A tunnel structure disease identification and safety risk early warning method and system

Abstract

This invention belongs to the field of tunnel structural health monitoring technology and discloses a method and system for identifying tunnel structural defects and providing early warning of safety risks. The method collects surface image data, structural status monitoring data, and environmental operation data of the tunnel lining, and constructs an evaluation object dataset according to tunnel station number, monitoring section, and circumferential location of the lining. It extracts surface defect characteristic parameters such as cracks, leakage, and spalling from the surface image data, extracts structural response characteristic parameters from the structural status monitoring data, and combines historical time-series data to obtain the defect development trend and structural response change trend. It obtains the surface damage level through a defect level rule base, obtains the physical safety level through a physical assessment model, and then determines the current safety level through level envelope fusion. Based on the current safety level, development trend, environmental operation data, circumferential location of the lining, and the relationship between multiple defects, it determines the risk level and generates graded early warning information.

Description

Technical Field

[0001] This invention belongs to the field of tunnel structure health monitoring technology, specifically a method and system for identifying tunnel structure defects and providing early warning of safety risks. Background Technology

[0002] During long-term operation, tunnel linings are continuously subjected to the effects of surrounding rock pressure, groundwater, traffic loads, temperature and humidity changes, and vibration disturbances, which can easily lead to surface defects such as cracks, leakage, and spalling. These defects are accompanied by structural responses such as increased convergence deformation, abnormal stress and strain, and changes in seepage pressure. Current operation and maintenance typically obtain relevant data through manual inspection, image acquisition and recognition, convergence deformation monitoring, stress and strain monitoring, seepage pressure monitoring, and vibration monitoring. Then, based on individual indicator thresholds, manual experience, or statistical results, defects are recorded and graded for early warning.

[0003] Existing image-based tunnel defect detection technologies can identify defects such as cracks, leaks, and spalling from lining surface images and extract geometric parameters such as crack width, crack length, leakage area, and spalling area. However, their judgment objects are mainly focused on the surface morphology of the lining, which is difficult to reflect the actual stress state and bearing capacity changes of the lining at the same location. Existing sensor-based structural health monitoring technologies can acquire data such as convergence deformation, stress and strain, seepage pressure, and vibration, and can indicate abnormalities based on monitoring thresholds. However, such data is usually managed according to measurement points or cross sections, and there is a lack of a unified evaluation object between it and the specific defect areas obtained by image recognition, making it difficult to establish a correspondence between apparent damage, structural response, and safety level.

[0004] While some multi-source data risk identification methods incorporate image, sensor, and environmental data, they often employ general fusion models, empirical thresholds, or statistical scoring methods. The data processing process tends to focus on output results, lacking a clear judgment chain between apparent disease levels, structural safety levels, environmental operational effects, and the relationships between multiple disease combinations. When the same lining section exhibits severe apparent disease but the structural response is not yet obvious, or when the structural response is abnormal but the apparent disease is not prominent, existing methods are prone to problems such as inconsistent evaluation objects, scattered early warning basis, and unclear sources of risk levels. It is difficult to trace the final risk result back to specific disease parameters, structural response parameters, and environmental operational effect data, and it is also inconvenient to verify the early warning results during subsequent maintenance and treatment. Summary of the Invention

[0005] The technical problem this invention aims to solve is that in existing tunnel structural defect monitoring, apparent image detection, structural condition monitoring, and risk warning are usually processed separately. It is difficult to establish a correspondence between apparent defect information such as cracks, leaks, and spalling and structural response information such as lining deformation, stress-strain, and seepage pressure at the same monitoring location. This results in a fragmented basis for safety assessment, and the warning results are prone to remain at the level of a single threshold judgment or a single defect identification. In particular, when the degree of apparent defect and the degree of structural response anomaly are inconsistent at the same location, existing methods cannot simultaneously reflect the appearance of the defect and the structural bearing capacity, affecting the risk assessment in tunnel operation and maintenance.

[0006] To address the aforementioned problems, this invention provides a method for identifying tunnel structural defects and providing early warning of safety risks. This method collects surface image data, structural condition monitoring data, and environmental operational data of the tunnel lining. The structural condition monitoring data includes convergence deformation data and at least one of stress-strain data and seepage pressure data. Subsequently, evaluation objects are determined according to the tunnel station number, monitoring section, and circumferential location of the lining. The surface image data, structural condition monitoring data, and environmental operational data corresponding to the same evaluation object are associated to form an evaluation object dataset, ensuring that subsequent defect identification, structural response analysis, grade evaluation, and risk warning are all performed around the same evaluation object.

[0007] After forming the evaluation object dataset, the apparent disease characteristic parameters of at least one type of apparent disease, such as cracks, leakage, and spalling, are extracted from the apparent image data in the evaluation object dataset. The structural response characteristic parameters are extracted from the structural condition monitoring data in the evaluation object dataset. The apparent disease characteristic parameters are used to characterize the scale, distribution, and area proportion of the lining surface disease, while the structural response characteristic parameters are used to characterize the lining deformation and the collected stress-strain and seepage pressure changes. Then, the historical time series data of the same evaluation object are called to obtain the disease development trend and the structural response change trend, so that the risk assessment has both current location data and historical change data as the basis.

[0008] After the apparent disease characteristic parameters are input into the disease level rule base, the disease level rule base outputs the corresponding single disease level, and the apparent damage level is determined by the single disease level. After the structural response characteristic parameters are input into the physical assessment model, the physical assessment model determines the actual stress state of the tunnel lining, and compares the actual stress state with the tunnel lining bearing capacity envelope to obtain the physical safety level. The apparent damage level reflects the severity of the disease morphology and distribution, and the physical safety level reflects the relationship between the lining structural response and bearing capacity. The two form evaluation results separately and then are fused together.

[0009] After obtaining the apparent damage level and physical safety level, the two are fused using a level envelope, and the level with the higher risk level is taken as the current safety level. Then, the risk level is determined based on the current safety level, the trend of disease development, the trend of structural response change, environmental operation data, the circumferential location of the evaluation object in the lining, and the relationship of multiple disease combinations, and graded early warning information is generated. Thus, in cases where the apparent disease is relatively minor but the structural response is abnormal, the apparent disease is obvious but the structural response has not yet reached a high risk level, or there are multiple disease combinations in the same evaluation object, a risk judgment result can be formed based on the data source under the same evaluation object.

[0010] Furthermore, the circumferential lining components include the arch crown, arch waist, sidewalls, and invert arch. Each evaluation object is assigned a unique object number, which is used to associate the apparent defect characteristic parameters, structural response characteristic parameters, historical time series data, risk level, and graded early warning information of the same evaluation object.

[0011] Furthermore, the extraction process of apparent disease feature parameters includes: converting the apparent image data to grayscale, histogram equalization, filtering and denoising, and performing illumination correction to obtain a preprocessed image; performing pixel-level classification on the preprocessed image to obtain a mask image of the disease region corresponding to the identified apparent diseases; performing morphological processing on the mask image and extracting the contour of the disease region based on connected component labels; when the identified apparent diseases include cracks, performing skeletonization processing on the crack region, measuring the crack width distribution along the crack centerline normal to obtain the maximum crack width and the average crack width. The total crack length and crack line density are calculated as follows: when the identified apparent defects include leakage, the pixel area of ​​the leakage area is converted to obtain the leakage area and the leakage area percentage; when the identified apparent defects include peeling, the pixel area of ​​the peeling area is converted to obtain the peeling area and the peeling area percentage. The crack line density is the ratio of the total crack length to the area of ​​the corresponding detection area of ​​the evaluation object; the leakage area percentage is the ratio of the leakage area to the area of ​​the corresponding detection area of ​​the evaluation object; and the peeling area percentage is the ratio of the peeling area to the area of ​​the corresponding detection area of ​​the evaluation object.

