Digital observation of surface rock movement at working face and intelligent prediction method for safety of passing structures

Abstract

This invention discloses a method for digitally observing surface rock movement at a working face and intelligently predicting the safety of over-structure mining. This method, which belongs to the fields of mining and information technology, includes the following steps: determining a safety warning value for structure movement and deformation; collecting and preprocessing movement and deformation data; analyzing the dynamic patterns of surface movement and deformation using finite element analysis; establishing a structure movement and deformation prediction model; training and validating the structure movement and deformation prediction model, and using the trained model to predict structure movement and deformation; comprehensively evaluating safety and stability; and revising the prediction model parameters and continuously evaluating until safe over-structure mining is achieved. This invention provides a highly promising new approach for accurately predicting rock movement trends and ensuring the safety of important facilities such as structures.

Description

Technical Field

[0001] The present invention belongs to the fields of mining and information technology, and specifically relates to a method for digitally observing surface rock movement at a working face and intelligently predicting the safety of passing structures. Background Art

[0002] When coal mining areas are near critical infrastructure such as structures, the potential safety risks posed by surface rock movement "passing over" them increase dramatically. "Passing over" refers to the dynamic process by which surface rock movement gradually spreads from areas away from the structures, directly impacting them. For example, gas stations often store large quantities of flammable and explosive oils, and their facilities are extremely sensitive to surface deformation. When surface rock movement "passes over" structures, even relatively small ground displacements can cause compression, stretching, or distortion of the station's structural components, leading to cracks in walls and uneven foundation settlement. It can also cause collisions and compression of oil storage equipment, causing deformation or even rupture. It can also pull and bend oil pipelines, loosening joints or causing pipe ruptures. These damages can easily lead to oil leaks, which, if ignited by open flames or static electricity, can cause serious accidents such as fires or even explosions, resulting in significant economic losses and serious threats to the surrounding environment and human safety.

[0003] Traditional methods for observing and analyzing surface rock movement mainly include methods based on empirical formulas and some simple statistical analysis methods. Empirical formula-based methods are often developed under specific geological conditions and have poor versatility. When faced with different geological environments or complex mining conditions, the predicted results may deviate significantly from the actual situation. While simple statistical analysis methods can process historical data to a certain extent, they cannot effectively mine the deep-level information in the data and struggle to capture the nonlinear characteristics and long-term dependencies in surface rock movement data. Traditional methods have many limitations when processing complex time series data such as surface rock movement. For example, traditional methods struggle to accurately predict and model characteristics such as periodic fluctuations and trend changes in time series. Furthermore, as new data is generated over time, traditional methods are unable to adaptively update the model to adapt to new data patterns, resulting in a gradual decrease in prediction accuracy over time.

[0004] With the recent advancements in machine learning and deep learning, long short-term memory (LSTM) networks and their variants have demonstrated remarkable performance in processing time series data. Stacked LSTMs, in particular, stack multiple LSTM layers together to progressively extract multi-level features from complex time series data. Each LSTM layer can be considered a feature extractor. Lower-level LSTMs capture local features and short-term dependencies in the data. As the number of layers increases, higher-level LSTMs further integrate these local features to uncover longer-term, more complex dependencies and patterns. The Transformer's multi-head attention mechanism processes the feature sequences output by the LSTM layers in parallel, spanning different locations and time points, comprehensively capturing long-range dependencies in the data and effectively compensating for the information attenuation that LSTMs can experience when processing long sequences. This multi-level feature extraction approach offers significant advantages in processing complex, multi-factor-influenced time series data such as surface rock movement, providing a promising new approach for accurately predicting rock movement trends and ensuring the safety of critical infrastructure such as structures. Summary of the Invention

[0005] In order to solve the above problems, the present invention proposes a method for digital observation of surface rock movement at the working face and intelligent prediction of the safety of passing structures. It mainly uses stacked long short-term memory networks and Transformer to establish a structure movement deformation prediction model based on movement deformation observation data mining, and uses this model to predict the movement deformation of the structure; according to the prediction results of the structure movement deformation, the safety and stability of the structure are comprehensively evaluated, and when the predicted value exceeds the warning value, corresponding technical measures are taken.

[0006] The technical solutions of the present invention are as follows:

[0007] A method for digitally observing rock movement on a working face and intelligently predicting the safety of passing structures includes the following steps:

[0008] Step 1: Determine the safety warning value of structure movement and deformation;

[0009] Step 2: Collect motion deformation data and perform preprocessing;

[0010] Step 3: Use finite element analysis to analyze the dynamic changes of surface movement and deformation;

[0011] Step 4: Establish a structure movement and deformation prediction model;

[0012] Step 5: Train and verify the structure movement and deformation prediction model, and use the trained model to predict the structure movement and deformation;

[0013] Step 6: Comprehensively evaluate safety and stability;

[0014] Step 7: Modify the prediction model parameters and continue to evaluate until the structure is safely mined.

[0015] Furthermore, the specific process of step 1 is as follows:

[0016] Step 1.1: Comprehensively consider the structural characteristics of the structure, its geological conditions, and the surrounding environment to identify key factors affecting the stability and safety of the structure. Key factors include the type of foundation, the properties of the building materials, the overall rigidity of the structure, the geological structure characteristics, and other surrounding engineering activities.

[0017] Step 1.2: Identify the key indicators that affect the movement and deformation state of the structure, including displacement, inclination, curvature, and strain; combine the structure's design specifications, usage requirements, and relevant industry standards to determine the threshold value of each key indicator. This threshold value is the safety warning value for the movement and deformation of the structure.

[0018] Furthermore, the specific process of step 2 is as follows:

[0019] Step 2.1: Install various types of sensors at key structural locations of the structure to collect daily surface movement and deformation data in the mined area in front of the structure. The sensors include displacement sensors, inclination sensors, curvature sensors, and strain sensors.

[0020] Step 2.2: Detect and process outliers.

[0021] The process of outlier detection is as follows: define the displacement data sequence collected by the displacement sensor as , calculate the mean of the displacement data series and standard deviation , the formula is as follows:

[0022] ;

[0023] ;

[0024] in, is the total number of displacement data points; For the displacement data points; For the displacement data points; for The displacement data points of the interval are marked as outliers; is the standard deviation The income from the prescription;

[0025] The process of outlier processing is as follows:

[0026] For data with a stable trend, the linear interpolation method is used to handle outliers. Assume that at the first time point and the third time point The displacement data points are and , and the second time point If the displacement data points are abnormal or missing, Displacement data points It is calculated by the linear interpolation formula, which is:

[0027] ;

[0028] For data with unstable trend, the weighted linear interpolation method is used to deal with outliers. Middle displacement data points The displacement data points identified as outliers and removed are called missing value points. Interpolate the position; let the time series corresponding to the displacement data series be , For the A point in time, is the total number of time points, one displacement data point corresponds to one time point, and the total number of time points corresponds to the total number of displacement data points; the weighted linear interpolation formula is:

[0029] ;

[0030] ;

[0031] in, for The interpolation result of the location; for The weight of For the time points To time points Time distance; For the time points To time points Time distance; For the time points The displacement data points;

[0032] Step 2.3: Normalize the data. The formula is:

[0033] ;

[0034] in, is the normalized displacement data point; is the displacement data point; 、 are the minimum and maximum values ​​of the displacement data in the entire data set;

[0035] For inclination, curvature, and strain, outlier detection and processing, and data normalization are performed in the same way as displacement data.

[0036] Furthermore, the specific process of step 3 is as follows:

[0037] Step 3.1: Construct a geometric model and use the unit partitioning strategy to discretize the units to obtain a discretized model. The specific process is as follows: First, construct a two-dimensional or three-dimensional geometric model based on geological survey data and the design blueprint of the structure; then, use the unit partitioning strategy to discretize the two-dimensional or three-dimensional geometric model to obtain a discretized model; the process of the unit partitioning strategy is as follows: let the total number of units be , the total number of nodes is ; For each unit, the three nodes are numbered 、 、 , No. The displacement vector of any node in a unit is , is the horizontal displacement, is the vertical displacement, 、 Respectively The horizontal displacement and vertical displacement of any node in a unit; ,node ,node The displacement vectors are 、 、 ; 、 Node Horizontal displacement and vertical displacement; 、 Node Horizontal displacement and vertical displacement; 、 Node Horizontal displacement and vertical displacement; is the transposed sign; the displacement vectors are connected through shape functions; for linear triangle elements, the nodes The shape function is expressed as:

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] in, For nodes The shape function of is the area of ​​the triangular unit cell; For computing nodes The shape function is the intermediate parameter constructed; For computing nodes The correlation coefficient when the shape function is ; For computing nodes The correlation coefficient when the shape function is ; is the abscissa representing the horizontal position in the plane rectangular coordinate system; is the ordinate representing the vertical position in the plane rectangular coordinate system; The nodes in the triangle element The horizontal axis of The nodes in the triangle element The horizontal axis of is the node in the triangle element The vertical coordinate of The nodes in the triangle element The vertical coordinate of

[0043] node Shape function ,node Shape function With node The shape functions of are expressed in the same way;

[0044] Final The displacement vector of any node in a unit is expressed as:

[0045] ;

[0046] At the same time, the strain-displacement relationship is:

[0047] ;

[0048] in, For the The strain vector of any node in the element; is the strain-displacement matrix, which is:

[0049] ;

[0050] Step 3.2, establish the constitutive equation, equilibrium equation and overall stiffness matrix;

[0051] Step 3.3: Solve the overall stiffness matrix and analyze the results to obtain the dynamic change law of surface movement and deformation.

