Subway tunnel multi-direction monitoring method and system, electronic device and storage medium

Abstract

This application discloses a method, system, electronic device, and storage medium for multi-directional monitoring of subway tunnels. The method includes acquiring multi-source heterogeneous monitoring data of the subway tunnel structure and surrounding geological environment; preprocessing the multi-source heterogeneous monitoring data to obtain a standardized monitoring sequence; determining the initial reference value for each monitoring point based on the standardized monitoring sequence, and dynamically updating the initial reference value according to the structural service status; independently analyzing and cross-validating the deformation response and environmental disturbance parameters of the tunnel structure based on the standardized monitoring sequence and the initial reference value, including: quantifying the correlation between different monitoring indicators to identify abnormal causes and obtaining structural health status assessment results; and outputting a graded early warning signal when the structural health status assessment results meet preset early warning conditions. This application meets the comprehensive requirements of modern urban rail transit for the safety monitoring of subway tunnel structures in terms of comprehensiveness, real-time performance, and accuracy.

Description

Technical Field

[0001] This application relates to the field of engineering monitoring technology, and in particular to a method, system, electronic device and storage medium for multi-directional monitoring of subway tunnels. Background Technology

[0002] Currently, during the construction and long-term operation of subway tunnels, the structure is continuously subjected to the coupled effects of multiple factors, including the evolution of complex geological conditions, disturbances from adjacent projects, and cyclical traffic loads. This easily induces typical deformation responses such as horizontal displacement of shield tunnel segments, vertical settlement, crown settlement, and convergence of cross-sectional clearance. Simultaneously, the surrounding rock-support system may experience damage phenomena such as crack propagation and localized spalling. Critical structural components face the risk of stress concentration and cumulative deterioration, seriously threatening the tunnel's service safety and the stability of the surrounding environment.

[0003] Current mainstream methods for monitoring the health of tunnel structures are mostly based on the discrete deployment of single-type sensors, lacking multi-physics sensing capabilities and system-level integrated architecture. These methods generally suffer from problems such as limited monitoring dimensions, insufficient spatiotemporal resolution, and poor data synchronization, making it difficult to achieve collaborative sensing and fusion analysis of multi-source heterogeneous parameters such as structural deformation, crack evolution, and stress state. Furthermore, existing monitoring systems have not yet established a unified characterization framework for various typical defects such as uneven settlement and local instability, leading to delayed anomaly identification and difficulties in tracing the root causes. This makes it difficult to meet the comprehensive requirements of modern urban rail transit for the safety monitoring of subway tunnel structures in terms of comprehensiveness, real-time performance, and accuracy. Summary of the Invention

[0004] In order to overcome the shortcomings of existing technologies and meet the comprehensive requirements of modern urban rail transit for the safety monitoring of subway tunnel structures in terms of comprehensiveness, real-time performance and accuracy, this application provides a multi-directional monitoring method, system, electronic equipment and storage medium for subway tunnels.

[0005] Firstly, the objective of this invention is achieved through the following technical solution: Multi-directional monitoring methods for subway tunnels include: Acquire multi-source heterogeneous monitoring data of the subway tunnel structure and surrounding geological environment; the multi-source heterogeneous monitoring data is collected at different monitoring points at preset frequencies by various types of sensors deployed on the subway tunnel structure and surrounding environment. The multi-source heterogeneous monitoring data is preprocessed, including: removing statistical outliers, correcting system errors based on environmental auxiliary parameters, and performing dimensional normalization to obtain a standardized monitoring sequence. Based on the standardized monitoring sequence, the initial reference value of each monitoring point is determined, and the initial reference value is dynamically updated according to the structural service status. Based on the standardized monitoring sequence and the initial benchmark value, the deformation response and environmental disturbance parameters of the tunnel structure are independently analyzed and cross-validated, including: calculating the deviation and time-series change rate of each monitoring index relative to the benchmark state, comparing the deviation and time-series change rate with preset safety thresholds respectively, and quantifying the correlation between different monitoring indices to identify abnormal causes and obtain the structural health status judgment result. When the structural health status determination result meets the preset early warning conditions, a graded early warning signal is output.

[0006] By adopting the above technical solution, this invention addresses the problem of fragmented multi-dimensional state perception caused by isolated sensors and asynchronous sampling in traditional monitoring. It deploys a multi-type sensor network covering the tunnel structure and surrounding geological environment, synchronously collecting data at a preset frequency. This constructs a multi-source heterogeneous monitoring network with high spatiotemporal resolution and high monitoring efficiency, thereby achieving coordinated characterization of key parameters such as structural deformation, crack evolution, and environmental disturbances. Furthermore, by performing outlier removal, system error correction based on environmental auxiliary parameters such as temperature / water level, and dimensional normalization on the raw data, false anomalies introduced by sensor drift, environmental interference, and dimensional differences are effectively eliminated, significantly improving the signal-to-noise ratio and cross-index comparability of subsequent analyses. Based on this, an initial benchmark value dynamically updated according to the structural service status is introduced. This not only avoids false alarms or missed alarms caused by benchmark failure during long-term monitoring but also enables the health assessment baseline to adaptively evolve with long-term structural creep or major disturbance events. This invention also compares the deviation and rate of change of each monitoring indicator with static / dynamic safety thresholds, and performs cross-validation using multi-indicator correlation analysis. This not only identifies single out-of-limit events but also distinguishes between structural damage and non-hazardous environmental responses through the coupling relationships between indicators (such as the strong correlation between arch subsidence and sidewall convergence), thereby accurately locating the causes of anomalies. Finally, this invention achieves graded early warning triggered by the linkage of data criteria across all dimensions, significantly improving the accuracy of early warnings and the efficiency of operation and maintenance response.

[0007] In a preferred embodiment of this application, the independent analysis of the deformation response of the tunnel structure includes: Quantitative assessment of horizontal displacement and vertical settlement: For any monitoring point i, calculate the horizontal displacement deviation. Vertical settlement deviation ,in These represent the current measured horizontal displacement value and the current measured vertical settlement value, respectively. The corresponding baseline horizontal displacement and baseline vertical settlement values ​​are used; a least squares linear regression model is employed. Fitting or The slope of the deformation trend is obtained from the sequence of changes over time t. ,when When the deformation rate exceeds the threshold allowed by the specification, it is determined to be an unstable state; Co-operational analysis of crown settlement and sidewall convergence: Calculation of crown settlement ,in This represents the current measured settlement of the arch. The arch crown settlement is used as a baseline value; and horizontal displacement data of the same cross section are acquired simultaneously. The Pearson correlation coefficient between the arch crown settlement and the horizontal displacement data is calculated. When the Pearson correlation coefficient is greater than a first preset correlation threshold and the arch crown settlement exceeds the standard threshold, it is determined that the overall structure is dominated by compression deformation; when the Pearson correlation coefficient is less than a second preset correlation threshold and the arch crown settlement increases significantly, it is determined that the local arch crown is unstable. Identification of the distribution characteristics of clearance convergence: Calculate the relative convergence values ​​between each measuring point within the cross-section. ,in Let i be the baseline convergence value between monitoring point i and monitoring point j. The current measured convergence value is used to identify the net clearance convergence mode of the current cross section, in order to distinguish between uniform convergence mode and local abrupt change.

