Tunnel health state dynamic evaluation method and system based on simulation and data driving

Abstract

The application provides a tunnel health state dynamic evaluation method and system based on simulation and data driving, belongs to the technical field of highway tunnel engineering structure health monitoring and service performance evaluation, and comprises the following steps: collecting current tunnel structure apparent images and sensor monitoring data; establishing a physical simulation model to obtain underlying mechanical performance indexes; constructing a quantitative mapping mechanism between the underlying mechanical performance indexes and criterion layer data, converting the underlying mechanical performance indexes into criterion layer virtual disease data matrix, and performing feature alignment and data fusion to construct a multivariate double-driving time sequence feature data set; inputting into a time sequence prediction network based on a self-attention mechanism, and outputting prediction data of the index layer; calculating dynamic weight distribution coefficients of each criterion layer prediction data, calculating a tunnel health grade score, and obtaining a health state dynamic evaluation result. The application realizes scientific evaluation and prospective dynamic early warning of the tunnel health grade.

Description

Technical Field

[0001] This invention belongs to the technical field of structural health monitoring and service performance evaluation of highway tunnel engineering, and particularly relates to a dynamic evaluation method and system for tunnel health status based on simulation and data-driven approaches. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of transportation infrastructure, the number and operating time of highway tunnels are increasing daily. Due to the complex geological conditions, groundwater erosion, material aging, and the long-term coupling effects of vehicle dynamic loads, many operating tunnels are experiencing various structural defects, including lining cracks, water leakage, voids behind the lining, and lining deformation. The continuous accumulation of these defects not only damages the integrity and durability of the tunnel structure but can also lead to collapses and water inrushes, threatening traffic safety. Conducting scientific and accurate health assessments of tunnels can play a role in preventative maintenance, preventing safety accidents, and extending the service life of the structure. Traditional tunnel health assessment methods mostly rely on regular manual inspections or sampling with non-destructive testing equipment, which is not only time-consuming and labor-intensive but also makes it difficult to dynamically predict the evolution and development of tunnel defects based on static data analysis.

[0004] In recent years, machine learning and deep learning models have been introduced into the field of tunnel health assessment. These models, trained on large datasets, capture the complex nonlinear relationships between various disease indicators and structural health levels, thus improving assessment efficiency. However, existing data-driven assessment methods face the following technical bottlenecks in practical engineering applications: First, most existing machine learning models are purely data-driven, treating tunnel structures as "black boxes" without physical constraints. Because there are few actual measured samples of severely damaged tunnels on the verge of collapse in operating tunnels, and the distribution of field monitoring data is unbalanced, data-driven models often produce inaccurate predictions when faced with extreme conditions exceeding historical datasets, leading to insufficient model generalization ability. Second, purely data-driven models lack support from underlying physical and mechanical mechanisms. Cracks, water leakage, and spalling on the tunnel lining surface are essentially macroscopic physical manifestations of internal stress concentration, plastic strain accumulation, and the outward evolution of microscopic damage. Existing technologies ignore the relationship between underlying mechanical damage decay and apparent monitoring data, causing assessment results to often deviate from the actual mechanical evolution logic of engineering projects. Third, existing assessment systems mostly conduct static assessments of the cross-sectional state of a tunnel at a specific moment, resulting in significant lag in health assessment results, which makes it difficult to meet the needs of modern transportation hubs for advanced dynamic early warning of structural safety. Summary of the Invention

