Tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, early warning analysis method and implementation method

Abstract

The invention provides a tunnel full-section comprehensive monitoring system based on a distributed optical fiber sensing technology, an early warning analysis method and an implementation method, and relates to the field of tunnel engineering.The tunnel full-section comprehensive monitoring system comprises a terminal processing system and a cloud AI processing system, and the terminal processing system and the cloud AI processing system are in signal connection to jointly construct a two-stage interaction system architecture; the terminal processing system is used for quantifying risk factors of the whole section of the tunnel into numeric feature vectors and calculating state differences of the same tunnel monitoring section at different time points according to the numeric feature vectors through a specific algorithm so as to obtain target deformation data; and determining the current early warning state of the tunnel section according to the execution degree # imgabs0 #. The system has the following beneficial effects: firstly, the distributed optical fiber sensing technology is utilized, and the multi-section data in the tunnel can be monitored in real time; and secondly, a feedback mechanism is introduced, and a set of intelligent verification system for continuous learning and optimization is constructed through mutual fusion of professional experience of constructors and an advanced deep learning technology.

Description

Technical Field

[0001] The present invention relates to the field of tunnel engineering, and in particular to a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, and an early warning analysis method and an implementation method. Background Art

[0002] As highway construction gradually develops towards the west, mountain tunnel projects account for an increasing proportion. Compared with other projects, tunnel projects have their own special properties, such as concealment, complexity and uncertainty. As most mountain tunnels are buried deep, long distances, and the geological conditions along the line are complex, the safety issues of tunnel surrounding rock structures and support structures have become increasingly prominent during tunnel construction. Therefore, it is particularly important to strengthen daily monitoring of tunnels in all aspects during construction. Through all-round monitoring, various deformation and damage phenomena in the tunnel excavation section during daily construction can be discovered in time, and accurate and effective reinforcement treatment can be carried out on unstable positions in a timely manner to avoid further serious dangers in the future, resulting in huge costs for repairs or even shutdowns, which will affect the safety and progress of the project.

[0003] At present, tunnel monitoring is mainly based on traditional point monitoring, which is achieved by installing monitoring equipment at a certain point in the tunnel and then obtaining equipment data manually on a daily basis. However, the disadvantage of point monitoring is that it can only measure data at a certain point in space, with fewer data points, and cannot cover the entire tunnel monitoring section. The monitoring coverage is small and the accuracy is insufficient. It is very likely that there will be no abnormal data measured at a certain point, but it is actually seriously deformed. Due to the long distance between data points, monitoring equipment needs to be connected to a long cable. The wiring inside the tunnel is complicated and the construction process is easily cut by large construction equipment, resulting in failure. In contrast, fiber grating sensors can be buried or adhered to the monitored object, and can measure data at all points along the deployed sensor line, making up for the shortcomings of point monitoring.

[0004] Although this technology is now very mature, it has been applied in tunnel projects for a short time. How to deploy and protect the sensors to maintain stable operation during their service life and how to post-process the data are still difficult problems that need to be solved. The existing tunnel monitoring systems and methods based on optical fiber sensors can perform more intensive linear monitoring, but cannot monitor the monitored data in real time. When the changes in the monitored data exceed the safety threshold, an alarm cannot be issued immediately. Summary of the invention

[0005] In view of the shortcomings of the above-mentioned prior art, the technical problem to be solved by the present invention is to provide a tunnel full-section comprehensive monitoring system and an early warning analysis method and an implementation method based on distributed fiber optic sensing technology, which can realize comprehensive and real-time monitoring during tunnel construction and operation, and significantly improve the monitoring efficiency and safety.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows: the present invention provides a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, including a terminal processing system and a cloud AI processing system, wherein the terminal processing system and the cloud AI processing system are mutually signal-connected to jointly construct a two-level interactive system architecture; The terminal processing system is used to quantify the risk factors of the entire tunnel section into a numerical feature vector, and calculate the state difference of the same tunnel monitoring section at different time points according to the numerical feature vector through a specific algorithm to obtain the target deformation data, and determine the execution degree according to the target deformation data. ; According to the execution Determine the current warning status of the tunnel section.

[0007] In a preferred solution, the terminal processing system includes a data acquisition module, a network connection module, a data analysis module and a self-checking module; the data acquisition module, the network connection module, the data analysis module and the self-checking module are electrically connected to each other; Data acquisition module, including a fiber optic sensor array and an intelligent numbering junction box, is used to collect electrical signal data from multiple distributed fiber optic sensors at each monitoring section to monitor temperature, strain and displacement parameters in the tunnel; The network connection module is used to connect the system to the external network and transmit the collected electrical signal data to the data analysis module; The data analysis module is used to process and analyze the received electrical signal data, obtain the monitoring results of various monitoring items inside the tunnel, and extract numerical feature vectors; The system self-check module is used to check whether the entire monitoring system network is in a normal connection state and whether each module is working normally. It determines the system status by sending and receiving specific detection electrical signals and comparing them with the preset normal electrical signal range; The cloud AI processing system includes a data warning module and a visualization module; the data warning module and the visualization module are electrically connected to each other; The data early warning module is used to predict the subsequent trend of the data obtained from monitoring, and combine the data with tunnel safety to make safety early warnings; The visualization module is used to dynamically and intuitively display the changes in monitoring data of each part of each monitoring section.

[0008] In a preferred embodiment, the present invention further provides an early warning analysis method for a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, which is characterized by comprising the following steps: S1. Data acquisition: The electrical signal data of multiple distributed optical fiber sensors at each monitoring section are collected through the optical fiber sensor sequence in the data acquisition module of the terminal processing system to obtain the electrical parameters related to temperature, strain and displacement in the tunnel; the distributed optical fiber sensor sequence converts physical quantities into measurable electrical signals using a specific electrical sensing principle, and connects each sensor to the corresponding intelligent numbered junction box interface, and connects to the main cable interface along the tunnel, transmits the data to the optical fiber demodulator through the main cable, and performs signal source grouping and secondary grouping through the network connection module and the data analysis module; S2, data transmission: using the network connection module of the terminal processing system to transmit the collected data to the local indoor server; the network connection module adopts a redundant network connection design, including wired network connection and wireless network connection, and connects to the local network through the on-site PC and router; S3. Data analysis: In the data analysis module of the terminal processing system, the received electrical sensor data is processed and analyzed using the supporting signal processing software to obtain the monitoring results of various monitoring items inside the tunnel, and the numerical feature vector is extracted from the received monitoring results to obtain the target deformation data feature vector at a certain moment. ; The target deformation data is data of at least one deformation feature vector of a displacement feature, a deformation rate feature, a cumulative displacement feature, and a relative displacement feature, and the displacement feature, the deformation rate feature, the cumulative displacement feature, and the relative displacement feature are all calculated based on the conversion of electrical signal data collected by the distributed optical fiber sensor; S4, system self-check: by using the abnormality check port of the supporting signal processing software to check the sensor connection status and data capture measurement status, when the self-check result is normal, the monitoring data is transmitted to the local data storage; if there is an abnormality, it returns to the supporting signal processing software to report the specific abnormal point; S5. Early warning judgment: In the data early warning module of the cloud AI processing system, the monitoring data in the local data storage is obtained through the network server, and the corresponding monitoring data is transmitted to the big data platform. The monitoring data along the distributed optical fiber sensor is learned and predicted based on the deep learning model, and the predicted data is compared with the execution degree. Combined to determine different safety warning levels, the data is stored in the cloud database. When the monitoring value approaches or exceeds the safety threshold, the warning channel is activated and different forms of warning notifications are sent to relevant personnel, including SMS, email, sound and light alarms; Among them, a support vector machine, a random forest or a neural network deep learning model is selected, and the deep learning model is obtained by training historical monitoring data and corresponding tunnel status data; S6. Visualization display: The visualization module of the cloud AI processing system is used to dynamically and intuitively display the changes in the monitored electrical data of each part of each monitoring section, and the stored data is transmitted to the three-dimensional visualization software through the cloud database to dynamically display the status information of the tunnel in the form of visual charts or three-dimensional models.

