Highway pavement settlement monitoring and predicting method based on multi-source data

Abstract

The invention discloses a highway pavement settlement monitoring and predicting method based on multi-source data, relates to the technical field of traffic safety, and solves the problems of long monitoring period, high cost and incapability of real-time dynamic monitoring in the prior art. The settlement problem of the road cannot be found in time, and the road surface damage speed is increased. The method comprises the steps of obtaining pavement settlement signals of a road; performing feature extraction on the pavement settlement signal to obtain feature information of pavement settlement; comparing the settlement threshold value with the feature information, and determining a settlement section of the marked road; obtaining multi-source heterogeneous data; fusing the multi-source heterogeneous data to obtain a multi-source time sequence matrix; predicting the settlement road section according to the multi-source time sequence matrix to obtain prediction information; the road surface settlement information can be detected in time, certain measures are taken, the safety of road surface driving is guaranteed, and the road surface damage speed is reduced.

Description

Technical Field

[0001] The present invention belongs to the field of traffic safety, relates to a highway pavement settlement monitoring and prediction technology, and specifically is a highway pavement settlement monitoring and prediction method based on multi-source data. Background Art

[0002] Highways are an important part of the modern transportation network. Highway pavement settlement refers to the local or overall sinking of the road surface, which is one of the common diseases in highway engineering. It not only affects driving comfort and safety, but also accelerates road damage and increases maintenance costs.

[0003] Traditional methods of monitoring road subsidence mainly rely on leveling, GPS and other technologies. Although traditional leveling and GPS technologies have high accuracy, they require regular manual operation and cannot reflect subsidence changes in real time. In addition, they have disadvantages such as long monitoring cycles, high costs, and the inability to conduct real-time dynamic monitoring. Therefore, subsidence problems on highways cannot be discovered in time, causing the speed of road damage to accelerate. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a monitoring and prediction method for highway pavement settlement based on multi-source data, which is used to solve the technical problems of the prior art such as long monitoring cycle, high cost, inability to conduct real-time dynamic monitoring, inability to timely detect settlement problems occurring on highways, resulting in accelerated road damage.

[0005] To achieve the above objectives, a first aspect of the present invention provides a method for monitoring and predicting highway pavement settlement based on multi-source data, comprising:

[0006] Obtain road surface settlement signals;

[0007] Extract features from road subsidence signals to obtain characteristic information of road subsidence;

[0008] Compare the subsidence threshold and feature information to identify the subsidence sections of the marked highway;

[0009] Acquire multi-source heterogeneous data; fuse the multi-source heterogeneous data to obtain a multi-source time series matrix;

[0010] The subsidence section is predicted based on the multi-source time series matrix to obtain prediction information.

[0011] Preferably, the road surface settlement signal collection method includes:

[0012] Divide the highway into several sections of roads to be tested based on their length; obtain risk scores of the several roads to be tested, and determine the step length of setting the measuring points according to the risk scores;

[0013] The acquisition station is set on the road to be measured according to the setting step of the measuring point; the multi-physical field sensor of the acquisition station is used to collect the road surface settlement signal on the road according to the preset collection frequency; wherein, the road surface settlement signal is the integration of the time series signals collected by each sensor.

[0014] It should be noted that the sensor is a multi-physics field sensor, consisting of a differential sensor, a three-component magnetic sensor and an electrode; the risk score is the prediction result of the previous round of monitoring; the survey line is laid along the direction of the highway; each collection station should include a data acquisition module that can collect sensor data in real time. The collection station must have protective functions such as waterproof, moisture-proof and shockproof.

[0015] The present invention divides a highway into several sections to be tested, sets the spacing between measuring points according to the risk scores of the sections to be tested; sets collection stations on the highway to be tested according to the distances between the measuring points, and collects data using sensors at a set collection frequency; the number of collection stations can be set according to the risk scores, and intensive collection is performed on sections with high risks, which is conducive to timely detection of road subsidence and taking corresponding measures to reduce the rate of road damage.

[0016] Preferably, the feature extraction based on the road surface settlement signal includes:

[0017] Preprocessing the road subsidence signal; wherein the preprocessing includes: missing value filling, denoising and removing outliers;

[0018] The pre-processed road surface settlement signal is converted into digital form to obtain a settlement digital signal; the settlement digital signal is processed using a signal processing algorithm to obtain characteristic information; wherein the signal processing algorithm includes: Fourier transform or wavelet transform.

[0019] Preferably, the marking of the subsided road sections according to the subsidence threshold and characteristic information includes:

[0020] comparing the feature information with a sedimentation threshold;

[0021] When the characteristic information is greater than the settlement threshold, the corresponding highway will be marked to obtain the settlement section and an abnormal signal will be generated; otherwise, the road surface settlement signal of the highway will be continuously collected and the abnormal signal will be sent to the corresponding technician.

[0022] The present invention pre-processes the subsidence signal to ensure the integrity and quality of the subsidence signal, and performs format conversion on the subsidence signal to unify the format of the signal, which is convenient for subsequent analysis and processing; utilizes Fourier transform or wavelet transform to extract features of the subsidence digital signal, and obtains the subsidence section based on the feature information and the subsidence threshold analysis, and generates an abnormal signal to notify the corresponding technical personnel; can respond to the subsidence section in time and process the subsidence section in time.

