Subway tunnel wall stress self-adaptive monitoring method and device and medium

Abstract

The invention discloses a subway tunnel wall stress self-adaptive monitoring method and device and a medium, and the method comprises the steps: obtaining a vibration signal collected by a train vibration sensor and an optical signal collected by a distributed optical fiber sensor in real time, demodulating the vibration signal and the optical signal, and calculating the actual stress under the thermal-mechanical coupling effect, dynamic displacement of the tunnel wall is extracted according to the vibration signal, dynamic stress is calculated, and then actual stress under the heat-force-vibration multi-field coupling effect is calculated; carrying out noise reduction processing on the stress signal monitored in real time by using wavelet transform, and predicting actual stress data under the multi-field coupling action by using Kalman filtering according to the stress signal subjected to noise reduction processing to obtain corrected stress data; and inputting the corrected stress data into the LSTM-SVM model to obtain a final stress prediction result. According to the method, the stress state of the subway tunnel wall can be comprehensively and continuously monitored, and meanwhile, the monitoring precision, stability and robustness in a complex and dynamic tunnel environment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of stress monitoring, and in particular to a subway tunnel wall stress adaptive monitoring method, device and medium. Background Art

[0002] Traditional tunnel stress monitoring typically uses point-to-point sensors such as stress sensors for data collection. Stress sensors typically employ resistance strain gauges and piezoelectric sensors. While they can provide stress data locally, they are limited by the number and location of sensors deployed, resulting in a limited monitoring range and spatial distribution. This makes it impossible to comprehensively monitor stress changes throughout the entire tunnel, and thus unable to fully reflect the health status of the tunnel structure. This is especially true in long tunnels or under complex geological conditions, where there are many blind spots. Furthermore, these sensors have long monitoring cycles, making it difficult to meet the needs of real-time monitoring. They are also often sensitive to environmental factors (such as temperature and electromagnetic interference) and are prone to measurement errors. Consequently, these sensors present problems such as poor real-time performance, narrow coverage, inability to continuously monitor, and low monitoring accuracy.

[0003] Fiber optic sensors utilize the optical properties of optical fibers to sense physical quantities such as temperature, pressure, and vibration, enabling comprehensive and continuous capture of changing signals. However, existing fiber optic sensors are typically used to monitor tunnel deformation. Furthermore, direct use of fiber optic sensors for tunnel wall stress monitoring only detects mechanical stress. However, during actual subway tunnel operation, the stress experienced by tunnel walls is affected by the coupling of multiple physical fields, such as train vibration and temperature fluctuations. Consequently, the mechanical stress detected by direct fiber optic sensors deviates from the actual stress value, making accurate stress monitoring difficult.

[0004] During subway operations, it's often necessary to analyze the stress distribution of subway tunnel walls to identify areas of stress concentration and potential weak links in advance, thereby guiding tunnel design optimization. However, sensors only capture real-time stress states. Existing techniques for predicting stress distribution typically use large amounts of stress measurement data to train neural network models. However, tunnel stress changes are influenced by a variety of complex factors, exhibiting nonlinearity, time-dependent dependencies, and uncertainty. Traditional neural network models often struggle to cope with high-noise, complex time-series data, resulting in unstable and inaccurate predictions in complex environments. Summary of the Invention

[0005] The technical problem to be solved by the present invention is: in response to the technical problems existing in the prior art, the present invention provides a subway tunnel wall stress adaptive monitoring method, device and medium with simple implementation method, high monitoring accuracy and stability, strong environmental adaptability and robustness, which can comprehensively and continuously monitor the stress state of the subway tunnel wall, and at the same time improve the monitoring accuracy, stability and robustness in complex and dynamic tunnel environments.