[0008] By adopting the above technical solutions, and targeting typical tunnel deformation modes, refined modeling is performed on horizontal / vertical displacement, crown settlement-sidewall convergence synergy, and cross-sectional clearance convergence distribution characteristics. By fitting the deformation trend slope with least squares, the risk of accelerated deformation can be quantitatively identified. The correlation threshold between crown settlement and horizontal displacement is used to distinguish between overall compression deformation and local crown instability. The relative convergence values ​​between measuring points are used to identify uniform convergence and local abrupt changes, thereby achieving accurate identification of different failure modes of the tunnel structure.

[0009] In a preferred embodiment, this application includes the analysis of environmental disturbance parameters of a tunnel structure, comprising: Nonlinear prediction of crack propagation: Calculating crack width increment And adopt an exponential growth model The crack development trend is fitted, where A and B are fitting parameters. If the crack width increment in the predicted future period exceeds the critical value, an early warning is issued that the crack is accelerating its expansion. Coupled analysis of stress state: Calculation of actual stress value and stress change Define the regional stress concentration factor ,in The average stress value of the region is given. A stress concentration factor greater than 1.5 indicates the presence of stress concentration. The causes of stress anomalies are analyzed by correlating the crown settlement and seepage pressure values ​​of the region. The pre-built causal inference model outputs the probability of the occurrence of abnormal events based on the multi-source heterogeneous monitoring data.

[0010] By adopting the above technical solutions, nonlinear fitting of crack width increment based on the exponential growth model can predict the accelerated crack propagation trend in advance and overcome the conservatism of linear extrapolation; defining regional stress concentration coefficients and setting thresholds can effectively identify potential high-stress areas; further, stress anomalies are correlated with parameters such as crown settlement and seepage pressure, and the causal probability is output by combining the pre-constructed causal inference model, realizing the acquisition of information from phenomenon observation to quantification of causes.

[0011] In a preferred embodiment of this application, the formula for measuring the groundwater level monitoring data is: in, It is the first The groundwater level depth value obtained from this measurement. This represents the elevation of the location of the water level probe. For the first The frequency of the water level probe read by the frequency reader during the next measurement. is the reference frequency of the water level probe under the reference zero water level condition, and k is the frequency-depth conversion coefficient determined by the probe material properties and installation conditions.

[0012] By adopting the above technical solution, the frequency readings are accurately converted into water level depth, eliminating systematic errors caused by individual sensor differences or installation deviations, and significantly improving the accuracy of groundwater level monitoring.

[0013] In a preferred embodiment of this application, the step of dynamically updating the initial reference value based on the structural service status includes: A sliding window mechanism is adopted, using a continuous standardized monitoring sequence of length N as the evaluation window, where N is the number of days in the monitoring period; When the mean drift of the monitored indicators within the window exceeds twice the corresponding historical standard deviation, or when the cumulative service time of the structure reaches the preset maintenance node, the benchmark value update process is triggered, and the sliding average value of the new window is used as the updated initial benchmark value.

[0014] By adopting the above technical solution, the initial baseline value is dynamically updated and a dual update trigger condition for the initial baseline value is set, so that the baseline value can not only track the normal creep trend of the structure, but also be reset in time during major disturbances or service phase transitions.

[0015] In a preferred embodiment of this application: after obtaining the structural health status assessment result, the method further includes: Based on historical multi-source heterogeneous monitoring data and structural design parameters, a digital twin of the tunnel is constructed, and a physical information neural network is embedded in the digital twin of the tunnel to fuse monitoring data and mechanical control equations. When a high-risk abnormal event is detected, the tunnel digital twin is invoked to perform a virtual disturbance injection experiment to simulate the impact of different causes on the current structural state and obtain virtual disturbance response information. The actual monitoring response information and the virtual disturbance response information are compared for similarity to obtain the matching degree; if the matching degree is higher than the preset matching degree threshold, the abnormal cause type is confirmed and the cause confidence score is output. The construction of the tunnel digital twin includes: The BIM structural model serves as the geometric framework, and is associated with geological stratification, lining material properties, and support structure information. The standardized monitoring sequence is input as a boundary condition into the finite element solver to update the internal stress-strain field in real time. The mechanical transmission path between monitoring points is modeled by a spatiotemporal graph convolutional network to generate a structural response propagation map, which is used to identify potential cascading failure regions.

[0016] By adopting the above technical solution, a tunnel digital twin is constructed that integrates BIM geometry, material properties and real-time monitoring data, and a Physical Information Neural Network (PINN) is embedded to integrate data and mechanical equations. When a high-risk anomaly occurs, the structural response under different causes (such as sudden surge, excavation and vibration) is simulated through virtual disturbance injection experiments and compared with the actual monitoring response, so as to upgrade the traditional correlation analysis to a physically explainable causal confirmation, which significantly improves the accuracy of attribution of abnormal events.

[0017] In a preferred embodiment of this application, the quantification of the correlation between different monitoring indicators to identify abnormal causes includes: Calculate the dynamic Pearson correlation coefficient between any two monitoring indicator sequences to generate a multi-indicator correlation matrix; When the absolute value of the Pearson correlation coefficient of a certain indicator pair exceeds the first preset threshold, and exhibits synchronous mutation characteristics or lag response characteristics in time, it is marked as a candidate correlation pair; For each candidate association pair, time-delay mutual information analysis and Granger causality test are performed to determine whether there is a statistically significant one-way or two-way causal relationship, and then a local causal subgraph is constructed. All local causal subgraphs are merged into a global multivariate causal graph model, where nodes are monitoring indicators and edge weights are determined by a weighted average of causal strength and time delay. A structural learning algorithm is used to extract the dominant causal chain from the multivariate causal graph model, and typical failure modes are matched in combination with a preset expert rule base. Based on the global causal graph model, causal paths ending with structural deformation indices are identified, and each causal path is mapped to the corresponding abnormal cause type by combining a pre-built expert rule base. Based on the number and combined strength of currently activated causal paths, the identification results and confidence levels of abnormal causes are output.

[0018] By adopting the above technical solutions, "hidden chain failure areas" that have not yet shown macroscopic deformation but have already redistributed internal forces can be identified in advance, realizing a leap from monitoring the apparent state to early warning of the internal mechanical mechanism. Specifically, the Spatiotemporal Graph Convolutional Network (ST-GCN) models the mechanical transmission path between monitoring points and generates a structural response propagation spectrum, which can dynamically capture the transmission law of stress / deformation in the circumferential and longitudinal directions of the tunnel and the energy accumulation area; combined with the finite element solver to update the internal stress-strain field in real time, the digital twin not only has geometric realism, but also mechanical computability.

[0019] Secondly, the objective of this invention is achieved through the following technical solution: The subway tunnel multi-directional monitoring system includes: The multi-source data acquisition module is used to acquire multi-source heterogeneous monitoring data of the subway tunnel structure and surrounding geological environment; the multi-source heterogeneous monitoring data is collected at different monitoring points at a preset frequency by various types of sensors deployed on the subway tunnel structure and surrounding environment. The data preprocessing module is used to preprocess the multi-source heterogeneous monitoring data, including: removing statistical outliers, correcting system errors based on environmental auxiliary parameters, and performing dimensional normalization to obtain standardized monitoring sequences. The dynamic benchmark management module is used to determine the initial benchmark value of each monitoring point based on the standardized monitoring sequence, and to dynamically update the initial benchmark value according to the structural service status. The health status analysis module is used to independently analyze and cross-validate the deformation response and environmental disturbance parameters of the tunnel structure based on the standardized monitoring sequence and the initial benchmark value. This includes: calculating the deviation and time-series change rate of each monitoring index relative to the benchmark state, comparing the deviation and time-series change rate with preset safety thresholds respectively, and quantifying the correlation between different monitoring indices to identify abnormal causes and obtain the structural health status judgment result. The graded early warning output module is used to output a graded early warning signal when the structural health status determination result meets the preset early warning conditions.

