Tunnel construction dynamic risk grading management and control method based on multi-source monitoring data fusion

Abstract

This invention belongs to the field of tunnel construction safety technology and provides a dynamic risk classification and control method for tunnel construction based on multi-source monitoring data fusion. The method includes collecting multi-source monitoring data for tunnel construction and constructing a risk feature vector using information entropy; calculating the Mahalanobis distance between the risk feature vector and the feature centers of various risk modes, obtaining the probability of occurrence of each risk type through membership degrees, and determining the highest probability risk type; determining a risk propagation path set based on a directed risk propagation network, starting with the highest probability risk type; recursively calculating the probability of occurrence of each node along the propagation path using propagation probability, and weighting the damage risk assessment value of each node with damage performance feature data; multiplying the two values ​​to obtain a comprehensive damage risk value; taking the maximum value of the comprehensive damage risk value for all path nodes as the extreme value node, and performing risk classification and control accordingly. This invention achieves dynamic quantitative assessment and classification control of tunnel construction risks.

Description

Technical Field

[0001] This invention belongs to the field of tunnel construction safety technology, and in particular relates to a method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion. Background Technology

[0002] The tunnel construction environment is characterized by complex geological conditions, variable surrounding rock properties, and frequent construction disturbances, which leads to the nonlinear evolution and chain transmission of safety risks.

[0003] Traditional risk management methods mainly rely on threshold alarm mechanisms based on single monitoring indicators, such as setting warning lines solely based on the absolute value of crown settlement or convergence displacement. Such methods can only identify local anomalies and cannot reflect the coupling relationships and dynamic propagation patterns among multiple risk factors. Some engineering practices employ risk matrix methods based on expert experience for assessment; however, this method is highly dependent on subjective judgment, has a fixed assessment cycle, and is updated slowly, making it difficult to adapt to the dynamic construction scenario of continuous tunnel excavation.

[0004] While some studies have attempted to improve assessment accuracy by introducing fuzzy comprehensive evaluation or analytic hierarchy process (AHP), these calculations are typically completed before construction, constituting offline static analysis and unable to adaptively adjust risk status based on real-time monitoring data. In research on risk propagation mechanisms, existing technologies often employ fault tree analysis or event tree analysis for causal reasoning. However, these methods focus only on the deduction of a single pre-defined path, failing to construct a complete directed propagation network covering the interactions of multiple risk types, resulting in a one-sided characterization of the risk chain diffusion process. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion, thus solving the aforementioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a dynamic risk classification and control method for tunnel construction based on multi-source monitoring data fusion, which specifically includes:

[0007] Collect multi-source monitoring data during tunnel construction, and determine risk feature vectors based on the multi-source monitoring data during tunnel construction;

[0008] Based on the risk feature vector, determine the probability of occurrence of each tunnel construction risk type, and based on the probability of occurrence of each tunnel construction risk type, determine the tunnel construction risk type with the highest probability.

[0009] Construct a directed network for the propagation of tunnel construction risk types, and traverse the directed network for the propagation of tunnel construction risk types starting from the tunnel construction risk type with the highest probability to determine the set of risk propagation paths;

[0010] The damage manifestation characteristic data of each node in the risk propagation path is obtained. Based on the damage manifestation characteristic data of each node in the risk propagation path and the probability of occurrence of the tunnel construction risk type, the comprehensive damage risk value of each node in the risk propagation path is determined. Among them, the damage manifestation characteristic data refers to the damage situation that can reflect when the risk type occurs.

[0011] Based on the comprehensive damage risk value of each node in the risk propagation path, tunnel construction risk is classified and controlled.

[0012] Based on the above technical solutions, the present invention also provides the following optional technical solutions:

[0013] Further technical solution: The determination of the risk feature vector specifically includes:

[0014] The multi-source monitoring data is standardized to determine the standardized values ​​of the monitoring data; the standardization method for the multi-source monitoring data is Min-Max standardization.

[0015] The information entropy of the monitoring data is determined based on the standardized values ​​of the monitoring data;

[0016] Based on the information entropy of the monitoring data, the risk feature vector is determined.

[0017] Further technical solution: The determination of the information entropy of the monitoring data specifically includes:

[0018] Through the formula: ;

[0019] Determine the information entropy of the monitoring data ;

[0020] In the above formula, This refers to the information entropy of the i-th monitoring data. This refers to the standardized value of the i-th monitoring data at time t, and N refers to the total number of times.

[0021] Further technical solution: The determination of the risk feature vector also includes:

[0022] Through the formula: ;

[0023] Determine the components of the risk feature vector ;

[0024] In the above formula, This refers to the i-th component of the risk feature vector. This refers to the standardized value of the i-th monitoring data at the current moment. This refers to the fusion weight;

[0025] Based on the components of the risk feature vector Determine the risk feature vector;

[0026] The method for obtaining the fusion weights is as follows:

[0027] Through the formula: ;

[0028] Generate fusion weights ;

[0029] In the above formula, This refers to the information entropy of the j-th monitoring data. This refers to the information entropy of the i-th monitoring data, and M refers to the number of monitoring data.

[0030] Further technical solution: The determination of the probability of occurrence of tunnel construction risk types specifically includes:

[0031] Based on the risk feature vector, determine the Mahalanobis distance between the risk feature vector and the feature center vector of each type of risk mode in tunnel construction;

[0032] The membership degree of a risk pattern is determined based on the Mahalanobis distance between the risk feature vector and the feature center vector of each risk type.

[0033] Based on the membership degree of the risk model, determine the probability of occurrence of the risk type in tunnel construction.

[0034] Further technical solution: The determination of the membership degree of the risk mode specifically includes:

[0035] Through the formula: ;

[0036] Determine the membership degree of risk models ;

[0037] In the above formula, This refers to the membership degree of the k-th risk pattern. This refers to the Mahalanobis distance between the risk feature vector and the feature center vector of the k-th risk type. This refers to the preset scaling parameters;

[0038] The determination of the probability of occurrence of risk types in tunnel construction also includes:

[0039] Through the formula: ;

[0040] Determine the probability of occurrence of different risk types in tunnel construction. ;

[0041] In the above formula, This refers to the probability of the k-th type of risk occurring during tunnel construction. This refers to the membership degree of the r-th risk pattern, and m refers to the number of risk types.

[0042] Further technical solution: The determination of the comprehensive damage risk value of each node in the risk propagation path specifically includes:

[0043] Based on the damage performance characteristics data of each node in the risk propagation path, determine the damage risk assessment value of each node in the risk propagation path.

[0044] Based on the probability of occurrence of different types of tunnel construction risks, determine the probability of occurrence of each node in the risk propagation path;

[0045] Based on the damage risk assessment value of each node in the risk propagation path and the occurrence probability of each node in the risk propagation path, the comprehensive damage risk value of each node in the risk propagation path is determined.

