Multi-source tunnel monitoring data acquisition and processing method and system

Abstract

The invention provides a multi-source tunnel monitoring data acquisition and processing method and system, and relates to the technical field of tunnel monitoring data processing, and the method comprises the steps: obtaining audio frequency resonance spectrum data of a tunnel lining; determining a plurality of audio frequency feedback standard sections and a plurality of audio frequency feedback variation sections based on the audio frequency resonance spectrum data of the tunnel lining; determining a plurality of degradation feature aggregation segments and a void anomaly mapping graph of each degradation feature aggregation segment based on the plurality of audio frequency feedback standard segments and the plurality of audio frequency feedback variation segments; obtaining mechanical impedance sampling information of each collaborative mechanical impedance sampling point of each degradation feature gathering section; and determining the grouting repair entity cavity of the tunnel lining based on the plurality of mechanical impedance sampling points of each degradation characteristic gathering section and the plurality of collaborative mechanical impedance sampling points of each degradation characteristic gathering section. According to the method, the grouting repair entity cavity caused by the tunnel lining void disease can be efficiently and accurately determined.

Description

Technical Field

[0001] This invention relates to the field of tunnel monitoring data processing technology, and in particular to a method and system for acquiring and processing multi-source tunnel monitoring data. Background Technology

[0002] Tunnel lining health monitoring is a core component of ensuring the safe operation and maintenance of transportation infrastructure. The accurate identification and location of internal voids directly impacts the scientific validity of long-term stability assessments and maintenance decisions for tunnel linings. Currently, precision monitoring equipment based on physical principles such as mechanical impedance testing is widely used in lining quality assessment, enabling effective detection of defects in localized areas of the tunnel lining. However, these methods still have systemic limitations in practical applications. Faced with the massive monitoring needs of tunnels, existing methods typically rely on uniformly distributed or equally spaced inspection strategies. While this ensures the accuracy of individual point data, it results in a huge volume of data collected across the entire tunnel domain, high monitoring costs, and long operation cycles. Furthermore, existing technologies lack progressive screening and verification mechanisms, making it difficult to quickly focus on and accurately locate cavities caused by internal voids in the tunnel lining. This leads to a lack of precise basis for repair decisions, failing to meet the demands of efficient, low-cost, and precise safe operation and maintenance of tunnels.

[0003] Therefore, how to efficiently and accurately determine the cavities in the tunnel lining caused by voids and require grouting repair is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to efficiently and accurately determine the cavities in the tunnel lining caused by voids and require grouting repair.

[0005] According to a first aspect, the present invention provides a method for acquiring and processing multi-source tunnel monitoring data, comprising: acquiring acoustic resonance spectrum data of the tunnel lining; determining multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining; determining multiple deterioration feature clustering segments and a voiding anomaly mapping map of each deterioration feature clustering segment based on the multiple acoustic feedback standard segments and the multiple acoustic feedback variation segments; determining multiple mechanical impedance sampling points of each deterioration feature clustering segment based on the voiding anomaly mapping map of each deterioration feature clustering segment, and acquiring mechanical impedance sampling information of each mechanical impedance sampling point of each deterioration feature clustering segment; and based on each mechanical impedance sampling point of each deterioration feature clustering segment... The mechanical impedance sampling information of the mechanical impedance sampling points determines multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment; based on the multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment, multiple cooperative mechanical impedance sampling points are determined for each deterioration feature cluster segment, and the mechanical impedance sampling information of each cooperative mechanical impedance sampling point in each deterioration feature cluster segment is obtained; based on the multiple mechanical impedance sampling points and multiple cooperative mechanical impedance sampling points in each deterioration feature cluster segment, the grouting repair solid cavity of the tunnel lining is determined.

[0006] In one possible implementation, determining multiple degradation feature clusters and a voiding anomaly mapping for each degradation feature cluster based on the multiple audio feedback standard segments and the multiple audio feedback aberration segments includes: constructing an audio feedback spectrum, which includes multiple nodes and multiple edges between the nodes; the multiple nodes include multiple audio feedback standard segment nodes and multiple audio feedback aberration segment nodes; each audio feedback aberration segment node is connected to multiple audio feedback standard segment nodes; the node feature of each audio feedback standard segment node is the audio resonance spectrum data of each audio feedback standard segment; and the node feature of each audio feedback aberration segment node is the audio resonance spectrum data of each audio feedback aberration segment; and processing the audio feedback spectrum based on a graph neural network to determine the multiple degradation feature clusters and the voiding anomaly mapping for each degradation feature cluster.

[0007] In one possible implementation, determining the grouting repair cavity of the tunnel lining based on multiple mechanical impedance sampling points of each deterioration feature cluster segment and multiple collaborative mechanical impedance sampling points of each deterioration feature cluster segment includes: clustering the mechanical impedance sampling information of each mechanical impedance sampling point of each deterioration feature cluster segment and the mechanical impedance sampling information of each collaborative mechanical impedance sampling point of each deterioration feature cluster segment to obtain multiple impedance response association clusters for each deterioration feature cluster segment, each impedance response association cluster containing mechanical impedance sampling information of multiple mechanical impedance sampling points and multiple collaborative mechanical impedance sampling points; based on the multiple impedance sampling points of each deterioration feature cluster segment... Multiple core deterioration feature clusters are identified by the anti-response correlation cluster and the void anomaly mapping map of each deterioration feature cluster. Based on the mechanical impedance sampling information of multiple mechanical impedance sampling points of each core deterioration feature cluster, the mechanical impedance sampling information of multiple cooperative mechanical impedance sampling points of each core deterioration feature cluster, and the acoustic resonance spectrum data of each core deterioration feature cluster, multiple suspected cavity anomaly regions of each core deterioration feature cluster are identified, and ground-penetrating radar analysis data of multiple suspected cavity anomaly regions of each core deterioration feature cluster are obtained. Based on the ground-penetrating radar analysis data of multiple suspected cavity anomaly regions of each core deterioration feature cluster, the grouting repair solid cavity of the tunnel lining is determined.

[0008] In one possible implementation, the input of the graph neural network is the audio feedback spectrum, and the output of the graph neural network is multiple deterioration feature clusters and a void anomaly mapping map of each deterioration feature cluster.

[0009] According to a second aspect, the present invention provides a multi-source tunnel monitoring data acquisition and processing system, comprising: an acquisition module for acquiring acoustic resonance spectrum data of tunnel lining; an acoustic feedback analysis module for determining multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining; a feature recognition module for determining multiple deterioration feature clusters and a voiding anomaly mapping map of each deterioration feature cluster based on the multiple acoustic feedback standard segments and the multiple acoustic feedback variation segments; a mechanical impedance sampling module for determining multiple mechanical impedance sampling points of each deterioration feature cluster based on the voiding anomaly mapping map of each deterioration feature cluster, and acquiring mechanical impedance sampling information of each mechanical impedance sampling point of each deterioration feature cluster; and an impedance feature analysis module for... Based on the mechanical impedance sampling information of each mechanical impedance sampling point in each deterioration feature cluster segment, multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment are determined; a collaborative sampling module is used to determine multiple collaborative mechanical impedance sampling points in each deterioration feature cluster segment based on the multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment, and to acquire the mechanical impedance sampling information of each collaborative mechanical impedance sampling point in each deterioration feature cluster segment; a cavity determination module is used to determine the grouting repair entity cavity of the tunnel lining based on the multiple mechanical impedance sampling points and the multiple collaborative mechanical impedance sampling points in each deterioration feature cluster segment.

