A method and system for dynamic prediction of safety risk of subway tunnel structure

Abstract

The present invention discloses a method and system for dynamically predicting the structural safety risk of a subway tunnel, and relates to the technical field of tunnel risk prediction. The method for dynamically predicting the structural safety risk of a subway tunnel uses optical fiber sensing technology to acquire and preprocess monitoring data in real time, and can detect tiny structural changes. Structural risk sensitive factors are extracted based on monitoring data, feature labels are set, and feature labels with high sensitivity are identified through correlation coefficients and preset thresholds. These sensitive factors are used to calculate the comprehensive structural risk impact, set safety risk grading indicators, and obtain structural risk responsibility parameters in combination with tunnel data to ultimately determine the structural risk level. A data set training prediction model is constructed, and structural risk trends are obtained through new monitoring data, and a comprehensive prediction of multiple risk sensitive factors is performed. This improves the efficiency of formulating and implementing risk response strategies. According to risk trends, dual early warning prompts are provided to staff to ensure passenger safety and safe operation of tunnels.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel risk prediction, and in particular to a method and system for dynamically predicting the structural safety risk of a subway tunnel. Background Art

[0002] Subway is an important public transportation tool in modern urban construction. It not only responds to the travel concept of low-carbon and environmental protection, but also provides citizens with convenient travel options. However, in subway construction, there are multiple subway lines intersecting to form transfer points. Although it is convenient for passengers to transfer, it also brings burdens to the underground structure. In particular, the subway brings super-strong loads to the tunnel structure during the period of passing. There may be a serious risk of landslides, which will not only threaten the life safety of passengers, but also increase the cost of urban construction and affect the progress of subsequent construction. Therefore, it is necessary to monitor the safety risks of subway tunnels in real time. However, the monitoring objects of traditional risk prediction systems are relatively simple, such as monitoring lighting, ventilation, and fire. It fails to fully monitor the overall structural risk factors of the transfer point tunnel to obtain the risk level category of historical monitoring data to predict the tunnel structure risk level in the future. Therefore, it is difficult to meet the requirements of accurate and comprehensive risk prediction.

[0003] In summary, how to accurately and comprehensively predict the safety risks of subway tunnel structures is an urgent problem to be solved and optimized in the dynamic prediction system of subway tunnel structure safety risks. Summary of the invention

[0004] The present invention provides a method and system for dynamically predicting the safety risks of subway tunnel structures, which solve the technical problem of how to more accurately and comprehensively predict the safety risks of subway tunnel structures in related technologies.

[0005] In order to solve the above technical problems, the present invention provides a method and system for dynamically predicting the safety risk of subway tunnel structures. The specific technical solutions are as follows:

[0006] In a first aspect, a method for dynamically predicting the structural safety risk of a subway tunnel comprises the following steps:

[0007] Obtain monitoring data; based on the monitoring data, obtain structural risk sensitivity factors;

[0008] Among them, corresponding feature labels are set based on the type of monitoring data;

[0009] Obtain the correlation coefficient between monitoring data and feature labels; obtain the preset threshold;

[0010] Based on the correlation coefficient and the preset threshold, a highly sensitive feature label is obtained;

[0011] Feature labels with high sensitivity are structural risk sensitive factors;

[0012] Based on the structural risk sensitivity factors, the comprehensive structural risk impact is obtained, and the safety risk classification index is set; based on the structural risk sensitivity factors and tunnel data, the structural risk responsibility parameters are obtained; based on the safety risk classification index and the structural risk responsibility parameters, the structural risk level is obtained;

[0013] Building a data set based on the monitoring data and the structural risk level, training and obtaining a prediction model;

[0014] Obtain new monitoring data, and obtain structural risk trends based on the new monitoring data and the prediction model;

[0015] The change of the structural risk level is obtained according to the structural risk trend, and the early warning information is output in a preset form.

[0016] As a further optimization scheme of the present invention, the monitoring data and the new monitoring data are both obtained in real time through optical fiber sensing signals; wherein the method for obtaining the monitoring data includes:

[0017] Substitute the interfering fiber signal into the wavelet packet inner product space basis function system Perform linear superposition, expressed as: ; Where f(t) is the useful optical fiber characteristic signal, c λ is the interference signal processing coefficient, A is the threshold of the spatial basis function system to remove redundant interference fiber signals;

[0018] Decompose the useful signal features into low-frequency and high-frequency parts, and obtain the decomposition coefficients through multi-layer wavelet packet decomposition, and set thresholds for each group of decomposition coefficients, including: mean value Ei, variance σi and comprehensive value Ti;

[0019] According to the set threshold values ​​of each group of decomposition coefficients, multiple wavelet packet decomposition layers are screened one by one to obtain time domain features and frequency domain features;

[0020] Obtaining an optical fiber enhancement characteristic signal according to the acquired time domain characteristics and frequency domain characteristics;

[0021] The acquired time domain and frequency characteristic information is concentrated in the coefficients with large modulus values. The corresponding regular statistical differences are obtained through the time domain characteristics and frequency domain characteristics. The statistical difference characteristic quantity is used to distinguish the vibration segment and the non-vibration segment in the fiber-enhanced characteristic signal. The vibration behavior type is identified according to the signals of different vibration bands, thereby obtaining the tunnel environment data X. env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist .

[0022] As a further optimization scheme of the present invention, the method for obtaining the structural risk sensitivity factor includes:

[0023] Based on the pre-processed tunnel environment data X env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist ,pass: , obtain the correlation coefficient between each monitoring data and the feature label corresponding to each monitoring data;

[0024] Where, X j Represents the monitoring data after preprocessing, Y i Represents the characteristic labels of each monitoring data, ρ(X j , Y i ) represents data X j With label Y i The correlation coefficient, Cov(X j , Y i ) represents data X j With label Y i The covariance of j Represents feature X j The standard deviation of i Indicates label Y i The standard deviation of

[0025] Based on the correlation coefficient between each monitoring data and the feature label corresponding to each monitoring data, through: , obtain the sensitivity of feature labels in each monitoring data;

[0026] In the formula, I Xj represents the sensitivity coefficient of each monitoring data after preprocessing, T represents the total number of monitoring data, Imp Xj，t represents the feature sensitivity in the t-th monitoring data;

[0027] The sensitivity of the feature label is the structural risk feature sensitivity coefficient;

[0028] The preset threshold is a set correlation coefficient, a feature label whose absolute value of the correlation coefficient is greater than the preset threshold is obtained, a structural risk feature sensitivity coefficient corresponding to the feature label is obtained, and a structural risk sensitivity factor is obtained based on the size of the structural risk feature sensitivity coefficient;

[0029] The obtained structural risk sensitive factors are interactively combined to obtain the comprehensive structural risk impact.