[0052] Furthermore, in step 3.2, The constitutive equation of each unit is:

[0053] ;

[0054] in, For the The stress tensor of any node in an element; is the elasticity matrix, expressed as:

[0055] ;

[0056] in, is the elastic modulus; is Poisson's ratio;

[0057] The specific expression of the equilibrium equation is:

[0058] ;

[0059] in, It is The volume of a unit; For virtual strain; is the strain vector; is the stress tensor; is the volume; It is The surface of each unit; is the surface force vector; For the surface; is the displacement vector; is a node The force vector of

[0060] The overall stiffness equation is:

[0061] ;

[0062] in, is the overall stiffness matrix; is the node displacement vector; is the force vector of the node;

[0063] No. The stiffness matrix of each element The calculation formula is:

[0064] ;

[0065] Global stiffness matrix The assembly process is to superimpose the stiffness matrix of each unit to the corresponding position of the overall stiffness matrix according to the corresponding relationship between the unit node and the overall node;

[0066] In step 3.3, the overall stiffness equation is solved using direct or iterative methods. Get the node displacement vector ; After obtaining the node displacement vector, the unit strain vector and stress tensor are further calculated; the unit strain vector is calculated based on the strain-displacement relationship, and the unit stress tensor is calculated by the constitutive equation.

[0067] Furthermore, in step 4, the input variables of the structure movement and deformation prediction model are the relevant parameters in the surface movement and deformation observation data, and the output is the movement and deformation value of the structure; the relevant parameters include displacement, inclination, curvature, and strain;

[0068] The structure movement and deformation prediction model consists of multiple stacked LSTM layers, a Transformer layer, and a fully connected layer. LSTM is a long short-term memory network. Transformer is a sequence model based on the attention mechanism.

[0069] The working process of the structure movement and deformation prediction model is as follows:

[0070] Step 4.1: Input the relevant parameters of the surface movement and deformation observation data into the first layer of LSTM. The working process of the first layer of LSTM is as follows:

[0071] Step 4.1.1: Preprocess the input parameters. After preprocessing, the input parameters are a vector at each time point. Define the first vector after preprocessing. The input data for each time step is ,in is the dimension of the input data; For the time step, The Data of dimensions;

[0072] Step 4.1.2: Calculate the forget gate. The formula is:

[0073] ;

[0074] in, is the forget gate in the first layer of LSTM Output of time steps; is the Sigmoid activation function; is the weight matrix of the forget gate in the first layer of LSTM; is the first layer of LSTM The hidden state of the time step; is the bias vector of the forget gate in the first layer of LSTM;

[0075] Step 4.1.3: Perform input gate calculation to generate the input gate control signal and candidate cell state. The formula is:

[0076] ;

[0077] ;

[0078] in, is the first layer of LSTM The control signal of the input gate in time steps; is the weight matrix of the input gate in the first layer of LSTM; is the bias vector of the input gate in the first layer of LSTM; is the first layer of LSTM Candidate cell states for time steps; is the tanh activation function; is the weight matrix used to calculate the candidate cell state in the first layer of LSTM; is the bias vector used to calculate the candidate cell state in the first layer of LSTM;

[0079] Step 4.1.4. Update the cell state according to the output of the forget gate and input gate:

[0080] ;

[0081] in, is the first layer of LSTM Update cell state for each time step; Represents element-wise multiplication operation; is the first layer of LSTM The cell state at each time step;

[0082] Step 4.1.5: Perform output gate calculation to generate the output gate control signal and output hidden state. The formula is:

[0083] ;

[0084] ;

[0085] in, is the first layer of LSTM The control signal of the output gate in time steps; is the weight matrix of the output gate in the first layer of LSTM; is the bias vector of the output gate in the first layer of LSTM; is the first layer of LSTM The hidden state of the time step;

[0086] Step 4.2, Enter the second layer LSTM. The calculation formula of the second layer LSTM is as follows:

[0087] ;

[0088] ;

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] in, The forget gate in the second layer LSTM Output of time steps; is the weight matrix of the forget gate in the second layer of LSTM; is the bias vector of the forget gate in the second layer of LSTM; is the first The hidden state of the time step;

[0094] is the first The control signal of the input gate in time steps; is the weight matrix of the input gate in the second layer of LSTM; is the bias vector of the input gate in the second layer of LSTM; is the first Candidate cell states for time steps; It is the weight matrix used to calculate the candidate cell state in the second layer of LSTM; is the bias vector used to calculate the candidate cell state in the second layer of LSTM;

[0095] is the first Update cell state for each time step; is the first The cell state at each time step;

[0096] is the first The control signal of the output gate in time steps; is the weight matrix of the output gate in the second layer of LSTM; is the bias vector of the output gate in the second layer of LSTM;

[0097] In the second LSTM layer The hidden state of time steps Will serve as the input of the third layer LSTM;

[0098] Step 4.3, The input of the LSTM layer is The output hidden state of the layer; The calculation formula of the layer LSTM is as follows:

[0099] ;

[0100] ;

[0101] ;

[0102] ;

[0103] ;

[0104] ;

[0105] in, For the Forgetting level in layer LSTM Output of time steps; It is The weight matrix of the forget gate in the layer LSTM; It is The bias vector of the forget gate in the layer LSTM;

[0106] For the In the LSTM layer The control signal of the input gate in time steps; It is The weight matrix of the input gate in the layer LSTM; It is The bias vector of the input gate in the layer LSTM;

[0107] For the In the LSTM layer Candidate cell states for time steps; It is The weight matrix used to calculate the candidate cell state in the LSTM layer; It is The relevant bias vector used to calculate the candidate cell state in the layer LSTM;

[0108] For the In the LSTM layer Update cell state for each time step; It is In the LSTM layer The cell state at each time step;

[0109] For the In the LSTM layer The control signal of the output gate in time steps; It is The weight matrix of the output gate in the layer LSTM; It is The bias vector of the output gate in the layer LSTM;

[0110] For the In the LSTM layer The hidden state of the time step;

[0111] is the total number of time steps, and the final output of the entire stacked LSTM layer is In the LSTM layer The hidden state of time steps ;

[0112] Step 4.4, Input Transformer layer and get output matrix ;

[0113] Step 4.5, Input the fully connected layer to predict the movement and deformation of the structure.

[0114] Furthermore, in step 4, the Transformer layer adopts a multi-head attention mechanism and a feedforward neural network; the specific process is as follows:

[0115] Step 4.4.1. First, the final output of the stacked LSTM layer Perform linear transformation to obtain the query matrix , key matrix Sum Matrix ;set up 、 、 They are 、 、 The weight matrix is:

[0116] ;

[0117] ;

[0118] ;

[0119] Then, suppose 、 、 For the The query matrix, key matrix, and value matrix of each head, then The attention scores of the individuals are:

[0120] ;

[0121] in, For the Attention score of size; is the softmax function; yes Dimensions; is the transpose symbol; is the number of heads;

[0122] Finally, the output matrix of the multi-head attention for:

[0123] ;

[0124] in, For splicing operation; It is the weight matrix used to fuse multi-head information; For the Attention score of size;

[0125] Step 4.4.2, Perform feedforward neural network processing, the formula is:

[0126] ;

[0127] in, It is a feedforward neural network; 、 are two different weight matrices of a feedforward neural network; 、 are two different bias vectors for the feedforward neural network;

[0128] In step 4.5, the fully connected layer will The features are converted into specific prediction values; The predicted value of the time step The calculation formula is:

[0129] ;

[0130] in, is the weight matrix of the fully connected layer; is the bias vector of the fully connected layer.

[0131] Furthermore, the specific process of step 5 is as follows:

[0132] Step 5.1: Train the model by minimizing the mean square error loss function; let the predicted value sequence be , the true value sequence is , then the calculation formula of the mean square error loss function is:

[0133] ;

[0134] in, is the mean square error loss function value; For the The true value of the time step; For the The predicted value of each time step;

[0135] At the same time, the mean absolute error loss function is used as a supplementary evaluation indicator, and the calculation formula is:

[0136] ;

[0137] in, is the mean absolute error loss function value;

[0138] Step 5.2: Update the model parameters using the adaptive moment estimation optimization algorithm. The parameters include the weight matrix and bias vector of each LSTM layer, Transformer layer, and output layer. The update process is as follows:

[0139] First, calculate the update of the first-order moment vector:

[0140] ;

[0141] in, For the The first-order moment vector of time steps; For the The first-order moment vector of time steps; For the The gradient of time steps;

[0142] Then, calculate the update of the second-order moment vector:

[0143] ;

[0144] in, For the The second-order moment vector of time steps; For the The second-order moment vector of time steps; and are two different hyperparameters used to control momentum;

[0145] Next, the first-order moment vector and the second-order moment vector are corrected for deviation:

[0146] ;

[0147] ;

[0148] in, 、 Respectively First-order moment vector deviation and second-order moment vector deviation of each time step;

[0149] Finally, update the parameters:

[0150] ;

[0151] in, 、 Respectively time step, Parameters of time steps; is the learning rate; is a non-zero constant;

[0152] Step 5.3, hyperparameters include learning rate , the number of hidden units in the LSTM layer and the number of heads in the Transformer layer , set a series of value, Value and The evaluation indicators are mean square error and mean absolute error, and the best hyperparameter combination is selected. The number of training rounds when the loss of the validation set reaches the minimum is selected as the final number of training rounds. The learning rate scheduling strategy is used to adjust the learning rate.

