A tunnel collapse risk prediction method and system based on multi-parameter fusion

Abstract

The application provides a tunnel collapse risk prediction method and system based on multi-parameter fusion, and relates to the technical field of tunnel construction. The method comprises the following steps: collecting various monitoring data of a tunnel to be predicted; constructing a tunnel collapse risk index system; calculating a collapse risk index of the tunnel to be predicted according to the tunnel collapse risk index system and the various monitoring data; judging whether the collapse risk index is greater than or equal to a collapse risk index threshold value; if yes, constructing a graph data structure according to the various monitoring data; combining a one-dimensional convolutional neural network and a graph convolutional network to construct a tunnel collapse risk prediction model; inputting the various monitoring data and the graph data structure into the tunnel collapse risk prediction model to predict the collapse risk of the tunnel to be predicted, and outputting a tunnel collapse risk level. The application considers the superposition of geological conditions, construction methods and environmental factors, does not rely on single monitoring data, improves the rationality of the sample set, and is suitable for tunnel collapse risk prediction in complex environments.

Description

Technical Field

[0001] This invention relates to the field of tunnel construction technology, and in particular to a method and system for predicting tunnel collapse risk based on multi-parameter fusion. Background Technology

[0002] Increasingly, highways and railways in my country extend into mountainous areas prone to geological disasters. Tunnels, as underground structures traversing towering mountains, face numerous safety hazards during construction. These hazards arise from navigating complex terrains, variable hydrogeological conditions, and rock formations rife with adverse geological problems such as fractures, karst, and hazardous water bodies. Tunnel collapse, a serious and common safety accident, is related to multiple factors including geological conditions, construction methods, and environmental changes. Collapses caused by the combined effects of various factors, such as sudden water and mud inrushes, rock bursts, and lining cracking, are unpredictable, rapid, and highly destructive, posing serious safety risks. They can also lead to casualties, project delays, and significant economic losses and social impacts. Therefore, tunnel collapse prevention and risk assessment play a crucial role in disaster avoidance during tunnel construction.

[0003] In existing technologies, traditional tunnel collapse risk prediction methods mainly rely on expert experience and traditional engineering analogies, neglecting the superimposed effects of multiple factors such as tunnel environment, geological conditions, construction methods, and support structures. This makes it difficult to comprehensively and accurately reflect the actual risks of tunnels under different conditions and construction stages.

[0004] Furthermore, traditional methods for predicting tunnel collapse risks primarily rely on single types of monitoring data, such as support structure stress, temperature, and humidity. While these data can provide some early warnings, tunnel collapses are the result of multiple factors, and these indicators are interconnected. This often limits the accuracy and real-time performance of predictions based on a single factor, making it impossible to comprehensively describe tunnel risks in complex environments. Summary of the Invention

[0005] To address the technical problems of existing tunnel collapse risk prediction methods that rely primarily on expert experience and traditional engineering analogies, neglecting the cumulative effects of multiple factors such as tunnel environment, geological conditions, construction methods, and support structures, making it difficult to comprehensively and accurately reflect the actual risks of tunnels under different conditions and construction stages, and relying mainly on single types of monitoring data, the accuracy and real-time performance of predictions are often limited, and they cannot fully describe the risks of tunnels in complex environments, this invention provides a tunnel collapse risk prediction method and system based on multi-parameter fusion.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] First aspect:

[0008] This invention provides a method for predicting tunnel collapse risk based on multi-parameter fusion, comprising:

[0009] S1: Collect various monitoring data of the tunnel to be predicted;

[0010] S2: Construct a tunnel collapse risk indicator system;

[0011] S3: Calculate the collapse risk index of the tunnel to be predicted based on the tunnel collapse risk index system and various monitoring data;

[0012] S4: Determine whether the collapse risk index is greater than or equal to the collapse risk index threshold; if yes, proceed to S5; otherwise, return to S2.

[0013] S5: Construct a graph data structure based on various monitoring data;

[0014] S6: Combine one-dimensional convolutional neural networks and graph convolutional networks to construct a tunnel collapse risk prediction model;

[0015] S7: Input various monitoring data and graph data structures into the tunnel collapse risk prediction model, predict the collapse risk of the tunnel to be predicted, and output the tunnel collapse risk level.

[0016] The second aspect:

[0017] This invention provides a tunnel collapse risk prediction system based on multi-parameter fusion, comprising:

[0018] processor;

[0019] The memory stores computer-readable instructions that, when executed by the processor, implement the tunnel collapse risk prediction method based on multi-parameter fusion as described in the first aspect.

[0020] Third aspect:

[0021] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the tunnel collapse risk prediction method based on multi-parameter fusion as described in the first aspect.

[0022] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0023] In this embodiment of the invention, by constructing a tunnel collapse risk index system and calculating the collapse risk index of the tunnel to be predicted based on the tunnel collapse risk index system and various monitoring data, the prediction of tunnel collapse risk no longer relies on expert experience and traditional engineering analogy methods. Considering the superimposed effects of multiple factors such as tunnel environment, geological conditions, construction methods, and support structures, it is difficult to comprehensively and accurately reflect the actual risks of the tunnel under different conditions and construction stages. By constructing a graph data structure and inputting various monitoring data and graph data structure into the tunnel collapse risk prediction model, the collapse risk of the tunnel to be predicted is predicted, and the tunnel collapse risk level is output. It no longer relies on a single type of monitoring data, so that the accuracy and real-time performance of tunnel collapse risk prediction are no longer limited, and the risk of the tunnel can be comprehensively described in complex environments. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating a tunnel collapse risk prediction method based on multi-parameter fusion provided in an embodiment of the present invention;

[0026] Figure 2 This is a schematic diagram of a tunnel collapse risk prediction system based on multi-parameter fusion, provided as an embodiment of the present invention. Detailed Implementation

[0027] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0028] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0029] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0030] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0031] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0032] Reference manual attached Figure 1 The diagram shows a flowchart of a tunnel collapse risk prediction method based on multi-parameter fusion provided by an embodiment of the present invention.