[0006] In order to solve the above technical problems, the technical solution proposed by the present invention is: A method for adaptively monitoring subway tunnel wall stress, comprising the following steps: Real-time monitoring of vibration signals collected by train vibration sensors and optical signals collected by distributed optical fiber sensors installed on subway tunnel walls, demodulation of the monitored optical signals to obtain stress signals, and extraction of temperature changes and Bragg wavelength changes. Calculation of actual stress under thermal-mechanical coupling based on the extracted temperature changes and Bragg wavelength changes. Extraction of dynamic displacement of the tunnel wall caused by train vibration based on the monitored vibration signals, calculation of dynamic stress based on the dynamic displacement of the tunnel wall, and calculation of actual stress data under thermal-mechanical-vibration multi-field coupling based on the actual stress under thermal-mechanical coupling and the dynamic stress calculation. The stress signal monitored in real time by the distributed optical fiber sensor is subjected to noise reduction processing using wavelet transform, the stress signal after noise reduction processing is used as observation data, the actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the initial stress state data, and Kalman filtering is used to perform prediction to obtain corrected thermal-mechanical-vibration coupled stress data; The corrected thermal-mechanical-vibration coupled stress data is input into a pre-trained stress prediction model to obtain the final stress prediction result output of the subway tunnel wall. The stress prediction model is obtained by training an LSTM-SVM model using a training dataset composed of historical stress data. The LSTM (Long Short-Term Memory)-SVM (Support Vector Machine) model uses the LSTM model to obtain preliminary stress prediction results. The output of the LSTM model is used as the input of the SVR model, and the SVR model corrects the output of the LSTM model to obtain the final stress prediction result.

[0007] Furthermore, the actual stress data under the thermal-mechanical-vibration multi-field coupling is obtained by calculating the actual stress under the thermal-mechanical coupling and the dynamic stress, including: The dynamic displacement of the tunnel wall is extracted from the collected vibration signal, and the real-time temperature change and Bragg wavelength change are extracted from the demodulated stress signal; Calculate the actual stress under thermal-mechanical coupling based on the change in Bragg wavelength and real-time temperature change; Calculating dynamic strain based on the real-time dynamic displacement of the tunnel wall, and then calculating dynamic stress caused by train vibration based on the calculated dynamic strain; The actual stress under the thermal-mechanical coupling action and the dynamic stress caused by the train vibration are superimposed to obtain the actual stress under the thermal-mechanical-vibration multi-field coupling action.

[0008] Furthermore, the calculation expression for calculating the actual stress under the thermal-mechanical coupling is: , in, represents the actual stress under the thermal-mechanical coupling effect, represents the effective elastic-optical coefficient, represents the effective refractive index, represents the elastic modulus of the optical fiber, represents the total change in Bragg wavelength, represents the Bragg wavelength, represents the thermal expansion coefficient of the optical fiber material, represents the thermo-optical coefficient, Indicates the temperature change.

[0009] Furthermore, the calculation expression of the dynamic stress caused by train vibration is calculated as: , in, represents the dynamic stress caused by train vibration, represents the elastic modulus of the optical fiber, represents dynamic strain; Actual stress under the coupling of thermal, mechanical and vibration fields The calculation expression is: , When the train vibration induces the elliptical deformation of the tunnel wall, the dynamic stress caused by the train vibration is the mean tangential stress : , in, is the tunnel radius, is the train vibration amplitude, is the frequency, is the Poisson's ratio of the material, For time.

[0010] Furthermore, the noise reduction processing of the stress data monitored in real time by the distributed optical fiber sensor using wavelet transform includes: A wavelet basis is selected according to the frequency characteristics and noise distribution of the stress data, and the selected wavelet basis is used to perform multi-layer wavelet decomposition on the stress signal monitored in real time by the distributed optical fiber sensor. In each layer of decomposition, the input stress signal is convolved with the scaling function to obtain low-frequency coefficients, and the input stress signal is convolved with the wavelet basis function to obtain low-frequency coefficients. The next level further decomposes the low-frequency coefficients obtained by the previous level of decomposition, and in each layer of decomposition, a preset dynamic threshold is used to perform denoising on the high-frequency coefficients to remove high-frequency noise. The dynamic threshold is dynamically set according to the parameters of the high-frequency coefficients of each layer. Finally, the low-frequency coefficients obtained by the multi-layer wavelet decomposition and the high-frequency coefficients after denoising are subjected to inverse wavelet transform to obtain the stress data after denoising. The stress data after the primary denoising is subjected to an adaptive filter to remove low-frequency noise, and the weight of the filter is continuously adjusted to adaptively track changes in the low-frequency noise, thereby obtaining stress data after the secondary denoising.

