Engineering geologic body multi-parameter coupling risk assessment method

Abstract

The invention discloses an engineering geologic body multi-parameter coupling risk assessment method, belongs to the technical field of geological safety monitoring, and can solve the problem that an existing geologic body risk assessment method cannot dynamically and automatically update water conservancy and hydropower engineering geologic body risk assessment along with multi-source real-time monitoring data and cannot predict a future risk trend. The method comprises the following steps: S1, generating initial risk distribution according to initial geological data of a target engineering area and a preset coupling rule; s2, determining a target parameter prediction value and a risk trend prediction value within a preset time length according to the real-time monitoring data flow, the geological updating data and the initial risk distribution; s3, generating a dynamic risk threshold according to the target parameter prediction value and the risk trend prediction value in combination with a preset engineering safety constraint condition; and S4, comparing the current measurement value of the real-time monitoring data flow with the dynamic risk threshold value, and determining the dynamic risk level of the target engineering area. The method is used for risk assessment of the water conservancy and hydropower engineering geologic body.

Description

Technical Field

[0001] The invention relates to a multi-parameter coupling risk assessment method for an engineering geological body, and belongs to the technical field of geological safety monitoring. Background Art

[0002] As the transition to clean energy accelerates, the scale and complexity of water conservancy and hydropower projects, as a core form of large-scale energy storage, are increasing day by day. Project safety is highly dependent on accurate risk assessments of the stability of the engineering geological bodies in key areas such as underground powerhouses and water transmission systems. However, the long construction cycles of such projects and the frequent switching of operating conditions result in dynamic changes in the loads, stresses, and seepage environments borne by the geological bodies. Geological conditions often change or encounter "unknown tunnel sections" during construction. This makes it difficult for static risk assessments based on limited data from the survey and design phase to reflect the real-time evolution of the geological body's status and risks during project advancement and operation. This leads to key technical bottlenecks such as delayed risk warnings and decreased accuracy as the project progresses.

[0003] Existing geological risk assessment techniques typically rely on numerical simulation combined with deterministic or probabilistic analysis frameworks (such as the Monte Carlo method), or comprehensive assessment models based on expert scoring and hierarchical analysis. Some methods attempt to integrate selected monitoring data (such as displacement, stress, and seepage pressure) from the construction or operation period to update the model. However, these updates often rely on manual intervention by engineers to periodically recalculate or adjust parameters, a cumbersome and time-consuming process with slow response. Alternatively, they rely solely on fixed thresholds to determine whether the current state exceeds a limit, lacking the ability to predict future risk trends based on data streams or adaptive dynamic early warning mechanisms. Existing methods fail to fully utilize the streaming nature of real-time monitoring data, making it impossible to establish an automated, real-time / near-real-time risk assessment dynamic update closed loop. Summary of the Invention

[0004] The present invention provides a multi-parameter coupling risk assessment method for engineering geological bodies, which can solve the problem that existing geological body risk assessment methods cannot achieve dynamic and automatic update of risk assessment of water conservancy and hydropower engineering geological bodies along with multi-source real-time monitoring data and predict future risk trends.

[0005] The present invention provides a multi-parameter coupling risk assessment method for engineering geological bodies, the method comprising:

[0006] S1. generating an initial risk distribution of the target engineering area based on initial geological data of the target engineering area and preset coupling rules;

[0007] S2. Acquire real-time monitoring data streams and geological update data of the target project area, and determine target parameter prediction values ​​and risk trend prediction values ​​within a preset time period based on the real-time monitoring data streams, the geological update data, and the initial risk distribution;

[0008] S3. Generate a dynamic risk threshold based on the target parameter prediction value and the risk trend prediction value, in combination with preset engineering safety constraints;

[0009] S4. Compare the current measurement value of the real-time monitoring data stream with the dynamic risk threshold to determine the dynamic risk level of the target engineering area.

[0010] Optionally, the S1 specifically includes:

[0011] Construct a three-dimensional geological numerical model based on the initial geological data of the target project area;

[0012] Calculating the stability quantitative index of the geological body of each grid unit in the three-dimensional geological numerical model according to a preset coupling rule;

[0013] An initial risk distribution of the target project area is generated based on the spatial distribution of the stability quantitative index.

[0014] Optionally, the step of obtaining the real-time monitoring data stream and geological update data of the target engineering area in S2 specifically includes:

[0015] Continuously collecting monitoring data of surrounding rock displacement, anchor stress and pore water pressure in the target engineering area through a distributed sensor network to form a real-time monitoring data stream;

[0016] Obtain geological radar re-survey data of the fault zone in the target engineering area as geological update data.

[0017] Optionally, determining the target parameter prediction value and the risk trend prediction value within a preset time period according to the real-time monitoring data stream, the geological update data, and the initial risk distribution in S2 specifically includes:

[0018] Using the initial risk distribution to train a long short-term memory network to obtain a deep time series prediction model, and determining a three-dimensional input tensor based on the real-time monitoring data stream, the geological update data and the initial risk distribution;

[0019] The three-dimensional input tensor is input into the deep time series prediction model to obtain the target parameter prediction value and risk trend prediction value within a preset time length.

[0020] Optionally, determining a three-dimensional input tensor according to the real-time monitoring data stream, the geological update data, and the initial risk distribution specifically includes:

[0021] Performing spatial registration and fusion on the geological update data and the initial risk distribution to generate spatial enhancement features;

[0022] The real-time monitoring data stream is subjected to sliding window normalization processing, and the processed real-time monitoring data stream is spliced ​​with the spatial enhancement feature in the time dimension to obtain a three-dimensional input tensor.

[0023] Optionally, the S3 specifically includes:

[0024] Calculating the fluctuation range of the predicted value of the target parameter at a preset confidence level;

[0025] Adjusting the tolerance coefficient of the fluctuation range according to the change slope of the risk trend prediction value;

[0026] Generate dynamic risk thresholds based on the adjusted tolerance factor and preset engineering safety constraints.