[0009] In the preferred solution, in step S3, to construct the target deformation data feature vector The weight of the target deformation data feature vector The features in are dynamically weighted to evaluate the relative importance of the tunnel deformation state, and the tunnel operation time is assumed to be , Tunnel operation time The time variable is used to adjust the weight of the displacement feature according to the different operation stages of the tunnel. The displacement feature weight is The calculation formula is: ; in, , It is an adjustment coefficient used to control the rate at which the weight changes over time. Different values ​​will cause the weight to grow or decrease at different rates at different time stages. , As the time dividing point, It divides the initial and middle stages of tunnel operation. It divides the middle and late stages; The strain energy density is calculated based on the strain signal data collected by the distributed optical fiber sensor in the data acquisition module. , and in the target deformation data feature vector Strain energy density Characteristic index, strain energy density The formula used is as follows: ; in, is the strain data after the strain signal is converted, is the elastic modulus of the material, and the strain energy density As a new feature Add to target deformation data feature vector In the above example, the target deformation data feature vector is , strain energy density An indicator used to evaluate the energy accumulation inside the tunnel structure. When the strain energy density exceeds a certain limit, it means that the tunnel section structure is in a high stress state and there is a greater risk of deformation or damage.

[0010] In the preferred solution, in step S5, the hierarchical deformation data feature vector is obtained. , and calculate the target deformation data feature vector With the classified deformation data feature vector The Euclidean distance between , normalize different eigenvalues, map eigenvalues ​​of different dimensions and magnitudes to the interval [0,1], and set the target deformation data eigenvector and the classified deformation data feature vector The Features and ; The normalized eigenvalue formula is as follows: ; ; The target deformation data feature vector Normalize the minimum and maximum values ​​in the historical data to obtain the normalized eigenvalues , the formula is as follows: ; in, and are the target deformation data feature vectors With the classified deformation data feature vector Middle The eigenvalues ​​after feature normalization, is the feature vector Middle The eigenvalues ​​after feature normalization, , Characteristics The minimum and maximum values ​​in the historical data, , Characteristics The minimum and maximum values ​​in the historical data, the normalized eigenvalues Used to analyze the strain energy density in different time periods, different tunnel parts or different working conditions Data mining to ensure accurate and representative minimum and maximum values; Substitution , and , then the normalized target deformation data feature vector and the normalized classified deformation data feature vector The Euclidean distance between The formula is: ; Among them, the Euclidean distance By measuring the difference between the current tunnel state and the predefined graded deformation state, that is, the normalized target deformation data feature vector and the normalized classified deformation data feature vector The spatial distance between the tunnels can be used to assess the safety of the tunnel and to classify the warning levels accordingly; If the Euclidean distance The value of is small, then the target deformation data feature vector at the current moment With the classified deformation data feature vector The smaller the difference between them, the higher the similarity, that is, the higher the execution high; If the Euclidean distance The larger the value of is, the greater the difference between the target deformation data feature vector at the current moment and the graded deformation data feature vector is, and the lower the similarity is, the lower the execution degree is. Low.

[0011] In the preferred solution, in step S5, according to the normalized feature vector With the eigenvector The Euclidean distance between To determine the degree of execution ; Set execution degree The formula is as follows: ; in, is the maximum value of the Euclidean distance, is represented as a normalized distance metric with a threshold between [0,1]; if At this time, the execution degree , that is, the actual state of the current tunnel fully matches the graded security state, and the tunnel is in the safest state. On the contrary, if The value increases, As the value of increases, the execution The value of decreases, which means that the difference between the current actual state of the tunnel and the classified safety state is getting bigger and bigger, and the degree to which the tunnel deviates from the safety state is increasing; Then, the execution degree The warning levels are as follows: ; Among them, the execution Euclidean distance The closer the value is to 1, the closer it is to the graded safety status. The scope is divided into warning levels. It is the warning level threshold set according to tunnel safety.

[0012] In the preferred solution, in step S5, when there are risk factors around the entire tunnel section, different influencing factors have different degrees of influence on the tunnel displacement characteristics, and the displacement characteristic threshold adjustment coefficient for: ; in, For the The influence index of each influencing factor, the value range , is the number of influencing factors, through all Sum the influencing indicators of the influencing factors and divide them by the number of influencing factors , and obtain the displacement characteristic threshold adjustment coefficient reflecting the overall impact , displacement feature threshold adjustment coefficient It is used to measure the demand for adjustment of displacement feature threshold due to changes in the surrounding environment; The adjusted threshold The calculation formula is: ; in, is the original threshold, the original threshold It is a standard value used to determine that the tunnel displacement is within a safe range. When there are risk factors in the surrounding area, the displacement characteristic threshold adjustment coefficient Original threshold Adjust to get, if ,but equal , which means that the overall impact of surrounding factors is large, and the safety judgment standard remains unchanged; if ,but Less than , which means that the early warning mechanism is sensitive to the surrounding factors and issues an early warning; if , then the strictness of safety judgment is improved; Set execution degree The sequence is: ; in, for time point; Calculating Moving Averages The formula is as follows: ; Calculate moving standard deviation The formula is as follows: ; in, Indicates at a point in time The moving average of execution degree, is the moving window size, moving standard deviation It is used to measure the discreteness of the execution data in the moving window to reflect the execution value around its moving average. The size of fluctuations; like and , then adjust the warning level; in, is the trend threshold, is the standard deviation threshold, Indicates the current time point Compared with the previous time point The difference between the execution moving averages, if the moving average An upward trend and moving standard deviation If the moving average is small, the tunnel status is deteriorating and the change is stable. On the contrary, if the moving average is stable And the moving standard deviation If it is larger, it is determined that there is data fluctuation interference.

[0013] In the preferred solution, in step S5, a hybrid algorithm of the back propagation algorithm and the loss function is used, and the deformation degree of the tunnel section predicted by the deep learning model is set to , and its calculation formula is: ; in, represents the prediction function of the deep learning model, is the parameter set of the model, For the moment The eigenvector of The eigenvector of Input to a The predicted value of the deformation degree of the tunnel section at that moment is obtained in the deep learning model ; Calculate deformation trend ,like The absolute value of is greater than the set trend threshold , then the deformation trend is determined to be abnormal; When abnormal deformation is determined Existence, that is ,but ,otherwise ; in, is the abnormal deformation threshold; Warning signs The determination formula is as follows: ; in, is the safety threshold, if the warning signal , then the execution issues a warning, otherwise , confirm that the system is in normal condition and no alarm is required.