[0012] Furthermore, the extraction process of structural response characteristic parameters includes: performing timestamp correction, outlier removal, filtering and noise reduction, and missing value interpolation on the structural condition monitoring data; obtaining the convergence deformation amount and convergence deformation rate based on the processed convergence deformation data; when the structural condition monitoring data includes stress and strain data, obtaining the stress and strain change amount based on the processed stress and strain data; when the structural condition monitoring data includes seepage pressure data, obtaining the seepage pressure change amount and seepage pressure change trend based on the processed seepage pressure data.

[0013] Furthermore, historical time-series data includes apparent disease characteristic parameters and structural response characteristic parameters formed by the same evaluation object in different evaluation periods. Disease development trends include at least one of the following: the rate of change of maximum crack width, the rate of change of total crack length, the rate of change of leakage area, and the rate of change of spalling area. Structural response change trends include at least one of the following: the rate of convergence deformation, the rate of change of stress and strain, and the rate of change of seepage pressure. Each rate of change is determined according to the ratio of the parameter difference between adjacent evaluation periods of the same evaluation object to the time interval of the evaluation period.

[0014] Furthermore, the disease severity rule base stores the rules for determining the severity of a single disease using production rules. The crack severity is determined based on the maximum crack width, total crack length, and crack line density; the leakage severity is determined based on the leakage area and the proportion of the leakage area; and the spalling severity is determined based on the spalling area and the proportion of the spalling area. When there are two or more single disease severity levels for the same evaluation object, the single disease severity level with the highest risk is determined as the apparent damage level. When there is only one single disease severity level for the same evaluation object, that single disease severity level is determined as the apparent damage level.

[0015] Furthermore, when the structural condition monitoring data includes convergence deformation data, stress-strain data, and seepage pressure data, the physical assessment model determines the bending moment and axial force of the tunnel lining section based on the surrounding rock pressure, water pressure, convergence deformation, stress-strain changes, and seepage pressure changes. It then determines the physical safety factor based on the axial force-bending moment correlation envelope and determines the physical safety level based on the level range of the physical safety factor. The physical safety factor is the ratio of the ultimate bending moment under the corresponding axial force on the bearing capacity envelope to the actual bending moment.

[0016] Furthermore, the risk level is determined using a multi-factor scoring method. The scoring factors include the current safety level, the trend of disease development, the trend of structural response change, the location of the evaluated object within the lining circumferentially, environmental operational data, and the relationship between multiple diseases. Specifically, the current safety level is obtained by fusion of level envelopes; the trends of disease development and structural response change are obtained from historical time-series data; the location of the evaluated object within the lining circumferentially is determined by the evaluated object itself; the relationship between multiple diseases is obtained from a single disease level; and the environmental operational data is based on collected data from rainfall intensity, traffic load, temperature and humidity changes, and abnormal vibration events to form an environmental operational score. When the crack level and leakage level of the same evaluated object both reach level two or above, or when the crack level and spalling level both reach level two or above, the risk level based on the scoring factors will be upgraded by one level. When the crack level, leakage level, and spalling level of the same evaluated object all reach level two or above, the risk level will be classified as extremely high risk.

[0017] The present invention also provides a tunnel structure defect identification and safety risk early warning system, including a data acquisition module, an evaluation object construction module, an apparent defect feature extraction module, a structural response feature extraction module, a trend analysis module, a defect level evaluation module, a physical model evaluation module, a level envelope fusion module, and a risk early warning module.

[0018] The data acquisition module collects visual image data, structural condition monitoring data, and environmental operation data of the tunnel lining. The evaluation object construction module determines the evaluation object according to the tunnel station, monitoring section, and circumferential location of the lining, and associates the visual image data, structural condition monitoring data, and environmental operation data of the same evaluation object into an evaluation object dataset. The apparent defect feature extraction module outputs apparent defect feature parameters from the visual image data in the evaluation object dataset. The structural response feature extraction module outputs structural response feature parameters from the structural condition monitoring data in the evaluation object dataset. The trend analysis module outputs defect data based on the historical time series data of the evaluation object. The module for evaluating disease development trends and structural response changes is used to output single disease levels and apparent damage levels based on apparent disease characteristic parameters. The module for evaluating physical models is used to output physical safety levels based on structural response characteristic parameters. The module for fusion of levels and envelopes is used to output the current safety level based on apparent damage levels and physical safety levels. The module for risk warning is connected to the module for fusion of levels and envelopes, the module for trend analysis, and the module for constructing evaluation objects. It is used to receive current safety levels, disease development trends, structural response change trends, environmental operation data, the circumferential location of the evaluation object in the lining, and the relationship between multiple disease combinations, and output risk levels and graded warning information.

[0019] Furthermore, the data acquisition module includes an image acquisition unit, a convergence deformation monitoring unit, a stress-strain monitoring unit, a seepage pressure monitoring unit, and an environmental operation effect acquisition unit. The system also includes an edge computing node and a cloud server. The edge computing node is used to preprocess the apparent image data and structural status monitoring data, and to make threshold early warning judgments for at least one of the following: maximum crack width, convergence deformation rate, and leakage area. The cloud server is used to perform trend analysis, disease level evaluation, physical model evaluation, level envelope fusion, and risk level determination. When the early warning level output by the edge computing node is inconsistent with the early warning level output by the cloud server, the risk early warning module adopts the higher level of the two as the final early warning level.

[0020] The beneficial effects of this invention are as follows:

[0021] 1. This invention uses tunnel station number, monitoring section, and circumferential lining location to jointly determine the evaluation object, and associates the apparent image data, structural status monitoring data, and environmental operation data of the same evaluation object into an evaluation object dataset. This allows apparent defect parameters such as cracks, leakage, and spalling to be processed in the same spatial location and within the same evaluation period as structural response parameters such as convergence deformation, stress-strain, and seepage pressure. This avoids the inconsistency problem caused by the separate storage and judgment of existing image inspection data, sensor monitoring data, and environmental data. It provides a clear data source and location basis for subsequent defect level determination, physical safety assessment, and risk warning, facilitating the traceability and verification of warning results.

[0022] 2. This invention obtains a single disease level and an apparent damage level by inputting apparent disease characteristic parameters into a disease level rule base; simultaneously, it inputs structural response characteristic parameters into a physical assessment model to determine the actual stress state of the lining based on surrounding rock pressure, water pressure, convergence deformation, stress-strain changes, and seepage pressure changes, and obtains the physical safety level through the axial force-bending moment correlation envelope. The above processing allows for the separate evaluation of the severity of apparent disease and the structural bearing safety status, and then determines the current safety level through level envelope fusion. Compared with methods that rely solely on apparent image recognition results or solely on deformation and strain threshold alarms, this invention can determine the current safety status based on the side with higher risk when apparent disease is obvious but the structural response has not yet become abnormal, or when the structural response is abnormal but the apparent disease is not prominent, thus reducing the risk of missed judgments and underestimation caused by a single data source.