[0027] Optionally, the S4 specifically includes:

[0028] Calculating an absolute deviation value and a relative change rate between a current measurement value of the real-time monitoring data stream and the dynamic risk threshold;

[0029] The dynamic risk level of the target engineering area is determined according to the absolute deviation value and the relative change rate.

[0030] Optionally, determining the dynamic risk level of the target engineering area according to the absolute deviation value and the relative change rate specifically includes:

[0031] When the absolute deviation value exceeds the first-level deviation threshold or the relative change rate exceeds the first-level change rate threshold, it is determined to be a high risk level;

[0032] When the absolute deviation value does not exceed the secondary deviation threshold and the relative change rate does not exceed the secondary change rate threshold, it is determined to be a low risk level; otherwise, it is determined to be a medium risk level.

[0033] Optionally, after S4, the method further includes:

[0034] S5. Generate a multi-level warning signal according to the dynamic risk level of the target engineering area, and send the multi-level warning signal to the management and control platform.

[0035] Optionally, the preset confidence level is 90% to 95%.

[0036] The beneficial effects that the present invention can produce include:

[0037] The multi-parameter coupling risk assessment method for engineering geological bodies provided by the present invention determines the target parameter prediction values ​​and risk trend prediction values ​​within a preset time period in the future through the real-time monitoring data stream, geological update data and initial risk distribution of the target engineering area, realizes the fusion prediction of static and dynamic data, and improves the displacement prediction error.

[0038] The multi-parameter coupled risk assessment method for engineering geological bodies provided by the present invention first generates an initial risk distribution based on initial geological data and preset coupling rules. Then, the real-time monitoring data stream and geological update data of the water conservancy and hydropower engineering area are obtained, and the data and the initial risk distribution are input into a pre-trained deep time series prediction model to output the target parameter prediction value and risk trend prediction value within a specified time period in the future. The target parameter prediction value is associated with the preset monitoring items that affect the stability of the geological body. Combined with the preset engineering safety constraints, a dynamic risk threshold is generated based on the predicted value. The current measurement value of the real-time monitoring data stream is then compared with the threshold to determine the dynamic risk level and generate a corresponding multi-level warning signal output. This method solves the existing technical problem of being unable to realize the dynamic and automatic update of water conservancy and hydropower engineering geological body risk assessment along with multi-source real-time monitoring data and predict future risk trends, thereby realizing the dynamic tracking and early prediction of the coupled risk status of water conservancy and hydropower engineering geological bodies based on real-time monitoring data, significantly improving the timeliness and initiative of risk warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Flowchart of a multi-parameter coupling risk assessment method for engineering geological bodies provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention is described in detail below with reference to the embodiments, but the present invention is not limited to these embodiments.

[0041] The embodiment of the present invention provides a multi-parameter coupling risk assessment method for engineering geological bodies, such as Figure 1 As shown, the method includes:

[0042] S1. Generate an initial risk distribution of the target engineering area based on the initial geological data of the target engineering area and preset coupling rules.

[0043] Specifically, it includes: first, constructing a three-dimensional geological numerical model based on the initial geological data of the target engineering area; then calculating the stability quantitative indicators of the geological bodies of each grid unit in the three-dimensional geological numerical model according to the preset coupling rules; finally, generating the initial risk distribution of the target engineering area based on the spatial distribution of the stability quantitative indicators.

[0044] The specific implementation steps include:

[0045] 1. Establish a three-dimensional geological numerical model.

[0046] 1.1 Data input.

[0047] Import the initial geological data into the geological modeling software (using GOCAD or Midas GTS NX). The initial geological data includes:

[0048] Drill core data (lithology, RQD value, permeability coefficient);

[0049] Geological cross section (scale 1:500);

[0050] Geophysical exploration data (seismic velocity VP / VS ratio);

[0051] Ground stress measurement results (maximum principal stress azimuth and magnitude).

[0052] 1.2 Model construction.

[0053] 1.2.1 Use the inverse distance weighting method to perform spatial interpolation on discrete borehole data. The formula is: ;

[0054] in: is the attribute value of the grid point to be interpolated; For the The attribute value of a known point; is the Euclidean distance between the known point and the point to be interpolated; is the power parameter (the default value is 2).

[0055] 1.2.2 Divide the hexahedral mesh into units in the modeling software. The unit size is set according to the project partition as follows: Underground powerhouse area: grid side length ≤ 5 m; Water diversion tunnel area: grid side length ≤ 10 m; Reservoir basin area: grid side length ≤ 20 m.

[0056] 1.3 Model verification.

[0057] The model error was calculated using the cross-validation method: 10% of the drilling data was randomly retained as the validation set, and the root mean square error (RMSE) between the interpolation results and the measured values ​​was calculated, with the RMSE required to be ≤ 15%.

[0058] 2. Calculate quantitative indicators of the stability of geological bodies.

[0059] 2.1 Application of coupling rules: Execute the preset coupling rules (based on the Hoek-Brown criterion and the seepage-stress coupling equation) and perform iterative calculations on each grid cell.

[0060] 2.1.1 Calculation of rock mass strength parameters (using the Hoek-Brown criterion): ;

[0061] In the above formula, is the peak strength of the rock mass; is confining pressure; is the uniaxial compressive strength of the intact rock block (laboratory measurement value); , , All are functions of the geological strength index (GSI), determined by the characteristics of the rock mass structural surface.

[0062] 2.1.2 Seepage-stress coupling calculation (using Biot theory): ; in, is the rock mass permeability tensor (obtained from the field water pressure test); is the pore water pressure; is the water storage coefficient; is the Biot coefficient (range 0.5~0.9); is the rock mass displacement vector.

[0063] 2.2 Calculation of quantitative stability indicators.

[0064] Output three indicators for each grid cell: 2.2.1 Safety factor = shear strength / shear stress (using the Mohr-Coulomb criterion); 2.2.2 Osmotic stability index = critical hydraulic gradient / actual hydraulic gradient; 2.2.3 Deformation energy density (Unit: kJ / m³); where, Indicates that the action on the normal is Directional surface Directional stress components, express Direction and The strain components between the directions.

[0065] 3. Generate an initial risk distribution.

[0066] 3.1 Risk indicator integration.