[0046] The determination of the comprehensive damage risk value of each node in the risk propagation path further includes:

[0047] Through the formula: ;

[0048] Determine the comprehensive damage risk value of nodes in the risk propagation path. ;

[0049] In the above formula, This refers to the overall damage risk value of node e in the risk propagation path. This refers to the z-th node in the risk propagation path when node e is the endpoint. This refers to the number of nodes in the risk propagation path that terminate at node e. This refers to the probability of node e occurring in the risk propagation path.

[0050] Further technical solution: The determination of the damage risk assessment value of each node in the risk propagation path specifically includes:

[0051] Through the formula: ;

[0052] Determine the damage risk assessment value of each node in the risk propagation path. ;

[0053] In the above formula, This refers to the normalized value of the damage manifestation feature data of node e in the risk propagation path. This refers to the weighting coefficient of the l-th damage manifestation feature, where L refers to the number of damage manifestation feature data.

[0054] Further technical solution: The determination of the occurrence probability of each node in the risk propagation path specifically includes:

[0055] Through the formula: ;

[0056] Determine the probability of node e occurring in the risk propagation path. ;

[0057] In the above formula, This refers to the probability of occurrence of the tunnel construction risk type corresponding to the highest probability tunnel construction risk type. This refers to the node where node e is the endpoint in the risk propagation path. With nodes The probability of propagation between them This refers to the number of nodes in the risk propagation path when node e is the endpoint;

[0058] Among them, the propagation probability The specific method of obtaining it is as follows:

[0059] Retrieve nodes from historical data The frequency and nodes of the corresponding risk types When the corresponding risk type occurs, it causes a node The number of times it occurs;

[0060] Nodes in historical data The frequency and nodes of corresponding risk types When the corresponding risk type occurs, it causes a node The propagation probability is obtained by comparing the number of occurrences. .

[0061] Further technical solution: The method of classifying and controlling tunnel construction risks based on the comprehensive damage risk value of each node in the risk propagation path specifically includes:

[0062] Based on the comprehensive damage risk value of each node in the risk propagation path, extreme value nodes are determined; specifically, this includes: screening the maximum value of the comprehensive damage risk value of each node in all risk propagation paths, and marking the node corresponding to the maximum value of the comprehensive damage risk value as an extreme value node;

[0063] Based on the comprehensive damage risk value corresponding to the extreme node, tunnel construction risk is classified and controlled.

[0064] This invention provides a method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion, which has the following advantages compared with the prior art:

[0065] This invention effectively solves the problems of distortion in multi-source data fusion, static risk assessment, and incomplete damage quantification by introducing information entropy to quantify the quality of monitoring data, dynamically calculating the probability of risk types, and constructing a comprehensive damage model based on risk propagation paths. It has the advantages of being able to objectively quantify the information contribution of multi-source monitoring data, realize dynamic probability calculation of risk types, and conduct comprehensive damage risk assessment based on risk propagation paths, thereby improving the scientificity and real-time performance of risk classification and control. Attached Figure Description

[0066] Figure 1 This is a flowchart illustrating the dynamic risk classification and control method for tunnel construction based on multi-source monitoring data fusion provided by the present invention.

[0067] Figure 2 This is a schematic diagram of the process S10 provided by the present invention.

[0068] Figure 3 This is a schematic diagram of the process S20 provided by the present invention.

[0069] Figure 4 This is a schematic diagram of the S40 process provided by the present invention.

[0070] Figure 5 This is a schematic diagram of the S50 process provided by the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0072] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0073] Please see Figure 1 The present invention provides a method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion, comprising the following steps:

[0074] S10: Collect multi-source monitoring data during tunnel construction, and determine the risk feature vector based on the multi-source monitoring data during tunnel construction;

[0075] Among them, multi-source monitoring data refers to real-time or near-real-time data obtained during tunnel construction through various sensors, instruments and manual observation, reflecting the status of multiple aspects such as tunnel surrounding rock, support structure, groundwater, and construction environment. These data may include, but are not limited to, surrounding rock displacement, support stress, anchor bolt axial force, lining deformation, groundwater pressure, surface settlement, and environmental vibration. Their diversity aims to comprehensively capture the complex conditions of the construction site.

[0076] The risk feature vector is a mathematical vector formed by processing and fusing the collected multi-source monitoring data. It can quantify and characterize the current risk status of tunnel construction. Each component of the vector usually corresponds to a processed monitoring indicator or a combination thereof, which is used to comprehensively reflect multiple dimensions of potential risk information.

[0077] S20: Based on the risk feature vector, determine the probability of occurrence of each tunnel construction risk type, and based on the probability of occurrence of each tunnel construction risk type, determine the tunnel construction risk type with the highest probability.

[0078] S30: Construct a directed network for the propagation of tunnel construction risk types, starting with the tunnel construction risk type with the highest probability, and traverse the directed network for the propagation of tunnel construction risk types to determine the set of risk propagation paths;

[0079] S40: Obtain damage performance characteristic data of each node in the risk propagation path, and determine the comprehensive damage risk value of each node in the risk propagation path based on the damage performance characteristic data of each node in the risk propagation path and the probability of occurrence of tunnel construction risk types.

[0080] Among them, damage performance characteristic data refers to the damage situation that can reflect when the risk type occurs. That is, when a certain risk type occurs, it is a quantitative description of the specific damage or impact that may be caused. This data may include, but is not limited to, the amount of property loss, the time of project delay, the number of casualties, the degree of equipment damage, etc., and is used to assess the potential consequences of risk events.

[0081] S50: Based on the comprehensive damage risk value of each node in the risk propagation path, conduct risk classification and control for tunnel construction.

[0082] Specifically, in step S10, multi-source monitoring data during tunnel construction is first collected. This data can be collected by on-site personnel through manual inspection and visual observation, recording relevant parameters such as the width of surrounding rock cracks, the flow rate of seepage water, and the deformation of the support structure. Alternatively, data can be read periodically or in real-time using various sensors deployed inside the tunnel, such as displacement gauges, stress gauges, and pressure gauges. Subsequently, a risk feature vector is determined based on the collected multi-source monitoring data. This process can involve a simple arithmetic average of all collected data, or engineers can assign different weights to each data point based on its historical importance or the sensitivity of the current construction stage, and then perform a weighted average to form a risk feature vector that represents the current state of tunnel construction.

[0083] In step S20, the probability of occurrence of each tunnel construction risk type is determined based on the established risk feature vectors. This process can be based on historical accident statistics, by consulting a pre-established table of correspondences between risk patterns and feature vectors, to assess the likelihood of occurrence of various potential risks (such as rock instability, water inrush, and collapse), thereby obtaining the probability of occurrence for each risk type. Based on this, the risk type with the highest probability among all risk types is selected and determined as the highest-probability tunnel construction risk type.