[0023] 3. This invention further incorporates disease development trends, structural response change trends, environmental operation data, lining circumferential locations, and multi-disease combination relationships into the risk level determination process. It generates graded early warning information through multi-factor scoring, risk intervals, and combined disease correction rules. Disease development trends and structural response change trends reflect the changing direction of the same evaluation object within different evaluation periods. Environmental operation data reflects the impact of rainfall, traffic load, temperature and humidity changes, and abnormal vibrations on the evaluation object. Multi-disease combination relationships can correct for risks associated with cracks and leakage, cracks and spalling, and concurrent conditions of these three types of diseases. Therefore, the final risk level is no longer determined solely by a single detection value, but rather by the current safety level, development trends, location differences, and concurrent diseases. This provides the graded early warning information with clear triggering parameters, judgment rules, and handling criteria. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the overall process of tunnel structure defect identification and safety risk early warning in this invention.

[0025] Figure 2 This is a flowchart of the dual-channel security assessment sub-process of the present invention;

[0026] Figure 3 This is a data flow diagram of the hardware and software modules of the system of the present invention;

[0027] Figure 4 This is a schematic diagram illustrating the process of multi-source data acquisition, dual-channel security assessment, and hierarchical early warning 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] like Figures 1 to 4 As shown, this embodiment provides a method and system for identifying tunnel structural defects and providing early warning of safety risks. It is applicable to highway tunnels, railway tunnels, urban rail transit tunnels, and underground passages with lining structures during operation. The tunnel lining includes secondary lining and segment lining. This embodiment uses a highway tunnel during operation as an example for illustration. Other tunnel structures can be implemented according to the same data association, defect identification, structural response analysis, grade evaluation, and risk warning process.

[0030] In this embodiment, the tunnel is divided into several monitoring sections along the driving direction. Each monitoring section has a corresponding tunnel chainage range. At least one monitoring section is set in each monitoring section. Each monitoring section is divided into arch crown, arch waist, sidewall, and invert according to the circumferential part of the lining. An evaluation object is determined by the tunnel chainage, the monitoring section, and the circumferential part of the lining. For example, the arch waist of the first monitoring section at chainage K12+350 is one evaluation object, and the arch crown of the first monitoring section at chainage K12+350 is another evaluation object. Each evaluation object is assigned a unique object number. This unique object number is used to associate the appearance image data, structural status monitoring data, environmental operation data, historical time series data, risk level, and graded early warning information of the same evaluation object.

[0031] In this embodiment, the evaluation object dataset is stored according to a unified field. Each data record includes at least the object number, tunnel station number, monitoring section number, lining circumferential location, acquisition time, data type, acquisition equipment number, and data value. After the apparent image data, structural status monitoring data, and environmental operation effect data are written into the system, the evaluation object construction module first performs spatial matching according to the tunnel station number, monitoring section number, and lining circumferential location, and then performs temporal matching according to the evaluation period to which the acquisition time belongs. When both spatial and temporal matching are satisfied, the corresponding data are written into the same evaluation object dataset. Thus, subsequent apparent disease characteristic parameters, structural response characteristic parameters, historical time series data, risk level, and graded early warning information can all be traced back to the same evaluation object through the object number.

[0032] Object Number Evaluation Object Construction Module Data and results related to the same evaluation object Tunnel chainage On-site location information Determine the longitudinal position of the evaluation object Monitoring section number Monitoring deployment information Determine the cross section where the evaluation object is located. Lining circumferential section vault, arch, side wall, inverted arch Determine the circumferential position of the evaluation object Collection time Each acquisition unit Formation of evaluation cycle and historical time series data Data types Data acquisition module Distinguish between apparent images, convergent deformation, stress-strain, osmotic pressure, and environmental operational data. Data acquisition device number Each acquisition unit Tracing the source of data collection Data value Each acquisition unit or analysis module Store raw data, feature parameters, and evaluation results.

[0033] In this embodiment, the data acquisition module includes an image acquisition unit, a convergence deformation monitoring unit, a stress-strain monitoring unit, a seepage pressure monitoring unit, a vibration acceleration monitoring unit, and an environmental operation effect acquisition unit. The image acquisition unit is installed at least at one location on the tunnel sidewall, arch waist, and arch crown to acquire apparent image data of the lining area corresponding to the evaluation object. The image acquisition unit has a protection rating of at least IP65 to adapt to the humid, dusty, and water vapor environment inside the tunnel. The convergence deformation monitoring unit is located at the arch waist on both sides of the monitoring section to acquire convergence deformation data of the monitoring section where the evaluation object is located, with a measurement accuracy of at least 0.1m. m; The stress-strain monitoring unit adopts one of the following: vibrating wire strain gauge and fiber optic strain gauge, which is firmly installed on the inner wall of the tunnel lining to acquire lining stress-strain data; The seepage pressure monitoring unit adopts one of the following: vibrating wire piezometer, silicon pressure piezometer, and fiber optic piezometer, which is buried in a borehole 1-5m outside the tunnel lining, with at least 2 units arranged in each monitoring section; The vibration acceleration monitoring unit is rigidly installed on the tunnel arch or arch waist, with a sampling frequency of not less than 100Hz; The environmental operation effect acquisition unit is used to acquire at least one of the following: temperature and humidity data, rainfall intensity data, traffic load data, and abnormal vibration event data.

[0034] Each data acquisition unit collects data according to the set evaluation cycle; after collection, all types of data are written to the same time reference, and the data package records the acquisition time, tunnel station number, monitoring section, lining circumferential location, and equipment number; the evaluation object construction module matches the tunnel station number, monitoring section, and lining circumferential location, and associates the appearance image data, structural status monitoring data, and environmental operation effect data of the same evaluation object into an evaluation object dataset; subsequent appearance defect identification, structural response analysis, trend analysis, grade evaluation, and risk warning are all based on this evaluation object dataset.

[0035] Before being put into use, the image acquisition unit calibrates each fixed acquisition field of view. During calibration, a calibration reference of known size is set on the lining surface to obtain the conversion coefficient between pixel length and actual length. For fixed image acquisition units, the conversion coefficient is bound to the object number for storage. For mobile inspection acquisition methods, the conversion coefficient is determined based on the acquisition distance, lens parameters, and calibration reference.

[0036] Crack width, crack length, leakage area, and spalling area are all converted from pixel quantity to actual quantity according to the conversion factor; the area of ​​the detection area corresponding to the evaluation object is the effective area of ​​the lining corresponding to the evaluation object in the appearance image, which is converted by the same conversion factor.

[0037] The apparent disease feature extraction module receives apparent image data from the evaluation object dataset and performs grayscale conversion, histogram equalization, Gaussian filtering for noise reduction, and illumination correction on the apparent image data to obtain a preprocessed image. Grayscale conversion is used to unify image channels, histogram equalization is used to enhance the contrast of the lining surface texture, Gaussian filtering for noise reduction is used to reduce acquisition noise, and illumination correction is used to reduce the impact of uneven tunnel lighting on image segmentation.

[0038] Subsequently, threshold segmentation is performed on the preprocessed image to perform pixel-level classification, resulting in a binary mask image of the disease region corresponding to the identified apparent diseases. The apparent diseases include at least one type among cracks, seepage, and peeling. Morphological processing is performed on the binary mask image, including erosion, dilation, opening, and closing operations, to remove isolated noise points, fill small holes, and connect discontinuous disease regions. After connected component labeling, the contour and location of each disease region are obtained from the processed binary mask image.

[0039] When the identified apparent defects include cracks, the crack area is processed into a skeleton to obtain the crack centerline; the crack width distribution is measured along the normal direction of the crack centerline and converted into the actual width according to the pixel calibration coefficient to obtain the maximum crack width, the average crack width, and the total crack length; the crack line density is the ratio of the total crack length to the area of ​​the corresponding detection area of ​​the evaluation object; the crack tortuosity can also be calculated, which is the ratio of the actual crack length to the straight distance between the beginning and end of the crack.

