Slope real-time deformation monitoring system based on multi-sensor fusion

Abstract

The invention relates to the technical field of side slope monitoring, and particularly discloses a side slope real-time deformation monitoring system based on multi-sensor fusion, and the system comprises a data collection module which is disposed on a road side slope and is used for collecting multi-source monitoring data; the data transmission module is in communication connection with the data acquisition module and is used for remotely transmitting the multi-source monitoring data to a monitoring center; the data processing and early warning module is deployed in a monitoring center; a data fusion unit; a prediction and early warning unit; the system aims to solve the problem of large limitation of a single sensor monitoring means in a complex environment, the system realizes full-dimensional cooperative monitoring of a slope by fusing multi-source monitoring data of a foundation I nSAR, a fixed inclinometer array, a soil pressure sensor, a pore water pressure sensor, a rain gauge and the like, and the slope monitoring accuracy is improved through multi-sensor data fusion analysis. The error of performance reduction of a single sensor in a complex environment is effectively overcome, the comprehensive state of the slope is accurately obtained, and the reliability and precision of slope disaster early warning are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring technology, specifically to a real-time slope deformation monitoring system based on multi-sensor fusion. Background Technology

[0002] Traditional multi-sensor fusion often uses fixed weights or simple arithmetic averages, which cannot cope with the complex field environment in slope monitoring. For example, In-SAR accuracy drops sharply due to signal value issues when there is rain, snow, or severe multipath effects; inclinometers experience temperature drift when there are drastic temperature changes; and fixed weights can cause erroneous data to "contaminate" the fusion results when sensor performance deteriorates, leading to false alarms or missed alarms. To address these issues, we propose a real-time slope deformation monitoring system based on multi-sensor fusion. Summary of the Invention

[0003] The purpose of this invention is to provide a real-time slope deformation monitoring system based on multi-sensor fusion, in order to solve the problems mentioned in the background art, such as the sharp decrease in accuracy of In-SAR due to signal value issues when there is rain, snow, or severe multipath effects; the temperature drift of inclinometers when there are drastic temperature changes; and the fact that fixed weights can cause erroneous data to "contaminate" the fusion results when sensor performance deteriorates, leading to false alarms or missed alarms.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a real-time slope deformation monitoring system based on multi-sensor fusion, comprising a data acquisition module deployed on a highway slope for acquiring multi-source monitoring data. The data acquisition module includes: a ground-based InSAR module for monitoring slope surface displacement; a fixed inclinometer array for monitoring horizontal displacement at different depths within the slope; an earth pressure sensor for monitoring stress changes within the soil and rock mass; a pore water pressure sensor for monitoring pore water pressure changes; and a rain gauge for monitoring rainfall. A data transmission module is communicatively connected to the data acquisition module for remotely transmitting the multi-source monitoring data to a monitoring center. A data processing and early warning module is deployed at the monitoring center and includes: a data preprocessing unit for outlier removal and standardization of the received multi-source monitoring data; a data fusion unit for fusing the preprocessed multi-source monitoring data using an adaptive weighted fusion algorithm to generate comprehensive slope status information; and a prediction and early warning unit for calculating the future deformation trend of the slope based on the comprehensive slope status information using a time series prediction model and issuing an early warning when the deformation exceeds a preset threshold.

[0005] The placement and density of the ground-based InSAR module, fixed inclinometer, earth pressure sensor, and pore water pressure sensor are optimized based on engineering analogies and empirical judgment methods to identify potential slip surfaces and key deformation areas of highway slopes under rainfall conditions. The placement locations are focused on "convex slopes" and some soil-rock interfaces, and the placement density is divided into high-risk areas (sensor gap ≤ 30cm) and low-risk areas (sensor gap ≤ 50cm).

[0006] The data transmission module uses wireless transmission, and its network standard includes one or more combinations of LoRa or 4G / 5G, with satisfactory transmission quality.

[0007] The data fusion unit employs an adaptive weighted fusion algorithm, where the optimal weights are determined based on the inverse variance of the historical data from each sensor, in order to minimize the total mean square error after fusion.

[0008] The time series prediction model used by the prediction and early warning unit is the dynamic grey prediction model GM(1,1).

