Tunnel composite disaster intelligent early warning method and system based on propensity digital twinning

Abstract

The application provides a tunnel composite disaster intelligent early warning method and system based on a tendency digital twin, belongs to the technical field of tunnel gushing water data processing and analysis, and comprises the following steps: obtaining three-dimensional point cloud data of a tunnel structure and design parameters of the tunnel structure, constructing a digital twin model based on the three-dimensional point cloud data and the design parameters of the tunnel structure; collecting monitoring data in the tunnel structure in real time and integrating the collected data into the digital twin model; modeling the time-space features of the monitoring data in the tunnel structure collected in real time by using a space-time graph convolution network, outputting a time-space feature matrix, that is, generating a time-space feature representation of a disaster factor; inputting the time-space feature representation of the generated disaster factor into a Bayesian network, and the Bayesian network is used for further modeling the dependency relationship between different disaster factors and calculating a disaster occurrence probability; and the digital twin model identifies a potential disaster area in the tunnel based on the calculated disaster occurrence probability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of tunnel gushing water data processing and analysis, and particularly relates to a tunnel composite disaster intelligent early warning method and system based on a tendency digital twin. BACKGROUND

[0002] The statements in this section merely provide background information related to the application and do not necessarily constitute prior art.

[0003] Tunnel engineering is an important part of transportation and infrastructure construction. With the continuous expansion of its scale and the increasing complexity, the risk of tunnel disasters is also increasing. Especially in complex geological conditions, the tunnel interior may be affected by multiple disaster factors such as groundwater seepage and stress changes. These factors often have a mutual dependence relationship and may cause composite disasters such as water bursting and landslides, which seriously threaten the safety of tunnel structures and the reliability of engineering operation.

[0004] Existing tunnel disaster monitoring technology mainly relies on a single data source or simple monitoring means, which cannot comprehensively and accurately reflect the relationship between complex disaster factors inside the tunnel. In addition, traditional monitoring systems also have great limitations in real-time data processing and disaster tendency prediction, lacking efficient prediction and early warning capabilities for disasters, making it difficult to meet the needs of multi-source data fusion and real-time disaster assessment in modern tunnel engineering.

[0005] With the development of digital twin technology, it is possible to build realistic virtual models based on monitoring data of actual projects and update them in real time with sensor data, providing new possibilities for tunnel disaster monitoring. However, existing digital twin technology still has some deficiencies in disaster tendency prediction, especially when facing composite disaster scenarios. How to analyze multi-source monitoring data in real time and make accurate predictions remains a difficult problem to be solved. At the same time, traditional disaster monitoring methods rely mainly on static data or lagging risk assessment, and cannot provide dynamic and closely integrated emergency response capabilities with the on-site environment. Therefore, there are problems of real-time and accuracy in the tendency prediction of tunnel composite disasters. SUMMARY

[0006] To overcome the deficiencies of the prior art, the present application provides a tunnel composite disaster intelligent early warning method based on a tendency digital twin, which combines digital twin technology with a disaster tendency prediction model and uses augmented reality (AR) technology to realize real-time disaster monitoring and assessment, effectively improving the response speed and reliability of the early warning system, and is particularly suitable for disaster monitoring and emergency management in complex tunnel engineering.

[0007] To achieve the above purpose, one or more embodiments of the present application provide the following technical solutions:

[0008] In a first aspect, a tunnel composite disaster intelligent early warning method based on a tendency digital twin is disclosed, comprising:

[0009] obtaining three-dimensional point cloud data of a tunnel structure and design parameters of the tunnel structure, and constructing a digital twin model based on the three-dimensional point cloud data of the tunnel structure and the design parameters;

[0010] real-time collection of monitoring data in the tunnel structure and integration of the collected data into the digital twin model;

[0011] modeling the spatio-temporal features of the real-time collected monitoring data in the tunnel structure using a spatio-temporal graph convolution network, outputting a spatio-temporal feature matrix, i.e., generating a spatio-temporal feature representation of a disaster factor;

[0012] inputting the generated spatio-temporal feature representation of the disaster factor into a Bayesian network, the Bayesian network being used to further model the dependency relationship between different disaster factors and calculate a disaster occurrence probability;

[0013] the digital twin model identifies a potential disaster area in the tunnel based on the calculated disaster occurrence probability.

[0014] As a further technical solution, a three-dimensional laser is used to scan the tunnel to obtain the three-dimensional point cloud data of the tunnel structure.

[0015] As a further technical solution, the constructed digital twin model displays the spatial position information of all monitoring devices and the real-time data fed back by the monitoring devices.

[0016] The spatial position information of the monitoring devices includes the stake number, installation height, and depth of installation of the device inside the surrounding rock.