[0084] In step S30, a directed network for the propagation of tunnel construction risk types is constructed. This network can be built based on in-depth analysis of historical tunnel accident cases, identifying causal relationships and transmission mechanisms between different risk types, and mapping the directed connections between risk types. For example, rock instability may lead to support structure failure, which in turn may further trigger local collapses. This network visually demonstrates the evolution path of risks. Subsequently, starting with the tunnel construction risk type with the highest probability determined in step S20, the network is traversed. The traversal process can employ standard graph search algorithms, such as depth-first search or breadth-first search, to determine the set of risk propagation paths by starting with the highest probability tunnel construction risk type and traversing all subsequent risk type nodes reachable from the initial risk type.

[0085] In step S40, damage performance characteristic data for each node in the risk propagation path is acquired. This data can be obtained by consulting historical accident reports, engineering specifications, etc. For example, for the risk node "local collapse," its damage performance characteristic data may include the estimated amount of property loss and the number of days of potential construction delay. Subsequently, based on the acquired damage performance characteristic data for each node in the risk propagation path and the probability of occurrence of the tunnel construction risk type, the comprehensive damage risk value for each node in the risk propagation path is determined. This process can involve simply multiplying the damage performance characteristic data (e.g., property loss) of each node with the probability of occurrence of the corresponding risk type, and then summing the results to obtain a preliminary comprehensive damage risk value. For example, in the risk propagation path 'deformation → cracking → collapse,' the loss values ​​(after normalization) of each node (deformation, cracking, and collapse) are first calculated separately, multiplied by the probability of collapse, and then the three product results are summed to obtain the comprehensive damage risk value for the path.

[0086] In step S50, tunnel construction risk is classified and controlled based on the comprehensive damage risk value of each node in the risk propagation path. This process may involve comparing the calculated comprehensive damage risk value of each node with several preset fixed thresholds, for example, classifying the risk value into three levels: "low risk," "medium risk," and "high risk." Once the risk level is determined, corresponding measures can be taken for abnormal situations during the current construction phase according to the pre-established risk control plan. For example, for a node assessed as "high risk," the emergency plan can be activated immediately, requiring the construction unit to suspend operations and carry out emergency reinforcement.

[0087] Through the above technical solution, the present invention realizes the integration of multi-source data fusion and weighting, dynamic calculation of risk probability and comprehensive damage assessment of propagation path, which significantly improves the real-time, accuracy and comprehensiveness of tunnel construction risk management.

[0088] For preferred options, please refer to [link / reference]. Figure 2 The present invention further proposes the method for determining the risk feature vector, specifically including:

[0089] S11: Standardize the multi-source monitoring data to determine the standardized values ​​of the monitoring data; the standardization method for multi-source monitoring data is Min-Max standardization.

[0090] S12: Determine the information entropy of the monitoring data based on the standardized values ​​of the monitoring data;

[0091] S13: Determine the risk feature vector based on the information entropy of the monitoring data;

[0092] Determining the risk feature vector involves processing and transforming multi-dimensional monitoring data collected during tunnel construction to form a numerical set that comprehensively reflects the current construction risk status. This vector forms the basis for subsequent risk assessment and tiered management, and its accuracy directly affects the effectiveness of the entire method.

[0093] Standardizing multi-source monitoring data aims to eliminate the impact of differences in dimensions and orders of magnitude between different monitoring data, bringing them to the same scale. This prevents certain larger data from dominating subsequent calculations and ensures the fairness of each data's contribution to risk assessment.

[0094] Min-Max standardization is a commonly used data standardization method that maps raw data to the range [0,1] through a linear transformation. This method preserves the relative relationships of the original data and scales all data to a uniform interval, making it suitable for situations where the data distribution range is known and there are no outliers. Besides Min-Max standardization, Z-score standardization, Decimal Scaling standardization, and other methods can also be used. Standardized monitoring data values ​​are multi-source monitoring data that have undergone standardization processing. These values ​​are typically scaled to a specific range (e.g., [0,1] or [-1,1]) or converted to a distribution with a mean of 0 and a variance of 1 to facilitate subsequent mathematical calculations and model input.

[0095] Determining the information entropy of monitoring data refers to quantifying the amount of information or uncertainty contained in each monitoring data point by calculating its information entropy. A lower information entropy generally indicates lower uncertainty, a greater amount of effective information contained in the data, and potentially a higher contribution to risk assessment. Information entropy can be calculated based on various mathematical models such as probability distributions and frequency statistics.

[0096] Specifically, firstly, the collected multi-source monitoring data is standardized, converting it into standardized values ​​with uniform dimensions. This step effectively eliminates differences in numerical range and units between different types of monitoring data (such as displacement, stress, and settlement), ensuring fairness in subsequent analysis. Next, based on these standardized values, the information entropy of each monitoring data point is calculated. Information entropy, as an indicator of data uncertainty or information content, reflects the importance of different monitoring data to risk assessment. The lower the information entropy, the lower the uncertainty of the data, the greater the amount of effective information it contains, and the higher its contribution to risk characteristics. Finally, by combining the information entropy of the monitoring data, a risk feature vector can be more reasonably determined. For example, weights can be assigned to each standardized data point based on information entropy, with data having lower information entropy receiving higher weights, thus enabling the risk feature vector to more accurately reflect the true risk status of the current tunnel construction. This approach makes the construction of the risk feature vector more scientific and objective, providing a more reliable basis for determining the probability of subsequent risk types, thereby improving the accuracy and effectiveness of the entire risk classification and control method.

[0097] Through the above technical solution, the present invention avoids the deviation that may be caused by directly using raw data, so that the determined risk feature vector can more accurately and comprehensively reflect the real risk status of tunnel construction, significantly improve the accuracy of subsequent determination of the probability of occurrence of risk types, and thus improve the reliability and effectiveness of the entire dynamic risk classification and control method for tunnel construction.

[0098] Preferably, the present invention further proposes the method of determining the information entropy of the monitoring data, specifically including:

[0099] Through the formula: ;

[0100] Determine the information entropy of the monitoring data ;

[0101] In the above formula, This refers to the information entropy of the i-th monitoring data. This refers to the standardized value of the i-th monitoring data at time t, and N refers to the total number of times.

[0102] The above formula is a mathematical model for calculating information entropy. Information entropy, as an indicator of the uncertainty or randomness of information, is used here to evaluate the value or importance of the monitoring data source. The formula quantifies the information content of a single monitoring data sequence by performing a logarithmic weighted summation on the values ​​of the standardized monitoring data at different times and then normalizing the summation. Its role is to provide a quantitative, information theory-based evaluation basis for the subsequent construction of risk feature vectors.

