A Method and System for Safety Monitoring of Railway Blasting Sites Based on Vibration Monitors

Abstract

This invention relates to the field of engineering safety monitoring technology, and discloses a method and system for safety monitoring of railway blasting sites based on a vibration monitoring instrument. The method includes acquiring an initial three-dimensional surface model of the railway blasting area and dividing it into grid cells to extract surface geometric features to obtain a surface feature matrix; simultaneously acquiring raw vibration signals collected by sensors and performing integral smoothing transformation to obtain a displacement time series; performing weighted mapping, spatial interpolation, and deformation reconstruction on the two to obtain a continuous surface deformation model; comparing the model with the initial model at nodes to obtain a three-dimensional deformation vector, and combining this with safety standard threshold mapping to generate a risk distribution map; finally, extracting deformation time series data of high-risk areas, extracting dynamic parameters through first-order and second-order differences to deduce the evolution trend of instability anomalies, and distributing safety warnings. This invention achieves accurate monitoring of deformation at railway blasting sites.

Description

Technical Field

[0001] This invention relates to the field of engineering safety monitoring technology, and in particular to a method and system for safety monitoring of railway blasting sites based on a vibration monitoring instrument. Background Technology

[0002] Currently, in the construction, operation, and maintenance of railway infrastructure, safety monitoring of blasting operations along the line is a crucial link in ensuring the stability of the track structure and the safety of the surrounding environment. Given the complex terrain, precise monitoring of minute surface displacements and macroscopic deformations caused by blasting is an important means of preventing geological disasters and engineering accidents.

[0003] In a current technology, ground vibration monitors are typically deployed for localized point monitoring, while satellite remote sensing imagery is combined for big data analysis to assess surface changes over a wide area. This approach relies primarily on periodic macroscopic comparisons of remote sensing images or microscopic threshold alarm mechanisms from single-point devices to identify abnormal states. However, this monitoring approach faces significant limitations in practical applications. On the one hand, there is a severe spatiotemporal asynchrony between the area-level images acquired by satellite remote sensing and the point-level vibration data collected by ground sensors. On the other hand, existing solutions lack a deep fusion mechanism to address the spatiotemporal discrepancies of physical data when processing this multi-source heterogeneous information. They often only perform simple result overlay, failing to accurately reflect the evolution of dynamic surface displacement. This results in information silos between microscopic vibrations and macroscopic deformations, making it impossible to transform scattered data into unified dynamic monitoring results.

[0004] Existing technologies suffer from inaccurate dynamic perception of surface deformation in complex blasting environments and delayed early warning systems. Summary of the Invention

[0005] This invention provides a method and system for safety monitoring of railway blasting sites based on a vibration monitoring instrument, in order to solve the technical problems of inaccurate dynamic perception of surface deformation and delayed early warning in the complex blasting environment in the prior art.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a method for safety monitoring of railway blasting points based on a vibration monitoring instrument, comprising: An initial three-dimensional model of the surface of the railway blasting area is obtained, the initial three-dimensional model of the surface is divided into multiple grid units, and the surface geometric features of each grid unit are extracted to obtain the surface feature matrix. The original vibration signal collected by the sensor is acquired, and the original vibration signal is subjected to integral smoothing transformation to obtain the displacement time sequence; The displacement time series is weighted and mapped to the surface feature matrix to obtain discrete correlation features. Spatial interpolation is performed on the discrete correlation features to obtain the correlation feature matrix. Deformation reconstruction is performed on the correlation feature matrix to obtain the surface deformation model. The surface deformation model is compared with the initial three-dimensional surface model by node alignment to obtain a three-dimensional deformation vector. The three-dimensional deformation vector is then mapped to a risk level according to a preset safety standard threshold to obtain a risk distribution map. Deformation time series data of high-risk areas are extracted from the risk distribution map. Dynamic features are extracted from the deformation time series data to obtain deformation dynamic parameters. Evolution is deduced based on the deformation dynamic parameters to obtain the instability anomaly evolution trend. Based on the aforementioned unstable and abnormal evolution trend, safety early warnings are distributed to obtain early warning report data streams, thereby enabling safety monitoring of railway blasting sites.

[0007] Preferably, obtaining the initial three-dimensional surface model of the railway blasting area includes: UAV lidar point cloud data of the railway blasting area is acquired, and the lidar point cloud data is filtered and resampled to obtain a digital elevation model. The digital elevation model is then used as the initial three-dimensional surface model.

[0008] Preferably, the step of extracting surface geometric features from each grid cell to obtain a surface feature matrix includes: Calculate the surface normal vector of each grid cell in the initial three-dimensional surface model; Calculate the slope and aspect characteristics of each grid cell based on the surface normal vector; The slope features and aspect features are arranged in a matrix according to the spatial index corresponding to each grid cell to obtain the surface feature matrix.

[0009] Preferably, the step of performing an integral smoothing transformation on the original vibration signal to obtain a displacement time sequence includes: The original vibration signal is bandpass filtered to obtain a filtered vibration signal. The filtered vibration signal is then integrated twice to obtain an integrated vibration signal. The integrated vibration signal is then baseline-corrected to obtain initial displacement data. The initial displacement data were analyzed in the frequency domain using the Fast Fourier Transform algorithm to obtain the dominant frequency components of the blasting vibration. The initial displacement data is denoised by low-pass filtering based on the dominant frequency component of the blasting vibration to obtain the displacement time sequence.

