory and shooting angle, and uses image recognition algorithms to analyze whether there are abnormal signs such as cracks, collapse, leakage, etc. on the surface of the levee. Combined with the trend of sensor data changes, it is determined whether further measures need to be taken.

[0033] Medium-risk maintenance: In the case of medium-risk situations with local seepage anomalies or small displacements, a professional technical team consisting of water conservancy engineers, geotechnical engineers and structural engineers is organized to carry out on-site inspections with non-destructive testing equipment such as geological radar and acoustic wave detectors. Geological radar uses high-frequency electromagnetic waves to detect the internal structure of the levee, and acoustic wave detectors analyze internal defects by transmitting and receiving acoustic wave signals. According to the test results, the scope and severity of the problem are determined. For the seepage area, a new type of nano-scale anti-filtration material is used for treatment. Its pore size is precisely controllable, which can effectively prevent the loss of fine particles while ensuring smooth drainage; for the displacement area, carbon fiber composite materials are used for local reinforcement. The tensile strength of carbon fiber cloth is greater than 3500MPa. It is closely combined with the levee structure through the bonding process to enhance the bearing capacity of the levee structure.

[0034] High-risk maintenance: When the levee shows signs of overall instability, the emergency plan will be immediately activated. Evacuation notices will be sent to residents and units in potentially affected areas through the emergency broadcast system (covering a radius of 5 kilometers, with a sound intensity greater than 100dB) and the mobile phone SMS group messaging platform (cooperating with local operators to ensure timely delivery of information). At the same time, a rescue team of no less than 100 people will be organized and intelligent rescue equipment will be called. The automated riprap machine uses GPS positioning and automatic control technology, and can accurately throw 1 cubic meter of stones to the designated location within 10 minutes for riprap and footing; the mechanized sandbag stacking equipment can stack more than 500 sandbags per hour, quickly build a temporary protective embankment on the embankment, stabilize the embankment, and prevent breach accidents.

[0035] Finally, the data collection center uses big data analysis technology to conduct a comprehensive analysis of sensor data, on-site image and video data, and maintenance personnel operation records after the implementation of maintenance measures. A maintenance strategy evaluation model based on decision trees and neural networks is established, with the type of maintenance measures, implementation time, environmental conditions, etc. as input features, and the change of levee stability indicators as output targets. The model is trained to evaluate the effectiveness of the maintenance strategy. If the maintenance measures do not achieve the expected results, such as the displacement of the levee body continues to increase or the seepage situation does not improve, the risk assessment algorithm is reused to combine the latest data to assess the risk level, and optimization methods such as genetic algorithms are used to formulate new maintenance plans. At the same time, the actual data and experience during the maintenance process are fed back to the intelligent algorithm model, and the model parameters are adjusted and optimized to improve the accuracy and reliability of early warning and risk assessment.

[0036] Working principle: 1. System installation and debugging Sensor installation: Install sensors on the levee according to design requirements. When drilling holes in the levee to install pore water pressure sensors and displacement sensors, ensure that the drilling depth, verticality and aperture meet the standards. Use high-strength epoxy resin to fix the sensors to ensure that the sensors are tightly combined with the levee and do not affect the structural integrity of the levee. Seepage monitoring sensors and soil moisture sensors are installed in pre-excavated pits or pipes, and waterproof, moisture-proof and anti-interference treatments are done. The sensor data cable and power cable are connected, and initial calibration and testing are performed.

[0037] Communication network construction: When laying optical fiber, rationally plan the route along the dike, avoid possible sources of damage, and use underground or overhead laying to ensure stable and reliable optical fiber connection. In the area where wireless communication equipment is installed, configure the parameters of ZigBee and LoRa modules, establish an ad hoc network, test the wireless signal strength and transmission stability, and ensure that data can be accurately transmitted to the data collection center.

[0038] Data collection center setup: Build a data collection center computer room at a suitable location, install data processing servers, storage devices and network equipment, install and configure system software and databases, debug data collection programs, ensure that data from sensors can be received and processed in real time, and correctly stored and managed.

[0039] (II) Algorithm training and optimization Data collection and preprocessing: Collect at least one year of historical levee monitoring data, including data in different seasons, different water levels and different working conditions, and perform cleaning, filtering, normalization and missing value processing according to the data preprocessing algorithm to build a high-quality training data set.

[0040] Model training: Use training data sets to train anomaly detection and warning models and risk assessment models. For the composite warning model, use the back propagation algorithm (BP) to adjust the weights and biases of LSTM and MLP, set appropriate learning rates (such as 0.001) and training rounds (such as 1000 rounds), and verify model performance through the validation set to prevent overfitting. For the risk assessment model, determine the membership function of the fuzzy comprehensive evaluation method and the judgment matrix of the hierarchical analysis method based on expert experience and data characteristics, continuously optimize weight distribution, and improve model evaluation accuracy.

[0041] Model optimization and update: During the operation of the system, new monitoring data and maintenance feedback data are collected regularly (e.g. monthly) to perform incremental learning and optimization on the model. According to the characteristics of the new data and the model performance indicators, the algorithm parameters and structure are adjusted, such as increasing the number of hidden layer neurons in the LSTM network or improving the indicator system of the fuzzy comprehensive evaluation method, to ensure that the model can adapt to the dynamic changes in the safety status of the levee.

