A disease development evaluation method based on multi-source data and digital-physical fusion experiment

Abstract

This invention discloses a method for assessing tunnel lining defects based on multi-source data and data-physical fusion experiments. The method includes front-end multi-source data acquisition, numerical analysis and prediction of defect development, and physical testing machine-based reconstruction and prediction of defect conditions. This invention acquires multi-source data on tunnel lining defects and structural degradation using a linear array camera, laser vibrometer, and 3D laser scanner. The acquired data is processed and analyzed, and the development tendency of the main defects controlling lining damage is predicted based on the TCI circle. A tunnel lining-defect model is constructed based on the original tunnel lining operating conditions and the acquired defect information. A dynamic-static load lining pressure testing machine is used to simulate tunnel surrounding rock pressure, seismic loads, and traffic loads. Pressure tests are conducted on the lining-defect model to verify the development tendency of the main defects. The physical experimental results can be mapped to the actual defect development situation in the engineering field, providing a reference for the safe operation and maintenance of tunnel linings and disaster prevention and control.

Description

Technical Field

[0001] This invention relates to the field of tunnel lining maintenance and testing technology, specifically to a method for assessing disease development based on multi-source data and data-physical fusion experiments. Background Technology

[0002] Tunnel lining, as a permanent support structure in tunnel engineering, is a crucial guarantee for the normal operation and maintenance of traffic tunnels. With increasing service life, various internal and external factors lead to the emergence of various defects such as cracks, voids, water leakage, spalling, and deformation. Failure to promptly inspect, repair, and reinforce these defects will inevitably endanger the safety of traffic vehicles. Currently, the commonly used inspection method for subway tunnels is the traditional manual inspection method. This method is affected by the working conditions and experience level of the inspectors, resulting in inconsistent data quality. Furthermore, it is costly in terms of manpower and resources and inefficient, increasingly failing to meet the actual needs of tunnel inspection projects.

[0003] Tunnel lining defects are diverse, but related technologies are mostly limited to detecting single defects. For example, ultrasonic and ground-penetrating radar-based methods are suitable for detecting cavities inside the lining, infrared-based methods are suitable for detecting surface water leakage, and machine vision-based methods are suitable for detecting cracks. These methods all have limitations. For instance, ground-penetrating radar requires close proximity to the lining to ensure accuracy when detecting internal defects, thus limiting work efficiency. Existing detection methods typically involve fixed-point, periodic inspections of key areas, making it difficult to achieve intelligent, batch inspection of tunnels.

[0004] To address the above problems, this invention proposes a disease development assessment method based on multi-source data and data-physical fusion experiments. Summary of the Invention

[0005] This invention addresses the shortcomings of existing tunnel lining defect detection methods, such as limited accuracy and efficiency in defect quantification, lack of scientific theoretical support for defect development prediction, lack of physical experimental verification of prediction results, reliance on experience and poor targeting of defect prevention and control measures, and disconnect between data prediction and engineering practice. It aims to provide a defect development assessment method based on multi-source data and data-physical fusion experiments. This method comprehensively acquires multi-dimensional defect information of the lining through multi-source defect integrated acquisition equipment, achieves intelligent defect identification and accurate quantification using deep learning algorithms, accurately predicts the development direction and location of major lining defects based on TCI circle theory, and verifies the prediction results through dynamic-static load testing of a lining-defect model matched to actual working conditions. The experimental patterns are then mapped to the engineering site, achieving a closed-loop management of tunnel lining defects from comprehensive detection, intelligent quantification, accurate prediction, physical verification to scientific prevention and control. This fundamentally solves the industry pain points of tunnel lining defect analysis and operation and maintenance prevention and control, improves the accuracy and reliability of defect analysis and the targeting of prevention and control measures, ensures tunnel structural safety, extends service life, and reduces the total life-cycle operation and maintenance cost.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for assessing disease development based on multi-source data and data-physical fusion experiments, comprising the following steps: Step S1: Collect basic tunnel working condition information, standardize the quantifiable working condition parameters, and obtain standardized tunnel working condition data. Step S2: Using the standardized tunnel working condition data as the detection benchmark, multi-source defect comprehensive acquisition equipment is used to acquire multi-source detection data of surface defects, internal defects, structural abnormalities and structural deterioration of tunnel lining, forming multi-source original detection data of lining. Step S3: Using the original multi-source detection data of the lining as a sample, a deep learning algorithm is used to construct a lining disease database and train a disease detection model. The disease detection model is used to complete disease identification and quantification to obtain quantified disease data. Step S4: Based on the quantified disease data, screen the main control diseases of the lining, calculate and predict the development direction and location of the main control diseases of the lining based on the TCI circle theory, and obtain the prediction results of the development tendency of the main control diseases of the lining. Step S5: Based on standardized tunnel working condition data and multi-source original detection data of lining, fabricate lining molds, select similar materials, and pour and cure a tunnel lining-disease model containing equivalent defects and damages. Step S6: Place the tunnel lining-disease model in a dynamic-static load lining pressure testing machine to conduct pressure tests simulating surrounding rock pressure, seismic load, and traffic load. Use the test disease development data to verify the predicted development trend of the main lining disease and map the test patterns to the engineering site to form tunnel lining disease prevention and control measures.