[0010] Preferably, the step of spatially interpolating the discrete correlation features to obtain a correlation feature matrix, and then performing deformation reconstruction on the correlation feature matrix to obtain a surface deformation model, includes: A spatial variability function is constructed based on the spatial distribution differences of the discrete correlation features, and a Kriging estimation equation system is constructed based on the spatial variability function. The interpolation weights and estimation variances of each grid cell in the surface feature matrix are solved using the Kriging estimation equations. The discrete correlation features are then weighted and summed according to the interpolation weights to obtain the correlation feature matrix. Extract the feature value change sequence of each grid cell in the correlation feature matrix within a continuous time step, perform trend fitting on the feature value change sequence, and obtain the real-time deformation rate of each grid cell. Using the reciprocal of the estimated variance as the fitting weight, the real-time deformation rate is fitted using a preset spatial surface model to obtain the surface deformation model.

[0011] Preferably, the step of comparing the node alignment of the surface deformation model with the initial three-dimensional surface model to obtain a three-dimensional deformation vector, and mapping the three-dimensional deformation vector to a risk level according to a preset safety standard threshold to obtain a risk distribution map, includes: The real-time three-dimensional spatial coordinates of each grid cell in the surface deformation model are subtracted from the initial three-dimensional spatial coordinates of the corresponding grid node in the initial three-dimensional surface model to obtain the position deviation vector of each grid cell, and the position deviation vector is determined as the three-dimensional deformation vector. The modulus difference is determined based on the Euclidean norm of the three-dimensional deformation vector, and the deviation angle is determined based on the relative spatial angle between the three-dimensional deformation vector and the preset reference axis. A static multi-level deformation range is established based on the preset standard displacement limit, and the multi-level deformation range is determined as the preset safety standard threshold. The modulus difference and deviation angle of each grid cell are compared with the safety standard threshold to obtain the corresponding risk level, and the risk level is mapped to the corresponding spatial coordinates to obtain a risk distribution map.

[0012] Preferably, the step of extracting deformation time-series data of high-risk areas based on the risk distribution map, extracting dynamic features from the deformation time-series data to obtain deformation dynamic parameters, and performing evolutionary deduction based on the deformation dynamic parameters to obtain the instability anomaly evolution trend includes: Extract connected grid cells in the risk distribution map whose risk level exceeds the preset warning level to obtain high-risk areas; Extract the cumulative deformation displacement sequence of the high-risk area within a preset sliding time window to obtain deformation time series data; The deformation time series data is subjected to first-order difference to obtain the real-time deformation rate. The real-time deformation rate is subjected to second-order difference to obtain the real-time deformation acceleration. The real-time deformation rate and the real-time deformation acceleration are determined as deformation dynamics parameters. If the real-time deformation acceleration is continuously greater than zero and the real-time deformation rate exceeds the preset engineering instability critical threshold, then it is determined to be an unstable abnormal evolution trend.

[0013] Preferably, the step of distributing safety warnings based on the unstable anomaly evolution trend to obtain a warning report data stream includes: Extract the spatial coordinates of the high-risk areas where the aforementioned instability and abnormal evolution trend occurs to obtain the risk spatial coordinates; Based on the deformation dynamics parameters, a time recursion is performed to obtain the risk peak time window; By integrating the risk spatial coordinates, the risk peak time window, and the risk level corresponding to the risk distribution map, and encapsulating them according to a preset data format, early warning result data is obtained. The integrity verification calculation is performed on the warning result data to generate a data verification identifier, and the data verification identifier is added to the warning result data to obtain the warning report data stream.

[0014] Secondly, the present invention provides a railway blasting point safety monitoring system based on a vibration monitoring instrument, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains the initial three-dimensional surface model of the railway blasting area and extracts the surface geometric features by dividing the grid unit to obtain the surface feature matrix. Simultaneously, it obtains the original vibration signal collected by the sensor to obtain the displacement time sequence. Then, it performs weighted mapping, spatial interpolation and deformation reconstruction on the two to obtain the surface deformation model. By performing spatial topological weighted fusion of discrete sensor micro displacement point data and continuous surface grid matrix reflecting surface features, and performing Kriging completion and spatial surface fitting based on the spatial variogram, it realizes the spatiotemporal coupling and full grid coverage of multi-source heterogeneous point and surface data. Thus, it effectively fills the data gap in non-monitoring point areas, breaks the spatiotemporal barrier between macro geometric features and micro displacement data, realizes the efficient integration of multi-source heterogeneous information, improves the accuracy and global integrity of dynamic perception of surface deformation under complex blasting environment, and avoids the monitoring blind spot caused by a single data source.

[0017] (2) This invention obtains a three-dimensional deformation vector by comparing the nodes of the surface deformation model with the initial three-dimensional surface model, and calculates the difference in modulus and the deviation angle. Then, it maps the three-dimensional deformation vector to a risk level according to the preset safety standard threshold to obtain a risk distribution map. By finely depicting the real physical characteristics of terrain evolution in the multi-dimensional space of the Euclidean norm and the spatial angle, and directly comparing the calculated absolute deformation with the preset static multi-level deformation interval, it gets rid of the strong dependence on the characteristics of strain history disaster samples. Thus, it provides a standardized and objective physical scale for deformation damage in complex geological environments, effectively avoids the evaluation distortion caused by the lack of sample data or spatiotemporal unevenness, reduces the false alarm rate of the system, and improves the scientificity of risk level assessment and the reliability of early warning triggering. (3) This invention extracts deformation time series data of high-risk areas based on risk distribution map, performs first-order and second-order differences to obtain real-time deformation rate and real-time deformation acceleration, uses these as deformation dynamic parameters to determine the instability anomaly evolution trend, and combines the kinematic nonlinear evolution equation to perform time recursion to obtain early warning report data stream; by discretizing the cumulative deformation displacement of high-risk connected areas in the time domain, it accurately identifies the nonlinear accelerated creep characteristics with acceleration continuously greater than zero, locks the gradual failure stage of structural instability, and uses the kinematic equation to reversely calculate the time root of the remaining safe displacement margin to reach the limit displacement; thus, it successfully transforms the traditional static threshold or post-event passive alarm mechanism into a dynamic forward-looking prediction with a clear time dimension, significantly shortens the lag time of safety early warning, and provides a precise and quantitative countdown emergency response window for the prevention of sudden destructive disasters. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the safety monitoring method for railway blasting points based on a vibration monitoring instrument provided in the first embodiment of the present invention. Detailed Implementation

[0019] 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.