[0040] When the identified apparent defects include leakage, the pixel area of ​​the leakage area is statistically analyzed, and the leakage area is calculated based on the pixel calibration coefficient. The leakage area ratio is the ratio of the leakage area to the area of ​​the corresponding detection area of ​​the evaluation object. The length, width, equivalent diameter, and circularity of the circumscribed rectangle of the leakage area can also be calculated. The circularity is determined based on the relationship between the area of ​​the leakage area and the square of the perimeter of the leakage area.

[0041] When the identified apparent defects include peeling, the pixel area of ​​the peeling area is statistically analyzed and converted to obtain the peeling area. The peeling area ratio is the ratio of the peeling area to the area of ​​the corresponding detection area of ​​the evaluation object.

[0042] It can also calculate the length, width, equivalent diameter, and roundness of the circumscribed rectangle of the peeling area; the tortuosity, roundness, and equivalent diameter mentioned above are stored as auxiliary morphological parameters in the apparent disease characteristic parameter table for subsequent use in the disease level rule base.

[0043] The structural response feature extraction module receives structural state monitoring data from the evaluation object dataset. This data includes convergent deformation data and at least one of the following: stress-strain data and seepage pressure data. The structural state monitoring data undergoes timestamp correction, outlier removal, filtering and noise reduction, and missing value interpolation. Timestamp correction aligns data from different acquisition devices to the same evaluation period. Outlier removal eliminates data from acquisition failures, communication anomalies, and data exceeding the sensor's range. Filtering and noise reduction reduce environmental noise and transient interference. Missing value interpolation handles short-term data loss, employing either interpolation of data from adjacent evaluation periods or interpolation of historical data from the same evaluation object. After interpolation, structural state monitoring data is generated for calculation in the current evaluation period.

[0044] Based on the processed convergence deformation data, the convergence deformation amount and convergence deformation rate are calculated. The convergence deformation amount is the difference between the convergence measurement value in the current evaluation period and the initial benchmark value, and the convergence deformation rate is the ratio of the difference in convergence deformation amount between adjacent evaluation periods to the time interval between evaluation periods. When the structural condition monitoring data includes stress and strain data, the stress and strain change is obtained based on the processed stress and strain data, and the stress and strain change is the difference between the current stress and strain measurement value and the initial benchmark value.

[0045] When structural condition monitoring data includes seepage pressure data, the seepage pressure change and seepage pressure change trend are obtained from the processed seepage pressure data. The seepage pressure change is the difference between the current seepage pressure measurement and the baseline seepage pressure value, and the seepage pressure change trend is the ratio of the difference in seepage pressure change between adjacent evaluation periods to the time interval of the evaluation period. When structural condition monitoring data includes vibration acceleration data, the vibration peak value, vibration duration, and abnormal vibration event identifier are obtained from the processed vibration acceleration data, which serve as inputs for environmental operation performance scoring or rapid early warning judgment of edge computing nodes.

[0046] The trend analysis module calls the historical time series data of the evaluation object; the historical time series data includes the apparent disease characteristic parameters and structural response characteristic parameters formed by the same evaluation object in different evaluation periods; the disease development trend includes at least one of the following: the rate of change of maximum crack width, the rate of change of total crack length, the rate of change of leakage area, and the rate of change of spalling area; the structural response change trend includes at least one of the following: the rate of convergence deformation, the rate of change of stress and strain, and the rate of change of seepage pressure; each change rate is determined according to the ratio of the parameter difference between adjacent evaluation periods of the same evaluation object to the time interval of the evaluation period.

[0047] When performing trend prediction, the trend analysis module can use linear regression, exponential smoothing, or differential autoregressive moving average models. Linear regression is used when the changes in disease parameters are approximately linear. Exponential smoothing is used for data that fluctuates but does not have obvious periodicity. The differential autoregressive moving average model is used for monitoring data with autocorrelation. For evaluation objects that need to be predicted to reach the warning value, when the rate of change is greater than 0, the time window for reaching the preset warning value is obtained by dividing the difference between the preset warning value and the current parameter value by the rate of change. When the rate of change is less than or equal to 0, it is recorded as no development trend within the current evaluation period.

[0048] The disease severity assessment module receives apparent disease characteristic parameters and calls the disease severity rule base to determine the severity of a single disease. The disease severity rule base stores the severity determination rules for a single disease in production rules. Each production rule includes a rule number, disease type, applicable circumferential lining location, input parameter name, first threshold interval, second threshold interval, third threshold interval, fourth threshold interval, output level, and record field. The output level includes level 1, level 2, level 3, and level 4, with the risk level increasing sequentially.

[0049] The crack grade is determined based on the maximum crack width, total crack length, and crack line density; the leakage grade is determined based on the leakage area and the percentage of leakage area; the spalling grade is determined based on the spalling area and the percentage of spalling area. Taking crack grade determination as an example, when the maximum crack width, total crack length, and crack line density are all within the first threshold range, the crack grade is output as 1; when any of these parameters enters the second threshold range, the crack grade is output as 2; when any of these parameters enters the third threshold range, the crack grade is output as 3; when any of these parameters enters the fourth threshold range, the crack grade is output as 4.

[0050] Leakage level and spalling level are determined according to the same rules, based on the corresponding area and area ratio. Each threshold range is pre-stored in the disease level rule library. The threshold range is determined by tunnel design data, operation and maintenance procedures and historical inspection records.

[0051] When the disease level rule base outputs a single disease level, it simultaneously records the input parameters, hit rules, output level, and evaluation time. When there are more than two single disease levels for the same evaluation object, the single disease level with the highest risk is determined as the apparent damage level. When there is only one single disease level for the same evaluation object, that single disease level is determined as the apparent damage level.

[0052] The physical model evaluation module receives structural response characteristic parameters and calls the physical assessment model to output the physical safety level. The parameters required by the physical assessment model include surrounding rock pressure, water pressure, convergence deformation, stress-strain change, seepage pressure change, lining cross-sectional dimensions, and material strength parameters. The surrounding rock pressure is determined by the surrounding rock grade, burial depth, and cross-sectional dimensions in the tunnel design data. The water pressure is determined by the seepage pressure data collected by the seepage pressure monitoring unit. The convergence deformation is determined by the data collected by the convergence deformation monitoring unit. The stress-strain change is determined by the data collected by the stress-strain monitoring unit. The lining cross-sectional dimensions and material strength parameters are determined by the tunnel design data, as-built data, or maintenance records.

[0053] When the structural condition monitoring data includes convergence deformation data, stress-strain data, and seepage pressure data, the physical assessment model determines the bending moment and axial force of the tunnel lining section based on the surrounding rock pressure, water pressure, convergence deformation, stress-strain change, and seepage pressure change. The surrounding rock pressure is used to determine the external load on the lining, the water pressure is used to determine the external water action on the lining, the convergence deformation is used to correct the stress state of the lining, the stress-strain change is used to reflect the stress change of the lining section, and the seepage pressure change is used to reflect the water pressure change. After calling the above parameters, the physical assessment model generates the actual axial force and actual bending moment corresponding to the evaluation object and compares them with the bearing capacity envelope.