[0007] Furthermore, the multi-source disease comprehensive collection equipment in step S2 is composed of a mobile vehicle carrying a line array camera unit, a 3D laser scanner, a ground-penetrating radar, and a laser vibration meter.

[0008] Furthermore, the disease detection model described in step S3 can be adapted to use a semantic segmentation model, a target detection model, or an instance segmentation model based on the disease characteristics.

[0009] Furthermore, the formula for calculating the TCI value of the lining crack based on the TCI circle theory in step S4 is as follows: ; In the formula, The area to be lining; The number of cracks; For the first The width of the crack; For the first The length of the crack; This represents the TCI component along the tunnel length direction. The radial component of TCI in the tunnel; This represents the total TCI value. The weighted correlation coefficient for crack width; This is the weighted correlation coefficient for crack length.

[0010] Furthermore, , The calculation formula is: ; ; In the formula, For the first Strip crack normal and The included angle of the axis; For the first Strip crack normal and The included angle of the axis.

[0011] Furthermore, in step S4, based on the second-order TCI tensor, a TCI circle is constructed with reference to the Mohr stress circle, and the development direction of the main control crack and the circle radius are determined through the geometric relationship of the TCI circle.

[0012] Furthermore, the lining mold in step S5 is a detachable splicing structure, and the tunnel lining-disease model is 1:10 in size ratio to the original tunnel.

[0013] Furthermore, in step S5, hollow spheres and steel sheets are embedded inside the tunnel lining-disease model to proportionally recreate the actual defects and damage of the tunnel.

[0014] Furthermore, the dynamic-static load lining pressure testing machine described in step S6 is composed of an electro-hydraulic servo pressure application system and a vibration system.

[0015] Furthermore, in step S6, the prediction results of the development tendency of the main control disease of the lining are verified by the physical test results, and the development of the main control disease of the lining is directly mapped to the actual disease development at the engineering site.

[0016] The present invention provides a disease development assessment method based on multi-source data and data-physical fusion experiments, which achieves several technical advantages: 1. Achieve comprehensive, all-dimensional data collection of tunnel lining defects, overcoming the limitations of single detection methods. By integrating a multi-source defect collection system with a linear array camera unit, a 3D laser scanner, ground-penetrating radar, and a laser vibration meter, it acquires comprehensive, multi-source detection data on surface defects, internal defects, structural abnormalities, and structural deterioration of the tunnel lining in a single operation. This comprehensively covers all types of lining defects, completely resolving the issues of missed detections and incomplete coverage inherent in traditional single detection methods, and providing a complete and reliable data foundation for subsequent defect analysis.

[0017] 2. Achieve intelligent, efficient, and accurate quantification of defects, significantly improving the efficiency and accuracy of operation and maintenance detection. Deep learning algorithms are used to construct a dedicated database for internal and external defects in the lining, training semantic segmentation, target detection, or instance segmentation models that match defect features. This automatically completes defect identification and quantification, while damage indicators provide a quantitative description of the lining structure's deterioration. This replaces traditional manual / semi-automatic detection, significantly reducing human error and improving the efficiency and accuracy of defect quantification, adapting to the needs of large-scale tunnel operation and maintenance detection.