[0020] Reference Figure 1 The first embodiment of the present invention provides a method for safety monitoring of railway blasting sites based on a vibration monitoring instrument, including the following steps: S1, Obtain the initial three-dimensional surface model of the railway blasting area, divide the initial three-dimensional surface model into multiple grid units, extract the surface geometric features of each grid unit, and obtain the surface feature matrix; S2, acquire the original vibration signal collected by the sensor, perform integral smoothing transformation on the original vibration signal to obtain the displacement time sequence; S3, perform weighted mapping between the displacement time series and the surface feature matrix to obtain discrete correlation features, perform spatial interpolation on the discrete correlation features to obtain the correlation feature matrix, and perform deformation reconstruction on the correlation feature matrix to obtain the surface deformation model; S4, perform node alignment comparison between the surface deformation model and the initial three-dimensional surface model to obtain a three-dimensional deformation vector, and perform risk level mapping on the three-dimensional deformation vector according to the preset safety standard threshold to obtain a risk distribution map; S5. Extract deformation time series data of high-risk areas based on the risk distribution map, extract dynamic features from the deformation time series data to obtain deformation dynamic parameters, and perform evolution deduction based on the deformation dynamic parameters to obtain the instability anomaly evolution trend. S6. Based on the unstable and abnormal evolution trend, safety early warning is distributed to obtain early warning report data stream, thereby realizing safety monitoring of railway blasting points; In step S1, an initial three-dimensional surface model of the railway blasting area is obtained, the initial three-dimensional surface model is divided into multiple grid units, and surface geometric features are extracted from each grid unit to obtain a surface feature matrix.

[0021] The initial three-dimensional surface model of the railway blasting area is obtained, including: UAV lidar point cloud data of the railway blasting area is acquired, and the lidar point cloud data is filtered and resampled to obtain a digital elevation model. The digital elevation model is then used as the initial three-dimensional surface model.

[0022] In one implementation, a drone equipped with a lidar sensor deployed over the survey area performs a large-area scan to acquire surface point cloud data of the railway blasting area. This embodiment filters and resamples the lidar point cloud data. Specifically, a well-known multi-overhead window filtering algorithm is used to remove non-ground interference points caused by vegetation, buildings, or aerial noise, retaining the true ground point cloud. Next, a preset resampling resolution is set to perform spatial gridding resampling of the ground point cloud to obtain a digital elevation model (DEM), which is then used as the initial three-dimensional surface model.

[0023] It should be noted that the preset resampling resolution can be specifically set to 0.5 meters. The engineering rationale for choosing this value is as follows: the physical critical size of geometric features such as micro-landslides, micro-cracks, or local subsidence caused by blasting along the railway line is usually between 0.5 meters and 1.0 meters. Setting the resampling resolution to 0.5 meters can eliminate redundant high-frequency point cloud noise to the greatest extent possible without smoothing out these micro-geometric deformation features, thereby controlling the data scale of the digital elevation model, preventing memory overflow in subsequent matrix operations, and playing a key role in balancing spatial detail and computational efficiency.

[0024] Among them, surface geometric features are extracted from each grid cell to obtain the surface feature matrix, including: Calculate the surface normal vector of each grid cell in the initial three-dimensional surface model; Calculate the slope and aspect characteristics of each grid cell based on the surface normal vector; The slope features and aspect features are arranged in a matrix according to the spatial index corresponding to each grid cell to obtain the surface feature matrix.

[0025] In one implementation, this embodiment divides the initial 3D surface model into multiple grid cells with a resolution of 0.5 meters. For each grid cell, the 3D coordinates of its center node and adjacent grid cells are extracted. A local spatial tangent plane is constructed using a known least-squares fitting method, and the unit vector perpendicular to this spatial tangent plane is calculated and output as the surface normal vector of each grid cell.

[0026] It is worth noting that this embodiment calculates the slope and aspect characteristics of each grid cell based on the surface normal vector. Specifically, the projection components of the surface normal vector on the horizontal plane and the vertical axis components are extracted. By calculating the arctangent between the magnitudes of the vertical axis components and the projection components on the horizontal plane, the corresponding slope characteristics are obtained; simultaneously, the clockwise angle between the projection components on the horizontal plane and the preset north-direction axis is calculated, and the corresponding aspect characteristics are obtained. The slope characteristics characterize the steepness of the terrain, and the aspect characteristics characterize the potential sliding direction of deformation in the spatial gravity direction. Together, they constitute the core indicators reflecting the initial geometric vulnerability of the terrain. In this embodiment, according to the spatial index corresponding to each grid cell, the slope and aspect characteristics are transformed into multi-dimensional feature vectors, and arranged in a matrix according to the spatial geometric topology, outputting the surface feature matrix containing the background color characteristics of the entire terrain area.

[0027] In step S2, the original vibration signal collected by the sensor is acquired, and the original vibration signal is subjected to integral smoothing transformation to obtain the displacement time sequence.

[0028] The original vibration signal is subjected to integral smoothing transformation to obtain a displacement time sequence, including: The original vibration signal is bandpass filtered to obtain a filtered vibration signal. The filtered vibration signal is then integrated twice to obtain an integrated vibration signal. The integrated vibration signal is then baseline-corrected to obtain initial displacement data. The initial displacement data were analyzed in the frequency domain using the Fast Fourier Transform algorithm to obtain the dominant frequency components of the blasting vibration. The initial displacement data is denoised by low-pass filtering based on the dominant frequency component of the blasting vibration to obtain the displacement time sequence.