[0067] The comprehensive risk index is synthesized by linear weighting :

[0068] ; Weight coefficient assignment (based on engineering expert experience), where: =0.6 (structural stability weight); = 0.3 (osmotic stability weight); =0.1 (deformation energy weight).

[0069] 3.2 Spatial distribution rendering.

[0070] In the Paraview visualization platform, the comprehensive risk index of the grid unit is Mapping to RGB color values: Low risk ( ≤0.3): Green (RGB: 0,255,0); Medium risk (0.3< ≤0.6): Yellow (RGB: 255,255,0); High risk ( >0.6): Red (RGB: 255,0,0).

[0071] 3.3 Generate initial risk distribution map: Output the raster risk distribution map in GeoTIFF format, including:

[0072] Space coordinate system (using CGCS2000 coordinate system);

[0073] Metadata: Risk level percentage statistics (e.g., high-risk grids account for 15.7%).

[0074] This step converts initial geological data into a structured three-dimensional geological numerical model. Applying the Hoek-Brown criterion and the Biot seepage-stress coupling equation, the model quantitatively calculates rock mass stability indicators (safety factor, seepage stability index, and deformation energy density) for each grid cell. Finally, a weighted fusion process generates a visual initial risk distribution map. This process quantitatively expresses engineering risks from geological exploration data and establishes a spatialized risk baseline distribution model, providing a static geological background for subsequent real-time risk assessment. This overcomes the inaccuracy of traditional methods that rely on manual interpretation, improving initial risk identification accuracy by over 40%.

[0075] S2. Obtain the real-time monitoring data stream and geological update data of the target project area, and determine the target parameter prediction value and risk trend prediction value within a preset time period based on the real-time monitoring data stream, geological update data and initial risk distribution.

[0076] The acquisition of real-time monitoring data stream and geological update data of the target engineering area in the above S2 specifically includes: continuously collecting monitoring data of surrounding rock displacement, anchor stress and pore water pressure in the target engineering area through a distributed sensor network to form a real-time monitoring data stream; obtaining re-survey data of the fault zone in the target engineering area by geological radar as geological update data.

[0077] The implementation specifically includes the following steps:

[0078] 1. Collect real-time monitoring data streams through distributed sensor networks.

[0079] 1.1 Sensor deployment and parameter configuration 1.1.1 Surrounding rock displacement monitoring: Multi-point displacement meters are installed in the water conservancy and hydropower project area (including underground powerhouses and key sections of water transmission tunnels). The measurement directions include radial displacement ( ) and tangential displacement ( ).

[0080] Sensor type: vibrating wire displacement meter;

[0081] Sampling frequency: 0.1 Hz (1 sample every 10 seconds).

[0082] 1.1.2 Anchor stress monitoring: Install a steel bar stress meter on the support anchor to measure the axial stress value ( , unit: MPa).

[0083] Sensor type: vibrating wire rebar meter; Measuring range: 0–300 MPa.

[0084] 1.1.3 Pore water pressure monitoring: Piezometers are placed along the seepage path of the surrounding rock to measure the pore water pressure ( , unit: kPa).

[0085] Sensor type: Piezoresistive osmometer; Accuracy: ±0.1% FS.

[0086] 1.2 Real-time data stream processing.

[0087] 1.2.1 Data transmission protocol: The original voltage signal is converted into physical quantity through the industrial bus (Modbus RTU). The conversion formula is as follows: .

[0088] 1.2.2 Outlier processing.

[0089] Use the Hampel filter algorithm to remove outliers: (1) Calculate the median of the data in the sliding window ; (2) Calculate the absolute deviation ,in, For the window Real-time monitoring of raw data points; (3) If If the value is greater than 3×MAD (MAD is the median absolute deviation), it is considered an outlier (window size = 100 data points).

[0090] 2. Receive geological radar fault zone re-survey data.

[0091] 2.1 Retest data collection.

[0092] 2.1.1 Equipment and parameters.

[0093] Geological radar model: Ground penetrating radar (GPR) system with a center frequency of 100 MHz;

[0094] Survey line layout: parallel survey lines are laid out along the fault zone with a line spacing of ≤20 m;

[0095] Data format: Segy seismic data format (including time-depth profiles).

[0096] 2.2 Fault feature extraction.

[0097] 2.2.1 Data preprocessing

[0098] Bandpass filtering (50–150 MHz) to remove high-frequency noise;

[0099] Gain adjustment (AGC gain factor = 0.003).

[0100] 2.2.2 Speed ​​analysis.

[0101] The common center point velocity analysis method (CMP) is used to calculate the electromagnetic wave velocity: .

[0102] 2.2.3 Reverse time migration imaging.

[0103] The wave equation is solved using the finite difference method to generate the fault space coordinates ( ).

[0104] 2.3 Geological update data generation.

[0105] 2.3.1 Fault zone spatial model.

[0106] Output DXF file containing the following attributes: Fault attitude (dip ∠ inclination); fracture zone width (unit: m); filling material permeability coefficient (unit: m / s).

[0107] This step automatically collects real-time monitoring data streams of surrounding rock displacement, anchor stress, and pore water pressure at a 0.1 Hz frequency through a distributed sensor network. Outliers are cleaned in real time using the Hampel filter algorithm. Simultaneously, re-measured data from fault zones collected by geological radar are received and processed using bandpass filtering, CMP velocity analysis, and finite-difference reverse-time migration imaging to extract the three-dimensional spatial parameters of the faults. This process addresses the low frequency and high lag issues inherent in traditional manually collected data, providing high-quality real-time monitoring data with a frequency of ≥0.1 Hz and stratigraphic update data with meter-level accuracy for the deep time series prediction model. This improves the temporal and spatial resolution of the input data for subsequent prediction models by 90%, laying a reliable data foundation for dynamic risk assessment.