[0011] Furthermore, the denoising process using a preset dynamic threshold to remove high-frequency noise on high-frequency coefficients in each layer of decomposition includes: If the absolute value of the effective signal corresponding coefficient exceeding the specified ratio in the current high-frequency coefficient is greater than the dynamic threshold or the absolute value of the noise corresponding coefficient exceeding the specified ratio is less than the dynamic threshold, the hard threshold method is used for processing according to the dynamic threshold, otherwise the soft threshold method is used for processing according to the dynamic threshold. If the hard threshold method is used for processing, hour, , hour, Indicates the j The high-frequency coefficients after layer wavelet decomposition, Represents the high-frequency coefficient after denoising; If the soft threshold method is used for processing, hour, , When the high frequency coefficients of the current layer are After the reduction process, the high-frequency coefficients after the denoising process are obtained. ; Dynamic Threshold The calculation expression is: , in, For the The standard deviation of the high-frequency coefficients of the layer, For the The length of the high-frequency coefficients of the layer, is the influencing factor of tunnel vibration noise.

[0012] Furthermore, the stress signal after noise reduction is used as observation data, and the actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the stress state data at the initial moment, and prediction is performed using an extended Kalman filter, including: according to To predict the status, To predict the error covariance, Calculate the Kalman gain ,according to To update the status, follow Perform error covariance update, where is the Kalman gain at the kth moment, is the covariance of the stress state after the update at the kth moment, is the covariance of the predicted stress state at the kth moment, is the observation matrix, is the observed value at the kth moment, that is, the stress signal monitored in real time by the distributed optical fiber sensor at the kth moment is the stress signal after noise reduction obtained by wavelet transform. is the observation noise covariance, To predict the stress state at the kth moment, at the initial moment is the actual stress data under the thermal-mechanical-vibration multi-field coupling, is the identity matrix, is the updated stress state estimate at the kth moment, i.e., the corrected thermal-mechanical-vibration coupled stress data, and are the Jacobian matrices respectively.

[0013] Furthermore, the corrected thermal-mechanical-vibration coupled stress data is input into a pre-trained stress prediction model to obtain the stress prediction result output of the subway tunnel wall, including: Inputting the corrected thermal-mechanical-vibration coupled stress data into an LSTM model pre-trained using historical stress data, and the LSTM model predicting stress changes at future moments based on the historical stress data to obtain a stress prediction result; The output of the LSTM model is used as the input feature of the SVR model. The SVR model performs nonlinear optimization on the stress prediction value output by the LSTM model to obtain the optimized prediction result. The output of the LSTM model is fused with the output of the SVR model to obtain the fusion prediction result; The final stress prediction result is obtained according to the optimized prediction result and the fused prediction result.

[0014] Furthermore, it also includes the implementation of multi-field coupling stress over-limit warning according to the following decision rules:

[0015] in, Indicates the judgment indicator, 1 means triggering the over-limit warning, 0 means not triggering the over-limit warning, is the stress prediction value output by the LSTM-SVR model, is the baseline threshold, is the tunnel environmental sensitivity coefficient, is the historical stress standard deviation.

[0016] A subway tunnel wall stress adaptive monitoring device comprises a processor and a memory, wherein the memory is used to store a computer program, and the processor is used to execute the computer program to perform the above method.

[0017] A computer-readable storage medium storing a computer program, wherein the computer program implements the above method when executed by a processor.

[0018] Compared with the prior art, the present invention has the following advantages: the present invention utilizes distributed optical fiber sensors installed on the tunnel wall to monitor the stress signal of the tunnel wall in real time, demodulates the optical signal, and converts it into stress data by considering the thermal-mechanical-vibration multi-field coupling effect. The actual stress under the thermal-mechanical coupling effect and the dynamic stress caused by train vibration are comprehensively considered to obtain the actual stress. The mechanical stress collected by the optical fiber sensor can be accurately converted into the actual stress of the subway tunnel wall under the multi-physical field coupling effect. The calculated actual stress under the multi-physical field coupling effect is used as the initial state data, and the stress signal collected by the distributed optical fiber sensor is used as the observation data after noise reduction. Kalman filtering is used for prediction to obtain corrected thermal-mechanical-vibration coupling stress data, which can effectively correct errors to obtain more accurate multi-physical field coupling stress data. The stress prediction model constructed by the LSTM-SVM model is then used for stress prediction. The LSTM model and the SVM model can be combined to give full play to the advantages of both in time series dependence and nonlinear optimization, further improving the prediction accuracy, reliability and robustness, and meeting the real-time and accuracy requirements of subway tunnel stress monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the implementation process of the subway tunnel wall stress adaptive monitoring method of this embodiment.

