Tunnel state monitoring method and system based on cloud edge collaboration

Abstract

The invention discloses a tunnel state monitoring method and system based on cloud edge collaboration, and relates to the technical field of civil engineering tunnel detection. The monitoring method comprises the following steps: deploying a plurality of data acquisition networks and edge computing nodes in a target tunnel, pre-configuring monitoring partitions, establishing communication links between the edge computing nodes and each data acquisition network, configuring a lightweight data processing engine, and transmitting a transmission state monitoring data set correspondingly acquired in each monitoring partition; according to the technical scheme, the number of linkage partitions is dynamically determined through an offset comprehensive value, self-adaptive adjustment of a monitoring response range is achieved, missing detection or redundant calculation caused by fixed neighborhood analysis is avoided, and the utilization efficiency of edge computing resources is improved to a certain degree; and an abnormal parameter quantity correction mechanism is introduced, so that the situation that the abnormality extends backwards to the original partial partition which is not included in analysis can be found, the purpose of capturing the expansion of the hidden seepage channel in advance is achieved, and the recognition capability of the partial abnormality is enhanced.

Description

Technical Field

[0001] This invention relates to the field of civil engineering tunnel inspection technology, specifically to a tunnel condition monitoring method and system based on cloud-edge collaboration. Background Technology

[0002] Civil engineering tunnel inspection refers to the periodic or continuous monitoring of the structural condition during tunnel construction and operation through manual inspections or automated equipment to identify abnormalities such as cracks, leaks, deformation, and settlement. Traditional inspections often rely on periodic manual inspections combined with locally deployed sensors, such as installing convergence meters at the arch crown to measure deformation or burying piezometers at seepage points. Although some newly built tunnels have introduced automated monitoring systems, they mostly adopt a centralized architecture, with all sensor data uploaded to a central server for processing, and the edge devices only performing data acquisition functions.

[0003] Existing tunnel condition monitoring technologies have the following significant technical shortcomings:

[0004] Firstly, for example, after a long period of operation, a mountain highway tunnel experienced localized lining cracking. Because the monitoring zones were divided into 200-meter intervals, and the cracks happened to be located at the boundary of the surrounding rock grade (e.g., Grade III to Grade V), this critical location was not designated as a separate zone, and the sensors were sparsely deployed, resulting in the initial micro-crack propagation not being detected in time. When the displacement sensor finally triggered an alarm, the traditional monitoring system still conventionally retrieved data from only one zone before and after the cracks for analysis, without expanding the monitoring range and failing to detect that the anomaly had spread along the construction joint to adjacent sections. More seriously, due to the lack of intelligent edge-side discrimination capabilities, the traditional solution misjudged the instantaneous vibration caused by a heavy vehicle passing through as a structural anomaly, frequently uploading alarm information, which may cause maintenance personnel to become fatigued and ignore the real progressive deformation signals. Such problems expose the shortcomings of traditional methods in terms of coarse spatial division, rigid response mechanisms, and simplistic discrimination logic, making it difficult to adapt to the dynamic evolution monitoring needs of tunnel structures under complex geological conditions during long-term service. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The cloud-edge collaborative tunnel status monitoring method includes: deploying several data acquisition networks and edge computing nodes in the target tunnel, pre-configuring monitoring partitions, establishing communication links between the edge computing nodes and each data acquisition network, configuring a lightweight data processing engine, and transmitting the corresponding transmission status monitoring dataset collected in each monitoring partition.

[0007] Edge computing nodes receive status monitoring datasets, execute status assessment strategies for each monitoring partition according to time series, and obtain status instructions for the corresponding monitoring partitions; based on the status instructions, they select whether to trigger the linkage response analysis mechanism in the target monitoring partition to determine the linkage analysis domain.

[0008] Based on the results of the linkage response analysis mechanism, feature fusion and joint state assessment are performed, and the monitoring results are uploaded after judgment.

[0009] Furthermore, the data acquisition network includes at least: sensor components and a wireless communication module; the wireless communication module is integrated on the sensor components and is used to periodically report data to the edge computing nodes;

[0010] The configuration process of the monitoring zones is as follows: Based on the length L of the target tunnel, the tunnel is divided into zones at a set equal interval ΔL to form a preliminary zone sequence {P1, P2, ..., Pn}, where n is the number of preliminary monitoring zones; the characteristic segments of the target tunnel are acquired and mapped to the preliminary zone sequence. If any characteristic segment is contained in the initial zone, a secondary zone operation is performed to select the characteristic area as an independent zone and use the remaining zones as standard zones to form a secondary zone sequence; otherwise, the corresponding initial zone is directly used as a standard zone. For independent zones, the preset interval of its data acquisition network is Q1, and the sampling frequency is K1; for standard zones, the preset interval of its data acquisition network is Q2, and the sampling frequency is K2; where Q1 < Q2, it indicates that the sensor components deployed in the independent zone are more densely packed, and K1 > K2.