[0033] This invention provides a tunnel collapse risk prediction method based on multi-parameter fusion. This method can be implemented by a tunnel collapse risk prediction device based on multi-parameter fusion, which can be a terminal or a server. The processing flow of the tunnel collapse risk prediction method based on multi-parameter fusion may include the following steps:

[0034] S1: Collect various monitoring data of the tunnel to be predicted.

[0035] Optionally, the monitoring data includes: tunnel rock mass integrity coefficient, tunnel rock mass quality grade, tunnel longitudinal wave velocity, tunnel lighting brightness, tunnel rock mass stress, tunnel support structure stress, tunnel span and height, tunnel rock mass deformation, tunnel groundwater pressure, tunnel rock mass microseismic characteristics, and tunnel temperature, all collected by different monitoring devices at multiple times.

[0036] It should be noted that the tunnel rock mass integrity coefficient is an index measuring the structural type and integrity of the tunnel rock mass. It reflects the degree of development of fault fracture zones, weak interlayers, joints, and fissures, as well as the degree of weathering. The tunnel rock mass quality grade is a parameter used to evaluate the engineering characteristics of the rock mass. It reflects not only the rock mass strength but also the evaluation criteria related to engineering characteristics. The national standard classifies it into five categories: Class I, Class II, Class III, Class IV, and Class V (not elaborated upon here). The tunnel longitudinal wave velocity refers to the speed at which sound waves propagate in the tunnel rock mass, and it is closely related to the density and elastic properties of the rock mass. Tunnel lighting brightness refers to the light intensity inside the tunnel, which has a significant impact on visibility, safety, and driver perception. Tunnel rock mass stress refers to the stress value generated by the rock mass within the tunnel under the combined influence of gravity and tectonic stress. It affects the stability of the tunnel and the design of excavation and support. Tunnel support structure stress refers to the pressure or tension borne by the tunnel support structure during use, reflecting the critical failure value and bearing capacity of the support structure. Tunnel span and height refer to the horizontal distance between the tunnel's two side walls and the vertical distance between the arch and the floor slab. These determine the tunnel's excavation face size, excavation volume, and the degree of disturbance to the rock mass. Tunnel rock mass deformation refers to the relative change in the horizontal distance between the tunnel's two side walls and the relative change in the vertical distance between the arch and the floor slab. It reflects the stress release in the tunnel rock mass and early signs of rock mass instability. Tunnel groundwater pressure refers to the fissure water pressure within the tunnel rock mass. It affects the tunnel's waterproofing design, water level control during construction, and the weakening effect on the rock mass's structural surfaces. Tunnel rock mass microseismic characteristics refer to the vibration signals emitted when fissures expand and penetrate during stress release. They reflect the distance between the current state of the rock mass and the critical value for rupture and instability. Tunnel humidity refers to the water vapor content in the air inside the tunnel. It has a significant impact on tunnel corrosion, equipment maintenance, and air quality.

[0037] In this embodiment of the invention, by comprehensively collecting various monitoring data closely related to tunnel stability and combining them with detailed indicator definitions, the scientific nature, comprehensiveness, and operability of the risk prediction model's data foundation are ensured, thereby effectively improving the accuracy and practicality of tunnel collapse risk prediction.

[0038] S2: Construct a tunnel collapse risk indicator system.

[0039] It should be noted that the tunnel collapse risk index system is a set of indicators used to systematically assess the stability and safety of tunnels. Based on the tunnel's structural characteristics, environmental conditions, and operational status, key monitoring parameters reflecting tunnel collapse risk are selected and scientifically categorized into positive and negative indicators, with corresponding scoring ranges and weights assigned to each indicator. By comprehensively analyzing the monitoring data of various indicators, the tunnel's collapse risk level can be quantified, providing a quantitative basis for risk prediction and safety decision-making. This system not only considers single factors but also reflects the overall stress, environmental changes, and structural health status of the tunnel through multi-dimensional and multi-level information fusion, possessing comprehensiveness, scientific rigor, and dynamic adaptability.

[0040] Optionally, the tunnel collapse risk index system includes multiple tunnel collapse risk indicators and corresponding scoring ranges for each tunnel collapse risk indicator.

[0041] It should be noted that those skilled in the art can set the size of the scoring range corresponding to each tunnel collapse risk indicator according to actual needs, and this invention does not limit this.

[0042] Tunnel collapse risk indicators include: positive indicators and negative indicators.

[0043] It should be noted that positive indicators refer to factors that have a positive impact on tunnel safety and stability; the higher the value of these indicators, the lower the tunnel risk. Negative indicators refer to factors that have a negative impact on tunnel safety and stability; the higher the value of these indicators, the higher the tunnel risk.

[0044] Positive indicators include: tunnel rock mass integrity coefficient, tunnel rock mass quality grade, tunnel longitudinal wave velocity, and tunnel lighting brightness.

[0045] Inverse indicators include: tunnel rock mass stress, tunnel support structure stress, tunnel span and height, tunnel rock mass deformation, tunnel groundwater pressure, tunnel rock mass microseismic characteristics, tunnel rock mass microseismic characteristics, and tunnel temperature.

[0046] In this embodiment of the invention, by constructing a scientific and reasonable tunnel collapse risk indicator system, comprehensively integrating structural characteristics, environmental factors and operational status data, and clearly distinguishing between positive and negative indicators, the scientificity and accuracy of risk assessment are improved, and standardized data support is provided for subsequent risk prediction models based on deep learning. This significantly improves the reliability and practicality of tunnel collapse risk identification and decision support.

[0047] S3: Calculate the collapse risk index of the tunnel to be predicted based on the tunnel collapse risk index system and various monitoring data.

[0048] It should be noted that the collapse risk index is a value calculated by combining the tunnel collapse risk indicator system with monitoring data. This value reflects the potential risk level of tunnel collapse.

[0049] In this embodiment of the invention, by constructing a collapse risk index based on a tunnel collapse risk indicator system and monitoring data, it is possible to achieve quantitative fusion and unified expression of multi-source monitoring data of tunnels, improve the scientificity, accuracy and dynamic adaptability of tunnel risk assessment, and provide a solid data foundation and decision support for subsequent risk prediction modeling and intelligent operation and maintenance management.