[0020] Figure 2 This is a schematic diagram of the implementation process of extracting optical signals and calculating stress values ​​based on optical fiber sensing technology in this embodiment.

[0021] Figure 31 is a schematic diagram of the implementation flow of wavelet transform in this embodiment.

[0022] Figure 4 3 is a flow chart of the prediction and update implementation of the Kalman filter in this embodiment.

[0023] Figure 5 3 is a schematic diagram of the implementation process of the stress prediction model training and prediction in this embodiment. DETAILED DESCRIPTION

[0024] The present invention will be further described below in conjunction with the accompanying drawings and specific preferred embodiments, but the scope of protection of the present invention is not limited thereby.

[0025] like Figure 1 As shown, the steps of the subway tunnel wall stress adaptive monitoring method of this embodiment include: Step S01. Real-time monitoring of vibration signals collected by train vibration sensors and optical signals collected by distributed optical fiber sensors installed on subway tunnel walls, demodulation of the monitored optical signals to obtain stress signals, and extraction of temperature changes and Bragg wavelength changes. Calculation of actual stress under thermal-mechanical coupling based on the extracted temperature changes and Bragg wavelength changes, extraction of dynamic displacement of the tunnel wall caused by train vibration based on the monitored vibration signals, and calculation of dynamic stress based on the dynamic displacement of the tunnel wall. Actual stress data under thermal-mechanical-vibration multi-field coupling is obtained based on the actual stress under thermal-mechanical coupling and the dynamic stress calculation.

[0026] Step S101: Fiber optic sensor data acquisition and demodulation In this embodiment, distributed fiber optic sensors are installed on the subway tunnel walls to monitor stress signals in real time. Light signals within the optical fibers are collected in real time, and data collected over a long monitoring period is extracted from the fiber optic sensors. A fiber optic demodulator is used to demodulate the information in the optical signals into stress signals. By using fiber-optic sensing technology to survey subway tunnel walls, full coverage of the tunnel walls can be easily achieved, providing long-term, continuous data support. This facilitates continuous, real-time stress monitoring throughout the entire tunnel structure, improving the comprehensiveness and real-time nature of the data, enabling high-precision, real-time stress monitoring of subway tunnels while also adapting to complex environmental changes.

[0027] This embodiment utilizes a stress monitoring system based on fiber-optic sensing technology to achieve high-precision, real-time monitoring of subway tunnel walls. The distributed deployment of fiber-optic sensors within the tunnel structure enables comprehensive and continuous capture of stress change signals, making it particularly suitable for complex and dynamic tunnel environments. Because fiber-optic sensing technology is insensitive to external electromagnetic interference and boasts long-distance transmission, it can continuously collect high-quality data over long periods. Extracting stress information through a fiber-optic interrogator ensures high accuracy and stability of the stress data, providing reliable foundational data for subsequent analysis.

[0028] This embodiment also adds a vibration sensor to the monitoring system to monitor key parameters of train vibration in real time, such as vibration frequency. , amplitude et al., combined with the vibration data of the vibration sensor to correct the stress value calculated directly based on the fiber Bragg grating (FBG) principle.

[0029] Step S102. Stress calculation After the stress signal is demodulated in step S101, further stress calculation is performed to convert it into stress data. During actual subway tunnel operation, the stress on the tunnel wall is not simply mechanical stress, but is affected by the coupling of multiple physical fields, such as train vibration and temperature changes. This embodiment fully considers the thermal-mechanical-vibration multi-field coupling and calculates the actual stress under this multi-field coupling based on the vibration signal and real-time temperature changes, generating real-time monitored stress data.

[0030] In this embodiment, calculating the actual stress under the thermal-mechanical-vibration multi-field coupling includes: Step S121. Extracting the dynamic displacement of the tunnel wall based on the collected vibration signal, and extracting the real-time temperature change based on the demodulated stress signal; Step S122. Calculate the actual stress under the thermal-mechanical coupling according to the Bragg wavelength and the real-time temperature change; Step S123. Calculate the dynamic strain based on the real-time dynamic displacement of the tunnel wall, and then calculate the dynamic stress caused by the train vibration based on the calculated dynamic strain; Step S124. Superimpose the actual stress under the thermal-mechanical coupling and the dynamic stress caused by the train vibration to obtain the actual stress under the thermal-mechanical-vibration multi-field coupling.