[0009] The detection steps include: Step S1: Based on engineering analogy and empirical judgment, analyze and determine the potential slip surface and key deformation area of ​​the highway slope;

[0010] Step S2: Based on the analysis results of Step S1, optimize the deployment of multiple types of sensors in the key deformation area. The sensors include at least those used to monitor surface displacement, deep displacement, stress, pore water pressure, and rainfall.

[0011] Step S3: Collect multi-source monitoring data from each sensor and transmit it remotely;

[0012] Step S4: Preprocess the received multi-source monitoring data, including outlier removal and data standardization;

[0013] Step S5: Use a data fusion algorithm to fuse the preprocessed multi-source monitoring data to obtain the comprehensive status information of the slope.

[0014] Step S6: Based on the fused data, use the prediction model to predict the slope deformation trend;

[0015] Step S7: Based on the comparison between the prediction results and the preset safety threshold, execute the graded warning.

[0016] In step S4, outlier removal during the preprocessing of the received multi-source monitoring data employs the 3σ criterion. The specific steps are as follows:

[0017] Step S11: Select the raw monitoring data within a time window from each sensor to form a data sequence: X = [x1, x2, x3, ..., xn ];

[0018] Step S12: Calculate the mean μ of the sequence. The mean reflects the central tendency of the data. The formula is:

[0019] Step S13: Calculate the standard deviation σ of the data sequence. The standard deviation reflects the dispersion of the data, and the formula is:

[0020] Step S14: For each data point x in the data sequence i If the following conditions are met, it is considered an outlier and removed: |x i -μ|>3σ.

[0021] In step S4, the standardization of the received multi-source monitoring data during preprocessing adopts the standard score method, and the specific steps are as follows:

[0022] Step S21: Select historical valid data after outliers have been removed to form the original data sequence: D i =[d i1 ,d i2 ,d i3 ,...,d in ];

[0023] Among them, D i It is the raw monitoring data sequence of the i-th type of sensor; d ij It is the j-th raw data from the i-th type of sensor; n is the data sample size;

[0024] Step S22: Calculate the mean of the original data sequence:

[0025] Step S23: Calculate the sample standard deviation of the original data sequence:

[0026] Step S24: Calculate the standard score:

[0027] in This refers to the standardized result of the j-th data from the i-th type of sensor;

[0028] d ij -μ i The deviation of the original data from the mean reflects the degree to which the data deviates from the center.

[0029] Divide by σ i This refers to eliminating the differences in the magnitude of the original data, so that data from different sensors are on the same distribution scale.

[0030] In step S5, a data fusion algorithm is used to fuse the preprocessed multi-source monitoring data. The specific steps are as follows:

[0031] Step S31: Define the object of calculation, the standardized sensor data sequence:

[0032]

[0033] Step S32: Because the slope monitoring data is a finite sample, it is necessary to calculate the sample variance:

[0034]

[0035] in, It is the sample variance of the standardized data of the i-th type of sensor;

[0036] This refers to the mean of the standardized data of the i-th type of sensor;

[0037] The smaller the variance, the more stable the data after standardization from the sensor.

[0038] Step S33: Calculate the fusion weights based on variance (and environmental parameters): where the data quality parameter C j Reflecting the stability of sensor data: (The smaller the variance, the higher C) j The closer to 1, the higher the credibility.

[0039] Environmental interference coefficient E j Reflecting the impact of the environment on the sensor; setting E j =min(0.8,0.1R), where R is the rainfall;

[0040] The initial weights are: Where, ω j (0) refers to the fusion weight of the j-th type of sensor in the initial state, which satisfies m refers to the total number of sensors participating in the fusion;

[0041] Step S34: Consistency optimization, obtaining the final weights:

[0042] Where, δ j For residuals, Standardized sensor data for the current moment. The mean of the standardized data from all sensors at the current moment;

[0043] This is a penalty term; the larger the residual, the smaller this value, and the greater the penalty imposed on the initial weights.

[0044] k is the adjustment coefficient, and m is the number of sensor types participating in the fusion;

[0045] Step S35: Fusion computation, Where v syn For the overall deformation rate, v j The deformation rate is the result of preprocessing using a single sensor.