[0153] Furthermore, the specific process of step 6 is: comparing the model prediction value with the safety warning value determined in step 1. When the prediction value does not exceed the warning value, it indicates that the structure is currently in a safe state; when the prediction value exceeds the warning value, it indicates that there is a safety risk in the structure.

[0154] Furthermore, the specific process of step 7 is as follows:

[0155] Step 7.1: During the mining process at the working face, continuously obtain the next stage of movement and deformation observation data of the surface and structures in the mined area;

[0156] Step 7.2: Substitute the newly acquired observation data into the prediction model and use the Adam optimization algorithm to modify the parameters of the prediction model so that the model can better adapt to the new data characteristics and actual deformation conditions.

[0157] Step 7.3: Use the revised prediction model to predict the movement and deformation of the structure again and re-evaluate its safety and stability;

[0158] Step 7.4: Repeat step 7.3, continuously adjusting the model and assessing the safety status until the working face safely passes through the structure, ensuring that the safety impact on the structure is within a controllable range throughout the mining process.

[0159] The beneficial technical effects brought about by the present invention: The present invention proposes a method for digital observation of surface rock movement at the working face and intelligent prediction of the safety of over-structures. By utilizing the powerful processing capabilities of stacked LSTM and Transformer for complex time series data, it can deeply explore the multi-level features in the surface rock movement data. By stacking multiple layers of LSTM and Transformer, the model can gradually extract long-term change trends from short-term rock movement fluctuations. This multi-level information extraction method makes the prediction of rock movement more accurate. Compared with traditional prediction methods, it can more accurately capture subtle changes in rock movement, reduce prediction errors, and provide a reliable basis for subsequent safety assessments. By comparing the predicted rock movement data with the safety threshold of the structure in real time, potential safety threats can be discovered in a timely and accurate manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0160] Figure 1 This is a flow chart of the method for digital intelligent observation of surface rock movement at the working face and intelligent prediction of safety of passing structures according to the present invention. DETAILED DESCRIPTION

[0161] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0162] This invention aims to achieve reasonable and accurate dynamic safety warnings for structures. It proposes a dynamic safety warning and protection technology for structures based on movement deformation prediction, correction, and re-prediction. The specific technical approach is as follows: First, the impact of surface movement deformation on the target structure is analyzed, the main indicators affecting the structure are identified, and the thresholds for each indicator, namely, the structure movement deformation safety warning values, are determined. Secondly, based on daily surface movement deformation observation data from the mined area ahead of the structure, the dynamic changes in surface movement deformation are analyzed and used to predict structure movement deformation. Secondly, a structure movement deformation prediction model is established based on mining of movement deformation observation data to predict structure movement deformation. Based on the predicted structure movement deformation results, the safety and stability of the structure are comprehensively evaluated. When the predicted value exceeds the warning value, appropriate technical measures are implemented. Finally, based on the surface movement deformation observation data from the mined area and the structure in the next stage, the prediction model parameters are corrected. Then, structure movement deformation prediction and safety and stability evaluation are performed again until the working face safely passes through the structure.

[0163] like Figure 1 As shown in FIG, a method for digitally observing surface rock movement at a working face and intelligently predicting the safety of overpass structures includes the following steps:

[0164] Step 1: Determine the safety warning value of structure movement and deformation. This includes the following steps:

[0165] Step 1.1: Detailed analysis of the impact of ground movement and deformation on the target structure. By comprehensively considering the structure's structural characteristics, geological conditions, and surrounding environment, key factors influencing its stability and safety were identified. The key factors ultimately determined were: the structure's foundation type (e.g., shallow versus deep foundations; different foundation types have different load-bearing and adaptability to ground deformation); building material properties (e.g., concrete strength grade, steel yield strength; material properties determine their performance under load and deformation); the overall structural stiffness (including the cross-sectional dimensions and layout of components such as beams and columns; stiffness affects the structure's ability to resist deformation); geological characteristics (e.g., ground stability, the presence of faults or weak interlayers; geological structures can exacerbate or mitigate the transmission of ground movement and deformation to the structure); and other surrounding engineering activities (e.g., the presence of other underground construction projects and traffic loads in the adjacent area; these external factors may have a cumulative effect on ground movement and deformation caused by coal mining).

[0166] Step 1.2: Identify key indicators that influence the movement and deformation of structures, including displacement (horizontal and vertical), inclination, curvature, and strain. Based on the structure's design specifications, usage requirements, and relevant industry standards, determine thresholds for each key indicator. These thresholds serve as safety warning values ​​for structure movement and deformation. For example, for ordinary brick-concrete residential buildings, the horizontal displacement threshold can be set at 10 mm, the vertical displacement threshold at 15 mm, the inclination threshold at 0.003 (calculated based on the structure's height), the curvature threshold at 0.0002 (indicating the degree of bending per unit length), and the strain threshold at 0.001 (reflecting the degree of deformation of the structural material). For large steel frames in industrial plants, given their relatively high structural stiffness but demanding high deformation accuracy, the horizontal displacement threshold can be set at 8 mm, the vertical displacement threshold at 12 mm, the inclination threshold at 0.002, the curvature threshold at 0.00015, and the strain threshold at 0.0008.

[0167] Step 2: Collect movement deformation data and perform preprocessing. The specific process is as follows:

[0168] Step 2.1: Install various types of sensors at key structural locations of the structure, including but not limited to high-precision displacement sensors (such as laser displacement sensors and capacitive displacement sensors), tilt sensors (such as dual-axis inclinometers), curvature sensors, and strain sensors (such as fiber Bragg grating strain sensors). These sensors can collect real-time or periodic data on daily surface movement and deformation in the mined area in front of the structure, covering a sufficiently long time period and a variety of operating conditions. The sensor placement must be carefully designed based on the structural mechanics model of the structure to ensure a comprehensive and accurate representation of the structure's deformation state. For large structures such as high-rise buildings and large bridges, sensor placement must adhere to specific structural mechanics principles and monitoring objectives. For example, sensors can be placed at key locations such as bridge piers, mid-span and ends of main beams, and pylons to comprehensively capture deformation information under different operating conditions. For high-rise buildings, sensors can be installed at corner columns, around the core tube, and at key beams and columns on different floors. The frequency of data collection depends on the type of structure and monitoring requirements. For large, slowly changing structures, the frequency can be once every hour or even several hours. For structures in complex environments or those sensitive to deformation, the frequency may be increased to once every minute or even higher. At the same time, the acquisition timestamp of each data point must be recorded to facilitate the subsequent construction of accurate time series data.

[0169] Step 2.2: Perform outlier detection and processing.

[0170] Step 2.2.1, outlier detection. Define the displacement data sequence collected by the displacement sensor as , assuming that under normal circumstances, the displacement data sequence obeys the normal distribution . Calculate the mean of the displacement data series and standard deviation , the formula is as follows:

[0171] ;

[0172] ;

[0173] in, is the total number of displacement data points; For the displacement data points; For the displacement data points;

[0174] According to the properties of normal distribution, about 99.7% of the displacement data points should fall in Displacement data points outside this interval are initially marked as outliers. is the standard deviation The result of taking the square root.

[0175] Step 2.2.2, outlier processing. When the sensor has communication failure, power supply problem or other reasons that cause abnormal or missing data, the outlier needs to be processed. For short-term outliers (such as several consecutive time points), if the data change trend is relatively stable, a simple linear interpolation method is used. Assume that at the first time point and the third time point The displacement data points are and , and the second time point If the displacement data points are abnormal or missing, Displacement data points It can be calculated by the linear interpolation formula, which is:

[0176] ;

[0177] In order to improve the accuracy of interpolation, weighted linear interpolation is considered, which assigns different weights according to the time distance between the displacement data points and the outliers.

[0178] Assume that the displacement data sequence Middle displacement data points The displacement data points identified as outliers and removed are called missing value points. Interpolation is performed at the position where the displacement data series is located. Assume that the time series corresponding to the displacement data series is , For the A point in time, is the total number of time points, which corresponds to the total number of displacement data points. time points The displacement data points are , , No. The time point is , is the time point corresponding to the abnormal value point or missing value point, , No. time points To time points The time distance is .calculate Weight ,The weight is usually calculated based on a distance function, such as , For the time points To time points The time distance. Here the denominator is a normalization factor that ensures that the sum of all weights is 1, i.e. The weighted linear interpolation formula is:

[0179] ;

[0180] in, for The interpolation result of the position; the meaning of this formula is that the interpolation result Value is another valid data point The weighted sum of It depends on the time distance between the data point and the missing value point. The closer the time distance is, the greater the weight of the data point is, and the greater the contribution to the interpolation result is.

[0181] For inclination, curvature, and strain, outlier detection and processing can be performed in the same way as displacement.

[0182] Step 2.3: Perform data normalization. The specific process involves normalizing each selected feature—data features collected by various sensors that reflect key indicators of the structure's movement and deformation, including displacement (horizontal and vertical), inclination, curvature, and strain—separately. Furthermore, it's important to consider dimensional differences and data ranges between sensor data to ensure that the normalization process retains the original information.

[0183] The minimum-maximum normalization method is used to map the data to the interval [0, 1]. Let the minimum value of the displacement data in the entire data set be , the maximum value is For each displacement data point in the dataset , and the normalized result is obtained by the formula Calculated, The displacement data points are normalized. This normalization method makes feature data of different magnitudes comparable and helps improve the efficiency and stability of model training. For example, a displacement data point may be at the millimeter level, while the strain data may be a small proportional value. Normalization can bring them into the same numerical range. When calculating the minimum and maximum values, ensure that the dataset does not contain outliers, otherwise the normalization results may be inaccurate. If outliers exist in the dataset, they should be processed before normalization.