[0031] This embodiment considers the influence of multi-physical field coupling, such as train vibration and temperature change, and incorporates the effect of temperature change on the Bragg wavelength into the stress calculation. This can more accurately reflect the actual stress of the tunnel wall under thermal-mechanical coupling. At the same time, the dynamic stress caused by train vibration is calculated based on the dynamic displacement of the tunnel wall caused by train vibration. The actual stress under thermal-mechanical coupling and the dynamic stress caused by train vibration are combined to obtain the final actual stress under thermal-mechanical-vibration multi-field coupling. This can comprehensively consider the thermal-mechanical-vibration multi-field coupling, effectively improve the accuracy of subway tunnel wall stress calculation, and achieve accurate calculation of subway tunnel wall stress.

[0032] To calculate the actual stress under the coupled effects of heat, force, and vibration, this embodiment performs the following analysis: First, let’s analyze the traditional stress calculation process based on the principle of fiber Bragg grating (FBG): Bragg wavelength and the period of the fiber Bragg grating and the effective refractive index The following relationship exists: (1) in, represents the grating period.

[0033] When the fiber Bragg grating is subjected to stress, its period and effective refractive index will change, resulting in the Bragg wavelength According to the theory of elastic optics, the relationship between strain, effective refractive index and Bragg wavelength change is expressed as: (2) in, is the change in Bragg wavelength, is the effective elastic-optical coefficient of the optical fiber, for example, for quartz optical fiber, The value of is usually around 0.22, Indicates strain.

[0034] According to Hooke's law in material mechanics, stress and strain The following relationship exists: (3) in, is the elastic modulus of the optical fiber, for example, for silica optical fiber, The value of is approximately Pa.

[0035] According to the above formula, the stress Variation of Bragg wavelength The relationship between: (4) Then, the stress value calculated based on the fiber Bragg grating (FBG) principle is corrected, and the static stress σ′ calculated by the thermal-mechanical coupling model is combined with the dynamic stress caused by train vibration. The actual stress under the thermal-mechanical-vibration multi-field coupling is obtained by superposition During actual operation of subway tunnels, the tunnel walls are affected by the coupling of multiple physical fields, such as train vibration and temperature changes. When the temperature changes, the effective refractive index and grating period of the optical fiber will change, causing the Bragg wavelength to drift. The temperature change ΔT will affect the effective refractive index and grating period. The change in effective refractive index due to temperature change can be expressed as: (5) in, is the thermo-optical coefficient, represents the effective refractive index of the optical fiber, Indicates that due to temperature changes The resulting change in the effective refractive index.

[0036] The change in grating period due to temperature change can be expressed as: (6) in, is the thermal expansion coefficient of the optical fiber material, represents the grating period, Indicates that due to temperature changes The resulting change in the grating period is Indicates the temperature change.

[0037] Combining the above two formulas, we can deduce that the temperature changes The change in Bragg wavelength caused by : (7) After sorting, we get: (8) According to the FBG principle, the change in Bragg wavelength caused by stress is , then the Bragg wavelength changes caused by temperature change and stress are superimposed to obtain the total Bragg wavelength change + .

[0038] The original stress before the thermal-mechanical coupling correction is Variation of Bragg wavelength The relationship is as follows: (9) in, is the effective elastic-optical coefficient, E is the elastic modulus of the optical fiber, Represents the effective refractive index.

[0039] Actual stress under thermal-mechanical coupling after correction for thermal-mechanical coupling (static stress σ′) Total change with Bragg wavelength The relationship between them is: (10) in, represents the actual stress under the thermal-mechanical coupling effect, represents the effective elastic-optical coefficient, represents the effective refractive index, represents the elastic modulus of the optical fiber, represents the total change in Bragg wavelength, represents the Bragg wavelength, represents the thermal expansion coefficient of the optical fiber material, represents the thermo-optical coefficient, Indicates the temperature change.

[0040] This embodiment constructs a thermal-mechanical coupling model shown in the above formula (10). The model incorporates the influence of temperature change on the Bragg wavelength into the stress calculation, which can more accurately reflect the actual stress of the tunnel wall under the action of thermal-mechanical coupling.