[0010] In one possible implementation, the feature recognition module is further configured to: construct an audio feedback graph, the audio feedback graph including multiple nodes and multiple edges between the multiple nodes, the multiple nodes including multiple audio feedback standard segment nodes and multiple audio feedback variant segment nodes, each audio feedback variant segment node being connected to multiple audio feedback standard segment nodes, the node feature of each audio feedback standard segment node being the audio resonance spectrum data of each audio feedback standard segment, and the node feature of each audio feedback variant segment node being the audio resonance spectrum data of each audio feedback variant segment; and process the audio feedback graph based on a graph neural network to determine multiple degradation feature clusters and a voiding anomaly mapping map of each degradation feature cluster.

[0011] In one possible implementation, the cavity determination module is further configured to: cluster the mechanical impedance sampling information of each mechanical impedance sampling point of each deterioration feature cluster segment and the mechanical impedance sampling information of each cooperative mechanical impedance sampling point of each deterioration feature cluster segment to obtain multiple impedance response association clusters for each deterioration feature cluster segment, each impedance response association cluster containing the mechanical impedance sampling information of multiple mechanical impedance sampling points and the mechanical impedance sampling information of multiple cooperative mechanical impedance sampling points; and determine the cavity determination based on the multiple impedance response association clusters of each deterioration feature cluster segment and the cavity removal anomaly mapping of each deterioration feature cluster segment. Multiple core deterioration feature clusters are identified. Based on the mechanical impedance sampling information of multiple mechanical impedance sampling points in each core deterioration feature cluster, the mechanical impedance sampling information of multiple coordinated mechanical impedance sampling points in each core deterioration feature cluster, and the acoustic resonance spectrum data of each core deterioration feature cluster, multiple suspected cavity anomaly regions in each core deterioration feature cluster are determined, and ground-penetrating radar analysis data of multiple suspected cavity anomaly regions in each core deterioration feature cluster are obtained. Based on the ground-penetrating radar analysis data of multiple suspected cavity anomaly regions in each core deterioration feature cluster, the grouting repair solid cavity of the tunnel lining is determined.

[0012] In one possible implementation, the input of the graph neural network is the audio feedback spectrum, and the output of the graph neural network is multiple deterioration feature clusters and a void anomaly mapping map of each deterioration feature cluster.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring acoustic resonance spectrum data of a tunnel lining; determining multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining; determining multiple deterioration feature cluster segments and a voiding anomaly mapping map of each deterioration feature cluster segment based on the multiple acoustic feedback standard segments and the multiple acoustic feedback variation segments; determining multiple mechanical impedance sampling points of each deterioration feature cluster segment based on the voiding anomaly mapping map of each deterioration feature cluster segment, and acquiring each mechanical impedance of each deterioration feature cluster segment. The mechanical impedance sampling information of the anti-sampling points; based on the mechanical impedance sampling information of each mechanical impedance sampling point in each deterioration feature cluster segment, determine multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment; based on the multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment, determine multiple cooperative mechanical impedance sampling points in each deterioration feature cluster segment, and obtain the mechanical impedance sampling information of each cooperative mechanical impedance sampling point in each deterioration feature cluster segment; based on the multiple mechanical impedance sampling points and multiple cooperative mechanical impedance sampling points in each deterioration feature cluster segment, determine the grouting repair solid cavity of the tunnel lining.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned multi-source tunnel monitoring data acquisition and processing method. The method includes: acquiring acoustic resonance spectrum data of the tunnel lining; determining multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining; determining multiple deterioration feature cluster segments and a voiding anomaly mapping map for each deterioration feature cluster segment based on the multiple acoustic feedback standard segments and the multiple acoustic feedback variation segments; determining multiple mechanical impedance sampling points for each deterioration feature cluster segment based on the voiding anomaly mapping map for each deterioration feature cluster segment, and acquiring the mechanical impedance of each mechanical impedance sampling point for each deterioration feature cluster segment. Anti-sampling information; based on the mechanical impedance sampling information of each mechanical impedance sampling point in each deterioration feature cluster segment, determine multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment; based on the multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions in each deterioration feature cluster segment, determine multiple cooperative mechanical impedance sampling points in each deterioration feature cluster segment, and obtain the mechanical impedance sampling information of each cooperative mechanical impedance sampling point in each deterioration feature cluster segment; based on the multiple mechanical impedance sampling points and multiple cooperative mechanical impedance sampling points in each deterioration feature cluster segment, determine the grouting repair solid cavity of the tunnel lining.

[0015] This invention provides a method and system for acquiring and processing multi-source tunnel monitoring data. The method includes acquiring acoustic resonance spectrum data of the tunnel lining; determining multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining; determining multiple deterioration feature cluster segments and a voiding anomaly mapping map for each deterioration feature cluster segment based on the multiple acoustic feedback standard segments and the multiple acoustic feedback variation segments; determining multiple mechanical impedance sampling points for each deterioration feature cluster segment based on the voiding anomaly mapping map of each deterioration feature cluster segment, and acquiring mechanical impedance sampling information for each mechanical impedance sampling point of each deterioration feature cluster segment; and acquiring mechanical impedance sampling information for each mechanical impedance sampling point of each deterioration feature cluster segment. This method identifies multiple impedance gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions for each deterioration feature cluster segment. Based on these regions, multiple coordinated mechanical impedance sampling points are determined for each segment, and mechanical impedance sampling information for each point is acquired. Finally, based on these points, the grouting repair cavity of the tunnel lining is determined. This method can efficiently and accurately identify the grouting repair cavity caused by tunnel lining voids. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a multi-source tunnel monitoring data acquisition and processing method provided in an embodiment of the present invention;

[0017] Figure 2 This is a schematic diagram of a tunnel lining provided in an embodiment of the present invention;

[0018] Figure 3 This is a schematic flowchart illustrating a process for determining multiple clusters of deterioration features and a voiding anomaly mapping map for each cluster of deterioration features, as provided in an embodiment of the present invention.

[0019] Figure 4 A schematic diagram of a mechanical impedance tester provided in an embodiment of the present invention;

[0020] Figure 5 A schematic diagram of a process for determining the grouting repair cavity of a tunnel lining, provided as an embodiment of the present invention;

[0021] Figure 6 This is a schematic diagram of a multi-source tunnel monitoring data acquisition and processing system provided in an embodiment of the present invention.

[0022] Figure 7This is a schematic diagram of the structure of an electronic device provided in one embodiment of this specification. Detailed Implementation

[0023] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0024] As used herein, the term "comprising" and its variations are open terms meaning "including but not limited to". The term "based on" means "at least partially based on". The terms "one embodiment" and "an embodiment" mean "at least one embodiment". The term "another embodiment" means "at least one other embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other definitions, whether explicit or implicit, may be included below. Unless explicitly indicated by the context, the definition of a term shall remain consistent throughout the specification.

[0025] In this embodiment of the invention, the following are provided: Figure 1 The method for acquiring and processing multi-source tunnel monitoring data, as shown, includes steps S1 to S7:

[0026] Step S1: Obtain the acoustic resonance spectrum data of the tunnel lining.

[0027] The acoustic resonance spectrum data of the tunnel lining is obtained by applying an excitation signal to the tunnel lining structure through an excitation device, collecting the acoustic wave signal fed back by the lining structure in real time by an acoustic sensor, and then processing it through a fast Fourier transform to obtain the frequency domain distribution data. Figure 2 This is a schematic diagram of a tunnel lining provided in an embodiment of the present invention.

[0028] The acoustic resonance spectrum data of tunnel lining includes the amplitude, phase, and resonance peak distribution data of the lining structure in different frequency bands within the corresponding frequency range.

[0029] The acoustic resonance spectrum data of the tunnel lining can provide a preliminary indication of the density and structural integrity of the lining.

[0030] Resonance peak distribution data is a quantitative record of the distribution of different resonance peaks within the corresponding frequency range in the acoustic resonance spectrum data of tunnel lining. The resonance peak distribution data includes the specific frequency location corresponding to each resonance peak, as well as the interval distance and amplitude difference between different resonance peaks.