[0184] For features such as inclination, curvature, and strain, data normalization can be performed in the same way as displacement features.

[0185] Step 3. Use finite element analysis to analyze the dynamic changes of surface movement and deformation. The core principle of finite element analysis is to discretize the continuous research object (such as the area containing surface and underground geological structures) into a finite number of small units, and finally solve the mechanical response of the entire system under given loads and boundary conditions by establishing the mechanical equilibrium equation at the unit level and considering the connection and interaction between units, so as to obtain the dynamic changes of surface movement and deformation. It is based on the variational principle or the weighted residual method to convert the partial differential equations describing surface deformation into a set of algebraic equations for solution. This process can simulate the stress-strain state of complex geological bodies under the influence of various factors (such as mining activities, groundwater level changes, structure loads, etc.) and the resulting surface displacement changes. The specific process is as follows:

[0186] Step 3.1: Construct a geometric model and use the unit partitioning strategy to discretize the units to obtain a discretized model. The specific process is as follows:

[0187] Step 3.1.1. Construct a 2D or 3D geometric model based on detailed geological survey data, including information on stratum distribution, lithologic parameters (such as elastic modulus, Poisson's ratio, and density), geological structures (such as faults and folds), and precise structural design blueprints (including structural form, dimensions, and material properties). For example, when studying the impact of coal mining on surface structures, it is necessary to determine the depth, thickness, and inclination of the coal seam; the shape and extent of the goaf; and the location, foundation type, and height of the surface structure. For 3D models, it is also necessary to account for topographical fluctuations. This can be accomplished by obtaining terrain elevation information from survey data or a geographic information system (GIS).

[0188] Step 3.1.2, the process of unit division strategy is: select the appropriate unit type and divide it according to the characteristics of the research problem and the accuracy requirements. After the division is completed, the unit discretization is realized, and then the discretized model is obtained. For complex shapes and stress concentration areas, triangular units are more advantageous. Let the total number of units be , the total number of nodes is For each unit, the three nodes are numbered 、 、 , No. The displacement vector of any node in a unit is , is the horizontal displacement, is the vertical displacement, 、 Respectively The horizontal displacement and vertical displacement of any node in a unit; ,node ,node The displacement vectors are 、 、 ; 、 Node Horizontal displacement and vertical displacement; 、 Node Horizontal displacement and vertical displacement; 、 Node Horizontal displacement and vertical displacement; is the transposed symbol. The displacement vector is connected through the shape function, which is a function based on the coordinates of the unit node. It has the characteristics of taking values ​​between 0 and 1 within the unit and taking values ​​of 1 at the corresponding node. For linear triangular elements, the node The shape function can be expressed as:

[0189] ;

[0190] ;

[0191] ;

[0192] ;

[0193] in, For nodes The shape function of is the area of ​​the triangular unit cell; For computing nodes The shape function is the intermediate parameter constructed; For computing nodes The correlation coefficient when the shape function is ; For computing nodes The correlation coefficient when the shape function is ; is the abscissa representing the horizontal position in the plane rectangular coordinate system; is the ordinate representing the vertical position in the plane rectangular coordinate system; The nodes in the triangle element The horizontal axis of The nodes in the triangle element The horizontal axis of The nodes in the triangle element The vertical coordinate of The nodes in the triangle element The vertical coordinate of .

[0194] node Shape function ,node Shape function With node The shape functions of are expressed in the same way;

[0195] Final The displacement vector of any node in a unit can be expressed as:

[0196] ;

[0197] At the same time, the strain-displacement relationship is:

[0198] ;

[0199] in, For the The strain vector of any node in the element; is the strain-displacement matrix, which consists of the partial derivatives of the shape function with respect to the coordinates and is formulated as:

[0200] ;

[0201] Step 3.2: Establish the constitutive equation, equilibrium equation and overall stiffness matrix. The specific process is as follows:

[0202] Step 3.2.1, determine the constitutive equation. Determine the constitutive equation based on the characteristics of the geological material. For elastic materials, the stress-strain relationship follows the generalized Hooke's law, and the constitutive equation can be expressed as:

[0203] ;

[0204] in, is the stress tensor; is the strain tensor; is the elastic matrix. For isotropic elastic materials, the elastic matrix The expression is:

[0205] ;

[0206] in, is the elastic modulus; is Poisson's ratio; these parameters are determined through geological surveys and material tests.

[0207] No. The constitutive equation of each unit is:

[0208] ;

[0209] in, For the The stress tensor of any node in an element;

[0210] Step 3.2.2, establish the equilibrium equation and overall stiffness matrix. Based on the principle of virtual work, establish the equilibrium equation. For the discretized model obtained in step 3.1, in the virtual displacement Under this condition, the virtual work of external forces is equal to the virtual work of internal forces. The virtual work of external forces includes the virtual work done by surface forces and concentrated forces, while the virtual work of internal forces is the work done by the internal stress of the unit on the virtual strain. The specific expression of the equilibrium equation is:

[0211] ;

[0212] in, It is The volume of a unit; For virtual strain; is the strain vector; is the stress tensor; is the volume; It is The surface of each unit; is the surface force vector; For the surface; is the displacement vector; is a node The force vector.

[0213] The stress-strain relationship Substitute into the equilibrium equation and use the strain-displacement relationship , the unit integral is converted into the relationship between the nodal force and displacement through the shape function, and the overall stiffness equation is obtained:

[0214] ;

[0215] in, It is the overall stiffness matrix, which is assembled from the stiffness matrices of each unit. The elements in the overall stiffness matrix are Representation node Unit displacement at the node The force generated on is the node displacement vector; is the force vector of the node. The stiffness matrix of each element The calculation formula is:

[0216] ;

[0217] Global stiffness matrix The assembly process is to superimpose the stiffness matrix of each unit to the corresponding position of the overall stiffness matrix according to the corresponding relationship between the unit node and the overall node. For example, if The three nodes of a unit are numbered 、 、 , each element in the The stiffness matrix of each element The location is , these elements are added to the overall stiffness matrix The corresponding position Among them, For the Nodes of units Unit displacement at the node The force generated on

[0218] Step 3.3: Solve the overall stiffness matrix and analyze the results to obtain the dynamic change law of surface movement and deformation. The specific process is as follows:

[0219] Step 3.3.1. Select the solution method. Solve the overall stiffness equation Get the node displacement vector Direct methods (such as Gaussian elimination) or iterative methods (such as the conjugate gradient method and the Jacobi iteration method) are typically used for solving problems. Direct methods are more efficient for small and medium-scale problems, but for large sparse matrices (common in finite element analysis), iterative methods are often more advantageous because they can exploit the sparsity of the matrices to reduce computational complexity and storage requirements. During the solution process, boundary conditions must be considered. For example, for fixed boundaries on the ground, the corresponding node displacement components are set to zero. For symmetric boundaries, symmetry conditions can be used to reduce computational complexity.

[0220] Step 3.3.2, result analysis. After obtaining the node displacement vector, the strain vector and stress tensor of the unit can be further calculated. The unit strain vector can be calculated based on the strain-displacement relationship, and the unit stress tensor is calculated by the constitutive equation. For surface movement deformation analysis, focus on the displacement of the surface nodes. For example, when calculating the vertical displacement of a point on the surface, it is obtained by interpolating the node displacement and shape function of the unit where the point is located. Suppose the coordinates of a point on the surface are , the unit is , then the vertical displacement of the point , For nodes Finite element analysis at different time points or under different working conditions can reveal the dynamic patterns of surface movement and deformation, such as the development of surface subsidence basins and the direction and magnitude of displacement vectors. These results can be exported as data files for subsequent use as input in a stacked LSTM-based structure movement and deformation prediction model. Furthermore, the distribution and changes in stress concentration areas can be analyzed to determine whether they will lead to geological damage or instability of the structure foundation, providing a basis for engineering design and safety assessment. Based on the patterns derived from the analysis, combined with factors such as the relative position of the structure to the mining area and the similarity of geological conditions, the movement and deformation of the structure can be predicted.

[0221] Step 4: Establish a structure movement and deformation prediction model. Establish a structure movement and deformation prediction model based on movement and deformation observation data mining. When constructing the model, the relevant parameters in the surface movement and deformation observation data are used as input variables, and the movement and deformation value of the structure is used as the output variable. The relevant parameters include displacement (horizontal and vertical displacement), inclination, curvature, strain, etc.

[0222] When predicting the movement and deformation of structures, the deformation process is influenced by a variety of complex factors, including the structure's inherent structural characteristics (including its materials and structural form), external loads (wind loads, earthquakes, thermal stresses caused by temperature fluctuations), and geological variations (foundation settlement, soil creep, etc.). These factors intertwine over time, resulting in complex time series characteristics in structure movement and deformation data. The stacked LSTM model effectively captures these complex time series relationships, enabling accurate deformation prediction. While LSTMs can handle long-term dependencies in time series data, their performance may be limited when dealing with very long sequences or complex interactions between multiple factors. Combining the Transformer layer with the LSTM layer leverages the Transformer layer's strengths to complement the LSTM's shortcomings. For example, while LSTMs may gradually forget early information when processing long sequences, the Transformer layer's ability to capture long-range dependencies allows it to recover this important early information and incorporate it into subsequent feature representations. This combination allows the model to more comprehensively and accurately consider the impact of various factors at different times when predicting structure movement and deformation, thereby improving prediction accuracy.