[0017] As a further technical solution, the collected data is integrated into the digital twin model, specifically comprising:

[0018] real-time collection of monitoring data in the tunnel structure by monitoring devices arranged in the tunnel structure, transmission of the data collected by the monitoring devices through an Internet of Things device, and mapping of the data with virtual monitoring points in the digital twin model.

[0019] As a further technical solution, the spatio-temporal graph convolution network is a spatio-temporal graph structure, modeling the tunnel monitoring points as nodes in the graph, and reflecting the spatial correlation between the nodes through edges.

[0020] The spatio-temporal graph convolution network includes a graph convolution layer and a time convolution layer; the graph convolution layer is used to capture the spatial dependency relationship of stress or environmental data between different regions of the tunnel; and the time convolution layer is used to extract the time variation trend of the monitoring data, thereby generating a spatio-temporal feature representation of a disaster factor.

[0021] As a further technical solution, the Bayesian network constructs a causal relationship graph between the disaster factors through nodes and directed edges, each node representing a disaster factor, including tunnel stress, underground water level, and the directed edges representing the causal relationship between different factors.

[0022] Through conditional probability calculation of the Bayesian network, the risk of each disaster factor is predicted according to the current data, wherein the Bayesian network can model the process of rainstorm leading to the rise of underground water level and further affecting the change of tunnel stress as a probability dependent chain, and calculate the probability of tunnel structure damage.

[0023] As a further technical solution, it also includes connecting the digital twin model to the AR device, superimposing the three-dimensional tunnel structure in the digital twin model, real-time monitoring data, and disaster prediction results into the real scene, and the AR device enables the tunnel manager to directly view the potential risk area, monitoring device location and real-time data on site.

[0024] In a second aspect, a tunnel composite disaster intelligent early warning system based on propensity digital twin is disclosed, comprising:

[0025] The digital twin model construction module is configured to obtain three-dimensional point cloud data of the tunnel structure and design parameters of the tunnel structure, and construct a digital twin model based on the three-dimensional point cloud data of the tunnel structure and the design parameters.

[0026] Real-time monitoring data in the tunnel structure is collected and integrated into the digital twin model;

[0027] The spatio-temporal feature representation module is configured to model the spatio-temporal features of the real-time collected monitoring data in the tunnel structure using a spatio-temporal graph convolution network, output a spatio-temporal feature matrix, and generate a spatio-temporal feature representation of the disaster factor.

[0028] The disaster occurrence probability calculation module is configured to input the generated spatio-temporal feature representation of the disaster factor into a Bayesian network, which is used to further model the dependency relationship between different disaster factors and calculate the disaster occurrence probability.

[0029] The disaster area identification module is configured to identify the potential disaster area in the tunnel based on the calculated disaster occurrence probability by the digital twin model.

[0030] The above one or more technical solutions have the following beneficial effects:

[0031] The method of the present application combines a spatio-temporal graph convolution network (ST-GCN) and a Bayesian network to extract spatio-temporal features and model the causal relationship of disaster factors from the multi-source monitoring data of the tunnel. The ST-GCN is used to extract the spatio-temporal dynamic change features in the tunnel monitoring point data, capture the spatial correlation between different regions of the tunnel and its time change trend, and the Bayesian network is used to calculate the interaction of each disaster factor and its influence on the safety of the tunnel through conditional probability calculation. Based on this, the present application can accurately predict the probability of disaster occurrence in the multiple disaster situation of the tunnel, and dynamically adjust the prediction result, ensuring the real-time and accuracy of the disaster tendency prediction.

[0032] The technical scheme of the present application can realize real-time monitoring and intelligent early warning of composite disasters, combine digital twin technology and advanced disaster prediction model, and use augmented reality (AR) technology to seamlessly integrate virtual monitoring results with the real environment, solving the problem of tunnel disaster monitoring.

[0033] The advantages of the additional aspects of the present application will be partially given in the following description, partially become obvious from the following description, or be known through the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0034] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application, and the illustrative embodiments of the present application and their description serve the purpose of explaining the present application, and do not constitute improper limitations on the present application.

[0035] Fig. 1 The present application is a digital twin system overall structure diagram for the embodiment of the present application.

[0036] Fig. 2 The present application is a composite disaster risk prediction and emergency response flowchart for the embodiment of the present application. DETAILED DESCRIPTION

[0037] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as generally understood by those skilled in the art to which the present application belongs.

[0038] It should be noted that the terms used herein are only for the purpose of describing the specific embodiments, and are not intended to limit the exemplary embodiments according to the present application.