[0005] To overcome the shortcomings of the existing technologies, this invention proposes a dynamic assessment method and system for tunnel health status based on simulation and data-driven approaches. This addresses the technical deficiencies of existing pure data-driven assessment models, which are detached from the laws of physical and mechanical decay, prone to failure in extreme condition predictions, and lack of forward-looking early warning capabilities in static assessments. By establishing a quantitative mathematical mapping mechanism between multi-physics simulation mechanical indicators and apparent criterion layer data, and combining it with a self-attention temporal prediction network, this invention deeply mines the temporal evolution laws and coupling aggravation effects of multi-source disease characteristics in the time dimension, thereby achieving scientific assessment of health levels and forward-looking dynamic early warning.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, this invention discloses a dynamic assessment method for tunnel health status based on simulation and data-driven approaches, comprising: A hierarchical model for tunnel health assessment is constructed, defining the target layer, criterion layer, and indicator layer from top to bottom. Collect current tunnel structure appearance images and sensor monitoring data; A three-dimensional fluid-structure interaction physical simulation model including the surrounding rock and secondary lining structure was established using finite element analysis software to obtain multi-dimensional bottom mechanical performance indicators of the tunnel structure under different levels of adverse chemical conditions. A quantitative mapping mechanism is constructed between the underlying mechanical performance indicators and the criteria layer data. The underlying mechanical performance indicators are transformed into a virtual disease data matrix of the criteria layer. The virtual disease data matrix of the criteria layer is then aligned with the real data matrix of the current indicator layer and fused with features to construct a multivariate dual-drive time-series feature dataset. The dual-drive time series feature dataset is input into the trained time series prediction network based on the self-attention mechanism, and the prediction data of various indicator layers for the future prediction period are output. The dynamic weight allocation coefficients of the prediction data of each criterion layer are calculated by using a combination of subjective and objective weighting methods. The tunnel health level score is calculated by combining the prediction data. Based on the level score, the dynamic assessment result of the current health status of the tunnel in the future prediction period is obtained.

[0007] Secondly, this invention discloses a dynamic assessment system for tunnel health status based on simulation and data-driven methods, comprising: The model building module is used to construct a hierarchical model for tunnel health assessment, defining the target layer, criterion layer, and indicator layer from top to bottom. The data acquisition module is used to collect current tunnel structure appearance images and sensor monitoring data; The physical simulation module is used to establish a three-dimensional fluid-structure interaction physical simulation model including the surrounding rock and secondary lining structure using finite element analysis software, and to obtain multi-dimensional underlying mechanical performance indicators of the tunnel structure under different levels of adverse chemical conditions. The feature construction module is used to construct a quantitative mapping mechanism between the underlying mechanical performance indicators and the criterion layer data. It transforms the underlying mechanical performance indicators into a criterion layer virtual disease data matrix, and performs feature alignment and data fusion between the criterion layer virtual disease data matrix and the current indicator layer real data matrix to construct a multivariate dual-driven time-series feature dataset. The indicator prediction module is used to input the dual-drive time series feature dataset into the trained time series prediction network based on the self-attention mechanism, and output the prediction data of each indicator layer for the future prediction period. The dynamic evaluation module is used to calculate the dynamic weight allocation coefficients of the prediction data of each criterion layer using a combination of subjective and objective weighting methods, calculate the tunnel health level score in combination with the prediction data, and obtain the dynamic evaluation result of the current health status of the tunnel in the future prediction period based on the level score.

[0008] Thirdly, the present invention discloses an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, complete the steps of the above-mentioned dynamic assessment method for tunnel health status based on simulation and data-driven methods.

[0009] Fourthly, the present invention discloses a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-mentioned dynamic assessment method for tunnel health status based on simulation and data-driven methods.

[0010] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention creatively proposes to establish a quantitative communication bridge between the underlying mechanical decay mechanism and apparent monitoring data using multiphysics simulation methods such as three-dimensional fluid-structure interaction. By extracting multi-dimensional underlying physical simulation indicators such as tensile and compressive damage factors, pore water pressure gradient, and normal opening displacement of the contact surface, a precise mathematical mapping is achieved with apparent defects such as cracks, water leakage, and underlying cavities. This mechanism injects prior physical knowledge consistent with the logic of structural engineering and geotechnical mechanics into the purely data-driven evaluation model.

[0011] The dual-drive mode of physical simulation and real data constructed in this invention effectively supplements the data shortage of severely damaged fault samples in field measurements by using virtual samples of extreme adverse working conditions generated by physical simulation. It fundamentally solves the technical pain point of unbalanced high-risk data in the engineering testing field and significantly improves the robustness and generalization ability of time series prediction networks in dealing with complex and harsh working conditions and unknown deformation evolution.

[0012] This invention represents a technological advancement in the assessment of the health status of highway and railway tunnels, moving from "post-hoc static slicing and scoring" to "pre-hoc dynamic time-series prediction." By using a dual-driven time-series dataset to power a self-attention mechanism network, it outputs quantitative indicators of future tunnel deterioration. Furthermore, it incorporates a modified hierarchical analysis weighting method combining subjective and objective approaches to calculate dynamic health scores. This allows for accurate and forward-looking early warnings of tunnel health degradation trends, providing transportation management departments with ample time to develop preventative maintenance and reinforcement plans.

