Intelligent safety and stability assessment method, system and equipment for underground cavern and storage medium

Abstract

The invention provides an intelligent safety and stability assessment method, system and device for an underground cavern and a medium, and the method comprises the following steps: S1, building a three-dimensional geological model of the underground cavern according to initial geological exploration data; s2, multiple monitoring instruments are pre-buried in key parts of the underground cavern, and monitoring data are collected in real time; s3, constructing an underground cavern safety and stability evaluation model based on a deep learning algorithm; s4, inputting real-time monitoring data into an evaluation model, combining historical monitoring data and geological model information, analyzing and comparing with a three-dimensional numerical calculation result, adjusting geological parameters, performing reconstruction evaluation, and outputting a safety and stability level; and S5, according to an evaluation result, automatically generating early warning information and processing suggestions. According to the method, the problems of poor real-time performance and low data utilization rate of traditional evaluation under complex geological conditions are solved, and the intelligent level of construction safety monitoring of underground cavern groups such as hydropower stations and pumped storage projects can be effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of underground engineering safety assessment, and specifically relates to an intelligent safety and stability assessment method, system, equipment and storage medium for underground caverns. Background Art

[0002] In underground engineering construction, especially during the construction of underground cavern complexes in large hydropower stations and pumped-storage power plants, surrounding rock stability remains a key concern for project safety due to complex geological conditions and harsh construction environments. Traditional safety assessment methods rely primarily on empirical formulas and simple analysis of on-site monitoring data. These methods lack the ability to deeply mine monitoring data and conduct real-time dynamic assessments, making them of limited value for on-site safety analysis and construction guidance.

[0003] With the development of deep learning technology, its successful application in image recognition, data prediction, and other fields has provided new insights into underground engineering safety assessment. However, applying deep learning technology to underground cavern safety assessment still faces many challenges, such as the high dimensionality of monitoring data and the modeling of complex geological conditions. Summary of the Invention

[0004] The first purpose of the present invention is to provide an intelligent safety and stability assessment method for underground caverns in response to the above-mentioned problems.

[0005] To this end, the above-mentioned purpose of the present invention is achieved through the following technical solutions:

[0006] An intelligent safety and stability assessment method for underground caverns comprises the following steps:

[0007] S1. Build a three-dimensional geological model of the underground cavern based on the initial geological survey data;

[0008] S2. Pre-install a variety of high-precision monitoring instruments, including strain gauges, pressure sensors, and water level gauges, in key structural locations within the underground caverns to collect real-time dynamic monitoring data on surrounding rock deformation, anchor stress, anchor cable load, and groundwater status. The collected data is systematically screened using the Pairwise algorithm to remove noise and outliers, ensuring high data quality and reliability. Based on this screened data and the GSI (geological strength index) parameters of the underground cavern's main surrounding rock, a real-time displacement increment database is established to dynamically analyze surrounding rock stability, predict potential risks, and guide project construction and maintenance.

[0009] S3. Based on deep learning algorithms, feature extraction and analysis are performed on the processed monitoring data. Real-time online inversion of rock mass parameters and background deformation prediction are performed to build a safety and stability assessment model for underground caverns.

[0010] S4. Input the real-time monitoring data into the assessment model, combine it with historical monitoring data and geological model information, and conduct real-time analysis and comparison with the three-dimensional numerical calculation results. Continuously adjust the geological parameters to conduct real-time reconstruction and assessment of the safety and stability status of the underground cavern, and output the safety and stability level.

[0011] S5. Based on the evaluation results, early warning information and corresponding treatment suggestions are automatically generated to achieve intelligent monitoring and early warning of the safety and stability status of underground caverns.

[0012] While adopting the above technical solutions, the present invention may also adopt or combine the following technical solutions:

[0013] As a preferred technical solution of the present invention: in step S1, the three-dimensional geological model includes the physical and mechanical parameters of the surrounding rock, structural surface information, initial ground stress field, lithology of the surrounding rock, development of joints and fissures, and groundwater flow path.