[0054] In one specific calculation method, the physical evaluation model adopts an elastic foundation circular ring model, discretizing the tunnel lining along the circumference into 36 beam elements of equal length, with each node having 2 translational degrees of freedom and 1 rotational degree of freedom. Foundation springs are applied to each node in the normal and tangential directions, and the foundation spring coefficient k is determined based on the surrounding rock grade and foundation stiffness value. When measured foundation stiffness values ​​are unavailable, values ​​are obtained from a table corresponding to the elastic resistance coefficients of the surrounding rock grade in the "Highway Tunnel Design Code." Taking Grade 4 surrounding rock as an example, the foundation spring coefficient k is taken as 150 MPa / m, the unit weight of the surrounding rock is taken as 22 kN / m³, and the lateral pressure coefficient is taken as 0.3. The physical evaluation model solves for the bending moment and axial force of each section based on the tunnel burial depth, surrounding rock pressure, water pressure, and structural response characteristic parameters.

[0055] The physical assessment model determines the physical safety factor based on the axial force-bending moment correlation envelope. The axial force-bending moment correlation envelope represents the ultimate bending moment of the lining section under different axial forces. The physical safety factor K is the ratio of the ultimate bending moment Mu to the actual bending moment M under the corresponding axial force N on the bearing capacity envelope, i.e., K = Mu / M. Specifically, after determining the actual axial force of the lining section, the ultimate bending moment corresponding to the axial force is obtained from the bearing capacity envelope. The ultimate bending moment is divided by the actual bending moment to obtain the physical safety factor K. Then, the physical safety level is determined according to the level range in which the physical safety factor K is located.

[0056] K≥2.0 Level 1 20 points 1.5≤K＜2.0 Level 2 50 points 1.0≤K＜1.5 Level 3 80 points K＜1.0 Level 4 100 points

[0057] When the structural condition monitoring data does not simultaneously include stress-strain data and seepage pressure data, the physical assessment model adopts the assessment path corresponding to the collected data. When convergent deformation data and stress-strain data are available, the actual stress state of the lining is determined based on the convergent deformation amount and stress-strain change amount, and the physical safety level is determined according to the bearing capacity envelope. When convergent deformation data and seepage pressure data are available, the actual stress state of the lining is determined based on the convergent deformation amount and seepage pressure change amount, and the physical safety level is determined according to the bearing capacity envelope. The above assessment paths use the same physical safety level range, and the output results are all physical safety levels.

[0058] The grade envelope fusion module receives the apparent damage grade and the physical safety grade. The apparent damage grade reflects the severity of apparent defects such as cracks, leaks, and spalling in the same evaluation object, while the physical safety grade reflects the relationship between the actual stress state and bearing capacity of the lining in the same evaluation object. During grade envelope fusion, if the apparent damage grade and the physical safety grade are consistent, the consistent grade is taken as the current safety grade; if the apparent damage grade and the physical safety grade are inconsistent, the grade with the higher risk level is taken as the current safety grade. For example, if the apparent damage grade is level 2 and the physical safety grade is level 3, then the current safety grade is level 3; if the apparent damage grade is level 4 and the physical safety grade is level 2, then the current safety grade is level 4.

[0059] The risk warning module determines the risk level based on the current safety level, the trend of disease development, the trend of structural response changes, environmental operation data, the location of the evaluation object in the lining circumferential region, and the combination of multiple diseases. The risk level is determined using a multi-factor scoring method. The current safety level, trend, lining circumferential region, environmental operation data, and combination of multiple diseases are each converted into a score value, and then summed according to preset weights to obtain the comprehensive risk score R.

[0060] The overall risk score R is calculated using the following formula:

[0061] R=0.35A+0.25B+0.15C+0.10D+0.15E

[0062] Among them, A is the current safety level score; B is the trend score; C is the circumferential part score of the lining; D is the environmental operation effect score; and E is the multi-disease combination relationship score.

[0063] Current security level rating: A Hierarchical envelope fusion results 0.35 Level 1 is worth 20 points, Level 2 is worth 50 points, Level 3 is worth 80 points, and Level 4 is worth 100 points. Trend Score B Disease development trend and structural response change trend 0.25 20 points are awarded for a predicted timeframe of >90 days to reach the warning threshold, 50 points for 30-90 days, 80 points for 7-30 days, and 100 points for <7 days. Lining circumferential section score C Evaluation object construction results 0.15 The arch is worth 50 points, the side walls 60 points, the arch waist 80 points, and the arch crown 100 points. Environmental operation role rating D Rainfall intensity, traffic load, temperature and humidity changes, and abnormal vibration events 0.10 The ratio of current environmental load to design load is as follows: <0.8: 20 points; 0.8–1.0: 50 points; 1.0–1.2: 80 points; >1.2: 100 points. Multi-disease combination relationship score E Single disease level combination 0.15 No concurrent diseases: 0 points; Level 1 upgrade: 25 points; Level 2 upgrade: 50 points; Upgrade directly to the highest level: 75 points.

[0064] The environmental load level L is defined as the ratio of the current combined load to the design load, and is calculated using the following formula:

[0065]

[0066] Where T is the current traffic load (equivalent number of axle loads). For design traffic load; R is the current water head height. The design head is V; V is the current average speed of the vehicle. The weighting coefficients are set by default for the designed vehicle speed. This can be adjusted by the system administrator based on the actual operation of the tunnel. When a certain monitoring data is missing, the ratio for that item is set to 1.0.

[0067] The overall risk score R ranges from 0 to 100. The risk level is determined according to the table below:

[0068] R＜30 Low risk Blue Alert 30≤R＜60 Medium risk Yellow Alert 60≤R＜85 High risk Orange alert R≥85 Extremely high risk Red Alert

[0069] In this embodiment, the combination relationship of multiple diseases is determined by the combination of crack level, leakage level and spalling level in the same evaluation object. The disease level evaluation module first outputs the crack level, leakage level and spalling level respectively, and then determines the apparent damage level and the combination relationship of multiple diseases based on the single disease level corresponding to the identified disease.

[0070] In this embodiment, the combination relationship of multiple diseases is determined by the combination of crack level, leakage level and spalling level in the same evaluation object. The disease level evaluation module first outputs the crack level, leakage level and spalling level respectively, and then determines the apparent damage level and the combination relationship of multiple diseases based on the single disease level corresponding to the identified disease.

[0071] Cracks, leaks, and peeling are all classified as Level 1. No correction Level 1 The crack level is level 2 or above, the leakage level is level 2 or above, and the spalling level is level 1. Risk level raised by 1 level Level 3 The crack level is level 2 or higher, the spalling level is level 2 or higher, and the leakage level is level 1. Risk level raised by 1 level Level 3 The crack level is level 3 or above, the leakage level is level 2 or above, and the spalling level is level 1. Risk level raised by 2 levels Level 4 The crack level is level 3 or above, the spalling level is level 3 or above, and the leakage level is level 1. Risk level raised by 2 levels Level 4 The crack level, leakage level, and spalling level all reached level 2 or above. Directly classified as extremely high risk Level 4 Any disease level reaches level 3 or above and triggers a combination correction. The risk level has been raised by one level; if the original risk level was level 2, it will be raised to level 3. Level 4

[0072] If the theoretical level after combination correction exceeds level 4, then level 4 is directly taken as the final level. When the combination of multiple diseases directly triggers extremely high risk, the risk warning module covers the calculation results of the weighted scoring method and directly determines the risk level as extremely high risk.

[0073] In another implementation, the risk warning module uses the risk matrix method to determine the risk level. The risk matrix is ​​constructed with the probability level of disease occurrence as one dimension and the severity level of consequences as another dimension. The probability level of disease occurrence is determined based on the disease development rate, the rate of change of structural response, and environmental operation data. The severity level of consequences is determined based on the current safety level, the circumferential location of the evaluation object in the lining, and the combination relationship of multiple diseases.