[0013] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0014] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0015] Figure 1 This is a flowchart of the dynamic assessment method for tunnel health status based on simulation and data-driven approaches described in Embodiment 1 of the present invention.

[0016] Figure 2 This is a hierarchical structure diagram of tunnel service status assessment as described in Embodiment 1 of the present invention.

[0017] Figure 3 This is a schematic diagram illustrating the quantitative mapping mechanism between the underlying mechanical performance indicators and the criterion layer data as described in Embodiment 1 of the present invention. Detailed Implementation

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0020] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0021] Example 1 In one or more embodiments, a dynamic assessment method for tunnel health status based on simulation and data-driven approaches is disclosed, such as... Figure 1 As shown, it includes the following steps: Step S1: Construct a hierarchical model for tunnel health assessment, defining the target layer, criterion layer, and indicator layer from top to bottom.

[0022] Specifically, in establishing a dynamic assessment index system for tunnel health status, the principles of scientific rigor, completeness, and independence are followed to overcome the problems of overly complex and overlapping dimensions in traditional factor layer divisions. The Analytic Hierarchy Process (AHP) is used to simplify and precisely divide the entire index system into three levels, as shown in the tunnel service status assessment hierarchy. Figure 2 As shown, specifically: The first layer is the target layer, which is the overall health level of the tunnel in service. According to the current tunnel maintenance technical specifications, a four-level classification method is adopted, which strictly defines the overall health status of the tunnel structure as Level 1 (no defects or minor defects, in a safe state), Level 2 (general defects, requiring enhanced monitoring), Level 3 (relatively serious defects, requiring planned maintenance) and Level 4 (serious defects, requiring immediate reinforcement and emergency repair measures).

[0023] The second layer is the criteria layer, which covers five core defect dimensions that directly affect the safety and durability of tunnel structures. Specifically, these are: lining cracking, water leakage, material deterioration, voids behind the lining, surface spalling, and lining deformation.

[0024] The third layer is the indicator layer, which performs high-resolution quantitative decomposition of the criterion layer. The indicator layer is set with 12 monitoring indicators: crack length, crack width, leakage area, leakage status, lining strength ratio, lining thickness, void area, void depth; spalling area, spalling depth, lining radial deformation, and longitudinal misalignment.

[0025] Furthermore, the indicator layer data corresponding to each criterion layer includes: Lining cracking: crack length and crack width; Water leakage problems: leakage area, leakage condition; Material degradation: lining strength ratio, lining thickness ratio; Voiding behind lining: void area and void depth; Surface peeling: peeling area, peeling depth; Lining deformation: radial deformation and longitudinal misalignment of the lining.

[0026] Step S2: Collect the current tunnel structure appearance image and sensor monitoring data, use computer vision algorithms to extract and quantify the current indicator layer real data, and obtain the current criterion layer real data matrix.

[0027] Specifically, during the routine operation phase of the tunnel, a comprehensive inspection vehicle equipped with a high-definition industrial line array camera and a 3D laser scanner is used to collect panoramic cloud images and high-resolution images of the tunnel's secondary lining surface, and simultaneously read time-series data from distributed sensor networks such as displacement gauges and piezometers.

[0028] On the data processing side, conventional camera calibration techniques are used to determine the pixel equivalent of the image. A deep convolutional neural network is then used for instance segmentation and edge detection of the apparent disease area, statistically analyzing the pixel distribution of disease features. This allows for the automated calculation of the actual topological length and maximum crack width of the lining surface cracks, as well as the actual physical area of ​​water leakage and surface spalling. It should be understood that the quantification of the monitoring data employs image recognition techniques from deep learning to transform unquantifiable monitoring data into intuitive quantitative data. All 12 real indicator layer data extracted at the current moment are normalized to construct the real state feature vector for the current moment. .

[0029] Step S3: Use finite element analysis software to establish a three-dimensional fluid-structure interaction physical simulation model that includes the surrounding rock and secondary lining structure, and obtain multi-dimensional bottom mechanical performance indicators of the tunnel structure under different levels of adverse chemical conditions.