[0031] The resonance peak distribution data can intuitively reflect the response intensity distribution law of the lining structure at each resonance frequency point under acoustic excitation.

[0032] Optionally, the acoustic resonance spectrum data of the tunnel lining can provide a preliminary reflection of the density and structural integrity of the lining. A healthy, dense lining has specific spectral characteristics (such as a distinct main peak and regular intervals); while a lining with voids or looseness will show anomalies in its spectrum, such as a shift in the main peak, the appearance of new peaks, attenuation of existing peaks, or splitting.

[0033] Step S2: Based on the acoustic resonance spectrum data of the tunnel lining, determine multiple acoustic feedback standard segments and multiple acoustic feedback variation segments.

[0034] In some embodiments, an audio frequency analysis model can be used to determine multiple standard audio frequency feedback segments and multiple variable audio frequency feedback segments. The audio frequency analysis model is a Transformer model. The input to the audio frequency analysis model is the audio frequency resonance spectrum data of the tunnel lining, and the output of the audio frequency analysis model is the multiple standard audio frequency feedback segments and multiple variable audio frequency feedback segments.

[0035] The Transformer model is a deep learning architecture based on a self-attention mechanism. Its core consists of an encoder and a decoder. The encoder, through multi-head attention, can process the input sequence in parallel and capture long-range dependencies within the sequence. The feedforward neural network performs non-linear transformations on the features. When processing temporally or logically related sequence data, the Transformer model can preserve the structural information of the original signal through positional encoding. The Transformer model exhibits extremely high efficiency in feature extraction and sequence modeling.

[0036] The audio feedback standard section is a tunnel lining section whose spectral characteristics conform to the design specifications and have no defective signal characteristics, determined by the audio analysis model.

[0037] The acoustic resonance spectrum data corresponding to the acoustic feedback standard section shows that the resonance frequency fluctuation is small and the energy distribution sequence is uniform, which indicates that the tunnel lining corresponding to the acoustic feedback standard section is in a dense or intact state.

[0038] The audio feedback anomaly section is the tunnel lining section where the spectral energy deviates abnormally, the main frequency shifts, and stray peaks appear, as determined by the audio analysis model.

[0039] The audio feedback anomaly segment corresponds to the possibility of voids inside the lining.

[0040] Voiding refers to the separation of the interface between the back of the tunnel lining structure and the surrounding rock during the construction of the tunnel lining structure due to reasons such as incomplete filling, concrete shrinkage, or settlement of the surrounding rock.

[0041] The section with abnormal audio feedback is a candidate section of the tunnel lining that will be the focus of subsequent monitoring.

[0042] The acoustic resonance spectrum data of tunnel lining contains frequency distribution characteristics reflecting the dynamic response of the lining structure. The mass state of the tunnel lining is directly mapped onto the spectral distribution of the acoustic signal. Peak shift, bandwidth variation, and energy attenuation coefficient in the spectrum are key features distinguishing the standard acoustic feedback segment from the variable acoustic feedback segment. By analyzing the acoustic resonance spectrum data of tunnel lining, the model can extract sensitive feature indicators characterizing the internal medium distribution and interface coupling state of the lining structure, thereby achieving accurate identification of differences in the structural properties of tunnel lining.

[0043] The Transformer model can perform a full-sequence scan of the acoustic resonance spectrum data of the input tunnel lining. Through a self-attention mechanism, the Transformer model can calculate the correlation of spectral characteristics at different locations, thereby automatically identifying sequence distributions that conform to the spectral characteristics under standard construction quality and labeling them as multiple standard acoustic feedback segments. Simultaneously, the Transformer model can sensitively capture weak non-steady-state changes in the spectrum. By comparing with a global spectral benchmark, the Transformer model can accurately locate intervals of abnormal energy feedback, thus identifying multiple acoustic feedback anomaly segments.

[0044] Step S3: Based on the multiple audio feedback standard segments and the multiple audio feedback variation segments, determine multiple degradation feature cluster segments and a voiding anomaly mapping map for each degradation feature cluster segment.

[0045] In some embodiments, Figure 3 This is a flowchart illustrating a method for determining multiple degradation feature clusters and a void anomaly mapping for each degradation feature cluster, as provided in an embodiment of the present invention. The determination of the multiple degradation feature clusters and the void anomaly mapping for each degradation feature cluster includes steps S31-S32:

[0046] Step S31: Construct an audio feedback graph. The audio feedback graph includes multiple nodes and multiple edges between the nodes. The multiple nodes include multiple standard audio feedback segment nodes and multiple variable audio feedback segment nodes. Each variable audio feedback segment node is connected to multiple standard audio feedback segment nodes. The node feature of each standard audio feedback segment node is the audio resonance spectrum data of each standard audio feedback segment. The node feature of each variable audio feedback segment node is the audio resonance spectrum data of each variable audio feedback segment.

[0047] An audio feedback map is a graphical structure that represents the relationship between standard and variable audio feedback segments. Through the combination of nodes and edges, the audio feedback map clearly presents the correlation between different data segments, providing structured data for analyzing the deterioration characteristics of tunnel lining structures.

[0048] Multiple nodes are the basic building blocks of the audio feedback spectrum. These nodes include multiple standard audio feedback segment nodes and multiple variable audio feedback segment nodes. Each standard audio feedback segment node corresponds to one standard audio feedback segment, and each variable audio feedback segment node corresponds to one variable audio feedback segment. The node characteristic of each standard audio feedback segment node is the audio resonance spectrum data of that standard segment, and the node characteristic of each variable audio feedback segment node is the audio resonance spectrum data of that variable segment.

[0049] The acoustic resonance spectrum data of the acoustic feedback standard section is the spectrum data of a specific segment corresponding to the acoustic resonance spectrum data of the tunnel lining.

[0050] The acoustic resonance spectrum data of the acoustic feedback variation section is the spectrum data of a specific segment corresponding to the acoustic resonance spectrum data of the tunnel lining.

[0051] The edges in the audio feedback graph can be used to establish the association between the audio feedback variation segment nodes and the audio feedback standard segment nodes.

[0052] The features of each edge include the distance between the audio feedback variant segment and the corresponding audio feedback standard segment, as well as the degree of difference between their audio resonance spectrum data.

[0053] In some embodiments, a deep neural network can be used to determine the degree of difference in the acoustic resonance spectrum data between each acoustic feedback variation segment node and each acoustic feedback standard segment node.

[0054] Step S32: Process the audio feedback spectrum based on graph neural network to determine multiple degradation feature clusters and the voiding anomaly mapping map of each degradation feature cluster.

[0055] Graph Neural Networks (GNNs) are deep learning models that can directly operate on graphs. GNNs capture spatial dependencies between nodes through message passing mechanisms. By aggregating features from neighboring nodes and updating them in conjunction with their own state, GNNs can learn complex topological features in the graph. GNNs can process data in non-Euclidean space and have significant advantages in tasks such as node classification, link prediction, and subgraph recognition. The input to the GNN is the audio feedback graph, and the output is multiple degraded feature clusters and a de-empty anomaly mapping for each degraded feature cluster.

[0056] Multiple deterioration feature clusters were selected from multiple audio feedback anomaly segments using graph neural networks, and are key tunnel lining sections suspected of being delamination to be investigated.

[0057] The deterioration feature clustering section corresponds to the key local area in the tunnel lining where defects may exist.

[0058] The voiding anomaly mapping map for each deterioration feature cluster segment is output by a graph neural network, representing the spatial distribution of the probability of voiding damage occurring at various locations within each deterioration feature cluster segment. The voiding anomaly mapping map of the deterioration feature cluster segment numerically calibrates the probability of voiding risk at different points within the cluster segment.