[0103] Information entropy of monitoring data refers to the amount of information or degree of uncertainty contained in the i-th monitoring data sequence. In tunnel construction risk assessment, different monitoring data (such as surface settlement, surrounding rock stress, and support structure deformation) have different fluctuations and information contents. By calculating information entropy, the importance of each monitoring data in reflecting changes in the tunnel construction status can be objectively assessed. The larger the information entropy value, the higher the uncertainty of the monitoring data, or in other words, the more effective information it contains, and the greater its potential contribution to risk assessment. For example, the information entropy of a monitoring data sequence (such as surface settlement data at a certain measuring point) collected over a period of time can be calculated using the above formula; alternatively, information entropy can be calculated separately for different types of monitoring data (such as temperature, humidity, and pressure) to assess their respective weights in risk assessment.

[0104] The standardized value of the i-th monitoring data at time t refers to the value of the i-th monitoring data at a specific time t after standardization. The purpose of standardization is to eliminate the differences between data of different dimensions and orders of magnitude, so that they are on a uniform scale, thereby avoiding the dominance of some large data in the calculation and ensuring the fairness and accuracy of information entropy calculation.

[0105] The total number of time points N refers to the total number of time points included in the monitoring data sequence used to calculate information entropy. This parameter determines the time window or data sample size for information entropy calculation. A larger N value generally means a more comprehensive data sample for calculation, resulting in a more stable and reliable information entropy calculation, and better reflecting the long-term information characteristics of the monitoring data sequence. N can be set to a fixed time window length, such as the number of monitoring data points within the past 24 hours, week, or construction cycle; or it can be dynamically adjusted, selecting the most recent N data points for calculation based on data acquisition frequency and real-time requirements.

[0106] Specifically, this invention first obtains the standardized value of the i-th monitoring data at time t. These standardized values ​​are data that have been preprocessed to eliminate the influence of dimensions. Then, using a preset total number of times, these standardized values ​​are substituted into a formula. This formula, based on information theory principles, objectively calculates the information entropy of the i-th monitoring data by weighted summation of each standardized value and its logarithm, followed by normalization. This calculation method ensures that the information content of each monitoring data sequence can be accurately and consistently quantified, providing a solid mathematical foundation for the subsequent construction of risk feature vectors. Through this precise information entropy calculation, the importance of different monitoring data in reflecting tunnel construction risks can be more accurately assessed, enabling the final generated risk feature vector to more realistically reflect the current risk status, thereby improving the accuracy and reliability of the entire dynamic risk classification and control method for tunnel construction.

[0107] Through the above technical solution, the present invention solves the problems of ambiguity and subjectivity that may exist when determining the information entropy of monitoring data, ensuring the objectivity and accuracy of the information entropy calculation results. In turn, it provides a reliable quantitative basis for the subsequent construction of risk feature vectors, and significantly improves the scientificity and accuracy of dynamic risk assessment of tunnel construction.

[0108] Preferably, the present invention further proposes that the determination of the risk feature vector also includes:

[0109] Through the formula: ;

[0110] Determine the components of the risk feature vector ;

[0111] In the above formula, This refers to the i-th component of the risk feature vector. This refers to the standardized value of the i-th monitoring data at the current moment. This refers to the fusion weight;

[0112] Based on the components of the risk feature vector Determine the risk feature vector;

[0113] The method for obtaining the fusion weights is as follows:

[0114] Through the formula: ;

[0115] Generate fusion weights ;

[0116] In the above formula, This refers to the information entropy of the j-th monitoring data. This refers to the information entropy of the i-th monitoring data, and M refers to the number of monitoring data.

[0117] In this context, each component of the risk feature vector represents the contribution of the i-th monitoring data point. Its function is to combine the standardized monitoring data with fusion weights reflecting its importance, thereby more accurately representing the actual impact of that monitoring data in the overall risk assessment. This component can be obtained by multiplying the standardized value of the monitoring data at the current moment with the corresponding fusion weight. Besides direct multiplication, a linear model can also be constructed, using the standardized value as input, the fusion weights as model parameters, and the output being the components of the risk feature vector.

[0118] Fusion weight is a coefficient used to measure the relative importance or contribution of the i-th monitoring data in risk assessment. Its function is to weight different monitoring data so that monitoring data with large information content and low uncertainty receive higher weights, while data with small information content and high uncertainty receive lower weights, thereby improving the accuracy and reliability of risk feature vectors.

[0119] The formula for calculating the fusion weight reflects the principle that the lower the information entropy (i.e., the lower the uncertainty of the data and the greater the amount of information), the greater the corresponding weight. This formula allows for the dynamic allocation of weights to each monitoring data point, ensuring that data with richer and more reliable information dominates the risk feature vector. The generation of these weights can be achieved through data analysis software or a custom program. First, the information entropy of all monitoring data is calculated; then, the fusion weight for each monitoring data point is calculated according to the formula.

[0120] Specifically, the proposed solution introduces fusion weights and combines them with standardized values ​​of the current monitoring data to determine the components of the risk feature vector. These fusion weights are calculated based on the information entropy of the monitoring data; the lower the information entropy, the lower the uncertainty of the monitoring data, the greater the amount of information it contains, and therefore the higher its weight. In this way, the proposed solution can dynamically assess the importance of different monitoring data sources and weight them to construct a risk feature vector that more accurately reflects the current risk status of tunnel construction. This weighting mechanism ensures that the risk feature vector is not simply composed of standardized values, but fully considers the information quality and reliability of each data source, thereby improving the representativeness and accuracy of the risk feature vector and providing a more solid foundation for determining the probability of subsequent risk types.

[0121] Through the above technical solution, this application not only improves the representativeness and reliability of the risk feature vector, but also provides a more accurate input for determining the probability of occurrence of subsequent risk types, thereby significantly improving the accuracy and effectiveness of dynamic risk classification and control in tunnel construction.

[0122] For preferred options, please refer to [link / reference]. Figure 3 The present invention further proposes the method for determining the probability of occurrence of tunnel construction risk types, specifically including:

[0123] S21: Based on the risk feature vector, determine the Mahalanobis distance between the risk feature vector and the feature center vector of each type of risk mode in tunnel construction;

[0124] S22: Determine the membership degree of the risk pattern based on the Mahalanobis distance between the risk feature vector and the feature center vector of each risk type;

[0125] S23: Determine the probability of occurrence of risk types in tunnel construction based on the membership degree of the risk model;

[0126] Among them, determining the Mahalanobis distance between the risk feature vector and the feature center vector of each type of risk mode in tunnel construction refers to calculating the Mahalanobis distance between the currently collected risk feature vector and the feature center vectors of various types of tunnel construction risk modes established in advance. As a distance metric that takes into account the data covariance, Mahalanobis distance can effectively handle the correlation between data of different dimensions and is not sensitive to scale, thus more accurately measuring the similarity or deviation between the current risk feature vector and the known risk mode.

[0127] Pre-established risk pattern feature center vectors can be obtained by clustering historical monitoring data (e.g., using K-means or spectral clustering algorithms), with each cluster center representing a typical risk pattern. Alternatively, adaptive learning algorithms, such as those based on Gaussian Mixture Models (GMM) or Support Vector Machines (SVM), can be used to dynamically update the risk pattern feature center vectors and calculate the Mahalanobis distance between the current risk feature vector and these dynamically updated center vectors, adapting to the dynamic changes in the tunnel construction environment.