[0046] This invention possesses at least the following beneficial effects: The real-time slope deformation monitoring system based on multi-sensor fusion efficiently adapts to the safety control needs of highway slopes (especially mountainous and rain-prone slopes prone to instability), highlighting its core advantages. Its monitoring dimensions are comprehensive, covering the entire range of slope surface-deep displacement, internal stress, and pore water pressure through the collaborative use of multiple devices such as ground-based InSAR, inclinometers, and earth pressure sensors. This avoids the limitations of single sensors and accurately captures changes in slope condition. It exhibits strong environmental adaptability; the sensors are waterproof and anti-interference capable, able to cope with complex field environments. The transmission method is flexible and supports breakpoint resumption, ensuring no data loss. The early warning system is scientifically reliable; multi-sensor collaborative verification reduces false alarms and missed alarms. Operation and maintenance are convenient; the system operates automatically without complex manual intervention, and its modular design adapts to slopes of different scales and risk levels, reducing manpower and maintenance costs and providing full-process support for slope safety control. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the overall architecture of the present invention;

[0048] Figure 2 This is a flowchart of the monitoring steps of the present invention;

[0049] Figure 3 This is a flowchart of the data acquisition and early warning process of this invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please see Figure 1-3This invention provides a technical solution: a real-time slope deformation monitoring system based on multi-sensor fusion, comprising a data acquisition module deployed on a highway slope for collecting multi-source monitoring data. The data acquisition module includes: a ground-based InSAR module for monitoring slope surface displacement; a fixed inclinometer array for monitoring horizontal displacement at different depths within the slope; an earth pressure sensor for monitoring stress changes within the soil and rock mass; a pore water pressure sensor for monitoring pore water pressure changes; and a rain gauge for monitoring rainfall. A data transmission module, communicatively connected to the data acquisition module, is used to remotely transmit the multi-source monitoring data to a monitoring center. A data processing and early warning module, deployed at the monitoring center, includes: a data preprocessing unit for outlier removal and standardization of the received multi-source monitoring data; a data fusion unit for fusing the preprocessed multi-source monitoring data using an adaptive weighted fusion algorithm to generate comprehensive slope status information; and a prediction and early warning unit for calculating the slope's deformation trend over future periods based on the comprehensive slope status information using a time-series prediction model, and issuing an early warning when the deformation exceeds a preset threshold.

[0052] The placement and density of the ground-based InSAR module, fixed inclinometer, earth pressure sensor, and pore water pressure sensor are optimized based on engineering analogies and empirical judgment methods to identify potential slip surfaces and key deformation areas of highway slopes under rainfall conditions. The placement locations are focused on "convex slopes" and some soil-rock interfaces, and the placement density is divided into high-risk areas (sensor gap ≤ 30cm) and low-risk areas (sensor gap ≤ 50cm).

[0053] The data transmission module uses wireless transmission, and its network standard includes one or more combinations of LoRa or 4G / 5G, with satisfactory transmission quality.

[0054] The data fusion unit uses an adaptive weighted fusion algorithm, and the optimal weights are determined based on the inverse of the variance of the historical data of each sensor, so as to minimize the total mean square error after fusion.

[0055] The time series prediction model used by the prediction and early warning unit is the dynamic grey prediction model GM(1,1).

[0056] The detection steps include: Step S1: Based on engineering analogy and empirical judgment, analyze and determine the potential slip surface and key deformation area of ​​the highway slope;

[0057] Step S2: Based on the analysis results of Step S1, optimize the deployment of multiple types of sensors in the key deformation area. The sensors include at least those used to monitor surface displacement, deep displacement, stress, pore water pressure, and rainfall.

[0058] Step S3: Collect multi-source monitoring data from each sensor and transmit it remotely;

[0059] Step S4: Preprocess the received multi-source monitoring data, including outlier removal and data standardization;

[0060] Step S5: Use a data fusion algorithm to fuse the preprocessed multi-source monitoring data to obtain the comprehensive status information of the slope.

[0061] Step S6: Based on the fused data, use the prediction model to predict the slope deformation trend;

[0062] Step S7: Based on the comparison between the prediction results and the preset safety threshold, execute the graded warning.