[0041] Further analysis of the dynamic stress caused by train vibration: This embodiment is based on the theory of structural dynamics to establish a vibration model that fits the structural characteristics of subway tunnels. A simplified single-degree-of-freedom vibration model is used as an example. Specifically, the dynamic displacement x(t) of the tunnel wall caused by train vibration can be expressed as: (11) Due to dynamic stress and dynamic strain The dynamic strain is related to the dynamic displacement. Assuming that the elastic modulus of the tunnel wall material is E, the dynamic strain can be obtained by taking the derivative of the dynamic displacement. , and then calculate the dynamic stress caused by train vibration : (12) in, represents the dynamic stress caused by train vibration, represents the elastic modulus of the optical fiber, represents dynamic strain.

[0042] Then the static stress calculated by the thermal-mechanical coupling model is Dynamic stress caused by train vibration By superposition, the actual stress of the tunnel wall under the coupling of thermal, mechanical and vibration fields is obtained. : (13) Therefore, the actual stress under the thermal-mechanical-vibration multi-field coupling can be obtained The calculation expression is: (14) Furthermore, in the scenario of coupling between elliptical deformation of subway tunnel and train vibration, assuming that the tunnel radius is R, the train vibration amplitude is A, the frequency is f, and the material Poisson's ratio is , is time, then when the train vibration causes the tunnel wall to ellipse, its radial displacement can be expressed as: ,in is the circumferential angle, which results in a radial strain of , the tangential strain is According to material mechanics, the tangential stress , then considering the circumferential symmetry of the elliptical deformation, the mean tangential stress is taken as: (15) That is, in the scenario where train vibration causes elliptical deformation of the tunnel wall, the dynamic stress caused by the train vibration The mean tangential stress : (16) The above formula (15) is obtained by introducing R, A, f, , it is possible to quantify the coupling effect between the tunnel's elliptical deformation and the train's vibration, and to better fit the dynamic stress characteristics of subway tunnels.

[0043] The mean tangential stress After superimposing the thermal-mechanical coupling stress, the actual stress under the thermal-mechanical-vibration multi-field coupling of the elliptical deformation of the fusion tunnel in this scenario can be obtained. : (17) like Figure 2 As shown, this embodiment obtains the optical signal collected by the optical fiber sensor installed on the subway tunnel wall, demodulates the optical signal using an optical fiber demodulator, and calculates the actual stress value according to the above formula (14) or (17) based on the Bragg wavelength change, temperature change and vibration data. By adopting the above method, the actual stress value under the thermal-mechanical-vibration multi-field coupling is constructed. The calculation model can comprehensively consider the thermal-mechanical-vibration multi-field coupling and realize the accurate calculation of subway tunnel wall stress.

[0044] Step S02. The stress signal monitored in real time by the distributed optical fiber sensor is subjected to noise reduction processing using wavelet transform. The stress signal after noise reduction processing is used as the observation data, and the actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the stress state data at the initial moment and predicted using Kalman filtering to obtain the corrected thermal-mechanical-vibration coupling stress data.

[0045] In tunnel stress monitoring, collected signals often contain significant noise due to factors such as environmental noise and equipment interference. Especially in complex environments, factors such as temperature fluctuations and mechanical interference can cause signal errors and distortion. While traditional signal processing methods can denoise, their effectiveness is limited and they are prone to introducing processing errors. The Kalman filter uses a recursive algorithm to gradually optimize the signal estimate, effectively removing high-frequency noise. The wavelet transform, through multi-resolution analysis, can separate the noise component of the signal from the actual signal, further improving signal quality. This embodiment combines the wavelet transform with the Kalman filter to denoise the wavelength variation signal obtained from the fiber optic sensor. The wavelet transform first removes high-frequency noise, and the Kalman filter further smoothes the signal, preventing the Kalman filter from being affected by excessive high-frequency noise. The wavelet transform also separates the high- and low-frequency components of the signal, and the Kalman filter uses a model to further optimize the signal. This not only extracts the valid signal and effectively removes high-frequency noise from the signal, but also optimizes the signal through the Kalman filter, improving signal quality and accuracy.