[0128] Determining the membership degree of a risk pattern refers to transforming Mahalanobis distance into the degree to which the current risk feature vector belongs to a specific risk pattern. This membership degree transforms a hard distance metric into a soft degree of belonging, making risk assessment more flexible and robust, especially when the boundaries of risk patterns are ambiguous or multiple risks overlap. This membership degree can be determined by inputting the Mahalanobis distance into a decay function (e.g., Gaussian, exponential, or Sigmoid function); the smaller the distance, the higher the membership degree; the larger the distance, the lower the membership degree. Alternatively, a fuzzy logic inference system can be used to construct fuzzy rules based on Mahalanobis distance and other auxiliary information (e.g., historical cases), and output the membership degree of the current risk feature vector to various risk patterns through a fuzzy inference mechanism.

[0129] Determining the probability of occurrence of risk types in tunnel construction refers to quantifying the likelihood of each potential risk type occurring under the current tunnel construction environment. It forms the basis for subsequent risk classification and management, directly impacting the accuracy and timeliness of decision-making. This probability can be obtained by normalizing the membership degrees of all risk modes; for example, dividing the membership degree of each risk mode by the sum of the membership degrees of all risk modes ensures that the sum of the probabilities of occurrence for all risk types is 1. Alternatively, it can be calculated using Bayesian inference methods, combining prior knowledge and membership degrees. For instance, prior probabilities for various risk types can be established based on historical data, and then the posterior probabilities can be updated using Bayesian formulas, using membership degrees as evidence, to obtain a more accurate probability of occurrence for each risk type.

[0130] This application's solution introduces Mahalanobis distance to measure the similarity between risk feature vectors and preset risk patterns, effectively overcoming the limitations of traditional Euclidean distance in handling multidimensional, correlated data, thus making risk pattern identification more accurate. Subsequently, the Mahalanobis distance is transformed into risk pattern membership degrees, realizing a shift from hard classification to soft attribution, enhancing the model's ability to handle uncertainty and ambiguity. Finally, by normalizing the membership degrees, the probability of occurrence for each risk type is obtained, providing a quantitative and reliable basis for subsequent risk classification and control. This method, in determining the probability of occurrence of tunnel construction risk types, not only considers the inherent correlation of multi-source monitoring data but also improves the refinement of risk assessment through the concept of membership degrees, making the determination of tunnel construction risk type occurrence probabilities more accurate and robust, laying a solid foundation for subsequent risk propagation analysis and classification control.

[0131] Through the above technical solutions, the present invention enables risk assessment results to be more refined and reliable, providing a more solid data foundation for subsequent risk propagation analysis and dynamic hierarchical management, thereby improving the scientificity and effectiveness of tunnel construction risk management.

[0132] Preferably, the present invention further proposes the method of determining the membership degree of risk patterns, specifically including:

[0133] Through the formula: ;

[0134] Determine the membership degree of risk models ;

[0135] In the above formula, This refers to the membership degree of the k-th risk pattern. This refers to the Mahalanobis distance between the risk feature vector and the feature center vector of the k-th risk type. This refers to the preset scaling parameters;

[0136] The determination of the probability of occurrence of risk types in tunnel construction also includes:

[0137] Through the formula: ;

[0138] Determine the probability of occurrence of different risk types in tunnel construction. ;

[0139] In the above formula, This refers to the probability of the k-th type of risk occurring during tunnel construction. This refers to the membership degree of the r-th risk pattern, and m refers to the number of risk types.

[0140] Among them, the risk pattern membership degree refers to the degree of similarity or belonging between the current risk feature vector and a specific risk pattern. It quantifies the matching degree between the current tunnel construction status and a certain known risk pattern. This membership degree can be represented by a variety of mathematical functions. For example, in addition to the exponential decay function used in this application, it can also be a distance-based linear decay function or a polynomial decay function.

[0141] The formula for calculating the membership degree of a risk pattern is a Gaussian membership function. Its function is to calculate the membership degree based on the Mahalanobis distance between the risk feature vector and the feature center vector of the k-th risk type. This formula ensures that the closer the distance, the higher the membership degree, and the membership degree value is between 0 and 1.

[0142] Mahalanobis distance is a distance metric that takes into account the covariance of data. It can effectively handle the correlation between features in multidimensional data, thus more accurately reflecting the similarity between two vectors.

[0143] The preset scaling parameter controls the rate at which membership decays with distance, i.e., controls the "width" or "steepness" of the membership function curve. A smaller scaling parameter means that membership decays rapidly with increasing distance, indicating a more stringent matching requirement for risk patterns; a larger scaling parameter indicates a more lenient matching requirement. This parameter can be set by those skilled in the art.

[0144] The probability of occurrence of a risk type in tunnel construction refers to the actual likelihood of the k-th risk type occurring under the current tunnel construction conditions. It transforms the membership degree into a statistically significant probability value. The calculation formula normalizes the membership degree of all risk modes to obtain the probability of occurrence of each risk type. This normalization ensures that the sum of the probabilities of occurrence of all risk types is 1, making the probability value have clear statistical significance, which is convenient for comparison and decision-making.

[0145] Specifically, firstly, using a Gaussian membership function, based on the Mahalanobis distance between the risk feature vector and the feature center vectors of various risk modes, and a preset scaling parameter, the similarity between the current risk state and each risk mode is quantified. Mahalanobis distance effectively handles the correlation between multidimensional data, while the scaling parameter flexibly controls the rate of similarity decay. Subsequently, by normalizing the membership degrees of all risk modes, the membership degrees are transformed into probability of occurrence of risk types with clear physical meaning. This normalization ensures that the sum of the probabilities of occurrence of all risk types is 1, thus providing a clear and comparable probability distribution. In this way, when determining the probability of occurrence of tunnel construction risk types, this application not only considers the distance relationship between the risk feature vector and known risk modes, but also introduces adjustments to the sensitivity of membership degree decay through the scaling parameter, and finally, through probability normalization, makes the risk assessment results more scientific and interpretable, providing a solid data foundation for subsequent risk classification and management.

[0146] Through the above technical solution, this application overcomes the ambiguity and subjectivity that may exist in traditional methods when quantifying risks, making the risk assessment results more objective and reliable. This provides a more refined and scientific decision-making basis for dynamic risk classification and control in tunnel construction, and effectively improves the accuracy of risk identification and prediction.

[0147] For preferred options, please refer to [link / reference]. Figure 4 The present invention further proposes a comprehensive damage risk value for each node in the risk propagation path, specifically including:

[0148] S41: Determine the damage risk assessment value of each node in the risk propagation path based on the damage performance characteristics data of each node in the risk propagation path.