[0108] The above S2 determines the target parameter prediction value and risk trend prediction value within a preset time period based on the real-time monitoring data stream, geological update data and initial risk distribution, specifically including: (1) Using the initial risk distribution to train a long short-term memory network, a deep time series prediction model is obtained, and a three-dimensional input tensor is determined based on the real-time monitoring data stream, geological update data, and the initial risk distribution; Specifically, the geological update data and the initial risk distribution can be spatially aligned and fused to generate spatial enhancement features; the real-time monitoring data stream can then be normalized using a sliding window, and the processed real-time monitoring data stream and the spatial enhancement features can be spliced ​​in the time dimension to obtain a three-dimensional input tensor.

[0109] (2) Input the three-dimensional input tensor into the deep time series prediction model to obtain the target parameter prediction value and risk trend prediction value within the preset time length.

[0110] Among them, the target parameter prediction value is used to indicate the preset monitoring items that affect the stability of the target engineering geological body.

[0111] The implementation specifically includes the following steps: 1. Spatial registration and fusion generate spatial enhancement features.

[0112] 1.1 Standardization of spatial coordinate system.

[0113] 1.1.1 Obtain the initial risk distribution (format: GeoTIFF, coordinate system: CGCS2000) and geological update data (format: DXF, including fault zone spatial coordinates) obtained above.

[0114] 1.1.2 Perform affine transformation to unify the coordinate system on the ArcGIS platform. The transformation formula is: ;

[0115] in,( ) is the original coordinate point; ( , ) is the coordinate point after transformation; is the rotation angle (calculated based on the control points); , is the translation amount (the difference in the measured control point coordinates).

[0116] 1.2 Grid feature fusion.

[0117] 1.2.1 Perform the following operations for each spatial grid cell.

[0118] Extract the comprehensive risk index of the grid in the initial risk distribution .

[0119] Extract the fault zone parameters covering the grid: fault width , permeability coefficient , inclination angle ,inclination .

[0120] Generate 5-channel spatial feature vector: [ ].

[0121] 1.2.2 Output size is The spatial enhancement feature matrix ( is the total number of grids).

[0122] 2. Real-time monitoring of data stream timing processing.

[0123] 2.1 Sliding window normalization processing.

[0124] 2.1.1 Window settings.

[0125] Window length: 600 time points (covering 100 minutes, since the sensor sampling rate is 0.1 Hz, i.e., 1 every 10 seconds).

[0126] Slide step: 60 time points (one move every 10 minutes).

[0127] 2.1.2 Normalization method.

[0128] For each monitoring parameter (displacement, stress, seepage pressure) in the window, execute: ;

[0129] in, is the original parameter value; is the parameter mean within the window; is the standard deviation of the parameters within the window; To prevent division by zero, take a small constant ( ).

[0130] 2.2 Three-dimensional input tensor construction.

[0131] 2.2.1 The spatial enhancement feature matrix (size ) Copy in the time dimension Second-rate( =600), generating Tensor .

[0132] 2.2.2 The normalized real-time monitoring data stream (size ×3, 3 corresponds to displacement / stress / osmotic pressure) is expanded to Tensor (Through spatial broadcasting, that is, copying data to all grids at the same time).

[0133] 2.2.3 Concatenating Tensors Along the Channel Dimension and , generating a three-dimensional input tensor .

[0134] 3. Pre-trained deep time series prediction model.

[0135] 3.1 Model Pre-training

[0136] 3.1.1 Input data preparation: The initial risk distribution obtained above is compressed into a 128-dimensional feature vector through a fully connected layer .

[0137] 3.1.2 Pre-training task: Construct an autoencoder to pre-train the initial weights of the long short-term memory network (LSTM).

[0138] Encoder: Input historical risk series (length 1000), output 128-dimensional state vector through two layers of LSTM .

[0139] Fusion layer: and Splicing, generating fusion features through the fully connected layer .

[0140] Decoder: Input , risk sequence reconstructed by single-layer LSTM .

[0141] Loss function: sequence reconstruction mean squared error (MSE).

[0142] 3.2 Prediction model architecture.

[0143] 3.2.1 Input layer: three-dimensional input tensor (size ).

[0144] 3.2.2 Spatiotemporal feature extraction module.

[0145] (1) Spatial convolution layer: using 3×3 convolution kernel Convolution of spatial dimensions (channels: 64, stride: 1, padding: SAME).

[0146] Output feature map size: .

[0147] (2) Time series processing layer.

[0148] Long Short-Term Memory Network (LSTM) Structure:

[0149] First layer: 128 neurons, time steps =600.

[0150] The second layer: 64 neuron units, receiving the hidden state output of the first layer.

[0151] Activation function: hyperbolic tangent function (tanh).

[0152] Output: 64-dimensional time feature vector (Last time step state).

[0153] 3.2.3 Prediction output layer.

[0154] Fully connected layer 1: 64 neurons (ReLU activation), input .

[0155] Fully connected layer 2: 4 neurons (linear activation), outputting 4 predictions: : predicted radial displacement value (unit: mm); : predicted value of anchor stress (unit: MPa); : predicted pore water pressure (unit: kPa); : Risk trend forecast value (risk index change rate in the next 6 hours).

[0156] This step unifies the spatial coordinates (rotation angles) by affine transformation and translation 、 Determine the grid correspondence), fuse the initial risk distribution and fault parameters into a spatial enhancement feature matrix; use the sliding window standard deviation normalization process on the real-time monitoring data stream to construct the time-space joint dimension The system uses a pre-trained deep time series prediction model (a two-layer LSTM architecture: 128 neurons in the first layer → 64 neurons in the second layer, with tanh activation) to extract spatiotemporal features and output predicted values ​​for key monitoring parameters (i.e., target parameters such as displacement, stress, and seepage pressure) and the risk change rate for the next six hours. This solves the challenge of coupled modeling of multidimensional geological spatiotemporal data, limiting the six-hour displacement prediction error to within ±1.2 mm (measured standard deviation), and increasing the accuracy of risk trend prediction to 91%.

[0157] S3. Generate a dynamic risk threshold based on the target parameter prediction value and risk trend prediction value, combined with the preset engineering safety constraints.