[0020] Thirdly, the objective of this invention is achieved through the following technical solution: An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described multi-directional monitoring method for subway tunnels.

[0021] Fourthly, the objective of this invention is achieved through the following technical solution: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described multi-directional monitoring method for subway tunnels.

[0022] In summary, this application includes at least one of the following beneficial technical effects: 1. A multi-source heterogeneous sensor network covering the tunnel structure and surrounding geological environment was constructed, and outlier removal, environmental parameter compensation and dimension normalization were performed on the collected data, which effectively improved the reliability and cross-dimensional comparability of the original monitoring data; an initial benchmark value mechanism based on dynamic updates of service status was introduced, which enabled the health assessment baseline to be adaptively adjusted with long-term structural creep or external disturbances, avoiding misjudgment caused by fixed benchmarks. 2. By further combining deviation, rate of change and correlation analysis of multiple indicators, independent identification and cross-validation of deformation response and environmental disturbance can be achieved. This not only detects abnormal states, but also preliminarily distinguishes between structural damage and non-hazardous environmental response, significantly improving the accuracy and timeliness of structural health assessment. Attached Figure Description

[0023] Figure 1 This is a flowchart of a multi-directional monitoring method for subway tunnels according to one embodiment of this application; Figure 2 This is a flowchart following step S5 in a multi-directional monitoring method for subway tunnels according to an embodiment of this application; Figure 3 This is a schematic diagram of a device according to one embodiment of this application. Detailed Implementation

[0024] The present application will be further described in detail below with reference to the accompanying drawings.

[0025] In one embodiment, such as Figure 1 As shown, this application discloses a multi-directional monitoring method for subway tunnels, which specifically includes the following steps: S1: Acquire multi-source heterogeneous monitoring data of the subway tunnel structure and surrounding geological environment; the multi-source heterogeneous monitoring data is collected at different monitoring points at preset frequencies by various types of sensors deployed on the subway tunnel structure and surrounding environment.

[0026] In this embodiment, multi-source heterogeneous monitoring data refers to a collection of monitoring information from different types of physical quantities, different sampling frequencies, and different spatial locations; including but not limited to: tunnel structure response data and environmental disturbance parameters. Tunnel structure response data includes, for example, segment horizontal displacement, vertical settlement, crown settlement, clearance convergence, crack width, and support axial force; environmental disturbance parameters include, for example, groundwater level, surface settlement, tilt of adjacent buildings, and blasting vibration velocity. The preset frequency is dynamically set according to the engineering risk level, for example, up to once per hour during shield tunneling through existing lines, and once per day during normal operation.

[0027] Specifically, referencing the experience of a certain metro line 7 Phase II project, static levels were deployed in the shield tunnel section to monitor arch settlement, with a sampling frequency of once every 6 hours; string-type crack gauges (50mm range) were installed at the segment joints to record crack opening and closing in real time; forced centering prisms were deployed on the tunnel sidewalls and track bed, and horizontal / vertical displacements were automatically observed by a surveying robot (such as a Leica TS15); simultaneously, settlement markers were deployed on the ground surface above the tunnel, at the corners of adjacent buildings, and at underground pipeline nodes, and were periodically remeasured using an electronic level (Trimble DiNi) with second-order leveling accuracy. In addition, groundwater levels were collected using an LRK-SW-RYY01 automatic water level gauge, and blasting vibrations were recorded using a TC-4850N wireless vibration meter to record the three-component velocity. All sensor data was transmitted back to the "Smart Monitoring Cloud Platform" via 4G / fiber optic cable, achieving unified timestamp alignment and centralized storage. The Smart Monitoring Cloud Platform is an intelligent terminal used by users or managers.

[0028] S2: Preprocess the multi-source heterogeneous monitoring data, including: removing statistical outliers, correcting systematic errors based on environmental auxiliary parameters, and performing dimensional normalization to obtain standardized monitoring sequences.

[0029] In this embodiment, statistical outliers refer to data points that significantly deviate from the normal fluctuation range due to sensor malfunction, electromagnetic interference, or accidental impacts, and are identified using the 3σ principle. Environmental auxiliary parameters refer to external variables that affect sensor readings but are not part of the structural response itself, such as the effect of temperature on the frequency drift of a string instrument and the disturbance of atmospheric pressure on the liquid level in a hydrostatic level. Dimensional normalization aims to eliminate the interference of different physical dimensions (mm, kPa, Hz, etc.) on subsequent multi-index fusion analysis. Min-Max or Z-score methods are used to map the data to [0, 1] or the standard normal distribution interval, forming a comparable "standardized monitoring sequence".

[0030] Specifically, for the raw frequency data F of the string-type crack gauge, abrupt outliers are first eliminated using the 3σ principle; then, the frequency is corrected for temperature drift based on the actual measured temperature T on site. Where α is the temperature correction factor. The reference temperature is F, and the measured frequency is F; then the corrected frequency is... Converted to crack width w, and finally normalized using Min-Max: For hydrostatic leveling data on sedimentation, the liquid column height variation is corrected by combining it with concurrent air pressure data.

[0031] Furthermore, the multi-source heterogeneous monitoring data includes groundwater level monitoring data, and the measurement formula for groundwater level monitoring data is: in, It is the first The groundwater level depth value obtained from this measurement. This represents the elevation of the location of the water level probe. For the first The frequency of the water level probe read by the frequency reader during the next measurement. is the reference frequency of the water level probe under the reference zero water level condition, and k is the frequency-depth conversion coefficient determined by the probe material properties and installation conditions.

[0032] Specifically, because the output of a string-type water level gauge is a resonant frequency signal, its value is affected by the probe's own physical characteristics and the on-site installation conditions, and cannot directly represent the water level depth. These physical characteristics include elastic modulus and cavity volume; on-site installation conditions include verticality and siltation. Therefore, it is necessary to establish a linear mapping relationship between frequency change and water head height through physical calibration, i.e., introducing a frequency-depth conversion coefficient k. For example, during the construction of a connecting tunnel for a rail transit line, a dedicated monitoring hole was drilled 10m outside the tunnel, and an LRK-SW-RYY01 string-type water level gauge with a range of 0-50m and a resolution of 1mm was lowered. After installation, with the probe fully exposed to the atmosphere (i.e., water level at zero), its initial reference frequency was recorded. =2856.3Hz; simultaneously, the absolute elevation of the center point of the probe's sensing film was accurately measured using a total station. =12.75m. Then, on-site calibration was performed: Clean water of a known depth was injected into the borehole, such as 10.0m, and the frequency F was recorded as 2832.1Hz. The value of k was then calculated as ( - ) / ( -F)=(12.75−2.75) / (2856.3−2832.1)≈0.413m / Hz, where, =2.75 represents the water level depth. During formal monitoring, the sampling frequency is... Then, substitute into the formula This allows you to obtain the groundwater level depth in real time.