[0011] Furthermore, the transmission status monitoring dataset includes at least: vibration acceleration values, strain values, crack width growth rate, spatial displacement of key points in the tunnel, and pore water pressure.

[0012] Furthermore, the operation process of the state assessment strategy is as follows: the current state monitoring data is processed by feature processing, and the obtained feature parameter set is compared with the standard deviation σ of the target monitoring zone. If it exceeds the range of ±3σ, it is marked as a potential anomaly. The number S of potential anomalies marked in the target monitoring zone is counted. When 0 < S < 3, it is classified as a warning; when S = 0, it is classified as a normal; when S > 3, it is classified as an alarm. The feature parameter set includes at least: structural vibration dominant frequency offset, root mean square strain value, crack width growth rate, crown settlement rate, and seepage pressure gradient. An improved lightweight classification model is called to discriminate the multidimensional feature parameters and output the current state probability distribution of the target monitoring zone, including the probability distribution of the three state labels: normal, warning, and alarm. A state judgment threshold is set to determine whether to generate a warning or alarm command.

[0013] Furthermore, the feature parameter set, after normalization, is used as input to the improved lightweight classification model. This model receives feature vector sequences from N consecutive sampling periods using a sliding time window, constructing an N×D two-dimensional input matrix; where D is the feature dimension and N is the time step. Forward inference is used to calculate the probability distribution of the target monitoring zone's state as normal, warning, or alarm. State determination thresholds are set: when the alarm probability ≥ 0.7, an alarm command is generated; when the warning probability ≥ 0.6 and the alarm probability < 0.3, a warning command is generated; all other cases are classified as normal. The feature parameter set is normalized by calculating the standardized offset exponent η for each feature parameter of each monitoring zone. q，j (t).

[0014] Furthermore, the process of selecting whether to trigger the linkage response analysis mechanism in the target monitoring partition based on the status command is as follows: When any target monitoring partition generates an early warning command, the linkage response analysis mechanism is triggered: Let the target monitoring partition be: the first abnormal partition. The process of triggering the linkage response analysis mechanism is as follows: After comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring partition, the quantity exceeding ±3σ is obtained. After weighted calculation, the result is: the first offset comprehensive value. According to the preset mapping table, the number of adjacent partitions K_1 of the first abnormal partition is determined based on the first offset comprehensive value. The linkage analysis domain is formed by taking K_1 / 2 adjacent partitions before and after the target monitoring partition.

[0015] When any target monitoring partition generates an alarm command, the linkage response analysis mechanism is triggered. The target monitoring partition is denoted as the second abnormal partition. The process of the triggered linkage response analysis mechanism is as follows: After comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring partition, the quantity exceeding ±3σ is obtained. After weighted calculation, the number S of potential anomalies marked in the processed target monitoring partition is used as a correction factor. The result is the second offset comprehensive value. According to the preset mapping table, the number K_2 of adjacent partitions of the second abnormal partition is determined based on the second offset comprehensive value. The linkage analysis domain is formed by taking K_2 / 2 adjacent partitions before and after the target monitoring partition.

[0016] Furthermore, the process of performing feature fusion analysis is as follows: Calculate the state difference degree between adjacent target partitions within the linkage analysis domain: In the formula, D(q, q+1) represents the state difference between the target monitoring partition q and its adjacent partition q+1, and k i The weight coefficient of the parameter in the i-th feature parameter set is defined, and its value range is [0, 1]. It represents the Euclidean distance between the target monitoring partition q and its neighboring partition q+1 at N time points on the i-th type of feature parameter set.

[0017] Furthermore, the process of performing joint state assessment is as follows: A spatial consistency index C_s is introduced, and a joint assessment is performed combining state difference and temporal trend. A judgment result is generated according to a preset rule engine, and this judgment result is uploaded to the cloud as the detection result. The spatial consistency index C_s is calculated based on the following formula: C_s = 1 - σ η / μ η In the formula, μ η Represents all η within the linkage domain q，i (t) is the average value in the spatiotemporal dimension; σ η Represents all η within the linkage domain q，i The standard deviation corresponding to (t).