[0018] 3. Based on the TCI circle theory, accurate prediction of major defects is achieved, filling the technical gap in the prediction of defect development trends. Through TCI value calculation, second-order tensor component decomposition, and TCI circle construction, major defects in tunnel lining are accurately screened, and the development direction and location of major defects are scientifically predicted. This breaks through the limitation of existing technologies that can only assess the current severity of defects, providing theoretical support for the early prevention and control of tunnel defects, and realizing the upgrade of tunnel operation and maintenance from post-maintenance to pre-prediction.

[0019] 4. Construct a multi-source data-physical experiment fusion mapping system to ensure the engineering reliability of prediction results. By casting a lining-disease model that matches actual tunnel working conditions, dynamic-static load pressure tests are conducted to simulate the development of defects under surrounding rock pressure, earthquakes, and traffic loads. The accuracy of data prediction conclusions is verified by physical test results, breaking down the barriers between data models and engineering reality. This addresses the shortcomings of existing technologies that rely solely on data model predictions and lack physical verification, significantly improving the engineering applicability and reliability of defect prediction results. Attached Figure Description

[0020] Figure 1This is a flowchart of the disease development assessment method based on multi-source data and data-physical fusion experiments of the present invention.

[0021] Figure 2 This is a complete technical flowchart of the disease development assessment method based on multi-source data and data-physical fusion experiments of the present invention.

[0022] Figure 3 This is a schematic diagram of the molar stress circle for the plane stress state of the present invention.

[0023] Figure 4 This is a direct geometric representation of the TCI circle of the present invention.

[0024] Figure 5 This is a plan view of the tunnel's surface.

[0025] Figure 6 This is an image showing the results of crack extraction.

[0026] Figure 7 (a), (b), and (c) are respectively field test diagrams, experimental diagrams, and simulation diagrams performed according to the method of the present invention.

[0027] Figure 8 This is a schematic diagram of the calculation results of the present invention. Detailed Implementation

[0028] The following detailed implementation of a disease development assessment method based on multi-source data and data-physical fusion experiments according to the present invention is described in detail with reference to specific embodiments. The embodiments of the present invention are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0029] Example 1 This invention provides a method for assessing disease development based on multi-source data and data-physical fusion experiments. Figure 1 This is a flowchart of the method of the present invention. Figure 2 The complete technical flow diagram of the present invention includes the following steps: Step S1 involves collecting basic tunnel working condition information and standardizing quantifiable working condition parameters to obtain standardized tunnel working condition data. The specific content includes: The survey obtained basic information about the tunnel, including its name, design dimensions, mileage, construction technology, service life, geological conditions, surrounding rock grade, and lining concrete grade, and processed the main quantifiable information.

[0030] Step S2: Using the standardized tunnel working condition data as the detection benchmark, multi-source defect comprehensive acquisition equipment is used to acquire multi-source detection data of surface defects, internal defects, structural abnormalities and structural deterioration of the tunnel lining, forming multi-source original detection data of the lining.

[0031] In step S2, the multi-source disease comprehensive acquisition equipment is composed of a mobile vehicle carrying a line-scan camera unit, a 3D laser scanner, a ground-penetrating radar, and a laser vibrometer. Specifically: Based on the multi-source disease comprehensive acquisition equipment, the tunnel lining damage and internal and external disease conditions are obtained. The multi-source disease comprehensive acquisition equipment is a comprehensive detection equipment composed of a line array camera unit, a three-dimensional laser scanner, a ground radar and a laser vibration meter. It can obtain the surface disease conditions such as cracks, water leakage and spalling of the tunnel lining, the structural abnormality of the lining deformation, the internal disease conditions such as lining voids and insufficient lining, and the structural deterioration conditions such as lining deterioration and insufficient concrete strength.

[0032] Preferably, the line array camera unit consists of alternating line array cameras and laser light sources, and the laser vibration meter can realize mobile vibration detection.

[0033] Step S3: Using the original multi-source detection data of the lining as samples, a deep learning algorithm is used to construct a lining disease database and train a disease detection model. The model is used to complete disease identification and quantification to obtain quantified disease data.

[0034] Specifically, using the original multi-source detection data of the lining as samples, a disease database is constructed using deep learning algorithms, and a disease detection model is trained. The model is then used to complete disease identification and quantification. The disease detection model needs to be adapted to the disease characteristics, selecting either a semantic segmentation model, a target detection model, or an instance segmentation model. For lining distortion and deterioration, damage indicators are used for quantitative description.