[0158] Specifically, it includes: first calculating the fluctuation range of the target parameter prediction value under the preset confidence level; then adjusting the tolerance coefficient of the fluctuation range according to the changing slope of the risk trend prediction value; finally, generating a dynamic risk threshold based on the adjusted tolerance coefficient and the preset engineering safety constraints.

[0159] The preset confidence level is 90% to 95%. In actual applications, the preset confidence level can be set to 95%.

[0160] The implementation specifically includes the following steps:

[0161] 1. Calculate the fluctuation range of the predicted value of the target parameter.

[0162] 1.1 Prediction value distribution fitting.

[0163] Get the target parameter prediction value output in S2 (surrounding rock displacement , anchor stress , pore water pressure ).

[0164] For each parameter’s predicted value sequence within the next 6 hours (72 time points, 10 minutes apart), 1000 sets of samples were generated using the bootstrap method.

[0165] Calculate the 95% confidence interval for each time point: ;

[0166] in, for The sample mean of the predicted value at the moment; for The sample standard deviation of the moment-by-moment forecast value; is the critical value of the 95% confidence interval of the standard normal distribution, which is fixed at 1.96.

[0167] 1.2 Output fluctuation range.

[0168] Displacement fluctuation range: ;

[0169] in: is the displacement of surrounding rock at time point The mean of the predicted values; is the displacement of surrounding rock at time point The standard deviation of the predicted values.

[0170] Stress fluctuation range: ;

[0171] in, is the anchor stress at time point The mean of the predicted values; is the anchor stress at time point The standard deviation of the predicted values.

[0172] Osmotic pressure fluctuation range: ;

[0173] in, is the pore water pressure at time The mean of the predicted values; is the pore water pressure at time The standard deviation of the predicted values.

[0174] 2. Adjust tolerance factors based on risk trends.

[0175] 2.1 Calculation of risk trend slope.

[0176] Risk trend prediction value output from S2 (Unit: % / h), calculate its change slope: ;

[0177] in, For time point The risk trend prediction value; is the first time point (the current moment); is the last time point (6 hours later); It is the average rate of change of risk trend within 6 hours (unit: % / h²).

[0178] 2.2 Dynamically adjust the tolerance coefficient.

[0179] Initial tolerance factor: =3.0 (default).

[0180] Adjustment rules: ;

[0181] Boundary constraints: If <1.5, then take 1.5; if If >5.0, take 5.0.

[0182] 3. Generate dynamic risk thresholds.

[0183] 3.1 Setting of engineering safety constraint values.

[0184] According to the "Design Code for Water Conservancy and Hydropower Stations" NB / T 10072-2018:

[0185] Surrounding rock displacement constraint value : The maximum allowable displacement of Class III surrounding rock (Article 7.3.1 of the standard) is 20 mm.

[0186] Anchor stress constraint value : The 80% yield strength of HRB400 steel bar (Standard 9.2.3) is 0.8×400=320 MPa.

[0187] Pore ​​water pressure constraint value : The critical water pressure of shale formation (Standard Appendix C) is 800 kPa.

[0188] 3.2 Threshold calculation formula.

[0189] Displacement risk threshold (symbol: ): ;

[0190] Meaning: The smaller of the "Forecast Fluctuation Upper Limit" and the "Specification Constraint Value".

[0191] Stress risk threshold (symbol: ): .

[0192] Osmotic pressure risk threshold (symbol: ): .

[0193] This step uses the self-service sampling method to construct the 95% confidence interval of the target monitoring parameter, based on the slope of the risk trend. Dynamically adjust tolerance factor ,Finally, a dynamic risk threshold is generated by combining the engineering constraint values ​​specified in national standards.

[0194] S4. Compare the current measurement value of the real-time monitoring data stream with the dynamic risk threshold to determine the dynamic risk level of the target project area.

[0195] Specifically include:

[0196] (1) Calculate the absolute deviation and relative rate of change between the current measurement value of the real-time monitoring data stream and the dynamic risk threshold.

[0197] (2) Determine the dynamic risk level of the target project area based on the absolute deviation value and relative change rate.

[0198] Specifically: when the absolute deviation value exceeds the first-level deviation threshold or the relative change rate exceeds the first-level change rate threshold, it is judged as a high-risk level; when the absolute deviation value does not exceed the second-level deviation threshold and the relative change rate does not exceed the second-level change rate threshold, it is judged as a low-risk level; otherwise, it is judged as a medium-risk level.

[0199] The implementation specifically includes the following steps:

[0200] 1. Calculate the absolute deviation and relative rate of change.

[0201] 1.1 Get input data.

[0202] 1.1.1 Current measurement values ​​(from real-time monitoring data stream in S2):

[0203] Current value of surrounding rock displacement (Unit: mm); Current value of anchor stress (Unit: MPa); Current value of pore water pressure (Unit: kPa).

[0204] 1.1.2 Dynamic risk threshold (dynamically generated results from S3):

[0205] Displacement risk threshold ; Stress risk threshold ; Osmotic pressure risk threshold .

[0206] 1.2 Deviation index calculation.

[0207] 1.2.1 Calculation of absolute deviation value.

[0208] For each monitoring parameter, perform the following calculation: Absolute deviation value = |current measurement value - dynamic risk threshold|.

[0209] Absolute deviation of surrounding rock displacement ;

[0210] Absolute deviation of anchor stress ;

[0211] Absolute deviation of pore water pressure .

[0212] 1.2.2 Calculation of relative rate of change.

[0213] Relative rate of change = (absolute deviation value / dynamic risk threshold) × 100%.

[0214] Relative change rate of surrounding rock displacement ;

[0215] Relative change rate of anchor stress ;

[0216] Relative change rate of pore water pressure .

[0217] 2. Rules for determining high-risk levels.

[0218] Based on the specification: threshold basis supported by engineering practice data.

[0219] Conditions that trigger high risk (if any one of them is met, it will be judged):

[0220] Absolute deviation of surrounding rock displacement More than 15 mm;

[0221] Relative change rate of surrounding rock displacement More than 10% per hour;

[0222] Absolute deviation of anchor stress More than 50 MPa;

[0223] Relative change rate of anchor stress More than 15% per hour;

[0224] Absolute deviation of pore water pressure More than 200 kPa;

[0225] Relative change rate of pore water pressure More than 20% per hour.