[0030] As a further optimization scheme of the present invention, a method for obtaining the comprehensive structural risk impact degree includes:

[0031] The structural risk sensitivity factor is encoded as {A 1 i , A 2 i , A 3 i , …, A z i}, where i={1, 2, …, m} represents the number of monitoring times; A z Represents the code of the zth structural risk sensitive factor;

[0032] pass: Where I AB，quadratic represents the nonlinear interaction between sensitive factors, 2≤n≤6, and n is an integer, U n Expressed as the nth structural risk sensitivity factor, U 1 , U 2 , U 3 , U 4 , U 5 , U 6 Respectively with {g i , N i , K.O. i , V i , E i , H i}, thus obtaining a variety of interactive combination results; among them, g i is the coding vector of tunnel soil stress change, N i is the encoding vector of the crack cardinality, KO i is the encoding vector of the tunnel load compression coefficient, V i is the encoding vector of the Poisson's ratio of the tunnel soil layer, E i is the encoding vector of the compression modulus and H i is the encoding vector of water permeability;

[0033] According to the multiple interaction results, by: , where R is the comprehensive structural risk coefficient, β 0 is the influence intercept index, β 1 The coefficient of the interactive combination result of sensitive factors, the comprehensive structural risk coefficient is the comprehensive structural risk impact.

[0034] As a further optimization scheme of the present invention, the method for obtaining the structural risk level includes:

[0035] Dividing the comprehensive structural risk impact R into risk intervals specifically includes: dividing R into multiple risk intervals according to the number of interaction items of different risk sensitive factors, namely interval one: N≤k1; interval two: k1<N≤k2; interval three: k2<N≤k3, wherein N is the number of interaction items of sensitive factors, k1, k2 and k3 are respectively preset thresholds of interaction items of sensitive factors, and k1<k2<k3;

[0036] Based on multiple risk categories, through: ; In the formula, F is the safety risk classification index, R low is the low impact interval, R med. is the medium influence interval, R high It is a high impact interval, from which the security risk classification index is obtained;

[0037] The structural risk level is obtained by combining the obtained safety risk classification index with the structural risk responsibility parameter a.

[0038] As a further optimization solution of the present invention, the method for obtaining the structural risk level further includes:

[0039] Select n structural risk sensitive factors and randomly combine them into N interaction terms to form sensitive factor parameter data;

[0040] The sensitive factor parameter data is combined with the maximum values ​​of the tunnel vertical and lateral additional loads, displacements, and relative deformation curvature to obtain the structural risk responsibility parameter a;

[0041] According to the structural risk responsibility parameter weight a and the safety risk grading index, the structural risk weight G is obtained by: G=a·F; and the structural risk degree is divided according to the structural risk weight G to obtain the structural risk level, which are low risk level, medium risk level and high risk level.

[0042] As a further optimization scheme of the present invention, a method for predicting structural risk trends by constructing a prediction model according to the structural risk level includes:

[0043] The monitoring data and the structural risk level are combined to construct a data set, which is the training matrix K. When the training matrix K contains g different time states V t , so the training matrix is ;

[0044] V t In t j The n risk transformation values ​​observed at a time point are regarded as the observation vector ; Input the observation vector into the prediction model, including: ; To obtain the risk level category of monitoring data;

[0045] In the formula, H is the low-dimensional representation after encoding, f is the activation function, and W enc Encoding weight matrix, b enc Encoder bias vector, V t (t j ) input data, X is the reconstructed data, e is the activation function, W dec is the decoding weight matrix, b dec is the decoding bias vector, L is the loss function value, and D is the number of data points;

[0046] The new monitoring data is then input into the prediction model, and the new monitoring data risk level category is output and compared with the risk level category of the monitoring data to obtain the structural risk trend.

[0047] As a further optimization scheme of the present invention, a prediction model is constructed according to the structural risk level to predict the structural risk trend, which also includes:

[0048] Optimize the prediction model parameters by gradient descent: , where η is the learning rate, L is the gradient length, is the symbol for partial differentials.

[0049] As a further optimization solution of the present invention, the method of outputting warning information in a preset form includes:

[0050] According to the risk level category of the monitoring data output by the prediction model, the risk warning threshold is set to generate warning information;

[0051] Based on the risk level category of the monitoring data, when the preset risk warning threshold is reached, the first warning alarm is issued;

[0052] Based on the structural risk trend, the structural risk level changes of the monitoring data and the new monitoring data are obtained, triggering early warning information and issuing a secondary early warning alarm.

[0053] In a second aspect, the system provides an electronic device including a memory, a processor, and a subway tunnel structure safety risk dynamic prediction method program stored in the memory and executable on the processor, wherein the subway tunnel structure safety risk dynamic prediction method program is executed by the processor to implement the steps of the subway tunnel structure safety risk dynamic prediction method, and the system includes:

[0054] Data acquisition module 201: used to obtain monitoring data; obtain structural risk sensitivity factors based on the monitoring data; wherein corresponding feature labels are set based on the type of monitoring data; obtain the correlation coefficient between the monitoring data and the feature labels; obtain a preset threshold; obtain a feature label with high sensitivity based on the correlation coefficient and the preset threshold; the feature label with high sensitivity is the structural risk sensitivity factor;

[0055] Data analysis module 202: It is used to obtain the comprehensive structural risk impact based on the structural risk sensitivity factor and set the safety risk classification index; obtain the structural risk responsibility parameter based on the structural risk sensitivity factor and tunnel data; obtain the structural risk level based on the safety risk classification index and the structural risk responsibility parameter;

[0056] Risk prediction module 203: It is used to construct a data set based on the monitoring data and the structural risk level, train and obtain a prediction model; obtain new monitoring data, and obtain a structural risk trend based on the new monitoring data and the prediction model;

[0057] The early warning module 204 is used to obtain the change of the structural risk level according to the structural risk trend and output the early warning information in a preset form.