[0063] In step S4, outlier removal during the preprocessing of the received multi-source monitoring data employs the 3σ criterion. The specific steps are as follows:

[0064] Step S11: Select the raw monitoring data within a time window from each sensor to form a data sequence: X = [x1, x2, x3, ..., x n ];

[0065] Step S12: Calculate the mean μ of the sequence. The mean reflects the central tendency of the data. The formula is:

[0066] Step S13: Calculate the standard deviation σ of the data sequence. The standard deviation reflects the dispersion of the data, and the formula is:

[0067] Step S14: For each data point x in the data sequence i If the following conditions are met, it is considered an outlier and removed: |x i -μ|>3σ.

[0068] In step S4, the standardization of the received multi-source monitoring data during preprocessing is performed using the standard score method. The specific steps are as follows:

[0069] Step S21: Select historical valid data after outliers have been removed to form the original data sequence: D i =[d i1 ,d i2 ,d i3 ,...,d in ];

[0070] Among them, D i It is the raw monitoring data sequence of the i-th type of sensor; d ij It is the j-th raw data from the i-th type of sensor; n is the data sample size;

[0071] Step S22: Calculate the mean of the original data sequence:

[0072] Step S23: Calculate the sample standard deviation of the original data sequence:

[0073] Step S24: Calculate the standard score:

[0074] in This refers to the standardized result of the j-th data from the i-th type of sensor;

[0075] d ij -μ i The deviation of the original data from the mean reflects the degree to which the data deviates from the center.

[0076] Divide by σ i This refers to eliminating the differences in the magnitude of the original data, so that data from different sensors are on the same distribution scale.

[0077] In step S5, a data fusion algorithm is used to fuse the preprocessed multi-source monitoring data. The specific steps are as follows:

[0078] Step S31: Define the object of calculation, the standardized sensor data sequence:

[0079]

[0080] Step S32: Because the slope monitoring data is a finite sample, it is necessary to calculate the sample variance:

[0081]

[0082] in, It is the sample variance of the standardized data of the i-th type of sensor;

[0083] This refers to the mean of the standardized data of the i-th type of sensor;

[0084] The smaller the variance, the more stable the data after standardization from the sensor.

[0085] Step S33: Calculate the fusion weights based on variance (and environmental parameters): where the data quality parameter C j Reflecting the stability of sensor data: (The smaller the variance, the higher C) j The closer to 1, the higher the credibility.

[0086] Environmental interference coefficient E j Reflecting the impact of the environment on the sensor; setting E j =min(0.8,0.1R), where R is the rainfall;

[0087] The initial weights are: Where, ω j (0) refers to the fusion weight of the j-th type of sensor in the initial state, which satisfies m refers to the total number of sensors participating in the fusion;

[0088] Step S34: Consistency optimization, obtaining the final weights:

[0089] Where, δ j For residuals, Standardized sensor data for the current moment. The mean of the standardized data from all sensors at the current moment;

[0090] This is a penalty term; the larger the residual, the smaller this value, and the greater the penalty imposed on the initial weights.

[0091] k is the adjustment coefficient, and m is the number of sensor types participating in the fusion;

[0092] Step S35: Fusion computation, Where v syn For the overall deformation rate, v j The deformation rate is the result of preprocessing using a single sensor.

[0093] This system is suitable for long-term dynamic monitoring and risk early warning of highway slopes. It adopts a three-tier architecture model of on-site data collection, remote data transmission, and processing at the monitoring center. The specific implementation details are as follows:

[0094] 1. Sensor Selection and Functional Adaptation: The ground-based InSAR module uses a ground-based synthetic aperture radar (SAR) device with all-weather operation capability. During installation, the radar unit must be mounted on a stable area opposite the slope, ensuring the antenna faces the slope without obstruction (avoiding trees and buildings). A solar power supply system (including energy storage batteries) must also be installed to ensure continuous power supply. The fixed inclinometer array uses a high-precision sliding inclinometer with an IP68 waterproof rating, allowing for long-term underground monitoring of horizontal displacement at different depths within the slope. The inclinometer requires a PVC inclinometer tube with guide grooves on its inner wall to ensure the sensor slides along a fixed direction to collect data. The earth pressure sensor uses a vibrating wire earth pressure sensor with a corrosion-resistant stainless steel housing, allowing for direct burial. Inside the soil and rock mass, the sensor must be in close contact with the soil during installation to avoid measurement errors caused by gaps. Fine sand should be filled around the sensor to ensure uniform stress transmission. For the pore water pressure sensor, a differential pressure type with a measurement range of 0-200 kPa should be selected. It can monitor the pressure changes of rainwater in the soil pores in real time. The preferred installation locations are at the soil-rock interface and the permeable layer below cracks at the top of the slope to ensure the capture of the water pressure accumulation process after rainwater infiltration. A tipping bucket rain gauge with a measurement accuracy of 0.1 mm should be used. It can automatically record rainfall and rainfall intensity, and the data sampling frequency is synchronized with other sensors. During installation, the rain gauge should be fixed in a flat, unobstructed area at the top of the slope (such as the top of an observation pier) to prevent water accumulation and fallen leaves from obstructing the tipping bucket and affecting measurement accuracy.

[0095] 2. Specific Implementation of the Data Transmission Module: First, the system adopts wireless transmission. Depending on the network coverage of the slope area, one or more combinations of LoRa, 4G / 5G can be flexibly selected: LoRa Transmission: Suitable for mountainous areas without 4G / 5G signals. LoRa technology features long distance (5-10km per hop), low power consumption (sensor node battery life up to 6-12 months), and strong anti-interference capabilities. Data aggregation from multiple sensors can be achieved by deploying LoRa gateways, which are then connected to the monitoring center via wired connections (such as fiber optics), solving the transmission problems in remote areas. 4G / 5G Transmission: Suitable for areas along highways with operator network coverage. Sensor data is directly uploaded to the cloud server or local monitoring center via 4G / 5G modules, with a transmission latency of ≤10s, meeting real-time monitoring requirements. To ensure transmission stability, signal repeaters (such as 4G micro base stations) can be added in areas with weak signals to avoid signal attenuation caused by rainfall or terrain obstruction. Secondly, during data transmission, data encryption and breakpoint resumption methods can be used to prevent data loss and errors during transmission.