[0149] S42: Determine the probability of occurrence of each node in the risk propagation path based on the probability of occurrence of the type of tunnel construction risk;

[0150] S43: Determine the comprehensive damage risk value of each node in the risk propagation path based on the damage risk assessment value of each node in the risk propagation path and the occurrence probability of each node in the risk propagation path.

[0151] The determination of the comprehensive damage risk value of each node in the risk propagation path further includes:

[0152] Through the formula: ;

[0153] Determine the comprehensive damage risk value of each node in the risk propagation path. ;

[0154] In the above formula, This refers to the overall damage risk value of node e in the risk propagation path. This refers to the z-th node in the risk propagation path when node e is the endpoint. This refers to the number of nodes in the risk propagation path that terminate at node e. This refers to the probability of node e occurring in the risk propagation path;

[0155] This process involves determining the damage risk assessment value for each node in the risk propagation path based on its damage characteristics. The aim is to quantify the potential damage caused by each node in the risk propagation path during a risk event. This step transforms abstract damage information into calculable values, laying the foundation for subsequent comprehensive risk assessment. For example, an expert scoring method can be used, inviting multiple experts in tunnel engineering to evaluate and weight the damage characteristics (such as property loss, time loss, and the probability of casualties) of different risk nodes, and then summing the weighted values ​​to obtain the damage risk assessment value. Alternatively, a fuzzy comprehensive evaluation method can be employed, transforming the damage characteristics data into a fuzzy set and constructing a fuzzy relation matrix. Fuzzy operations can then be used to obtain the damage risk assessment value for each node, addressing the uncertainty and fuzziness of the data.

[0156] Based on the probability of occurrence of different tunnel construction risk types, the probability of occurrence at each node in the risk propagation path is determined, aiming to assess the actual likelihood of each specific node in the risk propagation path occurring. It refines the overall probability of occurrence of different tunnel construction risk types down to each node in the path, reflecting the dynamic evolution of risk during propagation. For example, a conditional probability model can be constructed based on historical accident data. Using the chain rule, the probability of occurrence of the highest-probability tunnel construction risk type is combined with the propagation probability between nodes to calculate the probability of occurrence at each node in the path. Alternatively, a Markov chain model can be used, treating the risk propagation path as a state transition process. A state transition matrix can be trained based on historical data to predict the probability of occurrence (i.e., occurrence or non-occurrence) of each node in the path given an initial probability of occurrence of a risk type.

[0157] The core of this solution is determining the comprehensive damage risk value of each node in the risk propagation path based on the damage risk assessment value and the probability of occurrence of each node. It combines the potential damage caused by a risk event (damage risk assessment value) with the likelihood of the event occurring (occurrence probability), thus obtaining a comprehensive index reflecting the severity of the risk. For example, the expected value calculation method can be used, directly multiplying the damage risk assessment value by the occurrence probability to obtain the comprehensive damage risk value. Alternatively, the risk matrix method can be used, dividing the damage risk assessment value and occurrence probability into different levels, and then finding the corresponding comprehensive risk level through a predefined risk matrix.

[0158] This invention first transforms abstract damage manifestation data into calculable damage risk assessment values. Then, based on the probability of occurrence of different tunnel construction risk types, it refines and determines the probability of occurrence for each node in the risk propagation path. Finally, a formula organically combines the damage risk assessment values ​​with the probability of occurrence to accurately calculate the comprehensive damage risk value for each node in the risk propagation path. This method not only considers the potential damage of individual nodes but also fully considers the cumulative effect and probability of occurrence of risks along the propagation path, thus enabling a more comprehensive and accurate assessment of the true risk level of each node in the risk propagation path. This refined risk quantification method provides more solid and reliable data support for subsequent risk classification and control in tunnel construction.

[0159] By combining the damage risk assessment values ​​and occurrence probabilities of each node in the risk propagation path, this application overcomes the shortcomings of traditional methods in terms of insufficient risk quantification. This allows the assessment of tunnel construction risks to move beyond qualitative or semi-qualitative levels and obtain a comprehensive damage risk value with practical significance. This precise risk quantification provides a more reliable basis for subsequent risk classification and control, enabling risk management measures to be more targeted and effective. This significantly improves the safety and controllability of tunnel construction and effectively avoids or mitigates potential losses from risk events.

[0160] Preferably, the present invention further proposes the damage risk assessment value for each node in the risk propagation path, specifically including:

[0161] Through the formula: ;

[0162] Determine the damage risk assessment value of each node in the risk propagation path. ;

[0163] In the above formula, This refers to the normalized value of the damage manifestation feature data of node e in the risk propagation path. This refers to the weighting coefficient of the l-th damage manifestation feature, where L refers to the number of damage manifestation feature data.

[0164] Among them, the damage risk assessment value of a node in the risk propagation path represents the potential loss or impact that a specific node in the risk propagation path may cause when a risk event occurs. It is a quantitative indicator used to measure the severity of the inherent damage of a single risk node without considering its probability of occurrence.

[0165] In the risk propagation path, the normalized value of the l-th damage manifestation feature data at node e refers to the result after standardizing the original damage manifestation feature data. This eliminates the impact of different dimensions and numerical ranges on the accuracy of the assessment and ensures the comparability of different feature data in the assessment. Normalization methods can include Min-Max normalization, Z-score standardization, etc.

[0166] The weight coefficient of the l-th damage manifestation feature reflects the relative importance of different damage manifestation features in risk assessment. By assigning a weight coefficient to each damage manifestation feature, the influence of expert experience, industry standards, or historical data analysis results on risk assessment can be reflected, making the assessment results more in line with actual needs and decision-making preferences. The weight coefficient can be determined based on expert scoring, analytic hierarchy process (AHP), entropy weighting, or principal component analysis, etc.

[0167] The number of damage manifestation feature data represents the total number of damage manifestation feature data used to assess the node damage risk in the risk propagation path. These feature data together constitute a comprehensive description of the node damage situation.

[0168] Specifically, when determining the damage risk assessment value based on the damage manifestation characteristic data of each node in the risk propagation path, the normalized value of the l-th damage manifestation characteristic data of node e in the risk propagation path is first obtained. This normalized value ensures the comparability of damage manifestation characteristic data with different dimensions in the assessment. Subsequently, a weight coefficient for the l-th damage manifestation characteristic is assigned to each damage manifestation characteristic data. This weight coefficient reflects the relative importance of different damage characteristics in the risk assessment, enabling the assessment results to fully consider the differences in the contribution of different damage types to the overall risk. Finally, the damage risk assessment value of each node in the risk propagation path is obtained by multiplying each normalized damage manifestation characteristic data with its corresponding weight coefficient and summing all the products. This weighted summation calculation method ensures that the damage risk assessment value not only reflects the degree of damage but also comprehensively considers the importance of different damage types, thereby enabling a more accurate and objective quantification of the inherent damage risk of each node in the risk propagation path. In the aforementioned method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion, step S41 determines the damage risk assessment value of each node in the risk propagation path based on the damage performance characteristics data of each node. This forms the basis for the subsequent step S43 to determine the comprehensive damage risk value of each node in the risk propagation path. The solution in this application normalizes the damage performance characteristics data and introduces weighting coefficients, making the calculation of the damage risk assessment value more precise and reasonable. This not only solves the assessment bias problem that may result from directly using raw damage data, but also, through the introduction of weighting coefficients, allows the assessment results to better reflect the degree of attention paid to different damage types in actual risk management. This provides a more reliable input for the subsequent calculation of the comprehensive damage risk value, thereby improving the accuracy and practicality of the entire risk classification and control method.