[0058] The present invention has at least the following beneficial effects: the present invention obtains monitoring data in real time through optical fiber sensing and pre-processes the monitoring data; optical fiber sensing technology can be used to obtain high-precision monitoring results to detect tiny structural changes, and non-destructive noise reduction processing is performed on optical fiber interference signals, thereby obtaining more accurate optical fiber signal characteristics;

[0059] Based on the monitoring data, the structural risk sensitive factor is obtained; wherein, the corresponding feature label is set based on the monitoring data type; the correlation coefficient between the monitoring data and the feature label is obtained; the preset threshold is obtained; based on the correlation coefficient and the preset threshold, a feature label with high sensitivity is obtained; the feature label with high sensitivity is the structural risk sensitive factor;

[0060] Based on the structural risk sensitivity factor, the comprehensive structural risk impact is obtained, and the safety risk classification index is set; based on the structural risk sensitivity factor and tunnel data, the structural risk responsibility parameter is obtained; based on the safety risk classification index and the structural risk responsibility parameter, the structural risk level is obtained; and the characteristic label with high sensitivity is obtained as the structural risk sensitivity factor, which can improve the accuracy of the early warning system. This means that the early warning system can send out signals earlier before the risk occurs, providing decision makers with more time to formulate response strategies, thereby improving the efficiency of tunnel structure risk management and providing reliable technical support for ensuring the safe operation of the tunnel;

[0061] Based on the monitoring data and the structural risk level, a data set is constructed to train and obtain a prediction model; new monitoring data is obtained, and based on the new monitoring data and prediction model, the structural risk trend is obtained; thus, the structural risk level is comprehensively predicted for the various risk-sensitive factors obtained according to the prediction model, thereby improving the accuracy and comprehensiveness of the predicted risk; through the detailed analysis and classification of structural risks, the organization can more effectively formulate and implement risk response strategies and improve the ability to respond to sudden risks;

[0062] Then, the change in structural risk level is obtained based on the structural risk trend, and warning information is output in a preset form. Further, two warning reminders are given to the staff based on the results of the predicted risks to avoid the risk from expanding and delaying and threatening the lives of passengers. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic flow chart of a method for dynamically predicting the safety risk of a subway tunnel structure provided by an embodiment of the present invention;

[0064] Figure 2 It is a schematic diagram of a subway tunnel structure safety risk dynamic prediction system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The present application is further described in detail below in conjunction with the accompanying drawings. It is necessary to point out here that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technical personnel in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0066] This embodiment provides a method and system for dynamically predicting the structural safety risk of a subway tunnel, and the specific implementation methods are as follows:

[0067] like Figure 1 As shown, a method for dynamic prediction of subway tunnel structure safety risk includes:

[0068] Obtain monitoring data; based on the monitoring data, obtain structural risk sensitivity factors;

[0069] Among them, corresponding feature labels are set based on the type of monitoring data;

[0070] Obtain the correlation coefficient between monitoring data and feature labels; obtain the preset threshold;

[0071] Based on the correlation coefficient and the preset threshold, a highly sensitive feature label is obtained;

[0072] Feature labels with high sensitivity are structural risk sensitive factors;

[0073] Based on the structural risk sensitivity factors, the comprehensive structural risk impact is obtained, and the safety risk classification index is set; based on the structural risk sensitivity factors and tunnel data, the structural risk responsibility parameters are obtained; based on the safety risk classification index and the structural risk responsibility parameters, the structural risk level is obtained;

[0074] Building a data set based on the monitoring data and the structural risk level, training and obtaining a prediction model;

[0075] Obtain new monitoring data, and obtain structural risk trends based on the new monitoring data and the prediction model;

[0076] Acquire changes in structural risk levels according to the structural risk trend, and output warning information in a preset form;

[0077] It is necessary to explain in detail in this embodiment that the method for obtaining monitoring data in this embodiment is to obtain monitoring data in real time through optical fiber sensing signals; in other embodiments, tunnel structure data can also be monitored and obtained by directly entering the tunnel structure monitoring data set or other monitoring data equipment; the method for real-time acquisition of optical fiber sensing signals adopted in this embodiment has optical fiber sensing technology that provides high-precision monitoring results and can detect tiny structural changes, which is crucial for early identification of potential problems. At the same time, optical fiber sensors have excellent resistance to chemical corrosion and physical wear, and are very suitable for long-term monitoring in harsh environments and are not affected by electromagnetic interference. Due to the low maintenance requirements and long-term durability of optical fiber sensors, optical fiber sensing technology usually shows high economic benefits in life cycle cost analysis. The optical fiber sensing network can be customized according to the specific needs of the tunnel, is easy to expand, and can adapt to tunnel structures of different lengths and shapes;

[0078] In this embodiment, to obtain monitoring data, it is first necessary to obtain monitoring data of subway tunnels from various monitoring systems. These data may include various forms, such as temperature, humidity, strain, acceleration, water seepage, and geological information. These data are collected in real time by various sensors installed in the tunnel and transmitted to the central database, thereby providing comprehensive environmental, structural and operational data, which is helpful for comprehensive analysis of the health status of the tunnel structure. Real-time monitoring data can detect abnormalities in a timely manner and improve the safety of tunnel operations;

[0079] Based on the monitoring data, the structural risk monitoring data of the tunnel structure in the past month is obtained, and the sensitive factors that have a significant impact on the structural risk are extracted by using the monitoring data. According to different types of monitoring data, corresponding feature labels are set, including but not limited to:

[0080] Set tunnel environment dataX env. The temperature, humidity, water level, air pressure and water permeability are the first feature tags;

[0081] Set structure health data X struct The median displacement, stress-strain, vertical and lateral additional loads, tunnel soil stress change, crack base, tunnel load compression coefficient, tunnel soil Poisson’s ratio, deformation curvature and compression modulus are the second characteristic tags;

[0082] Set subway operation data X oper The train speed, load and frequency are the third feature labels,

[0083] Set historical data X hist Past fault records and maintenance records are the fourth feature tag;

[0084] By extracting sensitive factors and focusing on the variables that have the greatest impact on structural health, the accuracy of analysis can be improved, and key feature labels can be determined to simplify the data processing process and effectively reduce the consumption of computing resources;

[0085] Set corresponding feature labels based on the type of monitoring data. According to different types of monitoring data, feature labels can be regarded as classification and labeling of monitoring data for subsequent feature extraction and analysis. Using a systematic approach to manage monitoring data is helpful for subsequent data analysis and processing. Clear labels help machine learning models understand and process data, improving model performance and accuracy.

[0086] Obtain the correlation coefficient between the monitoring data and the feature label, and calculate the correlation coefficient between the monitoring data and different feature labels. Select a threshold (set the correlation coefficient to 0.9) to determine that the correlation between the feature label and the target variable is strong; through correlation analysis, the features that have a greater impact on structural risks can be screened out, the accuracy and calculation efficiency of the model can be improved, and only the features with strong correlation are retained as risk sensitivity factors for further analysis;

[0087] The preset threshold is obtained, and the preset correlation coefficient is 0.9, which is used to determine whether the correlation coefficient is significant enough. When the correlation coefficient between a feature label and the monitoring data is greater than 0.9, the feature is considered to have high sensitivity, thereby improving the standardization and uniformity of the feature screening process and ensuring that the selected features are of practical significance in structural risk assessment;

[0088] Based on the correlation coefficient and the preset threshold, the feature labels with high sensitivity are obtained. The actual calculated correlation coefficient is compared with the preset threshold, and the feature labels with correlation coefficients higher than the threshold are screened out, that is, the feature labels with high sensitivity, which accurately screen out the features that are critical to structural risk assessment, improve the effectiveness of the model, focus more on key variables, and reduce redundant information interference;

[0089] Defining highly sensitive feature labels as structural risk sensitive factors as input for subsequent risk assessment, using streamlined sensitive factors as input helps improve the reliability and interpretability of analysis results and facilitates comprehensive structural risk impact assessment;

[0090] Based on the structural risk sensitivity factors, the comprehensive structural risk impact is obtained, and the safety risk classification index is set. Based on the structural risk sensitivity factors, the comprehensive structural risk impact is obtained, and the safety risk classification index is set. The comprehensive impact evaluation can more intuitively display the health status of the structure, thereby providing clear risk level indicators, which is helpful to formulate corresponding risk management strategies;

[0091] Based on the structural risk sensitivity factors and tunnel data, the structural risk responsibility parameters are obtained, and then the structural risk responsibility parameters are obtained by randomly combining multiple sensitive factor interaction items through multiple sensitive factors. The structural risk responsibility parameters can reflect the risk sources in the structure and their importance. By providing specific parameter values, it is convenient for detailed analysis and evaluation, which can help identify high-risk areas and implement targeted repair measures.