[0023] Preferably, fusing multi-source heterogeneous data to obtain a multi-source time series matrix includes:

[0024] Retrieve multi-source heterogeneous data; where multi-source heterogeneous data includes: micro-motion sensor data, electromagnetic sensor data, and natural potential data;

[0025] The features of multi-source heterogeneous data are extracted separately to obtain multi-source feature data; the multi-source feature data are fused using feature-level fusion and spatiotemporal alignment mechanisms to obtain a multi-source time series matrix.

[0026] The present invention analyzes the characteristics of multi-source heterogeneous data respectively to obtain multi-source characteristic data; and fuses the multi-source characteristic data to obtain a multi-source time series matrix; it can obtain dynamic information of road surface settlement from different angles; the micro-motion sensor uses natural source surface wave information to efficiently detect underground loose soil; it does not require an artificial seismic source, is safe and environmentally friendly; the electromagnetic sensor can quickly detect the internal structure of the roadbed and the location of the cavity; the natural potential sensor can monitor the groundwater flow and pore water pressure changes in real time, reflecting the impact of fluid migration on the roadbed stability; the fusion of the three sensor data provides more comprehensive monitoring data and improves the reliability of settlement warning.

[0027] Preferably, extracting features of multi-source heterogeneous data to obtain multi-source feature data includes:

[0028] Retrieving micro-motion sensor data and electromagnetic sensor data from multi-source heterogeneous data; extracting time domain features, frequency domain features, and spatial features of the micro-motion sensor data and electromagnetic sensor data respectively;

[0029] Retrieve the natural potential data from multi-source heterogeneous data; extract the time domain features and spatial features of the natural potential data;

[0030] The time domain features, frequency domain features and spatial features of the extracted micro-motion sensor data and electromagnetic sensor data, as well as the time domain features and spatial features of the extracted natural potential data are integrated into multi-source feature data.

[0031] Preferably, the fusing of multi-source feature data using feature-level fusion and spatiotemporal alignment mechanisms includes:

[0032] Retrieve multi-source feature data; unify the timestamps of sub-data in multi-source feature data through GPS / NTP protocol; map sensor coordinates to the UTM grid coordinate system and construct a local spatial grid;

[0033] The time domain, frequency domain, and spatial features of each sensor are aligned according to the time window to form a multi-dimensional input matrix; each feature dimension is separately standardized to obtain a multi-source time series matrix.

[0034] Preferably, the predicting of the subsidence section according to the multi-source time series matrix includes:

[0035] Retrieve multi-source time series matrices and pre-trained prediction models;

[0036] Input the multi-source time series matrix into the prediction model to obtain prediction information; wherein the prediction information includes: prediction value, risk score and early warning signal;

[0037] Obtain a preset threshold; generate a visualization interface based on the prediction information and the preset threshold, and display the visualization interface to the corresponding technical personnel.

[0038] The present invention inputs a multi-source time series matrix into a pre-trained prediction model to obtain prediction information; and generates a visualization interface based on preset thresholds and prediction information, and displays the visualization interface to corresponding technical personnel; it not only improves the efficiency and accuracy of highway settlement monitoring, but also significantly improves the safety of highway operation and maintenance efficiency through scientific early warning and timely response.

[0039] Preferably, the method for obtaining the preset threshold includes:

[0040] Obtain historical characteristic information of historical road surface settlement, sort the historical characteristic information in chronological order, and divide the historical characteristic information into several groups according to a set period;

[0041] Calculate the baseline threshold for each group: p threshold =μ+2σ; where μ is the mean value of each set of feature information; σ is the variance of each set of feature information;

[0042] An environmental correction factor is obtained, and the threshold is adjusted by calculating the product of the environmental correction factor and the baseline threshold to obtain a preset threshold.

[0043] A second aspect of the present invention provides an electronic device, comprising: a communication unit and a processing unit;

[0044] A communication unit, used to obtain road surface settlement signals of the highway;

[0045] The processing unit is used to extract features from road subsidence signals to obtain characteristic information of road subsidence; compare subsidence thresholds with characteristic information to determine subsidence sections of marked highways; obtain multi-source heterogeneous data; fuse the multi-source heterogeneous data to obtain a multi-source time series matrix; and predict the subsidence sections based on the multi-source time series matrix to obtain prediction information.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. The present invention divides a highway into several sections of highway to be tested, and sets the spacing between measuring points according to the risk scores of the several sections of highway to be tested; sets collection stations on the highway to be tested according to the distance between the measuring points, and collects data according to the set collection frequency; can set the number of collection stations according to the risk score, and conduct intensive collection on sections with high risks, which is conducive to timely discovering the settlement of the road surface and taking corresponding measures to reduce the speed of road damage; pre-processes the settlement signal to ensure the integrity and quality of the settlement signal, and converts the format of the settlement signal to unify the format of the signal, which is convenient for subsequent analysis and processing; uses Fourier transform or wavelet transform to extract features of the settlement digital signal, and obtains the settlement section according to the feature information and the settlement threshold analysis, and generates an abnormal signal to notify the corresponding technical personnel; can respond to the settlement section in time and process the settlement section in time.