[0033] S3: Based on the standardized monitoring sequence, determine the initial reference value of each monitoring point, and dynamically update the initial reference value according to the service status of the structure.

[0034] In this embodiment, the initial baseline value refers to the average value of monitoring indicators of the structure during the initial stage of stable service (e.g., no significant disturbance within 3 months after construction completion), representing the reference level under healthy conditions. "Structural service status" includes cumulative operating time, major events experienced (e.g., heavy rainfall, adjacent foundation pit excavation, earthquake), and the statistical characteristics of the monitoring sequence itself. Dynamic updates aim to adaptively adjust the baseline value according to the long-term creep or periodic disturbances of the structure.

[0035] For example, the average settlement data from days 30 to 60 after tunnel boring machine (TBM) completion is selected as the initial baseline value. A sliding window mechanism is then used, with a window length of N = 15 days, and the average value within the window is calculated daily. Compared with historical standard deviation ;like If a significant state transition occurs (such as due to a sudden rise in the water level of Wuyong), a baseline update is triggered, and the new window mean is set as the updated baseline. Simultaneously, if the structure's cumulative service reaches a preset maintenance milestone, such as one year, the baseline is forcibly reset to incorporate long-term creep effects, even without significant drift.

[0036] Specifically, the initial baseline values ​​are dynamically updated based on the structural service status, including: S31: A sliding window mechanism is adopted, using a continuous standardized monitoring sequence of length N as the evaluation window, where N is the number of days in the monitoring period.

[0037] In this embodiment, N is set to 7-30, with 7 days applicable to high-risk periods (such as crossing existing lines or during periods of heavy rainfall) and 30 days applicable to stable operation periods.

[0038] S32: When the mean drift of the monitored index within the window exceeds twice the corresponding historical standard deviation, or when the cumulative service time of the structure reaches the preset maintenance node, the benchmark value update process is triggered, and the sliding average value of the new window is used as the updated initial benchmark value.

[0039] In this embodiment, the historical standard deviation refers to the standard deviation of the structure calculated during the initial stabilization period (e.g., the first 90 days). Preset maintenance milestones are key time points set based on engineering experience and specifications, such as 1 year, 3 years, and 5 years of operation, used to force a reset of the benchmark to incorporate long-term cumulative effects.

[0040] S4: Based on standardized monitoring sequences and initial benchmark values, independently analyze and cross-validate the deformation response and environmental disturbance parameters of the tunnel structure, including: calculating the deviation and time-series change rate of each monitoring index relative to the benchmark state, comparing the deviation and time-series change rate with preset safety thresholds respectively, and quantifying the correlation between different monitoring indicators to identify abnormal causes and obtain the structural health status judgment results.

[0041] In this embodiment, independent analysis refers to judging the exceedance and rate of a single monitoring indicator; cross-validation refers to verifying the authenticity of anomalies and inferring their causes through the coupling relationship between multiple indicators (such as the synchronicity of settlement and water level); preset safety thresholds are set based on the "Technical Specification for Monitoring of Urban Rail Transit Engineering" (GB50911-2013) and engineering experience, such as a cumulative settlement value of 20mm for tunnel segments and a rate of 3mm / d. Quantitative correlation is achieved using Pearson correlation coefficient, mutual information, or Granger causality test to distinguish between structural damage and environmental disturbances.

[0042] Specifically, the independent analysis of the deformation response of the tunnel structure includes: S401: Quantitative assessment of horizontal displacement and vertical settlement: For any monitoring point i, calculate the horizontal displacement deviation. Vertical settlement deviation ,in These represent the current measured horizontal displacement value and the current measured vertical settlement value, respectively. The corresponding baseline horizontal displacement and baseline vertical settlement values ​​are used; a least squares linear regression model is employed. Fitting or The slope of the deformation trend is obtained from the sequence of changes over time t. ,when When the deformation rate exceeds the threshold allowed by the specification, it is determined to be an unstable state.

[0043] In this embodiment, the allowable deformation rate threshold is 0.5 mm / d, and the time t is approximately 7 days.

[0044] S402: Co-operational analysis of crown settlement and sidewall convergence: Calculation of crown settlement ,in This represents the current measured settlement of the arch. The base value for the crown settlement is used; and the horizontal displacement data of the same section is acquired simultaneously. The Pearson correlation coefficient between the crown settlement and the horizontal displacement data is calculated. When the Pearson correlation coefficient is greater than the first preset correlation threshold and the crown settlement exceeds the standard threshold, it is determined that the overall structure is dominated by compression deformation. When the Pearson correlation coefficient is less than the second preset correlation threshold and the crown settlement increases significantly, it is determined that the local crown instability is caused by compression deformation.

[0045] In this embodiment, the subsidence of the arch and the horizontal displacement of the sidewalls reflect the deformation behavior of the tunnel's top and waist, respectively. If the two are highly synchronized (strong positive correlation), it indicates that the surrounding rock is uniformly convergent, representing overall compression deformation; if the arch continues to subside while the sidewall displacement is weak (weak correlation), it suggests that the arch support has failed or that local soil has collapsed, indicating high-risk local instability. Specifically, the first preset correlation threshold is 0.8, and the second preset correlation threshold is 0.5.

[0046] S403: Identification of the distribution characteristics of clearance convergence: Calculate the relative convergence values ​​between each measuring point within the cross-section. ,in Let i be the baseline convergence value between monitoring point i and monitoring point j. The current measured convergence value is used to identify the net clearance convergence mode of the current cross section, in order to distinguish between uniform convergence mode and local abrupt change.

[0047] In this embodiment, clearance convergence refers to the reduction in distance between any two points within the tunnel cross-section, reflecting the overall closing trend of the cross-section. For example, in a circular shield tunnel cross-section, eight measuring points (numbered 1-8, where 1 is the arch crown and 5 is the invert) are arranged circumferentially. Reference convergence matrix. Established from initial steady-state period measurements. Current field measurements show that the convergence increments of 1-5 (crown-invert) are... =12.3mm, 3–7 (left waist–right waist) =11.8mm, with convergence differences in all directions <10%, is judged as a "uniform convergence mode," consistent with normal formation consolidation characteristics. However, in the section crossing the fault zone, the convergence increment at measuring points 2–6 (upper right–lower left) reaches 18.7mm, while that at 1–5 is only 6.2mm. Furthermore, a sudden increase in the fracture gauge reading occurs near measuring point 2, which the system identifies as a "local abrupt change mode." The risk area is located on the right shoulder, and it is recommended to focus on checking the sealing of the segment joints and the fullness of the grouting behind them.

[0048] In step S4, the analysis of environmental disturbance parameters of the tunnel structure includes: S404: Nonlinear prediction of crack propagation: Calculating crack width increment And adopt an exponential growth model The crack development trend is fitted, where A and B are fitting parameters. If the crack width increment in the predicted future period exceeds the critical value, an early warning is issued that the crack is accelerating its expansion.

[0049] S405: Coupled analysis of stress state: Calculation of actual stress values and stress change Define the regional stress concentration factor ,in The average stress value of the region is given. A stress concentration factor greater than 1.5 indicates the presence of stress concentration. The causes of stress anomalies are analyzed by correlating the crown settlement and seepage pressure values ​​of the region.