[0030] In this embodiment, to address the problem of data imbalance caused by the severe lack of extreme severe disease samples (i.e., high-risk data) in the real monitoring database, a three-dimensional fluid-structure interaction simulation model including the surrounding rock strata, the initial support system, and the secondary lining is established using nonlinear finite element software such as ABAQUS.

[0031] The stress simulation of concrete materials adopts the plastic damage constitutive model (CDP), which introduces a tensile damage factor. With pressure damage factor This method accurately characterizes the tensile and compressive stiffness degradation behavior of materials under complex stress states. In the finite element boundary condition settings, external working conditions ranging from slight degradation to extreme severity are artificially applied, including excessive asymmetric bias pressure on the surrounding rock, sudden rise and fall of groundwater level, and large-scale uneven settlement of the foundation. A fully coupled transient implicit solution is then performed to obtain the mechanical performance indicators of the underlying layer.

[0032] Step S4: Construct a quantitative mapping mechanism between the underlying mechanical performance indicators and the criterion layer data, transform the underlying mechanical performance indicators into a criterion layer virtual disease data matrix, and perform feature alignment and data fusion between the criterion layer virtual disease data matrix and the current indicator layer real data matrix in the time series to construct a multivariate dual-driven time series feature dataset.

[0033] Constructing a quantitative mapping mechanism between underlying mechanical performance indicators and indicator layer data, including: Extract the distribution cloud map of tensile damage factor in the lining space field of the three-dimensional fluid-structure interaction physical simulation model, and reconstruct the geometric features and volume integral of the continuous mesh element set that exceeds the set critical damage threshold, and map it into virtual lining crack length and width data. The pore water pressure gradient and local seepage velocity vector in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual seepage area and state data using Darcy's law of permeability. Extract the lining stiffness degradation parameters and failure element erosion depth from the three-dimensional fluid-structure interaction physical simulation model, and map them into virtual lining strength ratio and lining thickness ratio data. The normal opening displacement and contact stress of the contact surface element between the lining and the surrounding rock in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual void depth data behind the lining. The compressive damage factor and equivalent plastic compressive strain concentration region in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual surface spalling area and depth data. The relative displacement vectors and displacement time history derivatives of each calculation node in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual radial deformation and longitudinal misalignment data of the lining.

[0034] In this embodiment, a quantitative mapping mechanism is constructed between the underlying mechanical performance indicators and the criterion layer data, such as... Figure 3 As shown, specifically: First, the tensile damage factor in the spatial field of the lining grid is extracted. Set the critical damage threshold for macroscopic cracking. When the cell contains At that point, mechanically, it was determined that the concrete had completely lost its tensile strength and macroscopic cracking had occurred. Extract all continuous damaged mesh element sets that meet the conditions. By integrating the volume elements, it is precisely mapped to the virtual equivalent length of the apparent crack. Equivalent width of crack :

[0035]

[0036] In the formula, For equivalent plastic strain, The shape function is used to transform the damage factor and plastic principal strain into crack geometric parameters. The characteristic length of the mesh element that undergoes macroscopic cracking. The equivalent plastic tensile strain is used. The crack length is obtained by volume element integration, and the crack width is obtained by multiplying the maximum equivalent plastic strain by the element characteristic length.

[0037] Secondly, the pore water pressure gradient field obtained from the fluid-structure interaction model is extracted. According to Darcy's Law ,in Given the permeability coefficient tensor, calculate the local seepage velocity vector. A critical infiltration velocity was set to characterize the water flow rate exceeding the threshold of the drainage system. The area of ​​the region with an abnormally high flow velocity is integrally mapped and transformed into a virtual leakage area. :

[0038] In the formula, This refers to the inner surface area of ​​the lining. For surface pore water pressure, Let the area be a micro-element. The local maximum seepage velocity within this leakage area. and in accordance with the flow rate thresholds set by current standards ( and area threshold The flow velocity and the integral area are determined by a conditional decision function. The joint mapping is applied to specific apparent leakage conditions (defining a set of conditions {wetting, dripping, linear flow, gushing}), and the specific mapping logic is as follows: (1) When and When the mapping is 'wet', it is determined that the state is 'wet'. (2) When When the time is right, the mapping is determined to be in the 'drip' state; (3) When When the time is right, the mapping is determined to be in the 'linear flow' state; (4) When or The mapping is determined to be in the 'surging' state; Third, the tensile-compressive combined stiffness degradation parameter SDEG and the normal erosion depth of the bearing capacity failure elements (i.e., mesh elements with SDEG approaching 1) of the lining material elements in the three-dimensional fluid-structure interaction physical simulation model are extracted. Virtual lining strength ratio. Compared to thickness The mapping formula is:

[0039]

[0040] In the formula, For the lining volume, For the initial design thickness, For cross-sectional loss thickness, To meet The spatial distribution length of the extremely severe damage element along the cross-section normal.