[0092] Obtain the structural risk level based on the safety risk classification index and the structural risk responsibility parameter. By combining the safety risk classification index and the structural risk responsibility parameter, determine whether the structural risk level obtained from the historical data meets the low risk level, medium risk level or high risk level. When the medium or high risk level is obtained, an early warning alert is required to ensure an accurate assessment of the overall structural risk, so that decision makers can understand the structural health status, provide graded risk results, and help formulate priority repair plans;

[0093] Based on the monitoring data and structural risk level, a data set is constructed to train and obtain a prediction model. The machine learning data set is constructed using historical monitoring data and structural risk level. The prediction model is trained to predict future structural risk trends. The model is trained with historical data to improve the prediction accuracy and provide prediction results for future trends, so that preventive measures can be taken in advance.

[0094] Obtain new monitoring data, obtain structural risk trends based on the new monitoring data and the prediction model, input the new monitoring data into the trained prediction model, obtain the structural risk level for the next month through the prediction model based on the structural risk level obtained from historical monitoring data, compare the structural risk level of the historical monitoring data with the structural risk level of the new monitoring data to obtain the structural risk trend, update the prediction results in real time, provide dynamic structural risk assessment, so as to help operation and maintenance personnel respond to possible structural risk changes in a timely manner, and issue secondary alerts based on the risk level obtained from historical detection data: when the historical monitoring data is at a low or high risk level, and the new monitoring data predicts a medium or high risk level, the risk trend is in an upward state, and a secondary early warning alert is required; conversely, no early warning alert is required if the risk trend decreases.

[0095] According to the structural risk trend, the change of structural risk level is obtained, and the warning information is output in a preset form. Based on the structural risk trend, the change of structural risk level is judged, and the warning information is output in a preset form (such as alarm, report, etc.), providing timely risk warning, improving response speed, reducing potential risks and hidden dangers, and ensuring the long-term safe operation of the tunnel.

[0096] In another preferred embodiment of the present invention, the following steps may also be included:

[0097] Step 1: obtain monitoring data in real time through optical fiber sensing signals and pre-process the monitoring data, specifically including: decomposing the interference signal of the optical fiber signal monitoring data by wavelet packet to obtain the optical fiber enhanced characteristic signal, and collecting tunnel environment data X according to the band signal change characteristics of different optical fiber enhanced characteristic signals env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist ;Preprocess the collected monitoring data to obtain data of uniform scale;

[0098] Step 2, analyzing the preprocessed monitoring data to obtain the structural risk sensitivity factor, specifically comprising: comparing the correlation coefficient between each preprocessed monitoring data and the feature label of each monitoring data with a preset threshold, and sorting the structural risk feature sensitivity coefficients based on the comparison result to retain the structural risk features with high sensitivity;

[0099] Step 3, setting a safety risk classification index according to the structural risk sensitivity factor, and obtaining a structural risk level through a structural risk responsibility parameter;

[0100] Step 4: construct a prediction model based on the structural risk level to predict the structural risk trend;

[0101] Step 5: Dynamically display the changes in security risks through visualization based on the structural risk trends, and formulate an early warning mechanism.

[0102] In the embodiment of the present invention, by using the optical fiber sensor to obtain and process data in real time through the optical fiber signal in step 1, it can be ensured that the decision-making information is the latest and most accurate. By standardization processing, the differences between different data features can be eliminated, making the subsequent analysis process more stable and reliable;

[0103] Then, we clean the data to handle missing values ​​and outliers, and calculate the mean μ of each feature. j and standard deviation σ j To standardize the data: ; In the formula, z ij is the standardized data point, x ijis the original data point, the value of the i-th row and the j-th column, and the standardized data is represented by the matrix Z as: ; Wherein, N and M are positive integers.

[0104] According to the monitoring data obtained in step 1, the features that have a greater impact on structural risks are obtained through step 2, and then the sensitivity judgment of the structural risk factor features is used to select sensitive factors, so as to identify and extract the factors that are most sensitive to structural risk changes, which can improve the accuracy of subsequent risk analysis and prediction, and simplify the data model and improve the calculation efficiency by reducing irrelevant or redundant features; according to the interaction items of the structural risk sensitive factors collected in step 2, the risk classification indicators are set, and then the risk responsibility parameters are obtained to classify the structural risks. The clear risk classification allows different response measures to be taken for risks of different levels. By assigning reasonable weights to each sensitive factor, the comprehensive risk can be evaluated more scientifically; based on the structural risk level obtained in step 3, a prediction model is constructed. The structural risk trend is predicted by the prediction model. Through an effective prediction model, the risk trend can be identified in advance, and more time can be gained for taking preventive measures. Dynamic monitoring and real-time updating of the model can improve the sensitivity and accuracy of the prediction. Then, according to the structural risk prediction trend data obtained by the prediction model in step 4, the visualization tool is used to dynamically display the changing trend of the risk level, and the early warning mechanism is set according to the risk level threshold. When the real-time predicted risk level exceeds the preset threshold, the early warning is triggered, and maintenance or repair measures are taken to avoid the risk. Therefore, the use of visual dynamic display allows decision makers to intuitively understand the risk change trend, which is convenient for real-time monitoring. Through the early warning mechanism, potential risks can be responded to in a timely and effective manner, the probability of accidents can be reduced, and the safety of the tunnel structure can be improved.