[0029] In one implementation, a triaxial piezoelectric accelerometer deployed on the railway track bed or subgrade slope collects transient impact vibrations caused by blasting in real time, obtaining the original vibration signal, which is a discrete time-domain acceleration sequence. This embodiment performs bandpass filtering on the original vibration signal to obtain a filtered vibration signal. It should be noted that the cutoff frequency range of the bandpass filter can be specifically defined as 0.5 Hz to 150.0 Hz. The engineering rationale for choosing this frequency range is that the core frequency band of the effective impact vibration energy wave of the rock and soil mass caused by engineering blasting is strictly concentrated between 0.5 Hz and 150.0 Hz. Setting the cutoff frequency range within this physical interval can accurately filter out ultra-low frequency foundation sway interference caused by railway traffic itself, as well as high-frequency environmental electromagnetic clutter interference, playing a controlling role in filtering irrelevant environmental background noise at the data source.

[0030] In one implementation, this embodiment utilizes the complex trapezoidal numerical integral formula to perform continuous quadratic numerical integration on the filtered vibration signal in the time domain, obtaining the integrated vibration signal. Since the inherent zero-point drift error of the sensor is amplified in the integration calculation, generating a nonlinear divergent trend term over time, this embodiment performs baseline correction on the integrated vibration signal. A univariate high-order trend polynomial is constructed, and the integrated vibration signal is curve-fitted using the least squares method to extract the low-frequency constant drift and linear divergent trend term. Subsequently, the divergent trend term is subtracted from the integrated vibration signal to eliminate accumulated error, obtaining the physically accurate initial displacement data.

[0031] It is worth noting that this embodiment uses a Fast Fourier Transform (FFT) algorithm to perform frequency domain analysis on the initial displacement data to obtain the dominant frequency component of the blasting vibration. First, the initial displacement data is windowed and truncated using a Hanning window function to suppress spectral leakage. Then, the windowed data is input into the FFT operator to transform the signal from the time domain to the frequency domain, and the amplitude spectrum is calculated. Within a preset blasting dominant frequency search range, specifically set to 5.0 Hz to 50.0 Hz, this range corresponds to the dominant frequency band where the transient destructive energy of the blasting is most concentrated. The maximum value of the amplitude spectrum is retrieved, and the frequency value corresponding to the largest amplitude is determined as the dominant frequency component of the blasting vibration.

[0032] In one implementation, the initial displacement data is low-pass filtered and denoised based on the dominant frequency component of the blasting vibration to obtain a displacement time series. This embodiment constructs a low-pass digital filter whose cutoff frequency is adaptively set to 1.5 times the dominant frequency component of the blasting vibration. The rationale for choosing 1.5 times as the adaptive cutoff coefficient is that the dynamic displacement response information of blasting vibration mainly consists of the dominant frequency and its low-order harmonics. Controlling the cutoff frequency at 1.5 times the dominant frequency not only preserves the energy intensity and temporal contour of the blasting main shock waveform but also accurately removes the minute glitch noise remaining in the high-frequency band after quadratic integration and baseline correction. This prevents high-frequency glitch from causing severe Runge phenomenon and pseudo-deformation in subsequent spatial surface fitting, thus playing a crucial role in balancing time-domain waveform fidelity and high-frequency glitch suppression. Through this low-pass filtering process, a smooth displacement time series with a high signal-to-noise ratio is output.

[0033] In step S3, the displacement time series is weighted and mapped to the surface feature matrix to obtain discrete correlation features. Spatial interpolation is performed on the discrete correlation features to obtain the correlation feature matrix. Deformation reconstruction is performed on the correlation feature matrix to obtain the surface deformation model.

[0034] Specifically, spatial interpolation is performed on the discrete correlation features to obtain a correlation feature matrix, and deformation reconstruction is performed on the correlation feature matrix to obtain a surface deformation model, including: A spatial variability function is constructed based on the spatial distribution differences of the discrete correlation features, and a Kriging estimation equation system is constructed based on the spatial variability function. The interpolation weights and estimation variances of each grid cell in the surface feature matrix are solved using the Kriging estimation equations. The discrete correlation features are then weighted and summed according to the interpolation weights to obtain the correlation feature matrix. Extract the feature value change sequence of each grid cell in the correlation feature matrix within a continuous time step, perform trend fitting on the feature value change sequence, and obtain the real-time deformation rate of each grid cell. Using the reciprocal of the estimated variance as the fitting weight, the real-time deformation rate is fitted using a preset spatial surface model to obtain the surface deformation model.

[0035] In one implementation, a weighted mapping is performed between the displacement time series and the surface feature matrix to obtain discrete correlation features. The sensor node set is defined as follows: Each sensor at any time The displacement time series value is The set of grid cell center points in the surface feature matrix is The surface feature vector corresponding to each grid cell is: .

[0036] First, calculate the values ​​for each sensor node. To the center of the grid cell Spatial Euclidean distance Based on the physical characteristic that blasting vibration energy decreases inversely with distance, the distance attenuation weight is defined as:

[0037] Then, the displacement values ​​of each sensor are spatially weighted and fused using distance weights to obtain the fused displacement scalar of the grid cell: The fused displacement scalar is the discrete correlation feature.

[0038] It is worth noting that in this embodiment, a spatial variogram is constructed based on the spatial distribution differences of the discrete correlation features, and a Kriging estimation equation system is constructed based on the spatial variogram.