[0014] In the preferred solution, in step S5, the deep learning model is regularly updated and trained using new data. Time collects new monitoring data from the last update to the current time and its corresponding real tunnel section status information , build a new training dataset , using the training dataset Retrain the deep learning model and use the mean square error loss function as the optimization objective function; During the training process, the difference between the predicted results and the actual results is calculated, and the model parameters are continuously adjusted through the back propagation algorithm. , using the mean square error loss function The minimization formula is as follows: ; in, It is represented by the number of samples of the collected new monitoring data and its corresponding real tunnel section status information, Expressed as at time The feature vector formed by the collected new monitoring data after preprocessing, is represented as a set of parameters representing a deep learning model, Represented as a deep learning model based on the input feature vector and current parameters The prediction results made, Represented as the input feature vector The corresponding real tunnel section status information; In the data early warning module, when the relevant monitoring data Perform analysis and verification to determine the false positives, and set the true status determined by analysis and verification to ; If the false report is , the real anomaly is , then the formula for determining the false alarm is as follows: ; If the warning signal And the real state ,but , it is determined to be a false alarm. If the warning signal or And the real state ,but , it is determined to be a real abnormal situation or a normal situation without warning.

[0015] In a preferred embodiment, the present invention provides another implementation method of a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, characterized in that it includes the following steps: S1. Inside the surrounding rock: after excavation, drill a hole, place and break the bagged anchor, insert the rubber sleeve with the optical fiber sensor into the anchor, and inject quick-setting mud to fix it; S2, between surrounding rock and primary support: attach the optical fiber sensor to the primary support steel mesh, and lay it from the lower left corner along the outer side to the lower right corner; S3, between primary support and secondary lining: After the primary support is completed, the optical fiber sensor is attached to the inner side of the primary support, also from the lower left corner to the lower right corner; S4. Cable management: All sensor cables converge at the lower right corner, are pulled to one side of the tunnel through the channel reserved at the lower left corner, and are connected to the data collector.

[0016] The present invention provides a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, an early warning analysis method and an implementation method. Through the cooperation between the above structures, the following beneficial effects are achieved: First, distributed fiber optic sensing technology can be used to monitor multiple cross-section data in the tunnel in real time. Multiple sensor data in the cross-section are intelligently numbered and uniformly transmitted and demodulated through the main cable, which reduces the difficulty of laying monitoring equipment for hidden projects such as tunnels. The platform automatically monitors data changes and conducts intelligent prediction and early warning based on the big data platform. Second, a feedback mechanism was introduced. By integrating the professional experience of construction workers with advanced deep learning technology, a set of intelligent verification system with continuous learning and optimization was constructed. This feedback mechanism makes the system not only a static auxiliary tool, but also a dynamically developing intelligent platform, which can continuously improve its judgment ability with the verification feedback of construction workers during the monitoring construction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a main structural diagram of the present invention; Figure 2 It is a schematic diagram of the structure of the sensor laying of the present invention; Figure 3 The present invention Figure 2 The overall appearance structure diagram in ; Figure 4 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0018] In order to better understand the purpose, structure and function of the present invention, the embodiments and features in the embodiments of the present application can be combined with each other without conflict. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0019] Example 1 like Figure 4 As shown, the present invention provides a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, including a terminal processing system and a cloud AI processing system, wherein the terminal processing system and the cloud AI processing system are mutually signal-connected to jointly construct a two-level interactive system architecture; The terminal processing system is used to quantify the risk factors of the entire tunnel section into a numerical feature vector, and calculate the state difference of the same tunnel monitoring section at different time points according to the numerical feature vector through a specific algorithm to obtain the target deformation data, and determine the execution degree according to the target deformation data. ; According to the execution Determine the current warning status of the tunnel section.

[0020] In a preferred solution, the terminal processing system includes a data acquisition module, a network connection module, a data analysis module and a self-checking module; the data acquisition module, the network connection module, the data analysis module and the self-checking module are electrically connected to each other; Data acquisition module, including a fiber optic sensor array and an intelligent numbering junction box, is used to collect electrical signal data from multiple distributed fiber optic sensors at each monitoring section to monitor temperature, strain and displacement parameters in the tunnel; Among them, each monitoring section in the tunnel includes a sequence of fiber optic sensors for multiple different monitoring projects. Furthermore, such as displacement sensors, stress sensors or temperature sensors, each sensor is connected to the corresponding intelligent numbered junction box interface, and connected to the main cable interface along the tunnel. The main cable then passes through the tunnel exit to connect to the fiber optic demodulator, and connects to the PC end to group the signal sources of the sections according to the different junction box numbers through the supporting software. After the secondary grouping of the monitoring projects, the software is used to identify and demodulate each layer.

[0021] The network connection module is used to connect the system to the external network and transmit the collected electrical signal data to the data analysis module; Among them, the network connection module is connected to the local network through the on-site PC and router.

[0022] The data analysis module is used to process and analyze the received electrical signal data, obtain the monitoring results of various monitoring items inside the tunnel, and extract numerical feature vectors; Among them, after the data analysis module is successfully connected to the network, the data processed by the fiber optic demodulator will be transmitted to the local indoor server through the local network, and then processed by the supporting signal processing software to obtain the results.

[0023] The system self-check module is used to check whether the entire monitoring system network is in a normal connection state and whether each module is working normally. It determines the system status by sending and receiving specific detection electrical signals and comparing them with the preset normal electrical signal range; Among them, the system self-check module uses the software's built-in abnormality check port to check whether the sensor is connected normally and whether the data can be captured and measured normally. If normal, the data will be stored locally, and if there is an abnormality, the abnormal point will be reported.

[0024] The cloud AI processing system includes a data warning module and a visualization module; the data warning module and the visualization module are electrically connected to each other; The data warning module is used to receive the monitoring data transmitted by the terminal processing system, and predict the subsequent trend of the monitoring data based on the deep learning model. It also combines the predicted data with the tunnel safety to make different safety warning levels, and sends different forms of warning notifications according to different safety warning levels. The deep learning model is obtained by training the historical monitoring data and the corresponding tunnel status data; Among them, the data warning module transmits the corresponding test data to the big data platform, connects to the machine learning algorithm to learn and predict a large amount of data along the sensor line, sets security classification for data changes through specifications, and stores the data in the cloud database. When the monitoring value is approaching the safety threshold, the warning channel is activated and an alarm text message is sent to the relevant personnel with the reserved phone number for safety warning notification.

[0025] The visualization module is used to dynamically and intuitively display the changes in monitoring data of each part of each monitoring section.

[0026] The visualization module transmits the stored data to the 3D visualization software through the cloud database, and displays the monitoring data dynamically. Through mobile devices, such as mobile phones and tablets; and PCs, such as notebooks and desktop computers, the data can be logged in and viewed anytime and anywhere through the APP installed on the mobile device and the software installed on the PC. Specifically, the visualization module can choose to view detailed information on electrical monitoring data for a specific monitoring section or a specific time period, as well as view tunnel deformation simulation animations based on electrical data; the visualization module can also generate customized monitoring reports based on user needs. The report content includes statistical analysis results of electrical data, tunnel status assessment results, and early warning information.