[0014] As a preferred technical solution of the present invention: in step S2, the monitoring instrument includes a multi-point displacement meter, an anchor stress meter, an anchor dynamometer, a piezometer, and a sonicator.

[0015] As a preferred technical solution of the present invention: in step S3, the Pairwise algorithm effectively identifies redundant data through pairwise similarity calculation and threshold determination. The core of the Pairwise algorithm uses cosine similarity, and the formula is as follows:

[0016]

[0017] Where sim(x,y) is the cosine similarity between vectors x and y, ranging from -1 to 1. Values ​​closer to 1 indicate more similar directions; values ​​closer to -1 indicate more opposite directions; and 0 indicates orthogonality. x, y are the two data vectors to be compared, typically numerical feature vectors. n is the dimension of the vector, i.e., the number of components in each vector. xi, yi are the i-th components of vectors x and y (1 ≤ i ≤ n).

[0018] As a preferred technical solution of the present invention: in step S4, the deep learning algorithm uses a convolutional neural network CNN or a recurrent neural network RNN ​​and its variants to comprehensively analyze the time series characteristics and spatial distribution characteristics of the monitoring data.

[0019] As a preferred technical solution of the present invention: in step S5, the security stability level is divided into multiple levels, including: stable, basically stable, unstable, and corresponding warning thresholds and processing measures are formulated according to different levels.

[0020] As a preferred technical solution of the present invention: in step S6, the warning information is sent to relevant personnel through various means, including: text messages, emails, and mobile phone APP push.

[0021] The second object of the present invention is to provide an intelligent safety and stability assessment system for underground caverns, comprising the following modules:

[0022] The data acquisition module is used to pre-embed a variety of monitoring instruments in key locations of underground caverns to collect real-time monitoring data, including surrounding rock deformation, anchor stress, anchor cable load, and groundwater status.

[0023] The data processing module is used to pre-process the monitoring data collected by the data acquisition module;

[0024] The model building module uses the Pairwise algorithm to screen and preprocess the GSI parameters of the main surrounding rock of the underground cavern, establishes a data preprocessing database based on real-time displacement increments, extracts and analyzes features of the processed monitoring data based on a deep learning algorithm, and performs real-time online inversion of rock mass parameters and background deformation prediction to build a safety and stability assessment model for the underground cavern.

[0025] The evaluation module is used to input the real-time monitoring data into the underground cavern safety and stability evaluation model constructed by the model construction module, and combine the historical monitoring data and geological model information to conduct a real-time evaluation of the safety and stability status of the underground cavern and output the safety and stability level;

[0026] The early warning module is used to automatically generate early warning information and corresponding processing suggestions based on the safety and stability level obtained by the evaluation module, and send them to relevant personnel.

[0027] The third object of the present invention is to provide an electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus.

[0028] a memory for storing a computer program;

[0029] A processor is used to execute the computer program stored in the memory to implement the steps of the intelligent safety and stability assessment method for underground caverns as described above.

[0030] Another object of the present invention is to provide a non-volatile storage medium, in which an executable program is stored. When the executable program is executed by a processor, the steps of the intelligent safety and stability assessment method for underground caverns as described above are implemented.

[0031] Compared with existing technologies, the present invention has the following beneficial effects: It achieves dynamic monitoring of construction safety by integrating a three-dimensional dynamic geological model, real-time acquisition of multi-source monitoring data, and a deep learning algorithm. It uses the Pairwise algorithm combined with cosine similarity to filter high-dimensional monitoring data and construct a real-time displacement increment database driven by surrounding rock GSI parameters. It employs a CNN / RNN neural network to perform convolution operations (including local receptive fields and weight sharing techniques) and feature extraction on spatiotemporal monitoring data to achieve rock mass parameter inversion and deformation prediction. Finally, a dynamic assessment model is used to classify safety states into multiple levels: "safe-warning-dangerous." Surrounding rock deformation thresholds are set based on engineering experience, and graded warnings are delivered via multiple channels, such as text messages and apps. This invention addresses the problems of poor real-time performance and low data utilization of traditional assessments under complex geological conditions. It can effectively enhance the intelligent level of construction safety monitoring for underground cavern complexes, such as hydropower stations and pumped storage projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a conceptual diagram of the design of the intelligent safety and stability assessment method for underground cavern groups of the present invention.

