Karst tunnel large-scale cave treatment method

Abstract

The application discloses a karst tunnel large-scale cave treatment method, relates to the technical field of building construction, and acquires the accurate position and attribute data of the underground cave through an advanced detection robot, adopts an artificial intelligence algorithm and a data analysis platform, automatically identifies and classifies the cave, generates a three-dimensional model of the cave, improves the efficiency and accuracy of treating the cave, reduces the labor and time cost, presents the position, size and attribute of the cave in a visual manner, and automatically grouts the device and intelligently plans the path to make the grouting engineering more accurate and controllable, the grouting strategy can be adjusted in real time, adapts to the change of the geological condition and the distribution of the cave, reduces waste, issues an alarm immediately when an abnormality occurs through real-time monitoring of the surface subsidence, the underground water level and the grouting effect, and the real-time monitoring and feedback mechanism is helpful to timely find problems and take necessary measures, so that the safety of the engineering is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of building construction, in particular to a large-scale karst cave treatment method for karst tunnel. BACKGROUND

[0002] Tunnel construction is an engineering project that is highly dependent on geological conditions. When it passes through karst areas, various sizes and shapes of karst caves, their locations, filling conditions, and filling properties can have a significant impact on tunnel construction. During the construction of underground projects, karst caves make the mechanical effects of surrounding rock more complex. When a tunnel is excavated and encounters a karst cave, it is easy to cause local stress concentration in the surrounding rock, especially when the karst cave is irregular in shape and close to the tunnel face. The stress concentration is often higher in the sharp corners of the karst cave, which increases the deformation of the tunnel and can even cause geological phenomena such as falling and collapse.

[0003] Through the efforts of tunnel builders over the years, China has accumulated rich experience in dealing with complex karst geology tunnel construction technology and has reached the world's advanced level. By using advanced detection, curtain grouting, pipe roof pre-support, high-pressure water-rich cavity energy release and pressure reduction, detour, bridge crossing, disaster prevention and early warning, rescue and escape systems, and other comprehensive construction techniques and measures such as "exploration", "blocking", "draining", "detouring", "crossing", and "preventing", the ability to safely handle large-scale karst during tunnel construction has been greatly enhanced. According to the "multi-source coordination principle", a set of effective advanced geological prediction technology is selected to determine the size, accurate location, filling properties, and connectivity of hidden faults and karst in front of the tunnel excavation face. By taking targeted treatment measures, not only can tunnel collapse, sudden mud and water disasters be avoided, but also the tunnel construction progress can be significantly accelerated, engineering construction costs can be saved, and the economic and social benefits of the project can be improved.

[0004] However, traditional karst cave treatment methods often use hand-drawn paper, which cannot provide real-time three-dimensional models of karst caves. Construction teams have difficulty accurately understanding the location, size, and properties of underground karst caves, leading to inaccuracies in construction planning and increasing the risk of accidents during construction. In addition, traditional karst cave treatment methods cannot timely detect surface subsidence, underground water level, and abnormal grouting effects, increasing the safety hazards of the project. If timely measures are not taken, it may lead to accidents and damage to the tunnel project, requiring additional manpower and time to handle unexpected situations, resulting in low engineering efficiency and increased costs. Therefore, there is an urgent need for a large-scale karst cave treatment method for karst tunnels that can automatically identify and classify karst caves and provide abnormal alerts to solve these problems. SUMMARY

[0005] (I) Technical problems solved

[0006] In view of the deficiencies of the prior art, the karst tunnel large-scale cave treatment method is provided, which solves the problems that the prior art cannot provide a real-time three-dimensional model of the cave, and cannot timely discover the abnormality of the ground subsidence, the underground water level and the grouting effect.

[0007] (II) Technical scheme

[0008] To achieve the above object, the karst tunnel large-scale cave treatment method is provided, which comprises the following technical solutions.

[0009] Geological survey and advanced detection, geological survey is carried out to obtain geological data of the tunnel area, including the position, size and distribution of the cave, high-precision detection is carried out in the shield machine using an advanced detection robot, and the position and attributes of the cave are identified and recorded;

[0010] Cave three-dimensional model identification, the geological survey and advanced detection data are imported into a data analysis platform, computer vision and machine learning algorithms are used to automatically identify and classify the cave, and a three-dimensional model of the cave is generated;

[0011] Grouting path planning, an automatic grouting device is used to generate a grouting path according to the advanced detection data and artificial intelligence classification results, the automatic grouting device carries out accurate grouting in front of the shield machine, and adjusts according to real-time monitoring data;

[0012] Grouting monitoring, based on a real-time monitoring system, the surface subsidence, underground water level and grouting effect are monitored through sensors, and the monitoring data are integrated into a control center to realize remote monitoring and real-time feedback.