[0039] Specifically, the spatial relative distances between known sensor locations are extracted from the discrete correlation features. As the independent variable, the semivariance of the corresponding feature values ​​is extracted. As the dependent variable, this embodiment uses a spherical semivariogram model for fitting, the expression of which is:

[0040] in, It is the nugget constant (characterizing measurement error and microscale variability). For the partial sill value, For variable range (effective distance characterizing spatial correlation). Utilizing known sensor point pairs. The three parameters in the above model are calculated using the weighted least squares method, along with their corresponding semivariance values. , , The specific objective function to be constructed is:

[0041] in, This represents the number of point pairs within the distance interval; For sensor point pairs Spatial distance between them; Distance interval The measured semivariance value at the location; For the spherical semivariogram model in The theoretical semivariance value at the specified location. The optimal parameter values ​​are obtained by solving using the Newton-Raphson iterative method. Subsequently, based on the grid cells to be interpolated... Based on the spatial geometric topological relationships between the sensor locations and known sensor sites in the neighborhood, and combined with the determined variogram parameters, a system of ordinary kriging equations is constructed:

[0042] in, The interpolation weights to be determined are: For Lagrange multipliers, For sensors , The distance between them For sensors Interpolation point The distance between them.

[0043] Using the Kriging estimation equations, the interpolation weights and estimation variances of each grid cell in the non-monitored area of ​​the surface feature matrix are calculated. Then, the known discrete correlation features within the neighborhood are weighted and summed using the interpolation weights to fill in the numerical gaps in the non-monitored area, resulting in the correlation feature matrix covering the entire grid area.

[0044] In one implementation, this embodiment extracts the feature value change sequence of each grid cell in the associated feature matrix within a continuous time step, performs trend fitting on the feature value change sequence, and obtains the real-time deformation rate of each grid cell. Specifically, using the time offset as the independent variable and the feature value as the dependent variable, a linear regression trend fitting is performed, and the regression coefficient obtained by fitting, i.e., the slope, is determined as the real-time deformation rate of the grid cell.

[0045] It should be noted that the preset duration corresponding to the continuous time step can be specifically defined as 24 hours. The physical and engineering rationale for choosing this value is that the response of the rock and soil mass caused by blasting operations includes both transient strong vibration impacts and nonlinear creep that is slowly released over time. Setting the time span for trend fitting to 24 hours ensures that the algorithm can fully capture the slow creep rate of the surface structure within a complete regular day-night cycle after the blasting; at the same time, it can effectively filter out the second-level high-frequency local vibration interference caused by a single transient blasting impact, as well as the periodic thermal drift error of the sensor caused by the day-night temperature difference, playing a key role in controlling the process between creep feature extraction and transient noise smoothing.

[0046] It is worth noting that in this embodiment, the reciprocal of the estimated variance is used as the fitting weight, and a pre-defined spatial surface model is used to perform surface fitting on the real-time deformation rate to obtain the surface deformation model. Specifically, the reciprocal of the estimated variance is determined as the observation weight of each grid cell when performing weighted fitting. The real-time deformation rate is used as the observation value, and a weighted normal equation system is constructed by combining the horizontal and vertical plane coordinates of each grid cell and the observation weight. This system is then substituted into the pre-defined spatial surface model for fitting and solving. Specifically, considering that the surface deformation caused by blasting has highly nonlinear characteristics and is prone to producing discontinuous abrupt terrain such as cracks and steep slopes in local areas, the pre-defined spatial surface model adopts a radial basis function surface fitting equation:

[0047] in, This represents the predicted deformation rate value of the grid cell. The planar coordinate vector representing the target mesh cell , This represents the planar coordinate vector of the i-th known observation grid cell, where N is the number of control points. The fitting weight coefficients are to be solved. To characterize spatially local nonlinear abrupt changes, the radial basis kernel function can be, for example, a Gaussian kernel function. To coordinate the low-order polynomial of the global trend, this embodiment specifically uses a first-order linear polynomial, namely... ,in , , Here, the inverse of the estimated variance is used as a regularization constraint weight and substituted into the surface fitting equation for matrix solving, thereby reconstructing the deformation state of each grid cell in the entire region and outputting the surface deformation model. By using the inverse of the estimated variance as a weight, the interpolation reliability is high, meaning that grid cells with small estimated variance dominate the surface reconstruction, while the reliability is low, meaning that grid cells with large estimated variance have a weaker constraint on the surface. This plays a role in filtering high-value information and reducing uncertainty errors at the data fusion level.

[0048] In step S4, the surface deformation model and the initial three-dimensional surface model are compared and aligned at nodes to obtain a three-dimensional deformation vector. The three-dimensional deformation vector is then mapped to a risk level based on a preset safety standard threshold to obtain a risk distribution map, including: The real-time three-dimensional spatial coordinates of each grid cell in the surface deformation model are subtracted from the initial three-dimensional spatial coordinates of the corresponding grid node in the initial three-dimensional surface model to obtain the position deviation vector of each grid cell, and the position deviation vector is determined as the three-dimensional deformation vector. The modulus difference is determined based on the Euclidean norm of the three-dimensional deformation vector, and the deviation angle is determined based on the relative spatial angle between the three-dimensional deformation vector and the preset reference axis. A static multi-level deformation range is established based on the preset standard displacement limit, and the multi-level deformation range is determined as the preset safety standard threshold. The modulus difference and deviation angle of each grid cell are compared with the safety standard threshold to obtain the corresponding risk level, and the risk level is mapped to the corresponding spatial coordinates to obtain a risk distribution map.

[0049] In one implementation, this embodiment subtracts the real-time three-dimensional spatial coordinates of each grid cell in the surface deformation model from the initial three-dimensional spatial coordinates of the corresponding grid cell in the initial surface three-dimensional model to obtain the position deviation vector of each grid cell. This position deviation vector is then determined as the three-dimensional deformation vector. Next, the Euclidean norm of the three-dimensional deformation vector is calculated to obtain the modulus difference representing the absolute displacement scalar. Simultaneously, the relative spatial angle between the three-dimensional deformation vector and a preset reference axis is calculated to obtain the deviation angle representing the deviation of the deformation displacement slip direction.