[0074] The likelihood level of disease occurrence is determined as follows: Extremely low when the disease development rate is <0.01 mm / day and the environmental load level is <0.8; Low when the disease development rate is 0.01–0.05 mm / day and the environmental load level is 0.8–1.0; Medium when the disease development rate is 0.05–0.1 mm / day and the environmental load level is 1.0–1.2; High when the disease development rate is >0.1 mm / day and the environmental load level is >1.2; Extremely high when sudden abnormal monitoring data occurs. The severity level is determined according to the current safety level and the circumferential location of the lining: Level 1 and the evaluated object is located on the sidewall or invert arch is Low; Level 2 and the evaluated object is Located on the sidewall or invert arch is Low. The risk level is determined by the following criteria: if the object is located on the side wall or inverted arch, or if it is at level 1 and located at the arch waist, it is considered medium; if it is at level 3 and located at the arch waist, or if it is at level 2 and located at the arch crown, it is considered high; if it is at level 4 and located anywhere, or if it is at level 3 and located at the arch crown, it is considered extremely high. The risk warning module determines the risk level in a preset 5×5 risk matrix based on the probability level of disease occurrence and the severity level of consequences. If the same object triggers the multi-disease combination correction rule, the risk level obtained by the multi-disease combination correction rule is used as the final risk level. When the multi-disease combination relationship directly triggers extremely high risk, the risk matrix output result is overwritten by the extremely high risk determination result.

[0075] The risk warning module generates tiered warning information based on risk levels. These tiered warnings are output in the form of warning records. Each record includes the object number, tunnel chainage, monitoring section number, lining circumferential location, data collection time, apparent defect type, single defect level, apparent damage level, physical safety level, current safety level, comprehensive risk score, risk level, warning level, trigger parameters, trigger rules, and recommended remedial measures. Trigger parameters include at least one of the following: maximum crack width, total crack length, leakage area, spalling area, convergence deformation rate, stress-strain change, seepage pressure change, and environmental operation impact score. Trigger rules include defect level rule library hit rules, level envelope fusion rules, multi-defect combination correction rules, and risk level interval rules.

[0076] In a specific early warning and response method, the triggering condition for a blue alert is that a single disease reaches level 1 and has no obvious development trend. The recommended response is to monitor and maintain the normal monitoring frequency. The triggering conditions for a yellow alert include that a single disease reaches level 2 and the development rate is >0.02 mm / day, or the risk adjustment indicated by the disease combination effect rule reaches level 1 or above. The recommended response is to strengthen monitoring, check once every half month, and assess the disease development trend. The triggering conditions for an orange alert include that a single disease reaches level 3, or the risk adjustment indicated by the disease combination effect rule reaches level 1 or above, or the probability is medium and the consequence is high in the risk matrix method. The recommended response is to arrange internal inspection within 15 days, formulate a maintenance plan, and restrict the passage of heavy vehicles. The triggering conditions for a red alert include that a single disease reaches level 4, or cracks, leakage, and peeling diseases coexist and all reach level 2 or above, or the probability is high and the consequence is extremely high in the risk matrix method. The recommended response is to immediately close traffic, implement temporary support, activate the emergency plan, and organize experts for on-site assessment.

[0077] After the warning information is generated, the system will send it to the designated person in charge through at least one of the following methods: SMS, application push, monitoring center pop-up, and email. The warning information can also be highlighted in the corresponding color on the tunnel's two-dimensional plan or three-dimensional model, and the warning time and processing status will be recorded. When the warning level obtained by the multi-factor weighted scoring method and the risk matrix method are inconsistent, the warning level with the higher risk level shall prevail.

[0078] When using a multi-factor weighted scoring method, the comprehensive risk index R is calculated using the following formula:

[0079]

[0080] In the formula:

[0081] The severity of the disease is scored based on the current safety level; Level I (minor) is scored 20 points, Level II (attention) is scored 50 points, Level III (severe) is scored 80 points, and Level IV (extremely severe) is scored 100 points.

[0082] The disease development rate is scored, and the score is a weighted score of urgency determined based on the timeliness of prediction. Based on the timeliness of prediction: 20 points for the time to reach the warning value > 90 days, 50 points for 30 to 90 days, 80 points for 7 to 30 days, and 100 points for < 7 days.

[0083] The importance coefficient of the structural part where the disease is located is scored. This score is an importance weight determined based on the structural mechanical characteristics. Based on the structural mechanical characteristics, the following scores are given: inverted arch 50 points, side wall 60 points, arch waist 80 points, and arch crown 100 points.

[0084] The environmental load impact score is determined based on the external action weights based on the environmental-load coupling effect; the ratio of the current environmental load to the design load is <0.8, scoring 20 points; 0.8 to 1.0, scoring 50 points; 1.0 to 1.2, scoring 80 points; and >1.2, scoring 100 points.

[0085] The combined effect of disease is scored, where crack level, leakage level, and spalling level are determined based on the apparent disease characteristic parameters extracted in step S2. After determining the level of each disease individually according to their respective preset severity thresholds, the combination of crack level, leakage level, and spalling level is used as input conditions and mapped according to the following rules: 0 points for no concurrent diseases; 25 points for increasing by one level if crack level ≥ II and leakage level ≥ II, or crack level ≥ II and spalling level ≥ II; 50 points for increasing by two levels if crack level ≥ III and leakage level ≥ II, or crack level ≥ III and spalling level ≥ III; 75 points for increasing to the highest level if all three diseases are ≥ II; when the "risk level directly increased to the highest" condition is triggered, the weighted scoring method is not applied, and the risk level is directly determined as extremely high risk.

[0086] ω1, ω2, ω3, ω4 The weighting coefficients are 0.35, 0.25, 0.15, 0.1, and 0.15, respectively; where ω1=0.35 reflects that the current safety status is the most dominant factor, ω2=0.25 reflects the timeliness of the development trend, ω3=0.15 reflects the mechanical importance of the structural parts, ω4=0.10 reflects the external induction effect of environmental loads, and ω5=0.15 reflects the amplification effect of multiple diseases coupled together.

[0087] The comprehensive risk index R ranges from 0 to 100, and risk levels are classified according to the following thresholds: R<30 is low risk, 30≤R<60 is medium risk, 60≤R<85 is high risk, and R≥85 is extremely high risk.

[0088] When the disease combination effect rule triggers the "risk level directly increased to the highest level", the system will overwrite the calculation result of the weighted scoring method and directly determine the risk level as extremely high risk.

[0089] When using the risk matrix method, a 5×5 risk matrix is ​​constructed with the probability level of disease occurrence (vertical axis) and the severity level of consequences (horizontal axis). The probability level is determined comprehensively based on the disease development rate and environmental load level: rate <0.01mm / day and load level <0.8 is extremely low, 0.01~0.05mm / day and load level 0.8~1.0 is low, 0.05~0.1mm / day and load level 1.0~1.2 is medium, >0.1mm / day and load level >1.2 is high, and the occurrence of sudden abnormal monitoring data is extremely high.

[0090] The severity level of the consequences is determined comprehensively based on the current severity of the disease and the importance of the structural part: Level I + sidewall / inverted arch is low, Level II + sidewall / inverted arch or Level I + arch waist is medium, Level III + arch waist or Level II + arch crown is high, and Level IV + any part or Level III + arch crown is extremely high.

[0091] The risk level is determined by the intersection of the matrix, and the matrix's preset rule base stores the output levels corresponding to 5×5 combinations.