[0042] (III) Maintenance operation process Daily inspection and monitoring: During normal operation, according to the maintenance strategy requirements, inspection personnel and drones regularly inspect and monitor the levees. The data collection center analyzes and processes data in real time, and intelligent algorithms determine the safety status of the levees. If any abnormalities are found, early warnings will be issued to notify maintenance personnel in a timely manner.

[0043] Risk response and maintenance implementation: When receiving an early warning or risk assessment result, the corresponding maintenance strategy is quickly initiated. According to the task instructions received by the smart wearable device, the maintenance personnel bring tools and materials to the site, perform maintenance operations according to the predetermined plan, and use AR technology to assist construction to ensure the accurate implementation of maintenance measures. During the maintenance process, the sensor data is continuously monitored, the maintenance operation details and data changes are recorded, and the relevant information is fed back to the data collection center after the maintenance is completed.

[0044] Strategy adjustment and experience summary: The data collection center uses big data analysis and intelligent algorithms to evaluate and adjust maintenance strategies based on maintenance feedback data, summarize successful experiences and lessons from failures, update algorithm models and maintenance knowledge bases, and continuously improve the overall performance and maintenance effects of the system.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A levee safety monitoring and maintenance system based on multi-sensor fusion, characterized in that: It includes a pore water pressure sensor installed inside the embankment body, which is used to monitor the water pressure change inside the embankment body; Horizontal displacement sensors installed on the top and slope of the embankment are used to monitor the lateral displacement of the embankment; Vertical displacement sensors installed at key locations of the embankment foundation and embankment body are used to monitor the settlement of the embankment body; Seepage monitoring sensors installed on the embankment are used to monitor the seepage conditions of the embankment; The data acquisition and transmission module is used to transmit the data collected by the sensor to the data acquisition center.

2. The levee safety monitoring and maintenance system based on multi-sensor fusion according to claim 1 is characterized by: The pore water pressure sensor is a high-precision MEMS pore water pressure sensor, and its measurement accuracy is within ±0.1 kPa.

3. The levee safety monitoring and maintenance system based on multi-sensor fusion according to claim 1 is characterized by: The horizontal displacement sensor is a laser interference displacement sensor, and its resolution is not less than 0.01 mm.

4. The levee safety monitoring and maintenance system based on multi-sensor fusion according to claim 1 is characterized by: The vertical displacement sensor is a high-precision static level with an accuracy within ±0.05mm.

5. The levee safety monitoring and maintenance system based on multi-sensor fusion according to claim 1 is characterized by: The seepage monitoring sensor is an intelligent pressure measuring tube with a built-in microprocessor and a water flow sensor, which can sense changes in water flow speed and direction and make a preliminary judgment on the type of seepage anomaly.

6. The levee safety monitoring and maintenance system based on multi-sensor fusion according to claim 1 is characterized by: The data acquisition and transmission module adopts a hybrid communication network, which includes: The transmission rate of optical fiber communication as the backbone is not less than 10Gbps; As a supplementary ZigBee wireless sensor network, the operating frequency band is 2.4GHz and the transmission distance is 100 meters; As a supplementary LoRa wireless communication module, the operating frequency band is 433MHz / 868MHz / 915MHz, and the transmission distance is 15 kilometers; Ad hoc networking and adaptive routing algorithms are used to ensure reliable data transmission.

7. The levee safety monitoring and maintenance system based on multi-sensor fusion according to claim 1 is characterized by: The data collection center includes: High-performance multi-core processors and data processing servers; Real-time data processing software based on Linux operating system; The Hadoop distributed file system is used to store data, and multiple copies of data blocks are set; MySQL database, used to record data collection time, sensor location and equipment status.

8. A dike safety monitoring and maintenance method based on multi-sensor fusion, applied to a dike safety monitoring and maintenance system based on multi-sensor fusion as described in any one of claims 1 to 7, characterized in that: The following steps are involved: S1. Use the data acquisition and transmission module to obtain data collected by the water pressure sensor, horizontal displacement sensor, vertical displacement sensor and seepage monitoring sensor; S2, using data preprocessing algorithms to process the collected raw data, including filtering and normalization processing; S3, using anomaly detection and early warning algorithms to detect anomalies in the processed data and issue early warning information; S4. Use risk assessment algorithms to assess levee risks; S5. Develop and implement maintenance strategies based on the assessment results.

9. A dike safety monitoring and maintenance method based on multi-sensor fusion according to claim 8, characterized in that: In step S2, the data preprocessing algorithm includes: a joint filtering algorithm, which first uses a wavelet filtering algorithm to remove high-frequency noise, and then uses a Kalman filtering algorithm for state prediction and correction; a dynamic normalization algorithm, which normalizes data of different physical dimensions to the [0,1] interval according to the real-time change range; a data missing value processing algorithm, when data is missing, an interpolation algorithm based on support vector regression is used to fill in the missing values.

10. A levee safety monitoring and maintenance system and method based on multi-sensor fusion according to claim 8, characterized in that: In step S3, the anomaly detection and warning algorithm includes: performing cluster analysis in combination with the DBSCAN density clustering algorithm and the K-Means clustering algorithm to determine the distribution range of normal data; a composite warning model based on the long short-term memory network LSTM and the multi-layer perceptron MLP takes the preprocessed sensor data as input and outputs the safety status information of the levee.

Images