[0027] Example 2 Further illustrate with reference to Example 1, Figure 1-3 The structure shown in the figure, the early warning analysis method of the tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology includes the following steps: S1. Data acquisition: The electrical signal data of multiple distributed optical fiber sensors in each monitoring section are collected through the optical fiber sensor sequence in the data acquisition module of the terminal processing system to obtain the electrical parameters related to temperature, strain and displacement in the tunnel; the distributed optical fiber sensor sequence uses a specific electrical sensing principle to convert physical quantities into measurable electrical signals, and in the acquisition process, a specific filter circuit is used to filter the collected original signal to eliminate noise signals caused by external electromagnetic interference and other factors, so as to improve the signal-to-noise ratio of the signal; the filter parameters of the filter circuit can be adaptively adjusted according to the electromagnetic characteristics of the tunnel environment and the electrical characteristics of the sensor, each sensor is connected to the corresponding intelligent numbered junction box interface, connected to the main cable interface along the tunnel direction, and the data is transmitted to the optical fiber demodulator through the main cable, and the signal source is grouped and re-grouped through the network connection module and the data analysis module; S2. Data transmission: The network connection module of the terminal processing system is used to transmit the collected data to the local indoor server; the network connection module adopts a redundant network connection design, including wired network connection and wireless network connection, and connects to the local network through the on-site PC and router. When one of the network connections fails, it automatically switches to another network connection to ensure the stability and continuity of data transmission; the network connection module also has a data caching function. When the network is temporarily interrupted, the collected electrical data is cached in the local storage unit, and the cached data is automatically uploaded after the network is restored; S3. Data analysis: In the data analysis module of the terminal processing system, the received electrical sensor data is processed and analyzed using the supporting signal processing software to obtain the monitoring results of various monitoring items inside the tunnel, and the numerical feature vector is extracted from the received monitoring results to obtain the target deformation data feature vector at a certain moment. ; The target deformation data is data of at least one deformation feature vector of a displacement feature, a deformation rate feature, a cumulative displacement feature, and a relative displacement feature, and the displacement feature, the deformation rate feature, the cumulative displacement feature, and the relative displacement feature are all calculated based on the conversion of electrical signal data collected by the distributed optical fiber sensor; S4, system self-check: by using the abnormality check port of the supporting signal processing software to check the sensor connection status and data capture measurement status, when the self-check result is normal, the monitoring data is transmitted to the local data storage; if there is an abnormality, it returns to the supporting signal processing software to report the specific abnormal point; S5. Early warning judgment: In the data early warning module of the cloud AI processing system, the monitoring data in the local data storage is obtained through the network server, and the corresponding monitoring data is transmitted to the big data platform. The monitoring data along the distributed optical fiber sensor is learned and predicted based on the deep learning model, and the predicted data is compared with the execution degree. Combined to determine different safety warning levels, the data is stored in the cloud database. When the monitoring value approaches or exceeds the safety threshold, the warning channel is activated and different forms of warning notifications are sent to relevant personnel, including SMS, email, sound and light alarms; Among them, a support vector machine, a random forest or a neural network deep learning model is selected, and the deep learning model is obtained by training historical monitoring data and corresponding tunnel status data; S6. Visual display: The visualization module of the cloud AI processing system is used to dynamically and intuitively display the changes in the monitored electrical data of each part of each monitoring section, and the stored data is transmitted to the three-dimensional visualization software through the cloud database to dynamically display the status information of the tunnel in the form of a visual chart or a three-dimensional model; the visualization module has an interactive function, and the user can remotely operate through the terminal device to choose to view the detailed monitoring data of a specific monitoring section and a specific time period, as well as view the tunnel deformation simulation animation based on the monitoring data; the visualization module can also generate customized monitoring reports according to user needs, and the content of the monitoring report includes the statistical analysis results of electrical data, the results of tunnel status assessment and early warning information.

[0028] In the preferred solution, in step S3, to construct the target deformation data feature vector The weight of the target deformation data feature vector The features in are dynamically weighted to evaluate the relative importance of the tunnel deformation state, and the tunnel operation time is assumed to be , Tunnel operation time The time variable is used to adjust the weight of the displacement feature according to the different operation stages of the tunnel. The displacement feature weight is The calculation formula is: ; in, It is expressed as the weight of the displacement feature when constructing the target deformation data feature vector. The size of the weight determines the relative importance of the displacement feature in evaluating the deformation state of the tunnel. It is expressed as the operation time of the tunnel in years. This time variable is used to adjust the weight of the displacement feature according to the different operation stages of the tunnel. , It is an adjustment coefficient used to control the rate at which the weight changes over time. Different values ​​will cause the weight to increase or decrease at different speeds at different time stages. The specific value can be set according to the actual situation and experience of the tunnel. , As the time dividing point, It divides the initial and middle stages of tunnel operation. Divides the boundary between the middle and late stages. When Calculate the weight, which is relatively low and increases slowly over time; When When calculating the weights, the weights continue to change and tend to be stable but maintain a high weight to better reflect the impact of long-term displacement accumulation on the tunnel state; And in the feature vector Strain energy density Characteristic index, strain energy density The calculation is performed based on the strain signal data collected by the distributed optical fiber sensor in the data acquisition module. The calculation formula is as follows: ; Among them, the strain energy density An indicator used to evaluate the energy accumulation inside the tunnel structure. When the strain energy density exceeds a certain limit, it means that the tunnel section structure is in a high stress state and there is a risk of greater deformation or damage. is the strain data after the strain signal is converted, Expressed as the calculated strain energy density The number of strain data points, strain data It is converted from the strain signal collected by the optical fiber sensor. is the elastic modulus of the material, and the strain energy density As a new feature Add to target deformation data feature vector In the above example, the target deformation data feature vector is .

[0029] In the preferred solution, in step S5, the hierarchical deformation data feature vector is obtained. , and calculate the eigenvector in With the eigenvector The Euclidean distance between , normalize different eigenvalues, map eigenvalues ​​of different dimensions and magnitudes to the interval [0,1], and set the eigenvector and the eigenvector The Features and ; The normalized eigenvalue formula is as follows: ; ; At the same time, the strain energy density As a new feature Substitute into the feature vector , then the eigenvector , and for the newly added features According to the same logic, the minimum and maximum values ​​in the historical data are normalized to obtain the normalized eigenvalues. , the formula is as follows: ; in, and The eigenvectors are With the eigenvector Middle The eigenvalues ​​after feature normalization, is the feature vector Middle The eigenvalues ​​after feature normalization, , Characteristics The minimum and maximum values ​​in the historical data, , Characteristics Minimum and maximum values ​​in historical data; Substitution , and , then the normalized target deformation data feature vector and the normalized classified deformation data feature vector The Euclidean distance between The formula is: ; Among them, the Euclidean distance By measuring the difference between the current tunnel state and the predefined graded deformation state, that is, the normalized target deformation data feature vector and the normalized classified deformation data feature vector The difference between the strain energy density and the As a new feature Add to target deformation data feature vector After that, the entire data processing and analysis process will undergo a series of changes.

[0030] For normalization, as mentioned earlier, the new feature The normalized formula ,Sure and It is necessary to conduct a comprehensive sorting and statistical analysis of the strain energy density data in the historical data, involving the strain energy density in different time periods, different tunnel parts or different working conditions. The data is mined to ensure accurate and representative minimum and maximum values.

[0031] Specifically, if the Euclidean distance The value of is small, then the target deformation data feature vector at the current moment With the classified deformation data feature vector The smaller the difference between them, the higher the similarity, that is, the higher the execution high; If the Euclidean distance The larger the value of is, the greater the difference between the target deformation data feature vector at the current moment and the graded deformation data feature vector is, and the lower the similarity is, the lower the execution degree is. Low.