[0092] This embodiment also provides a tunnel structure defect identification and safety risk early warning system. The system includes a data acquisition module, an evaluation object construction module, an apparent defect feature extraction module, a structural response feature extraction module, a trend analysis module, a defect level evaluation module, a physical model evaluation module, a level envelope fusion module, and a risk early warning module. The modules are connected sequentially according to the data flow. The data acquisition module outputs apparent image data, structural status monitoring data, and environmental operation effect data. The evaluation object construction module receives the above data and outputs an evaluation object dataset. The apparent defect feature extraction module receives the apparent image data in the evaluation object dataset and outputs apparent defect feature parameters. The structural response feature extraction module receives the structural status monitoring data in the evaluation object dataset and outputs structural response feature parameters.

[0093] The trend analysis module receives historical time-series data, apparent disease characteristic parameters, and structural response characteristic parameters, and outputs the disease development trend and structural response change trend. The disease level evaluation module receives apparent disease characteristic parameters and outputs the single disease level and apparent damage level. The physical model evaluation module receives structural response characteristic parameters and outputs the physical safety level. The level envelope fusion module receives the apparent damage level and physical safety level and outputs the current safety level. The risk warning module receives the current safety level, disease development trend, structural response change trend, environmental operation data, the circumferential location of the evaluated object in the lining, and the relationship of multiple disease combinations, and outputs the risk level and graded warning information.

[0094] In this embodiment, the system is equipped with a data storage unit, which includes a raw data table, an evaluation object data table, an apparent disease characteristic parameter table, a structural response characteristic parameter table, a historical time series data table, a disease level rule table, a physical assessment parameter table, a risk scoring rule table, and an early warning record table.

[0095] The data acquisition module writes raw data into the raw data table; the evaluation object construction module generates the evaluation object dataset and writes it into the evaluation object data table; the apparent disease feature extraction module and the structural response feature extraction module generate corresponding feature parameters respectively; the trend analysis module calls the historical time series data table to calculate the changing trend; the disease level evaluation module calls the disease level rule table to output the apparent damage level; the physical model evaluation module calls the physical assessment parameter table to output the physical safety level; the level envelope fusion module outputs the current safety level; and the risk warning module calls the risk scoring rule table and the multi-disease combination correction rule to generate warning records.

[0096] In one specific implementation, the system further includes an edge computing node and a cloud server. The edge computing node is located at the tunnel site data collection terminal, and the cloud server is located at the monitoring center. The data collected by the image acquisition unit, convergence deformation monitoring unit, stress and strain monitoring unit, seepage pressure monitoring unit, vibration acceleration monitoring unit, and environmental operation effect acquisition unit are first transmitted to the edge computing node.

[0097] After receiving the apparent image data, the edge computing node performs grayscale conversion, filtering and noise reduction, illumination correction and image size normalization. It also performs single-factor threshold early warning judgment on at least one of the following: maximum crack width, convergence deformation rate and leakage area. When the maximum crack width, convergence deformation rate or leakage area reaches the corresponding preset warning value, the edge computing node generates a primary early warning message and uploads the object number, trigger parameters, collection time and field data of the evaluation object to the cloud server.

[0098] After receiving the data uploaded by the edge computing nodes, the cloud server retrieves the historical time-series data of the evaluation object according to the evaluation object number, and performs apparent disease feature extraction, structural response feature extraction, trend analysis, disease level evaluation, physical model evaluation, level envelope fusion, and risk level determination. When the warning level output by the edge computing node is inconsistent with the warning level output by the cloud server, the risk warning module adopts the higher risk level as the final warning level. The edge computing node is used for rapid on-site screening, and the cloud server is used for comprehensive analysis and final risk determination.

[0099] In a specific application example, the arch waist of the first monitoring section at chainage K12+350 of an operating highway tunnel is taken as the evaluation object. The evaluation object construction module assigns a unique object number to this location and associates the apparent image data, convergence deformation data, stress-strain data, seepage pressure data, rainfall intensity data and traffic load data of this location into the evaluation object dataset.

[0100] After processing the surface image data, the surface defect feature extraction module identifies two types of surface defects: cracks and leaks. The crack area is processed into a skeleton to obtain the crack centerline, and the crack width distribution is measured along the normal of the crack centerline to obtain the maximum crack width, average crack width, total crack length, and crack line density. The leak area is converted into the leak area and the leak area percentage after pixel area conversion.

[0101] The structural response feature extraction module performs timestamp correction, outlier removal, filtering and noise reduction, and missing value interpolation on the structural condition monitoring data to obtain the convergence deformation amount, convergence deformation rate, stress-strain change amount, seepage pressure change amount, and seepage pressure change trend. The trend analysis module calls the data of the evaluation object in the past multiple evaluation periods to calculate the maximum crack width change rate, leakage area change rate, convergence deformation rate, and seepage pressure change rate.

[0102] The disease level evaluation module inputs the maximum crack width, total crack length, crack linear density, leakage area, and leakage area ratio into the disease level rule base to obtain the crack level and leakage level, and determines the single disease level with the higher risk as the apparent damage level. The physical model evaluation module determines the bending moment and axial force of the tunnel lining section based on the surrounding rock pressure, water pressure, convergence deformation, stress-strain change, and seepage pressure change, and obtains the physical safety factor K based on the axial force-bending moment correlation envelope, and then obtains the physical safety level based on the level range in which the physical safety factor K is located.

[0103] The grade envelope fusion module compares the apparent damage grade and the physical safety grade, and takes the grade with the higher risk level as the current safety grade. The risk warning module calculates the comprehensive risk score R based on the current safety grade, the development trend of the disease, the trend of structural response change, the arch waist score, rainfall intensity, traffic load, and the combination relationship between cracks and leakage. If both the crack grade and the leakage grade reach level 2 or above, the risk grade corresponding to the comprehensive risk score R is increased by 1 level, and a graded warning information containing the evaluation object number, disease type, trigger parameters, risk grade, and warning level is generated.

[0104] In this embodiment, apparent defect identification, structural response analysis, physical model evaluation, and risk warning are all performed around the same evaluation object. Apparent image data is used to extract apparent defect characteristic parameters of cracks, leaks, and spalling, and is not directly used as the final safety level. Structural condition monitoring data is used to extract convergence deformation, stress-strain change, seepage pressure change, and vibration acceleration related parameters, and is then entered into the physical evaluation model to obtain the physical safety level. Environmental operation effect data participates in the comprehensive judgment as environmental operation effect score in the risk scoring stage. Apparent defect characteristic parameters, structural response characteristic parameters, historical trends, environmental operation effect data, and multi-defect combination relationships under the same evaluation object are all entered into the risk judgment process, forming a continuous processing chain from data collection, object association, feature extraction, level evaluation, level envelope to graded warning. Based on the above input data, processing rules, parameter sources, module connection relationships, and output fields, those skilled in the art can realize the tunnel structure defect identification and safety risk warning described in this embodiment.