[0050] In one possible implementation, S3 specifically includes sub-steps S301 to S305:

[0051] S301: Based on each scoring interval, calculate the first weighting coefficient of various monitoring data using the Delphi method.

[0052] It should be noted that the Delphi Method is a systematic approach to forecasting and decision-making based on the opinions of a group of experts. It typically involves multiple rounds of anonymous questionnaires to collect expert judgments and suggestions on a specific issue. After each round, the results are statistically analyzed and feedback is provided, encouraging experts to gradually reach a consensus through repeated reflection and mutual reference. This method effectively integrates expert experience, reduces the influence of personal biases on decision-making, and is widely used in complex systems analysis, indicator weight determination, and future trend forecasting. It is scientific, systematic, and highly reliable.

[0053] In one possible implementation, S301 specifically includes sub-steps S3011 to S3017:

[0054] S3011. Set the initial number of experts.

[0055] S3012: Based on each scoring range, various monitoring data are scored by various experts.

[0056] S3013: Calculate the average score of various monitoring data based on the scoring results of various monitoring data.

[0057] Optionally, the average score of various monitoring data can be calculated according to the following formula:

[0058]

[0059] in, Indicates the first j The average score of the monitoring data. I Indicates the number of experts. A ij Indicates the first i The expert on the firstj Scoring of various monitoring data.

[0060] S3014: Calculate the standard deviation of various monitoring data based on the mean score.

[0061] Optionally, the standard deviation of the scores for various monitoring data can be calculated using the following formula:

[0062]

[0063] in, B j Indicates the first j The standard deviation of the scoring data.

[0064] S3015: Calculate the average standard deviation of the scores based on the standard deviation of the scores of various monitoring data.

[0065] Optionally, the standard deviation of the average rating can be calculated using the following formula:

[0066]

[0067] in, Indicates the standard deviation of the average score. J This indicates the amount of monitoring data.

[0068] S3016: Determine if the average standard deviation of the ratings is greater than the rating standard deviation threshold. If yes, increase the number of experts and return to S3012. Otherwise, proceed to S3017.

[0069] It should be noted that those skilled in the art can set the value of the standard deviation threshold for scoring according to actual needs, and this invention does not limit this.

[0070] Optionally, the number of experts can be increased according to the following formula:

[0071]

[0072] in, I new This indicates the number of newly added experts. Represents the ceiling function. B target This represents the standard deviation threshold for the scoring.

[0073] S3017: Calculate the first weighting coefficient for various monitoring data based on the average score.

[0074] Optionally, the first weighting coefficient for various monitoring data is calculated according to the following formula:

[0075]

[0076] in, Indicates the first j The first weighting coefficient for this type of monitoring data.

[0077] In this embodiment of the invention, a multi-round expert scoring and dynamic supplementation mechanism dynamically increases the number of experts when the average score standard deviation exceeds a preset threshold. This not only quantifies the differences of opinion among experts and corrects deviations in a timely manner, but also ensures that the final average score truly reflects the consensus of multiple experts. The first weighting coefficient calculated based on this average score takes into account the guiding significance of expert experience, and improves the stability and credibility of weighting through an iterative feedback mechanism. This lays a scientific and reliable foundation for subsequent subjective and objective fusion weighting combined with the entropy weight method, significantly improving the accuracy, robustness, and generalizability of tunnel collapse risk assessment results.

[0078] S302: Normalize various monitoring data.

[0079] S303: Based on the normalized monitoring data, calculate the second weight coefficient of each monitoring data using the entropy weight method.

[0080] It should be noted that the entropy weight method is an objective weighting method based on information entropy theory to determine the weights of indicators. It measures the contribution of each indicator to the overall uncertainty of the system by analyzing the dispersion of its data. The greater the change in indicator data, the more information it provides, the smaller its entropy value, and the greater its weight. Conversely, the smaller the change, the less information it provides, and the smaller its weight. The entropy weight method avoids interference from subjective human factors, objectively reflects the relative importance of each indicator in multi-indicator decision-making problems, and is widely used in fields such as comprehensive evaluation, risk analysis, and multi-attribute decision-making. It is characterized by its scientific rigor, impartiality, and ease of operation.

[0081] In one possible implementation, S303 specifically includes sub-steps S3031 to S3034:

[0082] S3031. Calculate various entropy values ​​for different monitoring data after normalization.

[0083] Optionally, the entropy value includes: information entropy value and fuzzy entropy value.

[0084] Furthermore, the specific formula for calculating information entropy is as follows:

[0085]

[0086] in, Represents the normalized th j The information entropy value of the monitoring data, ln Represents the natural logarithm function. p ( Z j) represents the normalized i-th j The probability distribution of various monitoring data.

[0087] Furthermore, the specific formula for calculating the fuzzy entropy value is as follows:

[0088]

[0089] in, Represents the normalized th j Fuzzy entropy values ​​of monitoring data m Indicates the embedding dimension. H ( ) represents the Heaviside step function. r Indicates the tolerance threshold. Represents the normalized th j The monitoring data and the first The distance between various monitoring data Represents the normalized th Various monitoring data.

[0090] It should be noted that those skilled in the art can set the size of the embedding dimension and tolerance threshold according to actual needs, and this invention does not limit them.

[0091] S3032: Calculate various coefficients of variation for normalized monitoring data based on various entropy values.

[0092] Optionally, the coefficient of variation includes: a first coefficient of variation and a second coefficient of variation.

[0093] Furthermore, the formula for calculating the coefficient of variation is as follows:

[0094]

[0095] in, Represents the normalized th j The first coefficient of variation of the monitoring data, Represents the normalized th j The second coefficient of variation of the monitoring data.

[0096] S3033: Calculate various importance indices for normalized monitoring data based on various coefficients of variation.

[0097] Optionally, the coefficient of variation includes: the first importance index and the second importance index.