[0018] Furthermore, the judgment results generated by the preset rule engine are based on the following criteria: when only the standardized offset index of the target monitoring partition is abnormal, the adjacent partitions D(q, q+1) in the neighboring area exceed the standard, and the spatial consistency index C_s in the linkage domain does not exceed the degradation threshold, it is judged as a local defect; when the combined state difference D increases in any direction and the standardized offset index of at least two consecutive partitions increases, it is judged as directional spread; when the spatial consistency index C_s exceeds the degradation threshold, it is judged as coordinated deterioration.

[0019] A cloud-edge collaborative tunnel status monitoring system includes:

[0020] Data monitoring and deployment module: Deploy several data acquisition networks and edge computing nodes within the target tunnel, and pre-configure monitoring zones. The edge computing nodes establish communication links with each data acquisition network and are configured with a lightweight data processing engine to transmit the corresponding transmission status monitoring datasets collected within each monitoring zone.

[0021] Linkage Response Analysis Module: Edge computing nodes receive status monitoring datasets, execute status assessment strategies for each monitoring partition according to time series, and obtain status instructions for the corresponding monitoring partitions; based on the status instructions, they select whether to trigger the linkage response analysis mechanism in the target monitoring partition to determine the linkage analysis domain;

[0022] Results evaluation and uploading module: Based on the results of the linkage response analysis mechanism, it performs feature fusion and joint state evaluation, and uploads the monitoring results after judgment.

[0023] This invention provides a method and system for monitoring tunnel status based on cloud-edge collaboration, which has the following beneficial effects:

[0024] (1) This scheme achieves differentiated and precise allocation of monitoring resources by implementing secondary zoning along the tunnel, and improves and adjusts the density and sampling frequency of sensors configured in the main areas to ensure the monitoring sensitivity of high-risk sections; at the same time, it can also avoid the resource waste caused by the traditional uniform and dense deployment along the entire line to a certain extent, and achieve dynamic matching between monitoring investment and structural risk.

[0025] (2) This scheme dynamically determines the number of linkage partitions by offset comprehensive value, realizes adaptive adjustment of monitoring response range, avoids missed detection or redundant calculation caused by fixed neighborhood analysis, and improves edge computing resource utilization efficiency to a certain extent; on the other hand, the introduction of abnormal parameter number correction mechanism can detect the extension of abnormalities to the original un-analyzed partitions, realize the purpose of early capture of the expansion of hidden seepage channels, and enhance the ability to identify some abnormalities.

[0026] (3) This scheme constructs a corresponding matrix in the linkage analysis domain and introduces state difference degree and spatial consistency index for joint evaluation. It not only identifies the abnormal propagation direction by weighted Euclidean distance to effectively determine the direction of directional spread, but also distinguishes the coordinated deterioration caused by local defects and environmental factors by using spatial consistency index to analyze from the perspective of regional evolution, thereby ensuring that the results are more accurate and effective. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the overall steps of the cloud-edge collaborative tunnel status monitoring method in this invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1:

[0030] Please see Figure 1 This embodiment provides a tunnel status monitoring method based on cloud-edge collaboration. This method is for auxiliary status monitoring operations in tunnel construction in the field of civil engineering. The specific steps of this monitoring method are as follows:

[0031] S1. Deploy several data acquisition networks and edge computing nodes within the target tunnel to transmit status monitoring datasets;

[0032] S1.1 Deploy a data acquisition network along the target tunnel at preset intervals and pre-configure monitoring zones. The area formed by all monitoring zones covers the area inside the target tunnel.

[0033] The data acquisition network includes sensor components and wireless communication modules;

[0034] The sensor components include: vibration sensors, strain gauges, crack gauges, displacement sensors, and piezometers, etc.; the wireless communication module is integrated on the sensor components and is used to periodically report data to the edge computing nodes.

[0035] The configuration process for the monitoring partition is as follows:

[0036] S1.1.1. Based on the length L of the target tunnel, divide it into initial partitions at a set equal interval ΔL to form a preliminary partition sequence {P1, P2, ..., Pn}, where n is the number of preliminary monitoring partitions; S1.1.2. Obtain the characteristic segments of the target tunnel and map them to the preliminary partition sequence. If the initial partition contains any characteristic segment, perform a secondary partitioning operation to select the characteristic area as an independent partition and use the remaining partitions as standard partitions to form a secondary partition sequence; otherwise, use the corresponding initial partition directly as a standard partition; finally, each monitoring partition contains several independent partitions and standard partitions.

