A tunnel health monitoring method and system based on distributed optical fiber sensing

Through distributed fiber sensing technology and pre-trained models, the association and potential risks between tunnel monitoring points are identified and risk warning signals are generated, which solves the problems of real-time monitoring and risk assessment in tunnel health monitoring, and improves the safety management of tunnels and Maintenance efficiency.

CN118670458BActive Publication Date: 2024-11-01CHENGDU CONSTR ENG GROUP CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411132021.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-01
Estimated Expiration
2044-08-19

AI Technical Summary

Technical Problem

Existing tunnel health monitoring methods are difficult to achieve comprehensive and real-time monitoring of tunnels, and it is difficult to accurately judge the health status correlation between adjacent monitoring points, resulting in insufficient risk assessment and early warning.

Method used

Using distributed fiber sensing method, the real-time data of each monitoring point in the tunnel is obtained and input into a pre-trained health correlation analysis model. Combining the tunnel physical model and propagation mode prediction model, the potential impact domain of the abnormal monitoring point is identified. and abnormal propagation intensity to generate risk warning signals.

Benefits of technology

It realizes accurate judgment of the health status correlation between tunnel monitoring points, improves the efficiency of risk assessment and early warning, and significantly improves the safety management level and maintenance efficiency of the tunnel.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118670458B_ABST
    Figure CN118670458B_ABST
Patent Text Reader

Abstract

The present application relates to a tunnel health monitoring method and system based on distributed optical fiber sensing, which belongs to the technical field of tunnel engineering monitoring. The health monitoring method includes: obtaining real-time monitoring data of each monitoring point in the tunnel; inputting the real-time monitoring data of each monitoring point into a pre-trained health association analysis model to obtain the health status result of each monitoring point; in response to the health status result of the monitoring point being an abnormal state, determining the monitoring point as an abnormal monitoring point; based on the pre-trained tunnel physical model and propagation mode prediction model, predicting the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point; obtaining the risk monitoring point and the corresponding risk level according to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring point; generating a risk warning signal according to the risk monitoring point and the corresponding risk level. The present application can accurately judge the health status association between adjacent monitoring points, and conduct effective risk assessment and warning accordingly.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of tunnel engineering monitoring, and in particular to a tunnel health monitoring method and system based on distributed optical fiber sensing. Background Art

[0002] Tunnel structures are affected by complex surrounding rock and geological and hydrological conditions, and the safety of their structures has always been a very important issue in engineering construction, operation and maintenance. The purpose of tunnel health monitoring is to detect tunnel damage and degradation and assess its safety status. It can effectively predict tunnel performance changes, provide reference and basis for tunnel repair and damage control, improve tunnel operation efficiency, and reduce accidents and the resulting loss of life and property.

[0003] At present, traditional tunnel health monitoring methods rely on manual periodic inspections and simple sensor measurements, which are not only time-consuming and labor-intensive, but also difficult to achieve comprehensive and real-time monitoring of tunnels. With the development of science and technology, especially the progress of distributed fiber optic sensing technology, machine learning and big data analysis technology, new solutions have been provided for tunnel health monitoring.

[0004] However, as a continuous structure, the various monitoring points of the tunnel are interdependent. How to accurately judge the health status correlation between adjacent monitoring points and conduct effective risk assessment and early warning based on this is a key issue that needs to be solved urgently. Summary of the invention

[0005] In order to accurately determine the health status association between adjacent monitoring points and conduct effective risk assessment and early warning based on this, the present application provides a tunnel health monitoring method and system based on distributed optical fiber sensing.

[0006] In a first aspect, the present application provides a tunnel health monitoring method based on distributed optical fiber sensing, which adopts the following technical solution:

[0007] A tunnel health monitoring method based on distributed optical fiber sensing, the health monitoring method comprising:

[0008] Obtain real-time monitoring data for each monitoring point in the tunnel;

[0009] The real-time monitoring data of each monitoring point is input into the pre-trained health association analysis model to obtain the health status results of each monitoring point;

[0010] In response to the health status result of the monitoring point being an abnormal state, determining the monitoring point as an abnormal monitoring point;

[0011] Based on the pre-trained tunnel physical model and propagation mode prediction model, the corresponding potential impact domain and abnormal propagation intensity are predicted according to the abnormal monitoring point;

[0012] According to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring point, a risk monitoring point and a corresponding risk level are obtained;

[0013] A risk warning signal is generated based on the risk monitoring points and the corresponding risk levels.

[0014] By adopting the above technical solution, the status data of each monitoring point is captured in real time based on the advantages of distributed fiber optic sensing technology. The correlation between monitoring points is identified and abnormal monitoring points are detected through the pre-trained health association analysis model. The propagation path prediction results output by the tunnel physical model and the propagation mode prediction results output by the propagation mode prediction model are combined to determine the potential impact domain and abnormal propagation intensity of the abnormal monitoring point, identify other risk monitoring points and risk levels that may be affected, and accurately judge the health status association between adjacent monitoring points, and conduct effective risk assessment and early warning based on this, greatly improving the safety management level and maintenance efficiency of the tunnel.