[0033] Figure 2 This is a flowchart of three-dimensional intelligent monitoring and safety early warning analysis of rock mass based on deep learning.

[0034] Figure 3a Schematic diagram of the actual monitoring instrument layout.

[0035] Figure 3b Schematic diagram of the measurement point arrangement for FLAC3D numerical calculation.

[0036] Figure 4 The process of automatic inversion of rock mass parameters is carried out for real-time monitoring data.

[0037] Figure 5 Automatically evaluate the inversion results of rock mass parameters.

[0038] Figure 6 It is the conversion between rock mass parameters of HB criterion and MC model.

[0039] Figure 7 It is the interface of the three-dimensional underground cavern real-time monitoring and early warning model system.

[0040] Figure 8 It is the interface of the real-time rock mass parameter inversion and early warning push system for layered excavation. DETAILED DESCRIPTION

[0041] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0042] like Figure 1-2 As shown, an intelligent safety and stability assessment method for underground caverns specifically includes the following steps:

[0043] S1. Build a three-dimensional geological model of the underground cavern based on the initial geological survey data;

[0044] The three-dimensional geological model includes the physical and mechanical parameters of the surrounding rock, structural surface information, initial ground stress field, etc. The coordinate system of the model coincides with the axis of the cavern, fully covering the main cavern and its excavation influence area. It also includes the lithology of the surrounding rock, the development of joints and fissures, the groundwater flow path, etc., to more accurately reflect the actual geological conditions of the underground cavern. The excavation steps in the model are simulated completely according to the actual excavation construction steps.

[0045] S2. Pre-install a variety of high-precision monitoring instruments, including strain gauges, pressure sensors, and water level gauges, in key structural locations within the underground caverns to collect real-time dynamic monitoring data on surrounding rock deformation, anchor stress, anchor cable load, and groundwater status. The collected data is systematically screened using the Pairwise algorithm to remove noise and outliers, ensuring high data quality and reliability. Based on this screened data and the GSI (geological strength index) parameters of the underground cavern's main surrounding rock, a real-time displacement increment database is established to dynamically analyze surrounding rock stability, predict potential risks, and guide project construction and maintenance.

[0046] The monitoring instruments include but are not limited to multi-point displacement meters, anchor stress gauges, anchor dynamometers, piezometers, sonicators, etc., and the frequency of monitoring data collection is dynamically adjusted according to the construction progress and geological conditions of the underground caverns. The arrangement of the monitoring instruments is consistent with the actual monitoring plan, support plan, and lithology and structural surface information to conduct excavation and support simulation analysis.

[0047] The Pairwise algorithm effectively identifies redundant data through pairwise similarity calculation and threshold judgment. The core of the Pairwise algorithm uses cosine similarity.

[0048]

[0049] Where sim(x,y) is the cosine similarity between vectors x and y, ranging from -1 to 1. Values ​​closer to 1 indicate more similar directions; values ​​closer to -1 indicate more opposite directions; and 0 indicates orthogonality. x, y are the two data vectors to be compared, typically numerical feature vectors. n is the dimension of the vector, i.e., the number of components in each vector. xi, yi are the i-th components of vectors x and y (1 ≤ i ≤ n).

[0050] S3. Based on deep learning algorithms, feature extraction and analysis are performed on the processed monitoring data. Real-time online inversion of rock mass parameters and background deformation prediction are performed to build a safety and stability assessment model for underground caverns.