[0059] By constructing an acoustic feedback spectrum, isolated acoustic feedback anomaly segments and standard acoustic feedback segments in tunnel lining can be integrated into a unified spatial correlation network. Since the health status of the lining structure often exhibits spatial continuity, the determination of local anomalies must be based on a comprehensive comparison with the baseline characteristics of surrounding standard segments. In the acoustic feedback spectrum, the node feature of each node is the acoustic resonance spectrum data of the corresponding segment, while multiple edges quantify the physical correlation attributes between nodes through the spatial relative distance between each acoustic feedback anomaly segment and its corresponding standard acoustic feedback segment, as well as the difference in their acoustic resonance spectrum data. This allows the model to more deeply analyze the evolution trend and correlation of spectral indicators between different segments, thereby significantly improving the accuracy of screening segments with clustered deterioration features. Using graph neural networks to process the acoustic feedback spectrum data can fully leverage the model's information transmission and representation capabilities in non-Euclidean space, effectively capturing the complex dependencies between lining segments, and thus more accurately identifying key tunnel lining sections with a risk of delamination.

[0060] Graph Neural Networks (GNNs) achieve deep fusion of node features by performing multi-layer convolution operations on the audio feedback spectrum. By learning the difference weights between each node in the audio feedback anomaly segment and its neighboring standard audio feedback segment nodes, GNNs can identify candidate regions with significant shifts in spectral features and define them as clusters of deteriorated features. Simultaneously, GNNs can extract hidden correlation features between nodes and map them onto a spatial coordinate system. By modeling the probability of voiding defects occurring at various locations within the clusters, GNNs can generate a voiding anomaly mapping map for each cluster of deteriorated features.

[0061] Step S4: Based on the desiccation anomaly mapping of each deterioration feature cluster segment, determine multiple mechanical impedance sampling points for each deterioration feature cluster segment, and obtain the mechanical impedance sampling information of each mechanical impedance sampling point for each deterioration feature cluster segment.

[0062] In some embodiments, a sampling point localization model can be used to determine multiple mechanical impedance sampling points for each degradation feature cluster segment. The sampling point localization model is a convolutional neural network. The input to the sampling point localization model is a void anomaly mapping map of each degradation feature cluster segment, and the output of the sampling point localization model is multiple mechanical impedance sampling points for each degradation feature cluster segment.

[0063] A convolutional neural network (CNN) is a type of neural network that incorporates convolutional computations. CNNs can automatically extract spatial hierarchical features from input data using convolutional kernels. A CNN consists of convolutional layers, pooling layers, and fully connected layers. The local connectivity and weight sharing characteristics of CNNs give them excellent robustness and computational efficiency when processing image or matrix-like data.

[0064] The multiple mechanical impedance sampling points in each degradation feature cluster are determined by the sampling point positioning model, which provides the precise physical coordinates for mechanical impedance monitoring within the degradation feature cluster.

[0065] The mechanical impedance sampling information of the mechanical impedance sampling points is the data information collected by deploying a mechanical impedance tester at the corresponding mechanical impedance sampling points.

[0066] In some embodiments, a controllable excitation force with continuous frequency sweep is applied to the lining surface using a mechanical impedance tester, and the dynamic response signal of the lining during the excitation process is captured by a sensor to obtain mechanical impedance sampling information. Figure 4 This is a schematic diagram of a mechanical impedance tester provided in an embodiment of the present invention.

[0067] The mechanical impedance sampling information at the mechanical impedance sampling point includes dynamic stiffness value and mechanical impedance curve.

[0068] Dynamic stiffness is a numerical value that indicates the degree of coupling between the tunnel lining and the surrounding rock behind it. Dynamic stiffness can be used to represent the tunnel lining's ability to resist deformation under dynamic loads.

[0069] The mechanical impedance curve is a function curve describing the resistance of a lining structure to controlled dynamic excitation signals as a function of frequency. Each sampling coordinate point in the mechanical impedance curve represents the ratio of the dynamic reaction force generated by the lining structure to the corresponding vibration velocity response at a specific excitation frequency.

[0070] The mechanical impedance sampling information of the sampling points can directly reflect the structural characteristics and density of the tunnel lining area corresponding to the sampling points.

[0071] The numerical distribution in the void anomaly mapping of the deterioration feature cluster section is highly correlated with the geometric location of the physical cavity on the back of the tunnel lining. The model can identify the local extreme value regions and abrupt boundary of the risk probability distribution through the void anomaly mapping of the deterioration feature cluster section, thereby accurately determining the key physical coordinate points that can characterize the degree of structural deterioration.

[0072] Convolutional neural networks (CNNs) can extract local extrema and edge mutation features from the voiding anomaly map of deterioration feature clusters through multiple layers of convolutional kernels, and can quantify the spatial heterogeneity of probability distributions. Through convolution operations, CNNs can capture the geometric features of probability peaks and gradient abrupt change points in the voiding anomaly map of deterioration feature clusters, and automatically identify representative central sites or probability mutation locations, thereby accurately determining multiple mechanical impedance sampling points for each deterioration feature cluster.

[0073] Step S5: Based on the mechanical impedance sampling information of each mechanical impedance sampling point of each degradation feature cluster segment, determine multiple impedance feature gradient drastic change regions and multiple impedance feature information non-homogeneous regions of each degradation feature cluster segment.

[0074] In some embodiments, an impedance characteristic analysis model can be used to determine multiple regions of abrupt changes in impedance characteristic gradients and multiple regions of non-homogeneous impedance characteristic information in each degradation feature cluster. The impedance characteristic analysis model is a Transformer model. The input to the impedance characteristic analysis model is the mechanical impedance sampling information of each mechanical impedance sampling point in each degradation feature cluster, and the output of the impedance characteristic analysis model is the multiple regions of abrupt changes in impedance characteristic gradients and multiple regions of non-homogeneous impedance characteristic information in each degradation feature cluster.

[0075] Multiple regions of abrupt changes in impedance characteristic gradients within the degradation feature clustering segment were determined by the impedance characteristic analysis model. These regions are located within the corresponding degradation feature clustering segment, and the mechanical impedance sampling information exhibits abrupt changes with spatial location.

[0076] The non-homogeneous regions of multiple impedance characteristic information in the degradation feature cluster segment were determined by the impedance characteristic analysis model. These regions are located within the corresponding degradation feature cluster segment, where the mechanical impedance sampling points exhibit high spatial dispersion and lack correlation of mechanical impedance sampling information.

[0077] By identifying regions of dramatic changes in impedance gradients and non-homogeneous impedance information, the abnormal constraint range of characteristic parameter evolution within each deterioration feature cluster can be established. By pinpointing the boundary locations where mechanical impedance sampling information experiences a magnitude jump, and the discrete intervals of abnormal characteristic parameter distribution, the core area of ​​lining structure performance degradation can be precisely located. This provides a spatial layout basis for subsequently adding collaborative mechanical impedance sampling points within this area, enabling high-density data acquisition of detailed characteristics of tunnel lining defects within the deterioration feature cluster.

[0078] The Transformer model excels at handling long-range dependencies and sequence modeling. It can spatially arrange the mechanical impedance sampling information of each mechanical impedance sampling point within each degradation feature cluster into an input sequence. Through a multi-head self-attention mechanism, the Transformer model can calculate the rate of change of the impedance gradient between sampling points and identify spatial ranges with drastic numerical fluctuations, defining these as multiple regions of dramatic impedance gradient changes within each degradation feature cluster. Simultaneously, the Transformer model can assess the dispersion of data points within the sequence and identify segments with anomalous signal dispersion, thereby determining multiple non-homogeneous regions of impedance feature information within each degradation feature cluster.