[0015] Optionally, the health monitoring method further includes a training step of the health association analysis model, and the training step includes:

[0016] Obtain historical monitoring data for each monitoring point in the tunnel;

[0017] Performing data preprocessing on the historical monitoring data;

[0018] Perform feature extraction on the historical monitoring data after data preprocessing to obtain the feature vector of the correlation between monitoring points;

[0019] Based on the association rule learning algorithm, determining the association rules between the monitoring points according to the feature vectors of the correlation between the monitoring points;

[0020] Constructing a monitoring point graph network of the tunnel, and obtaining a graph network analysis result based on graph convolutional network analysis;

[0021] Based on the association rules between the monitoring points and the graph network analysis results, a health association analysis model is trained.

[0022] By adopting the above technical solutions, the health association analysis model can identify and understand the interdependencies between different monitoring points and their normal behavior patterns by gaining insight into the complex relationships between monitoring points. When real-time monitoring data exhibits behavior inconsistent with these normal patterns, the model can identify the monitoring point as abnormal.

[0023] Optionally, based on a pre-trained tunnel physical model and a propagation mode prediction model, the step of predicting the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point includes:

[0024] Inputting the real-time monitoring data of the abnormal monitoring point into a pre-trained tunnel physical model to predict the stress propagation path and strain influence area of ​​the abnormal monitoring point;

[0025] Inputting the stress propagation path and strain influence area of ​​the abnormal monitoring point into a pre-trained propagation mode prediction model to predict the abnormal propagation mode of the abnormal monitoring point;

[0026] The stress propagation path, strain influence area and abnormal propagation mode of the abnormal monitoring point are data fused to obtain the potential influence domain and abnormal propagation intensity of the abnormal monitoring point.

[0027] By adopting the above technical solutions, combined with physical simulation and machine learning prediction, a precise and comprehensive method is provided to evaluate the potential impact domain and abnormal propagation intensity of abnormal monitoring points in the tunnel, so as to effectively guide the maintenance team to take targeted measures to improve the overall health and safety of the tunnel.

[0028] Optionally, the health monitoring method further includes a step of training the tunnel physical model, the step comprising:

[0029] Acquiring structural data of the tunnel and performing data preprocessing; wherein the structural data includes tunnel design parameters, construction material properties and historical monitoring data of multiple monitoring points;

[0030] Initializing parameters of the pre-built finite element model based on the tunnel design parameters and construction material properties;

[0031] Dividing the historical monitoring data of the plurality of monitoring points in the structural data into a first data set and a second data set;

[0032] Performing model parameter correction on the finite element model based on the first data set to obtain a calibrated finite element model;

[0033] The finite element model is cross-validated based on the second data set until the model performance meets a preset condition or the number of model iterations reaches a preset number, thereby obtaining the trained tunnel physical model.

[0034] By adopting the above technical solutions, the constructed tunnel physical model can accurately simulate the behavior of the tunnel under various conditions, providing a reliable analysis tool for tunnel health monitoring.

[0035] Optionally, the health monitoring method further includes a step of training the propagation mode prediction model, the step comprising:

[0036] Obtain monitoring data of historical abnormal events at multiple monitoring points in the tunnel;

[0037] Performing data preprocessing on the monitoring data of the historical abnormal events and dividing the data into a training set and a test set;

[0038] Training a pre-built machine learning model based on the training set, optimizing model parameters, and obtaining the propagation mode prediction model;

[0039] The propagation mode prediction model is tested and model parameters are corrected based on the test set until the loss function of the model meets a preset condition or the number of model iterations reaches a preset number, thereby obtaining the trained propagation mode prediction model.

[0040] By adopting the above technical solution, based on the machine learning algorithm, the propagation pattern of abnormal conditions of multiple monitoring points in the tunnel is learned, and then the trained and optimized model is deployed to the actual tunnel monitoring system, so as to accurately predict the abnormal propagation pattern of the monitoring points in the tunnel.

[0041] Optionally, the step of obtaining a risk monitoring point and a corresponding risk level according to the potential impact domain and the abnormal propagation intensity corresponding to the abnormal monitoring point includes:

[0042] Based on the monitoring point map network of the tunnel, the monitoring points located in the potential impact area are determined as risk monitoring points;

[0043] Determine the distance between the risk monitoring point and the abnormal monitoring point according to the position of the risk monitoring point in the monitoring point map network;

[0044] According to the size of the potential impact domain, the abnormal propagation intensity and the distance, a weighted calculation is performed based on preset weights to obtain the risk impact index of the risk monitoring point;

[0045] Based on the preset risk matrix, the corresponding risk level is determined according to the risk impact index of the risk monitoring point.

[0046] By adopting the above technical solutions, the monitoring points that may be affected can be quickly screened out, so as to concentrate resources and attention on those monitoring points that are most likely to face risks; at the same time, the obtained risk level provides decision makers with a clear risk classification, which helps decision makers to take corresponding prevention, preparation and response measures according to the risk level, so as to effectively manage and mitigate risks.