[0096] 3. Specific implementation of the data processing and early warning module: (1) Data preprocessing is carried out. First, outlier removal is performed. For the original data of each type of sensor, effective data of a certain time window is selected to form a data sequence. Through statistical analysis, abnormal data that deviates from the normal range (such as jump values ​​caused by sensor failure, sudden values ​​caused by instantaneous impact of rainfall) are identified and removed to avoid affecting the subsequent fusion results. For example, in the deep displacement data collected by a certain inclinometer, there is a "20mm displacement value" that is far beyond the normal range. After being judged as an outlier by the 3σ criterion, it is removed to ensure the rationality of the data sequence. Then, data standardization is performed. Since the data dimensions of different sensors are very different (such as displacement unit is mm, stress unit is kPa), all data need to be converted into dimensionless standard fractions. In specific operations, first calculate the mean and dispersion of the historical valid data of each type of sensor, and then convert the original data into a standard value of a uniform scale according to the degree of deviation of the data from the mean to eliminate the influence of the dimension; (2) The implementation of the data fusion unit, the weight allocation is mainly based on two factors - sensor data stability and environmental interference. Among them, data stability is judged by the dispersion of the historical data of the sensor (the smaller the dispersion, the more stable the data, and the higher the weight); the environmental interference mainly refers to the rain gauge data (the greater the rainfall, the stronger the interference to the ground-based InSAR and other surface sensors, and its weight needs to be reduced). First calculate the "data quality parameter" according to the stability of each type of sensor (the parameter of stable sensors is close to 1, and the parameter of unstable sensors is close to 0), and then calculate the "environmental interference coefficient" according to the rainfall. The greater the rainfall, the greater the interference coefficient, and the weaker the sensor's anti-interference capability. Combining these two parameters, the initial weight of each type of sensor is calculated to ensure that the initial weight reflects the sensor's "intrinsic quality" and "external anti-interference capability". After the initial weight is calculated, it needs to be further corrected according to "multi-source data consistency". If the data of a certain sensor deviates too much from the data of most other sensors (such as the ground-based InSAR data deviating from the inclinometer data due to rainstorm interference), the weight of that sensor is reduced to avoid its erroneous data dominating the fusion result. Conversely, if the data of a certain sensor is highly consistent with other data, its weight is appropriately increased to enhance the reliability of the fusion result. The corrected final weight is multiplied by the standardized data of each sensor and then summed to obtain the comprehensive state information of the slope, providing basic data for subsequent trend prediction.(3) Implementation of the prediction and early warning unit: The GM(1,1) model is a time series model, which is particularly suitable for slope monitoring scenarios where data accumulation is limited and deformation trends are relatively stable. The specific steps are as follows: The input data is the "comprehensive deformation rate time series data" generated by the data fusion unit. Usually, the comprehensive deformation rate data of the most recent 4-10 days is selected to ensure that the input data is valid data after preprocessing. Before making a prediction, the applicability of the GM(1,1) model needs to be verified first - the "development coefficient" (reflecting the stability of the deformation trend) is obtained through model calculation. If the absolute value of the development coefficient is less than 0.3, it means that the model has a high degree of fit to the current slope deformation and the prediction result is reliable; if the absolute value of the development coefficient is greater than or equal to 0.3, it means that the current deformation trend fluctuates too much and 1-2 more are needed. After re-modeling based on the latest data, the system ensures prediction accuracy. Once the model is validated, the "prediction duration" is input, and the model automatically outputs the daily comprehensive deformation rate and cumulative displacement for the future prediction duration, providing clear trend data for subsequent early warnings. To avoid deviations in prediction results due to new data, the system has a "daily update" mechanism. Every morning, the system automatically acquires the latest comprehensive deformation rate data from the previous day, retrains the GM(1,1) model, and updates the future prediction results, ensuring that the prediction results are always synchronized with the latest deformation state of the slope. The graded early warning system, based on the prediction results of the GM(1,1) model, combined with preset safety thresholds and multi-sensor collaborative verification, classifies slope risks into four levels (blue, yellow, orange, and red), with each level of early warning corresponding to clear judgment criteria and handling measures.

[0097] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0098] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time slope deformation monitoring system based on multi-sensor fusion, characterized in that, include: A data acquisition module, deployed on the highway slope, is used to collect multi-source monitoring data. The data acquisition module includes: Ground-based InSAR module for monitoring slope surface displacement; A fixed inclinometer array used to monitor horizontal displacement at different depths inside a slope. Earth pressure sensor used to monitor stress changes within soil and rock masses; A pore water pressure sensor used to monitor changes in pore water pressure; And rain gauges used to monitor rainfall; The data transmission module is communicatively connected to the data acquisition module and is used to remotely transmit the multi-source monitoring data to the monitoring center. The data processing and early warning module, deployed in the monitoring center, includes: The data preprocessing unit is used to remove outliers and standardize the received multi-source monitoring data; The data fusion unit is used to fuse the preprocessed multi-source monitoring data using an adaptive weighted fusion algorithm to generate comprehensive slope status information; The prediction and early warning unit is used to calculate the deformation trend of the slope in future periods based on the comprehensive slope status information and through a time series prediction model, and to issue an early warning message when the deformation exceeds a preset threshold.

2. The slope real-time deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: The placement and density of the ground-based InSAR module, fixed inclinometer, soil pressure sensor, and pore water pressure sensor are optimized based on engineering analogies and empirical judgment methods to identify potential slip surfaces and key deformation areas of the highway slope under rainfall conditions. The placement locations are focused on "convex slopes" and some soil-rock interfaces, and the placement density is divided into high-risk areas (sensor gap ≤ 30cm) and low-risk areas (sensor gap ≤ 50cm).

3. The slope real-time deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: The data transmission module adopts a wireless transmission method, and its network standard includes one or more combinations of LoRa or 4G / 5G, and its transmission quality is satisfactory.

4. The slope real-time deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: The data fusion unit employs an adaptive weighted fusion algorithm, where the optimal weights are determined based on the reciprocal of the variance of historical data from each sensor, in order to minimize the total mean square error after fusion.