[0105] In a preferred embodiment of the present invention, the above step 1 further includes:

[0106] Step 11: Substitute the interfered fiber signal into the wavelet packet inner product space basis function system Perform linear superposition, expressed as: ; Where f(t) is the useful optical fiber characteristic signal, c λ is the interference signal processing coefficient, A is the threshold of the spatial basis function system to remove redundant interference fiber signals;

[0107] Step 12, decomposing the useful signal features into low-frequency and high-frequency parts, and obtaining decomposition coefficients through multi-layer wavelet packet decomposition, and setting thresholds for each group of decomposition coefficients, including: mean value Ei, variance σi and comprehensive value Ti;

[0108] Step 13, screening multiple wavelet packet decomposition layers one by one according to the set decomposition coefficient thresholds of each group to obtain time domain features and frequency domain features;

[0109] Step 14, obtaining an optical fiber enhancement characteristic signal according to the acquired time domain characteristics and frequency domain characteristics;

[0110] Step 15, the acquired time domain and frequency characteristic information is concentrated in the coefficients with large modulus values, and the corresponding regular statistical differences are obtained through the time domain characteristics and frequency domain characteristics, and the statistical difference characteristic quantity is used to distinguish the vibration segment and the non-vibration segment in the optical fiber enhanced characteristic signal, and the vibration behavior type is identified according to the different vibration band signals, thereby obtaining the tunnel environment data X env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist .

[0111] In the embodiment of the present invention, the average value Ei obtained in step 12 is Where S 3，i is the wavelet decomposition system, N is the number of groups, j is the jth layer of wavelet packet decomposition; the variance σi is ; In the formula, S(j) is the expected value of the decomposition system; the comprehensive value Ti is , described by the signal denoising process: ; F(t) is the denoised optical fiber signal. According to the denoised optical fiber signal obtained: Perform hierarchical wavelet packet decomposition; where W is the decomposition result, WPT is the wavelet transform function, and c is the number of decompositions;

[0112] According to the average value Ei, variance σi and comprehensive value Ti obtained in step 12, the average value Ei of different components can represent the average level of the signal in different frequency bands in step 13. The time domain features are obtained as follows: ; The variance σi can be used to represent the change of the signal in different frequency bands and obtain the time domain characteristics: ; In the formula, t is the time domain feature, L is the length of the fiber signal after denoising, and v is the propagation speed of the fiber signal after denoising. is the dot product; the energy distribution and main frequency components of the signal in each frequency band can be characterized by the wavelet packet decomposition coefficient and its comprehensive value Ti, and the frequency domain characteristics are obtained as follows: ; where δ(f−f i ) represents the frequency domain position of the i-th frequency band.

[0113] According to step 13, in step 14, the extracted time domain and frequency domain features are combined to form an enhanced feature signal: ; F(t) z is the optical fiber enhanced characteristic signal, T is the time domain characteristic, is the frequency domain feature, α and β are the signal enhancement weight coefficients;

[0114] Finally, use the Brillouin scattering frequency shift in step 15: ; In the formula, Δv B is the Brillouin frequency shift, C T , C ϵ is the constant coefficient, T is the temperature data, The temperature distribution is measured by the optical signal in the optical fiber. The Brillouin scattering frequency shift changes with the temperature and the gauge factor. The temperature can be inferred by measuring the frequency shift. Then, using Brillouin scattering, the Brillouin frequency shift distribution of the optical fiber length is measured. Brillouin scattering is not only sensitive to temperature, but also to the gauge factor. The strain distribution can also be measured by the change in the Brillouin frequency shift, and then used ; to sense the vibration along the optical fiber, and monitor the health status of the tunnel structure, such as cracks and deformation, by analyzing the vibration distribution of the frequency shift signal. 0 is the frequency of scattered light, P is the pressure of the optical fiber, c is the speed of light, T 0 for temperature, E is the elastic modulus, Δ T is the temperature change, Δ σ The stress changes; then the optical fiber sensor combined with the optical fiber acoustic wave obtains the tunnel train running signal, through: Real-time monitoring of tunnel strain caused by train passing, thereby obtaining the structural bending and displacement caused by train operation, where: λ B is the optical fiber reflection wavelength, λ 0 is the initial wavelength, Δλ is the wavelength shift and the relationship between temperature T and gauge factor The original data in the database is combined with optical fiber sensors to continuously record environmental, structural health and subway operation data, which are stored in the central database: The historical data can be retrieved at any time for historical trend analysis. The historical data includes long-term records of parameters such as temperature, strain, and vibration. In the formula, T i is the temperature parameter at the i-th moment, is the strain coefficient and V at the ith moment i is the vibration data at the i-th moment, t i is the timestamp, from which the tunnel environment data X is obtained env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist .

[0115] In a preferred embodiment of the present invention, the above step 2 may include:

[0116] Step 21, based on the pre-processed tunnel environment data X env., Structural health data X struct 、Subway operation dataX oper And historical data X hist ,pass: , obtain the correlation coefficient between each monitoring data and the feature label corresponding to each monitoring data;

[0117] Where, X j Represents the monitoring data after preprocessing, Y i Represents the characteristic labels of each monitoring data, ρ(X j , Y i ) represents data X j With label Y i The correlation coefficient, Cov(X j , Y i ) represents data X j With label Y i The covariance of j Represents feature X j The standard deviation of i Indicates label Y i The standard deviation of

[0118] Step 22, based on the correlation coefficients of each monitoring data and the feature labels corresponding to each monitoring data, by: , obtain the sensitivity of feature labels in each monitoring data;

[0119] In the formula, I Xj represents the sensitivity coefficient of each monitoring data after preprocessing, T represents the total number of monitoring data, Imp Xj，t represents the feature sensitivity in the t-th monitoring data;

[0120] Step 23, the sensitivity of the feature label is the structural risk feature sensitivity coefficient;

[0121] Step 24, the preset threshold is a set correlation coefficient, a feature label whose absolute value of the correlation coefficient is greater than the preset threshold is obtained, a structural risk feature sensitivity coefficient corresponding to the feature label is obtained, and a structural risk sensitivity factor is obtained based on the size of the structural risk feature sensitivity coefficient;

[0122] Step 25, interactively combine the obtained structural risk sensitive factors to obtain the comprehensive structural risk impact.

[0123] In this embodiment, step 21 can reveal the relationship and dependency between each data item and its feature label by calculating the correlation coefficient between them, which is helpful to understand the role of different monitoring data items in structural health and risk assessment. Analyzing the correlation coefficient can help identify redundant features. Merging highly correlated features or eliminating redundant features can reduce data dimensions and improve computational efficiency. The streamlined feature set can reduce model complexity and improve the generalization ability of the model. Using the correlation coefficient to screen out important features can improve the prediction performance of the machine learning model, improve the quality and annotation effect of the data set, and make model training and evaluation more accurate. Feature interactions, combined features or derived features can be constructed based on the correlation results to provide data support for subsequent modeling and analysis steps, which helps to improve the overall analysis effect; then, step 22 is used to evaluate the sensitivity of the features and identify the features that are most sensitive to structural risk changes, which helps to further screen and weight the role of each feature in risk assessment; then, step 23 is used to select the most sensitive features based on the tunnel environment data X. env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist The correlation coefficient between each monitoring data and the feature label of each monitoring data is compared with the preset threshold value 0.9 to obtain the comparison result. According to the comparison result of step 23, the feature sensitivity structural risk assessment results are sorted through step 24, and the most sensitive feature is selected as the structural risk sensitive factor, thereby retaining the features with the greatest impact on the structural risk, improving the prediction effect and accuracy of the model, and making the risk prediction more focused on key factors. Finally, step 25 considers the interaction between features to obtain the comprehensive impact of structural risk, further improves the accuracy of the prediction model, and the generated comprehensive features can capture complex nonlinear relationships and provide more information for risk prediction.