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for identifying tunnel structural defects and providing early warning of safety risks, characterized in that: include: S1. Collect apparent image data, structural condition monitoring data and environmental operation data of the tunnel lining. The structural condition monitoring data includes convergence deformation data and at least one of stress-strain data and seepage pressure data. S2. Determine the evaluation object according to the tunnel chainage, monitoring section and circumferential part of the lining, and associate the appearance image data, structural status monitoring data and environmental operation data of the same evaluation object into an evaluation object dataset; S3. Extract at least one type of apparent disease characteristic parameter from the apparent image data of cracks, leakage, and spalling from the apparent image data of the evaluation object dataset, and extract structural response characteristic parameters from the structural state monitoring data. S4. Call the historical time series data of the evaluation object to obtain the disease development trend and structural response change trend; S5. Input the apparent disease characteristic parameters into the disease level rule base to obtain a single disease level, and determine the apparent damage level based on the single disease level. S6. Input the structural response characteristic parameters into the physical evaluation model, determine the actual stress state of the tunnel lining by the physical evaluation model, and compare the actual stress state with the bearing capacity envelope of the tunnel lining to obtain the physical safety level. S7. Perform level envelope fusion on the apparent damage level and the physical safety level, and take the level with higher risk as the current safety level. S8. Determine the risk level based on the current safety level, disease development trend, structural response change trend, environmental operation data, the circumferential location of the evaluation object in the lining, and the combination relationship of multiple diseases, and generate graded early warning information.

2. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 1, characterized in that: In step S2, the circumferential lining portion includes the arch crown, arch waist, sidewalls, and invert arch; each evaluation object is assigned a unique object number, which is used to associate the apparent disease characteristic parameters, structural response characteristic parameters, historical time series data, risk level, and graded early warning information of the same evaluation object.

3. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 2, characterized in that: In step S3, the extraction of the apparent disease characteristic parameters includes: The apparent image data is converted to grayscale, histogram equalization, filtered for noise reduction, and illumination correction to obtain a preprocessed image; The preprocessed image is classified at the pixel level to obtain a mask image of the disease area corresponding to the identified apparent disease; The mask image is subjected to morphological processing, and the contour of the diseased area is extracted based on the connected component labeling; When the apparent defects, including cracks, have been identified, the crack area is skeletonized, and the crack width distribution is measured along the crack centerline normal to obtain the maximum crack width, average crack width, total crack length and crack line density. When the identified apparent defects include leakage, the pixel area of ​​the leakage area is converted to obtain the leakage area and the percentage of leakage area. When the identified apparent defects include peeling, the pixel area of ​​the peeling area is converted to obtain the peeling area and the percentage of the peeling area. Among them, crack linear density is the ratio of the total crack length to the area of ​​the corresponding detection area of ​​the evaluation object, leakage area ratio is the ratio of the leakage area to the area of ​​the corresponding detection area of ​​the evaluation object, and peeling area ratio is the ratio of the peeling area to the area of ​​the corresponding detection area of ​​the evaluation object.

4. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 3, characterized in that: In step S3, the extraction of the structural response feature parameters includes: The structural condition monitoring data is subjected to timestamp correction, outlier removal, filtering and noise reduction, and missing value imputation. The convergence deformation amount and convergence deformation rate are obtained from the processed convergence deformation data. When structural condition monitoring data includes stress and strain data, the stress and strain change is obtained from the processed stress and strain data. When structural condition monitoring data includes seepage pressure data, the amount and trend of seepage pressure change are obtained from the processed seepage pressure data.

5. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 4, characterized in that: In step S4, the historical time series data includes apparent disease characteristic parameters and structural response characteristic parameters formed by the same evaluation object in different evaluation periods; the disease development trend includes at least one of the following: the rate of change of maximum crack width, the rate of change of total crack length, the rate of change of leakage area, and the rate of change of spalling area. The structural response change trend includes at least one of the following: convergence deformation rate, stress-strain change rate, and seepage pressure change rate. Each rate of change is determined by the ratio of the parameter difference between adjacent evaluation periods for the same evaluation object to the time interval between evaluation periods.

6. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 5, characterized in that: In step S5, the disease severity rule base stores the rules for determining the severity of a single disease using production rules; the crack level is determined based on the maximum crack width, total crack length, and crack line density; the leakage level is determined based on the leakage area and the proportion of the leakage area; and the spalling level is determined based on the spalling area and the proportion of the spalling area. When there are two or more single disease levels for the same evaluation object, the single disease level with the highest risk is determined as the apparent damage level. When there is only one single disease level for the same evaluation object, the single disease level is determined as the apparent damage level.

7. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 6, characterized in that: In step S6, the structural condition monitoring data includes convergence deformation data, stress-strain data, and seepage pressure data; the physical assessment model determines the bending moment and axial force of the tunnel lining section based on the surrounding rock pressure, water pressure, convergence deformation, stress-strain change, and seepage pressure change, and determines the physical safety factor based on the axial force-bending moment correlation envelope, and then determines the physical safety level based on the level range of the physical safety factor. The physical safety factor is the ratio of the ultimate bending moment to the actual bending moment under the corresponding axial force on the bearing capacity envelope.

8. The method for identifying tunnel structural defects and providing early warning of safety risks according to claim 7, characterized in that: In step S8, the risk level is determined by a multi-factor scoring method. The scoring factors include the current safety level, the trend of disease development, the trend of structural response change, the circumferential position of the lining where the evaluation object is located, environmental operation data, and the combination relationship of multiple diseases. Among them, the current safety level is obtained by step S7, the trend of disease development and the trend of structural response change are obtained by step S4, the circumferential part of the lining where the evaluation object is located is obtained by step S2, the combination relationship of multiple diseases is obtained by step S5, and the environmental operation effect data are used to form an environmental operation effect score based on the collected data in rainfall intensity, traffic load, temperature and humidity changes and abnormal vibration events. When the crack level and leakage level of the same evaluation object reach level II or above, or when the crack level and spalling level reach level II or above, the risk level obtained based on the scoring factors will be upgraded by one level; when the crack level, leakage level and spalling level of the same evaluation object all reach level II or above, the risk level will be determined as extremely high risk.

9. A tunnel structure defect identification and safety risk early warning system, applied to the tunnel structure defect identification and safety risk early warning method according to any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to collect visual image data, structural status monitoring data, and environmental operation data of the tunnel lining. The evaluation object construction module is used to determine the evaluation object according to the tunnel station, monitoring section and circumferential part of the lining, and to associate the appearance image data, structural status monitoring data and environmental operation data of the same evaluation object into an evaluation object dataset. The apparent disease feature extraction module is used to output apparent disease feature parameters from the apparent image data in the evaluation object dataset. The structural response feature extraction module is used to output structural response feature parameters from the structural state monitoring data in the evaluation object dataset. The trend analysis module is used to output the disease development trend and structural response change trend based on the historical time series data of the evaluation object; The disease severity evaluation module is used to output a single disease severity level and apparent damage level based on the apparent disease characteristic parameters. The physical model evaluation module is used to output the physical safety level based on the structural response characteristic parameters. The graded envelope fusion module is used to output the current safety level based on the apparent damage level and the physical safety level; The risk warning module, connected to the grade envelope fusion module, trend analysis module, and evaluation object construction module, is used to receive the current safety level, disease development trend, structural response change trend, environmental operation data, the circumferential location of the evaluation object in the lining, and the combination relationship of multiple diseases, and outputs risk level and graded warning information.

10. The method and system for identifying tunnel structural defects and providing early warning of safety risks according to claim 9, characterized in that: The data acquisition module includes an image acquisition unit, a convergence deformation monitoring unit, a stress-strain monitoring unit, a seepage pressure monitoring unit, and an environmental operation effect acquisition unit; The system also includes edge computing nodes and cloud servers. The edge computing nodes are used to preprocess the appearance image data and structural status monitoring data, and to make threshold early warning judgments on at least one of the following: maximum crack width, convergence deformation rate, and leakage area. The cloud server is used to perform trend analysis, disease level evaluation, physical model evaluation, level envelope fusion, and risk level determination. When the warning level output by the edge computing node is inconsistent with the warning level output by the cloud server, the risk warning module adopts the higher level as the final warning level.

Images