[0047] Optionally, after the step of generating a risk warning signal according to the risk monitoring point and the corresponding risk level, the step further includes:

[0048] According to the risk level, generating an emergency response plan corresponding to the risk level;

[0049] Based on the location information of the risk monitoring point, the risk warning signal and the emergency response plan are sent to a maintenance terminal corresponding to the risk monitoring point.

[0050] By adopting the above technical solutions, timely risk warnings and the push of emergency response plans can ensure that maintenance personnel respond quickly, allowing relevant personnel to take prompt measures to prevent further damage to the tunnel structure, thereby improving the safety of tunnel operations.

[0051] In the second aspect, the present application provides a tunnel health monitoring system based on distributed optical fiber sensing, which adopts the following technical solutions:

[0052] A tunnel health monitoring system based on distributed optical fiber sensing, the health monitoring system comprising:

[0053] An acquisition module is used to obtain real-time monitoring data of each monitoring point in the tunnel;

[0054] A health status result generation module is used to input the real-time monitoring data of each monitoring point into a pre-trained health association analysis model to obtain the health status result of each monitoring point;

[0055] an abnormal monitoring point determination module, configured to determine that the monitoring point is an abnormal monitoring point in response to the health status result of the monitoring point being an abnormal state;

[0056] An abnormal propagation prediction module, used to predict the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point based on the pre-trained tunnel physical model and propagation mode prediction model;

[0057] A risk monitoring point determination module is used to obtain the risk monitoring point and the corresponding risk level according to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring point;

[0058] The risk warning module is used to generate a risk warning signal of a corresponding level according to the risk monitoring point and the corresponding risk level and send it to the maintenance terminal.

[0059] In a third aspect, the present application provides a computer device, which adopts the following technical solution:

[0060] A computer device comprises a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method as described in the first aspect.

[0061] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0062] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any one of the methods in the first aspect.

[0063] In summary, the present application includes at least one of the following beneficial technical effects: by accurately judging the health status association between adjacent monitoring points and conducting effective risk assessment and early warning based on this, the safety management level and maintenance efficiency of the tunnel are greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a first flow chart of a tunnel health monitoring method according to one of the embodiments of the present application.

[0065] Figure 2 It is a second flow chart of the tunnel health monitoring method according to one of the embodiments of the present application.

[0066] Figure 3 This is a schematic diagram of the model structure of a graph convolutional network model of one of the embodiments of the present application.

[0067] Figure 4 It is a third flow chart of the tunnel health monitoring method according to one of the embodiments of the present application.

[0068] Figure 5 It is a fourth flow chart of the tunnel health monitoring method according to one of the embodiments of the present application.

[0069] Figure 6 It is a schematic diagram of the model structure of a finite element model of one of the embodiments of the present application.

[0070] Figure 7 It is a fifth flow chart of the tunnel health monitoring method according to one of the embodiments of the present application.

[0071] Figure 8 It is a schematic diagram of the model structure of a propagation mode prediction model of one of the embodiments of the present application.

[0072] Fig. 9 It is a sixth flow chart of the tunnel health monitoring method according to one of the embodiments of the present application.

[0073] Fig.10 It is the seventh flow chart of the tunnel health monitoring method of one of the embodiments of the present application. DETAILED DESCRIPTION

[0074] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1-10 It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0075] The embodiment of the present application discloses a tunnel health monitoring method based on distributed optical fiber sensing.

[0076] Reference Figure 1 , a tunnel health monitoring method based on distributed optical fiber sensing, the health monitoring method comprising:

[0077] Step S101, obtaining real-time monitoring data of each monitoring point in the tunnel;

[0078] In some embodiments, by deploying distributed optical fiber sensors in the tunnel, continuous monitoring can be performed over the entire length of the optical fiber to capture changes in parameters such as temperature, strain, displacement, humidity, etc.; when structural changes (such as cracks, deformation, etc.) occur at the monitoring points in the tunnel, the changes in these physical quantities can be captured by the optical fiber sensors;

[0079] It is understandable that distributed fiber optic sensing technology realizes real-time monitoring of tunnel structural integrity, can timely capture the status changes of each monitoring point, and provides a high-precision data basis for tunnel health monitoring.

[0080] Step S102, inputting the real-time monitoring data of each monitoring point into a pre-trained health association analysis model to obtain the health status result of each monitoring point;

[0081] Among them, the health association analysis model compares real-time monitoring data with normal behavior patterns and uses the learned correlations between monitoring points to identify deviations in data patterns, thereby detecting abnormal monitoring points, reducing reliance on manual judgment, and improving the efficiency and accuracy of identifying abnormal conditions.

[0082] Step S103, in response to the health status result of the monitoring point being an abnormal state, determining the monitoring point as an abnormal monitoring point;

[0083] Among them, when the real-time data of the monitoring point deviates significantly from the normal behavior pattern of the health association analysis model, the system marks this monitoring point as an abnormal monitoring point;

[0084] It can be understood that by quickly confining potential structural problems to specific abnormal monitoring points, it provides a clear location for subsequent detailed inspection and maintenance work, avoiding the time and cost of a comprehensive inspection.