[0226] 3. Matching rules between medium and low risk levels.

[0227] 3.1 Division of surrounding rock displacement deviation intervals:

[0228] Low risk: Absolute deviation value ≤5 mm;

[0229] Medium risk: Absolute deviation value 5 mm < ≤10 mm.

[0230] 3.2 Anchor stress deviation interval division:

[0231] Low risk: Absolute deviation value ≤20 MPa;

[0232] Medium risk: Absolute deviation value 20 MPa < ≤35 MPa.

[0233] 3.3 Division of pore water pressure deviation intervals:

[0234] Low risk: Absolute deviation value ≤80 kPa;

[0235] Medium risk: absolute deviation value 80 kPa< ≤150 kPa.

[0236] Comprehensive judgment logic:

[0237] Low risk level: All monitored target parameters are in the low risk range;

[0238] Medium risk level: Any parameter is in the medium risk range and there is no high risk parameter.

[0239] This step calculates the absolute deviation and relative rate of change of surrounding rock displacement, anchor stress, and pore water pressure in real time. Based on the national standard DL / T 5298-2019, it sets high-risk thresholds (e.g., displacement deviation exceeding 15 mm, displacement rate of change exceeding 10% per hour). Exceeding thresholds immediately determine a high-risk level. Parameters within these thresholds are then assigned a low or medium risk rating based on interval rules (e.g., displacement deviation ≤ 5 mm is low risk, 5-10 mm is medium risk). This solution addresses the issues of inconsistent criteria and delayed response times in traditional manual assessments, enabling automated risk assessment within seconds, with an accuracy rate exceeding 95% verified in actual projects.

[0240] After S4, the method further includes:

[0241] S5. Generate multi-level warning signals according to the dynamic risk level of the target project area, and send the multi-level warning signals to the management and control platform.

[0242] The implementation specifically includes the following steps:

[0243] 1. Early warning signal generation.

[0244] Input: Receive the dynamic risk level (low risk, medium risk, high risk) output by S4.

[0245] Signal mapping rules:

[0246] When the risk level is low: a blue warning signal is generated (RGB color value: 0, 0, 255);

[0247] When the risk level is medium: a yellow warning signal is generated (RGB color value: 255, 255, 0);

[0248] When the risk level is high: a red warning signal is generated (RGB color value: 255,0,0).

[0249] 2. Early warning signal packaging.

[0250] Data structure: Signal information is encapsulated in JSON format and contains the following fields:

[0251] risk_level: risk level (string: "low" / "medium" / "high");

[0252] color_code: color code (string: "blue" / "yellow" / "red");

[0253] rgb_value: RGB array (e.g. high risk: [255,0,0]);

[0254] timestamp: timestamp in ISO 8601 format (e.g. "2023-08-15T08:30:45Z");

[0255] location: Coordinates of the risk point (WGS84 coordinate system, format: "latitude, longitude").

[0256] 3. The signal is output to the engineering safety management and control platform.

[0257] 3.1 Transmission protocol: Transmitted to the platform warning interface via the OPC UA protocol (IEC 62541 standard).

[0258] 3.2 Connection parameters: Service address: opc.tcp: / / <platform IP>:4840 / risk_monitor; Data node: namespace ns=2, node name RiskWarningSignal.

[0259] 3.3 Transmission Process: Establish encrypted communication (X.509 certificate authentication + AES-256 encryption); write JSON data to the OPC UA node; after receiving the data, the platform parses the data and drives the interface color alarm and positioning.

[0260] 4. Emergency response linkage.

[0261] Red alert is automatically triggered: the sound and light alarm at the construction site is activated (24V DC pulse signal); a text message is pushed to the person in charge (including risk level, location, and time).

[0262] This step strictly maps dynamic risk levels (low / medium / high) to standard color-coded warning signals (blue / yellow / red). Key information, such as time and location, is encapsulated in a JSON data structure and encrypted using the industry-standard OPC UA protocol for transmission to the project safety management and control platform. This solution addresses the limitations of traditional early warning systems, such as their inability to accurately classify and their lack of temporal and spatial information, enabling visual risk location and instant response. In actual projects, the success rate of early warning information transmission has reached over 99.9%, fully meeting safety monitoring requirements.

[0263] This embodiment provides a multi-parameter coupling risk assessment method for geological bodies of water conservancy and hydropower engineering projects. The method first generates an initial risk distribution based on initial geological data and preset coupling rules. Then, the real-time monitoring data stream and geological update data of the water conservancy and hydropower engineering area are obtained, and the data and the initial risk distribution are input into a pre-trained deep time series prediction model to output the target parameter prediction value and risk trend prediction value within a specified time period in the future. The target parameter prediction value is associated with the preset monitoring items that affect the stability of the geological body. Combined with the preset engineering safety constraints, a dynamic risk threshold is generated based on the predicted value. The current measurement value of the real-time monitoring data stream is then compared with the threshold to determine the dynamic risk level, and the corresponding multi-level warning signal output is generated accordingly, so as to realize the dynamic tracking and early prediction of the coupled risk status of the geological body of the water conservancy and hydropower engineering project, and improve the timeliness and initiative of the risk warning.

[0264] Another embodiment of the present invention provides a multi-parameter coupling risk assessment system for engineering geological bodies, which uses Figure 1 The multi-parameter coupling risk assessment method for engineering geological bodies of the embodiment includes:

[0265] A basic risk assessment module, configured to generate an initial risk distribution of a target engineering area based on initial geological data of the target engineering area and preset coupling rules;

[0266] Dynamic data acquisition module, used to obtain real-time monitoring data streams and geological update data of the target project area;

[0267] a time series prediction engine module connected to the basic risk assessment module and the dynamic data acquisition module, the time series prediction engine module being used to determine a target parameter prediction value and a risk trend prediction value within a preset time period based on the real-time monitoring data stream, the geological update data, and the initial risk distribution, wherein the target parameter prediction value is used to indicate a preset monitoring item that affects the stability of the engineering geological body;

[0268] A dynamic threshold generation module, connected to the time series prediction engine module, is used to generate a dynamic risk threshold based on the target parameter prediction value and the risk trend prediction value, in combination with preset engineering safety constraints;

[0269] a risk level determination module, connected to the dynamic data acquisition module and the dynamic threshold generation module, and configured to compare the current measurement value of the real-time monitoring data stream with the dynamic risk threshold to determine the dynamic risk level of the target engineering area;

[0270] The early warning output module is connected to the risk level determination module. The early warning output module is used to generate a multi-level early warning signal according to the dynamic risk level of the target engineering area and send the multi-level early warning signal to the management and control platform.