[0035] The implementation process of deep learning algorithms involves data collection, dataset construction, model training, and model application for defect detection. Taking semantic segmentation models as an example, the data preparation process before training is as follows: First, the original high-resolution image is cut into smaller sub-images to fit the model input. Then, a classifier based on convolutional neural networks is used to filter and classify the sub-images, establishing a dedicated crack defect dataset to provide data support for the training of subsequent crack detection algorithms such as DSC-CrackU.

[0036] The specific implementation of the deep learning algorithm in step S3 includes: 1. Data Preparation and Preprocessing First, the acquired lining images were segmented. The original high-resolution images (e.g., 899,942 pixels × 7562 pixels, consuming 6.33GB of memory) could not be directly used for model training. They needed to be manually cropped into 256-pixel × 256-pixel sub-images, with each large image segmented into 101,937 smaller images. Subsequently, a CNN classifier based on the ResNet34 architecture was trained to automatically filter the sub-images, classifying them into four categories: cracks, pipes, joints, and no defects. The crack classification accuracy reached 86.29%. From the filtered results, 580 crack images were selected, and pixel-level manual annotations were used to generate label images. Data augmentation was then performed through geometric transformations (rotation, cropping, mirroring, scaling) and pixel-level transformations (brightness, contrast, saturation adjustment, and Gaussian blur), expanding the dataset to 8700 images. These were then divided into training, validation, and test sets in an 8:1:1 ratio, forming a standardized crack disease database.

[0037] 2. Model Training Based on the DSC-CrackU architecture, this model incorporates U-Net with embedded dynamic serpentine convolutions, a feature fusion module, and a self-efficient channel module to adapt to the slender and curved shape of cracks. Training was conducted on an i9-14900KF + RTX4080 hardware environment, using the AdamW optimizer and a binary cross-entropy loss function. The initial learning rate was set to 0.01, and a total of 100 training epochs were performed, ultimately achieving a segmentation accuracy of 85.04% mIoU.

[0038] 3. Quantification of Diseases After completing pixel-level crack segmentation, the model establishes a mapping relationship between pixels and physical dimensions to calculate the area, length, width, and centroid coordinates of individual cracks. Based on the width, the severity of the cracks is categorized into four levels: blue, yellow, orange, and red. At the overall lining level, three core indicators—maximum crack width, total crack length, and crack density per unit area—are statistically analyzed for each segment. Safety level assessments are then performed according to Chinese highway tunnel maintenance standards, with a ±15% error margin introduced for rating stability verification. This forms a complete quantitative and risk assessment system from individual cracks to the entire lining section. The final output quantitative data has been converted from pixel dimensions to physical dimensions and can be directly used for engineering defect assessment and maintenance decisions.

[0039] Step S4: Based on the quantitative disease data, screen the main control diseases of the lining, calculate and predict the development direction and location of the main control diseases based on the TCI circle theory, and obtain the development tendency prediction results of the main control diseases.

[0040] Specifically, based on the quantitative defect data obtained in step S3, the main control defects of the lining are screened. The screening process follows industry standards such as the Technical Specifications for Highway Tunnel Maintenance and the Technical Regulations for the Classification and Detection of Structural Defects in Urban Rail Transit Shield Tunnels. Various defects are scored, and those with higher scores are selected as the main control defects. Given that the occurrence and development of most defects are accompanied by crack development, cracks are the primary focus in the actual analysis. The development tendency of the main control defects is predicted based on the TCI circle theory. Specifically, a second-order TCI tensor is constructed using crack data, and a TCI circle is constructed using the Mohr stress circle method. The geometric relationship of the TCI circle is used to determine the development direction of the main control cracks and the radius of the TCI circle, ultimately obtaining the predicted results of the development direction and location of the main control defects.

[0041] In step S4, the development tendency of the main tunnel defects is predicted based on the TCI circle theory. The specific method is as follows: Step S41: Lining cracks are the most direct representation of lining damage. Using the quantitative data obtained in step S3, the TCI value of the cracks within the lining section is calculated. The specific quantitative data obtained in step S3 includes length, width, angle, and lining section area. This allows for the assessment of the hazard and distribution of the lining cracks. The specific calculation formula is as follows: ; In the formula, The area to be lining; The number of cracks; For the first The width of the crack; For the first The length of the crack; This represents the TCI component along the tunnel length direction. The radial component of TCI in the tunnel; This represents the total TCI value. The weighted correlation coefficient for crack width is usually taken as 1 based on engineering experience; This is the weighted correlation coefficient for crack length, which is usually taken as 1 based on engineering experience.