[0050] It should be noted that the preset reference axis can be specifically defined as the gravity axis vector perpendicular to the ground plane, i.e., the [0,0,1] vector in the three-dimensional coordinate system. The physical reason for choosing this specific direction as the reference axis is that the core cause of the instability and failure of railway subgrade slopes or cutting retaining walls is that after the blasting vibration destroys the soil skeleton, it undergoes downward vertical settlement or downward sliding under its own gravity. Setting the gravity axis as the comparison benchmark for the deviation angle can accurately quantify the degree of subsidence and collapse of the surface deformation in the vertical gravity direction, as well as the tilting and overturning trend relative to the vertical plane. At the geometric morphological level, it plays a role in monitoring the evolution of the slope toward the critical state of landslide.

[0051] It is worth noting that this embodiment establishes a static multi-level deformation range based on a preset standard displacement limit, and determines the multi-level deformation range as a preset safety standard threshold. It should be noted that the preset standard displacement limit can be specifically set as a multi-level static rigid range according to the safety standards in the national "Railway Subgrade Design Code" or "Technical Specification for Green Protection of Subgrade Slopes". In this embodiment, the first-level critical value is set to 5.0 mm, and the second-level critical value is set to 15.0 mm.

[0052] It should be noted that the engineering rationale for selecting these specific physical values ​​is that, in the railway subgrade and slope safety management system, 5.0 mm is a well-known hard safety threshold for determining the onset of irreversible plastic creep deformation in a structure, while 15.0 mm is the ultimate failure and instability threshold that triggers macroscopic collapse of the subgrade slope structure or fracture of the retaining wall. Using these specific standard limit values ​​as the static safety standard thresholds of the system can provide an objective and unified benchmark for system risk assessment. The system compares the modulus difference and deviation angle calculated by each grid cell with the static safety standard thresholds. For example, a modulus difference less than 5.0 mm is determined to be at a safe level, between 5.0 and 15.0 mm is determined to be at a medium risk level, and greater than 15.0 mm is determined to be at a high risk level. The corresponding risk level is directly obtained, and the risk level is mapped and assigned to the corresponding spatial grid coordinates. Finally, the risk distribution map, which presents the overall geometric risk characteristic spectrum of the entire survey area, is output.

[0053] In step S5, deformation time-series data of high-risk areas are extracted based on the risk distribution map. Dynamic features are extracted from the deformation time-series data to obtain deformation dynamic parameters. Evolutionary deduction is performed based on these parameters to obtain the instability anomaly evolution trend, including: Extract connected grid cells in the risk distribution map whose risk level exceeds the preset warning level to obtain high-risk areas; Extract the cumulative deformation displacement sequence of the high-risk area within a preset sliding time window to obtain deformation time series data; The deformation time series data is subjected to first-order difference to obtain the real-time deformation rate. The real-time deformation rate is subjected to second-order difference to obtain the real-time deformation acceleration. The real-time deformation rate and the real-time deformation acceleration are determined as deformation dynamics parameters. If the real-time deformation acceleration is continuously greater than zero and the real-time deformation rate exceeds the preset engineering instability critical threshold, then it is determined to be an unstable abnormal evolution trend.

[0054] In one implementation, this embodiment extracts risk levels exceeding preset warning levels from the risk distribution map. For example, the preset warning level is labeled as a connected grid cell of the aforementioned medium or high risk level, resulting in a high-risk area. For this high-risk area, the cumulative deformation displacement sequence within a preset sliding time window is extracted to obtain deformation time series data. It should be noted that the preset sliding time window can be specifically set to 60 minutes. The engineering rationale for choosing 60 minutes is that, in geological creep mechanics, the stress redistribution and deformation manifestation of rock and soil after blasting disturbance usually require a certain physical response time. A 60-minute time window can contain enough data samples to smooth out instantaneous sporadic measurement noise, and is short enough to ensure the keen capture of precursors to sudden landslides, playing a key control role between data stability and dynamic early warning timeliness.

[0055] In one implementation, this embodiment specifically performs a discrete-time first-order difference calculation on the deformation time-series data to obtain the real-time deformation rate, which characterizes the rate of displacement change. Subsequently, a second-order time-domain difference calculation is further performed on the real-time deformation rate to obtain the real-time deformation acceleration, which characterizes the rate change trend. In this embodiment, the real-time deformation rate and the real-time deformation acceleration are uniformly defined as deformation dynamics parameters.

[0056] In one implementation, the system uses the three-stage creep theory in geotechnical mechanics—decaying creep, constant-rate creep, and accelerated creep—to evolve and deduce the deformation dynamics parameters. Specifically, the judgment logic involves determining whether the real-time deformation acceleration is consistently greater than zero and whether the real-time deformation rate exceeds a preset engineering instability critical threshold. It should be noted that the preset engineering instability critical threshold can be specifically defined as 2.0 mm / h according to the "Code for Design of Railway Subgrade Retaining Structures". If the acceleration is consistently greater than zero, physically it indicates that the slope or subgrade has entered the accelerated creep stage, which is unable to self-stabilize—a precursor stage of macroscopic nonlinear instability failure. Simultaneously, a rate exceeding 2.0 mm / h effectively eliminates false alarms from minor measurement oscillations or random deformation spikes. If both of the above physical kinematic conditions are met, the system clearly determines and outputs the abnormal evolution trend of instability, indicating a precursor to structural collapse failure.

[0057] In step S6, a safety warning is distributed based on the unstable and abnormal evolution trend to obtain a warning report data stream, thereby realizing the safety monitoring of railway blasting points.