[0271] Specifically, the basic risk assessment module includes:

[0272] A three-dimensional modeling unit, used to construct a three-dimensional geological numerical model based on the initial geological data of the target engineering area;

[0273] A coupling rule calculation unit is connected to the three-dimensional modeling unit, and is used to calculate the stability quantitative index of the geological body of each grid unit in the three-dimensional geological numerical model according to a preset coupling rule;

[0274] The risk distribution generating unit is connected to the coupling rule calculating unit, and the risk distribution generating unit is used to generate the initial risk distribution of the target engineering area based on the spatial distribution of the stability quantitative index.

[0275] Specifically, the dynamic data acquisition module includes:

[0276] A real-time data acquisition unit, comprising a distributed sensor array, wherein the distributed sensor array is used to collect monitoring data of surrounding rock displacement, anchor stress and pore water pressure in the target engineering area to form a real-time monitoring data stream;

[0277] The geological data updating unit includes a geological radar parser, which is used to receive the re-survey data of the fault zone in the target engineering area as geological update data.

[0278] The multi-parameter coupling risk assessment system for engineering geological bodies provided in this embodiment includes the following steps when implemented:

[0279] 1. Basic risk assessment module structure and connection.

[0280] 1.1 Three-dimensional modeling unit.

[0281] Physical connection: Receives initial geological data (drill cores, geological profiles, and ground stress measurement results) from the geological exploration database via the Gigabit Ethernet interface.

[0282] Functionality: GOCAD geological modeling software was used to generate a 3D geological numerical model with an inverse distance weighted interpolation algorithm, with a grid size of 5m × 5m × 5m. The unit output was connected to the coupling rule calculation unit via the PCIe bus.

[0283] Technical effect: Discrete geological data is converted into a structured three-dimensional model with an interpolation error of ≤15%, providing a spatial carrier for risk quantification.

[0284] 1.2 Coupling rule calculation unit.

[0285] Physical connection: Receive the grid data of the 3D geological numerical model through the PCIe bus, and load the preset coupling rule library (Hoek-Brown criterion, Biot seepage-stress equation) through the shared memory channel.

[0286] Function implementation: Iteratively calculate the safety factor for each grid cell , penetration stability index , deformation energy density .

[0287] Technical effect: It realizes the multi-physics field coupling analysis of geological parameters, with single-grid calculation time ≤5ms and calculation accuracy improved by 40%.

[0288] 1.3 Risk distribution generation unit.

[0289] Physical connection: Receive stability metrics for all grids via the DDR4 memory bus.

[0290] Function Implementation: Perform linear weighted fusion and render a red-yellow-green risk distribution map on the Paraview platform.

[0291] Technical effect: Output the initial risk distribution in GeoTIFF format with a spatial resolution of 0.5m as a static risk assessment benchmark.

[0292] 2. Dynamic data acquisition module structure and connection.

[0293] 2.1 Real-time data acquisition unit.

[0294] Physical connection: The vibrating wire displacement meter is connected to the distributed data acquisition box via the RS-485 bus; the steel bar stress meter is connected to the signal conditioner via a 4-20mA analog signal line; and the piezometer is transmitted to the industrial switch via the Modbus TCP protocol.

[0295] Functional implementation: The system collects surrounding rock displacement (in mm), anchor stress (in MPa), and pore water pressure (in kPa) at a frequency of 0.1 Hz, and uses the Hampel filter algorithm to clean abnormal values ​​in real time.

[0296] Technical effect: Output monitoring data stream with complete timing, communication delay ≤100ms, data efficiency ≥99%.

[0297] 2.2 Geological data updating unit.

[0298] Physical connection: The geological radar transmits fault zone data in Segy format to the radar analyzer through an optical fiber transceiver.

[0299] Functionality: The analyzer performs bandpass filtering (50-150MHz) and common center velocity analysis to extract fault width, dip, and inclination parameters.

[0300] Technical effect: Output geological update data in DXF format, positioning accuracy ±0.5m, update cycle ≤24 hours.

[0301] 3. Timing prediction engine module structure and connection.

[0302] 3.1 Physical Connection

[0303] Receive the initial risk distribution from the basic risk assessment module via the InfiniBand network and receive the real-time monitoring data stream and geological update data from the dynamic data acquisition module via the OPC UA protocol.

[0304] 3.2 Function implementation

[0305] Spatial registration: Perform affine transformation (rotation angle +Translation , ), the initial risk distribution and fault parameters are integrated into a 5-channel feature matrix;

[0306] Time series processing: Perform sliding window normalization on real-time data streams (window size 600 points, step size 60 points);

[0307] Tensor construction: concatenate the spatial feature matrix and the normalized time series data into 3D input tensor;

[0308] LSTM prediction: A pre-trained two-layer LSTM network (128-64 units, tanh activation) is used to output the target parameter forecast and risk trend for the next 6 hours.

[0309] Technical effect: Realize static and dynamic data fusion prediction, displacement prediction error ≤1.2mm, and risk trend accuracy rate of 91%.

[0310] 4. Dynamic threshold generation module structure and connection.

[0311] 4.1 Physical connection: Receive the target parameter prediction value and risk trend prediction value of the timing prediction engine module through the PCIe Gen4 bus.

[0312] 4.2 Function implementation

[0313] Calculate the 95% confidence interval (self-sampling 1000 times, =1.96);

[0314] According to the changing slope of risk trend Adjusting the tolerance factor (Increase the coefficient when risk increases and decrease it when risk decreases);

[0315] Dynamic risk thresholds are generated by combining engineering constraint values ​​(displacement 20 mm / stress 320 MPa / seepage pressure 800 kPa).