[0039] In the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0040] Embodiment one

[0041] Referring to the accompanying drawings Figs. 1-2As shown, the embodiment discloses a tunnel composite disaster intelligent early warning method based on tendency digital twinning, which includes the following steps:

[0042] Step one: obtain the three-dimensional point cloud data of the tunnel structure and the design parameters of the tunnel structure, and construct a digital twinning model based on the three-dimensional point cloud data of the tunnel structure and the design parameters;

[0043] Step two: real-time collection of monitoring data in the tunnel structure and integration of the collected data into the digital twinning model;

[0044] Step three: modeling the spatio-temporal features of the real-time collected monitoring data in the tunnel structure using a spatio-temporal graph convolution network, outputting a spatio-temporal feature matrix, i.e. generating a spatio-temporal feature representation of the disaster factors;

[0045] Step four: input the generated spatio-temporal feature representation of the disaster factors into a Bayesian network, which is used to further model the dependency relationship between different disaster factors and calculate the disaster occurrence probability;

[0046] Step five: the digital twinning model identifies the potential disaster area in the tunnel based on the calculated disaster occurrence probability.

[0047] In step one, specifically including:

[0048] 1-1) Basic construction of digital twinning model, including tunnel real scene model construction and monitoring device layout and data display;

[0049] 1-2) Real-time data integration and feedback of digital twinning model;

[0050] 1-3) Composite disaster tendency prediction and risk area identification, including disaster risk area identification and real-time risk update;

[0051] 1-4) Dynamic visualization display of disaster occurrence process, including dynamic disaster process simulation and user interactive experience;

[0052] 1-5) Integration and application of digital twinning and composite disaster prediction, including intelligent early warning and emergency response and historical data analysis and future trend prediction.

[0053] In this embodiment, 1-1) the basic construction of digital twinning model, the specific process is: the core basis of digital twinning model is to construct the real scene model of the tunnel through three-dimensional point cloud data. This process can obtain accurate three-dimensional data of the tunnel structure through laser scanning or other three-dimensional scanning technology. Import these point cloud data using Unity 3D engine, combine with the geometric design data of the tunnel, generate a highly realistic tunnel model.

[0054] Data types collected: 3D point cloud data, tunnel design parameters, material properties (type of surrounding rock, lining material), geological information.

[0055] Model construction process: Point cloud data is combined with design parameters to ensure that the virtual model is consistent with the structure of the actual tunnel, providing a reliable physical basis for subsequent dynamic simulation.

[0056] In this embodiment, the design parameters refer to the width and height information of the tunnel, with the aim of ensuring that the model constructed using point cloud data is consistent with the actual tunnel size, or has a very small error. Since point cloud data can only provide a spatial size model, it cannot obtain material properties and geological information, so design parameters need to be added.

[0057] Monitoring device layout display: The digital twin model not only displays the tunnel statically, but also includes the spatial position information of all monitoring devices and real-time data feedback. In Unity 3D, the devices need to be accurately laid out, and the positions of the monitoring devices and real-time monitoring data need to be labeled in the three-dimensional model; Monitoring device position information: The stake number, installation height, and depth of the device installed in the surrounding rock of the device need to be labeled in detail, directly positioning the specific spatial position of the device, ensuring that the actual position of each monitoring device is accurately reflected in the digital model.

[0058] In step 1-2), real-time data integration and feedback of the digital twin model, the specific process is: The key advantage of the digital twin system is its real-time synchronization capability with field data. Through continuous data collection of monitoring devices, the system can real-time feedback the current physical state of the tunnel and the working condition of the device, and integrate these data into the twin model. In this implementation patent, the monitoring devices include various sensors.

[0059] Real-time monitoring data display: Real-time monitoring data of various sensors (such as multi-point displacement meters, stress meters, and osmotic pressure meters) is directly fed back to the twin model through an interface. The system can visualize the monitoring results in the form of data windows, charts, or heat maps next to the position of the monitoring device, for example, real-time display of stress, displacement change, or osmotic pressure data.

[0060] Sensor data is transmitted through Internet of Things (IoT) devices and mapped to virtual monitoring points in the digital twin model. The data stream of each sensor can be input into the Unity 3D model through the MQTT (Message Queuing Telemetry Transport) data transmission protocol.

[0061] In step 1-3), composite catastrophe tendency prediction and risk area identification, the specific process is:

[0062] The composite catastrophe tendency prediction method is: combining a space-time graph convolution network (ST-GCN) and a Bayesian network to perform composite catastrophe tendency prediction, by modeling the probability dependence between space-time feature extraction and catastrophe factors, forming an innovative composite catastrophe prediction method. Specifically, the ST-GCN is used to extract the space-time dynamic change features in the multi-source monitoring data monitored by various sensors, and the Bayesian network is used to capture the causal relationship between the catastrophe factors, and the possibility of catastrophe occurrence is calculated through conditional probability.

[0063] The digital twin model can exhibit the prediction of catastrophe tendency based on the space-time graph convolution network ST-GCN and the Bayesian network, and further realize the visualization display of the disaster occurrence probability, even the disaster occurrence scale and form.

[0064] Firstly, the space-time graph convolution network ST-GCN models the space-time features of the tunnel monitoring data to form the basis of catastrophe tendency analysis.