[0037] For an independent partition, the preset interval of its data acquisition network is Q1, and the sampling frequency is K1;

[0038] For a standard partition, the preset interval of its data acquisition network is Q2, and the sampling frequency is K2;

[0039] Where Q1 < Q2, it indicates that the sensor components deployed in the independent zones are more densely packed, and K1 > K2, it indicates that the sensor components deployed in the independent zones collect signals more times per second; the characteristic section refers to any one or more of the following: geological condition change zone, construction joint zone, and historical disease high-incidence zone; where geological condition change zone refers to the range of ±50m from the interface of the change in surrounding rock grade; construction joint zone includes: the interface between open and cut excavation, the concentrated area of ​​shield assembly ring, and the interface of connecting passage; historical disease high-incidence zone refers to the section where the frequency of cracks and leakage events in the past three years is more than twice the average.

[0040] S1.2 Edge computing nodes are pre-deployed at the entrance and exit of the target tunnel, establish bidirectional communication links with each data acquisition network through industrial Ethernet, and are configured with a lightweight data processing engine to perform real-time data cleaning and outlier removal, and transmit the status monitoring data collected by the sensor components in each monitoring zone.

[0041] The condition monitoring dataset includes: vibration acceleration values, strain values, crack width growth rate, spatial displacement of key tunnel points, and pore water pressure. It should be noted that, in this embodiment, key tunnel points refer to the arch subsidence, and the spatial displacement of key tunnel points is the arch subsidence. The condition monitoring data collected from each monitoring zone are aggregated to the edge computing node to form the condition monitoring dataset.

[0042] This solution, by implementing a secondary zoning mechanism along the tunnel based on key stress points, geological abrupt changes, and historical damage areas, not only achieves differentiated and precise allocation of monitoring resources, but also allows for adjustments to the density and sampling frequency of sensors deployed in key areas to ensure monitoring sensitivity in high-risk sections. Furthermore, it avoids, to some extent, the resource waste caused by the traditional uniform and dense deployment along the entire line, achieving a dynamic match between monitoring investment and structural risk, demonstrating the engineering optimization effect of refined zoning management.

[0043] S2. Perform real-time assessment of monitoring status and initial anomaly judgment, triggering the linkage response analysis mechanism;

[0044] S2.1 The edge computing node receives the status monitoring dataset and executes a status assessment strategy for each monitoring partition according to the time series to obtain the status instructions for the corresponding monitoring partition. The operation process of the status assessment strategy is as follows: S2.1.1 The status monitoring data at the current moment is processed by feature processing. The obtained feature parameter set is compared with the standard deviation σ of the target monitoring partition. If it exceeds ±3σ, it is marked as a potential anomaly. The number S of potential anomalies marked in the target monitoring partition is counted. When 0 < S < 3, it is classified as a warning; when S = 0, it is classified as a normal; when S > 3, it is classified as an alarm. The feature parameter set includes: structural vibration dominant frequency offset, root mean square strain value, crack width growth rate, arch settlement rate, and seepage pressure gradient. S2.1.2 The locally deployed improved lightweight classification model is called to discriminate the multi-dimensional feature parameters and output the current status probability distribution of the target monitoring partition, including the probability distribution of the three status labels: normal, warning, and alarm. A status judgment threshold is set to determine whether to generate a warning or alarm instruction.

[0045] The specific process of S2.1.2 is as follows:

[0046] The feature parameter set, after normalization, is used as input to the lightweight classification model. The lightweight classification model receives a sequence of feature vectors from N consecutive sampling periods using a sliding time window, constructing an N×D two-dimensional input matrix; where D is the feature dimension and N is the time step, ranging from 5 to 10. Forward inference is used to calculate the probability distribution of the target monitoring zone's state (normal, warning, and alarm). State determination thresholds are set: when the alarm probability is ≥0.7, an alarm command is generated; when the warning probability is ≥0.6 and the alarm probability is <0.3, a warning command is generated; all other cases are classified as normal.