[0048] 2. The present invention analyzes the characteristics of multi-source heterogeneous data respectively to obtain multi-source characteristic data; and fuses the multi-source characteristic data to obtain a multi-source time series matrix; it can obtain dynamic information of road surface settlement from different angles; the micro-motion sensor uses natural source surface wave information to efficiently detect underground loose soil; it does not require an artificial seismic source, is safe and environmentally friendly; the electromagnetic sensor can quickly detect the internal structure of the roadbed and the location of the cavity; the natural potential sensor can monitor the flow of groundwater and the change of pore water pressure in real time, reflecting the impact of fluid migration on the stability of the roadbed; the three sensor data are integrated to provide more comprehensive monitoring data and improve the reliability of settlement warning; the multi-source time series matrix is ​​input into a pre-trained prediction model to obtain prediction information; and a visualization interface is generated according to the preset threshold and the prediction information, and the visualization interface is displayed to the corresponding technical personnel; it not only improves the efficiency and accuracy of highway settlement monitoring, but also significantly improves the safety and maintenance efficiency of highway operation through scientific warning and timely response. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 A schematic diagram of the overall method steps of the present invention;

[0051] Figure 2 Schematic diagram of the highway pavement settlement monitoring steps of the present invention;

[0052] Figure 3 Schematic diagram of the sedimentation signal acquisition architecture of the present invention;

[0053] Figure 4 Schematic diagram of the highway pavement settlement prediction steps of the present invention;

[0054] Figure 5 Schematic diagram of the electronic device of the present invention. DETAILED DESCRIPTION

[0055] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] See also Figure 1 The first embodiment of the present invention provides a method for monitoring and predicting highway pavement settlement based on multi-source data, comprising:

[0057] Obtain road surface settlement signals;

[0058] Extract features from road subsidence signals to obtain characteristic information of road subsidence;

[0059] Compare the subsidence threshold and feature information to identify the subsidence sections of the marked highway;

[0060] Acquire multi-source heterogeneous data; fuse the multi-source heterogeneous data to obtain a multi-source time series matrix;

[0061] The subsidence section is predicted based on the multi-source time series matrix to obtain prediction information.

[0062] See also Figure 2-3 , dividing the highway into several sections to be tested based on its length; obtaining risk scores for the several sections to be tested, and determining the setting step length of the measuring points according to the risk scores; setting acquisition stations on the sections to be tested according to the setting step length of the measuring points; using the multi-physics field sensors of the acquisition stations to collect road surface settlement signals on the highway according to a preset collection frequency; wherein the road surface settlement signal is the integration of the time series signals collected by each sensor.

[0063] In a specific embodiment, the data acquisition system consists of a central control console, a signal transmission system, a ground signal three-component electromagnetic sensor, a linear micro-motion array, electrodes, a ground data acquisition station, and a signal transmission line.

[0064] a.Electromagnetic sensor monitoring

[0065] Layout of measuring lines and measuring points: The measuring lines should be laid out along the direction of the highway to monitor the settlement changes of the entire road section; the spacing between measuring points should be designed according to the monitoring task, and the density of measuring points should be adjusted according to the geological conditions and settlement risks of the monitoring area. In areas with higher settlement risks, the density of measuring points should be appropriately increased; measuring points should be laid out inside the roadbed or under the shoulder to ensure that the sensors can accurately reflect the roadbed settlement.

[0066] Collection station and sensor layout: Each collection station should include a data acquisition module that can collect sensor data in real time; the collection station must have protective functions such as waterproof, moisture-proof, and shock-proof to adapt to the complex environment along the highway.

[0067] Control system layout: The central control console is arranged in the road command and dispatch control center or the ground survey department, and is manned by a dedicated person to issue monitoring commands.

[0068] Monitoring time: Determine the monitoring period based on project requirements and monitoring objectives; dynamically adjust the monitoring time based on real-time monitoring data and environmental changes; in areas with complex geological conditions or high subsidence risks, the monitoring frequency should be appropriately increased.

[0069] b. Micro-motion monitoring

[0070] Layout of measuring points and lines: Select a linear array, and the measuring lines should be laid out as close to the direction of the highway as possible to better monitor geological changes along the highway; increase the density of measuring lines in areas with complex geological conditions or high potential risks (such as landslides, collapses, and mined-out areas).

[0071] Collection station and sensor layout: Collection stations should be located in areas with flat and stable terrain; avoid being close to interference sources such as large machinery and high-voltage lines to ensure data stability and reliability; the equipment should have the function of on-site evaluation of the quality of collected data to ensure data quality.

[0072] Control system layout: The control system should be deployed in an area that is geologically stable and easy to maintain, and should be manually inspected regularly to ensure real-time continuous monitoring.

[0073] Monitoring time: Dynamically adjust monitoring time based on real-time monitoring data and environmental changes; when potential subsidence risks are discovered, increase monitoring frequency and duration to obtain more detailed data.

[0074] c. Natural potential monitoring

[0075] Electrode layout: The electrodes should be buried in moist soil as much as possible to avoid interference from debris such as gravel and grass. If the measuring point is located in a high-resistance medium such as rock or concrete ground, it needs to be grounded with moist soil. The electrode leads should be tightly connected to the measuring wires, and the connection points and exposed metal parts of the electrodes should be strictly prevented from contacting the ground or weeds to avoid electromagnetic interference.

[0076] Survey line layout: Survey lines are laid out along the direction of the highway to facilitate better real-time monitoring of ground subsidence.

[0077] Control system layout: When laying cables and installing equipment, effective anti-interference measures should be taken, such as using shielded cables and proper grounding.