[0032] In the preferred solution, in step S5, according to the normalized feature vector With the eigenvector The Euclidean distance between To determine the degree of execution ; Set execution degree The formula is as follows: ; in, is the maximum value of the Euclidean distance, It is expressed as a normalized distance metric with a threshold between [0,1]; if At this time, the execution degree , that is, the actual state of the current tunnel fully matches the graded security state, and the tunnel is in the safest state. On the contrary, if The value increases, As the value of increases, the execution The value of decreases, which means that the difference between the current actual state of the tunnel and the classified safety state is getting bigger and bigger, and the degree to which the tunnel deviates from the safety state is increasing; Calculated execution rate , the warning levels are divided into the following situations: when When the tunnel is in a safe state, it is judged to be in a safe state, indicating that the current tunnel state is very close to the classified safe state with a small difference. All monitoring indicators are within the normal range. The tunnel is in a safe and stable operating state and no special early warning measures are needed.

[0033] when , it is a mild warning. At this time, the tunnel status has begun to deviate from the safe state, but the degree of deviation is relatively small. It may be some small deformation or change trend, which requires certain attention, but has not yet reached a more serious level.

[0034] when When the warning is 0, it is a moderate warning, indicating that the deformation or state change of the tunnel has become obvious, and there may be certain safety hazards. Some measures need to be taken to strengthen monitoring, such as arranging professional personnel to conduct on-site inspections or inspect related equipment.

[0035] when When the warning is on, it is a high warning, indicating that the tunnel is in a dangerous state, the deformation is large or the change trend is serious, emergency measures need to be taken, and there is a great safety risk. A detailed inspection and assessment must be carried out immediately, and effective measures must be taken to reduce the risk, such as restricting traffic in the tunnel or preparing emergency repair plans.

[0036] when When the tunnel is in a dangerous state, it is judged as a dangerous state. At this time, the safety condition of the tunnel is already in a very dangerous situation. An accident is likely to occur. The tunnel faces a greater security threat. Measures must be taken immediately to ensure the safety of personnel and facilities to ensure the safe operation of the tunnel, such as evacuating personnel and stopping tunnel operations.

[0037] That is, execution The warning level classification can be expressed by the following formula: ; Among them, the execution Euclidean distance The closer the value is to 1, the closer it is to the graded safety status. The scope is divided into warning levels. It is the warning level threshold set according to tunnel safety.

[0038] Specifically, the execution degree is further calculated based on the Euclidean distance When Still applicable, but now Already includes new features The Euclidean distance calculation result indicates the execution degree The calculation results will comprehensively consider the impact of new features and determine the execution degree. Warning level threshold , , , When the new features are added, it is necessary to review and make corresponding adjustments, because the addition of new features changes the structure of the entire feature space and the data distribution. The thresholds previously determined based on data that do not contain new features are no longer applicable. By simulating and analyzing a large amount of historical data containing new features and combining the requirements and experience of actual tunnel safety management, reasonable warning level thresholds can be re-determined to ensure the accuracy and effectiveness of the early warning system.

[0039] In the preferred solution, in step S5, when there are risk factors around the entire tunnel section, different influencing factors have different degrees of influence on the tunnel displacement characteristics, and the displacement characteristic threshold adjustment coefficient for: ; in, For the The influence index of each influencing factor, the value range , is the number of influencing factors, then the adjusted threshold , is the original threshold; Specifically, the original threshold It is a standard value used to determine that the tunnel displacement is within a safe range. When there are risk factors in the surrounding area, the displacement characteristic threshold adjustment coefficient Original threshold Adjust to get, if ,but equal , which means that the overall impact of surrounding factors is relatively large, and the safety judgment standard remains unchanged; if ,but Less than , which means that the early warning mechanism becomes more sensitive under the influence of surrounding factors, and can detect possible abnormal displacement earlier, thereby issuing early warnings; if , the strictness of safety judgment is improved. When there are influencing factors around the tunnel, the adjustment coefficient is calculated by comprehensively considering the influencing indicators of each influencing factor. , dynamically adjust the threshold. The more influencing factors there are and the larger the influencing index is, the higher the adjustment coefficient will be. The larger the threshold The smaller it is, the higher the sensitivity of the early warning will be, ensuring that the tunnel can be monitored timely and effectively under special circumstances, and enabling the early warning mechanism to adapt to changes in the environment around the tunnel.

[0040] Set execution degree The sequence is: ; in, for time point; Then calculate the moving average The formula is as follows: ; Among them, the execution The sequence records the execution status of the tunnel at different time points, reflecting the degree of deviation of the tunnel from the safe state at different times. The moving average The calculation is to observe the execution The average level change trend within a certain time range is selected by including the current time point Recently Execution value at a time point Sum and divide by the window size , obtained It is the average execution degree within this window, which can smooth out some short-term and accidental data fluctuations, and more clearly show the long-term trend of execution degree over time, which helps to more accurately judge whether the tunnel status is gradually getting better or worse.

[0041] Calculate moving standard deviation The formula is as follows: ; in, is the moving window size, moving standard deviation It is used to measure the discreteness of the execution data in the moving window to reflect the execution value around its moving average. The fluctuation size of the system first calculates the execution degree at each time point. With the current moving average The squares of the differences are then summed over the moving window and divided by the window size. Then take the square root and you get the moving standard deviation. , a smaller mobile standard A small moving standard deviation means that the execution is relatively stable within this time window, with small fluctuations; a large moving standard deviation means that the execution value changes dramatically, with more unstable factors, and is affected by some temporary interference factors. Further analysis is needed to determine whether the fluctuation indicates that the real state of the tunnel has changed; like and , then adjust the warning level; in, is the trend threshold, is the standard deviation threshold.

[0042] Specifically, Indicates the current time point Compared with the previous time point When the difference is greater than the trend threshold When The obvious upward or downward trend means that the tunnel status is changing continuously, that is, it is getting worse or better. By calculating the moving average of the execution degree and moving standard deviation , analyze the changing trend and stability of execution degree in time series. If the moving average An upward trend and moving standard deviation If the moving average is small, it means that the tunnel status is deteriorating and the change is relatively stable, and the warning level needs to be increased; on the contrary, if the moving average is stable And the moving standard deviation If the data fluctuation is large, there may be data fluctuation interference, which needs further analysis and confirmation. Dynamic evaluation and early warning adjustment of the tunnel status from the perspective of time series ensures that this change trend is not caused by random fluctuations or unstable factors in the data, but is a relatively stable and reliable state change trend. Only when these two conditions are met at the same time will the warning level be adjusted, which can avoid misjudgment of the warning level due to individual abnormal data or short-term fluctuations, making the early warning mechanism more scientific and reasonable, and can accurately make corresponding early warning responses according to the actual state changes of the tunnel to ensure the safety of tunnel operations.

[0043] In the preferred solution, in step S5, a hybrid algorithm of the back propagation algorithm and the loss function is used, and the deformation degree of the tunnel section predicted by the deep learning model is set to , and its calculation formula is: ; in, Expressed as at time The deep learning model predicts the deformation degree of the tunnel section. This prediction value is used to assess the deformation degree of the tunnel at that moment, so as to facilitate subsequent status assessment and early warning. The prediction function of the deep learning model is a complex function built based on the deep learning algorithm. By learning a large amount of historical data, a mapping relationship from the input feature vector to the predicted value of the deformation degree is established. is the parameter set of the model. The parameters are adjusted by the back propagation algorithm during the model training process. They determine the specific prediction behavior of the model. For the moment The eigenvector of The eigenvector of Input to a The predicted value of the deformation degree of the tunnel section at that moment is obtained in the deep learning model ; Calculate deformation trend ,like The absolute value of is greater than the set trend threshold , then the deformation trend is determined to be abnormal; Among them, the deformation trend Expressed as at time The deformation trend of the tunnel section is approximated by calculating the difference between the predicted values ​​of deformation degree at adjacent moments. and Respectively expressed as time and time The difference between them reflects the change of deformation degree between these two moments.