[0051] Deep learning algorithms also include preprocessing of monitoring data, including data cleaning, filtering, normalization and other operations to improve the accuracy and reliability of the evaluation model;

[0052] The deep learning algorithm uses convolutional neural networks (CNN) or recurrent neural networks (RNN) and their variants. It first conducts a comprehensive analysis of the time series characteristics and spatial distribution characteristics of the monitoring data. Then, the displacement increments generated by the pre-buried multi-point displacers in each layer of the underground cavern are compared in real time with the displacement increments at the same location set in the numerical analysis. The geological parameters are continuously corrected to improve the accuracy and reliability of the assessment model. The CNN rock mass parameter prediction neural network efficiently extracts spatial features through local receptive fields and weight sharing. The core formula is as follows:

[0053] Convolution operation: Capturing local patterns:

[0054]

[0055] Nonlinear activation: Introducing model expressiveness:

[0056]

[0057] Pooling operation: reduces computational complexity and enhances translation invariance:

[0058]

[0059] Loss function and optimization: driving model parameter updates:

[0060]

[0061] Where O(i, j) is the value of the output feature map at position (i, j); X is the input feature map (size H in ×W in , may contain multiple channels); W is the convolution kernel (size k h ×k w , one kernel per channel); b is the bias term (scalar, one per output channel); s is the stride, which controls the sliding stride of the convolution kernel (e.g., s=1 or s=2); k h ,k w where yi,c is the true label of the i-th sample (one-hot encoding, where only the true class is 1); pi,c is the model-predicted probability that the i-th sample belongs to class c. L is the cross-entropy loss; and N is the batch size.

[0062] S4. Input the real-time monitoring data into the assessment model, combine it with historical monitoring data and geological model information, and conduct real-time analysis and comparison with the three-dimensional numerical calculation results. Continuously adjust the geological parameters to conduct real-time reconstruction and assessment of the safety and stability status of the underground cavern, and output the safety and stability level.

[0063] Based on the quantitatively corrected geological parameters provided in step S3 and the real-time updated geological exposure structural surface, the discrete element system is automatically triggered periodically (for example, at 6 pm) to perform background analysis and real-time update of the 3D visualization platform data. A feedback model is constructed based on the monitoring data, geological conditions, and construction information. Excavation simulation is performed in combination with the actual support plan and monitoring points. By comparing the simulation results with the field data, the mechanical parameters are adjusted and the subsequent excavation is predicted. At the same time, the surrounding rock stability classification warning standards and support adjustments are updated, thereby achieving timely and systematic real-time feedback analysis during the construction period of the underground cavern group.

[0064] Based on the geological structure exposed during excavation of each layer of the cavern and the spatial properties of the 3D blocks, a projection relationship module between the local working plane and the global coordinate system is established. This allows for the rapid implementation of uniform and effective support arrangements, enabling visual operation processes, customized support parameters, and reinforcement optimization solutions to ensure that the 3D blocks are anchored in the intact surrounding rock.

[0065] The safety and stability levels are divided into multiple levels, including: stable, basically stable, and unstable. Corresponding warning thresholds and treatment measures are formulated according to different levels. The division of safety and stability levels also includes setting specific warning standards for monitoring indicators such as surrounding rock deformation and anchor axial force. The warning standards combine the evaluation of surrounding rock deformation stability and macro indicators extracted from engineering experience.

[0066] S5. Based on the evaluation results, early warning information and corresponding treatment suggestions are automatically generated to achieve intelligent monitoring and early warning of the safety and stability status of underground caverns.

[0067] Early warning information is sent to relevant personnel through various means, including text messages, emails, mobile APP push, etc., to ensure that relevant personnel are notified in time to take corresponding measures.