[0098] Furthermore, the specific formula for calculating the importance index is as follows:

[0099]

[0100] in, Represents the normalized th j The most important index of this monitoring data, Represents the normalized th j The second most important index of the monitoring data.

[0101] S3034: Calculate the second weighting coefficient for various monitoring data based on various importance indices.

[0102] Optionally, the second weighting coefficients for various monitoring data are calculated according to the following formula:

[0103]

[0104] in, Indicates the first j The second weighted index of the monitoring data.

[0105] In this embodiment of the invention, by calculating the information entropy value and fuzzy entropy value of the normalized monitoring data respectively, the two entropy values ​​are converted into corresponding coefficients of variation, and further normalized to obtain two importance indices. These indices are then averaged to generate a second weighting coefficient. This approach can simultaneously capture the differences in the overall probability distribution and local similarity of the monitoring data, thereby comprehensively reflecting the information contribution and fuzzy differences of each indicator. It reduces the bias of the single entropy method when dealing with noise or outliers, achieving a more objective, fair, and robust weighting. This provides a solid foundation for subsequent fusion with subjective weights and input to deep learning models, significantly improving the accuracy and stability of tunnel collapse risk assessment.

[0106] S304: Calculate the final weighting coefficients for various monitoring data based on the first and second weighting coefficients.

[0107] Optionally, the final weighting coefficients for various monitoring data are calculated according to the following formula:

[0108]

[0109] in, W j Indicates the first j The final weighting coefficients for various monitoring data.

[0110] S305: Based on the final weighting coefficients of various monitoring data, the collapse risk index of the tunnel to be predicted is calculated using the ideal point method.

[0111] It should be noted that the ideal point method is a multi-attribute decision-making method. It ranks solutions by constructing ideal and negative ideal solutions and measuring their proximity to the ideal state. The ideal solution consists of the optimal values ​​of each evaluation index, while the negative ideal solution consists of the worst values. Actual solutions are then ranked based on their distance from both the ideal and negative ideal solutions. Solutions closer to the ideal solution and further away from the negative ideal solution are considered superior. The ideal point method comprehensively considers the influence of multiple indicators, intuitively reflecting the overall merits of solutions. It is widely used in engineering evaluation, risk assessment, and decision support, and is characterized by its clear logic and simple calculation.

[0112] In this embodiment of the invention, by combining the Delphi method and the entropy weight method to assign initial and objective weights to the monitoring data, and using normalization to eliminate dimensional differences, the collapse risk index of the tunnel is quantified and calculated using the ideal point method. This allows multi-source heterogeneous monitoring information to be integrated into an intuitive and comparable risk value, taking into account both expert experience and data characteristics, improving the scientific and rational nature of weight allocation, and enhancing the accuracy and stability of the risk index through standardization and comprehensive evaluation. This provides a high-quality data foundation and labels for subsequent deep learning-based tunnel risk prediction models, significantly improving the model's predictive performance and dynamic early warning capabilities, and providing reliable decision support for tunnel safety management.

[0113] In one possible implementation, S305 specifically includes sub-steps S3051 to S3055:

[0114] S3051: Calculate the positive and negative ideal points of various monitoring data.

[0115] Optionally, the positive and negative ideal points for various monitoring data can be calculated using the following formulas:

[0116]

[0117]

[0118] in, Indicates the first j The ideal point of the monitoring data max This indicates taking the maximum value. f kj Indicates in k The first time the data was collected j Such monitoring data, β 1 indicates a positive indicator. min This indicates taking the minimum value. β 2 indicates a contrarian indicator. Indicates the first j The negative ideal point of the monitoring data.

[0119] S3052: Calculate the dimensional standardization length of various monitoring data.

[0120] Optionally, the dimensionally standardized length of various monitoring data can be calculated according to the following formula:

[0121]

[0122] in, L j Indicates the first j The standardized length of the monitoring data is represented by ||, which indicates taking the absolute value. Indicates the first j The maximum value of the monitoring data, Indicates the first j The minimum value of the monitoring data.

[0123] S3053: Based on the dimensional standardization length and final weighting coefficient of various monitoring data, calculate the distance between each monitoring data point and the positive ideal point as the first distance.

[0124] Optionally, the first distance is specifically:

[0125]

[0126] in, D 1 represents the distance between the time point and the ideal point, i.e., the first distance.

[0127] S3054: Based on the dimensional standardization length and final weighting coefficient of various monitoring data, calculate the distance between each monitoring data point and the negative ideal point as the second distance.

[0128] Optionally, the second distance is specifically:

[0129]

[0130] in, D 2 represents the distance between the time point and the negative ideal point, i.e., the second distance.

[0131] S3055: Calculate the collapse risk index of the tunnel to be predicted based on the first distance and the second distance.

[0132] Optionally, the collapse risk index of the tunnel to be predicted can be calculated according to the following formula:

[0133]

[0134] in, D This indicates the collapse risk index of the tunnel to be predicted.

[0135] In this embodiment of the invention, positive and negative ideal points are calculated for each monitoring data point, and different indicators are uniformly measured using dimensional standardization. The deviation between the current state and the positive / negative ideal points is obtained using weighted Euclidean distance. Then, the tunnel collapse risk index is calculated using a relative proximity formula. This approach takes into account both the optimal / worst states of positive and negative indicators, and eliminates the dimensional differences among monitoring data points. This makes the risk index more sensitive to key risk factors and less affected by noise or outliers, achieving unified quantification and intuitive expression of multi-dimensional indicators. Furthermore, this method can dynamically update the risk index based on the latest monitoring data, providing reliable quantitative labels for subsequent deep learning models and significantly improving the scientific rigor, stability, and generalizability of tunnel risk assessment.

[0136] S4: Determine if the collapse risk index is greater than or equal to the collapse risk index threshold. If yes, proceed to S5. Otherwise, return to S2.

[0137] Specifically, when the collapse risk index is greater than or equal to the collapse risk index threshold, it is determined that the tunnel to be predicted has an anomaly and needs further prediction. When the collapse risk index is less than the collapse risk index threshold, it is determined that the tunnel to be predicted does not have an anomaly and returns to S1 to reacquire monitoring data.