[0169] Through the above technical solution, the present invention enables the damage risk assessment value to be more accurate, comprehensive and objective, avoiding assessment deviations caused by different data dimensions or differences in importance, thus providing a more solid foundation for the subsequent calculation of comprehensive damage risk value, and thereby improving the scientific nature and decision support effectiveness of the entire tunnel construction dynamic risk classification and control method.

[0170] Preferably, the present invention further proposes the determination of the occurrence probability of each node in the risk propagation path, specifically including:

[0171] Through the formula: ;

[0172] Determine the probability of node e occurring in the risk propagation path. ;

[0173] In the above formula, This refers to the probability of occurrence of the tunnel construction risk type corresponding to the highest probability tunnel construction risk type. This refers to the node where node e is the endpoint in the risk propagation path. With nodes The probability of propagation between them This refers to the number of nodes in the risk propagation path when node e is the endpoint;

[0174] Among them, the propagation probability The specific method of obtaining it is as follows:

[0175] Retrieve nodes from historical data The frequency and nodes of the corresponding risk types When the corresponding risk type occurs, it causes a node The number of times it occurs;

[0176] Nodes in historical data The frequency and nodes of corresponding risk types When the corresponding risk type occurs, it causes a node The propagation probability is obtained by comparing the number of occurrences. ;

[0177] Specifically, this application first uses the frequency of risk types occurring and the number of times they mutually trigger each other in historical data to calculate the conditional probability of risk propagation between adjacent nodes. This calculation method based on actual occurrence frequency ensures high objectivity and reliability of the propagation probability. Then, using the occurrence probability corresponding to the highest probability tunnel construction risk type as the starting point of the risk propagation chain, this initial probability is multiplied by the propagation probabilities between all adjacent nodes along the risk propagation path to obtain the occurrence probability of a specific node e in the risk propagation path. This multiplication mechanism reflects the characteristic of risk events gradually attenuating or accumulating in the propagation chain, allowing the final calculated node occurrence probability to comprehensively consider the intensity of the initial risk and the path dependence of risk propagation. In this way, the scheme of this application provides key input parameters for determining the comprehensive damage risk value of each node in the risk propagation path, making the assessment of the comprehensive damage risk value no longer dependent on subjective judgment or rough estimation, but based on objective historical data and a rigorous probability model. This significantly improves the accuracy and reliability of risk assessment, providing a more solid data foundation for subsequent risk classification and control.

[0178] Through the above technical solution, this application can overcome the limitations of subjective judgment or experience estimation in traditional risk assessment, making the calculation of the probability of occurrence of each node in the risk propagation path more objective, accurate and data-supported, improving the assessment accuracy of the comprehensive damage risk value of each node in the risk propagation path, thereby making the risk classification and control of tunnel construction more scientific and reasonable, more effectively identifying high-risk nodes, and formulating targeted control measures, significantly enhancing the refinement and effectiveness of dynamic risk management in tunnel construction.

[0179] For preferred options, please refer to [link / reference]. Figure 5 The present invention further proposes a method for risk classification and control in tunnel construction based on the comprehensive damage risk value of each node in the risk propagation path, specifically including:

[0180] S51: Determine the extreme value nodes based on the comprehensive damage risk value of each node in the risk propagation path; specifically, this includes: filtering the maximum value of the comprehensive damage risk value of each node in all risk propagation paths, and marking the node corresponding to the maximum value of the comprehensive damage risk value as the extreme value node.

[0181] S52: Based on the comprehensive damage risk value corresponding to the extreme value node, conduct risk classification and control for tunnel construction.

[0182] The purpose of identifying extreme nodes is to pinpoint the critical nodes with the highest overall damage risk value from among numerous nodes in the risk propagation path. This helps to concentrate limited risk management resources on the most critical risk points. This can be achieved by iterating through the overall damage risk values ​​of all nodes and finding the maximum value; or by using a sorting algorithm to sort the overall damage risk values ​​of all nodes in descending order and selecting the node at the top of the list. Maximum value selection is a specific method for identifying extreme nodes, which involves comparing the overall damage risk values ​​of all nodes in all risk propagation paths and finding the one with the largest value.

[0183] Marking the node corresponding to the maximum value of the comprehensive damage risk as an extreme value node is a clear indication of the screening results. Risk points marked as extreme value nodes mean that they have the highest potential comprehensive damage risk and need to be prioritized for handling and control.

[0184] Based on the comprehensive damage risk value corresponding to the extreme node, the risk classification and control of tunnel construction is based on the most critical risk point identified and corresponding risk management measures are taken. Its role is to transform the abstract risk value into an operable control instruction to ensure that the risk can be effectively controlled.

[0185] This can be achieved by comparing the comprehensive damage risk value of extreme nodes with multiple preset risk level thresholds, thereby classifying them into different risk levels, such as "low risk," "medium risk," "high risk," or "emergency risk." Comparing the comprehensive damage risk value with thresholds is the fundamental operation of risk classification. By setting different thresholds, continuous comprehensive damage risk values ​​can be discretized into different risk levels.

[0186] Beyond simple threshold comparisons, fuzzy logic, expert systems, or machine learning models can be used to classify risks more finely. Risk classification is the process of categorizing risks into different levels. Risk classification helps managers clearly understand the severity of current risks and provides a basis for decision-making regarding subsequent control measures.

[0187] In addition to risk classification based on numerical thresholds, a comprehensive classification can also be achieved by combining multiple dimensions such as the likelihood of risk occurrence, the scope of impact, and the difficulty of recovery. Classifying and managing anomalies during construction according to risk classification is the specific action taken after risk classification. Different risk levels correspond to different management strategies and resource investments. For example, low risk may only require routine monitoring; medium risk may require enhanced inspections and early warnings; and high risk may require immediate emergency measures such as work stoppages and reinforcement.

[0188] Specifically, the scheme first filters the maximum value of the comprehensive damage risk at each node in all risk propagation paths, thereby accurately identifying the most potentially hazardous risk points in the current tunnel construction, i.e., extreme value nodes. This process allows risk management to shift from a general focus on all risks to a focus on the most critical and urgent risk sources. Subsequently, the scheme compares the comprehensive damage risk value corresponding to the identified extreme value nodes with preset risk level thresholds, thereby achieving a refined classification of tunnel construction risks. This classification mechanism enables risk managers to formulate and implement differentiated control measures based on the severity of the risks. For example, for high-risk extreme value nodes, emergency plans can be activated immediately, and emergency reinforcement or work stoppage measures can be taken; while for lower-risk extreme value nodes, preventative measures such as enhanced monitoring and optimized construction plans can be implemented.