[0058] The process of distributing safety warnings based on the aforementioned unstable and abnormal evolution trend yields a warning report data stream, including: Extract the spatial coordinates of the high-risk areas where the aforementioned instability and abnormal evolution trend occurs to obtain the risk spatial coordinates; Based on the deformation dynamics parameters, a time recursion is performed to obtain the risk peak time window; By integrating the risk spatial coordinates, the risk peak time window, and the risk level corresponding to the risk distribution map, and encapsulating them according to a preset data format, early warning result data is obtained. The integrity verification calculation is performed on the warning result data to generate a data verification identifier, and the data verification identifier is added to the warning result data to obtain the warning report data stream.

[0059] In one implementation, once the system identifies an unstable anomaly evolution trend, it immediately extracts the three-dimensional grid index of the high-risk area where the unstable anomaly evolution trend occurs in geographic space to obtain the risk spatial coordinates. Subsequently, to accurately determine the time node of failure, this embodiment uses time recursion based on the nonlinear creep theory of soil and rock. In the accelerated creep stage of soil and rock, its deformation rate usually exhibits an exponential growth physical evolution characteristic with time, i.e., the deformation rate equation is:

[0060] in, The current real-time deformation rate, The creep acceleration index characterizes the degree of soil and rock degradation. By differentiating the rate equation in the time domain, the real-time deformation acceleration under the initial state can be obtained. Then, the creep acceleration index was calculated. To ensure that the recursive equations conform to real engineering boundary conditions, this embodiment calculates the difference between the preset collapse limit deformation (i.e., the aforementioned 15.0 mm ultimate failure instability threshold) and the current real-time cumulative deformation displacement of the high-risk area, and determines this difference as the remaining safe displacement margin. The deformation rate equation is integrated over time and set equal to the remaining safety displacement margin. Construct the exponential creep evolution equation:

[0061] The calculated Substitute into the above equation and solve for the time root in reverse. :

[0062] The system uses the above formula to calculate the countdown time required for the evolution from the current state to the gauge violation limit. The countdown time obtained from the solution is added to the current timestamp, and the peak risk time window when the surface deformation is expected to reach the standard failure limit is recursively calculated. Through this time recursion based on the nonlinear exponential creep law, a high-precision quantitative response time reference can be provided for on-site engineering emergency rescue.

[0063] It is worth noting that this embodiment integrates the risk spatial coordinates, the risk peak time window, and the risk level corresponding to the risk distribution map, and encapsulates them in a structured manner according to a preset data format, such as the standard lightweight JSON or XML data structure format commonly used between systems, to obtain the early warning result data. The control function of using a preset data format is to ensure that the generated early warning information can be seamlessly integrated into the various information-based command screens or automated train dispatching systems of the railway bureau, ensuring real-time information exchange across multiple terminals and platforms.

[0064] Finally, to ensure the tamper-proof and non-repudiable nature of the early warning instructions during network transmission and archiving, this embodiment performs integrity verification calculations on the early warning result data. A hash digest algorithm is used to process the entire early warning result data structure block, generating a unique corresponding data verification identifier, i.e., a digital fingerprint. The system directly embeds or appends this data verification identifier to a designated verification field of the early warning result data, ultimately outputting a standardized early warning report data stream. This mechanism ensures that the alarm data possesses irrefutable original evidentiary legal effect in subsequent engineering accident tracing and legal liability determination. Thus, accurate monitoring and proactive safety prevention of deformation states at railway blasting sites are achieved.

[0065] In summary, this invention constructs a continuous three-dimensional surface deformation model by spatially interpolating and reconstructing the surface feature matrix and displacement time series acquired by sensors. Furthermore, through rigorous comparison of the three-dimensional deformation vector with safety standard thresholds, it accurately maps the risk distribution map in macroscopic space. By combining the first and second-order dynamic difference extrapolation of deformation displacement in the time domain, it achieves advanced identification of macroscopic nonlinear creep instability trends and precise recursion of emergency response time windows. This invention realizes proactive forward-looking prediction and precise dynamic perception of surface deterioration trends at complex railway blasting sites, breaking down the information barrier between microscopic vibrations and macroscopic deformation, and providing objective and quantitative data and decision support for engineering disaster prevention and mitigation and railway safety.

[0066] The second embodiment of the present invention provides a railway blasting point safety monitoring system based on a vibration monitoring instrument, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0067] It should be noted that the railway blasting point safety monitoring system based on a vibration monitor provided in this embodiment of the invention is used to execute all the process steps of the railway blasting point safety monitoring method based on a vibration monitor in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0068] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for safety monitoring of railway blasting sites based on a vibration monitoring instrument, characterized in that, include: An initial three-dimensional model of the surface of the railway blasting area is obtained, the initial three-dimensional model of the surface is divided into multiple grid units, and the surface geometric features of each grid unit are extracted to obtain the surface feature matrix. The original vibration signal collected by the sensor is acquired, and the original vibration signal is subjected to integral smoothing transformation to obtain the displacement time sequence; The displacement time series is weighted and mapped to the surface feature matrix to obtain discrete correlation features. Spatial interpolation is performed on the discrete correlation features to obtain the correlation feature matrix. Deformation reconstruction is performed on the correlation feature matrix to obtain the surface deformation model. The surface deformation model is compared with the initial three-dimensional surface model by node alignment to obtain a three-dimensional deformation vector. The three-dimensional deformation vector is then mapped to a risk level according to a preset safety standard threshold to obtain a risk distribution map. Deformation time series data of high-risk areas are extracted from the risk distribution map. Dynamic features are extracted from the deformation time series data to obtain deformation dynamic parameters. Evolution is deduced based on the deformation dynamic parameters to obtain the instability anomaly evolution trend. Based on the aforementioned unstable and abnormal evolution trend, safety early warnings are distributed to obtain early warning report data streams, thereby enabling safety monitoring of railway blasting sites.