[0068] The present invention also provides an intelligent safety and stability assessment system for underground caverns, comprising the following modules:

[0069] The data acquisition module is used to pre-embed a variety of monitoring instruments in key locations in the underground cavern to collect real-time monitoring data, including surrounding rock deformation, anchor stress, anchor cable load, and groundwater status. It also includes a data verification function to perform real-time verification of the collected monitoring data to ensure its accuracy and completeness.

[0070] The data processing module is used to pre-process the monitoring data collected by the data acquisition module and collect monitoring data such as surrounding rock deformation, anchor stress, anchor cable load, groundwater status, etc. in real time;

[0071] The model building module uses the Pairwise algorithm to screen and preprocess the GSI parameters of the main surrounding rock of the underground cavern, establishes a data preprocessing database based on real-time displacement increments, extracts and analyzes features of the processed monitoring data based on a deep learning algorithm, and performs real-time online inversion of rock mass parameters and background deformation prediction to build an underground cavern safety and stability assessment model. It also includes a model optimization function, dynamically optimizing and adjusting the assessment model based on feedback from the assessment results to improve the model's accuracy and adaptability.

[0072] The evaluation module is used to input the real-time monitoring data into the underground cavern safety and stability evaluation model constructed by the model construction module, and combine the historical monitoring data and geological model information to conduct a real-time evaluation of the safety and stability status of the underground cavern and output the safety and stability level;

[0073] The early warning module is used to automatically generate early warning information and corresponding processing suggestions based on the safety and stability level obtained by the evaluation module, and send them to relevant personnel. It also includes an early warning recording function to record and store the generated early warning information for subsequent query and analysis.

[0074] The present invention also provides an electronic device, which includes a processor, a communication interface, a memory and a communication bus. The processor, the communication interface and the memory communicate with each other via the communication bus.

[0075] Memory, memory is used to store computer programs;

[0076] The processor is used to execute the computer program stored in the memory to implement the steps of the intelligent safety and stability assessment method for underground caverns as described above.

[0077] The present invention also provides a non-volatile storage medium, in which an executable program is stored. When the executable program is executed by a processor, the steps of the intelligent safety and stability assessment method for underground caverns as described above are implemented.

[0078] The examples are as follows:

[0079] S1: 3D geological dynamic modeling and parameter initialization

[0080] Establish a 3D geological model: Taking the underground cavern complex of a general hydropower station or pumped-storage power station as an example, establish a 3D geological model of the main caverns, including the main and auxiliary powerhouse caverns, main transformer room, busbar cavern, tailrace cavern, and diversion tunnel. The y-axis of the model's coordinate system coincides with the axis of the powerhouse cavern, and the z-axis is the vertical direction.

[0081] Dynamically simulate the excavation process according to actual construction steps (layered excavation and support sequence).

[0082] Preset rock mass parameters: define the GSI (geological strength index) range of different surrounding rock types (e.g. Class I surrounding rock GSI = 75-90, Class IV surrounding rock GSI = 10-30).

[0083] Initial layered excavation calculations were performed to calculate the excavation response deformation increments for different rock layer GSI combinations, and the same monitoring points as those in actual monitoring were arranged in the 3D model, such as Figure 3a and Figure 3b As shown in Figure 2, the displacement increments of different virtual measuring points under different GSI combinations are obtained, which is convenient for the subsequent prediction of actual rock mass parameters.

[0084] For example, in a cavern with Class II / Class III surrounding rock, a nine-story excavation is performed, and 10 pre-buried multi-point displacement meters are installed. If the GSI is 60-75 for Class II and 45-60 for Class III, the displacement increment sample library will contain 10 (instruments) × 9 (layers) × 15 (75-60) × 15 (60-45) = 20,250 sample data.

[0085] S2: Dynamic collection and cleaning of multi-source monitoring data

[0086] Deployment of monitoring instruments: Multi-point displacement meters, anchor stress meters and other equipment are pre-buried and installed in key locations in the cavern to collect data such as surrounding rock deformation, anchor axial force, and groundwater status in real time.