[0085] Step S104, based on the pre-trained tunnel physical model and propagation mode prediction model, predict the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point;

[0086] In some specific embodiments, a tunnel physical model may be used to calculate possible propagation paths and ranges of abnormal conditions in the tunnel structure, and a propagation mode prediction model may be used to predict the propagation mode of the abnormal condition, and the two may be combined to obtain a more comprehensive impact domain assessment.

[0087] Among them, the potential impact domain prediction results of abnormal monitoring points provide an assessment of the potential spread and intensity of the abnormal impact, which helps to evaluate the stability of the entire tunnel structure and guide maintenance personnel to focus on the areas most likely to be affected.

[0088] Step S105, obtaining risk monitoring points and corresponding risk levels according to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring points;

[0089] Among them, by analyzing the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring points, the system can identify other risk monitoring points that may be affected in addition to the abnormal monitoring points, and assign corresponding risk levels to them.

[0090] Step S106, generating a risk warning signal according to the risk monitoring point and the corresponding risk level.

[0091] Among them, based on the predicted risk monitoring points and risk levels, the system can automatically generate a trigger risk warning mechanism to prevent and mitigate possible damage in advance and improve the safety and reliability of tunnel operations.

[0092] In the above implementation, based on the advantages of distributed fiber optic sensing technology, the status data of each monitoring point is captured in real time, the correlation between the monitoring points is identified and the abnormal monitoring points are detected through the pre-trained health association analysis model, and the propagation path prediction results output by the tunnel physical model and the propagation mode prediction results output by the propagation mode prediction model are combined to determine the potential impact domain and abnormal propagation intensity of the abnormal monitoring point, identify other risk monitoring points and risk levels that may be affected, so as to accurately judge the health status association between adjacent monitoring points, and conduct effective risk assessment and early warning based on this, which greatly improves the safety management level and maintenance efficiency of the tunnel.

[0093] Reference Figure 2 As a further implementation of the health monitoring method, the health monitoring method further includes a training step of a health association analysis model, and the training step includes:

[0094] Step S201, obtaining historical monitoring data of each monitoring point in the tunnel;

[0095] Among them, the historical monitoring data of the tunnel monitoring points can be collected from the monitoring database, including parameters such as strain, displacement, temperature, and humidity.

[0096] Step S202, performing data preprocessing on historical monitoring data;

[0097] Among them, the collected data will be cleaned and normalized to remove outliers and noise to ensure data quality.

[0098] Step S203, extracting features from the historical monitoring data after data preprocessing to obtain a feature vector of the correlation between monitoring points;

[0099] Among them, features that are helpful in identifying the correlation between monitoring points, such as trends and periodicity in time series analysis, are extracted from the preprocessed data, while those features that are not statistically significant for establishing the association between monitoring points are eliminated.

[0100] Step S204, determining the association rules between the monitoring points based on the association rule learning algorithm and the feature vectors of the correlation between the monitoring points;

[0101] In some possible embodiments, an association rule learning algorithm, such as Apriori or FP-growth, may be used to discover frequent patterns and strong rules between monitoring points, so as to learn the association rules between monitoring points.

[0102] Step S205, constructing a monitoring point graph network of the tunnel, and obtaining a graph network analysis result based on graph convolutional network analysis;

[0103] Specifically, the nodes in the graph convolutional network represent monitoring points, and the edges between the nodes represent the strength of the association between the monitoring points. The graph convolutional network (GCN) analysis algorithm can be used to further analyze the health status association between the monitoring points.

[0104] Reference Figure 3 , the graph convolution network (GCN) of the embodiment of the present application includes an input layer, a graph network construction, a GCN layer, a fully connected layer and an output layer. Among them, the input layer mainly includes monitoring point data (feature data of each monitoring point, such as stress, strain, displacement, etc.) and monitoring point association (the strength of association between monitoring points, which defines the edge in the graph network); the graph network construction is to construct a monitoring point graph network, taking the monitoring points as nodes and the strength of association between monitoring points as edges to construct the graph network; the GCN layer includes graph convolution layer 1 (extracting local features from the graph network through graph convolution operation), activation function (applying nonlinear activation function (such as ReLU) to increase the nonlinear expression ability of the model), graph convolution layer 2 (further extracting advanced features from the graph network), activation function (applying activation function again); the fully connected layer is mainly used to perform full connection operations on the features extracted by the convolution layer, integrate feature information, and apply nonlinear activation functions to enhance the expression ability of the model; the output layer is used to output the health status association between monitoring points, which is used to determine whether the monitoring point is abnormal and its degree of association with other monitoring points.

[0105] Step S206: Based on the association rules between monitoring points and the graph network analysis results, a health association analysis model is trained.

[0106] In the above implementation, the health association analysis model can identify and understand the interdependencies between different monitoring points and their normal behavior patterns by gaining insight into the complex relationships between the monitoring points. When the real-time monitoring data exhibits behavior inconsistent with these normal patterns, the model can identify the monitoring point as abnormal.