[0079] Step S6: Based on the multiple impedance characteristic gradient drastic change regions and the multiple impedance characteristic information non-homogeneous regions of each degradation feature cluster segment, determine multiple cooperative mechanical impedance sampling points for each degradation feature cluster segment, and obtain the mechanical impedance sampling information of each cooperative mechanical impedance sampling point for each degradation feature cluster segment.

[0080] In some embodiments, a collaborative sampling analysis model can be used to determine multiple collaborative mechanical impedance sampling points for each degradation feature cluster segment. The collaborative sampling analysis model is a deep neural network. The inputs to the collaborative sampling analysis model are multiple impedance characteristic gradient abrupt change regions and multiple impedance characteristic information non-homogeneous regions for each degradation feature cluster segment; the outputs of the collaborative sampling analysis model are multiple collaborative mechanical impedance sampling points for each degradation feature cluster segment.

[0081] A deep neural network (DNN) is a type of neural network composed of multiple hidden layers. DNNs can learn highly abstract features of input data. Through nonlinear activation functions and hierarchical structures, DNNs can simulate extremely complex function mappings. Each neuron in a layer extracts deep semantic information from the previous layer through weighted aggregation of information from the previous layer. DNNs can be applied to complex classification and regression analysis tasks.

[0082] The multiple cooperative mechanical impedance sampling points for each degradation feature cluster segment are determined by the cooperative sampling analysis model. These auxiliary measurement positions are added to the mechanical impedance sampling points within the degradation feature cluster segment to eliminate isolated point deviations and perform multi-point verification.

[0083] Coordinated mechanical impedance sampling points can improve the accuracy and robustness of monitoring results.

[0084] The mechanical impedance sampling information of the coordinated mechanical impedance sampling point is a set of data obtained by monitoring the mechanical impedance at the determined location of the coordinated mechanical impedance sampling point using a mechanical impedance tester.

[0085] The mechanical impedance sampling information includes dynamic stiffness values ​​and mechanical impedance curves.

[0086] By using the mechanical impedance sampling information from the coordinated mechanical impedance sampling points, it is possible to perform high-density monitoring of the evolution details of characteristic parameters in the region of dramatic changes in impedance characteristic gradient and the region of non-homogeneous impedance characteristic information, thereby improving the analytical accuracy of the boundary of lining structure defects and the distribution characteristics of internal media in the deterioration feature accumulation section.

[0087] The regions of dramatic changes in impedance gradient and the non-homogeneous regions of impedance characteristic information establish the abnormal constraint ranges for the evolution of characteristic parameters within each degradation feature cluster segment. These regions identify the boundary locations of lining structure performance degradation and the discrete intervals of abnormal characteristic parameter distributions, constituting the spatial search guidance constraints for the deep neural network.

[0088] The multiple impedance gradient abrupt change zones within each degradation feature cluster reflect areas of rapid abrupt changes in mechanical impedance parameters, while the multiple non-homogeneous impedance feature information zones represent areas of irregular distribution of mechanical impedance parameters. Both types of regions are critical areas where potential hazards exist in the tunnel lining structure. Initial mechanical impedance sampling points may not fully cover the critical locations in these areas, nor fully capture the structural details within them. Therefore, these two types of regions ensure that coordinated mechanical impedance sampling points can accurately target the core areas of structural anomalies.

[0089] Deep neural networks can fuse features of geometry and impedance intensity in regions of abrupt changes in impedance gradients and regions of non-homogeneous impedance information. Through nonlinear mapping in fully connected layers, deep neural networks can assess the coverage defects of existing mechanical impedance sampling point density and calculate spatial locations that minimize uncertainty in defect assessment, thereby determining multiple collaborative mechanical impedance sampling points for each cluster of deterioration features. Leveraging its deep feature learning capabilities, deep neural networks can fully utilize the feature information of both regions—those with abrupt changes in impedance gradients and those with non-homogeneous impedance information—to ensure that the setting of collaborative sampling points is both efficient and targeted.

[0090] Step S7: Determine the grouting repair cavity of the tunnel lining based on multiple mechanical impedance sampling points of each deterioration feature cluster segment and multiple coordinated mechanical impedance sampling points of each deterioration feature cluster segment.

[0091] In some embodiments, Figure 5 This is a schematic flowchart illustrating the process of determining the grouting repair cavity of a tunnel lining according to an embodiment of the present invention. The determination of the grouting repair cavity of the tunnel lining includes steps S71 to S74:

[0092] Step S71: Based on the mechanical impedance sampling information of each mechanical impedance sampling point of each degradation feature cluster segment and the mechanical impedance sampling information of each cooperative mechanical impedance sampling point of each degradation feature cluster segment, clustering is performed to obtain multiple impedance response association clusters for each degradation feature cluster segment. Each impedance response association cluster contains the mechanical impedance sampling information of multiple mechanical impedance sampling points and the mechanical impedance sampling information of multiple cooperative mechanical impedance sampling points.

[0093] The clustering method used is K-means clustering, an unsupervised learning algorithm that groups data points with similar characteristics into the same cluster by calculating the Euclidean distance between samples and cluster centers. During operation, the K-means clustering algorithm continuously updates the positions of the cluster centers until the objective function converges, thus achieving automatic classification of unlabeled data.

[0094] The multiple impedance response correlation clusters of each degradation feature cluster segment are obtained by clustering the mechanical impedance sampling information of mechanical impedance sampling points and co-mechanical impedance sampling points within the degradation feature cluster segment using the K-means clustering algorithm.

[0095] Each impedance response cluster consists of mechanical impedance sampling information from multiple mechanical impedance sampling points with similar impedance characteristics within its own degradation feature cluster segment, as well as mechanical impedance sampling information from multiple cooperative mechanical impedance sampling points.

[0096] The sampling information within the same impedance response cluster shows a high degree of consistency in key parameters such as amplitude, phase, and dynamic stiffness extracted from the mechanical impedance curve.

[0097] The process of clustering mechanical impedance sampling information of each mechanical impedance sampling point in each degradation feature cluster segment and mechanical impedance sampling information of each co-factor mechanical impedance sampling point in each degradation feature cluster segment using the K-means clustering algorithm is as follows: K cluster centers are randomly initialized. The Euclidean distance from each sampling information to each cluster center is calculated, and each sampling information is assigned to the cluster containing the nearest cluster center, thus forming multiple initial impedance response association clusters. Next, the mean of all sampling information within each cluster is calculated and used as the new cluster center. The distance from each sampling information to the new cluster center is calculated again, and the cluster relationships are reassigned. This cluster center update and cluster assignment process is repeated until the change in cluster centers is less than a preset threshold or the maximum number of iterations is reached, ultimately resulting in multiple stable impedance response association clusters.

[0098] Clustering effectively integrates complex tunnel lining impedance monitoring information. Due to the large number and diverse characteristics of mechanical impedance and co-mechanical impedance sampling points within each deterioration feature cluster, direct point-by-point analysis is highly complex. Clustering groups sampling points with similar impedance characteristics into multiple impedance response clusters, significantly simplifying the data structure and facilitating the extraction of valuable information reflecting structural properties. By dividing the sampling information within each deterioration feature cluster into multiple clusters, the different types and spatial distribution of lining structural states within a specific deterioration range can be visually displayed.

[0099] Step S72: Based on the multiple impedance response correlation clusters of each degradation feature cluster segment and the void anomaly mapping of each degradation feature cluster segment, determine multiple core degradation feature cluster segments.

[0100] In some embodiments, a core degradation segment screening model can be used to determine multiple core degradation feature clusters. The core degradation segment screening model is a convolutional neural network. The input to the core degradation segment screening model consists of multiple impedance response correlation clusters for each degradation feature cluster and a voiding anomaly mapping map for each degradation feature cluster. The output of the core degradation segment screening model is the multiple core degradation feature clusters.