[0065] The space-time graph structure of the space-time graph convolution network ST-GCN models the tunnel monitoring points as nodes in the graph, the nodes represent the arrangement positions of the monitoring devices, and the nodes are connected by edges to reflect the spatial correlation. The space-time graph structure is a graph for displaying the spatial position relationship of the tunnel monitoring points. The above space-time graph convolution network includes two layers of graph convolution and time convolution. The graph convolution layer is used to capture the spatial dependence relationship of stress or environmental data between different regions of the tunnel, and the time convolution layer is used to extract the time change trend of the monitoring data, and then generate the space-time feature representation of the catastrophe factors. Through this space-time joint modeling, the system can fully consider the dynamic changes of the environmental conditions inside and outside the tunnel, such as the gradual accumulation of stress or the long-term impact of rainfall increase on the tunnel structure.

[0066] In this embodiment, for a given N monitoring points in the tunnel structure, an undirected weighted graph G=(V,E,A) is constructed, where V={v1,v2,…,vN} represents N monitoring points; E represents the connection relationship between the monitoring points; A∈RN×N is an adjacency matrix that defines the connection strength between points and points, and is calculated as follows: n N×N

[0067] ;

[0068] where x i , x j are the coordinates of monitoring points i, j, and σ is a normalization factor. This method can ensure that the weights between adjacent sensors are high, and the connection between distant points is weak.

[0069] ​​By leveraging the convolutional operations of the graph convolutional layers in a graph network (GCN), information is propagated within the sensor network to extract spatial topological features. Graph convolution operations are used to aggregate information from neighboring nodes to update node features.

[0070] ;

[0071] in, It is the first l The node feature matrix of the layer; It is the first l Trainable weights of the layer; It is an adjacency matrix with self-loops; yes The degree matrix, ; This is the activation function.

[0072] The node feature matrix described above originates from the attribute information of nodes in the original data. Each node represents a monitoring point, and its corresponding features include monitoring information such as stress, displacement, and seepage pressure. These features are constructed into a feature matrix, which serves as the initial input to the GCN model.

[0073] An adjacency matrix with self-loops is an adjacency matrix A based on a graph plus an identity matrix (self-loop), where... A The adjacency structure from the input graph data is obtained based on the graph model (such as spatial adjacency, functional connectivity, etc.) constructed in actual applications.

[0074] After spatial feature extraction, one-dimensional convolution (T-Conv) is used for time series feature extraction:

[0075] ;

[0076] Among them, W t These are the temporal convolution weights, and the model's trainable parameters, which are automatically updated by the backpropagation algorithm during training. Their initial values ​​use either Xavier or Heinrich's initialization strategy, depending on the model architecture. This represents a one-dimensional convolution operation, where b is the bias term. is the activation function. X is the original input data tensor.

[0077] After processing by ST-GCN, the final output spatiotemporal feature matrix is:

[0078] ;

[0079] In the formula: H X is the spatiotemporal feature matrix, which will be used in a Bayesian network to establish the dependencies between different catastrophic factors; X is the original input data tensor; A is the adjacency matrix.

[0080] After obtaining the spatio-temporal characteristics of the disaster factors, Bayesian networks are used to further model the dependencies between different disaster factors. Bayesian networks construct a causal graph between disaster factors through nodes and directed edges, with each node representing a disaster factor, such as tunnel stress or groundwater level, and directed edges representing causal relationships between different factors. Through conditional probability calculations in Bayesian networks, the system can predict the risk of each disaster factor based on current data. For example, Bayesian networks can model the process of heavy rain causing groundwater levels to rise, which in turn affects changes in tunnel stress, as a chain of probabilistic dependencies and calculate the probability of tunnel structure failure.

[0081] Bayesian networks are a model based on probabilistic reasoning that can use conditional probability calculations to predict the risk of each disaster factor. The basic theoretical foundation of Bayesian networks is Bayes' theorem:

[0082] .

[0083] In disaster risk assessment, assume that multiple disaster factors in the system exist in the form of random variables, and they have causal relationships. Then the Bayesian network can be represented as a directed acyclic graph (DAG), where each node represents a random variable and each directed edge represents a causal relationship.

[0084] The joint probability distribution in the Bayesian network can be represented as:

[0085] ;

[0086] where represents the parent node set of node , i.e., the direct antecedents that affect the variable.

[0087] In risk prediction, the goal is to calculate the probability of a specific disaster event R occurring, given some observed data E. The posterior probability needs to be calculated:

[0088] .

[0089] If the conditional probabilities between disaster factors are known, Bayesian inference can be used for updating. For example, assume that the conditional probability between a disaster factor and the disaster risk R is:

[0090] ;

[0091] where the denominator can be calculated by the total probability formula:

[0092] .

[0093] In the Bayesian network, the maximum posterior estimation or the expectation maximization method can be used to infer the incomplete data, thereby improving the prediction accuracy of the catastrophe risk.