[0107] It is understood that the role of the health correlation analysis model in the tunnel monitoring system is not limited to directly identifying abnormal monitoring points, but also to understanding the health status correlation between monitoring points. This correlation is crucial to identify potential abnormal monitoring points. For example, if two monitoring points usually show a certain synchronization, and suddenly this synchronization disappears, even though the individual readings of the two monitoring points may still be within the normal range, this change may indicate a potential problem in a part of the tunnel that is not directly monitored.

[0108] Reference Figure 4 As an implementation of step S104, based on the pre-trained tunnel physical model and propagation mode prediction model, the step of predicting the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point includes:

[0109] Step S301, inputting the real-time monitoring data of the abnormal monitoring point into the pre-trained tunnel physical model, and predicting the stress propagation path and strain influence area of ​​the abnormal monitoring point;

[0110] Among them, the tunnel physical model is used to simulate the behavior of abnormal monitoring points in the physical environment, especially the stress propagation path and strain impact area, so as to accurately identify the scope of the affected area;

[0111] In some embodiments, the tunnel physical model can predict the stress transmission path in the tunnel structure when anomalies (such as cracks, deformations, etc.) occur in local areas through numerical simulation methods such as finite element analysis based on factors such as tunnel design parameters, material properties and geological conditions.

[0112] Step S302, inputting the stress propagation path and strain influence area of ​​the abnormal monitoring point into a pre-trained propagation mode prediction model to predict the abnormal propagation mode of the abnormal monitoring point;

[0113] Specifically, the data obtained from the physical model is input into a machine learning model that predicts abnormal propagation patterns. Based on historical data, the model learns different types of abnormal events and their propagation behaviors in the tunnel structure. It can predict the paths along which the abnormal situation may spread, as well as the speed of spread and the scope and intensity of potential impacts.

[0114] Understandably, this step can provide in-depth insights into the propagation trend of abnormal conditions, helping to predict the areas that may be affected in the future and the intensity of the impact. By understanding the abnormal propagation pattern, more targeted monitoring and maintenance strategies can be developed, effectively reducing the potential risks of tunnel structures.

[0115] Step S303: fusing the stress propagation path, strain influence area and abnormal propagation mode of the abnormal monitoring point to obtain the potential influence domain and abnormal propagation intensity of the abnormal monitoring point.

[0116] It can be understood that by combining predictions based on physical principles and statistical predictions based on historical data, taking into account the interactions and possible overlaps between them, the resulting potential impact domain prediction results not only include the scope of the area that may be affected, but also provide dynamic information about the spread of the anomaly, such as the possible propagation speed and direction.

[0117] In the above implementation, the combination of physical simulation and machine learning prediction provides an accurate and comprehensive method to evaluate the potential impact domain and abnormal propagation intensity of abnormal monitoring points in the tunnel, thereby effectively guiding the maintenance team to take targeted measures to improve the overall health and safety of the tunnel.

[0118] Reference Figure 5 As a further implementation of the tunnel health monitoring method, the health monitoring method further includes a training step of a tunnel physical model, the steps comprising:

[0119] Step S401, obtaining structural data of the tunnel and performing data preprocessing; wherein the structural data includes tunnel design parameters, construction material properties and historical monitoring data of multiple monitoring points;

[0120] Among them, the data preprocessing steps include data cleaning, normalization, feature extraction, etc., which improve data quality by removing noise and outliers; and provide direct input for model training by extracting key features such as the correlation between monitoring points and the changing trend of time series.

[0121] Step S402, initializing parameters of the pre-built finite element model based on tunnel design parameters and construction material properties;

[0122] Among them, parameter initialization includes setting material properties, boundary conditions, etc. to ensure that the model can accurately reflect the physical behavior of the actual tunnel.

[0123] Reference Figure 6The figure shows a schematic diagram of the model structure of the finite element model of the embodiment of the present application. The parameters and data (tunnel design parameters, construction material characteristics, geological conditions and historical monitoring data) provided by the input layer are preprocessed and used to construct a finite element model, including meshing, boundary condition setting and material property allocation, etc.; then, after the model is corrected and solved, the stress distribution and strain state can be output, and the reliability and accuracy of the model can be ensured by verifying the prediction results.

[0124] It should be noted that finite element models are usually mainly used to simulate the stress distribution and strain state of tunnels under various loads. Although finite element analysis is traditionally regarded as a tool for simulation analysis of responses under given conditions, it can also indirectly realize prediction functions by combining historical monitoring data and parameter optimization technology. For example, by integrating the design parameters and construction material characteristics of the tunnel with historical monitoring data, the changing trend of the structure under different working conditions or over time can be predicted by training the finite element model. This is actually a prediction based on physical principles, and the future state can be estimated to a certain extent by adjusting parameters and setting boundary conditions.

[0125] Step S403, dividing the historical monitoring data of the multiple monitoring points in the structure data into a first data set and a second data set;

[0126] Step S404, calibrating model parameters of the finite element model based on the first data set to obtain a calibrated finite element model;

[0127] The model is calibrated using historical monitoring data and the model parameters are adjusted to best reflect the actual measurement results at the monitoring points.

[0128] Step S405, cross-validating the finite element model based on the second data set until the model performance meets the preset conditions or the number of model iterations reaches the preset number, thereby obtaining a trained tunnel physical model.