[0041] Fourth, extract the normal opening displacement of the contact elements between the lining and the surrounding rock in the model. Contact stress attenuation rate with normal direction The mechanical condition for determining whether a gap occurs at the contact surface is: and Virtual void area With cavity depth The mapping formula is:

[0042]

[0043] In the formula, The total contact area, Let be the contact surface area of ​​a micro-element.

[0044] Fifth, extract the pressure damage factor from the model. Set the concrete crush damage threshold. Extract the surface layer crushed mesh cell set and map it into a virtual surface peeling area. With peeling depth :

[0045]

[0046] In the formula, For lining the inner surface, The geometric intrusion depth of the crush damage area along the lining normal is given.

[0047] Sixth, extract the relative displacement vector matrix of each computation node in the model space network in the global coordinate system. Virtual radial deformation of the lining Longitudinal misalignment The mapping formula is:

[0048]

[0049] In the formula, This is the radial unit vector from the cross-sectional measurement point to the geometric center of the tunnel; and For areas where longitudinal uneven settlement occurs, the spatial displacement vector of the corresponding nodes at the edges of two adjacent lining segments.

[0050] All virtual apparent defects obtained from physical simulation, including those under extreme adverse working conditions, are defined as virtual state feature vectors. .

[0051] Furthermore, the underlying mechanical performance indicators are transformed into a virtual disease data matrix at the indicator layer, and then fused and spliced ​​with the current real data matrix at the criterion layer. Specifically, this includes: The real data of the current criterion layer under normal or slightly damaged conditions extracted by computer vision and sensors is defined as the real feature state vector; the virtual disease data of the criterion layer generated by the three-dimensional fluid-structure interaction physical simulation model under multiple sets of extreme and severe boundary conditions is defined as the virtual feature state vector. By aligning the real feature state vector and the virtual feature state vector at the same time step and assembling them into a matrix, a dual-driven time-series feature dataset is constructed to compensate for the data imbalance defect in the actual monitoring database, which is extremely scarce in severely damaged samples.

[0052] Step S5: Input the dual-drive time series feature dataset into the trained time series prediction network based on the self-attention mechanism, explore the nonlinear coupling law between the simulated mechanical evolution data and the real apparent data in the time dimension, and output the prediction data of various indicator layers for the future prediction period.

[0053] The real-state feature vector containing actual environmental noise obtained from the on-site environment. , and the virtual state feature vector containing the underlying mechanical mechanism evolution logic In time step Dimensional alignment and matrix concatenation are then performed. This dual-drive combination effectively compensates for the long-tail defect of a single data stream, constructing a highly robust single-step fusion feature matrix. Selecting continuous past data A time window is formed to create a time-series input sequence matrix. ,in, The feature dimensions after fusion.

[0054] Will The input is fed into a time-series prediction network (Transformer architecture) based on a self-attention mechanism. The sequence matrix first preserves the temporal information of disease evolution through a positional encoding matrix, and then is input into the core computational layer of the multi-head attention mechanism.

[0055] In the formula, , , These are the query matrix, key matrix, and value matrix after linear mapping of the input sequence, respectively. The dimension of the key matrix is ​​given. This mechanism overcomes the limitation of traditional recurrent neural networks (RNNs) facing the vanishing gradient problem in long sequences, and can adaptively calculate and identify the nonlinear coupling aggravation effect between different types of disease indicators on the time axis from a global perspective. After multi-layer self-attention encoding and autoregressive decoding calculations, the network finally outputs a tunnel to the future time. Quantitative index vectors for predicting various disease criteria .