[0223] The model consists of multiple stacked LSTM layers, a Transformer layer, and a fully connected layer. The Transformer is a sequence model based on an attention mechanism. First, each LSTM layer performs preliminary processing on the input data using its unique gating mechanism. Leveraging its expertise in processing time series data, it captures local dependencies and short-term trends in the data. As the number of LSTM layers increases, the model gradually extracts more representative and abstract features, initially uncovering time series patterns in the data. Subsequently, the Transformer layer plays a key role. Its multi-head attention mechanism processes the feature sequences output by the LSTM layers in parallel, spanning different locations and time points, fully capturing long-range dependencies in the data and effectively compensating for the information attenuation that LSTMs may experience when processing long sequences. Furthermore, the Transformer layer uses a self-attention mechanism to assign different weights to different feature dimensions, focusing on more critical information in the data and further extracting global features and complex interactions. It integrates the local features extracted by the LSTM with broader contextual information, thereby building a more comprehensive and in-depth understanding of the data. This layered structure, combining stacked LSTMs and Transformers, forms an efficient hierarchical learning system for data features. Starting from raw multi-source sensor data, it gradually mines deep structural information and hidden patterns, effectively addressing the complex deformation patterns caused by the interweaving of multiple factors during the movement and deformation of structures, providing a solid foundation for accurately predicting movement and deformation trends.

[0224] The working process of the structure movement and deformation prediction model is as follows:

[0225] Step 4.1: Input the relevant parameters of the surface movement and deformation observation data into the first layer of LSTM. The working process of the first layer of LSTM is as follows:

[0226] Step 4.1.1, preprocessing and initialization. Preprocess the input parameters. After preprocessing, the input parameters are initialized at each time point. Is a vector that defines the first The input data for each time step is ,in is the dimension of the input data, i.e. the number of selected sensor data features; For the time step, The At the same time, initialize the initial hidden state of the first layer LSTM is a zero vector with dimension (For example, you can set , the specific value can be determined according to model adjustment and experiment). is the number of hidden units in the LSTM layer; the initial cell state Also initialized to a zero vector, the dimension is also The initialization value here is chosen to give the model a starting state. The zero vector is a simple and commonly used initialization method, but in some cases, random initialization or initialization based on a specific distribution can also be used to improve the performance of the model.

[0227] Step 4.1.2, perform forget gate calculation. The calculation of forget gate involves the concatenation of the input data at the current time point and the hidden state at the previous time point. Input data for time steps and the first layer of LSTM The hidden state of time steps Splice into a vector , whose dimensions are Then, we perform matrix multiplication with the weight matrix of the forget gate, add the bias vector, and finally use the Sigmoid activation function to get the output of the forget gate:

[0228] ;

[0229] in, is the forget gate in the first layer of LSTM Output of time steps; is the Sigmoid activation function; is the weight matrix of the forget gate in the first layer of LSTM, with dimension , The initial values ​​of the elements can be generated by random initialization methods (such as Xavier initialization or He initialization). These initialization methods can reasonably set the initial range of weights according to the input and output dimensions to help the model converge faster; is the bias vector of the forget gate in the first layer of LSTM, with a dimension of .

[0230] During training, and The value of will be continuously adjusted according to the gradient descent algorithm. Each element of The cell state at each time step A dimension of is used to determine the degree of retention of the dimension information, 0 means completely discarded, 1 means completely retained. For example, if If an element of is close to 0, then the corresponding The information in the dataset will be largely forgotten at the current time point. This forgetting mechanism is one of the keys to LSTM's ability to handle long-term dependencies. It can prevent old, unimportant information from accumulating over long time series, thereby affecting the model's predictive ability.

[0231] Step 4.1.3, perform input gate calculation to generate the input gate control signal and candidate cell state. The calculation method of the input gate control signal is similar to that of the forget gate. Multiply it with the weight matrix of the input gate, add the bias vector, and finally pass it through the Sigmoid activation function to get:

[0232] ;

[0233] in, is the first layer of LSTM The control signal of the input gate in time steps; is the weight matrix of the input gate in the first layer of LSTM, with dimension ; is the bias vector of the input gate in the first layer of LSTM; and The initialization method is similar to the parameters of the forget gate. The value of is between 0 and 1 and is used to control the degree to which new information flows into the cell state. When an element of is close to 1, it means that more new information is allowed to enter the cell state; when it is close to 0, it means that the inflow of new information is restricted.

[0234] Generate the candidate cell state at the current time point through a tanh activation function layer. Multiply it by the weight, add the bias vector, and then pass it through the tanh activation function to get:

[0235] ;

[0236] in, is the first layer of LSTM Candidate cell states for time steps, Contains the potential information contained in the current time step input data and the hidden state of the previous time step, and its value is between -1 and 1. For example, if a displacement feature value in the input data changes significantly, this change will be reflected in the multiplication of the weight matrix and the action of the tanh function. In this way, the model can dynamically adjust the candidate cell states according to the changes in input data and prepare for the update of the cell state. is the tanh activation function; is the weight matrix used to calculate the candidate cell state in the first layer of LSTM, with a dimension of ; is the bias vector used to calculate the candidate cell state in the first layer of LSTM, with a dimension of .

[0237] Step 4.1.4. Update the cell state according to the output of the forget gate and input gate:

[0238] ;

[0239] in, is the first layer of LSTM Update cell state for each time step; Represents element-wise multiplication operation; is the first layer of LSTM The cell state of the time step. This update process is to first selectively forget part of the information in the cell state of the previous time point according to the output of the forget gate, and then integrate the new candidate cell state information according to the output of the input gate. For example, if A dimension value of 0.3, If the value of this dimension is 5, then when updating, this part of information will become 0.3×5=1.5; at the same time, if The value in the corresponding dimension is 0.8, If the value of this dimension is 0.6, the newly incorporated information is 0.8 × 0.6 = 0.48, and the final updated cell state value for this dimension is 1.5 + 0.48 = 1.98. This cell state update method allows the model to flexibly adjust the degree of utilization of historical and new information at each time point, thereby better handling changes in time series data.

[0240] Step 4.1.5, perform output gate calculation to generate the output gate control signal and output hidden state. The calculation method of the output gate control signal is similar to that of the forget gate and input gate. Multiply it with the weight matrix of the output gate, add the bias vector, and finally pass it through the Sigmoid activation function to get:

[0241] ;

[0242] in, is the first layer of LSTM The control signal of the output gate in time steps; is the weight matrix of the output gate in the first layer of LSTM, with a dimension of ; is the bias vector of the output gate in the first layer of LSTM;

[0243] Will Through the tanh function processing, a value between -1 and 1 is obtained, and then Multiply element by element to get the hidden state at the current time point:

[0244] ;

[0245] in, is the first layer of LSTM The hidden state of the time step contains not only the input information of the current time point, but also the information of the previous time point. It will be used as one of the inputs of the next time step and also as the output of the first layer LSTM to the next layer LSTM. For example, if A dimension value of 0.7, If the value of this dimension is 0.8, then The value of this dimension is 0.7 × 0.8 = 0.56. Through the output gate, the model can control the information contained in the hidden state, so that only information relevant to the current prediction task is passed to the next layer or used as output, further improving the model's ability to filter and utilize information.

[0246] Step 4.2, Enter the second layer of LSTM. Its calculation process is similar to the first layer of LSTM, but the weight matrix and bias vector are different to adapt to the new input data dimension and further extract features. The calculation formula of the second layer of LSTM is as follows:

[0247] ;

[0248] ;

[0249] ;

[0250] ;

[0251] ;

[0252] ;

[0253] in, The forget gate in the second layer LSTM Output of time steps; is the weight matrix of the forget gate in the second layer LSTM, with dimension (Because the input is the hidden state of the previous layer, the dimension is ); is the bias vector of the forget gate in the second layer LSTM, with dimension . is the first The hidden state of time steps. Here and It is also generated by a suitable initialization method and adjusted during the training process. Similar to the forget gate in the first layer of LSTM, Each element determines the state of the second layer cell The degree of forgetting of the corresponding dimensional information is based on the output of the previous layer and the hidden state of the previous time step of the current layer. is the first The control signal of the input gate in time steps; is the weight matrix of the input gate in the second layer of LSTM; is the bias vector for the input gate in the second LSTM layer. These two parameters determine the degree of control over the flow of new information into the cell states of the second layer. Initialization and update methods are similar to those for the first layer, but are tailored to the dimensionality and structure of the input data in the second layer. is the first Candidate cell states for time steps; It is the weight matrix used to calculate the candidate cell state in the second layer of LSTM; is the bias vector used to calculate the candidate cell state in the second layer LSTM. Through the calculation method similar to the first layer, according to the current input (including the first layer LSTM The hidden state of time steps and the second LSTM layer The hidden state of time steps The candidate cell state is generated by concatenating the first layer and its value range is between -1 and 1, which contains the potential update content after further processing of the information passed from the first layer. is the first Update cell state for each time step; is the first This update process is similar to that of the first layer. Through the forget gate and input gate, the cell state of the second layer LSTM is updated, integrating partial information from the previous time step with the newly generated candidate information of the current time step. This allows the cell state of the second layer LSTM to capture more complex changes in time series features. These features are further abstracted from the original input data processed by the first layer LSTM. is the first The control signal of the output gate in time steps; is the weight matrix of the output gate in the second layer of LSTM; is the bias vector of the output gate in the second LSTM layer. It generates a control signal based on the current input and determines the output level of the current cell state.