[0129] Among them, through cross-validation, a data set different from that in the calibration phase is used to evaluate the accuracy and generalization ability of the model, and the model parameters are further optimized and adjusted to ensure that the model can provide accurate predictions under a wider range of conditions.

[0130] In the above implementation manner, the constructed tunnel physical model can accurately simulate the behavior of the tunnel under various conditions, providing a reliable analysis tool for tunnel health monitoring.

[0131] In addition, during the application process, the real-time monitoring data of the tunnel can be integrated into the tunnel physical model, and the model can be updated and recalibrated regularly to reflect the latest status of the tunnel, so as to use the latest monitoring data to fine-tune the model parameters and ensure that the model continues to adapt to tunnel conditions that change over time.

[0132] Reference Figure 7 As a further implementation of the tunnel health monitoring method, the health monitoring method further includes a step of training a propagation mode prediction model, the steps comprising:

[0133] Step S501, obtaining monitoring data of historical abnormal events at multiple monitoring points in the tunnel;

[0134] The monitoring data of historical abnormal events include, but are not limited to, stress, strain, displacement, and corresponding timestamps.

[0135] Step S502, preprocessing the monitoring data of historical abnormal events and dividing them into a training set and a test set;

[0136] Among them, data preprocessing includes data cleaning (removing outliers and missing values), data normalization or standardization (ensuring that the input data are of the same magnitude), and feature extraction (identifying key features that are useful for predicting anomaly propagation patterns).

[0137] Step S503, training the pre-built machine learning model based on the training set, optimizing the model parameters, and obtaining a propagation mode prediction model;

[0138] Reference Figure 8 In one of the embodiments of the present application, the pre-built machine learning model may adopt a convolutional neural network (CNN). Since tunnel monitoring data (such as stress, strain and displacement) has significant spatiotemporal characteristics, the convolutional neural network (CNN) can excellently capture and process such spatiotemporal data with complex patterns.

[0139] Among them, the structure of the convolutional neural network model of the embodiment of the present application includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; the input layer is used to input the monitoring data (stress, strain, displacement and timestamp, etc.) of historical abnormal events at multiple monitoring points in the tunnel; multiple convolution layers are used to extract spatiotemporal features in the monitoring data, and each convolution layer applies several convolution kernels to extract features through sliding window operations; each convolution layer is followed by a pooling layer for dimensionality reduction and feature selection, reducing the amount and complexity of data while retaining important features, and usually maximum pooling (Max Pooling) or average pooling (Average Pooling) can be used; the features extracted by the convolution layer and the pooling layer are input to the fully connected layer for further feature combination and pattern recognition; finally, the output layer gives the prediction results, including the abnormal propagation path, propagation speed and impact range.

[0140] It should be noted that the selection of the model needs to consider the complexity of the model, the prediction performance, and the characteristics of the training data to ensure that the selected model can effectively learn the pattern of anomaly propagation.

[0141] Step S504, testing the propagation mode prediction model and calibrating model parameters based on the test set until the loss function of the model meets a preset condition or the number of model iterations reaches a preset number, thereby obtaining a trained propagation mode prediction model.

[0142] Among them, an independent test data set is used to verify the trained model, evaluate its prediction accuracy and generalization ability, and the model is tuned through continuous iteration to improve the prediction performance.

[0143] In the above implementation, based on the machine learning algorithm, the propagation pattern of abnormal conditions of multiple monitoring points in the tunnel is learned, and then the trained and optimized model is deployed to the actual tunnel monitoring system, so as to facilitate the accurate prediction of the abnormal propagation pattern of the monitoring points in the tunnel.

[0144] Reference Fig. 9 As an implementation method of step S105, the step of obtaining the risk monitoring point and the corresponding risk level according to the potential impact domain and the abnormal propagation intensity corresponding to the abnormal monitoring point includes:

[0145] Step S601, based on the monitoring point map network of the tunnel, determining the monitoring points located in the potential impact area as risk monitoring points;

[0146] In some embodiments, the monitoring point map network of the tunnel is a network structure including all monitoring points and their interrelationships; when an abnormal event occurs, by analyzing the network, it is possible to quickly determine which monitoring points are located in the potential impact domain, and the monitoring points located in the impact domain are marked as risk monitoring points.

[0147] Step S602, determining the distance between the risk monitoring point and the abnormal monitoring point according to the position of the risk monitoring point in the monitoring point map network;

[0148] Among them, using the information in the monitoring point graph network, the distance between the risk monitoring point and the abnormal source (i.e., the abnormal monitoring point) can be calculated; this distance can be a physical distance or a network topological distance, that is, measured by the number of hops in the network path.

[0149] It is understandable that distance is one of the key factors in risk assessment. Generally speaking, the greater the distance, the smaller the impact of risk transmission. Incorporating distance factors into risk assessment improves the accuracy of the assessment and can more accurately predict the actual impact of risk transmission.