[0047] The feature parameter set is normalized as follows: the standardized offset exponent η for each feature parameter in each monitoring zone is calculated. q，i (t)=(|x q，i (t)-μ i |) / σ1 i ; where η q，j (t) represents the standardized offset exponent of the q-th monitoring zone and the i-th type of feature parameter at time t; x q，i (t) represents the current monitored value of the i-th type of feature parameter in the q-th partition; μ i σ1 represents the historical mean of the i-th type of feature parameter in the target monitoring zone; i The historical standard deviation of the i-th type of feature parameter in the target monitoring zone; this formula maps physical quantities of different dimensions to a dimensionless space, eliminates differences in magnitude, and makes the multi-source data contained in the feature parameter set comparable;

[0048] S2.2. Based on the status command, select whether to trigger the linkage response analysis mechanism in the target monitoring zone;

[0049] When any target monitoring zone generates an early warning command, the linkage response analysis mechanism is triggered. The target monitoring zone is denoted as the first abnormal zone. The process of the triggered linkage response analysis mechanism is as follows: After comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring zone, the quantity exceeding ±3σ is obtained. After weighted calculation, the result is the first offset comprehensive value. According to the preset mapping table, the number of adjacent zones K_1 of the first abnormal zone is determined based on the first offset comprehensive value. The linkage analysis domain is formed by taking K_1 / 2 adjacent zones before and after the target monitoring zone.

[0050] The calculation basis for the first offset composite value C1_dev is as follows: In the formula, i represents the category number of each type of data in the feature parameter set, i = 1, 2, ..., N1, where N1 is the number of each type of data in the feature parameter set, and w i The weight coefficient of the parameter in the i-th feature parameter set is defined, and its value range is [0, 1]. This represents the quantity of parameters in the i-th feature parameter set that exceed the ±3σ range; based on the first offset comprehensive value C1_dev, a preset mapping table is queried to determine the number of adjacent partitions K_1 of the linked first abnormal partition: when C1_dev∈(3,6], then K_1=2; when C1_dev∈(6,9], then K_1=4; when C1_dev>9, then K_1=6; the weighted first offset comprehensive value reflects the abnormal intensity, avoiding excessive response caused by false triggering of a single parameter; setting K_1 to an even number aims to ensure spatial symmetry and ensure balanced analysis of the influence range before and after;

[0051] When any target monitoring partition generates an alarm command, the linkage response analysis mechanism is triggered. The target monitoring partition is denoted as the second abnormal partition. The triggered linkage response analysis mechanism process is as follows: Based on the comparison of the feature parameter set with the historical statistical mean and standard deviation of the target monitoring partition, quantities exceeding ±3σ are obtained. After weighted calculation, the number S of potential anomalies marked within the processed target monitoring partition is used as a correction factor, resulting in the second offset comprehensive value. According to a preset mapping table, the number K_2 of adjacent partitions of the second abnormal partition is determined based on the second offset comprehensive value. The linkage analysis domain is formed by taking K_2 / 2 adjacent partitions before and after the target monitoring partition.

[0052] The calculation of the first offset comprehensive value C1_dev involves comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring zone to obtain quantities exceeding ±3σ. The weighted calculation steps are similar to those used in calculating the first offset comprehensive value C1_dev, so they will not be elaborated here. The number S of potential anomalies marked within the processed target monitoring zone is used as the correction factor, based on the formula: g = 1 + u × (max(S-3), 0); where g represents the correction factor, u is the adjustment coefficient, and its value range is [0, 1]. In this embodiment, u can be selected as 0.4. max(S-3, 0) is used to enhance the response intensity of multi-parameter collaborative anomalies. The calculation of the second offset comprehensive value C2_dev is based on: C1_dev = g × C_dev; where C_dev = g × C_dev. `dev` represents the result obtained by weighting the values ​​exceeding ±3σ after comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring partition. The number of adjacent partitions K_2 of the linked second abnormal partition is determined by querying the preset mapping table based on the comprehensive value C2_dev: when C2_dev∈(6, 12], K_2=4; when C2_dev∈(12, 18], K_2=6; when C1_dev>18, K_1=8. The nonlinear correction of S is introduced to reflect that multi-parameter coordinated anomalies are more dangerous than single strong deviations. For example, if S>5, it is highly likely that the structure is deteriorating synchronously across multiple systems, indicating a potential instability precursor. In this case, it is necessary to expand the monitoring range to capture the hidden expansion path.

[0053] When any partition triggers a warning or alarm, the above scheme dynamically expands the linkage analysis domain based on the offset comprehensive value and introduces the number of potential anomalies as a correction factor. It adaptively adjusts the scope of neighboring partitions according to the anomaly intensity to avoid missed detections or redundant calculations caused by fixed neighborhoods.

[0054] By adopting the above technical solution, on the one hand, the number of linked partitions is dynamically determined by the offset comprehensive value, realizing the adaptive adjustment of the monitoring response range, avoiding missed detections or redundant calculations caused by fixed neighborhood analysis, and improving the utilization efficiency of edge computing resources to a certain extent; on the other hand, the introduction of an anomaly parameter number correction mechanism can detect anomalies extending backward to some partitions that were not originally included in the analysis, achieving the purpose of early detection of the expansion of hidden seepage channels, solving the technical problem that the traditional fixed neighborhood linkage mechanism cannot adapt to the complexity of anomaly propagation, and significantly enhancing the system's early identification capability for progressive structural failures.