[0138] It should be noted that those skilled in the art can set the value of the collapse risk index threshold according to actual needs, and this invention does not limit it.

[0139] In this embodiment of the invention, by introducing a judgment between the collapse risk index and the collapse risk index threshold, conventional monitoring and in-depth prediction are linked into a two-level process of "lightweight screening → in-depth analysis". This not only significantly saves computing and manpower costs, but also enables rapid early warning, dynamic threshold adjustment and system scalability, and comprehensively improves the real-time performance, robustness and safety assurance level of the tunnel monitoring system.

[0140] S5: Construct a graph data structure based on various monitoring data.

[0141] In this embodiment of the invention, by constructing a graph data structure, the temporal evolution relationship of monitoring data and the physical / topological association across indicators are organically integrated, enabling the graph convolutional network to make full use of multi-layer and multi-dimensional structured information, which significantly enhances the expressive power, robustness and accuracy of the tunnel collapse risk prediction model.

[0142] In one possible implementation, S5 specifically includes sub-steps S501 to S503:

[0143] S501: Construct multiple single-layer graph data structures based on single monitoring data.

[0144] Optionally, multiple single-layer graph data structures based on single monitoring data can be constructed according to the following formula:

[0145]

[0146]

[0147] in, Indicates in k Time and The first time the data was collected j The cosine similarity between monitoring data is the weight of the edges in a single-layer graph data structure. cos ( ) represents the cosine function. F ( ) represents the Fourier transform. Indicates in The first time the data was collected j Such monitoring data, ε j Indicates based on the first j The threshold of edges in a single-layer graph data structure constructed from monitoring data. num ( ) indicates the number of edges.

[0148] It should be noted that those skilled in the art can set the threshold value of the edges in a single-layer graph data structure according to actual needs, and this invention does not limit this.

[0149] S502: Construct a multi-layer graph data structure based on the combination of various monitoring data.

[0150] Optionally, a multi-layer graph data structure based on the combination of various monitoring data can be constructed according to the following formula:

[0151]

[0152] in, Indicates in k The first time the data was collected j The monitoring data and the first The cosine similarity between monitoring data is equivalent to the edge weights in a multi-layer graph data structure. Indicates in k The first time the data was collected Such monitoring data, T This indicates the topology of the monitoring device location.

[0153] S503: Construct a graph data structure based on each single-level graph data structure and multi-level graph data structure.

[0154] Optionally, the graph data structure is as follows:

[0155]

[0156] in, a This represents a hyperadjacency matrix, i.e., a graph data structure.

[0157] In this embodiment of the invention, by constructing single-layer graphs and multi-layer graphs respectively, and splicing the two into a graph data structure, the graph convolutional network can aggregate the temporal evolution information of each indicator and integrate the linkage relationship across indicators when transmitting messages. This not only suppresses temporal noise interference and reduces computational complexity, but also provides deep learning models with multi-dimensional, rich and robust structured features, thereby significantly enhancing the accuracy and reliability of tunnel collapse risk prediction.

[0158] S6: Combine one-dimensional convolutional neural networks and graph convolutional networks to construct a tunnel collapse risk prediction model.

[0159] It's important to note that a one-dimensional convolutional neural network (1D-CNN) is a deep learning model specifically designed for processing one-dimensional sequential data. By sliding convolutional kernels across the input data to extract local features, it effectively captures temporal patterns and local dependencies within the sequence. It typically consists of convolutional layers, pooling layers, and fully connected layers, featuring parameter sharing and local connectivity, which reduces computational complexity while enhancing feature extraction capabilities. 1D-CNNs are widely used in time series analysis, speech recognition, and signal processing, automatically learning key features from sequential data to improve modeling accuracy and generalization ability.

[0160] It's important to note that Graph Convolutional Networks (GCNs) are deep learning models specifically designed for processing graph-structured data. They capture the relationships and topological information between nodes in a graph by performing "convolution" operations between nodes and their neighbors, weighting and aggregating the features of each node with those of its neighbors. Unlike traditional convolutional neural networks, which can only operate on regular grids (such as images), GCNs can operate directly on arbitrarily sparse, non-Euclidean graphs. Through multi-layer stacking, nodes gradually fuse features from more distant neighbors, thereby learning the embedding representation of nodes within the global or local structure. GCNs are widely used for tasks such as node classification, graph classification, and link prediction.

[0161] In this embodiment of the invention, by combining 1D-CNN with GCN, it is possible to not only deeply mine the temporal patterns of tunnel monitoring data, but also capture the structured relationships between multiple indicators and multiple time points. The two complement each other, which not only enhances the model's adaptability and robustness to multi-source heterogeneous data, but also lays a solid foundation for subsequent attention fusion and deep prediction, thereby significantly improving the accuracy and reliability of tunnel collapse risk prediction.

[0162] S7: Input various monitoring data and graph data structures into the tunnel collapse risk prediction model, predict the collapse risk of the tunnel to be predicted, and output the tunnel collapse risk level.

[0163] Optionally, the tunnel collapse risk levels include no tunnel collapse risk, Level 1 tunnel collapse risk, Level 2 tunnel collapse risk, and Level 3 tunnel collapse risk.

[0164] It should be noted that "no tunnel collapse risk" means the tunnel is in a completely safe state. Level 1 tunnel collapse risk means the tunnel has a slight risk of collapse. Level 2 tunnel collapse risk means the tunnel has a moderate risk of collapse. Level 3 tunnel collapse risk means the tunnel has a high risk of collapse.

[0165] In this embodiment of the invention, by combining monitoring data with graph data structure to predict tunnel collapse risk and classify risk levels, the invention can provide an accurate, scientific, and dynamic risk assessment tool to help managers take appropriate measures according to different risk levels, achieve refined management of tunnel health status, ensure tunnel operation safety, and improve resource allocation and emergency response efficiency.