[0254] In the second LSTM layer The hidden state of time steps This layer further captures more complex time series features, which serve as input to the third LSTM layer. Compared to the first LSTM layer, the second LSTM layer can extract deeper information from the features extracted by the first layer. For example, it can capture more complex interactions between different sensor data features and long-term dependency patterns in the time series. This hierarchical feature extraction capability enables the model to better understand the complex patterns in structure movement and deformation data. As the number of layers increases, the ability to abstract data also increases.

[0255] Step 4.3, The calculation process of the first layer LSTM is similar to that of the second layer, except that the input becomes The output hidden state of the layer. The calculation formula of the layer LSTM is as follows:

[0256] ;

[0257] ;

[0258] ;

[0259] ;

[0260] ;

[0261] ;

[0262] in, For the Forgetting level in layer LSTM Output of time steps; It is The weight matrix of the forget gate in the LSTM layer has a dimension based on the In the LSTM layer The hidden state of time steps Dimensions and The hidden state dimension in the LSTM layer is determined by ; It is The bias vector of the forget gate in the LSTM layer has the dimension . Decided from the In the LSTM layer Time step cell state The calculation of which information to discard depends on the output of the L-1 layer and the hidden state of the L layer in the previous time step. The Sigmoid activation function is used to map the result to between 0 and 1 to control the forgetting of information. For the In the LSTM layer The control signal of the input gate in time steps; It is The weight matrix of the input gate in the layer LSTM; It is The bias vector of the input gate in the LSTM layer. This control signal is used to control the flow of new information into the first The degree of cell state in the layer is determined based on the current input (including the The output of the layer LSTM and the The hidden state of the previous time step in the LSTM layer is calculated using specific weights and biases and activated by the Sigmoid function. For the In the LSTM layer Candidate cell states for time steps; It is The weight matrix used to calculate the candidate cell state in the LSTM layer; It is The bias vectors used in the LSTM layer to calculate candidate cell states are generated using the tanh activation function. These values ​​reflect the potential update information contained in the current input, providing new context for updating the cell state. For the In the LSTM layer Update cell state for each time step; It is In the LSTM layer According to the same update logic as the first and second layers, the cell state is updated according to the output of the forget gate and the input gate, so that the first The cell state of the LSTM layer fuses historical information and new candidate information to further mine deep features in the data. For the In the LSTM layer The control signal of the output gate in time steps; It is The weight matrix of the output gate in the layer LSTM; It is The bias vector of the output gate in the LSTM layer. The control signal is calculated based on the current input to determine the output of the cell state. In the LSTM layer The hidden state of the time step The state serves as a higher-level feature representation of the entire stacked LSTM layer, which includes complex time series features extracted from the original input data after multi-layer processing. is the total number of time steps, and the final output of the entire stacked LSTM layer is In the LSTM layer The hidden state of time steps .

[0263] Step 4.4, Enter the Transformer layer. The Transformer layer uses a multi-head attention mechanism and a feedforward neural network. The specific process is:

[0264] Step 4.4.1. First, the final output of the stacked LSTM layer Perform linear transformation to obtain the query matrix , key matrix Sum Matrix .set up 、 、 They are 、 、 The weight matrix is:

[0265] ;

[0266] ;

[0267] ;

[0268] For example, if is a Matrix, assuming , , 、 、 are all 64×64 matrices, then 、 、 Both are 100×64 matrices.

[0269] Then, calculate the attention score:

[0270] ;

[0271] in, Score for attention; is the softmax function; yes Dimensions; is the transposed symbol. In order to obtain the feature representation of different subspaces, a multi-head attention mechanism is usually adopted. 、 、 Split into head (e.g. ), each head calculates the attention score and then concatenates them. 、 、 For the The query matrix, key matrix, and value matrix of each head, then The attention scores of the individuals are:

[0272] ;

[0273] in, For the Attention score of size;

[0274] Finally, the output matrix of the multi-head attention for:

[0275] ;

[0276] in, For splicing operation; It is the weight matrix used to fuse multi-head information; For the For example, 、 、 Split into 8 heads by column, each head has a dimension of 100×8, and calculate the attention score of each head to obtain 8 100×8 matrices. Assume It is a 64×64 matrix, and after splicing, a 100×64 output matrix is ​​obtained .

[0277] Step 4.4.2, Perform feedforward neural network processing, the formula is:

[0278] ;

[0279] in, It is a feedforward neural network; 、 are two different weight matrices of a feedforward neural network; 、 are two different bias vectors for the feedforward neural network. For example, if is a 100×64 matrix, is a 64×256 matrix, is a 256-dimensional vector, is a 256×64 matrix, Is a 64-dimensional vector, then first calculate Get a 100×256 matrix and pass it through the ReLU activation function Then, Multiply and add The final output is still of dimension 100×64.

[0280] The Transformer layer can further explore the global features and complex relationships in the data through the multi-head attention mechanism and feedforward neural network, supplementing the shortcomings of LSTM in processing long sequences.

[0281] Step 4.5, Input to the fully connected layer to predict the movement and deformation of the structure. Output matrix of the multi-head attention of the Transformer layer As a high-level feature representation of the entire model. According to the prediction target (such as predicting the displacement of a certain point of the structure, the tilt angle, etc.), add a fully connected layer (Dense layer) to The features are converted into specific prediction values. Let the weight matrix of the fully connected layer be , the dimension is (in is the dimension of the predicted target, for example, if only displacement is predicted, can be 1), the bias vector of the fully connected layer is , the dimension is . No. The predicted value of the time step The calculation formula is:

[0282] ;

[0283] When constructing a fully connected layer, the weight matrix The initialization of is also important. You can use a random initialization method and adjust it appropriately according to the specific activation function (if any) and input and output dimensions. Initialize to a small value, such as a constant close to 0. The fully connected layer maps the complex features contained in the output of the Transformer layer to the space of the prediction target, realizing the conversion from feature representation to specific prediction value.

[0284] Step 5: Train and verify the structure movement and deformation prediction model to ensure that the model has high accuracy and reliability. Use the trained model to predict structure movement and deformation. The specific process is as follows:

[0285] Step 5.1, loss function selection. For the problem of structure movement and deformation prediction, the commonly used loss function is the mean square error (MSE) loss function, which measures the average value of the square error between the predicted value and the true value. Suppose the predicted value sequence is , the true value sequence is , then the calculation formula of the mean square error loss function is:

[0286] ;

[0287] in, is the mean square error loss function value; For the The true value of the time step; For the The predicted value of each time step;

[0288] The model is trained by minimizing the MSE loss function, ensuring that its predictions are as close as possible to the actual structure movement and deformation. MSE squares the error, causing larger errors to account for a larger portion of the loss calculation, thus forcing the model to focus more on reducing large errors.

[0289] In addition, the mean absolute error (MAE) loss function is used as a supplementary evaluation indicator, and its calculation formula is:

[0290] ;

[0291] in, =MAE is the mean absolute error loss function; MAE is more robust to outliers because it does not perform a squaring operation and does not over-amplify the impact of large errors. Combining it with MSE provides a more comprehensive assessment of model performance. During model training, observe the trends of both MSE and MAE simultaneously. If MSE decreases while MAE increases, it may indicate a decline in the model's ability to handle outliers, necessitating further data analysis or model structure adjustments.

[0292] Step 5.2: Use the Adaptive Moment Estimation (Adam) optimization algorithm to update the model parameters (specifically, the weight matrix and bias vector of each LSTM layer, Transformer layer, and output layer). The Adam optimization algorithm combines the idea of ​​momentum method and adaptive learning rate. During the training process, the Adam algorithm automatically adjusts the learning rate of each parameter based on the first-order moment estimate and second-order moment estimate of the gradient, allowing the model to converge faster and more stably. Specifically, the Adam algorithm maintains two momentum vectors for each parameter: The first-order moment vector of the time step (similar to speed) and The second-order moment vector of time steps (Similar to acceleration). For the parameter (This includes all weight matrices and bias vectors), in the The gradient of the time step is , the learning rate is , and are two different hyperparameters used to control momentum (usually close to 1, such as , ), is a non-zero constant (to prevent division by zero, such as ), the update process is as follows:

[0293] First, calculate the update of the first-order moment vector:

[0294] ;

[0295] in, For the The first-order moment vector of the time step; this step is the exponentially weighted moving average of the gradient, and the new first-order moment vector is calculated based on the current gradient and the first-order moment vector of the previous time point, where Determines the degree of retention of past gradient information.

[0296] Then, calculate the update of the second-order moment vector:

[0297] ;

[0298] in, For the The second-order moment vector of time steps; this is the exponentially weighted moving average of the square of the gradient, which is used to measure the change of the gradient. It also determines the degree of retention of past information.

[0299] Next, the first-order moment vector and the second-order moment vector are corrected for deviation:

[0300] ;

[0301] ;

[0302] in, 、 Respectively The first-order moment vector deviation and the second-order moment vector deviation of each time step are calculated; since the setting of the initial value when calculating the moving average of the first-order moment vector and the second-order moment vector will cause deviations in the early estimation, the deviation correction step can reduce the impact of this deviation, especially in the early stage of training.

[0303] Finally, update the parameters:

[0304] ;

[0305] in, 、 Respectively time step, The parameters of the time step are updated according to the modified first-order moment vector and second-order moment vector and the learning rate, so that the parameters move in the direction of decreasing the loss function.