[0055] S3. Based on the results of the linkage response analysis mechanism, perform feature fusion analysis and joint state assessment, make a judgment, and then upload the monitoring results.

[0056] S3.1 Obtain the linkage analysis domain and perform feature fusion analysis based on the two-dimensional input matrix: Calculate the state difference between adjacent target partitions within the linkage analysis domain. In the formula, D(q, q+1) represents the state difference between the target monitoring partition q and its adjacent partition q+1, and k i This represents the weight coefficient of the parameter in the i-th feature parameter set, with a value range of [0, 1]; k at this point i Compared with the aforementioned w i The values ​​can be different, so different letters are used to represent them; The Euclidean distance between two partitions (i.e., target monitoring partition q and adjacent partition q+1) is the standardized offset exponential sequence at N time points on the i-th type of feature parameter set, reflecting their dynamic response differences; the weighted distance is used to measure the degree of state jump between adjacent partitions.

[0057] S3.2 Execute Joint State Assessment: Introduce the spatial consistency index C_s, combine it with the state difference degree D and the temporal trend for joint assessment, generate the judgment result according to the preset rule engine, and upload the judgment result as the detection result to the cloud; wherein, the spatial consistency index C_s is calculated based on: C_s=1-σ η / μ η In the formula, μ η Represents all η within the linkage domain q，i (t) is the average value in the spatiotemporal dimension; σ η Represents all η within the linkage domain q，i The standard deviation corresponding to (t); this spatial consistency index reflects whether the anomaly is a local isolated event or a systematic response; low C_s indicates a discrete response, which may be a local defect; high C_s combined with a multi-partition degradation trend suggests that environmental factors have caused the overall response;

[0058] Specifically, a corresponding matrix is ​​constructed within the linkage analysis domain, and state difference degree and spatial consistency index are introduced for joint evaluation. This not only identifies the direction of abnormal propagation by weighted Euclidean distance to effectively determine the direction of directional spread, but also distinguishes between local defects and synergistic deterioration caused by environmental factors by using spatial consistency index. This solves the technical deficiency of traditional methods in identifying the spatial correlation of anomalies, and analyzes from the perspective of regional evolution, thereby ensuring more accurate and effective results.

[0059] The judgment results are generated by the preset rule engine based on the following criteria:

[0060] When the standardized offset index of the target monitoring partition is abnormal, the adjacent partition D(q, q+1) in the neighboring area exceeds the standard, and the spatial consistency index C_s in the linkage domain does not exceed the degradation threshold, it is judged as a local defect; when the combined state difference D increases in any direction and the standardized offset index of at least two consecutive partitions increases, it is judged as directional spread; when the spatial consistency index C_s exceeds the degradation threshold, it is judged as coordinated deterioration.

[0061] It should be noted that the abnormal standardized offset index of the target monitoring partition refers to the fact that the standardized offset index of the target monitoring partition exceeds the standard threshold, while the standardized offset index of its adjacent partitions does not exceed the standard threshold. The standard threshold ranges from 0 to 3. In this embodiment, the standard threshold is usually set to 3. The abnormal standardized offset index of the target monitoring partition indicates that the anomaly is concentrated in the target monitoring partition and has not spread to neighboring partitions. The adjacent partition D(q, q+1) exceeding the standard indicates that the adjacent partition D(q, q+1) exceeds the preset threshold D_th. The preset threshold ranges from 0 to 2. In this embodiment, the preset threshold D_th can be set to 2. The spatial consistency index C_s in the linkage domain does not exceed the degradation threshold. The degradation threshold ranges from [0, 1]. In this embodiment, the value can be set to 0.6. When C_s does not exceed the degradation threshold, it indicates that the response differences between partitions are large and there is no coordination, which is consistent with the characteristics of local isolated anomalies.

[0062] Results Description: This solution achieves multi-source data fusion through normalization, identifies propagation paths using weighted difference, and distinguishes between local and overall degradation by leveraging spatial consistency indicators within the linkage domain. On one hand, it improves the accuracy of identifying abnormal spatial evolution trends and solves the problem that traditional fixed neighborhood analysis cannot dynamically adapt to abnormal expansion patterns. On the other hand, by unexpectedly capturing some cases of synchronous degradation in non-contact areas, it can effectively detect hidden environmental factors, thereby enhancing the overall solution's ability to identify complex tunnel anomalies.