[0078] Monitoring time: Adjust according to real-time data and external environment; during the monitoring time, analyze the collected data in real time, detect abnormal situations in time and adjust the monitoring strategy; reasonably allocate monitoring resources to avoid excessive monitoring during non-critical periods and improve monitoring efficiency.

[0079] Preprocessing the road surface settlement signal; wherein the preprocessing includes: filling missing values, denoising and removing outliers; performing analog-to-digital conversion on the preprocessed road surface settlement signal to obtain a settlement digital signal; processing the settlement digital signal using a signal processing algorithm to obtain characteristic information; wherein the signal processing algorithm includes: Fourier transform or wavelet transform.

[0080] Compare the characteristic information with the settlement threshold; when the characteristic information is greater than the settlement threshold, mark the corresponding highway to obtain the settlement section and generate an abnormal signal; otherwise, continue to collect the road surface settlement signal of the highway; and send the abnormal signal to the corresponding technician.

[0081] In a specific embodiment, data processing and transmission includes: a signal receiving unit, a signal processing unit, a data encoding and compression unit, a communication transmission unit, and a status monitoring and fault alarm unit.

[0082] (1) Signal receiving unit:

[0083] Function description: This unit is responsible for receiving the road subsidence signal collected by the ground subsidence signal monitoring part (such as the sensor array).

[0084] Technical Implementation: A high-precision analog-to-digital converter (ADC) is used to convert analog signals into digital signals for subsequent digital signal processing. At the same time, a signal amplification circuit is designed to enhance the ability to identify weak signals and ensure the accuracy of signal acquisition.

[0085] (2) Signal processing unit:

[0086] Function description: Filter, extract features, detect anomalies, and perform other processing on the received digital signal to identify the characteristic information of road subsidence.

[0087] Technical implementation: Filtering processing: Apply digital filters to remove noise interference and retain effective signal components.

[0088] Feature extraction: Use algorithms (such as fast Fourier transform FFT, wavelet transform, etc.) to extract the time domain and frequency domain features of the signal. These features can reflect information such as the type and thickness of the sediment.

[0089] Anomaly detection: Automatically identifies abnormal subsidence signals based on preset thresholds or machine learning models, marking road sections that may have subsidence problems.

[0090] (3) Data encoding and compression unit:

[0091] Function description: Encode and compress the processed signal data to reduce the amount of data transmission and improve transmission efficiency.

[0092] Technical Implementation: Highly efficient data compression algorithms (such as Huffman coding and LZW algorithm) are used to minimize data size while ensuring data integrity. Data encryption technology is also used to ensure security during data transmission.

[0093] (4) Communication transmission unit:

[0094] Function description: Transmit the encoded and compressed data to the backend data center via wireless or wired means.

[0095] Technical implementation: Wireless transmission: Utilize wireless communication technologies such as 4G / 5G, LoRa, and NB-IoT to achieve long-distance, low-power data transmission.

[0096] Wired transmission: In specific scenarios, such as when a fiber optic network has been laid along a highway, fiber optic communication can be used to ensure high-speed and stable data transmission.

[0097] (5) Status monitoring and fault alarm unit:

[0098] Function description: Real-time monitoring of the working status of the signal processing and transmission units. Once an abnormality is found (such as signal loss, transmission interruption), the alarm mechanism will be triggered immediately.

[0099] Technical implementation: Integrated status monitoring module, which detects the working status of each component through built-in self-diagnosis program and sends fault alarm information to management personnel through LED indicator lights, SMS or email, etc.

[0100] See also Figure 4, obtain multi-source heterogeneous data; wherein, the multi-source heterogeneous data includes: micro-motion sensor data, electromagnetic sensor data and natural potential data; respectively extract the time domain features, frequency domain features and spatial features of the micro-motion sensor data and electromagnetic sensor data; extract the time domain features and spatial features of the natural potential data; integrate the extracted time domain features, frequency domain features and spatial features of the micro-motion sensor data and electromagnetic sensor data, as well as the extracted time domain features and spatial features of the natural potential data into multi-source feature data; unify the timestamps of the neutron data in the multi-source feature data through the GPS / NTP protocol; map the sensor coordinates to the UTM grid coordinate system, and construct a local spatial grid; align the time domain, frequency domain and spatial features of each sensor according to the time window to form a multi-dimensional input matrix; perform separate standardization on each feature dimension to obtain a multi-source time series matrix.

[0101] Retrieve a multi-source time series matrix and a pre-trained prediction model; input the multi-source time series matrix into the prediction model to obtain prediction information; wherein the prediction information includes: prediction value, risk score and warning signal; obtain a preset threshold; generate a visualization interface based on the prediction information and the preset threshold, and display the visualization interface to the corresponding technical personnel.

[0102] It is worth noting that the methods for obtaining the preset thresholds include:

[0103] Obtain historical characteristic information of historical road surface settlement, sort the historical characteristic information in chronological order, and divide the historical characteristic information into several groups according to a set period;

[0104] Calculate the baseline threshold for each group: p threshold =μ+2σ; where μ is the mean value of each set of feature information; σ is the variance of each set of feature information;

[0105] An environmental correction factor is obtained, and the threshold is adjusted by calculating the product of the environmental correction factor and the baseline threshold to obtain a preset threshold.