[0013] The karst tunnel large-scale cave treatment method further comprises the following steps.

[0014] First, the tunnel construction area is determined, and existing geological and topographic data are collected, including stratum type, underground water level, geological and topographic map, and known position information of underground cave;

[0015] The existing data are analyzed to determine the karst geological characteristics, including the type and distribution of the cave;

[0016] The existing data include geological characteristics, including rock stratum type, color and texture;

[0017] The collected data are visually analyzed to make a geological map and a stratum profile;

[0018] The karst tunnel large-scale cave treatment method further comprises the following steps.

[0019] The geological survey and advanced detection data are converted from the original format to the data analysis platform processing format and are preprocessed;

[0020] Importing the pre-processed geological data and advanced detection data into a data repository in the platform;

[0021] Automatically identifying and classifying the karst caves based on computer vision and machine learning algorithms;

[0022] Based on the identified and classified karst cave data, generating a three-dimensional model of the karst caves using computer graphics technology, mapping the coordinates and attributes of the underground caves into a three-dimensional space;

[0023] Rendering the three-dimensional model and visualizing the shape, distribution and attributes of the karst caves;

[0024] The present application further provides that the step of automatically identifying and classifying the karst caves specifically includes:

[0025] Converting the geological data and advanced detection data obtained by geological exploration and advanced detection into numerical feature vectors and performing standardization processing;

[0026] Dividing the data into a training set and a test set;

[0027] Using support vector machines (SVM) and convolutional neural networks (CNN) to train the model to learn the features of the karst caves from the input data and perform classification;

[0028] The minimization loss function of the training process is defined as:

[0029]

[0030] Where L θ is the loss function, θ is the model parameter, N is the number of geological exploration training samples, y i is the actual survey data, is the output prediction model;

[0031] The present application further provides that the step of automatically identifying and classifying the karst caves further includes using the test set to evaluate the trained model:

[0032] Based on the machine learning model, predicting new geological data and detection data, the model assigns a class label to each data point, indicating whether it is a karst cave or not;

[0033] The class classification is based on the softmax function to convert the output of the model into a probability distribution, selecting the highest probability class, specifically:

[0034]

[0035] Where P(y i = j | x) represents the probability that data point i belongs to class j given input x, e zik is the output of the model, representing the score of data point i belonging to class j, and K is the total number of classes;

[0036] The application is further provided that the method for generating a three-dimensional model of the cave is:

[0037] Based on the identified and classified cave data obtained in the automatic identification and classification step, a three-dimensional coordinate system is established;

[0038] The coordinate information of each cave is mapped into the three-dimensional coordinate system, and the coordinates of each cave are set as (x i ,y i ,z i ), wherein x i , y i and z i represent the position coordinates of the cave, respectively;

[0039] Based on the graphics technology, a three-dimensional model is created, and according to the coordinate information of each cave, a corresponding geometric body is created in the three-dimensional model;

[0040] The classified cave attributes are mapped into the three-dimensional model, and different colors, textures and labels are used to represent caves of different categories;

[0041] The application is further provided that the method for generating a three-dimensional model of the cave is:

[0042] Initialize the path planning parameters, including the starting point (X s.t ,T s.t ) and the target point (X t.t ,T t.t );

[0043] Create empty open list O and closed list C for storing nodes to be explored and nodes that have been explored, and add the starting point to the open list O and set its initial cost to 0;

[0044] Enter the loop until the open list O reaches the end point, specifically:

[0045] Select the node with the lowest cost from the open list O as the current node;

[0046] If the current node (X s.t ,T s.t ) is the end point, the path planning is completed, and the loop is exited;

[0047] Move the current node (X s.t ,T s.t ) from the open list O to the closed list C;

[0048] Iterate through the neighbor nodes of the current node (X s.t ,T s.t );

[0049] From the end point, trace back along the parent node of each node until the starting point is reached, forming a path;

[0050] The application further provides that the sensors include a total station, a laser scanner, a water level monitoring sensor, and a grouting effect monitoring sensor for measuring the change in ground subsidence in real time, and the sensors are uniformly installed on the ground surface near the tunnel.

[0051] The groundwater level monitoring sensor is installed in the tunnel well for monitoring the change in groundwater level, and the change in groundwater level during grouting.