[0087] Dynamically adjust the acquisition frequency: Automatically adjust the data acquisition frequency according to the construction progress (such as blasting stage, after support) and geological conditions (such as encountering fault zones).

[0088] Data cleaning and screening: Pairwise algorithm (cosine similarity calculation) is used to screen the data twice:

[0089] Initial sample library stage: remove redundant displacement increment data (similarity > threshold).

[0090] Real-time monitoring stage: dynamically update valid data and retain feature difference data.

[0091] Data preprocessing: Filter and normalize the monitoring data to eliminate noise interference.

[0092] S3: Parameter Inversion and Prediction Based on Deep Learning

[0093] Build CNN / RNN neural network model:

[0094] The cleaned displacement increment data is input, and the spatial features are extracted through the convolutional layer (local receptive field, weight sharing), and the temporal features are processed in combination with RNN.

[0095] Convolution operation:

[0096]

[0097] Loss function optimization (cross entropy):

[0098]

[0099] Intelligent inversion of rock mass parameters:

[0100] According to the initially established displacement increment sample library and the connected monitoring displacement data, the predicted surrounding rock HB parameters (such as geological strength factor GSI and disturbance factor D) are output in real time, such as Figure 4 shown.

[0101] The reliability of the outputted surrounding rock HB parameters needs to be further evaluated and its accuracy needs to be determined based on the surrounding rock distribution of the monitoring section. Figure 5 shown

[0102] Rock mass parameters with high inversion reliability are automatically converted into MC criterion parameters (internal friction angle φ, cohesion c), such as Figure 6 shown.

[0103] S4: 3D Discrete Element Safety Assessment and Graded Warning

[0104] Numerical simulation and stability analysis:

[0105] The inversion parameters are input into the three-dimensional discrete element block model to simulate the deformation response of the surrounding rock and output the next layer of predicted deformation value and early warning value.

[0106] Combine the historical data and real-time monitoring results of the system to calculate the safety factor of the underground cavern. Figure 7 As shown, evaluate the stability:

[0107] Security level classification:

[0108] Safe (green): Deformation rate ≤ threshold 1, support optimization recommended.

[0109] Warning (yellow): Threshold 1 < deformation rate ≤ threshold 2, local reinforcement.

[0110] Danger (red): Deformation rate > threshold 2, emergency stop and investigation.

[0111] S5: Intelligent decision-making and early warning push

[0112] Dynamic response mechanism:

[0113] Automatically generate early warning information (including deformation trend diagram, risk location, and treatment suggestions) and push it to construction personnel via SMS and APP, such as Figure 8 shown.

[0114] The early warning information includes the specific values ​​of monitoring indicators such as surrounding rock deformation and anchor axial force, as well as the corresponding early warning levels and treatment suggestions.

[0115] Closed-loop feedback: Adjust subsequent excavation plans based on early warning results, update 3D model parameters, and iteratively optimize the evaluation process.

[0116] Thus far, the technical solutions of the present invention have been described in conjunction with the specific experimental processes shown in the accompanying drawings. However, the scope of protection of the present invention is not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A method, system, device and storage medium for intelligent safety and stability assessment of underground caverns, characterized in that: The steps include: S1. Build a three-dimensional geological model of the underground cavern based on the initial geological survey data; S2. A variety of high-precision monitoring instruments, including strain gauges, pressure sensors, and water level gauges, are pre-buried in key structural locations within the underground cavern to collect real-time dynamic monitoring data on surrounding rock deformation, anchor stress, anchor cable load, and groundwater status. The collected data is systematically screened using the Pairwise algorithm to remove noise and outliers, ensuring high data quality and reliability. Based on this screened data and combined with the GSI parameters of the underground cavern's main surrounding rock, a real-time displacement increment database is established to dynamically analyze surrounding rock stability, predict potential risks, and guide project construction and maintenance. S3. Based on deep learning algorithms, feature extraction and analysis are performed on the processed monitoring data. Real-time online inversion of rock mass parameters and background deformation prediction are performed to build a safety and stability assessment model for underground caverns. S4. Input the real-time monitoring data into the assessment model, combine it with historical monitoring data and geological model information, and conduct real-time analysis and comparison with the three-dimensional numerical calculation results. Continuously adjust the geological parameters to conduct real-time reconstruction and assessment of the safety and stability status of the underground cavern, and output the safety and stability level. S5. Based on the evaluation results, early warning information and corresponding treatment suggestions are automatically generated to achieve intelligent monitoring and early warning of the safety and stability status of underground caverns.