[0306] Step 5.3: Hyperparameter adjustment. The specific process is as follows:

[0307] Step 5.3.1. For hyperparameter tuning, use grid search to form different hyperparameter combinations. Adjust the learning rate , the number of hidden units in the LSTM layer and the number of heads in the Transformer layer , set a series of Value (such as 0.001, 0.005, 0.01, etc.), Values ​​(such as 32, 64, 128, etc.) and values ​​(such as 4, 8, 16, etc.), and then evaluate the performance of the model under different combinations on the validation set. The evaluation indicators are MSE and MAE, and the hyperparameter combination with the best performance is selected.

[0308] In step 5.3.2, to mitigate overfitting, add a Dropout layer after each LSTM layer. Set an appropriate probability threshold (e.g., 0.2-0.5). During training, these Dropout layers will randomly set the outputs of some neurons to 0, reducing interdependence between neurons and improving the model's generalization. Also, consider adding Dropout layers to certain parts of the Transformer layer (e.g., after the multi-head attention mechanism or after the feedforward neural network). The probability threshold can be the same as that for the Dropout layer after the LSTM layer or set independently. Determine whether to add Dropout layers to certain parts of the Transformer layer by evaluating them on the validation set. If the evaluation result exceeds the probability threshold, add the layer; otherwise, do not.

[0309] Step 5.3.3, determine the number of training rounds. The number of training rounds is an important hyperparameter. If the number of training rounds is too small, the model may not fully learn the patterns in the data, resulting in underfitting; if the number of training rounds is too large, it may lead to overfitting. Determine the appropriate number of training rounds by observing the loss curve on the validation set. Generally, the number of training rounds when the loss of the validation set reaches the minimum value is selected as the final number of training rounds. For example, during the training process, record the loss value of the validation set after each epoch and draw the loss curve. When the loss curve begins to rise or no longer decreases within multiple epochs, stop training and select the number of training rounds corresponding to the epoch with the lowest loss.

[0310] Step 5.3.4, setting up the learning rate scheduler. The learning rate scheduler adjusts the changes in the learning rate during the training process by adopting a learning rate scheduling strategy. Gradually reducing the learning rate as the number of training rounds increases can make the model converge more stably in the later stages of training. There are two types of learning rate scheduling strategies, namely the step decay strategy and the exponential decay strategy. For the step decay strategy, it is necessary to determine the decay step size (such as decaying once every 10 epochs) and the decay factor (such as 0.9). At each decay step, multiply the learning rate by the decay factor. The initial learning rate is ,go through After decay steps, the learning rate becomes For the exponential decay strategy, it is necessary to determine the decay rate (such as 0.95). The learning rate decreases exponentially with the number of training rounds, that is, , Represents the current training epoch. By evaluating the model's performance under different learning rate scheduling strategies and parameter settings on the validation set, we select the most appropriate learning rate scheduler parameters to further optimize the model's training process and improve its performance and generalization ability.

[0311] Step 6: Comprehensively evaluate safety and stability. The specific process is as follows: Based on the predicted values ​​of structure movement and deformation obtained by the above method, a comprehensive evaluation of the structure's safety and stability is conducted. The model's predicted values ​​are compared with the safety warning values ​​determined in Step 1. If the predicted values ​​do not exceed the warning values, the structure is currently in a relatively safe state. If the predicted values ​​exceed the warning values, it means that the structure has a safety risk and appropriate technical measures need to be taken, such as temporary reinforcement of the structure or adjustment of the mining plan.

[0312] Step 7: Modify the prediction model parameters and continue to evaluate until the structure is safely mined. This includes the following steps:

[0313] Step 7.1: During the mining process, continuously acquire the next stage of movement and deformation observation data of the ground surface and structures in the mined area. This new data includes the surface deformation closer to the structure during the mining process and the actual response of the structure itself;

[0314] Step 7.2: Substitute the newly acquired observation data into the prediction model and use the Adam optimization algorithm to modify the parameters of the prediction model so that the model can better adapt to the new data characteristics and actual deformation conditions.

[0315] Step 7.3: Use the revised prediction model to predict the movement and deformation of the structure again and re-evaluate its safety and stability;

[0316] Step 7.4: Repeat step 7.3, continuously adjusting the model and assessing the safety status until the working face safely passes through the structure, ensuring that the safety impact on the structure is within a controllable range throughout the mining process.

[0317] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the scope of protection of the present invention.

Claims

1. A method for digital intelligent observation of rock movement on working face and intelligent prediction of safety of passing structures, characterized by: The steps include: Step 1: Determine the safety warning value of structure movement and deformation; Step 2: Collect motion deformation data and perform preprocessing; Step 3: Use finite element analysis to analyze the dynamic changes of surface movement and deformation; the specific process is as follows: Step 3.1: Construct a geometric model and use the unit partitioning strategy to discretize the units to obtain a discretized model. The specific process is as follows: First, construct a two-dimensional or three-dimensional geometric model based on geological survey data and the design blueprint of the structure; then, use the unit partitioning strategy to discretize the two-dimensional or three-dimensional geometric model to obtain a discretized model; the process of the unit partitioning strategy is as follows: let the total number of units be , the total number of nodes is ; For each unit, the three nodes are numbered 、 、 , No. The displacement vector of any node in a unit is , is the horizontal displacement, is the vertical displacement, 、 Respectively The horizontal displacement and vertical displacement of any node in a unit; ,node ,node The displacement vectors are 、 、 ; 、 Node Horizontal displacement and vertical displacement; 、 Node Horizontal displacement and vertical displacement; 、 Node Horizontal displacement and vertical displacement; is the transpose symbol; the displacement vectors are connected through shape functions; For linear triangular elements, the nodes The shape function is expressed as: ； ； ； ； in, For nodes The shape function of is the area of ​​the triangular unit cell; For computing nodes The shape function is the intermediate parameter constructed; For computing nodes The correlation coefficient when the shape function is ; For computing nodes The correlation coefficient when the shape function is ; is the abscissa representing the horizontal position in the plane rectangular coordinate system; is the ordinate representing the vertical position in the plane rectangular coordinate system; The nodes in the triangle element The horizontal axis of The nodes in the triangle element The horizontal axis of The nodes in the triangle element The vertical coordinate of The nodes in the triangle element The vertical coordinate of node Shape function ,node Shape function With node The shape functions of are expressed in the same way; Final The displacement vector of any node in a unit is expressed as: ； At the same time, the strain-displacement relationship is: ； in, For the The strain vector of any node in the element; is the strain-displacement matrix, which is: ； Step 3.2, establish the constitutive equation, equilibrium equation and overall stiffness matrix; In step 3.2, The constitutive equation of each unit is: ； in, For the The stress tensor of any node in an element; is the elasticity matrix, expressed as: ； in, is the elastic modulus; is Poisson's ratio; The specific expression of the equilibrium equation is: ； in, It is The volume of a unit; For virtual strain; is the strain vector; is the stress tensor; is the volume; It is The surface of each unit; is the surface force vector; For the surface; is the displacement vector; is a node The force vector of The overall stiffness equation is: ； in, is the overall stiffness matrix; is the node displacement vector; is the force vector of the node; No. The stiffness matrix of each element The calculation formula is: ； Global stiffness matrix The assembly process is to superimpose the stiffness matrix of each unit to the corresponding position of the overall stiffness matrix according to the corresponding relationship between the unit node and the overall node; Step 3.3: Solve the overall stiffness matrix and analyze the results to obtain the dynamic variation law of surface movement and deformation; In step 3.3, the overall stiffness equation is solved using direct or iterative methods. Get the node displacement vector After obtaining the node displacement vector, the unit strain vector and stress tensor are further calculated; the unit strain vector is calculated based on the strain-displacement relationship, and the unit stress tensor is calculated by the constitutive equation; Step 4: Establish a structure movement and deformation prediction model; In step 4, the input variables of the structure movement and deformation prediction model are the relevant parameters in the surface movement and deformation observation data, and the output is the movement and deformation value of the structure; the relevant parameters include displacement, inclination, curvature, and strain; The structure movement and deformation prediction model consists of multiple stacked LSTM layers, a Transformer layer, and a fully connected layer. LSTM is a long short-term memory network. Transformer is a sequence model based on the attention mechanism. The working process of the structure movement and deformation prediction model is as follows: Step 4.1: Input the relevant parameters of the surface movement and deformation observation data into the first layer of LSTM. The working process of the first layer of LSTM is as follows: Step 4.1.1: Preprocess the input parameters. After preprocessing, the input parameters are a vector at each time point. Define the first vector after preprocessing. The input data for each time step is ,in is the dimension of the input data; For the time step, The Data of dimensions; Step 4.1.2: Calculate the forget gate. The formula is: ； in, is the forget gate in the first layer of LSTM Output of time steps; is the Sigmoid activation function; is the weight matrix of the forget gate in the first layer of LSTM; is the first layer of LSTM The hidden state of time steps; is the bias vector of the forget gate in the first layer of LSTM; Step 4.1.3: Perform input gate calculation to generate the input gate control signal and candidate cell state. The formula is: ； ； in, is the first layer of LSTM The control signal of the input gate in time steps; is the weight matrix of the input gate in the first layer of LSTM; is the bias vector of the input gate in the first layer of LSTM; is the first layer of LSTM Candidate cell states for time steps; is the tanh activation function; is the weight matrix used to calculate the candidate cell state in the first layer of LSTM; is the bias vector used to calculate the candidate cell state in the first layer of LSTM; Step 4.1.