[0106] In a specific implementation, using data for prediction includes five parts: feature engineering and multi-source data fusion, lightweight time series prediction model, threshold setting and early warning mechanism, model verification and optimization, deployment and monitoring.

[0107] (1) Feature Engineering and Multi-Source Data Fusion: Extract spatiotemporal features that are effective for predictive models from multi-source heterogeneous data. Use feature-level fusion and spatiotemporal alignment mechanisms to achieve multi-source data fusion and support model prediction. Design a phased feature extraction strategy based on sensor type and data characteristics to ensure the scientific nature and engineering practicality of the technical solution. For micro-motion sensor and electromagnetic sensor data features, extract their time domain features, frequency domain features, and spatial features; for natural potential data features, extract their time domain features and spatial features.

[0108] Example: 1. The steps for extracting multi-source heterogeneous data are as follows:

[0109] Micro-motion sensor data feature extraction

[0110] Time domain characteristics:

[0111] Hilbert-Huang transform (HHT): Where x(t) is the original signal. IMF i (t) is the eigenmode function, r n (t) is the residual term;

[0112] Extract instantaneous energy Peak-to-Peak, root mean square RMS; where A i (t) represents the instantaneous amplitude;

[0113] Sliding window statistics:

[0114]

[0115] Frequency domain characteristics:

[0116] Windowed FFT (Hanning window, window length N=256): Where, X(f) is the spectrum amplitude at frequency f, t is the time index, N is the window length, and x win (t) is the value of the windowed signal at time t, e -j2πft / N is the complex exponential function (the basis function of Fourier transform); extract the main frequency f peak , frequency band energy (0-10Hz, 10-20Hz).

[0117] Spatial features:

[0118] Kriging interpolation: in, The predicted value at the target point s0, λ i is the weight coefficient, z(s i ) is a known point s i The observation value at s0 is the target point, s iFor known points; combined with semivariogram modeling, output target point prediction value and spatial gradient.

[0119] b. Electromagnetic sensor data feature extraction

[0120] Time domain characteristics:

[0121] Abnormal pulse detection: Pulse strength = max(x(t)) - sliding mean(x(t)); where x(t) is the time domain data collected by the original signal sensor. max(x(t)) is the maximum instantaneous value in the positioning signal.

[0122] Trend fitting: Use third-order polynomial to fit signal trends Extract coefficients a, b, and c.

[0123] Frequency domain characteristics:

[0124] Multi-band energy analysis: Divide the frequency bands (0-5Hz, 5-15Hz, 15-30Hz) and calculate the energy contribution of each band.

[0125] Spatial features:

[0126] Measurement point correlation matrix: Among them, Cov(z(s i ),z(s j )): covariance of the observation values ​​of measurement point i and measurement point j, The standard deviation z(s i ):Observation value of measuring point i, s i ,s j : The spatial position of measuring point i and measuring point j;

[0127] Output the eigenvalues ​​of the correlation matrix (such as the maximum eigenvalue, matrix rank).

[0128] c. Feature extraction of spontaneous potential data

[0129] Time domain characteristics:

[0130] Trend and Seasonal Decomposition: Use STL decomposition (Seasonal-Trend Decomposition): z(t) = Trend(t) + Seasonal(t) + Residual(t); where z(t): raw data reflects actual observations. Trend(t): long-term trend quantifies slow geological evolution, Seasonal(t): cyclical fluctuations associated with external environmental factors (such as rainfall and temperature), and Residual(t): random noise or anomalies support real-time monitoring and early warning; extract trend slope

[0131] Spatial features:

[0132] Fluid diffusion model interpolation: Construct groundwater velocity field based on Darcy's law and predict the spatial distribution of potential: in, is the gradient sign, K is the permeability coefficient, h is the head height, and S is the water storage coefficient.

[0133] 2. Use feature-level fusion and spatiotemporal alignment mechanisms to achieve multi-source data fusion.

[0134] 1) Space-time alignment

[0135] (a) Time synchronization:

[0136] The timestamps are unified through the GPS / NTP protocol, high-frequency data (micro-motion) are downsampled to 10 Hz, and low-frequency data (spontaneous potential) are interpolated to 10 Hz.

[0137] Interpolation method: x interp (t) = CubicSpline(t raw ,x raw )(t); where t raw represents the original time point sequence; x raw represents the original observation sequence, CubicSpline represents the cubic spline interpolation function, t is the target interpolation time point, x interp (t): estimated value after interpolation.

[0138] (b) Spatial alignment:

[0139] The sensor coordinates are mapped to the UTM grid coordinate system to construct a local spatial grid (resolution 1m×1m).

[0140] Coordinate transformation formula: (x utm ,y utm )=UTM_from_WGS84(lon,lat); where lon,lat: geographic coordinates → used for global positioning. utm ,y utm Represents plane coordinates; used for local spatial analysis and engineering applications. UTM_from_WGS84 represents a coordinate transformation bridge, ensuring spatial alignment of multi-source data and supporting feature fusion and visualization.

[0141] b. Feature-level fusion

[0142] 1) Feature stitching:

[0143] Align the time domain, frequency domain, and spatial features of each sensor according to the time window to form a multidimensional input matrix:

[0144] X=[F 微动 ,F 电磁 ,F 电位]∈R T×D ; Where T is the number of time steps, D is the total feature dimension; R is a set of real numbers, emphasizing the numerical characteristics of the input data (continuous and computable).