[0042] Step S42, and The calculation method is as follows: ; ; In the formula, For the first Strip crack normal and The included angle of the axis; For the first Strip crack normal and The included angle of the axis.

[0043] Step S43, as follows Figure 3 As shown, Figure 3 This is a schematic diagram of the Mohr stress circle for the plane stress state of this invention. The TCI is a second-order tensor identical to the stress tensor. The TCI circle can be constructed using the Mohr stress circle method based on the relationship between principal stresses and shear stresses. The components of the TCI can be calculated from crack data. and These represent the principal stresses in tensor theory. and shear stress According to and The relationship between the TCI circle and the control crack development direction is calculated to obtain the main direction of the crack development.

[0044] The Mohr stress circle theory formula is: ; ; ; ; In the formula, For any Normal stress on an angled cross section; For any Shear stress on an angled cross section; The principal stress in the x-direction; The principal stress in the y-direction; This refers to the shear stress in the xy plane; The azimuth of the principal plane.

[0045] The above formula utilizes Mohr's Circle theory to transform tunnel lining crack data into the dominant crack development direction, describing the relationship between stress and original stress components on any inclined section, and constructing a stress circle accordingly. The center C represents the central location of the TCI stress state, corresponding to the mean normal stress. Geometrically, it is the reference point for determining the location of the TCI circle. Radius The radius represents the extreme range of the TCI stress state. A larger radius indicates a more complex stress state characterized by crack data and a higher risk gradient for crack development. On the TCI circle, the central angle is... The location corresponds to the principal stress point. In the analysis of tunnel lining defects, this angle... This refers to the main direction of the development of the controlling crack. The F and stress mentioned above are both tensors, and the Mohr stress circle can determine the direction of stress; therefore, the TCI circle can be derived by analogy.

[0046] like Figure 4The diagram shown is a direct geometric representation of the TCI circle of this invention. Based on the geometric relationships in the diagram, the following formula can be derived from the illustrated TCI circle method: ; ; ; In the formula, The development direction angle of the controlling crack; Let be the radius of the TCI circle.

[0047] Step S5: Based on standardized tunnel working condition data and multi-source original detection data of the lining, a lining mold is made, similar materials are selected, and a tunnel lining-disease model containing equivalent defects and damages is poured and cured.

[0048] Specifically, based on standardized tunnel working condition data and multi-source original detection data of the lining, a detachable and splicable lining mold is constructed. Similar materials are selected, and a tunnel lining-disease model with a scale of 1:10 to the original tunnel size is poured and cured. The model structure is adjusted according to the actual tunnel working conditions and can be circular or horseshoe-shaped. During the model pouring process, the actual damage is reconstructed proportionally based on the collected defect parameters. Hollow spheres are used to simulate cavitation-type defects, and steel plates are used to simulate crack-type defects, achieving accurate reproduction of equivalent defect damage.

[0049] Step S6: Place the tunnel lining-disease model in a dynamic-static load lining pressure testing machine to conduct pressure tests simulating surrounding rock pressure, seismic load, and traffic load. Use the test disease development data to verify the predicted results of the main control disease development tendency, and map the test patterns to the engineering site to form tunnel lining disease prevention and control measures.

[0050] Specifically, the tunnel lining-disease model poured in step S5 is placed in a dynamic-static load lining pressure testing machine. This machine consists of an electro-hydraulic servo pressure application system and a vibration system, which can simulate external forces such as surrounding rock pressure, seismic load, and traffic load. Through pressure testing, the development of defects in the model under the aforementioned load conditions is observed, and the predictive results of the main control defect development trend obtained in step S4 are verified using physical test data. Since the physical model and loading conditions are established based on actual working conditions, the defect development pattern after pressure testing can be directly mapped to the defect development situation at the actual engineering site. Based on this, targeted maintenance and reinforcement measures can be formulated to effectively control the development of tunnel defects.

[0051] Example 2 like Figure 5-8 As shown, based on the content of Example 1, which is a method for predicting and verifying the development of tunnel lining defects based on multi-source data-physical experiment fusion mapping, an experimental example is presented.