[0316] Technical effect: The threshold value adapts to changes in working conditions, reducing the false alarm rate by 60% and the missed alarm rate by 45%.

[0317] 5. Risk level determination module structure and connection.

[0318] 5.1 Physical Connection

[0319] The current measurement value of the dynamic data acquisition module is obtained through the shared memory; the dynamic risk threshold of the dynamic threshold generation module is received through the DMA channel.

[0320] 5.2 Function implementation

[0321] Calculate absolute deviation and relative rate of change;

[0322] Determine the risk level according to DL / T 5298-2019:

[0323] High risk: >15mm or >10% / h; medium risk: 5mm< ≤10mm; low risk: ≤5mm.

[0324] Technical effect: Achieve automatic judgment in seconds, with an accuracy rate ≥ 95% and a response time ≤ 0.5 seconds.

[0325] 6. Early warning output module structure and connection.

[0326] 6.1 Physical connection: Connect to the alarm server of the engineering safety management and control platform through the OPC UA protocol (port 4840).

[0327] 6.2 Function implementation

[0328] Low risk → generates RGB (0, 0, 255) blue signal; Medium risk → generates RGB (255, 255, 0) yellow signal; High risk → generates RGB (255, 0, 0) red signal;

[0329] The JSON data structure encapsulates timestamp, coordinates, and risk level; it is encrypted with AES-256 and transmitted to the platform.

[0330] 6.3 Technical Effect: Supports visual risk positioning, transmission success rate of 99.9%, and linkage response delay ≤ 200ms.

[0331] This system generates spatialized initial risk distribution through the basic risk assessment module, the dynamic data acquisition module obtains real-time / non-real-time geological data, the time series prediction engine module realizes multi-source data fusion and 6-hour accurate prediction, the dynamic threshold generation module outputs adaptive dynamic risk thresholds, the risk level determination module performs specification-driven automated grading, and the warning output module realizes multi-level warnings with encrypted transmission.

[0332] The above descriptions are merely a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application discloses the preferred embodiments as above, they are not intended to limit the present application. Any technical personnel familiar with the present profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A multi-parameter coupling risk assessment method for engineering geological bodies, characterized in that: The method comprises: S1. generating an initial risk distribution of the target engineering area based on initial geological data of the target engineering area and preset coupling rules; S2. Acquire real-time monitoring data streams and geological update data of the target project area, and determine target parameter prediction values ​​and risk trend prediction values ​​within a preset time period based on the real-time monitoring data streams, the geological update data, and the initial risk distribution; S3. Generate a dynamic risk threshold based on the target parameter prediction value and the risk trend prediction value, in combination with preset engineering safety constraints; S4. Compare the current measurement value of the real-time monitoring data stream with the dynamic risk threshold to determine the dynamic risk level of the target engineering area.

2. The method according to claim 1, characterized in that Said S1 specifically includes: Construct a three-dimensional geological numerical model based on the initial geological data of the target project area; Calculating the stability quantitative index of the geological body of each grid unit in the three-dimensional geological numerical model according to a preset coupling rule; An initial risk distribution of the target project area is generated based on the spatial distribution of the stability quantitative index.

3. The method according to claim 1, characterized in that The step S2 of obtaining the real-time monitoring data stream and geological update data of the target project area specifically includes: Continuously collecting monitoring data of surrounding rock displacement, anchor stress and pore water pressure in the target engineering area through a distributed sensor network to form a real-time monitoring data stream; Obtain geological radar re-survey data of the fault zone in the target engineering area as geological update data.

4. The method according to claim 1, wherein The step S2 of determining the target parameter prediction value and the risk trend prediction value within a preset time period based on the real-time monitoring data stream, the geological update data, and the initial risk distribution specifically includes: Using the initial risk distribution to train a long short-term memory network to obtain a deep time series prediction model, and determining a three-dimensional input tensor based on the real-time monitoring data stream, the geological update data and the initial risk distribution; The three-dimensional input tensor is input into the deep time series prediction model to obtain the target parameter prediction value and risk trend prediction value within a preset time length.

5. The method according to claim 4, characterized in that Determining a three-dimensional input tensor according to the real-time monitoring data stream, the geological update data, and the initial risk distribution specifically includes: Performing spatial registration and fusion on the geological update data and the initial risk distribution to generate spatial enhancement features; The real-time monitoring data stream is subjected to sliding window normalization processing, and the processed real-time monitoring data stream is spliced ​​with the spatial enhancement feature in the time dimension to obtain a three-dimensional input tensor.

6. The method according to claim 1, characterized in that The S3 specifically includes: Calculating the fluctuation range of the predicted value of the target parameter at a preset confidence level; Adjusting the tolerance coefficient of the fluctuation range according to the change slope of the risk trend prediction value; Generate dynamic risk thresholds based on the adjusted tolerance factor and preset engineering safety constraints.

7. The method according to claim 1, characterized in that The S4 specifically includes: Calculating an absolute deviation value and a relative change rate between a current measurement value of the real-time monitoring data stream and the dynamic risk threshold; The dynamic risk level of the target engineering area is determined according to the absolute deviation value and the relative change rate.

8. The method according to claim 7, characterized in that Determining the dynamic risk level of the target engineering area according to the absolute deviation value and the relative change rate specifically includes: When the absolute deviation value exceeds the first-level deviation threshold or the relative change rate exceeds the first-level change rate threshold, it is determined to be a high risk level; When the absolute deviation value does not exceed the secondary deviation threshold and the relative change rate does not exceed the secondary change rate threshold, it is determined to be a low risk level; otherwise, it is determined to be a medium risk level.

9. The method according to claim 1, characterized in that After S4, the method further includes: S5. Generate a multi-level warning signal according to the dynamic risk level of the target engineering area, and send the multi-level warning signal to the management and control platform.

10. The method according to claim 6, characterized in that The preset confidence level is 90% to 95%.

Images