[0056] Step S6: Calculate the dynamic weight allocation coefficients of the prediction data of each criterion layer using the subjective and objective combined weighting method, calculate the tunnel health level score in combination with the prediction data, map the score to the preset four-level health level, and output the dynamic assessment result of the current tunnel health status in the future prediction period.

[0057] In this embodiment, the dynamic weight allocation coefficients of the prediction data for each criterion layer are calculated using a subjective-objective combined weighting method, and the dynamic comprehensive health score of the tunnel is calculated in combination with the prediction data, including: By using expert scoring to establish a judgment matrix based on the relative importance of each evaluation indicator, solving the matrix eigenvectors and performing consistency checks, the subjective weight coefficients of each indicator are obtained. Based on the dispersion and information distribution characteristics of various indicators in the dual-drive time series feature dataset within the sample time window, the objective entropy weight coefficient is calculated using the principle of information entropy. By introducing a weight preference allocation coefficient, the subjective weight coefficient and the objective entropy weight coefficient are linearly combined to obtain the dynamic combined weight of each indicator. After standardizing and normalizing the degradation degree of the prediction data of each criterion layer in the future prediction period, the data is multiplied by the dynamic combination weights corresponding to each indicator and then summed layer by layer to obtain the dynamic comprehensive health score of the tunnel.

[0058] Specifically, based on the improved analytic hierarchy process (AHP) with a combination of subjective and objective weighting, the dynamic comprehensive health score and level of the tunnel are calculated, thereby avoiding the bias caused by the traditional AHP's over-reliance on expert subjective experience. This embodiment adopts an improved evaluation system with a combination of subjective and objective information weighting.

[0059] First, experts in tunnel engineering and geotechnical mechanics were invited to establish a subjective judgment matrix based on the impact of each assessment criterion on structural safety. This matrix was then used to verify consistency (i.e., consistency ratio). Under the premise of solving the matrix, the eigenvector corresponding to the largest eigenvalue is obtained, and after normalization, the subjective weight vector of each indicator is obtained. .

[0060] Secondly, based on the data dispersion of various indicators in the dual-driven time-series feature dataset within the sample time window, objective weighting is performed using the principle of information entropy. The calculation of the... Information entropy of the indicator :

[0061] In the formula, The total number of samples, For the first The sample at the th The contribution of features under each indicator. Based on this, and after normalization, an objective entropy weight vector driven purely by data distribution is derived. .

[0062] Introducing dynamic preference allocation coefficients Preferably, the value is between 0.5 and 0.6. The two weights mentioned above are linearly combined and weighted to obtain the dynamic combined weights of each disease index. :

[0063] In the formula, For the first Subjective weighting of each indicator; For the first Objective entropy weight of the indicator.

[0064] Obtain the vectors of predicted disease indicators for future moments from the network output in step S5. Based on current transportation industry engineering standards, this is converted into a dimensionless standardized degradation score matrix on a percentage basis. Ultimately, the tunnel at the target future time... Dynamic comprehensive health score The calculation is as follows:

[0065] In the formula, For dynamic weight combination, This is the normalized degradation score.

[0066] Dynamic comprehensive health score Mapped to a preset threshold evaluation range, the system automatically outputs a comprehensive health level warning result for the target layer in future time periods. Preferably, the definition is: Divided into Level 1 security status; Divided into 2 levels; Divided into 3 levels; It is divided into four levels of high-risk status.

[0067] This approach, based on both physical simulation and data-driven methods, connects the entire evaluation chain from "identifying underlying mechanical mechanisms" to "predicting future apparent states." It can accurately predict the deterioration inflection point of tunnel service performance months in advance, providing a valuable safety window for traffic management and decision-making departments to implement preventive maintenance and reinforcement.