5. A real-time slope deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: The time series prediction model used by the prediction and early warning unit is the dynamic grey prediction model GM(1,1).

6. The slope real-time deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: The following testing steps are included: Step S1: Based on engineering analogy and empirical judgment, analyze and determine the potential slip surface and key deformation area of ​​the highway slope; Step S2: Based on the analysis results of step S1, optimize the deployment of multiple types of sensors in the key deformation area. The sensors include at least sensors for monitoring surface displacement, deep displacement, stress, pore water pressure and rainfall. Step S3: Collect multi-source monitoring data from each sensor and transmit it remotely; Step S4: Preprocess the received multi-source monitoring data, including outlier removal and data standardization; Step S5: Use a data fusion algorithm to fuse the preprocessed multi-source monitoring data to obtain the comprehensive status information of the slope. Step S6: Based on the fused data, use the prediction model to predict the slope deformation trend; Step S7: Based on the comparison between the prediction results and the preset safety threshold, execute the graded warning.

7. A real-time slope deformation monitoring system based on multi-sensor fusion according to claim 6, characterized in that: In step S4, outlier removal during the preprocessing of the received multi-source monitoring data employs the 3σ criterion. The specific steps are as follows: Step S11: Select the raw monitoring data within a time window from each sensor to form a data sequence: X = [x1, x2, x3, ..., x n ]; Step S12: Calculate the mean μ of the sequence. The mean reflects the central tendency of the data. The formula is: Step S13: Calculate the standard deviation σ of the data sequence. The standard deviation reflects the dispersion of the data, and the formula is: Step S14: For each data point x in the data sequence i If the following conditions are met, it is considered an outlier and removed: |x i -μ|>3σ.

8. The slope real-time deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: In step S4, the standardization of the received multi-source monitoring data preprocessing adopts the standard score method, and the specific steps are as follows: Step S21: Select historical valid data after outliers have been removed to form the original data sequence: D i =[d i1 ,d i2 ,d i3 ,...,d in ]; Among them, D i It is the raw monitoring data sequence of the i-th type of sensor; d ij It is the j-th raw data from the i-th type of sensor; n is the data sample size; Step S22: Calculate the mean of the original data sequence: Step S23: Calculate the sample standard deviation of the original data sequence: Step S24: Calculate the standard score: in This refers to the standardized result of the j-th data from the i-th type of sensor; d ij -μ i The deviation of the original data from the mean reflects the degree to which the data deviates from the center. Divide by σ i This refers to eliminating the differences in the magnitude of the original data, so that data from different sensors are on the same distribution scale.

9. A real-time slope deformation monitoring system based on multi-sensor fusion according to claim 1, characterized in that: In step S5, a data fusion algorithm is used to fuse the preprocessed multi-source monitoring data. The specific steps are as follows: Step S31: Define the object of calculation, the standardized sensor data sequence: Step S32: Because the slope monitoring data is a finite sample, it is necessary to calculate the sample variance: in, It is the sample variance of the standardized data of the i-th type of sensor; This refers to the mean of the standardized data of the i-th type of sensor; The smaller the variance, the more stable the data after standardization from the sensor. Step S33: Calculate the fusion weights based on variance (and environmental parameters): where the data quality parameter C j Reflecting the stability of sensor data: (The smaller the variance, the higher C) j The closer to 1, the higher the credibility. Environmental interference coefficient E j Reflecting the impact of the environment on the sensor; setting E j =min(0.8,0.1R), where R is the rainfall; The initial weights are: Where, ω j (0) refers to the fusion weight of the j-th type of sensor in the initial state, which satisfies m refers to the total number of sensors participating in the fusion; Step S34: Consistency optimization, obtaining the final weights: Where, δ j For residuals, Standardized sensor data for the current moment. The mean of the standardized data from all sensors at the current moment; This is a penalty term; the larger the residual, the smaller this value, and the greater the penalty imposed on the initial weights. k is the adjustment coefficient, and m is the number of sensor types participating in the fusion; Step S35: Fusion computation, Where v syn For the overall deformation rate, v j The deformation rate is the value after preprocessing by a single sensor.

Images