[0044] When abnormal deformation is determined Existence, that is ,but ,otherwise ; in, The abnormal deformation threshold is a pre-set judgment value based on factors such as the tunnel design standard and historical monitoring data. When the predicted deformation value exceeds this threshold, it is judged that abnormal deformation exists. Indicates the time used to determine If the abnormal deformation , then there is abnormal deformation, if , the deformation is within the normal range. For and abnormal deformation threshold Make comparisons; Warning signs The determination formula is as follows: ; in, Indicates at time Is a warning signal issued? , a warning is issued if , it means safe and no warning is issued; is the safety threshold, used to predict the deformation degree Make comparisons; Specifically, the degree of deformation is taken into account , deformation trend and abnormal deformation Three factors, as long as any one of the conditions is met, that is, when Exceeding safety threshold When the deformation trend The absolute value of exceeds the trend threshold When abnormal deformation When there is an abnormality, the warning signal , indicating that an early warning is needed, otherwise , indicating that the system is in normal status and no alarm is needed.

[0045] In the preferred solution, in step S5, the deep learning model is regularly updated and trained using new data. Time collects new monitoring data from the last update to the current time and its corresponding real tunnel section status information , build a new training dataset , using the training dataset Retrain the deep learning model and use the mean square error loss function as the optimization objective function; During the training process, calculation With the true value The average of the squares of the differences between the two is used to continuously adjust the model parameters through the back propagation algorithm. , so that the loss function The minimization formula is as follows: ; in, It is expressed as the value of the loss function, which is used to measure the difference between the predicted results of the deep learning model and the actual results. During the model training process, the goal is to minimize this loss function by adjusting the model parameters. It is represented by the number of samples of newly collected monitoring data used to update the model and its corresponding real tunnel section status information, and Respectively represent the input feature vector of the sample (at time The preprocessed multi-source data feature vector) and the corresponding real tunnel section state value, is the model's input feature vector The prediction result is obtained by calculating the sum of the squares of the differences between the prediction result and the true result and taking the average to get the value of the loss function.

[0046] In the specific implementation, first for each sample , that is, from 1 to , calculate the model prediction results With the true value The square of the difference , and then sum these squared differences over all samples , and finally divided by the number of samples , get the average difference, that is, the mean square error loss function value During the training process, the goal is to continuously adjust the model parameters , using the back propagation algorithm to make this loss function As the value of gradually decreases, it means that the prediction results of the model are getting closer and closer to the true value, and the accuracy and reliability of the model are constantly improving.

[0047] Specifically, whether it is regularly using new data to update and train deep learning models, or determining false alarms and optimizing models after warning events, new features will become important considerations when updating training, new training data sets In If included , then in calculating the loss function When the model predicts the new features, the accuracy will directly affect the value of the loss function, thus affecting the model parameters. In the process of false alarm judgment and model optimization, the warning event is related to the new feature. For example, the cause of the false alarm may be the anomaly of the new feature data or the model's incorrect interpretation of the new feature. Therefore, when analyzing the cause of the false alarm and taking optimization measures, it is necessary to conduct in-depth analysis and adjustment of the data preprocessing algorithm of the new feature and the part of the model structure related to the new feature, so as to improve the system's adaptability to the new features and the reliability of the entire monitoring and early warning system.

[0048] In the preferred solution, in the data early warning module, when the relevant monitoring data Perform analysis and verification to determine the false positives, and set the true status determined by analysis and verification to ; If the false report is , the real anomaly is , then the formula for determining the false alarm is as follows: ; Specifically, After the relevant monitoring data After detailed analysis and verification, the true status is determined. , indicating that after further inspection and verification, there is actually no abnormal situation in the tunnel at present, that is, the previous warning was inaccurate and was a false alarm; If the warning signal And the real state When these two conditions are met at the same time, , it is determined to be a false alarm, while in other cases, if the warning signal or And the real state ,but , it is determined to be a real abnormal situation or a normal situation without warning; in, To determine the time False alarms that occur, It is represented by the actual state of the tunnel at the time determined by manual inspection and other auxiliary means. If , indicating that there is actually no abnormality; if , indicating that there is indeed an abnormality. But there is actually no abnormality It is considered a false alarm only when

[0049] In case of false reporting , then analyze the causes of false positives, such as modifying the threshold of data cleaning, improving the weight distribution of data fusion, etc., to improve the quality and accuracy of data and reduce the occurrence of false positives; or optimize the model structure, such as adding regularization terms to the network to prevent overfitting, adjusting the number of layers or nodes in the network, etc., to improve the performance and stability of the model; If the abnormal situation , then the processed abnormal data Incorporate into the training set to construct a new optimized training data set , which contains the original training data and the newly added data after abnormal processing. The model is optimized and trained again, with the goal of minimizing the loss function, and the model parameters are adjusted so that the model can better learn and identify abnormal situations, enhance the model's ability to identify special situations, and further improve the accuracy and reliability of monitoring and early warning.

[0050] Example 3 Further illustrate with reference to Example 1, Figure 1-3 The structure shown in the figure, the implementation method of the tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology includes the following steps: S1. Sensor laying inside surrounding rock: Drilling preparation: After tunnel excavation, drill holes at designated locations.

[0051] Fixing the sensor: First, place the bagged anchor into the drill hole and use a steel bar to push it to the very end and break it. Then, place the fiber optic monitoring sensor in a black rubber sheath and insert it into the drill hole until the very end is embedded in the anchor.

[0052] Fixing and sealing: Finally, quick-setting mud is injected into the borehole to ensure that the fiber optic sensor is firmly fixed inside the borehole.

[0053] S2. Sensor laying between surrounding rock and initial support: Installation on steel mesh: Due to the uneven surface of the surrounding rock, it is not feasible to directly attach the fiber optic sensor. Therefore, the fiber optic sensor is arranged along the steel mesh on the side of the initial support, and is laid out in a circle along the outer side from the lower left corner to the lower right corner of the tunnel.

[0054] Cable management: Finally, all external cables of the sensors are concentrated and placed in the lower right corner of the tunnel.

[0055] S3. Sensor laying between primary support and secondary lining: Attaching the initial support: After the initial support is completed, attach the fiber optic sensor to the side facing the inside of the tunnel, and also lay a circle along the outside from the lower left corner to the lower right corner of the tunnel.

[0056] Cable convergence: The external cables for these sensors also end up in the lower right corner of the tunnel, converging with cables from other locations.

[0057] S4. Comprehensive cable management: Reserved channel and access to data collector: All external cables of sensors located in the lower right corner of the tunnel are gathered together. When the secondary lining construction is carried out, a channel is reserved in the lower left corner of the tunnel, through which all cables are pulled to one side of the tunnel and finally connected to the data collector.

[0058] Example 4 Further described in conjunction with Embodiments 1 and 2, an embodiment of the present application provides an electronic device, which may include: A memory, a processor, and a computer program stored in the memory and executable on the processor.

[0059] When the processor executes the program, the comprehensive monitoring system for the entire tunnel section based on distributed optical fiber sensing technology, the early warning analysis method and the implementation method provided in the above-mentioned embodiments are implemented.

[0060] Furthermore, the electronic device further comprises: Communication interface, used for communication between memory and processor. Memory is used to store computer programs that can be run on the processor.