[0063] Example 2:

[0064] Based on Example 1, this embodiment also provides a tunnel status monitoring system based on cloud-edge collaboration, including:

[0065] Data monitoring and deployment module: Deploy several data acquisition networks and edge computing nodes within the target tunnel, and pre-configure monitoring zones. The edge computing nodes establish communication links with each data acquisition network and are configured with a lightweight data processing engine to transmit the corresponding transmission status monitoring datasets collected within each monitoring zone.

[0066] Linkage Response Analysis Module: Edge computing nodes receive status monitoring datasets, execute status assessment strategies for each monitoring partition according to time series, and obtain status instructions for the corresponding monitoring partitions; based on the status instructions, they select whether to trigger the linkage response analysis mechanism in the target monitoring partition to determine the linkage analysis domain;

[0067] Results evaluation and uploading module: Based on the results of the linkage response analysis mechanism, it performs feature fusion and joint state evaluation, and uploads the monitoring results after judgment.

[0068] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0069] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0070] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A tunnel status monitoring method based on cloud-edge collaboration, characterized in that, include: Several data acquisition networks and edge computing nodes are deployed within the target tunnel, and monitoring zones are pre-configured. The edge computing nodes establish communication links with each data acquisition network and are configured with a lightweight data processing engine to transmit the corresponding transmission status monitoring datasets collected within each monitoring zone. Edge computing nodes receive status monitoring datasets, execute status assessment strategies for each monitoring partition according to time series, and obtain status instructions for the corresponding monitoring partitions; based on the status instructions, they select whether to trigger the linkage response analysis mechanism in the target monitoring partition to determine the linkage analysis domain. Based on the results of the linkage response analysis mechanism, feature fusion and joint state assessment are performed, and the monitoring results are uploaded after judgment.

2. The tunnel status monitoring method based on cloud-edge collaboration according to claim 1, characterized in that: The data acquisition network includes at least: sensor components and a wireless communication module; the wireless communication module is integrated on the sensor components and is used to periodically report data to the edge computing nodes; The configuration process of the monitoring zones is as follows: Based on the length L of the target tunnel, the tunnel is divided into zones at a set equal interval ΔL to form a preliminary zone sequence {P1, P2, ..., Pn}, where n is the number of preliminary monitoring zones; the characteristic segments of the target tunnel are acquired and mapped to the preliminary zone sequence. If any characteristic segment is contained in the initial zone, a secondary zone operation is performed to select the characteristic area as an independent zone and use the remaining zones as standard zones to form a secondary zone sequence; otherwise, the corresponding initial zone is directly used as a standard zone. For independent zones, the preset interval of its data acquisition network is Q1, and the sampling frequency is K1; for standard zones, the preset interval of its data acquisition network is Q2, and the sampling frequency is K2; where Q1 < Q2, it indicates that the sensor components deployed in the independent zone are more densely packed, and K1 > K2.

3. The tunnel status monitoring method based on cloud-edge collaboration according to claim 1, characterized in that: The transmission status monitoring dataset includes at least: vibration acceleration values, strain values, crack width growth rate, spatial displacement of key points in the tunnel, and pore water pressure.

4. The tunnel status monitoring method based on cloud-edge collaboration according to claim 3, characterized in that: The operation process of the status assessment strategy is as follows: the status monitoring data at the current moment is processed by feature processing, and the obtained feature parameter set is compared with the standard deviation σ of the target monitoring partition. If it exceeds the range of ±3σ, it is marked as a potential anomaly. The number S of marked potential anomalies in the target monitoring partition is counted. When 0 < S < 3, it is regarded as an early warning class. When S=0, it is classified as normal; when S>3, it is classified as alarm. The feature parameter set includes at least: structural vibration dominant frequency offset, root mean square strain value, crack width growth rate, crown settlement rate, and seepage pressure gradient. An improved lightweight classification model is called to discriminate the multidimensional feature parameters and output the current state probability distribution of the target monitoring zone, including the probability distribution of the three state labels: normal, warning, and alarm. A state judgment threshold is set to determine whether to generate a warning or alarm command.