[0052] The grouting effect monitoring sensor is installed near the injection port of the grouting device for real-time monitoring of the flow rate and flow of the grouting material.

[0053] The application further provides that the sensors are connected to the central real-time monitoring system through wireless connection for real-time reception and recording of the data of the sensors.

[0054] The monitoring system is provided with an alarm threshold value, and an alarm is issued when the ground subsidence, groundwater level, and grouting effect exceed the safe range.

[0055] The application also provides a large-scale karst cave treatment system for a karst tunnel, characterized in that it comprises:

[0056] A data analysis platform for analyzing and visually displaying the shape, distribution, and properties of the karst cave based on computer vision and machine learning algorithms.

[0057] A real-time monitoring system with a total station, a laser scanner, a water level monitoring sensor, and a grouting effect monitoring sensor for real-time monitoring of ground subsidence, groundwater level, and grouting effect.

[0058] A control center based on monitoring data and built-in threshold values for issuing an alarm when the ground subsidence, groundwater level, and grouting effect exceed the safe range.

[0059] (Three) beneficial effects

[0060] The application provides a large-scale karst cave treatment method for a karst tunnel, which has the following beneficial effects:

[0061] For tunnel engineering under karst geological conditions, high precision of geological survey is realized, the accurate position and attribute data of underground karst cave are obtained by using advanced detection robot, the reliability of data is improved, the engineering team can better understand the underground situation and reduce the risk, the artificial intelligence algorithm and data analysis platform are used to automatically identify and classify the karst cave, and the three-dimensional model of the karst cave is generated, so that the efficiency and accuracy of processing the karst cave are improved, the manpower and time cost are reduced, the position, size and attribute of the karst cave are presented in a visual way, so that the construction team can accurately understand the underground situation of the tunnel construction area, and provide accurate construction planning basis.

[0062] In addition, the automatic grouting device and intelligent path planning make the grouting engineering more accurate and controllable, the grouting strategy can be adjusted in real time to adapt to the changes of geological conditions and karst cave distribution, the waste is minimized, and the resource utilization rate is improved.

[0063] At the same time, by installing total station, laser scanner, water level monitoring sensor and grouting effect monitoring sensor, the ground settlement, underground water level and grouting effect are monitored in real time, the sensors are connected to the central real-time monitoring system through wireless connection, and an alarm is sent immediately when an abnormality occurs, the real-time monitoring and feedback mechanism helps to find problems in time and take necessary measures, so as to improve the safety of the project.

[0064] Finally, the real-time monitoring system and sensors make the project monitoring and feedback more comprehensive and timely, improve the safety and controllability of the project, reduce the risk of accidents, and provide real-time data for the engineering team to respond to problems and adjust strategies in a timely manner.

[0065] The problems that the real-time three-dimensional model of the karst cave cannot be provided, and the ground settlement, underground water level and grouting effect cannot be found in time are solved. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 The flow chart of the karst tunnel large-scale karst cave treatment method of the present application. DETAILED DESCRIPTION

[0067] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0068] EMBODIMENT

[0069] Please refer to Figure 1 The present application provides a karst tunnel large-scale karst cave treatment method, comprising:

[0070] S1. Geological survey and advanced detection, conduct geological survey to obtain geological data of the tunnel area, including the location, size and distribution of karst caves, use advanced detection robots inside the shield machine for high-precision detection to identify and record the location and attributes of karst caves;

[0071] In the geological survey and advanced detection:

[0072] First, determine the tunnel construction area and collect existing geological and topographic data, including stratum type, groundwater level, geological and topographic maps, and known location information of underground caves;

[0073] Analyze existing data to determine karst geological characteristics, including karst cave types and distribution;

[0074] Existing data includes geological characteristics, including rock type, color, and texture;

[0075] Visualize the collected data and create geological maps and stratigraphic profiles;

[0076] S2. Karst cave three-dimensional model identification, import geological survey and advanced detection data into the data analysis platform, use computer vision and machine learning algorithms to automatically identify and classify karst caves, and generate three-dimensional models of karst caves;

[0077] Karst cave three-dimensional model identification steps:

[0078] Convert geological survey and advanced detection data from original format to data analysis platform processing format and perform data preprocessing;

[0079] Import preprocessed geological data and advanced detection data into the data storage library in the platform;

[0080] Based on computer vision and machine learning algorithms, automatically identify and classify karst caves;

[0081] Based on the identified and classified karst cave data, use computer graphics technology to generate three-dimensional models of karst caves, mapping the coordinates and attributes of underground caves to three-dimensional space;