2. The method according to claim 1, wherein: In step S1, the three-dimensional geological model includes the physical and mechanical parameters of the surrounding rock, structural surface information, initial ground stress field, lithology of the surrounding rock, development of joints and fissures, and groundwater flow path.

3. The method according to claim 1, wherein: In step S2, the monitoring instruments include a multi-point displacement meter, an anchor stress meter, an anchor dynamometer, a piezometer, and a sonicator.

4. The method according to claim 1, wherein: In step S2, the Pairwise algorithm effectively identifies redundant data through pairwise similarity calculation and threshold determination. The core of the Pairwise algorithm uses cosine similarity, and the formula is as follows: Where sim(x,y) is the cosine similarity between vectors x and y, and its value range is [-1,1]. The closer the value is to 1, the more similar the directions of the two vectors are; the closer the value is to -1, the more opposite the directions are; 0 indicates orthogonality. x, y: The two data vectors to be compared, usually numerical feature vectors; n: The dimension of the vector, that is, the number of components each vector contains. xi,yi: The i-th component of vectors x and y, where 1≤i≤n.

5. The method according to claim 1, wherein: In step S3, the deep learning algorithm uses convolutional neural network (CNN) or recurrent neural network (RNN) and its variants to comprehensively analyze the time series characteristics and spatial distribution characteristics of the monitoring data.

6. The method according to claim 1, wherein: In step S4, the security and stability level is divided into multiple levels, including: stable, basically stable, and unstable, and corresponding warning thresholds and processing measures are formulated according to different levels.

7. The method according to claim 1, wherein: In step S5, the warning information is sent to relevant personnel through various means, including: text message, email, and mobile phone APP push.

8. An intelligent safety and stability assessment system for underground caverns, characterized in that: Includes the following modules: The data acquisition module is used to pre-embed a variety of monitoring instruments in key locations of underground caverns to collect real-time monitoring data, including surrounding rock deformation, anchor stress, anchor cable load, and groundwater status. The data processing module is used to pre-process the monitoring data collected by the data acquisition module; The model building module uses the Pairwise algorithm to screen and preprocess the GSI parameters of the main surrounding rock of the underground cavern, establishes a data preprocessing database based on real-time displacement increments, extracts and analyzes features of the processed monitoring data based on a deep learning algorithm, and performs real-time online inversion of rock mass parameters and background deformation prediction to build a safety and stability assessment model for the underground cavern. The evaluation module is used to input the real-time monitoring data into the underground cavern safety and stability evaluation model constructed by the model construction module, and combine the historical monitoring data and geological model information to conduct a real-time evaluation of the safety and stability status of the underground cavern and output the safety and stability level; The early warning module is used to automatically generate early warning information and corresponding processing suggestions based on the safety and stability level obtained by the evaluation module, and send them to relevant personnel.

9. An electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus, wherein: a memory for storing a computer program; A processor, wherein the processor is used to execute a computer program stored in a memory to implement the steps of the intelligent safety and stability assessment method for an underground cavern as described in any one of claims 1 to 7.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores an executable program, and when the executable program is executed by the processor, the steps of the intelligent safety and stability assessment method for underground caverns as described in any one of claims 1 to 7 are implemented.

Images