[0150] Step S603, performing weighted calculation based on preset weights according to the size of the potential impact domain, the abnormal propagation intensity and the distance, to obtain the risk impact index of the risk monitoring point;

[0151] Among them, since each factor (the size of the potential impact domain, abnormal propagation intensity, and distance) contributes differently to the risk impact, it is necessary to assign a weight to each factor, and perform weighted calculations on each factor according to different weights to obtain a comprehensive risk impact index for the risk monitoring point, so as to quantify the risk level of each monitoring point, making risk assessment objective and easy to compare, and providing a quantitative basis for formulating risk management strategies.

[0152] Step S604: Based on a preset risk matrix, determine the corresponding risk level according to the risk impact index of the risk monitoring point.

[0153] Among them, the risk matrix can be formed by mapping the risk impact index to the specific index range and the corresponding risk level.

[0154] In some possible embodiments, the risk matrix may be set based on historical data, expert experience, or industry standards, with the goal of converting the risk impact index into an actual and actionable risk level.

[0155] In the above implementation, the monitoring points that may be affected are quickly screened out, so as to concentrate resources and attention on those monitoring points that are most likely to face risks; at the same time, the obtained risk level provides decision makers with a clear risk classification, which helps decision makers to take corresponding prevention, preparation and response measures according to the risk level, thereby effectively managing and mitigating risks.

[0156] Reference Fig.10 As an implementation of step S106, after the step of generating a risk warning signal according to the risk monitoring point and the corresponding risk level, the step further includes:

[0157] Step S701, generating an emergency response plan corresponding to the risk level according to the risk level;

[0158] The emergency response plan defines the actions to be taken in the event of warnings at different risk levels, such as local evacuation, further inspection or suspension of construction.

[0159] Step S702: Based on the location information of the risk monitoring point, the risk warning signal and the emergency response plan are sent to the maintenance terminal corresponding to the risk monitoring point.

[0160] The maintenance terminal may be a smart mobile terminal of the maintenance team, such as a mobile phone, a tablet computer, etc., or a computer terminal in the monitoring room.

[0161] In the above implementation, timely risk warning and push of emergency response plans can ensure rapid response by maintenance personnel, allowing relevant personnel to take prompt measures to prevent further damage to the tunnel structure, thereby improving the safety of tunnel operations.

[0162] The embodiment of the present application also discloses a tunnel health monitoring system based on distributed optical fiber sensing.

[0163] A tunnel health monitoring system based on distributed optical fiber sensing, the health monitoring system comprising:

[0164] An acquisition module is used to obtain real-time monitoring data of each monitoring point in the tunnel;

[0165] A health status result generation module is used to input the real-time monitoring data of each monitoring point into a pre-trained health association analysis model to obtain the health status result of each monitoring point;

[0166] an abnormal monitoring point determination module, for determining a monitoring point as an abnormal monitoring point in response to a health status result of the monitoring point being an abnormal state;

[0167] Anomaly propagation prediction module, which is used to predict the corresponding potential impact domain and anomaly propagation intensity according to the anomaly monitoring point based on the pre-trained tunnel physical model and propagation mode prediction model;

[0168] The risk monitoring point determination module is used to obtain the risk monitoring point and the corresponding risk level according to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring point;

[0169] The risk warning module is used to generate risk warning signals of corresponding levels according to risk monitoring points and corresponding risk levels and send them to the maintenance terminal.

[0170] In the above implementation, by accurately determining the health status association between adjacent monitoring points and performing effective risk assessment and early warning based on the association, the safety management level and maintenance efficiency of the tunnel are greatly improved.

[0171] The tunnel health monitoring system based on distributed optical fiber sensing in the embodiment of the present application can implement any of the above-mentioned health monitoring methods, and the specific working process of each module in the health monitoring system can refer to the corresponding process in the above-mentioned method embodiment.

[0172] In the several embodiments provided in this application, it should be understood that the provided methods and systems can be implemented in other ways. For example, the system embodiments described above are only illustrative; for example, the division of a certain module is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0173] The embodiment of the present application also discloses a computer device.

[0174] The computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the tunnel health monitoring method based on distributed optical fiber sensing as described above is implemented.

[0175] The embodiment of the present application also discloses a computer-readable storage medium.

[0176] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute any of the above-mentioned tunnel health monitoring methods based on distributed optical fiber sensing.

[0177] Among them, computer-readable storage media can be any tangible medium that contains or stores a program, which can be used by or in conjunction with an instruction execution system, apparatus or device; the program code contained on the computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0178] It should be noted that in the above embodiments, the description of each embodiment has different emphases, and for parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0179] The above are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Any feature disclosed in this specification (including the abstract and drawings), unless otherwise stated, can be replaced by other equivalent or alternative features with similar purposes. That is, unless otherwise stated, each feature is only an example of a series of equivalent or similar features.