[0052] This embodiment takes a subway tunnel as the research object. The tunnel is located in a complex and loose geological section. The design grade of the lining concrete is C30 and the surrounding rock grade is V. It has been subjected to heavy train loads, groundwater seepage and seasonal temperature stress for a long time. It has developed complex diseases such as lining cracks, water leakage and arch voids. The method of this invention is needed to predict and verify the development of the diseases.

[0053] Step S1: The tunnel is a single-bore double-track tunnel with an inner lining diameter of 5.5m, a single ring lining width of 1.5m, and a shield segment arc of 0.5m.

[0054] Step S2: As Figure 5 As shown, Figure 5 This is a plan view of the tunnel surface. A dedicated mobile inspection vehicle for subways, equipped with multi-source acquisition equipment, was used for on-site inspection to avoid operational interference. The focus was on collecting surface cracking and related data. The acquisition scheme based on a line-scan camera unit was as follows: a combination of high-definition visible light and infrared cameras was used to acquire surface images along the tunnel's longitudinal direction, accurately identifying 896 cracks on the lining surface. Simultaneously, areas with dense cracks and associated traces of surface spalling and water leakage were marked; 132 areas showed water leakage, and 217 areas showed spalling defects. Table 1 shows the on-site inspection and acquisition information.

[0055] Table 1 Step S3: As Figure 6 As shown, Figure 6 The image shows the crack extraction results. A database of subway tunnel defects was constructed, and a semantic segmentation model was used to segment crack images. The detection data was standardized and quantized to extract key crack parameters.

[0056] For the construction of the subway tunnel defect database, the data source is: raw tunnel data collected by the system mentioned above, mainly images, with the tunnel being the one described in this embodiment. Labeling: Defect images are filtered, and defect areas are labeled based on labels. Preprocessing: Data enhancement processing such as scaling, rotation, and mirroring is applied to the images. Storage: The data is placed on the detection device during acquisition and transferred to a local data storage terminal after acquisition.

[0057] The semantic segmentation model is specifically the DSC_CrackU model optimized based on U-Net. The training strategy of DSC_CrackU is based on the U-Net architecture and is customized around the crack morphology: it uses dynamic serpentine convolution DSConv to adapt to slender and curved crack structures, combined with the feature fusion module FFM and the self-efficient channel module SCM to enhance feature extraction; it uses the binary cross-entropy (BCE) loss function to handle class imbalance, and the optimizer is AdamW with an initial learning rate of 0.01, dynamically adjusted via ReduceLROnPlateau. The training was conducted for 100 rounds on a 14900KF and RTX 4080 hardware environment with a batch size of 4, while also utilizing data enhanced by rotation, mirroring, and brightness adjustment.

[0058] The model evaluation criteria are constructed from four aspects: volumetric metrics, distance metrics, complexity metrics, and speed metrics. The volumetric metrics use mIoU and Dice coefficients to measure segmentation accuracy; the distance metrics use HD95 to characterize boundary matching; the complexity metrics use parameter count and FLOPs to evaluate lightweightness; and the speed metrics verify inference efficiency through throughput (FPS) and average processing time per sample. Furthermore, the evaluation criteria are compared with FCN and U... Comparison and ablation experiments with Net series, DeepLabv3+ and other models comprehensively verify the advantages of the model in terms of accuracy, lightweight and real-time performance.

[0059] Standardizing and quantifying the detection data involves converting pixels to the scale of actual physical space and then classifying them according to industry standards. The extracted key crack parameters include three categories: basic quantitative parameters, geometric dimensional parameters, and distribution and evaluation parameters, as follows: Basic quantitative parameters: number of single cracks, number of intersecting cracks, number of endpoints, number of intersections; Geometric dimensional parameters: total crack length, crack area, maximum crack width, crack density; Distribution and evaluation parameters: crack distribution characteristics, TCI value, D value.

[0060] Step S4: As Figure 7 As shown, Figure 7 (a), (b), and (c) are respectively field inspection diagrams, experimental diagrams, and simulation diagrams based on the method of the present invention. Based on the extracted crack parameters, such as length, width, angle, and lining section area, the development direction of the main controlling cracks is calculated using the TCI circle. The prediction results indicate that the main controlling defect in this lining section is radial surface cracking, with a crack propagation direction of 2.41°. The cracks will continue to extend along the cracking direction and widen, with some cracks penetrating and expanding into the lining, leading to a further decrease in the bearing capacity of the arch lining. This is consistent with the concentrated distribution and continuous expansion characteristics of surface cracking and radial cracks detected in the field.