[0082] Render the three-dimensional model and visualize the shape, distribution and attributes of the karst caves;

[0083] The steps of automatically identifying and classifying karst caves include:

[0084] Convert the geological data and advanced detection data obtained by geological survey and advanced detection into numerical feature vectors and perform standardization processing;

[0085] Divide the data into training set and test set;

[0086] SVM and CNN are used for training, so that the model learns the features of the cave from the input data and classifies them;

[0087] The minimization loss function of the training process is defined as:

[0088]

[0089] where L θ is the loss function, θ is the model parameter, N is the number of geological survey training samples, y i is the actual survey data, is the output prediction model;

[0090] The automatic recognition and classification steps of the cave also include using the test set to evaluate the trained model:

[0091] Based on the machine learning model, new geological data and detection data are predicted, and the model assigns a class label to each data point, indicating whether it is a cave or not;

[0092] The class classification is based on the softmax function to convert the output of the model into a probability distribution, and the highest probability class is selected, specifically:

[0093]

[0094] where P(y i = j|x) represents the probability that data point i belongs to class j given input x, e z ik is the output of the model, indicating the score of data point i belonging to class j, and K is the total number of classes;

[0095] The method for generating a three-dimensional model of the cave is:

[0096] Based on the recognized and classified cave data obtained from the automatic recognition and classification steps, a three-dimensional coordinate system is established;

[0097] Map the coordinate information of each cave to the three-dimensional coordinate system. Let the coordinates of each cave be (x i ,y i ,z i ), then where x i , y i , and z i represent the position coordinates of the cave;

[0098] Based on the graphics technology, a three-dimensional model is created, and according to the coordinate information of each cave, a corresponding geometric body is created in the three-dimensional model;

[0099] Map the classified cave attributes to the three-dimensional model, and use different colors, textures, and markers to represent different classes of caves;

[0100] S3. Grouting path planning, using an automatic grouting device, generating a grouting path according to the advanced detection data and the artificial intelligence classification result, the automatic grouting device performing accurate grouting in front of the shield machine and adjusting according to real-time monitoring data;

[0101] The grouting path generation method is as follows:

[0102] Initializing path planning parameters, including the starting point (X s.t ,T s.t ) and the target point (X t.t ,T t.t );

[0103] Creating empty open list O and closed list C for storing nodes to be explored and explored nodes, adding the starting point to the open list O and setting its initial cost as 0;

[0104] Entering a loop until the open list O reaches the end point, specifically:

[0105] Selecting the node with the lowest cost from the open list O as the current node;

[0106] If the current node (X s.t ,T s.t ) is the end point, the path planning is completed, and the loop is exited;

[0107] Moving the current node (X s.t ,T s.t ) from the open list O to the closed list C;

[0108] Traversing the neighbor nodes of the current node (X s.t ,T s.t );

[0109] Calculating the cost from the current node to the starting point plus the cost from the neighbor node to the current node;

[0110] If the neighbor node is already in the closed list C, skip this node;

[0111] If the neighbor node is not in the open list O, add it to the open list O and record its parent node as the current node, and update the cost to the starting point;

[0112] If the neighbor node is already in the open list O and the new cost is lower, update the parent node and cost of the node;

[0113] Starting from the end point, tracing back along the parent node of each node until reaching the starting point, forming a path;

[0114] S4. Grouting monitoring, based on a real-time monitoring system, ground surface settlement, underground water level and grouting effect are monitored through sensors, and monitoring data is integrated into the control center to realize remote monitoring and real-time feedback;

[0115] The sensors include a total station, a laser scanner, a water level monitoring sensor, and a grouting effect monitoring sensor, for real-time measurement of changes in ground surface settlement, and the sensors are uniformly installed near the tunnel surface;

[0116] The underground water level monitoring sensor is installed in the tunnel well for monitoring changes in the underground water level, and changes in the underground water level during grouting;

[0117] The grouting effect monitoring sensor is installed near the injection port of the grouting device for real-time monitoring of the flow rate and flow of the grouting material;

[0118] The sensors are connected to the central real-time monitoring system through wireless connection, for real-time reception and recording of sensor data;

[0119] Alarm thresholds are set in the monitoring system to issue an alarm when the ground surface settlement, underground water level and grouting effect exceed the safe range;

[0120] The application also provides a large-scale karst cave treatment system for karst tunnels, comprising:

[0121] A data analysis platform analyzes and visually displays the shape, distribution and properties of the karst cave based on computer vision and machine learning algorithms;

[0122] A real-time monitoring system, with a total station, a laser scanner, a water level monitoring sensor and a grouting effect monitoring sensor, is used to monitor ground surface settlement, underground water level and grouting effect in real time;

[0123] A control center based on monitoring data and built-in thresholds issues an alarm when the ground surface settlement, underground water level and grouting effect exceed the safe range.