5. The tunnel status monitoring method based on cloud-edge collaboration according to claim 4, characterized in that: The normalized feature parameter set is used as input to the improved lightweight classification model. This model receives feature vector sequences from N consecutive sampling periods using a sliding time window, constructing an N×D two-dimensional input matrix; where D is the feature dimension and N is the time step. Forward inference is used to calculate the probability distribution of the target monitoring zone's state (normal, warning, and alarm). State determination thresholds are set: an alarm command is generated when the alarm probability ≥ 0.7; a warning command is generated when the warning probability ≥ 0.6 and the alarm probability < 0.3; all other cases are classified as normal. The normalized feature parameter set is calculated by calculating the standardized offset exponent η for each feature parameter of each monitoring zone. q，j (t).

6. The tunnel status monitoring method based on cloud-edge collaboration according to claim 5, characterized in that: The process of selecting whether to trigger the linkage response analysis mechanism in the target monitoring zone based on the status command is as follows: When any target monitoring zone generates an early warning command, the linkage response analysis mechanism is triggered. The target monitoring zone is denoted as the first abnormal zone. The process of the triggered linkage response analysis mechanism is as follows: After comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring zone, the quantity exceeding ±3σ is obtained. After weighted calculation, the result is the first offset comprehensive value. According to the preset mapping table, the number of adjacent zones K_1 of the first abnormal zone is determined based on the first offset comprehensive value. The linkage analysis domain is formed by taking K_1 / 2 adjacent zones before and after the target monitoring zone. When any target monitoring partition generates an alarm command, the linkage response analysis mechanism is triggered. The target monitoring partition is denoted as the second abnormal partition. The process of the triggered linkage response analysis mechanism is as follows: After comparing the feature parameter set with the historical statistical mean and standard deviation of the target monitoring partition, the quantity exceeding ±3σ is obtained. After weighted calculation, the number S of potential anomalies marked in the processed target monitoring partition is used as a correction factor. The result is the second offset comprehensive value. According to the preset mapping table, the number K_2 of adjacent partitions of the second abnormal partition is determined based on the second offset comprehensive value. The linkage analysis domain is formed by taking K_2 / 2 adjacent partitions before and after the target monitoring partition.

7. The tunnel status monitoring method based on cloud-edge collaboration according to claim 1, characterized in that: The process of performing feature fusion analysis is as follows: Calculate the state difference between adjacent target partitions within the linkage analysis domain: In the formula, D(q, q+1) represents the state difference between the target monitoring partition q and its adjacent partition q+1, and k i The weight coefficient of the parameter in the i-th feature parameter set is defined, and its value range is [0, 1]. It represents the Euclidean distance between the target monitoring partition q and its neighboring partition q+1 at N time points on the i-th type of feature parameter set.

8. The tunnel status monitoring method based on cloud-edge collaboration according to claim 5, characterized in that: The process of performing joint state assessment is as follows: A spatial consistency index C_s is introduced, and a joint assessment is performed combining state differences and temporal trends. A judgment result is generated according to a pre-defined rule engine, and this judgment result is uploaded to the cloud as the detection result. The spatial consistency index C_s is calculated as follows: C_s = 1 - σ η / μ η In the formula, μ η Represents all η within the linkage domain q，i (t) is the average value in the spatiotemporal dimension; σ η Represents all η within the linkage domain q，i The standard deviation corresponding to (t).

9. The tunnel status monitoring method based on cloud-edge collaboration according to claim 8, characterized in that: The judgment results generated by the preset rule engine are based on the following criteria: when only the standardized offset index of the target monitoring partition is abnormal, the adjacent partitions D(q, q+1) in the neighboring area exceed the standard, and the spatial consistency index C_s in the linkage domain does not exceed the degradation threshold, it is judged as a local defect; when the combined state difference D increases in any direction and the standardized offset index of at least two consecutive partitions increases, it is judged as directional spread; when the spatial consistency index C_s exceeds the degradation threshold, it is judged as coordinated deterioration.

10. A tunnel status monitoring system based on cloud-edge collaboration, characterized in that, include: Data monitoring and deployment module: Deploy several data acquisition networks and edge computing nodes within the target tunnel, and pre-configure monitoring zones. The edge computing nodes establish communication links with each data acquisition network and are configured with a lightweight data processing engine to transmit the corresponding transmission status monitoring datasets collected within each monitoring zone. Linkage Response Analysis Module: Edge computing nodes receive status monitoring datasets, execute status assessment strategies for each monitoring partition according to time series, and obtain status instructions for the corresponding monitoring partitions; based on the status instructions, they select whether to trigger the linkage response analysis mechanism in the target monitoring partition to determine the linkage analysis domain; Results evaluation and uploading module: Based on the results of the linkage response analysis mechanism, it performs feature fusion and joint state evaluation, and uploads the monitoring results after judgment.

Images