[0145] 2) Feature Normalization:

[0146] Perform Z-Score normalization independently for each feature dimension:

[0147]

[0148] Where x represents the original data value; μ represents the center position of the mean data set; σ represents the standard deviation; x norm Indicates the normalized value.

[0149] (2) Lightweight time series prediction model: The input layer is designed as a high-dimensional time series matrix after the fusion of multi-source features. A prediction model combining temporal convolutional network (TCN) and physical residual learning is adopted. The backbone network is determined to be TCN, which captures the temporal dependencies. Physical constraint branches are used to embed geomechanical equations to enhance the physical interpretability of the model.

[0150] Example: Time Series Forecasting Model Architecture

[0151] 1. Input layer design: high-dimensional time series matrix after multi-source feature fusion

[0152] Shape: X∈R T×D , where T is the time step (e.g., 300 steps, corresponding to a 5-minute window), and D is the total feature dimension (for example: 10 dimensions for micromotion + 8 dimensions for electromagnetics + 5 dimensions for natural potential → D = 23).

[0153] Normalization: Z-Score normalization is performed independently on each feature dimension: Among them, μ train ,σ train Calculated based on the training set.

[0154] 2.TCN backbone network: captures temporal dependencies and extracts high-order features.

[0155] Structure: 4 layers of dilated causal convolution (Temporal Convolutional Network, TCN) + residual connection + skip connection.

[0156] a. Dilated causal convolution formula:

[0157] The output feature H of layer l l Calculated as:

[0158] H l =ReLU(W l * d Hl-1 +b l ); where H l is the output of the lth layer, W l is the convolution kernel weight matrix of the i-th layer, * d represents the convolution operation with a dilation factor of d, b l is the bias term of the lth layer.

[0159] b. Residual connection:

[0160] The output of each layer is added to the input to prevent the gradient from disappearing:

[0161] in, is the final output of the lth layer;

[0162] c. Skip connection and feature aggregation:

[0163] The output of each layer is connected to the final layer after adjusting the number of channels through 1×1 convolution:

[0164] Among them, H skip The final output of the backbone network;

[0165] Final TCN output: H TCN =ReLU(H skip ).

[0166] 3. Physical constraint branch: embeds geomechanical equations to enhance the physical interpretability of the model.

[0167] a. Physical parameter input:

[0168] 1) Pore pressure change ΔP: Where g is the acceleration due to gravity, is the natural potential gradient, k is the permeability coefficient, and σ is the electrical conductivity of the soil.

[0169] 2) Soil layer thickness h:

[0170] Among them, α is the regional experience coefficient, S max It is the maximum allowable compression of soil or the design settlement limit.

[0171] 3) Elastic modulus E: E = ρ·v 2 , ρ is soil density, v is electromagnetic wave velocity

[0172] 4) Time constant τ:

[0173] Fitting the one-dimensional consolidation equation through the historical settlement curve:

[0174] b. Calculation of theoretical settlement:

[0175] Based on the one-dimensional consolidation theory formula:

[0176] Among them, τ is the one-dimensional consolidation equation fitted by the historical settlement curve; and output S phys ∈R T×1 .

[0177] 4. Residual Fusion Mechanism

[0178] a. Residual calculation: R = H TCN -S phys ;

[0179] b. Adaptive fusion:

[0180] Introduce a learnable weight α to control the residual contribution:

[0181] Initialized to 0.5 and optimized by backpropagation.

[0182] 5. Loss Function Design

[0183] Jointly optimize data fitting error, physical residual constraints, and time series trend consistency:

[0184] in, is the predicted settlement sequence, Y true is the true settlement sequence, R is the residual vector, ‖‖2 is the L2 norm, which is used to quantify the amplitude of the residual or vector. is the first-order difference of the predicted value, is the first-order difference of historical data.

[0185] 6. Output Layer

[0186] Mapped to sedimentation through the fully connected layer:

[0187] Output shape:

[0188] (3) Threshold setting and early warning mechanism: A dynamic threshold mechanism and early warning mechanism are set up. The output part includes predicted values, risk scores, early warning signals, and visual displays. Through a multi-level design, a full link closed loop from data to decision-making is achieved. Based on real-time multi-source data, environmental and traffic characteristics, continuous adjustments are made to adapt to environmental changes and mine data features. A three-level early warning mechanism is set up to assist decision-making and improve the efficiency and scientificity of addressing road sedimentation problems.

[0189] Threshold setting and early warning mechanism

[0190] Output content: predicted value, risk score, early warning signal and visual display.

[0191] 1. Risk score calculation

[0192] a. Historical baseline:

[0193] b. Real-time risk scoring:

[0194] Among them, p t For real-time risk scoring, is the predicted settlement at the current time step, is the predicted settlement at the i-th time step in the past 30 days.

[0195] 2. Dynamic thresholds and warning rules

[0196] a. Baseline threshold: p threshold =μ+2σ; where μ is the mean value of the corresponding group data and σ is the variance of the corresponding group data.