[0124] In summary, in this application:

[0125] For tunnel engineering under karst geological conditions, high precision of geological survey is achieved, accurate position and attribute data of underground karst caves are obtained through advanced detection robots, data reliability is improved, the engineering team can better understand the underground situation, reduce risks, artificial intelligence algorithms and data analysis platforms are used to automatically identify and classify karst caves, and three-dimensional models of karst caves are generated, thereby improving the efficiency and accuracy of treating karst caves, reducing labor and time costs, presenting the position, size and properties of karst caves in a visual way, enabling the construction team to accurately understand the underground situation of the tunnel construction area, and providing an accurate basis for construction planning.

[0126] In addition, the automatic grouting device and intelligent path planning make the grouting project more precise and controllable, and the grouting strategy can be adjusted in real time to adapt to the changes in geological conditions and cave distribution, thereby minimizing waste and improving resource utilization.

[0127] At the same time, by installing total station, laser scanner, water level monitoring sensor and grouting effect monitoring sensor, real-time monitoring of ground settlement, groundwater level and grouting effect is realized, the sensors are connected to the central real-time monitoring system through wireless connection, and an alarm is immediately sent when an abnormality occurs, the real-time monitoring and feedback mechanism helps to find problems in time and take necessary measures, thereby improving the safety of the project.

[0128] Finally, the real-time monitoring system and sensors make the project monitoring and feedback more comprehensive and timely, improve the safety and controllability of the project, and reduce the risk of accidents, providing real-time data for the engineering team to respond to problems and adjust strategies more timely.

[0129] Obviously, the above embodiments of the present application are only examples for clearly illustrating the present application, and are not intended to limit the embodiments of the present application. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the embodiments are not required to be exhausted. Any modification, equivalent replacement and improvement within the spirit and principle of the present application shall be included in the protection scope of the claims of the present application.

Claims

1. A method for large-scale cave treatment of a karst tunnel, characterized in that, The method comprises the following steps: Geological survey and advanced detection: conduct geological survey to obtain geological data of the tunnel area, including the location, size and distribution of karst caves, and use advanced detection robots to conduct high-precision detection inside the shield machine to identify and record the location and attributes of the karst caves; Karst cave three-dimensional model identification: import the geological survey and advanced detection data into a data analysis platform, use computer vision and machine learning algorithms to automatically identify and classify the karst caves, and generate a three-dimensional model of the karst caves; Grouting path planning: use an automated grouting device to generate a grouting path based on the advanced detection data and artificial intelligence classification results, and the automated grouting device conducts precise grouting in front of the shield machine and adjusts according to real-time monitoring data; Grouting monitoring: based on a real-time monitoring system, monitor ground subsidence, underground water level and grouting effect through sensors, and integrate the monitoring data into a control center to realize remote monitoring and real-time feedback; In the karst cave three-dimensional model identification step: Convert the geological survey and advanced detection data from the original format to the data analysis platform processing format and perform data preprocessing; Import the preprocessed geological data and advanced detection data into the data storage library in the platform; Based on computer vision and machine learning algorithms, automatically identify and classify the karst caves; Based on the identified and classified karst cave data, use computer graphics technology to generate a three-dimensional model of the karst caves, mapping the coordinates and attributes of the underground caves to the three-dimensional space; Render the three-dimensional model to visually display the shape, distribution and attributes of the karst caves; The automatic identification and classification of karst caves specifically includes: Convert the geological data and advanced detection data obtained by geological survey and advanced detection into numerical feature vectors and perform standardization processing; Divide the data into training set and test set; Use support vector machine SVM and convolutional neural network CNN to train the model to learn the features of the karst caves from the input data and perform classification; The minimization loss function of the training process is defined as: ， wherein is a loss function, is a model parameter, N is the number of geological survey training samples, is actual survey data, is an output prediction model; The automatic identification and classification of karst caves also includes using the test set to evaluate the trained model: Based on the machine learning model, predict new geological data and detection data, and the model assigns a class label to each data point, indicating whether it is a karst cave; The class classification converts the model's output into a probability distribution based on the softmax function, selecting the highest probability class, specifically: ， where represents the probability that data point i belongs to class j given input x, is the output of the model, representing the score of data point i belonging to class j, and K is the total number of classes.