[0124] In a preferred embodiment of the present invention, the above step 25 specifically further includes:

[0125] Step 251, encode the structural risk sensitivity factor as {A 1 i , A 2 i , A 3 i , …, A z i}, where i={1, 2, …, m} represents the number of monitoring times; A z Represents the code of the zth structural risk sensitive factor;

[0126] Step 252, by: Where I AB，quadratic represents the nonlinear interaction between sensitive factors, U nIt is expressed as the nth structural risk sensitive factor, and n is an integer, thereby obtaining multiple interactive combination results; where g i is the coding vector of tunnel soil stress change, N i is the encoding vector of the crack cardinality, KO i is the encoding vector of the tunnel load compression coefficient, V i is the encoding vector of the Poisson's ratio of the tunnel soil layer, E i is the encoding vector of the compression modulus and H i is the encoding vector of water permeability;

[0127] Step 253, according to the multiple interaction results, by: , where R is the comprehensive structural risk coefficient, β 0 is the influence intercept index, β 1 The coefficient of the interactive combination result of sensitive factors, the comprehensive structural risk coefficient is the comprehensive structural risk impact.

[0128] In this embodiment, the code of tunnel soil stress change g, crack cardinality N, tunnel load compression coefficient KO, tunnel soil Poisson's ratio V, compression modulus E and water permeability H is {g i , N i , K.O. i , V i , E i , H i}, so that every two risk-sensitive factors interact through: There are 15 interaction results, where each of the three risk-sensitive factors interacts through: Ten kinds of interaction results can be obtained, and each of the four risk-sensitive factors interacts through: There are 15 interaction results, where each of the five risk-sensitive factors interacts through: Six interaction results can be obtained, and the six factors can produce a set of interaction results.

[0129] In a preferred embodiment of the present invention, the above step 3 may include:

[0130] Step 31, dividing the comprehensive structural risk impact R into risk intervals, specifically comprising: dividing R into multiple risk intervals according to the number of interaction items of different risk sensitive factors, namely interval 1: N≤k1; interval 2: k1<N≤k2; interval 3: k2<N≤k3, wherein N is the number of interaction items of sensitive factors, k1, k2 and k3 are respectively preset thresholds of interaction items of sensitive factors, and k1<k2<k3;

[0131] Step 32, according to the multiple risk intervals, by: ; In the formula, F is the safety risk classification index, Rlow is the low impact interval, R med. is the medium influence interval, R high It is a high impact interval, from which the security risk classification index is obtained;

[0132] Step 33, obtaining the structural risk level according to the obtained safety risk classification index and the structural risk responsibility parameter a.

[0133] In this embodiment, step 31 uses the number of sensitive factor interaction items to divide the risk interval, which can clearly define each risk level and help better understand and monitor structural risks. This division method based on quantitative thresholds can flexibly adapt to risk assessment needs in different situations and has strong operability. According to step 32, F is introduced as a safety risk classification indicator to express the level of comprehensive structural risk impact, and by dividing the low, medium and high impact intervals, different levels of comprehensive structural risks can be clearly defined, providing a clear basis for subsequent decision-making. The classification indicator F provides a standardized and clear reference system for structural risk management; then, step 33 combines the comprehensive risk score with the structural risk responsibility parameter to make the risk assessment more scientific and authoritative. By setting the structural risk level, risk management and control are facilitated, providing a basis for early warning mechanisms and response measures.

[0134] In another preferred embodiment of the present invention, the above step 33 may include:

[0135] Step 331, select n structural risk sensitive factors and randomly combine them into N interaction items to form sensitive factor interaction data;

[0136] Step 332, the sensitive factor interaction data is passed through: , to obtain the structural risk liability parameter a, where N is the number of sensitive factor interaction terms, w i is the weight of the sensitive factor responsibility parameter, F i is the sensitive factor interaction data;

[0137] Step 333, according to the structural risk responsibility parameter a and the safety risk classification index, the structural risk weight G is obtained by: G = a·F, and the structural risk degree is calculated according to the structural risk weight G by: Multi-level classification is performed to obtain the structural risk level, which is low risk level, medium risk level and high risk level. In the formula, RiskLevel is the risk level, R total It is the risk sensitivity factor of the overall risk structure.

[0138] In this embodiment, 53 risk-sensitive factor interaction items can be obtained through step 2, of which there are 6 for one risk factor, 15 for two risk factors, 10 for three risk factors, 15 for four risk factors, 6 for five risk factors, and 1 for six risk factors. Therefore, when the risk responsibility parameter a is an interaction item of one or two risk factors, there are 21 results in the low impact interval, which are classified as low risk level by G=a·F, accounting for 39.6%; when the risk responsibility parameter a is an interaction item of three or four risk factors, there are 25 results in the medium impact interval, which are classified as low risk level by G=a·F, accounting for 39.6%; It is classified as medium risk level, accounting for 47.2%; when the risk responsibility parameter a is the interaction term of five and six risk factors, a total of 7 results are in the high impact interval, which is classified as high risk level through G=a·F, accounting for 13.2%, thus forming a structural risk level system.

[0139] In a preferred embodiment of the present invention, the above step 4 may include:

[0140] Step 41, construct a data set by combining the monitoring data and the structural risk level. The data set is the training matrix K. When the training matrix K contains g different time states V t , so the training matrix is ;

[0141] Step 42: V t In t j The n risk transformation values ​​observed at a time point are regarded as the observation vector ; Input the observation vector into the prediction model, including: ; To obtain the risk level category of monitoring data;

[0142] In the formula, H is the low-dimensional representation after encoding, f is the activation function, and W enc Encoding weight matrix, b enc Encoder bias vector, V t (t j ) input data, X is the reconstructed data, e is the activation function, W dec is the decoding weight matrix, b dec is the decoding bias vector, L is the loss function value, and D is the number of data points;

[0143] Step 43, the new monitoring data is input into the prediction model, the new monitoring data risk level category is output, and the risk level category of the new monitoring data is compared with the risk level category of the monitoring data to obtain the structural risk trend.