Claims

1. A tunnel health monitoring method based on distributed optical fiber sensing, characterized in that: The health monitoring method comprises: Obtain real-time monitoring data for each monitoring point in the tunnel; The real-time monitoring data of each monitoring point is input into the pre-trained health association analysis model to obtain the health status results of each monitoring point; In response to the health status result of the monitoring point being an abnormal state, determining the monitoring point as an abnormal monitoring point; Based on the pre-trained tunnel physical model and propagation mode prediction model, the corresponding potential impact domain and abnormal propagation intensity are predicted according to the abnormal monitoring point; According to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring point, a risk monitoring point and a corresponding risk level are obtained; Generate a risk warning signal according to the risk monitoring points and corresponding risk levels; The health monitoring method further includes a training step of the health association analysis model, wherein the training step includes: Obtain historical monitoring data for each monitoring point in the tunnel; Performing data preprocessing on the historical monitoring data; Perform feature extraction on the historical monitoring data after data preprocessing to obtain the feature vector of the correlation between monitoring points; Based on the association rule learning algorithm, determining the association rules between the monitoring points according to the feature vectors of the correlation between the monitoring points; Constructing a monitoring point graph network of the tunnel, and obtaining a graph network analysis result based on graph convolutional network analysis; Based on the association rules between the monitoring points and the graph network analysis results, a health association analysis model is trained; Based on the pre-trained tunnel physical model and propagation mode prediction model, the step of predicting the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point includes: Inputting the real-time monitoring data of the abnormal monitoring point into a pre-trained tunnel physical model to predict the stress propagation path and strain influence area of ​​the abnormal monitoring point; Inputting the stress propagation path and strain influence area of ​​the abnormal monitoring point into a pre-trained propagation mode prediction model to predict the abnormal propagation mode of the abnormal monitoring point; The stress propagation path, strain influence area and abnormal propagation mode of the abnormal monitoring point are data-fused to obtain the potential influence domain and abnormal propagation intensity of the abnormal monitoring point; The health monitoring method further comprises a step of training the tunnel physical model, the step comprising: Acquiring structural data of the tunnel and performing data preprocessing; wherein the structural data includes tunnel design parameters, construction material properties and historical monitoring data of multiple monitoring points; Initializing parameters of the pre-built finite element model based on the tunnel design parameters and construction material properties; Dividing the historical monitoring data of the plurality of monitoring points in the structural data into a first data set and a second data set; Performing model parameter correction on the finite element model based on the first data set to obtain a calibrated finite element model; Cross-validating the finite element model based on the second data set until the model performance meets a preset condition or the number of model iterations reaches a preset number, thereby obtaining the trained tunnel physical model; The health monitoring method further comprises a step of training the propagation mode prediction model, the step comprising: Obtain monitoring data of historical abnormal events at multiple monitoring points in the tunnel; Performing data preprocessing on the monitoring data of the historical abnormal events and dividing them into a training set and a test set; Training a pre-built machine learning model based on the training set, optimizing model parameters, and obtaining the propagation mode prediction model; The propagation mode prediction model is tested and model parameters are corrected based on the test set until the loss function of the model meets a preset condition or the number of model iterations reaches a preset number, thereby obtaining the trained propagation mode prediction model.

2. A tunnel health monitoring method based on distributed optical fiber sensing according to claim 1, characterized in that: The step of obtaining the risk monitoring point and the corresponding risk level according to the potential impact domain and the abnormal propagation intensity corresponding to the abnormal monitoring point includes: Based on the monitoring point map network of the tunnel, the monitoring points located in the potential impact area are determined as risk monitoring points; Determine the distance between the risk monitoring point and the abnormal monitoring point according to the position of the risk monitoring point in the monitoring point map network; According to the size of the potential impact domain, the abnormal propagation intensity and the distance, a weighted calculation is performed based on preset weights to obtain the risk impact index of the risk monitoring point; Based on the preset risk matrix, the corresponding risk level is determined according to the risk impact index of the risk monitoring point.

3. A tunnel health monitoring method based on distributed optical fiber sensing according to claim 2, characterized in that: According to the risk monitoring points and the corresponding risk levels, the step of generating a risk warning signal further includes: According to the risk level, generate an emergency response plan corresponding to the risk level; Based on the location information of the risk monitoring point, the risk warning signal and the emergency response plan are sent to a maintenance terminal corresponding to the risk monitoring point.

4. A tunnel health monitoring system based on distributed optical fiber sensing, used to execute a tunnel health monitoring method based on distributed optical fiber sensing according to any one of claims 1 to 3, characterized in that: The health monitoring system comprises: An acquisition module is used to obtain real-time monitoring data of each monitoring point in the tunnel; A health status result generation module is used to input the real-time monitoring data of each monitoring point into a pre-trained health association analysis model to obtain the health status result of each monitoring point; an abnormal monitoring point determination module, configured to determine that the monitoring point is an abnormal monitoring point in response to the health status result of the monitoring point being an abnormal state; An abnormal propagation prediction module, used to predict the corresponding potential impact domain and abnormal propagation intensity according to the abnormal monitoring point based on the pre-trained tunnel physical model and propagation mode prediction model; A risk monitoring point determination module is used to obtain the risk monitoring point and the corresponding risk level according to the potential impact domain and abnormal propagation intensity corresponding to the abnormal monitoring point; The risk warning module is used to generate a risk warning signal of a corresponding level according to the risk monitoring point and the corresponding risk level and send it to the maintenance terminal.

5. A computer device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 3 when executing the program.

6. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and execute the method according to any one of claims 1 to 3.