[0068] Example 2 In one or more embodiments, a dynamic assessment system for tunnel health status based on simulation and data-driven methods is disclosed, specifically including: The model building module is used to construct a hierarchical model for tunnel health assessment, defining the target layer, criterion layer, and indicator layer from top to bottom. The data acquisition module is used to collect current tunnel structure appearance images and sensor monitoring data; The physical simulation module is used to establish a three-dimensional fluid-structure interaction physical simulation model including the surrounding rock and secondary lining structure using finite element analysis software, and to obtain multi-dimensional underlying mechanical performance indicators of the tunnel structure under different levels of adverse chemical conditions. The feature construction module is used to construct a quantitative mapping mechanism between the underlying mechanical performance indicators and the criterion layer data. It transforms the underlying mechanical performance indicators into a criterion layer virtual disease data matrix, and performs feature alignment and data fusion between the criterion layer virtual disease data matrix and the current indicator layer real data matrix to construct a multivariate dual-driven time-series feature dataset. The indicator prediction module is used to input the dual-drive time series feature dataset into the trained time series prediction network based on the self-attention mechanism, and output the prediction data of each indicator layer for the future prediction period. The dynamic evaluation module is used to calculate the dynamic weight allocation coefficients of the prediction data of each criterion layer using a combination of subjective and objective weighting methods, calculate the tunnel health level score in combination with the prediction data, and obtain the dynamic evaluation result of the current health status of the tunnel in the future prediction period based on the level score.

[0069] Example 3 This embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps of the above-mentioned dynamic assessment method for tunnel health status based on simulation and data-driven methods.

[0070] Example 4 This embodiment provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described dynamic assessment method for tunnel health status based on simulation and data-driven approaches.

[0071] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0072] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0074] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0075] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic assessment method for tunnel health status based on simulation and data-driven approaches, characterized in that, include: A hierarchical model for tunnel health assessment is constructed, defining the target layer, criterion layer, and indicator layer from top to bottom. Collect current tunnel structure appearance images and sensor monitoring data; A three-dimensional fluid-structure interaction physical simulation model including the surrounding rock and secondary lining structure was established using finite element analysis software to obtain multi-dimensional bottom mechanical performance indicators of the tunnel structure under different levels of adverse chemical conditions. A quantitative mapping mechanism is constructed between the underlying mechanical performance indicators and the criteria layer data. The underlying mechanical performance indicators are transformed into a virtual disease data matrix of the criteria layer. The virtual disease data matrix of the criteria layer is then aligned with the real data matrix of the current indicator layer and fused with features to construct a multivariate dual-drive time-series feature dataset. The dual-drive time series feature dataset is input into the trained time series prediction network based on the self-attention mechanism, and the prediction data of various indicator layers for the future prediction period are output. The dynamic weight allocation coefficients of the prediction data of each criterion layer are calculated by using a combination of subjective and objective weighting methods. The tunnel health level score is calculated by combining the prediction data. Based on the level score, the dynamic assessment result of the current health status of the tunnel in the future prediction period is obtained.

2. The method for dynamic assessment of tunnel health status based on simulation and data-driven approaches as described in claim 1, characterized in that, The target layer is the comprehensive health level of the tunnel's service status; The criteria layer includes: lining cracking, water leakage, material deterioration, voids behind the lining, surface spalling, and lining deformation. The indicator layer quantifies and decomposes the criteria layer; the indicator layer data corresponding to each criteria layer includes: lining cracking: crack length, crack width; water leakage: water leakage area, water leakage status; material deterioration: lining strength ratio, lining thickness ratio; voids behind the lining: void area, void depth; surface spalling: spalling area, spalling depth; lining deformation: radial deformation of the lining, longitudinal misalignment.

3. The method for dynamic assessment of tunnel health status based on simulation and data-driven approaches as described in claim 1, characterized in that, Constructing a quantitative mapping mechanism between underlying mechanical performance indicators and indicator layer data specifically includes: Extract the distribution cloud map of tensile damage factor in the lining space field of the three-dimensional fluid-structure interaction physical simulation model, and reconstruct the geometric features and volume integral of the continuous mesh element set that exceeds the set critical damage threshold, and map it into virtual lining crack length and width data. The pore water pressure gradient and local seepage velocity vector in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual seepage area and state data using Darcy's law of permeability. Extract the lining stiffness degradation parameters and failure element erosion depth from the three-dimensional fluid-structure interaction physical simulation model, and map them into virtual lining strength ratio and lining thickness ratio data. The normal opening displacement and contact stress of the contact surface element between the lining and the surrounding rock in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual void depth data behind the lining. The compressive damage factor and equivalent plastic compressive strain concentration region in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual surface spalling area and depth data. The relative displacement vectors and displacement time history derivatives of each calculation node in the three-dimensional fluid-structure interaction physical simulation model are extracted and mapped into virtual radial deformation and longitudinal misalignment data of the lining.