[0050] For example, fiber optic strain gauges (FBGs) are deployed in a section of a tunnel to collect real-time circumferential stress on the tunnel segments. A 15m long monitoring section with 12 measuring points is selected. The maximum stress is measured at a certain moment. =18.6MPa, regional average stress =12.4MPa, calculated as follows =18.6 / 12.4≈1.499<1.5, not reaching the threshold. However, it was subsequently found that the settlement rate of the arch in this area suddenly increased to 3.8 mm / d (exceeding 3 mm / d), and the groundwater level dropped by 2.1 m, leading to intensified soil consolidation. After reassessment, it was found that the stress peak continued to rise, reaching its peak on the 5th day. =1.72>1.5, the system determined that stress concentration occurred. Further correlation analysis showed that the area was located within the influence range of the adjacent foundation pit dewatering well, and the surface settlement and the arch settlement were strongly correlated (Pearson coefficient 0.91), which was ultimately attributed to the chain effect of "groundwater level reduction causing stratum unloading - surrounding rock shrinkage - support structure tension".

[0051] S406: The pre-built causal inference model is based on multi-source heterogeneous monitoring data and outputs the probability of the cause of abnormal events.

[0052] In this embodiment, the pre-built causal inference model is an intelligent analysis engine trained based on historical multi-source monitoring data. It employs a Bayesian network, and the input includes multi-dimensional time series such as crack width, stress, settlement, water level, and vibration. The output is a "probability distribution of the causes of abnormal events". For example, if a crack suddenly increases, the model can output: "caused by a sudden drop in groundwater level (probability 65%)", "excited by vibration from a nearby blasting (probability 25%)", or "caused by cumulative material creep (probability 10%)".

[0053] Specifically, the causal inference model is a probabilistic graphical model built on a Bayesian network. The nodes in the causal inference model correspond to monitoring index variables, such as the crack width increment. , vault settlement rate Groundwater level change ΔD, support axial force N, blasting vibration velocity Regional stress concentration factors, etc.; the edges in the causal inference model represent the conditional dependencies between variables. The model input of the causal inference model is a multidimensional time series window of length T, X∈R. T×M Where M represents the number of monitoring indicators, e.g., M=8 indicates 8 types of monitoring indicators, including crown settlement, horizontal convergence, crack width, pore water pressure, surface settlement, vibration velocity, temperature, and humidity. N=7 represents the monitoring data for the past 7 days; the output is the probability distribution vector of the causes of the abnormal events. These correspond to four pre-defined causes: ① groundwater disturbance, ② adjacent construction excavation, ③ cumulative cyclic load from trains, and ④ material creep and degradation. The causal inference model is trained through the following steps: (1) Collect historical monitoring datasets containing at least 50 cases of abnormal events with confirmed causes, and label each case with the actual cause label; (2) Extract the multi-source heterogeneous monitoring sequences from the previous 7 days for each event as input features and perform standardization processing; (3) The conditional probability table (CPT) of the Bayesian network is learned by the maximum likelihood estimation method. The hyperparameters are set as follows: number of trees = 2000, minimum number of node samples = 20. (4) Evaluate the accuracy of the causal inference model for causal classification on the validation set. When the Top-1 accuracy is ≥80%, freeze the model parameters and deploy it to the smart monitoring cloud platform.

[0054] The input data of the causal inference model comes from the standardized monitoring sequence after preprocessing in step S2, and the deviation is calculated based on the benchmark value dynamically updated in step S3; the output causal probability P is used to weight and correct the structural health status judgment result in step S4.

[0055] For example, at a certain cross-section arch settlement point, the current deviation is calculated to be -18.2 mm, which does not exceed the cumulative threshold of 20 mm; however, the rate of change over the past 3 days is 3.5 mm / d, exceeding the rate threshold of 3 mm / d, triggering a yellow alert. Further cross-validation revealed that the groundwater level dropped by 2.1 m during the same period, and the Pearson correlation coefficient between the two was 0.87, > the preset correlation threshold of 0.8, indicating that the settlement was caused by groundwater desiccation and was considered an acceptable environmental disturbance, not structural instability. Conversely, if the settlement accelerates while the water level remains stable, accompanied by a sudden increase in sidewall convergence, and the Pearson correlation coefficient is >0.9, then the risk of support system failure is identified. The independent analysis and cross-validation process is automatically executed through the "Smart Monitoring Cloud Platform," outputting the structural health status as "Level of Concern, Caused by Groundwater Disturbance."

[0056] S5: When the structural health status assessment result meets the preset early warning conditions, output a graded early warning signal.

[0057] In this embodiment, the preset warning conditions are a logical combination of multi-dimensional criteria, for example, divided into three levels: Level 1 (blue) is a slight exceedance of a single limit; Level 2 (yellow) is multiple coordinated anomalies or a rapid deterioration of a single indicator; Level 3 (red) is a severe exceedance of a key indicator pointing to structural damage. The graded warning signal not only includes the alarm level, but also includes the location of the anomaly, possible causes, development trends, and suggested measures.

[0058] For example, when the monitoring shows that the settlement rate of the arch of the connecting passage reaches 4.2 mm / d, which exceeds the threshold of 3 mm / d and has a correlation coefficient of 0.92 with the surrounding surface settlement, the intelligent monitoring cloud platform system automatically determines it as a level-two warning and generates a report: "There is an accelerated settlement risk in the connecting passage area of ​​the intermediate ventilation shaft of a certain subway station. The cause is dewatering of the adjacent foundation pit. It is recommended to intensify monitoring and check the operation status of the dewatering wells." The report is then immediately pushed to the user terminal with the communication connection.

[0059] In one embodiment, such as Figure 2 As shown, after obtaining the structural health status assessment results, the multi-directional monitoring method for subway tunnels also includes: S61: Based on historical multi-source heterogeneous monitoring data and structural design parameters, a digital twin of the tunnel is constructed, and a physical information neural network is embedded in the tunnel digital twin to fuse monitoring data and mechanical control equations.

[0060] In this embodiment, the Physics-informed Neural Network (PINN) serves as a data-physics fusion engine. By embedding the elasticity control equations into the neural network loss function, it ensures that the prediction results both fit the measured data and satisfy the physical laws.

[0061] Specifically, constructing a digital twin of the tunnel includes: S611: Use the BIM structural model as the geometric skeleton and associate it with geological stratification, lining material properties and support structure information.

[0062] In this embodiment, taking two station sections of a certain rail transit line as an example, the Revit BIM model from the design phase is used as the geometric skeleton. This model includes structural parameters such as shield tunnel segment ring width of 1.5m, outer diameter of 6.2m, inner diameter of 5.5m, C50 concrete lining, and M30 bolt connections. Simultaneously, a three-dimensional stratigraphic model from the geological survey report is associated, including: ① artificial fill layer (0-2m), ② silty clay layer (2-10m, c=25kPa, φ=18°), ③ medium-coarse sand layer (10-15m, permeability coefficient k=1×10⁻⁶). -4 ④ Completely weathered granite (>15m). Support structure information includes the spacing of the steel arch frames of the connecting passage (0.75m) and the grouting range (3m), etc.

[0063] S612: Input the standardized monitoring sequence as a boundary condition into the finite element solver to update the internal stress-strain field in real time.

[0064] In this embodiment, the finite element solver CalculiX uses an 8-node hexahedral element (C3D8R) to discretely analyze the tunnel-surrounding rock system, with a total degree of freedom of approximately 1.2 million. A static analysis is automatically run every 24 hours to update the internal stress-strain field, and the results are stored in a time-series database.