[0094] The combination of ST-GCN and Bayesian network reflects the input of the spatio-temporal features extracted by ST-GCN into the Bayesian network as the conditional probability basis of each catastrophe factor. The features extracted by the spatio-temporal convolution network can more accurately represent the spatio-temporal dynamic changes of each catastrophe factor, and then the Bayesian network uses these spatio-temporal features to calculate the conditional probability between each catastrophe factor, thereby realizing the joint probability prediction of the catastrophe occurrence. This prediction method combines the spatio-temporal feature extraction capability of ST-GCN and the causal dependence modeling advantage of the Bayesian network, forming a highly accurate composite catastrophe prediction framework.

[0095] Through this method, the system can not only predict the catastrophe risk according to historical data and real-time data, but also dynamically update the conditional probability distribution of the Bayesian network according to new monitoring data, and real-time adjust the prediction result of the catastrophe tendency. In this way, the system can automatically adjust the evaluation of the possible future catastrophe when monitoring that the tunnel stress or groundwater level has changed significantly, ensuring the real-time and accuracy of the catastrophe tendency prediction.

[0096] In summary, by combining ST-GCN and Bayesian network, the present application can consider both spatio-temporal dynamic changes and the dependence relationship between catastrophe factors, significantly improving the accuracy of composite catastrophe tendency prediction.

[0097] Furthermore, the long-term accuracy and adaptability of the catastrophe tendency prediction model are improved through a self-learning mechanism. After each catastrophe prediction, the actual catastrophe occurrence and the prediction result are compared to adjust the parameters of the model to reduce the prediction error. Specifically, the system will construct a reward function to evaluate the prediction accuracy, and continuously optimize the model through deep reinforcement learning. This feedback mechanism is based on the accumulation of historical data, and gradually adjusts the network weights and parameters, so that the model has self-adaptive ability.

[0098] Regarding the construction of the reward function, a feedback mechanism is designed to evaluate the performance of the model based on the difference between the prediction result and the actual occurrence. Specifically, the reward function will optimize the prediction accuracy of the model to reduce the prediction error and improve the long-term accuracy and adaptability. Through this feedback mechanism, the system can adjust the model parameters after each catastrophe prediction, gradually enhancing the self-adaptive ability of the model.

[0099] For example, when the system detects a large error in a prediction, it adjusts the weights of relevant disaster factors and model parameters based on the actual situation. If a prediction indicates that the risk of stress imbalance in a tunnel structure is low, but a serious stress overload actually occurs, the model will increase the weight of structural stress in that area and optimize the stress transfer model to improve prediction accuracy. At the same time, the system will update the disaster prediction model in real time through online learning methods. Whenever new monitoring data is input, the system will immediately adjust the model parameters to ensure that the model can adapt to new environmental data and update the prediction results in real time.

[0100] Risk area identification: The digital twin system can identify potential disaster areas in the tunnel based on the disaster occurrence probability calculated by the ST-GCN and Bayesian network. For example, a specific area of a tunnel may face the dual risks of water inrush and collapse due to rising groundwater levels and increased tunnel stress, and this area will be identified as a high-risk area.

[0101] Probability value display: The probability of disaster occurrence will be displayed in percentage form next to the risk area in the digital twin model. For example, in a certain tunnel section, the system will display a 35% probability of collapse within the next 24 hours. These probability values are dynamically updated as monitoring data is input, ensuring real-time understanding of the likelihood of disaster occurrence.

[0102] Color and shape differentiation: Different disaster risk areas can be identified by different colors, such as red for high-risk areas, yellow for medium-risk areas, and green for low-risk areas. Each potential disaster type (such as collapse, water inrush, etc.) can be distinguished by different shapes (such as triangles, circles).

[0103] Dynamic update: With the update of monitoring data, the Bayesian network can adjust the probability of disaster occurrence in real time, and the identified areas in the digital twin model will also change dynamically, ensuring real-time reflection of the latest prediction results of disasters.

[0104] In addition, in this embodiment, the dynamic visualization of the disaster occurrence process is displayed, and the specific process is as follows: the combination of digital twin and AR can realize seamless connection between virtual and reality, and provide real-time disaster monitoring and risk assessment solutions. By superimposing the three-dimensional tunnel structure in the digital twin model, real-time monitoring data and disaster prediction results onto the real scene, AR technology enables tunnel managers to directly view potential risk areas, monitor equipment locations and real-time data on site. Such a combination greatly improves the efficiency and accuracy of emergency response, enabling managers to make faster and more accurate decisions in real environments. In addition, the powerful computing power of digital twin ensures the reliability of disaster prediction, and the convenient interaction of AR enhances the user's on-site operation experience. The combination of the two not only improves the practicality of the system, but also reduces the risk of misjudgment, making monitoring and early warning in complex disaster environments more intelligent and intuitive.