[0101] Multiple core deterioration feature clusters were identified through a core deterioration feature screening model, which further determined the tunnel lining sections with a high suspicion of cavitation from the initially screened deterioration feature clusters.

[0102] The core degradation feature cluster represents the lining structure in terms of acoustic resonance spectrum data, mechanical impedance sampling information, and void anomaly mapping. Figure 3 The disease exhibits highly consistent local characteristics across all dimensions.

[0103] Multiple impedance response clusters within each degradation feature cluster classify sampling points based on the similarity of their physical responses. The impedance features within each cluster quantify the dynamic stiffness and density patterns of the lining. The voiding anomaly map of each degradation feature cluster provides spatially continuous distribution information on risk probabilities. The convolutional neural network (CNN) uses its receptive field mechanism to extract local features from the voiding anomaly map of each degradation feature cluster, thereby identifying high-risk regions and gradient abrupt change edges in the map. By associating the mechanical impedance sampling information of multiple mechanical impedance sampling points within each degradation feature cluster (found in multiple impedance response clusters) with their corresponding spatial coordinates, the CNN can quantify and evaluate the dynamic stiffness performance of different regions under dynamic excitation. Through multi-layer nonlinear transformations, the CNN can calculate and match the numerical features in the voiding anomaly map of each degradation feature cluster with the mechanical response features in the mechanical impedance sampling information of multiple mechanical impedance sampling points within each degradation feature cluster. This process enables convolutional neural networks to pinpoint overlapping regions in the monitoring signals that exhibit anomalies in both probability distribution and dynamic stiffness values, thereby achieving precise targeting of multiple clusters of core degradation features.

[0104] In some embodiments, determining multiple core degradation feature clusters based on multiple impedance response correlation clusters of each degradation feature cluster and a void anomaly mapping of each degradation feature cluster includes steps S721~S723:

[0105] Step S721: Based on the multiple impedance response correlation clusters of each degradation feature cluster segment, determine the structural stiffness distribution map of each degradation feature cluster segment, the mechanical impedance gradient anomaly index of each impedance response correlation cluster, and the response feature anomaly level of each impedance response correlation cluster.

[0106] In some embodiments, deep neural networks can be used to determine the structural stiffness distribution map of each deterioration feature cluster, the mechanical impedance gradient anomaly index of each impedance response cluster, and the response feature anomaly level of each impedance response cluster.

[0107] The structural stiffness distribution map of each deterioration feature cluster segment is a spatial distribution image of the dynamic stiffness value distribution at various locations within each deterioration feature cluster segment, generated by a deep neural network based on discrete mechanical impedance sampling point data. The structural stiffness distribution map of each deterioration feature cluster segment uses numerical values ​​to identify the magnitude of the dynamic stiffness value of the lining structure at different physical coordinate points.

[0108] The mechanical impedance gradient variation index of the impedance response correlation cluster is determined by a deep neural network, which is a quantitative value of the degree of drastic change of mechanical impedance parameters within the impedance response correlation cluster with spatial location.

[0109] The anomaly level of the response characteristics of each impedance response cluster is a numerical level that is a health status classification label for each impedance response cluster by a deep neural network.

[0110] Deep neural networks (DNNs) can extract nonlinear features from multiple impedance response clusters within each degradation feature cluster, enabling them to uncover the deep evolution patterns in high-dimensional impedance curves and dynamic stiffness values. Leveraging their powerful function fitting capabilities, DNNs can map discretely distributed mechanical impedance sampling information into a structural stiffness distribution map for each degradation feature cluster, possessing spatially continuous properties. Simultaneously, DNNs can sensitively capture gradient changes in impedance features in the spatial or frequency domains, thereby accurately calculating the mechanical impedance gradient anomaly index for each impedance response cluster reflecting response feature fluctuations and automatically classifying the anomaly levels of each cluster's response features.

[0111] Step S722: Generate an abnormal feature superposition mapping map for each deterioration feature cluster based on the structural stiffness distribution map of each deterioration feature cluster, the mechanical impedance gradient anomaly index of each impedance response association cluster, the response feature anomaly level of each impedance response association cluster, and the voiding anomaly mapping map of each deterioration feature cluster.

[0112] In some embodiments, a deep neural network can be used to generate an abnormal feature overlay mapping map for each degraded feature cluster segment.

[0113] The anomaly feature overlay mapping map for each degradation feature cluster segment is output by a deep neural network. It is a spatial image of the degree of anomaly after the overlay of multi-source monitoring data features at each coordinate point within each degradation feature cluster segment. The anomaly feature overlay mapping map for each degradation feature cluster segment identifies the degree of anomaly at each physical coordinate point through different values.

[0114] Deep neural networks can correlate and map the structural stiffness distribution map of each deterioration feature cluster, the mechanical impedance gradient anomaly index of each impedance response correlation cluster, the response feature anomaly level of each impedance response correlation cluster, and the voiding anomaly mapping map of each deterioration feature cluster in a unified tunnel coordinate system. The deep neural network calculates the correlation of multiple indicators at each coordinate location, then uses a weighted algorithm to superimpose the values ​​of different terms, thereby generating an anomaly feature superposition mapping map for each deterioration feature cluster that reflects the overall anomaly degree at that location.

[0115] Step S723: Determine multiple core degradation feature clusters based on the superimposed mapping of abnormal features in each degradation feature cluster.

[0116] In some embodiments, a convolutional neural network can be used to determine multiple clusters of core degradation features.

[0117] Convolutional neural networks (CNNs) leverage their local connectivity and weight sharing to perform high-resolution spatial feature recognition on the superimposed mapping of anomalous features in each cluster of deteriorated features in the input. The model automatically extracts morphological features and edge gradients from high-brightness anomaly regions in the mapping. By performing connected component analysis and geometric centroid extraction on continuous high-anomaly regions in the image, the CNN can accurately determine the physical boundaries of anomalous features along the tunnel axis. Subsequently, through spatial clustering and logical segmentation operations, the model converges a large range of suspected segments to core regions with significant deterioration features, ultimately identifying multiple core clusters of deteriorated features with higher localization accuracy.

[0118] Step S73: Based on the mechanical impedance sampling information of multiple mechanical impedance sampling points in each core deterioration feature cluster segment, the mechanical impedance sampling information of multiple coordinated mechanical impedance sampling points in each core deterioration feature cluster segment, and the acoustic resonance spectrum data of each core deterioration feature cluster segment, determine multiple suspected cavity anomaly regions in each core deterioration feature cluster segment, and obtain ground-penetrating radar analysis data of multiple suspected cavity anomaly regions in each core deterioration feature cluster segment.

[0119] In some embodiments, a suspected anomaly determination model can be used to identify multiple suspected cavity anomaly regions in each core degradation feature cluster segment. The suspected anomaly determination model is a deep neural network. The inputs to the suspected anomaly determination model are mechanical impedance sampling information of multiple mechanical impedance sampling points in each core degradation feature cluster segment, mechanical impedance sampling information of multiple coordinated mechanical impedance sampling points in each core degradation feature cluster segment, and acoustic resonance spectrum data of each core degradation feature cluster segment. The output of the suspected anomaly determination model is multiple suspected cavity anomaly regions in each core degradation feature cluster segment.

[0120] Multiple suspected cavity anomaly areas in the core deterioration feature cluster section were identified using a suspected anomaly identification model, which indicates the specific geometric spatial range of possible back lining void defects within the core deterioration feature cluster section.

[0121] The suspected cavity anomaly area is marked with the specific area that needs to be verified using ground-penetrating radar.

[0122] The ground-penetrating radar (GPR) analysis data for multiple suspected cavity anomaly areas within each core deterioration feature cluster is obtained by analyzing the detection results after electromagnetic wave scanning of the suspected cavity anomaly areas using a ground-penetrating radar (GPR) instrument. The GPR analysis data includes the two-way travel time of the radar waves and the amplitude intensity of the reflected waves.