[0065] S613: The mechanical transmission path between monitoring points is modeled by a spatiotemporal graph convolutional network to generate a structural response propagation map, which is used to identify potential cascading failure regions.

[0066] In this embodiment, the graph structure definition of the Spatiotemporal Graph Convolutional Network (ST-GCN) includes nodes representing all sensor locations, for example, a total of 128 nodes; edges are jointly determined by spatial distance (<5m) and historical cross-correlation coefficient (>0.6); node features are standardized displacement, velocity, and acceleration triples; the time dimension is the past 7 days, forming a spatiotemporal tensor of T=7. ST-GCN contains two layers of spatiotemporal convolutional blocks, each containing temporal convolution (kernel=3) and graph convolution (Chebyshev order=2), outputting the response propagation weight vector for each node. High-weight paths can be identified through the response propagation weight vector. For example, if an abnormal signal from a certain arch node rapidly spreads along the path "arch top → right arch waist → invert arch," the system marks this path as a "potential cascading failure region."

[0067] Finally, based on the aforementioned BIM–FEA–ST-GCN fusion framework, a PINN module is embedded: PINN takes spatial coordinates (x, y, z) and time t as input and outputs a displacement field u(x, y, z, t); its loss function is: ,in, =128 represents the number of monitoring points. 10,000 represents the number of PDE sampling points. =1.0, =0.8 is the weighting coefficient. After training, PINN can generate a continuous displacement field in the entire space with an error RMSE of less than 0.25mm.

[0068] S62: When a high-risk abnormal event is detected, the tunnel digital twin is invoked to perform a virtual disturbance injection experiment to simulate the impact of different causes on the current structural state and obtain virtual disturbance response information.

[0069] In this embodiment, high-risk abnormal events refer to events that are identified as red alerts in step S4, such as settlement rate > 4 mm / d and stress concentration factor > 1.7. Virtual disturbance injection involves superimposing typical external loads on the current digital twin state to simulate the structural response under different inducing factors.

[0070] Specifically, when the system detects that the settlement rate of the arch of the Shuixi Station connecting passage reaches 4.3 mm / d (exceeding the threshold), it immediately invokes the digital twin to execute three sets of parallel virtual experiments: Groundwater inrush scenario: An instantaneous pore water pressure increment Δu = +60 kPa was applied to a sand layer unit (10-15 m depth) to simulate pipe rupture. The simulation lasted for 2 hours and coupled seepage-stress analysis (Biot consolidation theory) was used. Scenario of excavation of adjacent foundation pit: Remove 3m×15m×20m of soil 12m to the right of the tunnel to simulate the foundation pit of a new station. The unloading can be done in 3 steps, 1m each. Train vibration and impact scenario: A moving harmonic load F(t) = 8sin(2π⋅12t) kN / m is applied to the centerline of the track at a speed of 80 km / h for 45 seconds. Explicit dynamic analysis is used.

[0071] Each experiment uses the current stress-strain field reconstructed by PINN as the initial state. After running, it outputs the virtual disturbance response information of all monitoring points, including displacement time history, stress increment, and crack opening.

[0072] S63: Compare the actual monitoring response information with the virtual disturbance response information to obtain the matching degree; if the matching degree is higher than the preset matching degree threshold, confirm the type of abnormal cause and output the cause confidence score.

[0073] In this embodiment, the actual monitoring response information refers to the standardized displacement sequence 72 hours before the anomaly occurred; the matching degree is calculated using weighted dynamic time warping (W-DTW) distance, which assigns higher weight to critical time periods.

[0074] For example, obtain the actual crown settlement sequence (sampling interval 6 hours): =[−1.8, −2.2, −2.7, −3.1, −3.6, −4.0, −4.3] mm. Comparison with three sets of virtual responses: Groundwater inrush response: =[-1.7, -2.1, -2.6, -3.0, -3.5, -3.9, -4.2] Excavation response: =[-1.5, -1.8, -2.0, -2.2, -2.4, -2.5, -2.6] Train vibration response: =[−1.8±0.2,−2.2±0.2,…] Calculate the W-DTW distance (weight vector w = [1, 1, 1, 1.2, 1.5, 1.8, 2.0]): =0.08, =0.35, =0.42. The preset matching threshold is W-DTW distance < 0.15. Because =0.08 < 0.15, the system confirms the anomaly as caused by "sudden groundwater inrush". The cause confidence score is defined as: This confidence score, along with the cause type and cascading failure path map, is output to the "Smart Monitoring Cloud Platform" to generate a structural health assessment report.

[0075] In one embodiment, step S4, quantifying the correlation between different monitoring indicators to identify abnormal causes, includes: S411: Calculate the dynamic Pearson correlation coefficient between any two monitoring indicator sequences and generate a multi-indicator correlation matrix.

[0076] In this embodiment, the monitoring index sequence includes eight types of standardized time-series data: crown settlement s(t), horizontal convergence c(t), crack width ω(t), and regional stress concentration factor. Groundwater level depth D(t), surface subsidence Explosion vibration velocity Ambient temperature T(t). The dynamic Pearson correlation coefficient is calculated using a sliding window to determine the linear correlation between any two indicators within a local time period, avoiding global correlation from masking stage coupling; window length W = 7 days, step size 1 day.

[0077] Specifically, the system updates an 8×8 dynamic correlation matrix R(t) daily, where elements Let represent the correlation coefficient between the i-th and j-th indicators on days [t−6, t].

[0078] S412: When the absolute value of the Pearson correlation coefficient of a certain indicator pair exceeds the first preset threshold and exhibits synchronous mutation characteristics or lag response characteristics in time, it is marked as a candidate correlation pair.

[0079] In this embodiment, the first preset threshold is set to 0.8. Synchronous mutation refers to a jump of more than 2 standard deviations between two sequences at the same time point. Lag response refers to a significant response of one sequence within 1-3 days after a change in another sequence, determined by the peak position of the cross-correlation function.

[0080] S413: For each candidate association pair, perform time-delayed mutual information analysis and Granger causality test to determine whether there is a statistically significant one-way or two-way causal relationship, and then construct a local causal subgraph.

[0081] In this embodiment, Time Delay Mutual Information (TDMI) is used to capture nonlinear dependencies, and the formula is as follows: ,in This represents the value of the monitoring indicator X at the current time t. Let p(x, y) represent the value of another monitoring indicator Y in the past time t−τ; p(x, y) represents the joint probability distribution; p(x) is... The marginal probability distribution, i.e., the probability that X is x; p(y) is... The marginal probability score is the probability that Y takes the value y; the time delay τ∈{0,1,2,3} days is chosen to maximize I. The optimal lag is used. The Granger causality test is based on a vector autoregression (VAR) model. The null hypothesis is "Y does not Granger cause X". If the p-value of the F-test is <0.05, the null hypothesis is rejected, and causality is determined.

[0082] The continuous monitoring sequence was discretized using equal-frequency binning (10 bins per index) to estimate p(x,y); the Granger causality test used the AIC criterion to automatically select the optimal order of the VAR model (maximum lag of 3); all causality tests were conducted under the condition of controlling other variables, i.e., conditional Granger tests.

[0083] For example, regarding the indices of crown settlement and groundwater level depth: TDMI is maximum at τ=0, I=0.78; Granger test: p-value for water level → settlement = 0.003 < 0.05, p-value for settlement → water level = 0.42 > 0.05; it is determined to be a one-way causal relationship: water level → settlement.