[0105] Regarding the connection method of the model and the AR device: a cloud server can be used to process the digital twin model to ensure that computing and real-time data processing are not limited by performance. Data can be transmitted to the AR device through a wireless network to realize interaction between the model and reality.

[0106] Regarding data transmission and processing: fast and reliable data transmission protocols such as MQTT or WebSocket can be used to ensure that data transmission from the digital twin model to the AR device is real-time and stable.

[0107] Regarding interaction and updates: for user interaction with the AR device, the goal is to allow users to view disaster risk information in real time and respond accordingly. The AR device will display the necessary information based on the updated content of the digital twin model, helping users make quick decisions.

[0108] The digital twin model not only displays the current risk status, but also simulates the occurrence and development process of future disasters. Users can further understand the detailed information and dynamic evolution process of potential disasters by clicking on high-risk areas.

[0109] Clickable identification of risk areas: users can click on the marked risk areas in the model to view detailed information about the area.

[0110] Detailed information display: Each risk area should display information about the type of disaster, probability of occurrence, and spatiotemporal data of the current state. Disaster type: For example, show that the area is a water inrush, collapse, stress imbalance, or a composite area of multiple disasters; Disaster occurrence probability: Show the probability of disaster occurrence calculated by the Bayesian network, specifically in the form of percentage. For example, "There is a 35% probability of collapse in this area within the next 24 hours"; Current monitoring data: Show the current real-time monitoring data of the area, including stress, displacement, groundwater level, temperature, and other multi-source sensor collection information. For example, "The current tunnel stress is 250kPa, and the groundwater level has risen by 10cm".

[0111] Dynamic update and feedback: When real-time monitoring data changes, users can see the dynamic changes in the risk state, detailed information, and disaster tendency of the area in the model.

[0112] Interactive functions can also show the dynamic development process of potential disasters. When users click on a risk area, the system can simulate the evolution of the disaster process through visualization.

[0113] Timeline sliding function: Set a timeline to allow users to slide the timeline to view the future trend of the disaster. By predicting future features with ST-GCN, show how the disaster gradually develops from small-scale changes to global disasters. For example, users can view the expansion range and speed of water inrush within the next 12 hours.

[0114] Disaster propagation path: Use animations or heat maps to show the propagation path and impact range of the disaster. For example, when water inrush occurs, the system can display the propagation path of the water flow from the water inrush point to other areas of the tunnel, and update the dynamic data such as water flow speed and water accumulation in real time.

[0115] Emergency plan suggestions: Based on risk scores and disaster occurrence probabilities, the system can give corresponding emergency suggestions according to the dynamic process of the disaster. For example, suggest reinforcing a certain section of the tunnel or evacuating personnel from the relevant area.

[0116] Integration and application of digital twin and composite disaster prediction:

[0117] The digital twin system combines real-time data with the composite disaster tendency prediction model to become a complete tunnel disaster management platform. The system inputs real-time data from monitoring devices, dynamically interacts with the prediction model, and visualizes and simulates disaster risks in the model.

[0118] This embodiment of the technical solution embeds the composite disaster prediction model into the numerical twin model, so that the digital twin model not only has the function of data display, but also can combine data to predict disasters, and the disaster prediction results and disaster process can be visualized through the digital twin model.

[0119] Intelligent early warning and emergency response: When the risk of catastrophe reaches a certain threshold, the digital twin model will automatically trigger an early warning and issue a warning to the management personnel through a visual interface, and can provide emergency response plans according to the risk assessment.

[0120] Historical data analysis and future trend prediction: The system can store historical monitoring data to help analyze long-term trends. In addition, through compound catastrophe tendency prediction, the system can simulate possible future catastrophe scenarios to provide forward-looking support for tunnel safety maintenance.

[0121] In the technical solution of the embodiment, the digital twin model integrates three-dimensional point cloud data, tunnel design parameters, and monitoring device data to construct a realistic virtual tunnel model. Real-time monitoring data is fed back to the model to show the real-time changes of physical quantities such as stress, groundwater level, and osmotic pressure in the tunnel. With the help of the model, the system can reflect the internal operating state of the tunnel in real time and identify potential risk areas based on the results of catastrophe tendency prediction. At the same time, augmented reality (AR) technology is used to superimpose monitoring data and disaster warning information in the virtual model onto the real tunnel scene. Users can directly view risk areas, monitoring device locations, and their data states on site through mobile devices or AR glasses, greatly improving emergency response efficiency.

[0122] The advantage of the technical solution of the embodiment is that it realizes real-time visualization of complex tunnel structures and efficient integration of disaster prediction through digital twin technology. It improves the convenience of on-site operations and the accuracy of decision-making using augmented reality technology, enabling timely capture of subtle changes in catastrophe factors and response, thereby effectively reducing the risk of tunnel compound catastrophe.