[0123] The two-way travel time of a radar wave refers to the time it takes for a radar wave to travel from its emission to its reflection from the cavity interface. The two-way travel time of a radar wave can be used to calculate the specific depth and location of the cavity.

[0124] The amplitude intensity of a reflected wave is a quantitative indicator of the energy level of the reflected signal. When radar waves travel from concrete into the air, the amplitude increases significantly due to the large difference in dielectric constant. The amplitude intensity of the reflected wave can be used to determine the presence of a cavity.

[0125] The mechanical impedance sampling information and co-sampling information from multiple mechanical impedance sampling points in each core deterioration feature cluster provide high-density mechanical contact monitoring samples, thus reflecting the deformation resistance of the lining structure. Acoustic resonance spectrum data provides non-contact, wide-range acoustic response characteristics. The deep neural network, through its multi-layered hidden layer structure, can perform frequency domain feature encoding on the acoustic resonance spectrum and extract the resonance peak shift associated with lining voids. The deep neural network can cross-validate the mechanical impedance sampling information and quantify the changes in support conditions at the back of the lining by comparing the dynamic stiffness differences at different sampling locations. By applying nonlinear mapping to these features, the deep neural network can accurately define the specific boundaries of suspected abnormal areas with cavitation within each core deterioration feature cluster.

[0126] Step S74: Based on the ground-penetrating radar analysis data of multiple suspected cavity anomaly areas in each core deterioration feature cluster segment, determine the grouting repair solid cavity of the tunnel lining.

[0127] In some embodiments, a cavity determination model can be used to determine the grouting repair cavities of the tunnel lining. The cavity determination model is a deep neural network. The input to the cavity determination model is ground-penetrating radar analysis data of multiple suspected cavity anomaly areas in each core deterioration feature cluster segment, and the output of the cavity determination model is the grouting repair cavities of the tunnel lining.

[0128] The grouting repair cavity of the tunnel lining is determined by analyzing ground-penetrating radar data of suspected cavity anomaly areas using a cavity determination model. It is a physical void and precise spatial region located on the back of the lining structure, indicating the actual presence of voids. The grouting repair cavity of the tunnel lining includes the cavity's three-dimensional coordinates, geometric volume, and thickness.

[0129] Grouting repair of voids in tunnel lining can provide precise physical boundaries and engineering quantity basis for grouting construction, thereby achieving precise grouting repair of voids in tunnel lining.

[0130] Ground-penetrating radar (GPR) analysis data from multiple suspected cavity anomaly areas within each core deterioration feature cluster provides high-resolution reflection characteristics of the lining interior and the back medium interface. The reflection patterns of radar waves at different medium interfaces directly correspond to the spatial distribution of the physical entity, providing a physical basis for determining the final physical cavity.

[0131] Deep neural networks identify regions with enhanced reflected wave amplitude and phase reversal by analyzing ground-penetrating radar data from multiple suspected cavity anomaly areas within each core deterioration feature cluster. This allows them to pinpoint the physical interface between air and concrete. The deep neural network calculates the propagation speed and two-way travel time of radar waves in the lining medium, mapping the time axis data of electromagnetic signals to the spatial axis data of physical depth. By fusing data from multiple radar survey lines within the suspected area, the deep neural network extracts the envelope features reflecting the cavity boundary, ultimately calculating and determining the specific geometric dimensions and three-dimensional spatial location of the grouting repair cavity in the tunnel lining.

[0132] Please refer to the following. Figure 6 , Figure 6 A schematic diagram of a multi-source tunnel monitoring data acquisition and processing system provided in an embodiment of this specification is shown. It should be noted that... Figure 6 The multi-source tunnel monitoring data acquisition and processing system shown is used to execute this manual. Figure 1 The methods shown in the embodiments are illustrated for ease of explanation, showing only the parts related to the embodiments of this specification. For specific technical details not disclosed, please refer to this specification. Figure 1 The example shown.

[0133] Based on the same inventive concept Figure 6 This is a schematic diagram of a multi-source tunnel monitoring data acquisition and processing system provided in an embodiment of the present invention. The multi-source tunnel monitoring data acquisition and processing system includes:

[0134] Acquisition module 81 is used to acquire the acoustic resonance spectrum data of the tunnel lining;

[0135] The acoustic feedback analysis module 82 is used to determine multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining.

[0136] Feature recognition module 83 is used to determine multiple deterioration feature clusters and a voiding anomaly mapping map of each deterioration feature cluster based on the multiple audio feedback standard segments and the multiple audio feedback variation segments;

[0137] The mechanical impedance sampling module 84 is used to determine multiple mechanical impedance sampling points for each deterioration feature cluster based on the desiccation anomaly mapping map of each deterioration feature cluster, and to obtain mechanical impedance sampling information for each mechanical impedance sampling point of each deterioration feature cluster.

[0138] Impedance characteristic analysis module 85 is used to determine multiple impedance characteristic gradient drastic regions and multiple impedance characteristic information non-homogeneous regions of each degradation characteristic cluster based on the mechanical impedance sampling information of each mechanical impedance sampling point of each degradation characteristic cluster.

[0139] The collaborative sampling module 86 is used to determine multiple collaborative mechanical impedance sampling points for each degradation feature cluster based on multiple impedance feature gradient drastic regions and multiple impedance feature information non-homogeneous regions for each degradation feature cluster, and to obtain mechanical impedance sampling information for each collaborative mechanical impedance sampling point of each degradation feature cluster.

[0140] The cavity determination module 87 is used to determine the grouting repair cavity of the tunnel lining based on multiple mechanical impedance sampling points of each deterioration feature cluster segment and multiple cooperative mechanical impedance sampling points of each deterioration feature cluster segment.

[0141] See Figure 7 It shows a schematic diagram of the structure of an electronic device according to an embodiment of this specification, which can be used to implement... Figure 1 The method in the illustrated embodiment. (As shown) Figure 7 As shown, the electronic device 700 may include: at least one central processing unit 701, at least one network interface 704, user interface 703, memory 705, and at least one communication bus 702.

[0142] The communication bus 702 is used to enable communication between these components.

[0143] The user interface 703 may include a display screen and a camera. Optionally, the user interface 703 may also include a standard wired interface and a wireless interface.

[0144] The network interface 704 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0145] The processor 701 may include one or more processing cores. The processor 701 connects to various parts within the electronic device 700 using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 705, and by calling data stored in the memory 705. Optionally, the processor 701 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 701 may integrate one or a combination of several of the following: a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), and a modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 701 and may be implemented as a separate chip.

[0146] The memory 705 may include random access memory (RAM) or read-only memory. Optionally, the memory 705 may include a non-transitory computer-readable storage medium. The memory 705 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 705 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 705 may also be at least one storage system located remotely from the aforementioned processor 701. Figure 7 As shown, the memory 705, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and program instructions.

[0147] exist Figure 7 In the illustrated electronic device 700, the user interface 703 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 701 can be used to call the image-based interactive application stored in the memory 705.