[0189] Through the above technical solution, the solution of this application shifts the focus of risk management from "casting a wide net" to "precision strikes", which significantly improves the efficiency of risk identification and the pertinence of control measures, thereby effectively avoiding waste of resources and ensuring the timeliness and effectiveness of risk management.

[0190] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion, characterized in that, The method specifically includes: Collect multi-source monitoring data during tunnel construction, and determine risk feature vectors based on the multi-source monitoring data during tunnel construction; Based on the risk feature vector, determine the probability of occurrence of each tunnel construction risk type, and based on the probability of occurrence of each tunnel construction risk type, determine the tunnel construction risk type with the highest probability. Construct a directed network for the propagation of tunnel construction risk types, and traverse the directed network for the propagation of tunnel construction risk types starting from the tunnel construction risk type with the highest probability to determine the set of risk propagation paths; The damage manifestation characteristic data of each node in the risk propagation path is obtained. Based on the damage manifestation characteristic data of each node in the risk propagation path and the probability of occurrence of the tunnel construction risk type, the comprehensive damage risk value of each node in the risk propagation path is determined. Among them, the damage manifestation characteristic data refers to the damage situation that can reflect when the risk type occurs. Based on the comprehensive damage risk value of each node in the risk propagation path, tunnel construction risk is classified and controlled.

2. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion as described in claim 1, characterized in that, The determination of the risk feature vector specifically includes: The multi-source monitoring data is standardized to determine the standardized values ​​of the monitoring data; the standardization method for the multi-source monitoring data is Min-Max standardization. The information entropy of the monitoring data is determined based on the standardized values ​​of the monitoring data; Based on the information entropy of the monitoring data, the risk feature vector is determined.

3. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 2, characterized in that, The determination of the information entropy of the monitoring data specifically includes: Through the formula: ; Determine the information entropy of the monitoring data ; In the above formula, This refers to the information entropy of the i-th monitoring data. This refers to the standardized value of the i-th monitoring data at time t, and N refers to the total number of times.

4. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 2, characterized in that, The determination of the risk feature vector also includes: Through the formula: ; Determine the components of the risk feature vector ; In the above formula, This refers to the i-th component of the risk feature vector. This refers to the standardized value of the i-th monitoring data at the current moment. This refers to the fusion weight; Based on the components of the risk feature vector Determine the risk feature vector; The method for obtaining the fusion weights is as follows: Through the formula: ; Generate fusion weights ; In the above formula, This refers to the information entropy of the j-th monitoring data. This refers to the information entropy of the i-th monitoring data, and M refers to the number of monitoring data.

5. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 1, characterized in that, The determination of the probability of occurrence of various tunnel construction risk types specifically includes: Based on the risk feature vector, determine the Mahalanobis distance between the risk feature vector and the feature center vector of each type of risk mode in tunnel construction; The membership degree of a risk pattern is determined based on the Mahalanobis distance between the risk feature vector and the feature center vector of each risk type. Based on the membership degree of the risk model, determine the probability of occurrence of the risk type in tunnel construction.

6. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 5, characterized in that, The determination of the membership degree of the risk model specifically includes: Through the formula: ; Determine the membership degree of the risk model ; In the above formula, This refers to the membership degree of the k-th risk pattern. This refers to the Mahalanobis distance between the risk feature vector and the feature center vector of the k-th risk type. This refers to the preset scaling parameters; The determination of the probability of occurrence of risk types in tunnel construction also includes: Through the formula: ; Determine the probability of occurrence of different risk types in tunnel construction. ; In the above formula, This refers to the probability of the k-th type of risk occurring during tunnel construction. This refers to the membership degree of the r-th risk pattern, and m refers to the number of risk types.

7. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion as described in claim 1, characterized in that, The determination of the comprehensive damage risk value of each node in the risk propagation path specifically includes: Based on the damage performance characteristics data of each node in the risk propagation path, determine the damage risk assessment value of each node in the risk propagation path. Based on the probability of occurrence of different types of tunnel construction risks, determine the probability of occurrence of each node in the risk propagation path; Based on the damage risk assessment value of each node in the risk propagation path and the occurrence probability of each node in the risk propagation path, the comprehensive damage risk value of each node in the risk propagation path is determined. The determination of the comprehensive damage risk value of each node in the risk propagation path further includes: Through the formula: ; Determine the comprehensive damage risk value of nodes in the risk propagation path. ; In the above formula, This refers to the overall damage risk value of node e in the risk propagation path. This refers to the z-th node in the risk propagation path when node e is the endpoint. This refers to the number of nodes in the risk propagation path that terminate at node e. This refers to the probability of node e occurring in the risk propagation path.

8. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 7, characterized in that, The determination of the damage risk assessment value for each node in the risk propagation path specifically includes: Through the formula: ; Determine the damage risk assessment value of each node in the risk propagation path. ; In the above formula, This refers to the normalized value of the damage manifestation feature data of node e in the risk propagation path. This refers to the weighting coefficient of the l-th damage manifestation feature, where L refers to the number of damage manifestation feature data.

9. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 7, characterized in that, Determining the probability of occurrence of each node in the risk propagation path specifically includes: Through the formula: ; Determine the probability of node e occurring in the risk propagation path. ; In the above formula, This refers to the probability of occurrence of the tunnel construction risk type corresponding to the highest probability tunnel construction risk type. This refers to the node where node e is the endpoint in the risk propagation path. With nodes The probability of propagation between them This refers to the number of nodes in the risk propagation path when node e is the endpoint; Among them, the propagation probability The specific method of obtaining it is as follows: Retrieve nodes from historical data The frequency and nodes of the corresponding risk types When the corresponding risk type occurs, it causes a node The number of times it occurs; Nodes in historical data The frequency and nodes of corresponding risk types When the corresponding risk type occurs, it causes a node The propagation probability is obtained by comparing the number of occurrences. .

10. The method for dynamic risk classification and control of tunnel construction based on multi-source monitoring data fusion according to claim 1, characterized in that, The method of classifying and controlling tunnel construction risks based on the comprehensive damage risk value of each node in the risk propagation path specifically includes: Based on the comprehensive damage risk value of each node in the risk propagation path, extreme value nodes are determined; specifically, this includes: screening the maximum value of the comprehensive damage risk value of each node in all risk propagation paths, and marking the node corresponding to the maximum value of the comprehensive damage risk value as an extreme value node; Based on the comprehensive damage risk value corresponding to the extreme node, tunnel construction risk is classified and controlled.

Images