[0144] In this embodiment, the above implementation steps can provide insights into the changing trends of structural risks and predict potential structural problems in advance by comparing historical and current risk levels; structural risk trend analysis can provide a scientific basis for decision makers to help them formulate maintenance and repair strategies more reasonably. Combining risk trends with prediction models can build a more accurate early warning system, which helps prevent and reduce potential catastrophic events. By accurately monitoring and analyzing structural risks, it can optimize the allocation of maintenance and repair resources and improve operational efficiency.

[0145] In another preferred embodiment of the present invention, the above step 42 includes:

[0146] Step 421, optimize the prediction model parameters by gradient descent: , where

[0147] η is the learning rate, L is the gradient length, is the symbol for partial differentials.

[0148] In a preferred embodiment of the present invention, the above step 5 includes:

[0149] Step 51, setting a risk warning threshold according to the risk level category of the monitoring data output by the prediction model to generate warning information;

[0150] Step 52, based on the risk level category of the monitoring data, when a preset risk warning threshold is reached, a first warning alarm is issued;

[0151] Step 53, based on the structural risk trend, the structural risk level change between the monitoring data and the new monitoring data is obtained, warning information is triggered, and a secondary warning alarm is issued.

[0152] In this embodiment, a warning threshold is set according to the risk level, and data points with large reconstruction errors are identified as abnormal: Where D g is the risk threshold, A g The early warning mechanism for the i-th data point is the early warning information. 1 means that the Risklever risk level abnormal value of the historical monitoring data for one month is compared with the preset risk threshold. If the abnormal value is equal to or greater than the risk threshold, the early warning mechanism is triggered to issue an alarm. 0 means normal, alerting the staff to the risk level of the tunnel section structure obtained from the historical data;

[0153] Then, by inputting the new monitoring data into the prediction model, the risk level of the new monitoring data is obtained through: ; By comparing historical monitoring data with new monitoring data, where R historical is the historical risk level data, R' future is the future risk prediction result, b is the comparison weight, Rb The structural risk trend is predicted based on the risk level changes of historical monitoring data and new monitoring data. future >R historical If the structural risk trend is on the rise, a secondary alarm will be triggered to further alert the staff to the tunnel structure that needs to be paid special attention to in the future to avoid escalation of risks.

[0154] like Figure 2 As shown, a subway tunnel structure safety risk dynamic prediction system includes:

[0155] Data acquisition module 201: used to obtain monitoring data; obtain structural risk sensitivity factors based on the monitoring data; wherein corresponding feature labels are set based on the type of monitoring data; obtain the correlation coefficient between the monitoring data and the feature labels; obtain a preset threshold; obtain a feature label with high sensitivity based on the correlation coefficient and the preset threshold; the feature label with high sensitivity is the structural risk sensitivity factor;

[0156] Data analysis module 202: It is used to obtain the comprehensive structural risk impact based on the structural risk sensitivity factor and set the safety risk classification index; obtain the structural risk responsibility parameter based on the structural risk sensitivity factor and tunnel data; obtain the structural risk level based on the safety risk classification index and the structural risk responsibility parameter;

[0157] Risk prediction module 203: It is used to construct a data set based on the monitoring data and the structural risk level, train and obtain a prediction model; obtain new monitoring data, and obtain a structural risk trend based on the new monitoring data and the prediction model;

[0158] The early warning module 204 is used to obtain the change of the structural risk level according to the structural risk trend and output the early warning information in a preset form.

[0159] If the functions are implemented in the form of 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 the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, ROM, RAM, disk or optical disk, etc., various media that can store program codes.

[0160] Therefore, the purpose of the present invention can also be achieved by running a program or a group of programs on any computing device. The computing device can be a well-known general device. Therefore, the purpose of the present invention can also be achieved by simply providing a program product containing a program code that implements the method or device. That is to say, such a program product also constitutes the present invention, and the storage medium storing such a program product also constitutes the present invention. Obviously, the storage medium can be any well-known storage medium or any storage medium developed in the future. It should also be pointed out that in the device and method of the present invention, it is obvious that each component or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present invention. In addition, the steps of performing the above-mentioned series of processing can naturally be performed in chronological order according to the order of description, but it is not necessary to perform them in chronological order. Some steps can be performed in parallel or independently of each other.