2. The method for safety monitoring of railway blasting sites based on a vibration monitoring instrument according to claim 1, characterized in that, The acquisition of the initial three-dimensional surface model of the railway blasting area includes: UAV lidar point cloud data of the railway blasting area is acquired, and the lidar point cloud data is filtered and resampled to obtain a digital elevation model. The digital elevation model is then used as the initial three-dimensional surface model.

3. The method for safety monitoring of railway blasting points based on a vibration monitoring instrument according to claim 1, characterized in that, The step of extracting surface geometric features from each grid cell to obtain a surface feature matrix includes: Calculate the surface normal vector of each grid cell in the initial three-dimensional surface model; Calculate the slope and aspect characteristics of each grid cell based on the surface normal vector; The slope features and aspect features are arranged in a matrix according to the spatial index corresponding to each grid cell to obtain the surface feature matrix.

4. The method for safety monitoring of railway blasting points based on a vibration monitoring instrument according to claim 1, characterized in that, The integral smoothing transformation of the original vibration signal to obtain the displacement time sequence includes: The original vibration signal is bandpass filtered to obtain a filtered vibration signal. The filtered vibration signal is then integrated twice to obtain an integrated vibration signal. The integrated vibration signal is then baseline-corrected to obtain initial displacement data. The initial displacement data were analyzed in the frequency domain using the Fast Fourier Transform algorithm to obtain the dominant frequency components of the blasting vibration. The initial displacement data is denoised by low-pass filtering based on the dominant frequency component of the blasting vibration to obtain the displacement time sequence.

5. The method for safety monitoring of railway blasting sites based on a vibration monitoring instrument according to claim 1, characterized in that, The process of spatial interpolating the discrete correlation features to obtain a correlation feature matrix, and then performing deformation reconstruction on the correlation feature matrix to obtain a surface deformation model includes: A spatial variability function is constructed based on the spatial distribution differences of the discrete correlation features, and a Kriging estimation equation system is constructed based on the spatial variability function. The interpolation weights and estimation variances of each grid cell in the surface feature matrix are solved using the Kriging estimation equations. The discrete correlation features are then weighted and summed according to the interpolation weights to obtain the correlation feature matrix. Extract the feature value change sequence of each grid cell in the correlation feature matrix within a continuous time step, perform trend fitting on the feature value change sequence, and obtain the real-time deformation rate of each grid cell. Using the reciprocal of the estimated variance as the fitting weight, the real-time deformation rate is fitted using a preset spatial surface model to obtain the surface deformation model.

6. The method for safety monitoring of railway blasting points based on a vibration monitoring instrument according to claim 1, characterized in that, The process involves comparing the nodes of the surface deformation model with the initial 3D surface model to obtain a 3D deformation vector. Then, based on a preset safety standard threshold, the 3D deformation vector is mapped to a risk level to obtain a risk distribution map, including: The real-time three-dimensional spatial coordinates of each grid cell in the surface deformation model are subtracted from the initial three-dimensional spatial coordinates of the corresponding grid node in the initial three-dimensional surface model to obtain the position deviation vector of each grid cell, and the position deviation vector is determined as the three-dimensional deformation vector. The modulus difference is determined based on the Euclidean norm of the three-dimensional deformation vector, and the deviation angle is determined based on the relative spatial angle between the three-dimensional deformation vector and the preset reference axis. A static multi-level deformation range is established based on the preset standard displacement limit, and the multi-level deformation range is determined as the preset safety standard threshold. The modulus difference and deviation angle of each grid cell are compared with the safety standard threshold to obtain the corresponding risk level, and the risk level is mapped to the corresponding spatial coordinates to obtain a risk distribution map.

7. The method for safety monitoring of railway blasting sites based on a vibration monitoring instrument according to claim 1, characterized in that, The process involves extracting deformation time-series data of high-risk areas from the risk distribution map, extracting dynamic features from the deformation time-series data to obtain deformation dynamic parameters, and performing evolutionary deduction based on the deformation dynamic parameters to obtain the instability anomaly evolution trend, including: Extract connected grid cells in the risk distribution map whose risk level exceeds the preset warning level to obtain high-risk areas; Extract the cumulative deformation displacement sequence of the high-risk area within a preset sliding time window to obtain deformation time series data; The deformation time series data is subjected to first-order difference to obtain the real-time deformation rate. The real-time deformation rate is subjected to second-order difference to obtain the real-time deformation acceleration. The real-time deformation rate and the real-time deformation acceleration are determined as deformation dynamics parameters. If the real-time deformation acceleration is continuously greater than zero and the real-time deformation rate exceeds the preset engineering instability critical threshold, then it is determined to be an unstable abnormal evolution trend.

8. The method for safety monitoring of railway blasting points based on a vibration monitoring instrument according to claim 1, characterized in that, The step of distributing safety warnings based on the instability and abnormal evolution trend to obtain a warning report data stream includes: Extract the spatial coordinates of the high-risk areas where the aforementioned instability and abnormal evolution trend occurs to obtain the risk spatial coordinates; Based on the deformation dynamics parameters, a time recursion is performed to obtain the risk peak time window; By integrating the risk spatial coordinates, the risk peak time window, and the risk level corresponding to the risk distribution map, and encapsulating them according to a preset data format, early warning result data is obtained. The integrity verification calculation is performed on the warning result data to generate a data verification identifier, and the data verification identifier is added to the warning result data to obtain the warning report data stream.

9. A railway blasting point safety monitoring system based on a vibration monitor, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the program implements the method described in any one of claims 1 to 8.

Images