[0123] The technical solution of the embodiment integrates spatiotemporal graph convolutional networks (ST-GCN) and Bayesian networks to analyze tunnel multi-source monitoring data in real time, capture spatiotemporal dynamic changes of catastrophe factors, provide high-precision catastrophe tendency prediction, and update prediction results in real time when monitoring data changes. This ensures the real-time and accuracy of catastrophe prediction and provides more reliable support for tunnel disaster management.

[0124] The technical solution of the embodiment integrates real-time data from monitoring devices with virtual models through digital twin technology, enabling integrated analysis of multi-source data. Combined with Internet of Things (IoT) devices and augmented reality (AR) technology, tunnel management personnel can directly view risk areas and monitoring data in the actual environment, greatly improving the efficiency and accuracy of on-site emergency decision-making.

[0125] The feedback and self-learning mechanism of the disaster prediction model of the embodiment subtechnical solution optimizes the model parameters gradually through the accumulation of historical data, so that the prediction result is more accurate, and the misjudgment risk caused by the dependence on static data or lag data in the traditional monitoring method is reduced. Through the joint probability calculation of the disaster factors, the application can identify potential disasters in a complex environment, and help the management personnel to make correct judgments through AR display.

[0126] The embodiment subtechnical solution can not only identify potential high-risk areas in real time, but also dynamically display the process of disaster occurrence and expansion through visualization, and automatically provide emergency response suggestions according to risk assessment, so as to ensure that the management personnel can take effective measures in time before the disaster occurs, and reduce the impact of the disaster on the safety of the tunnel.

[0127] The embodiment subtechnical solution is suitable for various complex underground engineering projects, especially in the complex disaster scenarios such as gushing water and collapse in tunnel engineering. Through the combination of digital twinning technology and AR, the system not only improves the intelligent level of disaster monitoring, but also greatly expands the application field of digital twinning technology in disaster prediction and risk management.

[0128] Embodiment two

[0129] The purpose of the embodiment is to provide a computer device, including a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize the steps of the above method.

[0130] Embodiment three

[0131] The purpose of the embodiment is to provide a computer readable storage medium.

[0132] A computer readable storage medium has a computer program stored thereon, and the program is executed by a processor to execute the steps of the above method.

[0133] Embodiment four

[0134] The purpose of the embodiment is to provide a tunnel complex disaster intelligent early warning system based on tendency digital twinning, which includes:

[0135] The digital twinning model construction module is configured to obtain three-dimensional point cloud data of the tunnel structure and design parameters of the tunnel structure, and construct a digital twinning model based on the three-dimensional point cloud data of the tunnel structure and the design parameters;

[0136] Real-time monitoring data in the tunnel structure are collected and integrated into the digital twinning model;

[0137] The spatio-temporal feature representation module is configured to model spatio-temporal features of the monitoring data collected in real time in the tunnel structure by using a spatio-temporal graph convolution network, and output a spatio-temporal feature matrix, i.e., generate the spatio-temporal feature representation of the disaster factor;

[0138] The disaster occurrence probability calculation module is configured to input the spatio-temporal feature representation of the disaster factor into a Bayesian network, which is used to further model the dependency relationship between different disaster factors, and calculate the disaster occurrence probability.

[0139] The disaster area identification module is configured to identify the area of the potential disaster in the tunnel based on the calculated disaster occurrence probability by the digital twin model.

[0140] Embodiment five

[0141] The purpose of the embodiment is to provide a computer program product containing instructions, which, when running on a computer, causes the computer to execute the method and functions involved in any of the above embodiments.

[0142] The steps and method embodiments involved in the above embodiments correspond to embodiment one, and the specific embodiments can refer to the related description part of embodiment one. The term "computer readable storage medium" should be understood to include a single medium or multiple media of one or more instruction sets; it should also be understood to include any medium that can store, encode or carry instruction sets for execution by a processor and cause the processor to perform any method in the present application.

[0143] Those skilled in the art should understand that the above modules or steps of the present application can be realized by a general computer device, and alternatively, they can be realized by device executable program codes, so that they can be stored in a storage device for execution by a computing device, or they can be made into individual integrated circuit modules, or a plurality of modules or steps can be made into a single integrated circuit module. The present application is not limited to any specific combination of hardware and software.

[0144] The above describes the specific embodiments of the present application in combination with the accompanying drawings, but is not a limitation on the scope of protection of the present application. Those skilled in the art should understand that various modifications or variations made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the scope of protection of the present application.