[0166] In one possible implementation, S7 specifically includes sub-steps S701 to S710:

[0167] S701: Inputs various monitoring data into a one-dimensional convolutional neural network and outputs the first feature vector.

[0168] S702: Input the graph data structure into the graph convolutional network and output the second feature vector.

[0169] S703: Concatenate the first eigenvector and the second eigenvector to obtain the composite eigenvector.

[0170] Optionally, the comprehensive feature vector is specifically as follows:

[0171]

[0172] in, Represents the comprehensive feature vector. Represents the set of real numbers. B Indicates the total number of monitoring data. OC This represents the number of channels in a one-dimensional convolutional neural network. M Representing feature dimension, Represents the first eigenvector. This represents the second eigenvector.

[0173] S704: Based on the comprehensive feature vector, calculate the channel attention value through a feedforward neural network.

[0174] It's important to note that a feedforward neural network (FNN) is a fundamental and widely used artificial neural network architecture. It consists of an input layer, one or more hidden layers, and an output layer connected sequentially. Data flows in a single direction within the network, with no feedback or loops. Neurons in each layer process the input signal through weighted summation and activation functions, passing the results to the next layer, extracting higher-order features layer by layer. Feedforward neural networks are characterized by their simple structure, high computational efficiency, and ease of training. They are widely used in tasks such as classification, regression, and feature extraction, and are one of the fundamental frameworks of deep learning models.

[0175] Optionally, the channel attention value is calculated using a feedforward neural network according to the following formula:

[0176]

[0177] in, e c In a neural network, the first... c Channel attention values ​​for each channel. LeakyReLU ( ) represents the LeakyReLU activation function. This represents the weight matrix of the feedforward neural network. In a neural network, the first... c The combined feature vector of each channel.

[0178] S705: Normalize the attention score to obtain the channel attention coefficient.

[0179] Optionally, the attention score can be normalized according to the following formula to obtain the channel attention coefficient:

[0180]

[0181] in, α c In a neural network, the first... c Channel attention coefficients for each channel. softmax ( ) represents the softmax activation function. exp ( ) represents an exponential function.

[0182] S706: Based on the channel attention coefficient, the comprehensive feature vector is weighted and fused to obtain the first final feature vector.

[0183] Optionally, the comprehensive feature vectors are weighted and fused according to the following formula to obtain the first final feature vector:

[0184]

[0185] in, This represents the first final eigenvector.

[0186] S707: Based on the first final feature vector, determine the second final feature vector through a channel attention mechanism.

[0187] It's important to note that channel attention is a method used to enhance the feature representation capabilities of neural networks. It aims to adaptively allocate weights based on the importance of features from different channels. By globally modeling the features of each channel, the network learns the level of attention given to different channels, enabling it to highlight key features and suppress irrelevant or redundant information. Channel attention typically involves three steps: feature aggregation, weight calculation, and weighted fusion. This enhances the model's ability to perceive key information and is widely used in image processing, natural language processing, and time-series data analysis, effectively improving the model's expressive power and performance.

[0188] Optionally, the comprehensive feature vectors are weighted and fused according to the following formula to obtain the first final feature vector:

[0189]

[0190] in, This represents the second final eigenvector. CAM ( ) channel attention mechanism.

[0191] S708: Input the first final feature vector into the graph convolutional network and output the first initial predicted probability of each tunnel collapse risk level.

[0192] S709: Input the second final feature vector into a one-dimensional convolutional neural network to output the second initial prediction probability for each tunnel collapse risk level.

[0193] S710: Based on the first and second initial prediction probabilities of each tunnel collapse risk level, calculate the final prediction probability of each tunnel collapse risk level, and output the tunnel collapse risk level with the highest final prediction probability.

[0194] In this embodiment of the invention, by combining 1D-CNN and Graph Convolutional Network (GCN) and introducing a channel attention mechanism, different types of monitoring data can be effectively fused, improving the accuracy and robustness of the tunnel collapse risk prediction model. 1D-CNN is used to process time-series data and capture temporal patterns, while GCN is used to capture the spatial relationships between tunnel monitoring points, learning topological information between nodes through graph convolution operations. In the feature fusion stage, the channel attention mechanism adaptively assigns weights based on the importance of each feature channel, highlighting key features and suppressing redundant information, thereby further enhancing the model's expressive power. Finally, by weighted fusion of feature vectors and combining them with a feedforward neural network, the model can accurately predict the tunnel collapse risk level, providing reliable safety warnings and decision support, and offering a scientific basis for tunnel management and maintenance.

[0195] In one possible implementation, S710 specifically includes sub-steps S7101 to S7106:

[0196] S7101: The first and second initial prediction probabilities of each tunnel collapse risk level are fused to obtain the fused initial prediction probability.

[0197] Optionally, the initial predicted probabilities are fused as follows:

[0198]

[0199] in, P k Indicates in k The first time the data was collected j The initial prediction probability of the fusion of various monitoring data This represents the output of a one-dimensional convolutional neural network. k The first time the data was collected j The second initial prediction probability of the monitoring data, This represents the output of the graph convolutional neural network. k The first time the data was collected j The first initial prediction probability of the monitoring data.

[0200] S7102: Normalize the initial prediction probability of the fusion.

[0201] Optionally, the initial prediction probability of the fusion can be normalized according to the following formula:

[0202]

[0203] in, Indicates the normalized result. k The first time the data was collected j This type of monitoring data belongs to the first nThe fusion of initial predicted probabilities for each tunnel collapse risk level P kjn Indicates in k The first time the data was collected j This type of monitoring data belongs to the first n The fusion of initial predicted probabilities for each tunnel collapse risk level This indicates the total number of tunnel collapse risk levels.

[0204] S7103: Calculate the information entropy based on the fused initial prediction probability after normalization.

[0205] Optionally, the information entropy can be calculated according to the following formula:

[0206]

[0207] in, kj Indicates the normalized result. k The first time the data was collected j Information entropy of monitoring data.

[0208] S7104: Normalize the information entropy to obtain the risk weight.