Claims

1. A method for dynamic prediction of subway tunnel structural safety risk, characterized in that: The following steps are involved: Obtain monitoring data; Based on monitoring data, obtain structural risk sensitivity factors; Among them, corresponding feature labels are set based on the type of monitoring data; Obtain the correlation coefficient between monitoring data and feature labels; obtain the preset threshold; Based on the correlation coefficient and the preset threshold, a highly sensitive feature label is obtained; Feature labels with high sensitivity are structural risk sensitive factors; Based on the structural risk sensitivity factors, the comprehensive structural risk impact is obtained, and the safety risk classification index is set; based on the structural risk sensitivity factors and tunnel data, the structural risk responsibility parameters are obtained; based on the safety risk classification index and the structural risk responsibility parameters, the structural risk level is obtained; Building a data set based on the monitoring data and the structural risk level, training and obtaining a prediction model; Obtain new monitoring data, and obtain structural risk trends based on the new monitoring data and the prediction model; Acquire changes in structural risk levels according to the structural risk trend, and output warning information in a preset form; Wherein, the monitoring data and the new monitoring data are both obtained in real time through optical fiber sensing signals; wherein the method for obtaining the monitoring data includes: Substitute the interfering fiber signal into the wavelet packet inner product space basis function system Perform linear superposition, It is expressed as: ; Where f(t) is the useful optical fiber characteristic signal, c λ is the interference signal processing coefficient, A is the threshold of the spatial basis function system to remove redundant interference fiber signals; Decompose the useful signal features into low-frequency and high-frequency parts, and obtain the decomposition coefficients through multi-layer wavelet packet decomposition, and set thresholds for each group of decomposition coefficients, including: mean value Ei, variance σi and comprehensive value Ti; According to the set threshold values ​​of each group of decomposition coefficients, multiple wavelet packet decomposition layers are screened one by one to obtain time domain features and frequency domain features; Obtaining an optical fiber enhancement characteristic signal according to the acquired time domain characteristics and frequency domain characteristics; The acquired time domain and frequency characteristic information is concentrated in the coefficients with large modulus values, and the corresponding regular statistical differences are obtained through the time domain characteristics and frequency domain characteristics, and the statistical difference characteristic quantity is used to distinguish the vibration segment and the non-vibration segment in the optical fiber enhanced characteristic signal; the vibration behavior type is identified according to the different vibration band signals, and the tunnel environment data X is obtained through the vibration behavior type. env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist ; The method of obtaining the structural risk sensitive factor includes: Based on the pre-processed tunnel environment data X env. , Structural health data X struct 、Subway operation dataX oper And historical data X hist ,pass: , obtain the correlation coefficient between each monitoring data and the feature label corresponding to each monitoring data; In the formula, X j Represents the monitoring data after preprocessing, Y i Represents the characteristic labels of each monitoring data, ρ(X j , Y i ) represents data X j With label Y i The correlation coefficient, Cov(X j , Y i ) represents data X j With label Y i The covariance of j Represents feature X j The standard deviation of i Indicates label Y i The standard deviation of Based on the correlation coefficient between each monitoring data and the feature label corresponding to each monitoring data, through: , obtain the sensitivity of feature labels in each monitoring data; In the formula, I Xj represents the sensitivity coefficient of each monitoring data after preprocessing, T represents the total number of monitoring data, Imp Xj，t represents the feature sensitivity in the t-th monitoring data; The sensitivity of the feature label is the structural risk feature sensitivity coefficient; The preset threshold is a set correlation coefficient threshold, a feature label whose absolute value of the correlation coefficient is greater than the preset threshold is obtained, a structural risk feature sensitivity coefficient corresponding to the feature label is obtained, and a structural risk sensitivity factor is obtained based on the size of the structural risk feature sensitivity coefficient; The obtained structural risk sensitive factors are interactively combined to obtain the comprehensive structural risk impact; Among them, the method of obtaining the comprehensive structural risk impact includes: The structural risk sensitivity factor is encoded as {A 1 i , A 2 i , A 3 i , …, A z i }, where i={1, 2, …, m} represents the number of monitoring times; A z Represents the code of the zth structural risk sensitive factor; pass: Where, I AB，quadratic represents the nonlinear interaction between sensitive factors, 2≤n≤6, and n is an integer, U n It is expressed as the nth structural risk sensitive factor, U1, U2, U3, U4, U5, U6 are respectively related to {g i , N i , K.O. i , V i , E i , H i }, thus obtaining a variety of interactive combination results; among them, g i is the coding vector of tunnel soil stress change, N i is the encoding vector of the crack cardinality, KO i is the encoding vector of the tunnel load compression coefficient, V i is the encoding vector of the Poisson's ratio of the tunnel soil layer, E i is the encoding vector of the compression modulus and H i is the encoding vector of water permeability; According to the multiple interaction results, by: , where R is the comprehensive structural risk coefficient, β0 is the impact intercept index, β1 is the coefficient of the interactive combination of sensitive factors, and the comprehensive structural risk coefficient is the comprehensive structural risk impact; The method for obtaining the structural risk level includes: Dividing the comprehensive structural risk impact R into risk intervals specifically includes: dividing R into multiple risk intervals according to the number of interaction items of different risk sensitive factors, namely interval one: N≤k1; interval two: k1<N≤k2; interval three: k2<N≤k3, wherein N is the number of interaction items of sensitive factors, k1, k2 and k3 are respectively preset thresholds of interaction items of sensitive factors, and k1<k2<k3; Based on multiple risk categories, through: ; In the formula, F is the safety risk classification index, R low is the low impact interval, R med. is the medium influence interval, R high It is a high impact interval, from which the security risk classification index is obtained; The structural risk level is obtained by combining the obtained safety risk classification index with the structural risk responsibility parameter a.

2. A method for dynamic prediction of subway tunnel structural safety risk according to claim 1, characterized in that: The method for obtaining the structural risk level also includes: Select n structural risk sensitive factors and randomly combine them into N interaction terms to form sensitive factor parameter data; The sensitive factor parameter data is combined with the maximum values ​​of the tunnel vertical and lateral additional loads, displacements, and relative deformation curvature to obtain the structural risk responsibility parameter a; According to the structural risk responsibility parameter a and the safety risk classification index, the structural risk weight G is obtained through: G=a·F; and the structural risk degree is divided according to the structural risk weight G to obtain the structural risk level, which are low risk level, medium risk level and high risk level.

3. A method for dynamic prediction of subway tunnel structural safety risk according to claim 2, characterized in that: The method of building a prediction model based on the structural risk level to predict the structural risk trend includes: The monitoring data and the structural risk level are combined to construct a data set, which is the training matrix K. When the training matrix K contains g different time states V t , so the training matrix is ; V t In t j The n risk transformation values ​​observed at a time point are regarded as the observation vector ; Input the observation vector into the prediction model, including: ; To obtain the risk level category of monitoring data; In the formula, H is the low-dimensional representation after encoding, f is the activation function, and W enc Encoding weight matrix, b enc Encoder bias vector, V t (t j ) input data, X is the reconstructed data, e is the activation function, W dec is the decoding weight matrix, b dec is the decoding bias vector, L is the loss function value, and D is the number of data points; The new monitoring data is then input into the prediction model, and the new monitoring data risk level category is output and compared with the risk level category of the monitoring data to obtain the structural risk trend.

4. A method for dynamic prediction of subway tunnel structural safety risk according to claim 3, characterized in that: According to the structural risk level, a prediction model is constructed to predict the structural risk trend, including: Optimize the prediction model parameters by gradient descent: , where η is the learning rate, L is the gradient length, is the symbol for partial differentials.

5. A method for dynamic prediction of subway tunnel structural safety risk according to claim 4, characterized in that: The method for outputting warning information in a preset form includes: According to the risk level category of the monitoring data output by the prediction model, the risk warning threshold is set to generate warning information; Based on the risk level category of the monitoring data, when the preset risk warning threshold is reached, the first warning alarm is issued; Based on the structural risk trend, the structural risk level changes of the monitoring data and the new monitoring data are obtained, triggering early warning information and issuing a secondary early warning alarm.

6. A subway tunnel structure safety risk dynamic prediction system, characterized in that: The system is provided with an electronic device including a memory, a processor, and a subway tunnel structure safety risk dynamic prediction method program stored in the memory and executable on the processor. When the subway tunnel structure safety risk dynamic prediction method program is executed by the processor, the steps of the subway tunnel structure safety risk dynamic prediction method as claimed in any one of claims 1 to 5 are implemented. The system includes: Data acquisition module: used to obtain monitoring data; obtain structural risk sensitivity factors based on the monitoring data; set corresponding feature labels based on the type of monitoring data; obtain the correlation coefficient between the monitoring data and the feature labels; obtain a preset threshold; obtain a feature label with high sensitivity based on the correlation coefficient and the preset threshold; the feature label with high sensitivity is the structural risk sensitivity factor; Data analysis module: It is used to obtain the comprehensive structural risk impact based on the structural risk sensitivity factor and set the safety risk classification index; obtain the structural risk responsibility parameter based on the structural risk sensitivity factor and tunnel data; obtain the structural risk level based on the safety risk classification index and the structural risk responsibility parameter; Risk prediction module: It is used to construct a data set based on the monitoring data and the structural risk level, train and obtain a prediction model; obtain new monitoring data, and obtain the structural risk trend based on the new monitoring data and the prediction model; Early warning module: It is used to obtain the change of structural risk level according to the structural risk trend and output early warning information in a preset form.

Images