[0061] The memory may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0062] If the memory, processor and communication interface are implemented independently, the communication interface, memory and processor can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0063] The processor may include one or more processing units, for example, the processor may include an application processor (AP), an application specific integrated circuit (ASIC), a modem processor, a graphics processor (CPU), an image signal processor (ISP), a controller, a memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU). Among them, different processing units may be independent devices or integrated into one or more processors. Among them, the controller may be a neural center and a command center. The controller may generate an operation control signal according to the instruction opcode and the timing signal to complete the control of fetching and executing instructions. A memory may also be provided in the processor for storing instructions and data. In some embodiments, the memory in the processor is a high-speed cache memory. The memory may store instructions or data that have just been used or circulated by the processor. If the processor needs to use the instruction or data again, it may be directly called from the memory. Repeated access is avoided, the waiting time of the processor is reduced, and the efficiency of the system is improved.

[0064] A visualization module is used to display images, videos, etc. The visualization module may include a display panel, which may be a liquid crystal display (LCD), an organic light-emitting diode (OLED), an active-matrix organic light-emitting diode or an active-matrix organic light-emitting diode (AMOLED), a flexible light-emitting diode (FLED), Miniled, MicroLed, Micro-oLed, a quantum dot light-emitting diode (QLED), etc. In some embodiments, the visualization module may include 1 or N display modules, where N is a positive integer greater than 1.

[0065] Optionally, in a specific implementation, if the memory, processor and communication interface are integrated on a chip, the memory, processor and communication interface can communicate with each other through an internal interface.

[0066] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned tunnel full-section comprehensive monitoring system and early warning analysis method and implementation method based on distributed optical fiber sensing technology.

[0067] An embodiment of the present application also provides a computer program product, which can run computer instructions. When the computer instructions are executed by a processor, the above-mentioned tunnel full-section comprehensive monitoring system and early warning analysis method and implementation method based on distributed optical fiber sensing technology are implemented.

[0068] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology, characterized in that: It includes a terminal processing system and a cloud AI processing system, wherein the terminal processing system and the cloud AI processing system are connected to each other by signals to jointly construct a two-level interactive system architecture; The terminal processing system is used to quantify the risk factors of the entire tunnel section into a numerical feature vector, and calculate the state difference of the same tunnel monitoring section at different time points according to the numerical feature vector through a specific algorithm to obtain the target deformation data, and determine the execution degree according to the target deformation data. ; According to the execution Determine the current warning status of the tunnel section.

2. According to claim 1, the tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology is characterized in that: The terminal processing system includes a data acquisition module, a network connection module, a data analysis module and a self-checking module; the data acquisition module, the network connection module, the data analysis module and the self-checking module are electrically connected to each other; Data acquisition module, including a fiber optic sensor array and an intelligent numbering junction box, is used to collect electrical signal data from multiple distributed fiber optic sensors at each monitoring section to monitor temperature, strain and displacement parameters in the tunnel; The network connection module is used to connect the system to the external network and transmit the collected electrical signal data to the data analysis module; The data analysis module is used to process and analyze the received electrical signal data, obtain the monitoring results of various monitoring items inside the tunnel, and extract numerical feature vectors; The system self-check module is used to check whether the entire monitoring system network is in a normal connection state and whether each module is working normally. It determines the system status by sending and receiving specific detection electrical signals and comparing them with the preset normal electrical signal range; The cloud AI processing system includes a data warning module and a visualization module; The data warning module and the visualization module are electrically connected to each other; The data early warning module is used to predict the subsequent trend of the data obtained from monitoring, and combine the data with tunnel safety to make safety early warnings; The visualization module is used to dynamically and intuitively display the changes in monitoring data of each part of each monitoring section.

3. An early warning analysis method for a tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology according to any one of claims 1 to 2, characterized in that: The following steps are involved: S1. Data acquisition: The electrical signal data of multiple distributed optical fiber sensors at each monitoring section are collected through the optical fiber sensor sequence in the data acquisition module of the terminal processing system to obtain the electrical parameters related to temperature, strain and displacement in the tunnel; The distributed optical fiber sensor sequence uses a specific electrical sensing principle to convert physical quantities into measurable electrical signals, and connects each sensor to the corresponding intelligent numbered junction box interface, accesses the main cable interface along the tunnel, transmits data to the optical fiber demodulator through the main cable, and performs signal source grouping and secondary grouping through the network connection module and the data analysis module; S2, data transmission: using the network connection module of the terminal processing system to transmit the collected data to the local indoor server; The network connection module adopts a redundant network connection design, including wired network connection and wireless network connection, and is connected to the local network through the on-site PC and router; S3. Data analysis: In the data analysis module of the terminal processing system, the received electrical sensor data is processed and analyzed using the supporting signal processing software to obtain the monitoring results of various monitoring items inside the tunnel, and the numerical feature vector is extracted from the received monitoring results to obtain the target deformation data feature vector at a certain moment. ; The target deformation data is data of at least one deformation feature vector of a displacement feature, a deformation rate feature, a cumulative displacement feature, and a relative displacement feature, and the displacement feature, the deformation rate feature, the cumulative displacement feature, and the relative displacement feature are all calculated based on the conversion of electrical signal data collected by the distributed optical fiber sensor; S4, system self-check: by using the abnormality check port of the supporting signal processing software to check the sensor connection status and data capture measurement status, when the self-check result is normal, the monitoring data is transmitted to the local data storage; if there is an abnormality, it returns to the supporting signal processing software to report the specific abnormal point; S5. Early warning judgment: In the data early warning module of the cloud AI processing system, the monitoring data in the local data storage is obtained through the network server, and the corresponding monitoring data is transmitted to the big data platform. The monitoring data along the distributed optical fiber sensor is learned and predicted based on the deep learning model, and the predicted data is compared with the execution degree. Combined to determine different safety warning levels, the data is stored in the cloud database. When the monitoring value approaches or exceeds the safety threshold, the warning channel is activated and different forms of warning notifications are sent to relevant personnel, including SMS, email, sound and light alarms; Among them, a support vector machine, a random forest or a neural network deep learning model is selected, and the deep learning model is obtained by training historical monitoring data and corresponding tunnel status data; S6. Visualization display: The visualization module of the cloud AI processing system is used to dynamically and intuitively display the changes in the monitored electrical data of each part of each monitoring section, and the stored data is transmitted to the three-dimensional visualization software through the cloud database to dynamically display the status information of the tunnel in the form of visual charts or three-dimensional models.

4. According to claim 3, the tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology is characterized in that: In step S3, to construct the target deformation data feature vector The weight of the target deformation data feature vector The features in are dynamically weighted to evaluate the relative importance of the tunnel deformation state, and the tunnel operation time is assumed to be , Tunnel operation time The time variable is used to adjust the weight of the displacement feature according to the different operation stages of the tunnel. The displacement feature weight is The calculation formula is: ； in, , It is an adjustment coefficient used to control the rate at which the weight changes over time. Different values ​​will cause the weight to grow or decrease at different rates at different time stages. , As the time dividing point, It divides the initial and middle stages of tunnel operation. It divides the middle and late stages; The strain energy density is calculated based on the strain signal data collected by the distributed optical fiber sensor in the data acquisition module. , and in the target deformation data feature vector Strain energy density Characteristic index, strain energy density The formula used is as follows: ； in, is the strain data after the strain signal is converted, is the elastic modulus of the material, and the strain energy density As a new feature Add to target deformation data feature vector In the above example, the target deformation data feature vector is , strain energy density An indicator used to evaluate the energy accumulation inside the tunnel structure. When the strain energy density exceeds a certain limit, it means that the tunnel section structure is in a high stress state and there is a greater risk of deformation or damage.