[0209] Optionally, the initial prediction probability of the fusion can be normalized according to the following formula:

[0210]

[0211] in, ω kj Indicates the normalized result. k The first time the data was collected j Risk weights for various monitoring data.

[0212] S7105: Calculate the final predicted probability of each tunnel collapse risk level based on the risk weight and the normalized initial prediction probability.

[0213] Optionally, the final predicted probability of each tunnel collapse risk level is calculated according to the following formula:

[0214]

[0215] in, Indicates in k At this moment, it belongs to the first n The final predicted probability of each tunnel collapse risk level.

[0216] S7106: Select the tunnel collapse risk level with the highest final prediction probability for output.

[0217] Specifically, the tunnel collapse risk levels at the current moment are sorted in order of their final predicted probabilities from low to high, and the tunnel collapse risk level with the highest ranking is output.

[0218] In this embodiment of the invention, by combining the prediction results of 1D-CNN and Graph Convolutional Network (GCN), and employing information entropy, normalization, and risk weighting mechanisms, accurate prediction of tunnel collapse risk can be achieved. This method optimizes the accuracy and reliability of prediction results by fusing prediction results from different models and comprehensively considering the temporal and spatial characteristics of tunnel monitoring data. Simultaneously, information entropy calculation and risk weight assignment can adaptively adjust the importance of each risk level, improving the model's ability to handle uncertain data. Finally, through channel attention mechanisms and normalization, the model can accurately assess the risk level of tunnel collapse, providing scientific and quantitative decision support for tunnel safety management.

[0219] Reference manual attached Figure 2 The diagram shows a structural schematic of a tunnel collapse risk prediction system based on multi-parameter fusion provided by the present invention.

[0220] The present invention also provides a tunnel collapse risk prediction system 20 based on multi-parameter fusion, applied to the above-mentioned tunnel collapse risk prediction method based on multi-parameter fusion, comprising:

[0221] Processor 201;

[0222] The memory 202 stores computer-readable instructions, which, when executed by the processor 201, implement the tunnel collapse risk prediction method based on multi-parameter fusion as described in the method embodiment.

[0223] The tunnel collapse risk prediction system 20 based on multi-parameter fusion provided by the present invention can execute the tunnel collapse risk prediction method based on multi-parameter fusion described above and achieve the same or similar technical effects. To avoid duplication, the present invention will not elaborate further.

[0224] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0225] In this embodiment of the invention, by constructing a tunnel collapse risk index system and calculating the collapse risk index of the tunnel to be predicted based on the tunnel collapse risk index system and various monitoring data, the prediction of tunnel collapse risk no longer relies on expert experience and traditional engineering analogy methods. Considering the superimposed effects of multiple factors such as tunnel environment, geological conditions, construction methods, and support structures, it is difficult to comprehensively and accurately reflect the actual risks of the tunnel under different conditions and construction stages. By constructing a graph data structure and inputting various monitoring data and graph data structure into the tunnel collapse risk prediction model, the collapse risk of the tunnel to be predicted is predicted, and the tunnel collapse risk level is output. It no longer relies on a single type of monitoring data, so that the accuracy and real-time performance of tunnel collapse risk prediction are no longer limited, and the risk of the tunnel can be comprehensively described in complex environments.

[0226] It should be understood that the processor in the embodiments of the present invention can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0227] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0228] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0229] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0230] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0231] It should be understood that, in various embodiments of the present invention, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0232] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0233] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0234] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0235] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0236] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0237] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0238] This invention provides a computer-readable storage medium storing a computer program thereon, characterized in that, when executed by a processor, the program implements the tunnel collapse risk prediction method based on multi-parameter fusion as described in the method embodiment.

[0239] The present invention provides a computer-readable storage medium that can implement the steps and effects of the tunnel collapse risk prediction method based on multi-parameter fusion in the above-described method embodiments. To avoid repetition, the present invention will not elaborate further.

[0240] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0241] In this embodiment of the invention, by constructing a tunnel collapse risk index system and calculating the collapse risk index of the tunnel to be predicted based on the tunnel collapse risk index system and various monitoring data, the prediction of tunnel collapse risk no longer relies on expert experience and traditional engineering analogy methods. Considering the superimposed effects of multiple factors such as tunnel environment, geological conditions, construction methods, and support structures, it is difficult to comprehensively and accurately reflect the actual risks of the tunnel under different conditions and construction stages. By constructing a graph data structure and inputting various monitoring data and graph data structure into the tunnel collapse risk prediction model, the collapse risk of the tunnel to be predicted is predicted, and the tunnel collapse risk level is output. It no longer relies on a single type of monitoring data, so that the accuracy and real-time performance of tunnel collapse risk prediction are no longer limited, and the risk of the tunnel can be comprehensively described in complex environments.

[0242] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0243] The following points need to be explained:

[0244] (1) The accompanying drawings of the embodiments of the present invention only involve the structures involved in the embodiments of the present invention. Other structures can refer to the general design.

[0245] (2) For clarity, the thickness of layers or regions is enlarged or reduced in the drawings used to describe embodiments of the invention, i.e., these drawings are not drawn to scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being “above” or “below” another element, the element may be “directly” located “above” or “below” the other element or there may be intermediate elements.