[0197] b. Environmental correction factor:

[0198] When the rainfall is >50mm for 3 consecutive days: p threshold ←p threshold ×0.7

[0199] When the daily rise of groundwater level is greater than 10cm: threshold ←p threshold ×0.8

[0200] c. Multi-level warning rules:

[0201] Level 1 (monitoring level): p ≥ 0.4 or autoencoder abnormality persists for > 2 hours;

[0202] Level 2 (warning level): p ≥ 0.7 or Level 1 lasts for > 12 hours;

[0203] Level 3 (Emergency): p ≥ 0.9 or the RMS of the micro-signal increases by more than 50% week-on-week.

[0204] (4) Model validation and optimization: This includes offline validation datasets, phased training design, online active learning mechanisms, and lightweight model design. The training / validation / test sets are divided into an 8:1:1 ratio, time travel is strictly prohibited, and stratified sampling is performed in proportion to the training strategy.

[0205] Example: Model Validation and Optimization

[0206] The training / validation / test sets are divided into 8:1:1, time travel is strictly prohibited, and stratified sampling is performed in proportion.

[0207] 1. Training in stages

[0208] a. Physics branch pre-training (first 100 rounds):

[0209] Fix α=0 and optimize only the physical parameter mapping layer.

[0210] Loss function: Among them, except for the loss mixing coefficient of the physical parameter mapping layer which is non-zero, S phys is the collected physical value; y true is the pre-trained prediction value, and MSE is the mean square error.

[0211] b. Joint training (after 100 rounds):

[0212] Release all parameters and optimize the hybrid loss: in, is the predicted settlement sequence, Y true is the true settlement sequence, R is the residual vector, ‖‖2 is the L2 norm, which is used to quantify the amplitude of the residual or vector. is the first-order difference of the predicted value, is the first-order difference of historical data.

[0213] 2. Online active learning

[0214] Daily new data triggers model fine-tuning, using elastic weight consolidation (EWC) to prevent catastrophic forgetting: in, is the total loss function, F i is the diagonal element of the ith parameter in the Fisher information matrix, θ i is the current value of the i-th parameter of the model, θ i,old is the historical value of the i-th parameter of the model.

[0215] 3. Model lightweight design

[0216] a. Channel compression:

[0217] Add a channel compression module to the last layer of TCN: H compressed =ConvlD 1×1 (H TCN , channels = 16);

[0218] Among them, the backbone network adopts the time convolution network TCN; among them, Conv1D 1×1 To apply 1×1 convolution to the output of layer 1, H TCN is the output of the temporal convolutional network, channels=16 means the number of output channels is set to 16

[0219] b. Quantization-aware training:

[0220] Simulate 8-bit integer quantization during training (sample code):

[0221] python model=torch.quantization.quantize_dynamic(model,{nn.Conv1d},dtype=torch.qint8)

[0222] (5) Deployment and Monitoring: Design the edge layer, cloud layer, and visualization interface. The edge layer (road segment equipment) is used to run the autoencoder, calculate micro-motion temporal features, and cache 12 hours of raw data. The cloud layer (central server) is used for storage and computing. The visualization interface displays the risk score, threshold line, and historical comparison curve for each road segment. The visualization interface, combined with the threshold, can provide engineers with decision-making information and recommendations.

[0223] Example: Deployment and Monitoring

[0224] 1. Edge layer (road equipment)

[0225] Real-time calculation of micro-motion signal RMS and peak value: Window RMS: Calculated within a sliding window: Window

[0226] Peak-to-Peak: Divide the signal into sliding windows of fixed length. The difference between the maximum and minimum values ​​of the entire signal is: Peak-to-Peak = max(x(t)) - min(x(t))

[0227] Run the lightweight autoencoder to detect anomalies:

[0228] Anomaly Score=‖x-Decoder(Encoder(x))‖2

[0229] x: original sensor signal (such as micro-vibration waveform, electromagnetic pulse sequence);

[0230] z = Encoder(x): Extracts the core pattern of the signal (such as the main vibration frequency and pulse intensity);

[0231] Decoder reconstruction: reconstructs the original signal from the compressed features.

[0232] 2. Cloud layer (central server):

[0233] Storage: Historical data is stored in a time series database (InfluxDB) and a spatial database (PostGIS).

[0234] Computation: Distributed training of XGBoost auxiliary model, objective function:

[0235] Where i represents the i-th sample in the data set; y i represents the true sedimentation value of the i-th sample, represents the sedimentation prediction value of the XGBoost model for the i-th sample, represents the loss function, γ represents the leaf node splitting threshold; T represents the number of leaf nodes; λ represents the L2 regularization coefficient; w represents the leaf weight vector; ‖w‖ 2 Represents the squared L2 norm of the leaf weight.

[0236] 3. Visual interface:

[0237] Displays the risk score, threshold line, and historical comparison curve of each road section, providing engineers with decision-making recommendations.

[0238] See also Figure 5 , a second aspect of the present invention provides an electronic device, comprising: a communication unit and a processing unit;

[0239] A communication unit, used to obtain road surface settlement signals of the highway;

[0240] The processing unit is used to extract features from road subsidence signals to obtain characteristic information of road subsidence; compare subsidence thresholds with characteristic information to determine subsidence sections of marked highways; obtain multi-source heterogeneous data; fuse the multi-source heterogeneous data to obtain a multi-source time series matrix; and predict the subsidence sections based on the multi-source time series matrix to obtain prediction information.