[0084] For the index pair of crack width and ambient temperature: TDMI is maximum at τ=0, I=0.65; the p-values ​​of both-way Granger tests are <0.05, but physically, temperature change precedes crack response, so the temperature→crack relationship is retained based on domain knowledge. Each valid causal pair constitutes a local causal subgraph.

[0085] S414: Merge all local causal subgraphs into a global multivariate causal graph model, where nodes are monitoring indicators and edge weights are determined by a weighted average of causal strength and time delay.

[0086] In this embodiment, all local subgraphs are merged into a global directed graph G=(V, E), where the node set V represents 8 monitoring indicators, and the edge set E represents causal relationships. The edge weights are defined as follows: The mutual information weight α = 0.7. =3 days.

[0087] S415: Employs a structural learning algorithm to extract the dominant causal chain from a multivariate causal graph model and combines it with a pre-defined expert rule base to match typical failure modes.

[0088] In this embodiment, the structure learning algorithm uses the Peter-Clark (PC) algorithm to extract high-weight paths from the global graph. The expert rule base pre-defines four typical failure modes: Mode A (i.e., groundwater disturbance mode): Water level ↓ → Settlement ↑ → Stress ↑; Mode B (i.e., adjacent construction mode): Vibration ↑ → Surface settlement ↑ → Tunnel convergence ↑; Mode C (i.e., material creep mode): Time ↑ → Cracks ↑ → Stress redistribution; Mode D (i.e., temperature effect mode): Temperature ↑ → Crack ↓ (closing).

[0089] S416: Based on the global causal graph model, identify causal paths ending with structural deformation indicators, and combine them with a pre-built expert rule base to map each causal path to the corresponding abnormal cause type.

[0090] In this embodiment, the types of abnormal causes include groundwater disturbance, vibration from nearby construction, material creep degradation, or temperature stress effects; each causal path corresponds to a unique cause type. For example: the cause type for the path "water level → settlement → stress" is groundwater disturbance; the cause type for the path "vibration → surface settlement → convergence" is vibration from nearby construction; the cause type for the path "time → crack" is material creep degradation; and the cause type for the path "temperature → crack" is temperature stress effects.

[0091] S417: Based on the number and overall strength of the currently activated causal paths, output the identification results and confidence level of the abnormal causes.

[0092] In this embodiment, the overall strength is the geometric mean of the weights of the edges on the path; the confidence level is defined as: confidence level = (activation path strength / maximum strength of all possible paths) × pattern matching degree, where the pattern matching degree = 1.0 (complete match) or 0.5 (partial match).

[0093] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0094] In one embodiment, a multi-directional monitoring system for subway tunnels is provided, which corresponds to the multi-directional monitoring method for subway tunnels described in the above embodiments.

[0095] The multi-directional monitoring system for subway tunnels includes a multi-source data acquisition module, a data preprocessing module, a dynamic baseline management module, and a health status analysis module. Detailed descriptions of each functional module are as follows: The multi-source data acquisition module is used to acquire multi-source heterogeneous monitoring data of the subway tunnel structure and surrounding geological environment. The multi-source heterogeneous monitoring data is collected at different monitoring points at preset frequencies by various types of sensors deployed on the subway tunnel structure and surrounding environment. The data preprocessing module is used to preprocess multi-source heterogeneous monitoring data, including: removing statistical outliers, correcting system errors based on environmental auxiliary parameters, and performing dimensional normalization to obtain standardized monitoring sequences. The dynamic baseline management module is used to determine the initial baseline value of each monitoring point based on the standardized monitoring sequence, and to dynamically update the initial baseline value according to the service status of the structure. The health status analysis module is used to independently analyze and cross-validate the deformation response and environmental disturbance parameters of the tunnel structure based on standardized monitoring sequences and initial benchmark values. This includes: calculating the deviation and time-series change rate of each monitoring index relative to the benchmark state, comparing the deviation and time-series change rate with preset safety thresholds, quantifying the correlation between different monitoring indicators to identify abnormal causes, and obtaining the structural health status judgment result. The graded early warning output module is used to output graded early warning signals when the structural health status assessment result meets the preset early warning conditions.

[0096] For specific limitations regarding the multi-directional monitoring system for subway tunnels, please refer to the limitations of the multi-directional monitoring method for subway tunnels mentioned above, which will not be repeated here. Each module in the aforementioned multi-directional monitoring system for subway tunnels can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in the processor of a computer device in hardware form or independent of the processor, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0097] In one embodiment, an electronic device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the electronic device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores multi-source heterogeneous monitoring data and standardized monitoring sequences. The network interface communicates with external terminals via a network. When the computer program is executed by the processor, it implements a multi-directional monitoring method for subway tunnels.

[0098] In one embodiment, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of a multi-directional monitoring method for subway tunnels.

[0099] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program being executed by a processor to implement the steps of a multi-directional monitoring method for subway tunnels.

[0100] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0101] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0102] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multi-directional monitoring method for subway tunnels, characterized in that, include: Acquire multi-source heterogeneous monitoring data on subway tunnel structure and surrounding geological environment; The multi-source heterogeneous monitoring data is collected at preset frequencies at different monitoring points by various types of sensors deployed in the subway tunnel structure and surrounding environment. The multi-source heterogeneous monitoring data is preprocessed, including: removing statistical outliers, correcting system errors based on environmental auxiliary parameters, and performing dimensional normalization to obtain a standardized monitoring sequence. Based on the standardized monitoring sequence, the initial reference value of each monitoring point is determined, and the initial reference value is dynamically updated according to the structural service status. Based on the standardized monitoring sequence and the initial benchmark value, the deformation response and environmental disturbance parameters of the tunnel structure are independently analyzed and cross-validated, including: calculating the deviation and time-series change rate of each monitoring index relative to the benchmark state, comparing the deviation and time-series change rate with preset safety thresholds respectively, and quantifying the correlation between different monitoring indices to identify abnormal causes and obtain the structural health status judgment result. When the structural health status determination result meets the preset early warning conditions, a graded early warning signal is output; The independent analysis of the deformation response of the tunnel structure includes: Quantitative assessment of horizontal displacement and vertical settlement: For any monitoring point i, calculate the horizontal displacement deviation. Vertical settlement deviation ,in These represent the current measured horizontal displacement value and the current measured vertical settlement value, respectively. The corresponding baseline horizontal displacement and baseline vertical settlement values ​​are used; a least squares linear regression model is employed. Fitting or The slope of the deformation trend is obtained from the sequence of changes over time t. ,when When the deformation rate exceeds the threshold allowed by the specification, it is determined to be an unstable state; Co-operational analysis of crown settlement and sidewall convergence: Calculation of crown settlement ,in This represents the current measured settlement of the arch. The arch crown settlement is used as a baseline value; and horizontal displacement data of the same cross section are acquired simultaneously. The Pearson correlation coefficient between the arch crown settlement and the horizontal displacement data is calculated. When the Pearson correlation coefficient is greater than a first preset correlation threshold and the arch crown settlement exceeds the standard threshold, it is determined that the overall structure is dominated by compression deformation; when the Pearson correlation coefficient is less than a second preset correlation threshold and the arch crown settlement increases significantly, it is determined that the local arch crown is unstable. Identification of the distribution characteristics of clearance convergence: Calculate the relative convergence values ​​between each measuring point within the cross-section. ,in Let i be the baseline convergence value between monitoring point i and monitoring point j. The current measured convergence value is used to identify the net clearance convergence mode of the current cross section, in order to distinguish between uniform convergence mode and local abrupt change.