4. Update the cell state according to the output of the forget gate and input gate: ； in, is the first layer of LSTM Update cell state for each time step; Represents element-wise multiplication operation; is the first layer of LSTM The cell state at each time step; Step 4.1.5: Perform output gate calculation to generate the output gate control signal and output hidden state. The formula is: ； ； in, is the first layer of LSTM The control signal of the output gate in time steps; is the weight matrix of the output gate in the first layer of LSTM; is the bias vector of the output gate in the first layer of LSTM; is the first layer of LSTM The hidden state of the time step; Step 4.2, Enter the second layer LSTM. The calculation formula of the second layer LSTM is as follows: ； ； ； ； ； ； in, The forget gate in the second layer LSTM Output of time steps; is the weight matrix of the forget gate in the second layer of LSTM; is the bias vector of the forget gate in the second layer of LSTM; is the first The hidden state of the time step; is the first The control signal of the input gate in time steps; is the weight matrix of the input gate in the second layer of LSTM; is the bias vector of the input gate in the second layer of LSTM; is the first Candidate cell states for time steps; It is the weight matrix used to calculate the candidate cell state in the second layer of LSTM; is the bias vector used to calculate the candidate cell state in the second layer of LSTM; is the first Update cell state for each time step; is the first The cell state at each time step; is the first The control signal of the output gate in time steps; is the weight matrix of the output gate in the second layer of LSTM; is the bias vector of the output gate in the second layer of LSTM; In the second LSTM layer The hidden state of time steps Will serve as the input of the third layer LSTM; Step 4.3, The input of the LSTM layer is The output hidden state of the layer; The calculation formula of the layer LSTM is as follows: ； ； ； ； ； ； in, For the Forgetting level in layer LSTM Output of time steps; It is The weight matrix of the forget gate in the layer LSTM; It is The bias vector of the forget gate in the layer LSTM; For the In the LSTM layer The control signal of the input gate in time steps; It is The weight matrix of the input gate in the layer LSTM; It is The bias vector of the input gate in the layer LSTM; For the In the LSTM layer Candidate cell states for time steps; It is The weight matrix used to calculate the candidate cell state in the LSTM layer; It is The relevant bias vector used to calculate the candidate cell state in the layer LSTM; For the In the LSTM layer Update cell state for each time step; It is In the LSTM layer The cell state at each time step; For the In the LSTM layer The control signal of the output gate in time steps; It is The weight matrix of the output gate in the layer LSTM; It is The bias vector of the output gate in the layer LSTM; For the In the LSTM layer The hidden state of the time step; is the total number of time steps, and the final output of the entire stacked LSTM layer is In the LSTM layer The hidden state of time steps ; Step 4.4, Input Transformer layer and get output matrix ; The Transformer layer uses a multi-head attention mechanism and a feedforward neural network; the specific process is: Step 4.4.

1. First, the final output of the stacked LSTM layer Perform linear transformation to obtain the query matrix , key matrix Sum Matrix ;set up 、 、 They are 、 、 The weight matrix is: ； ； ； Then, suppose 、 、 For the The query matrix, key matrix, and value matrix of each head, then The attention scores of the individuals are: ； in, For the Attention score of size; is the softmax function; yes Dimensions; is the transpose symbol; is the number of heads; Finally, the output matrix of the multi-head attention for: ； in, For splicing operation; It is the weight matrix used to fuse multi-head information; For the Attention score of size; Step 4.4.2, Perform feedforward neural network processing, the formula is: ； in, It is a feedforward neural network; 、 are two different weight matrices of a feedforward neural network; 、 are two different bias vectors for the feedforward neural network; Step 4.5, Input the fully connected layer to predict the movement and deformation of the structure; In step 4.5, the fully connected layer will The features are converted into specific prediction values; The predicted value of the time step The calculation formula is: ； in, is the weight matrix of the fully connected layer; is the bias vector of the fully connected layer; Step 5: Train and verify the structure movement and deformation prediction model, and use the trained model to predict the structure movement and deformation; Step 6: Comprehensively evaluate safety and stability; Step 7: Modify the prediction model parameters and continue to evaluate until the structure is safely mined.

2. The method for digital intelligent observation of rock movement on working face and intelligent prediction of safety of overpassing structures according to claim 1 is characterized in that: The specific process of step 1 is: Step 1.1: Comprehensively consider the structural characteristics of the structure, its geological conditions, and the surrounding environment to identify the key factors affecting the stability and safety of the structure; Key factors include the type of foundation, building material properties, overall structural stiffness, geological characteristics, and other surrounding engineering activities; Step 1.2: Identify the key indicators that affect the movement and deformation state of the structure, including displacement, inclination, curvature, and strain; combine the structure's design specifications, usage requirements, and relevant industry standards to determine the threshold value of each key indicator. This threshold value is the safety warning value for the movement and deformation of the structure.

3. The method for digital intelligent observation of surface rock movement and intelligent prediction of safety of overpass structures according to claim 1 is characterized in that: The specific process of step 2 is: Step 2.1: Install various types of sensors at key structural locations of the structure to collect daily surface movement and deformation data in the mined area in front of the structure. The sensors include displacement sensors, inclination sensors, curvature sensors, and strain sensors. Step 2.2: Detect and process outliers. The process of outlier detection is as follows: define the displacement data sequence collected by the displacement sensor as , calculate the mean of the displacement data series and standard deviation , the formula is as follows: ； ； in, is the total number of displacement data points; For the displacement data points; For the displacement data points; for The displacement data points of the interval are marked as outliers; is the standard deviation The income from the prescription; The process of outlier processing is as follows: For data with a stable trend, the linear interpolation method is used to handle outliers. Assume that at the first time point and the third time point The displacement data points are and , and the second time point If the displacement data points are abnormal or missing, Displacement data points It is calculated by the linear interpolation formula, which is: ； For data with unstable trend, the weighted linear interpolation method is used to deal with outliers. Middle displacement data points The displacement data points identified as outliers and removed are called missing value points. Interpolate the position; let the time series corresponding to the displacement data series be , For the A point in time, is the total number of time points, one displacement data point corresponds to one time point, and the total number of time points corresponds to the total number of displacement data points; the weighted linear interpolation formula is: ； ； in, for The interpolation result of the location; for The weight of For the time points To time points Time distance; For the time points To time points Time distance; For the time points The displacement data points; Step 2.3: Normalize the data. The formula is: ； in, is the normalized displacement data point; is the displacement data point; 、 are the minimum and maximum values ​​of the displacement data in the entire data set; For inclination, curvature, and strain, outlier detection and processing, and data normalization are performed in the same way as displacement data.

4. The method for digital intelligent observation of surface rock movement and intelligent prediction of safety of overpass structures according to claim 3 is characterized in that: The specific process of step 5 is as follows: Step 5.1: Train the model by minimizing the mean square error loss function; let the predicted value sequence be , the true value sequence is , then the calculation formula of the mean square error loss function is: ； in, is the mean square error loss function value; For the The true value of the time step; For the The predicted value of each time step; At the same time, the mean absolute error loss function is used as a supplementary evaluation indicator, and the calculation formula is: ； in, is the mean absolute error loss function value; Step 5.2: Update the model parameters using the adaptive moment estimation optimization algorithm. The parameters include the weight matrix and bias vector of each LSTM layer, Transformer layer, and output layer. The update process is as follows: First, calculate the update of the first-order moment vector: ； in, For the The first-order moment vector of the time steps; For the The first-order moment vector of the time steps; For the The gradient of time steps; Then, calculate the update of the second-order moment vector: ； in, For the The second-order moment vector of time steps; For the The second-order moment vector of time steps; and are two different hyperparameters used to control momentum; Next, the first-order moment vector and the second-order moment vector are corrected for deviation: ； ； in, 、 Respectively First-order moment vector deviation and second-order moment vector deviation of each time step; Finally, update the parameters: ； in, 、 Respectively time step, Parameters of time steps; is the learning rate; is a non-zero constant; Step 5.3, hyperparameters include learning rate , the number of hidden units in the LSTM layer and the number of heads in the Transformer layer , set a series of value, Value and The evaluation indicators are mean square error and mean absolute error, and the best hyperparameter combination is selected. The number of training rounds when the loss of the validation set reaches the minimum is selected as the final number of training rounds. The learning rate scheduling strategy is used to adjust the learning rate.

5. The method for digital intelligent observation of rock movement on working face and intelligent prediction of safety of overpassing structures according to claim 1 is characterized in that: The specific process of step 6 is: comparing the model prediction value with the safety warning value determined in step 1. When the prediction value does not exceed the warning value, it indicates that the structure is currently in a safe state; when the prediction value exceeds the warning value, it indicates that there is a safety risk in the structure.

6. The method for digital intelligent observation of rock movement on working face and intelligent prediction of safety of overpass structures according to claim 5 is characterized in that: The specific process of step 7 is as follows: Step 7.1: During the mining process at the working face, continuously obtain the next stage of movement and deformation observation data of the surface and structures in the mined area; Step 7.2: Substitute the newly acquired observation data into the prediction model and use the Adam optimization algorithm to modify the parameters of the prediction model so that the model can better adapt to the new data characteristics and actual deformation conditions. Step 7.3: Use the revised prediction model to predict the movement and deformation of the structure again and re-evaluate its safety and stability; Step 7.4: Repeat step 7.3, continuously adjusting the model and assessing the safety status until the working face safely passes through the structure, ensuring that the safety impact on the structure is within a controllable range throughout the mining process.

Images