4. The method for dynamic assessment of tunnel health status based on simulation and data-driven approaches as described in claim 3, characterized in that, It is precisely mapped to the virtual apparent crack equivalent length by volume element integration. Equivalent width of crack : In the formula, For equivalent plastic strain, The shape function is used to transform the damage factor and plastic principal strain into crack geometric parameters. The characteristic length of the mesh element that undergoes macroscopic cracking. This is the equivalent plastic tensile strain.

5. The method for dynamic assessment of tunnel health status based on simulation and data-driven approaches as described in claim 1, characterized in that, The underlying mechanical performance indicators are transformed into a virtual disease data matrix at the indicator layer, and then fused and spliced ​​with the current real data matrix at the criterion layer. Specifically, this includes: The real data of the current criterion layer under normal or slightly damaged conditions extracted by computer vision and sensors is defined as the real feature state vector; the virtual disease data of the criterion layer generated by the three-dimensional fluid-structure interaction physical simulation model under multiple sets of extreme and severe boundary conditions is defined as the virtual feature state vector. By aligning the real feature state vector and the virtual feature state vector at the same time step and assembling them into matrices, a dual-drive time-series feature dataset is constructed.

6. The method for dynamic assessment of tunnel health status based on simulation and data-driven approaches as described in claim 1, characterized in that, The process of calculating dynamic weight allocation coefficients for each criterion-level prediction data using a subjective-objective combined weighting method, calculating a tunnel health level score based on the prediction data, and obtaining a dynamic assessment result of the tunnel's current health status for the future prediction period based on the score specifically includes: By using expert scoring to establish a judgment matrix based on the relative importance of each evaluation indicator, solving the matrix eigenvectors and performing consistency checks, the subjective weight coefficients of each indicator are obtained. Based on the dispersion and information distribution characteristics of various indicators in the dual-drive time series feature dataset within the sample time window, the objective entropy weight coefficient is calculated using the principle of information entropy. By introducing a weight preference allocation coefficient, the subjective weight coefficient and the objective entropy weight coefficient are linearly combined to obtain the dynamic combined weight of each indicator. After standardizing and normalizing the degradation degree of the prediction data of each criterion layer in the future prediction period, the data is multiplied by the dynamic combination weights corresponding to each indicator and then summed layer by layer to obtain the dynamic comprehensive health score of the tunnel.

7. The method for dynamic assessment of tunnel health status based on simulation and data-driven approaches as described in claim 6, characterized in that, Future target time The tunnel dynamic comprehensive health score is calculated as follows: in, This is the normalized degradation score; For dynamic comprehensive health scoring; The dynamic combination weights of various disease indicators.

8. A dynamic assessment system for tunnel health status based on simulation and data-driven methods, characterized in that, include: The model building module is used to construct a hierarchical model for tunnel health assessment, defining the target layer, criterion layer, and indicator layer from top to bottom. The data acquisition module is used to collect current tunnel structure appearance images and sensor monitoring data; The physical simulation module is used to establish a three-dimensional fluid-structure interaction physical simulation model including the surrounding rock and secondary lining structure using finite element analysis software, and to obtain multi-dimensional underlying mechanical performance indicators of the tunnel structure under different levels of adverse chemical conditions. The feature construction module is used to construct a quantitative mapping mechanism between the underlying mechanical performance indicators and the criterion layer data. It transforms the underlying mechanical performance indicators into a criterion layer virtual disease data matrix, and performs feature alignment and data fusion between the criterion layer virtual disease data matrix and the current indicator layer real data matrix to construct a multivariate dual-driven time-series feature dataset. The indicator prediction module is used to input the dual-drive time series feature dataset into the trained time series prediction network based on the self-attention mechanism, and output the prediction data of each indicator layer for the future prediction period. The dynamic evaluation module is used to calculate the dynamic weight allocation coefficients of the prediction data of each criterion layer using a combination of subjective and objective weighting methods, calculate the tunnel health level score in combination with the prediction data, and obtain the dynamic evaluation result of the current health status of the tunnel in the future prediction period based on the level score.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the simulation- and data-driven dynamic assessment method for tunnel health status as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the dynamic assessment method for tunnel health status based on simulation and data-driven methods as described in any one of claims 1-7.

Images