Claims

1. A tunnel compound disaster intelligent early warning method based on a propensity digital twin, characterized in that, The method comprises the following steps: obtaining three-dimensional point cloud data of a tunnel structure and design parameters of the tunnel structure, and constructing a digital twin model based on the three-dimensional point cloud data of the tunnel structure and the design parameters; real-time collection of monitoring data in the tunnel structure and integration of the collected data into the digital twin model; modeling of the time-space features of the real-time collected monitoring data in the tunnel structure by using a space-time graph convolution network, and output of a space-time feature matrix, i.e., generation of a space-time feature representation of a disaster factor; the space-time graph convolution network is a space-time graph structure, and the tunnel monitoring points are modeled as nodes in the graph, and the nodes are connected by edges to reflect the spatial correlation; the space-time graph convolution network comprises a graph convolution layer and a time convolution layer; the graph convolution layer is used to capture the spatial dependence of stress or environmental data between different regions of the tunnel; and the time convolution layer is used to extract the time variation trend of the monitoring data, and then generate the space-time feature representation of the disaster factor; the space-time feature representation of the disaster factor is input into a Bayesian network, the Bayesian network is used to further model the dependence between different disaster factors, and the probability of disaster occurrence is calculated; the Bayesian network constructs a causal relationship graph between disaster factors through nodes and directed edges, each node represents a disaster factor, including tunnel stress and underground water level, and the directed edges represent the causal relationship between different factors; the risk of each disaster factor is predicted according to the current data through conditional probability calculation of the Bayesian network, wherein the Bayesian network can model the process that heavy rain causes the underground water level to rise and further affects the change of the tunnel stress as a probability dependence chain, and calculate the probability of tunnel structure damage; the digital twin model identifies the area of potential disaster in the tunnel based on the calculated probability of disaster occurrence.

2. The tunnel compound catastrophe intelligent early warning method based on the tendency digital twin of claim 1, characterized in that, three-dimensional laser is used to scan the tunnel to obtain three-dimensional point cloud data of the tunnel structure; the digital twin model constructed displays the spatial position information of all monitoring devices and the real-time data fed back by the monitoring devices; the spatial position information of the monitoring devices comprises the stake number, installation height and depth of installation of the devices in the surrounding rock.

3. The tunnel compound catastrophe intelligent early warning method based on the tendency digital twin of claim 1, characterized in that, the collected data is integrated into the digital twin model, specifically comprising: real-time collection of monitoring data in the tunnel structure by using the monitoring devices arranged in the tunnel structure, and transmission of the data collected by the monitoring devices through an Internet of Things device and mapping of the data to virtual monitoring points in the digital twin model.

4. The tunnel compound catastrophe intelligent early warning method based on the tendency digital twin of claim 1, further comprising The method comprises the following steps: connecting the digital twin model to an AR device, superimposing the three-dimensional tunnel structure, real-time monitoring data and disaster prediction results in the digital twin model to the real scene, and enabling the tunnel management personnel to directly view the potential risk area, monitoring device position and real-time data on site through the AR device.

5. The tunnel compound disaster intelligent early warning system based on the tendency digital twin, characterized in that, The method comprises the following steps: a digital twin model construction module is configured to obtain three-dimensional point cloud data of a tunnel structure and design parameters of the tunnel structure, and construct a digital twin model based on the three-dimensional point cloud data of the tunnel structure and the design parameters; real-time collection of monitoring data in the tunnel structure and integration of the collected data into the digital twin model; The spatio-temporal feature representation module is configured to model the spatio-temporal features of the monitoring data collected in real time in the tunnel structure by using a spatio-temporal graph convolution network, and output a spatio-temporal feature matrix, i.e., generate the spatio-temporal feature representation of the disaster factor; The spatio-temporal graph convolution network is a spatio-temporal graph structure, modeling the tunnel monitoring points as nodes in the graph, and reflecting the spatial correlation between the nodes through edges; the spatio-temporal graph convolution network includes a graph convolution layer and a time convolution layer; The graph convolution layer is used to capture the spatial dependence of stress or environmental data between different regions of the tunnel; the time convolution layer is used to extract the time variation trend of the monitoring data, and further generate the spatio-temporal feature representation of the disaster factor; The disaster occurrence probability calculation module is configured to input the generated spatio-temporal feature representation of the disaster factor into a Bayesian network, which is used to further model the dependency relationship between different disaster factors and calculate the disaster occurrence probability; the Bayesian network constructs a causal relationship graph between disaster factors through nodes and directed edges, each node represents a disaster factor, including tunnel stress and underground water level, and the directed edges represent the causal relationship between different factors; the risk of each disaster factor is predicted according to the current data through the conditional probability calculation of the Bayesian network, wherein the Bayesian network can model the process that heavy rain causes the underground water level to rise and further affects the change of the tunnel stress as a probability dependency chain, and calculate the probability of tunnel structure damage; The disaster region identification module is configured to identify the region of potential disaster in the tunnel based on the calculated disaster occurrence probability by the digital twin model.

6. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the method of any one of claims 1-4.

7. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps of the method of any one of claims 1-4.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to perform the steps of the method of any one of claims 1-4.

Images