5. According to claim 4, the tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology is characterized in that: In step S5, the classified deformation data feature vector is obtained , and calculate the target deformation data feature vector With the classified deformation data feature vector The Euclidean distance between , normalize different eigenvalues, map eigenvalues ​​of different dimensions and magnitudes to the interval [0,1], and set the target deformation data eigenvector and the classified deformation data feature vector The Features and ; The normalized eigenvalue formula is as follows: ； ； The target deformation data feature vector Normalize the minimum and maximum values ​​in the historical data to obtain the normalized eigenvalues , the formula is as follows: ； in, and are the target deformation data feature vectors With the classified deformation data feature vector Middle The eigenvalues ​​after feature normalization, is the feature vector Middle The eigenvalues ​​after feature normalization, , Characteristics The minimum and maximum values ​​in the historical data, , Characteristics The minimum and maximum values ​​in the historical data, the normalized eigenvalues Used to analyze the strain energy density in different time periods, different tunnel parts or different working conditions Data mining to ensure accurate and representative minimum and maximum values; Substitution , and , then the normalized target deformation data feature vector and the normalized classified deformation data feature vector The Euclidean distance between The formula is: ； Among them, the Euclidean distance By measuring the difference between the current tunnel state and the predefined graded deformation state, that is, the normalized target deformation data feature vector and the normalized classified deformation data feature vector The spatial distance between the tunnels can be used to assess the safety of the tunnel and to classify the warning levels accordingly; If the Euclidean distance The value of is small, then the target deformation data feature vector at the current moment With the classified deformation data feature vector The smaller the difference between them, the higher the similarity, that is, the higher the execution high; If the Euclidean distance The larger the value of is, the greater the difference between the target deformation data feature vector at the current moment and the graded deformation data feature vector is, and the lower the similarity is, the lower the execution degree is. Low.

6. The tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology according to claim 5 is characterized in that: In step S5, according to the normalized feature vector With the eigenvector The Euclidean distance between To determine the degree of execution ; Set execution degree The formula is as follows: ； in, is the maximum value of the Euclidean distance, is represented as a normalized distance metric with a threshold between [0,1]; if At this time, the execution degree , that is, the actual state of the current tunnel fully matches the graded security state, and the tunnel is in the safest state. On the contrary, if The value increases, As the value of increases, the execution The value of decreases, which means that the difference between the current actual state of the tunnel and the classified safety state is getting bigger and bigger, and the degree to which the tunnel deviates from the safety state is increasing; Then, the execution degree The warning levels are as follows: ； Among them, the execution Euclidean distance The closer the value is to 1, the closer it is to the graded safety status. The scope is divided into warning levels. It is the warning level threshold set according to tunnel safety.

7. The tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology according to claim 6 is characterized in that: In step S5, when there are risk factors around the entire tunnel section, different influencing factors have different degrees of influence on the tunnel displacement characteristics, and the displacement characteristic threshold adjustment coefficient for: ； in, For the The influence index of each influencing factor, the value range , is the number of influencing factors. Sum the influencing indicators of the influencing factors and divide them by the number of influencing factors , and obtain the displacement characteristic threshold adjustment coefficient reflecting the overall impact , displacement feature threshold adjustment coefficient It is used to measure the demand for adjustment of displacement feature threshold due to changes in the surrounding environment; The adjusted threshold The calculation formula is: ； in, is the original threshold, the original threshold It is a standard value used to determine that the tunnel displacement is within a safe range. When there are risk factors in the surrounding area, the displacement characteristic threshold adjustment coefficient Original threshold Adjust to get, if ,but equal , which means that the overall impact of surrounding factors is large, and the safety judgment standard remains unchanged; if ,but Less than , which means that the early warning mechanism is sensitive to the surrounding factors and issues an early warning; if , then the strictness of safety judgment is improved; Set execution degree The sequence is: ； in, for time point; Calculating Moving Averages The formula is as follows: ； Calculate moving standard deviation The formula is as follows: ； in, Indicates at a point in time The moving average of execution degree, is the moving window size, moving standard deviation It is used to measure the discreteness of the execution data in the moving window to reflect the execution value around its moving average. The size of fluctuations; like and , then adjust the warning level; in, is the trend threshold, is the standard deviation threshold, Indicates the current time point Compared with the previous time point The difference between the execution moving averages, if the moving average An upward trend and moving standard deviation If the moving average is small, the tunnel status is deteriorating and the change is stable. On the contrary, if the moving average is stable And the moving standard deviation If it is larger, it is determined that there is data fluctuation interference.

8. The tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology according to claim 7 is characterized in that: In step S5, a hybrid algorithm of back propagation algorithm and loss function is used, and the deformation degree of the tunnel section predicted by the deep learning model is assumed to be , and its calculation formula is: ； in, represents the prediction function of the deep learning model, is the parameter set of the model, For the moment The eigenvector of The eigenvector of Input to a The predicted value of the deformation degree of the tunnel section at that moment is obtained in the deep learning model ; Calculate deformation trend ,like The absolute value of is greater than the set trend threshold , then the deformation trend is determined to be abnormal; When abnormal deformation is determined Existence, that is ,but ,otherwise ; in, is the abnormal deformation threshold; Warning signs The determination formula is as follows: ； in, is the safety threshold, if the warning signal , then the execution issues a warning, otherwise , confirm that the system is in normal condition and no alarm is required.

9. The tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology according to claim 8 is characterized in that: In step S5, the deep learning model is regularly updated with new data. Time collects new monitoring data from the last update to the current time and its corresponding real tunnel section status information , build a new training dataset , using the training dataset Retrain the deep learning model and use the mean square error loss function as the optimization objective function; During the training process, the difference between the predicted results and the actual results is calculated, and the model parameters are continuously adjusted through the back propagation algorithm. , using the mean square error loss function The minimization formula is as follows: ； in, It is represented by the number of samples of the collected new monitoring data and its corresponding real tunnel section status information, Expressed as at time The feature vector formed by the collected new monitoring data after preprocessing, is represented as a set of parameters representing a deep learning model, Represented as a deep learning model based on the input feature vector and current parameters The prediction results made, Represented as the input feature vector The corresponding real tunnel section status information; In the data early warning module, when the relevant monitoring data Perform analysis and verification to determine the false positives, and set the true status determined by analysis and verification to ; If the false report is , the real anomaly is , then the formula for determining the false alarm is as follows: ； If the warning signal And the real state ,but , it is determined to be a false alarm. If the warning signal or And the real state ,but , it is determined to be a real abnormal situation or a normal situation without warning.

10. The implementation method of the tunnel full-section comprehensive monitoring system based on distributed optical fiber sensing technology according to any one of claims 1 to 2 is characterized in that: The following steps are involved: S1. Inside the surrounding rock: after excavation, drill a hole, place and break the bagged anchor, insert the rubber sleeve with the optical fiber sensor into the anchor, and inject quick-setting mud to fix it; S2, between surrounding rock and primary support: attach the optical fiber sensor to the primary support steel mesh, and lay it from the lower left corner along the outer side to the lower right corner; S3, between primary support and secondary lining: After the primary support is completed, the optical fiber sensor is attached to the inner side of the primary support, also from the lower left corner to the lower right corner; S4. Cable management: All sensor cables converge at the lower right corner, are pulled to one side of the tunnel through the channel reserved at the lower left corner, and are connected to the data collector.

Images