[0061] Step S5: As Figure 8 The diagram shown illustrates the calculation results of this invention. A tunnel was constructed using a designed steel mold, and crack morphology molds were added to the mold during the pouring process to accurately simulate crack conditions. Concrete lining specimens were poured using materials such as cement mortar.

[0062] Step S6: Place the model in a dynamic-static load lining pressure testing machine to simulate the actual working conditions of the tunnel and monitor the surface crack propagation pattern. After the test, the crack propagation direction of the model surface was 0.64°, the deviation from the TCI circle prediction was 1.76°, and the deviation from the simulation was 0.34°. Table 2 is a comparison table of field detection, experimental, and simulation parameters.

[0063] Table 2 The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0064] The above embodiments are merely illustrative examples and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for assessing disease development based on multi-source data and data-physical fusion experiments, characterized in that, The method includes the following steps: Step S1: Collect basic tunnel working condition information, standardize the quantifiable working condition parameters, and obtain standardized tunnel working condition data. Step S2: Using the standardized tunnel working condition data as the detection benchmark, multi-source defect comprehensive acquisition equipment is used to acquire multi-source detection data of surface defects, internal defects, structural abnormalities and structural deterioration of tunnel lining, forming multi-source original detection data of lining. Step S3: Using the original multi-source detection data of the lining as a sample, a deep learning algorithm is used to construct a lining disease database and train a disease detection model. The disease detection model is used to complete disease identification and quantification to obtain quantified disease data. Step S4: Based on the quantified disease data, screen the main control diseases of the lining, calculate and predict the development direction and location of the main control diseases of the lining based on the TCI circle theory, and obtain the prediction results of the development tendency of the main control diseases of the lining. Step S5: Based on standardized tunnel working condition data and multi-source original detection data of lining, fabricate lining molds, select similar materials, and pour and cure a tunnel lining-disease model containing equivalent defects and damages. Step S6: Place the tunnel lining-disease model in a dynamic-static load lining pressure testing machine to conduct pressure tests simulating surrounding rock pressure, seismic load, and traffic load. Use the test disease development data to verify the predicted development trend of the main lining disease and map the test patterns to the engineering site to form tunnel lining disease prevention and control measures.

2. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, The multi-source disease comprehensive collection equipment in step S2 is composed of a mobile vehicle carrying a line array camera unit, a 3D laser scanner, a ground-penetrating radar, and a laser vibration meter.

3. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, The disease detection model described in step S3 is adapted to use a semantic segmentation model, a target detection model, or an instance segmentation model based on the disease characteristics.

4. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, The formula for calculating the TCI value of the lining crack based on the TCI circle theory in step S4 is as follows: ； In the formula, The area to be lining; The number of cracks; For the first The width of the crack; For the first The length of the crack; This represents the TCI component along the tunnel length direction. The radial component of TCI in the tunnel; This represents the total TCI value. The weighted correlation coefficient for crack width; This is the weighted correlation coefficient for crack length.

5. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 4, characterized in that, , The calculation formula is: ； ； In the formula, For the first Strip crack normal and The included angle of the axis; For the first Strip crack normal and The included angle of the axis.

6. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, In step S4, the TCI circle is constructed based on the second-order TCI tensor and with reference to the Mohr stress circle. The direction of the main crack development and the circle radius are determined by the geometric relationship of the TCI circle.

7. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, The lining mold mentioned in step S5 is a detachable splicing structure, and the tunnel lining-disease model is 1:10 in size ratio to the original tunnel.

8. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, In step S5, hollow spheres and steel sheets are placed inside the tunnel lining-disease model to proportionally recreate the actual defects and damage of the tunnel.

9. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, The dynamic-static load lining pressure testing machine described in step S6 is composed of an electro-hydraulic servo pressure application system and a vibration system.

10. The method for assessing disease development based on multi-source data and data-physical fusion experiments according to claim 1, characterized in that, In step S6, the physical test results are used to verify the prediction results of the main control disease development tendency of the lining, and the development of the main control disease of the lining is directly mapped to the actual disease development at the engineering site.

Images