[0241] Some of the data in the above formula are calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0242] The working principle of the present invention is as follows: the present invention obtains a road surface settlement signal of a highway; performs feature extraction on the road surface settlement signal to obtain characteristic information of the road surface settlement; compares the settlement threshold and the characteristic information to determine the settlement section of the marked highway; obtains multi-source heterogeneous data; fuses the multi-source heterogeneous data to obtain a multi-source time series matrix; and predicts the settlement section based on the multi-source time series matrix to obtain prediction information.

[0243] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for monitoring and predicting highway pavement settlement based on multi-source data, characterized in that: include: Obtain road surface settlement signals; Extract features from road subsidence signals to obtain characteristic information of road subsidence; Compare the subsidence threshold and feature information to identify the subsidence sections of the marked highway; Acquire multi-source heterogeneous data; fuse the multi-source heterogeneous data to obtain a multi-source time series matrix; The subsidence section is predicted based on the multi-source time series matrix to obtain prediction information.

2. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 1, characterized in that: The road surface settlement signal collection method includes: Divide the highway into several sections of roads to be tested based on their length; obtain risk scores of the several roads to be tested, and determine the step length of setting the measuring points according to the risk scores; The acquisition station is set on the road to be measured according to the setting step of the measuring point; the multi-physical field sensor of the acquisition station is used to collect the road surface settlement signal on the road according to the preset collection frequency; wherein, the road surface settlement signal is the integration of the time series signals collected by each sensor.

3. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 1, characterized in that: The feature extraction according to the road subsidence signal includes: Preprocessing the road subsidence signal; wherein the preprocessing includes: missing value filling, denoising and removing outliers; The pre-processed road surface settlement signal is converted into digital form to obtain a settlement digital signal; the settlement digital signal is processed using a signal processing algorithm to obtain characteristic information; wherein the signal processing algorithm includes: Fourier transform or wavelet transform.

4. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 1, characterized in that: The method of marking the subsided road section according to the subsidence threshold and characteristic information includes: comparing the feature information with a sedimentation threshold; When the characteristic information is greater than the settlement threshold, the corresponding highway will be marked to obtain the settlement section and an abnormal signal will be generated; otherwise, the road surface settlement signal of the highway will be continuously collected and the abnormal signal will be sent to the corresponding technician.

5. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 1, characterized in that: The multi-source heterogeneous data is fused to obtain a multi-source time series matrix, including: Retrieve multi-source heterogeneous data; where multi-source heterogeneous data includes: micro-motion sensor data, electromagnetic sensor data, and natural potential data; The features of multi-source heterogeneous data are extracted separately to obtain multi-source feature data; the multi-source feature data are fused using feature-level fusion and spatiotemporal alignment mechanisms to obtain a multi-source time series matrix.

6. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 5, characterized in that: The extracting features of multi-source heterogeneous data to obtain multi-source feature data includes: Retrieving micro-motion sensor data and electromagnetic sensor data from multi-source heterogeneous data; extracting time domain features, frequency domain features, and spatial features of the micro-motion sensor data and electromagnetic sensor data respectively; Retrieve the natural potential data from multi-source heterogeneous data; extract the time domain features and spatial features of the natural potential data; The time domain features, frequency domain features and spatial features of the extracted micro-motion sensor data and electromagnetic sensor data, as well as the time domain features and spatial features of the extracted natural potential data are integrated into multi-source feature data.

7. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 5, characterized in that: The method of fusing multi-source feature data using feature-level fusion and spatiotemporal alignment mechanisms includes: Retrieve multi-source feature data; unify the timestamps of sub-data in multi-source feature data through GPS / NTP protocol; map sensor coordinates to the UTM grid coordinate system and construct a local spatial grid; The time domain, frequency domain, and spatial features of each sensor are aligned according to the time window to form a multi-dimensional input matrix; each feature dimension is separately standardized to obtain a multi-source time series matrix.

8. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 1, characterized in that: The predicting of the subsidence section according to the multi-source time series matrix includes: Retrieve multi-source time series matrices and pre-trained prediction models; Input the multi-source time series matrix into the prediction model to obtain prediction information; wherein the prediction information includes: prediction value, risk score and early warning signal; Obtain a preset threshold; generate a visualization interface based on the prediction information and the preset threshold, and display the visualization interface to the corresponding technical personnel.

9. The method for monitoring and predicting road pavement subsidence based on multi-source data according to claim 8, characterized in that: The method for obtaining the preset threshold includes: Obtain historical characteristic information of historical road surface settlement, sort the historical characteristic information in chronological order, and divide the historical characteristic information into several groups according to a set period; Calculate the baseline threshold for each group: p threshold =μ+2σ; where μ is the mean value of each set of feature information; σ is the variance of each set of feature information; An environmental correction factor is obtained, and the threshold is adjusted by calculating the product of the environmental correction factor and the baseline threshold to obtain a preset threshold.

10. An electronic device, applied to the method for monitoring and predicting road pavement subsidence based on multi-source data as claimed in any one of claims 1 to 9, characterized in that: include: a communication unit and a processing unit; A communication unit, used to obtain road surface settlement signals of the highway; The processing unit is used to extract features from road subsidence signals to obtain characteristic information of road subsidence; compare subsidence thresholds with characteristic information to determine subsidence sections of marked highways; obtain multi-source heterogeneous data; fuse the multi-source heterogeneous data to obtain a multi-source time series matrix; and predict the subsidence sections based on the multi-source time series matrix to obtain prediction information.

Images