[0246] (3) Where there is no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0247] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting tunnel collapse risk based on multi-parameter fusion, characterized in that, include: S1: Collect various monitoring data of the tunnel to be predicted; The monitoring data includes: tunnel rock mass integrity coefficient, tunnel rock mass quality grade, tunnel longitudinal wave velocity, tunnel lighting brightness, tunnel rock mass stress, tunnel support structure stress, tunnel span and height, tunnel rock mass deformation, tunnel groundwater pressure, tunnel rock mass microseismic characteristics, and tunnel temperature, all collected by different monitoring devices at multiple times. S2: Construct a tunnel collapse risk indicator system; The tunnel collapse risk index system includes multiple tunnel collapse risk indicators and a scoring range corresponding to each tunnel collapse risk indicator; The tunnel collapse risk indicators include: positive indicators and negative indicators; The positive indicators include: the tunnel rock mass integrity coefficient, the tunnel rock mass quality grade, the tunnel longitudinal wave velocity, and the tunnel lighting brightness; The inverse indicators include: the tunnel rock mass stress, the tunnel support structure stress, the tunnel span and height, the tunnel rock mass deformation, the tunnel groundwater pressure, the tunnel rock mass microseismic characteristics, and the tunnel temperature; S3: Calculate the collapse risk index of the tunnel to be predicted based on the tunnel collapse risk index system and various monitoring data; Specifically, S3 includes: S301: Based on each of the aforementioned scoring intervals, calculate the first weighting coefficient for each of the aforementioned monitoring data using the Delphi method; S302: Normalize the various monitoring data; S303: Based on the normalized monitoring data, calculate the second weighting coefficient of each monitoring data using the entropy weighting method; S304: Calculate the final weighting coefficients for various monitoring data based on the first weighting coefficient and the second weighting coefficient; S305: Based on the final weighting coefficients of the various monitoring data, calculate the collapse risk index of the tunnel to be predicted using the ideal point method; S4: Determine whether the collapse risk index is greater than or equal to the collapse risk index threshold; if yes, proceed to S5; otherwise, return to S1. S5: Construct a graph data structure based on the various monitoring data; S6: Combine one-dimensional convolutional neural networks and graph convolutional networks to construct a tunnel collapse risk prediction model; S7: Input the various monitoring data and the graph data structure into the tunnel collapse risk prediction model, predict the collapse risk of the tunnel to be predicted, and output the tunnel collapse risk level.

2. The tunnel collapse risk prediction method based on multi-parameter fusion according to claim 1, characterized in that, S301 specifically includes: S3011; Set the initial number of experts; S3012: Based on each of the aforementioned scoring intervals, various experts score the various monitoring data. S3013: Calculate the average score of each monitoring data based on the scoring results of the various monitoring data; S3014: Calculate the standard deviation of the scores for each of the monitoring data based on the mean score; S3015: Calculate the average standard deviation of the scores based on the standard deviation of the scores of the various monitoring data; S3016: Determine whether the average standard deviation of the rating is greater than the standard deviation threshold of the rating; if yes, increase the number of experts and return to S3012; otherwise, proceed to S3017. S3017: Calculate the first weighting coefficient for each of the monitoring data based on the average score.

3. The tunnel collapse risk prediction method based on multi-parameter fusion according to claim 1, characterized in that, Specifically, S303 includes: S3031; Calculate various entropy values ​​of different monitoring data after normalization; S3032: Calculate various coefficients of variation of the normalized monitoring data based on the various entropy values; S3033: Calculate various importance indices for the normalized monitoring data based on the various coefficients of variation. S3034: Calculate the second weighting coefficient of each of the monitoring data based on the various importance indices.

4. The tunnel collapse risk prediction method based on multi-parameter fusion according to claim 1, characterized in that, Specifically, S305 includes: S3051: Calculate the positive and negative ideal points of various monitoring data; S3052: Calculate the dimensionally standardized length of all the monitoring data; S3053: Based on the dimensionally standardized length and final weighting coefficient of each of the monitoring data, calculate the distance between each of the monitoring data and the positive ideal point as the first distance; S3054: Based on the dimensional standardized length and final weighting coefficient of each of the monitoring data, calculate the distance between each of the monitoring data and the negative ideal point as a second distance; S3055: Calculate the collapse risk index of the tunnel to be predicted based on the first distance and the second distance.

5. The tunnel collapse risk prediction method based on multi-parameter fusion according to claim 1, characterized in that, S5 specifically includes: S501: Construct multiple single-layer graph data structures based on single monitoring data; S502: Construct a multi-layer graph data structure based on the combination of various monitoring data; S503: Construct the graph data structure based on each of the single-layer graph data structures and the multi-layer graph data structures.

6. The tunnel collapse risk prediction method based on multi-parameter fusion according to claim 1, characterized in that, The tunnel collapse risk levels include no tunnel collapse risk, Level 1 tunnel collapse risk, Level 2 tunnel collapse risk, and Level 3 tunnel collapse risk; Specifically, S7 includes: S701: Input the various monitoring data into the one-dimensional convolutional neural network and output the first feature vector; S702: Input the graph data structure into the graph convolutional network and output the second feature vector; S703: Concatenate the first feature vector and the second feature vector to obtain a comprehensive feature vector; S704: Based on the comprehensive feature vector, calculate the channel attention value using a feedforward neural network; S705: Normalize the attention score to obtain the channel attention coefficient; S706: Based on the channel attention coefficients, the comprehensive feature vector is weighted and fused to obtain the first final feature vector; S707: Based on the first final feature vector, determine the second final feature vector through a channel attention mechanism; S708: Input the first final feature vector into the graph convolutional network and output the first initial prediction probability of each tunnel collapse risk level; S709: Input the second final feature vector into the one-dimensional convolutional neural network and output the second initial prediction probability of each tunnel collapse risk level; S710: Calculate the final prediction probability of each tunnel collapse risk level based on the first and second initial prediction probabilities of each tunnel collapse risk level, and select the tunnel collapse risk level with the highest final prediction probability for output.

7. The tunnel collapse risk prediction method based on multi-parameter fusion according to claim 6, characterized in that, The S710 specifically includes: S7101: The first and second initial prediction probabilities of each of the tunnel collapse risk levels are fused to obtain the fused initial prediction probability; S7102: Normalize the initial prediction probability of the fusion; S7103: Calculate the information entropy based on the fused initial prediction probability after normalization. S7104: Normalize the information entropy to obtain the risk weight; S7105: Calculate the final prediction probability of each tunnel collapse risk level based on the risk weights and the normalized initial prediction probability. S7106: Select the tunnel collapse risk level with the highest predicted probability and output it.

8. A tunnel collapse risk prediction system based on multi-parameter fusion, characterized in that, include: processor; A memory storing computer-readable instructions, which, when executed by the processor, implement the tunnel collapse risk prediction method based on multi-parameter fusion as described in any one of claims 1 to 7.

Images