[0148] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0149] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for acquiring and processing multi-source tunnel monitoring data, characterized in that, include: Obtain the acoustic resonance spectrum data of the tunnel lining; Based on the acoustic resonance spectrum data of the tunnel lining, multiple acoustic feedback standard segments and multiple acoustic feedback variation segments are determined. Based on the multiple audio feedback standard segments and the multiple audio feedback variation segments, multiple degradation feature cluster segments are determined, and a de-emptying anomaly mapping diagram is obtained for each degradation feature cluster segment; Based on the delamination anomaly mapping map of each deterioration feature cluster segment, multiple mechanical impedance sampling points of each deterioration feature cluster segment are determined, and mechanical impedance sampling information of each mechanical impedance sampling point of each deterioration feature cluster segment is obtained. Based on the mechanical impedance sampling information of each mechanical impedance sampling point in each degradation feature cluster segment, multiple impedance feature gradient drastic change regions and multiple impedance feature information non-homogeneous regions in each degradation feature cluster segment are determined. Based on the multiple impedance characteristic gradient drastic change regions of each degradation feature cluster segment and the multiple impedance characteristic information non-homogeneous regions of each degradation feature cluster segment, multiple cooperative mechanical impedance sampling points of each degradation feature cluster segment are determined, and mechanical impedance sampling information of each cooperative mechanical impedance sampling point of each degradation feature cluster segment is obtained. The grouting repair cavity of the tunnel lining is determined based on multiple mechanical impedance sampling points of each deterioration feature cluster segment and multiple coordinated mechanical impedance sampling points of each deterioration feature cluster segment.

2. The multi-source tunnel monitoring data acquisition and processing method as described in claim 1, characterized in that, The determination of multiple degradation feature clusters based on the multiple audio feedback standard segments and the multiple audio feedback variation segments, and the de-empty anomaly mapping map of each degradation feature cluster segment include: A sound feedback spectrum is constructed, which includes multiple nodes and multiple edges between the nodes. The multiple nodes include multiple standard sound feedback segment nodes and multiple variable sound feedback segment nodes. Each variable sound feedback segment node is connected to multiple standard sound feedback segment nodes. The node feature of each standard sound feedback segment node is the sound resonance spectrum data of each standard sound feedback segment. The node feature of each variable sound feedback segment node is the sound resonance spectrum data of each variable sound feedback segment. The audio feedback spectrum is processed using a graph neural network to determine multiple degradation feature clusters and a voiding anomaly mapping map for each degradation feature cluster.

3. The multi-source tunnel monitoring data acquisition and processing method as described in claim 1, characterized in that, The determination of the grouting repair cavity of the tunnel lining based on multiple mechanical impedance sampling points of each deterioration feature cluster segment and multiple coordinated mechanical impedance sampling points of each deterioration feature cluster segment includes: Based on the mechanical impedance sampling information of each mechanical impedance sampling point in each degradation feature cluster segment and the mechanical impedance sampling information of each cooperative mechanical impedance sampling point in each degradation feature cluster segment, multiple impedance response association clusters are obtained for each degradation feature cluster segment. Each impedance response association cluster contains the mechanical impedance sampling information of multiple mechanical impedance sampling points and the mechanical impedance sampling information of multiple cooperative mechanical impedance sampling points. Multiple core degradation feature clusters are determined based on multiple impedance response correlation clusters of each degradation feature cluster and the void anomaly mapping of each degradation feature cluster. Based on the mechanical impedance sampling information of multiple mechanical impedance sampling points in each core deterioration feature cluster segment, the mechanical impedance sampling information of multiple coordinated mechanical impedance sampling points in each core deterioration feature cluster segment, and the acoustic resonance spectrum data of each core deterioration feature cluster segment, multiple suspected cavity anomaly areas in each core deterioration feature cluster segment are identified, and ground-penetrating radar analysis data of multiple suspected cavity anomaly areas in each core deterioration feature cluster segment are obtained. Based on the ground-penetrating radar analysis data of multiple suspected cavity anomaly areas in each core deterioration feature cluster, the grouting repair of the tunnel lining solid cavity is determined.

4. The multi-source tunnel monitoring data acquisition and processing method as described in claim 2, characterized in that, The input to the graph neural network is the audio feedback spectrum, and the output of the graph neural network is multiple deterioration feature clusters and a voiding anomaly mapping map of each deterioration feature cluster.

5. A multi-source tunnel monitoring data acquisition and processing system, characterized in that, include: The acquisition module is used to acquire the acoustic resonance spectrum data of the tunnel lining; The acoustic feedback analysis module is used to determine multiple acoustic feedback standard segments and multiple acoustic feedback variation segments based on the acoustic resonance spectrum data of the tunnel lining. The feature recognition module is used to determine multiple deterioration feature clusters and a de-empty anomaly mapping map of each deterioration feature cluster based on the multiple audio feedback standard segments and the multiple audio feedback variation segments. The mechanical impedance sampling module is used to determine multiple mechanical impedance sampling points for each deterioration feature cluster based on the desiccation anomaly mapping map of each deterioration feature cluster, and to obtain the mechanical impedance sampling information of each mechanical impedance sampling point for each deterioration feature cluster. The impedance characteristic analysis module is used to determine multiple impedance characteristic gradient drastic regions and multiple impedance characteristic information non-homogeneous regions of each degradation characteristic cluster based on the mechanical impedance sampling information of each mechanical impedance sampling point of each degradation characteristic cluster. The collaborative sampling module is used to determine multiple collaborative mechanical impedance sampling points for each degradation feature cluster based on multiple impedance feature gradient drastic regions and multiple impedance feature information non-homogeneous regions for each degradation feature cluster, and to obtain the mechanical impedance sampling information of each collaborative mechanical impedance sampling point for each degradation feature cluster. The cavity determination module is used to determine the grouting repair cavity of the tunnel lining based on multiple mechanical impedance sampling points of each deterioration feature cluster segment and multiple coordinated mechanical impedance sampling points of each deterioration feature cluster segment.

6. The multi-source tunnel monitoring data acquisition and processing system as described in claim 5, characterized in that, The feature recognition module is also used for: A sound feedback spectrum is constructed, which includes multiple nodes and multiple edges between the nodes. The multiple nodes include multiple standard sound feedback segment nodes and multiple variable sound feedback segment nodes. Each variable sound feedback segment node is connected to multiple standard sound feedback segment nodes. The node feature of each standard sound feedback segment node is the sound resonance spectrum data of each standard sound feedback segment. The node feature of each variable sound feedback segment node is the sound resonance spectrum data of each variable sound feedback segment. The audio feedback spectrum is processed using a graph neural network to determine multiple degradation feature clusters and a voiding anomaly mapping map for each degradation feature cluster.

7. The multi-source tunnel monitoring data acquisition and processing system as described in claim 5, characterized in that, The cavity determination module is also used for: Based on the mechanical impedance sampling information of each mechanical impedance sampling point in each degradation feature cluster segment and the mechanical impedance sampling information of each cooperative mechanical impedance sampling point in each degradation feature cluster segment, multiple impedance response association clusters are obtained for each degradation feature cluster segment. Each impedance response association cluster contains the mechanical impedance sampling information of multiple mechanical impedance sampling points and the mechanical impedance sampling information of multiple cooperative mechanical impedance sampling points. Multiple core degradation feature clusters are determined based on multiple impedance response correlation clusters of each degradation feature cluster and the void anomaly mapping of each degradation feature cluster. Based on the mechanical impedance sampling information of multiple mechanical impedance sampling points in each core deterioration feature cluster segment, the mechanical impedance sampling information of multiple coordinated mechanical impedance sampling points in each core deterioration feature cluster segment, and the acoustic resonance spectrum data of each core deterioration feature cluster segment, multiple suspected cavity anomaly areas in each core deterioration feature cluster segment are identified, and ground-penetrating radar analysis data of multiple suspected cavity anomaly areas in each core deterioration feature cluster segment are obtained. Based on the ground-penetrating radar analysis data of multiple suspected cavity anomaly areas in each core deterioration feature cluster, the grouting repair of the tunnel lining solid cavity is determined.

8. The multi-source tunnel monitoring data acquisition and processing system as described in claim 6, characterized in that, The input to the graph neural network is the audio feedback spectrum, and the output of the graph neural network is multiple deterioration feature clusters and a voiding anomaly mapping map of each deterioration feature cluster.

9. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the multi-source tunnel monitoring data acquisition and processing method as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the multi-source tunnel monitoring data acquisition and processing method as described in any one of claims 1 to 4.

Images