2. The method according to claim 1, wherein, In the geological survey and advanced detection step: First, determine the tunnel construction area and collect existing geological and topographic data, including stratum type, underground water level, geological and topographic map, and known location information of underground caves; Analyze the existing data to determine the karst geological characteristics, including the type and distribution of karst caves; The existing data includes geological characteristics, including rock type, color and texture; Visualize the collected data and create a geological map and stratum profile.

3. The method according to claim 1, wherein, The method for generating a three-dimensional model of the karst caves is: Based on the identified and classified karst cave data obtained from the automatic identification and classification step, establish a three-dimensional coordinate system; The coordinate information of each karst cave is mapped into a three-dimensional coordinate system, and the coordinates of each karst cave are set as , wherein , and respectively represent the position coordinates of the karst cave. Based on graphics technology, create a three-dimensional model according to the coordinate information of each karst cave in the three-dimensional model to create corresponding geometric bodies. The classified cave attributes are mapped into the three-dimensional model, using different colors, textures, and markers to represent different categories of caves.

4. The method according to claim 1, wherein, The grouting path generation method is as follows: initializing path planning parameters, including a start point and a target point ; An empty open list O and a closed list C are created for storing nodes to be explored and explored nodes, the starting point is added to the open list O, and its initial cost is set to 0; Enter the loop until the open list O reaches the end point, specifically: Select the node with the lowest cost from the open list O as the current node; If the current node is the end point, the path planning is completed, and the loop is exited. move the current node from the open list O to the closed list C; traversing neighbor nodes of the current node ; Starting from the end point, trace back along the parent nodes of each node until reaching the starting point to form a path.

5. The method according to claim 1, wherein, The sensors include a total station, a laser scanner, a water level monitoring sensor, and a grouting effect monitoring sensor, which are used to measure the changes in ground subsidence in real time, and the sensors are evenly installed on the ground surface near the tunnel; The groundwater level monitoring sensor is installed in the tunnel well to monitor the changes in groundwater level, as well as the changes in groundwater level during grouting; The grouting effect monitoring sensor is installed near the injection port of the grouting device to monitor the flow rate and flow of the grouting material in real time.

6. The method of claim 1, wherein the method is characterized by, The sensors are connected to a central real-time monitoring system through wireless connection, which is used to receive and record the data of the sensors in real time; Threshold values are set in the monitoring system to issue an alarm when the ground subsidence, groundwater level, and grouting effect exceed the safe range.

7. A karst tunnel large-scale cave treatment system, characterized in that, It includes: A data analysis platform that uses computer vision and machine learning algorithms to analyze and visualize the shape, distribution, and attributes of caves; The data analysis platform is used to: Convert geological survey and advanced detection data from the original format to the data analysis platform processing format and perform data preprocessing; Import the preprocessed geological data and advanced detection data into the data storage library in the platform; Based on computer vision and machine learning algorithms, automatically identify and classify caves; Based on the identified and classified cave data, use computer graphics technology to generate a three-dimensional model of the caves, mapping the coordinates and attributes of the underground caves to three-dimensional space; Render the three-dimensional model and visualize the shape, distribution, and attributes of the caves; A real-time monitoring system with a total station, laser scanner, water level monitoring sensor, and grouting effect monitoring sensor is used to monitor ground subsidence, groundwater level, and grouting effect in real time; The control center issues an alarm when the ground subsidence, groundwater level, and grouting effect exceed the safe range based on the monitoring data and built-in threshold values; The automatic identification and classification of caves specifically includes: Convert the data obtained by geological survey and advanced detection into numerical feature vectors and perform standardization processing; Divide the data into training and test sets; Use support vector machines (SVM) and convolutional neural networks (CNN) to train the model to learn the features of the caves from the input data and perform classification; ， wherein is a loss function, is a model parameter, N is the number of geological survey training samples, is actual survey data, is an output prediction model; The minimum loss function of the training process is defined as: The automatic identification and classification of caves also includes evaluating the trained model using the test set: Based on the machine learning model, predict new geological data and detection data, and assign a class label to each data point indicating whether it is a cave or not. The class classification converts the output of the model into a probability distribution based on the softmax function, and selects the class with the highest probability, specifically: ， where denotes the probability that data point i belongs to class j given input x, is the output of the model, denoting the score of data point i belonging to class j, and K is the total number of classes.

Images