2. The multi-directional monitoring method for subway tunnels according to claim 1, characterized in that, Analysis of environmental disturbance parameters of the tunnel structure includes: Nonlinear prediction of crack propagation: Calculating crack width increment And adopt an exponential growth model The crack development trend is fitted, where A and B are fitting parameters. If the crack width increment in the predicted future period exceeds the critical value, an early warning is issued that the crack is accelerating its expansion. Coupled analysis of stress state: Calculation of actual stress value and stress change Define the regional stress concentration factor ,in The average stress value of the region is given. A stress concentration factor greater than 1.5 indicates the presence of stress concentration. The causes of stress anomalies are analyzed by correlating the crown settlement and seepage pressure values ​​of the region. The pre-built causal inference model outputs the probability of the occurrence of abnormal events based on the multi-source heterogeneous monitoring data.

3. The multi-directional monitoring method for subway tunnels according to claim 1, characterized in that, The multi-source heterogeneous monitoring data includes groundwater level monitoring data, and the measurement formula for the groundwater level monitoring data is: in, It is the first The groundwater level depth value obtained from this measurement. This represents the elevation of the location of the water level probe. For the first The frequency of the water level probe read by the frequency reader during the next measurement. is the reference frequency of the water level probe under the reference zero water level condition, and k is the frequency-depth conversion coefficient determined by the probe material properties and installation conditions.

4. The multi-directional monitoring method for subway tunnels according to claim 1, characterized in that, The dynamic updating of the initial baseline value based on the structural service status includes: A sliding window mechanism is adopted, using a continuous standardized monitoring sequence of length N as the evaluation window, where N is the number of days in the monitoring period; When the mean drift of the monitored indicators within the window exceeds twice the corresponding historical standard deviation, or when the cumulative service time of the structure reaches the preset maintenance node, the benchmark value update process is triggered, and the sliding average value of the new window is used as the updated initial benchmark value.

5. The multi-directional monitoring method for subway tunnels according to claim 1, characterized in that, After obtaining the structural health status assessment result, the method further includes: Based on historical multi-source heterogeneous monitoring data and structural design parameters, a digital twin of the tunnel is constructed, and a physical information neural network is embedded in the digital twin of the tunnel to fuse monitoring data and mechanical control equations. When a high-risk abnormal event is detected, the tunnel digital twin is invoked to perform a virtual disturbance injection experiment to simulate the impact of different causes on the current structural state and obtain virtual disturbance response information. The actual monitoring response information and the virtual disturbance response information are compared for similarity to obtain the matching degree; if the matching degree is higher than the preset matching degree threshold, the abnormal cause type is confirmed and the cause confidence score is output. The construction of the tunnel digital twin includes: The BIM structural model serves as the geometric framework, and is associated with geological stratification, lining material properties, and support structure information. The standardized monitoring sequence is input as a boundary condition into the finite element solver to update the internal stress-strain field in real time. The mechanical transmission path between monitoring points is modeled by a spatiotemporal graph convolutional network to generate a structural response propagation map, which is used to identify potential cascading failure regions.

6. The multi-directional monitoring method for subway tunnels according to claim 5, characterized in that, The method of quantifying the correlation between different monitoring indicators to identify abnormal causes includes: Calculate the dynamic Pearson correlation coefficient between any two monitoring indicator sequences to generate a multi-indicator correlation matrix; When the absolute value of the Pearson correlation coefficient of a certain indicator pair exceeds the first preset threshold, and exhibits synchronous mutation characteristics or lag response characteristics in time, it is marked as a candidate correlation pair; For each candidate association pair, time-delay mutual information analysis and Granger causality test are performed to determine whether there is a statistically significant one-way or two-way causal relationship, and then a local causal subgraph is constructed. All local causal subgraphs are merged into a global multivariate causal graph model, where nodes are monitoring indicators and edge weights are determined by a weighted average of causal strength and time delay. The dominant causal chain is extracted from the global multivariate causal graph model using a structural learning algorithm, and typical failure modes are matched in combination with a preset expert rule base. Based on the global multivariate causal graph model, causal paths ending with structural deformation indices are identified, and each causal path is mapped to the corresponding abnormal cause type by combining a pre-built expert rule base. Based on the number and combined strength of currently activated causal paths, the identification results and confidence levels of abnormal causes are output.

7. A multi-directional monitoring system for subway tunnels, characterized in that, include: The multi-source data acquisition module is used to acquire multi-source heterogeneous monitoring data on the subway tunnel structure and surrounding geological environment; The multi-source heterogeneous monitoring data is collected at preset frequencies at different monitoring points by various types of sensors deployed in the subway tunnel structure and surrounding environment. The data preprocessing module is used to preprocess the multi-source heterogeneous monitoring data, including: removing statistical outliers, correcting system errors based on environmental auxiliary parameters, and performing dimensional normalization to obtain standardized monitoring sequences. The dynamic benchmark management module is used to determine the initial benchmark value of each monitoring point based on the standardized monitoring sequence, and to dynamically update the initial benchmark value according to the structural service status. The health status analysis module is used to independently analyze and cross-validate the deformation response and environmental disturbance parameters of the tunnel structure based on the standardized monitoring sequence and the initial benchmark value. This includes: calculating the deviation and time-series change rate of each monitoring index relative to the benchmark state, comparing the deviation and time-series change rate with preset safety thresholds respectively, and quantifying the correlation between different monitoring indices to identify abnormal causes and obtain the structural health status judgment result. The graded early warning output module is used to output a graded early warning signal when the structural health status determination result meets the preset early warning conditions; The independent analysis of the deformation response of the tunnel structure includes: Quantitative assessment of horizontal displacement and vertical settlement: For any monitoring point i, calculate the horizontal displacement deviation. Vertical settlement deviation ,in These represent the current measured horizontal displacement value and the current measured vertical settlement value, respectively. The corresponding baseline horizontal displacement and baseline vertical settlement values ​​are used; a least squares linear regression model is employed. Fitting or The slope of the deformation trend is obtained from the sequence of changes over time t. ,when When the deformation rate exceeds the threshold allowed by the specification, it is determined to be an unstable state; Co-operational analysis of crown settlement and sidewall convergence: Calculation of crown settlement ,in This represents the current measured settlement of the arch. The arch crown settlement is used as a baseline value; and horizontal displacement data of the same cross section are acquired simultaneously. The Pearson correlation coefficient between the arch crown settlement and the horizontal displacement data is calculated. When the Pearson correlation coefficient is greater than a first preset correlation threshold and the arch crown settlement exceeds the standard threshold, it is determined that the overall structure is dominated by compression deformation; when the Pearson correlation coefficient is less than a second preset correlation threshold and the arch crown settlement increases significantly, it is determined that the local arch crown is unstable. Identification of the distribution characteristics of clearance convergence: Calculate the relative convergence values ​​between each measuring point within the cross-section. ,in Let i be the baseline convergence value between monitoring point i and monitoring point j. The current measured convergence value is used to identify the net clearance convergence mode of the current cross section, in order to distinguish between uniform convergence mode and local abrupt change.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-directional monitoring method for subway tunnels as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-directional monitoring method for subway tunnels as described in any one of claims 1 to 6.

Images