[0046] In this embodiment, the stress data monitored in real time by the distributed optical fiber sensor is first subjected to noise reduction processing using wavelet transform, and the steps include: Step S201: Select a wavelet basis based on the frequency characteristics and noise distribution of the stress data. Use the selected wavelet basis to perform multi-layer wavelet decomposition on the stress data monitored in real time by the distributed optical fiber sensor. Apply a preset dynamic threshold to the low-frequency coefficients and high-frequency coefficients to remove high-frequency noise, obtaining denoised low-frequency and high-frequency coefficients. Perform an inverse wavelet transform to obtain the primary denoised stress data. Step S202: Using an adaptive filter to remove low-frequency noise from the stress data after primary denoising, and continuously adjusting the weight of the filter to adaptively track changes in the low-frequency noise, thereby obtaining stress data after secondary denoising.

[0047] Thermal coupling can introduce low-frequency noise into the signal due to temperature fluctuations and train vibrations. Furthermore, the fiber demodulation process and environmental interference can introduce high-frequency noise. Low-frequency noise can mask the true stress trends in the tunnel wall, while high-frequency noise can interfere with signal details. When vibration sensors collect train vibration signals, they are also affected by environmental vibrations and their own noise. During the noise reduction process, this embodiment first selects a wavelet basis with better time-frequency localization characteristics based on the signal's frequency characteristics and noise distribution. Wavelet transforms are then used to perform multi-scale decomposition of the signal, extracting low-frequency and high-frequency components. Threshold denoising methods (such as soft thresholding) are then used to remove high-frequency noise. This effectively preserves the signal's primary characteristics while reducing its impact on stress prediction. Furthermore, given the characteristics of low-frequency noise caused by temperature changes and train vibration, an adaptive filter is used to remove low-frequency noise. The filter weights are continuously adjusted to adaptively track changes in the low-frequency noise, effectively removing the low-frequency noise. Kalman filtering is then used to further optimize the signal. Prediction and updating steps eliminate residual noise, improving signal accuracy and stability. These noise reduction and optimization processes ensure the provision of higher-quality stress data and further optimize the accuracy and reliability of the stress signal. This ensures the acquisition of stable and accurate stress data even in complex environments, meeting the requirements of real-time monitoring. In a specific application embodiment, Figure 3 As shown, the specific steps of implementing the wavelet transform on the stress signal monitored in real time by the distributed optical fiber sensor in this embodiment include: Step S211. Select appropriate wavelet basis and decompose This embodiment selects a wavelet basis with better time-frequency localization characteristics based on the frequency characteristics and noise distribution of the signal, uses the selected wavelet basis to decompose the signal into a low-frequency (approximation) part and a high-frequency (detail) part, and extracts the multi-scale features of the signal through multi-level wavelet decomposition.

[0048] Assume that the stress signal monitored in real time by the distributed optical fiber sensor For a one-dimensional signal, the wavelet basis function is used when performing each level of decomposition and scaling function Convolve the signal where: Low frequency coefficient (approximation): by input stress signal and scaling function (Translation and scaling of the mother wavelet) to perform convolution and obtain the low-frequency coefficients : ; High frequency coefficients (details): By comparing the input stress signal with the wavelet basis function (Translation and scaling of the mother wavelet) to perform convolution and obtain the high-frequency coefficients : ; According to the above method, the stress signal Perform multi-layer wavelet decomposition to obtain approximations and details of multiple scales. The scale of the multi-layer wavelet decomposition is determined by the number of decomposition layers (e.g., the first-level decomposition corresponds to scale 2¹, the second-level decomposition corresponds to scale 2²). Decompose into low-frequency approximation part Aj and high-frequency detail part Dj of different frequency bands. The specific scale can be determined by selecting the wavelet basis based on the frequency characteristics of the stress signal and the noise distribution. Each level of decomposition will further decompose the low-frequency part, that is, the next level will further decompose the low-frequency coefficients obtained by the previous level decomposition. For example, the first level decomposition can obtain and , then Perform the second-level decomposition and obtain , and so on.

[0049] Step S212: Threshold denoising This embodiment effectively removes high-frequency noise by threshold processing of high-frequency coefficients. For the tunnel fiber monitoring scenario, when denoising is performed on the high-frequency coefficients in each layer of decomposition using a preset dynamic threshold to remove high-frequency noise, if the absolute value of the coefficient corresponding to the valid signal exceeding the specified proportion in the current high-frequency coefficients is greater than the dynamic threshold T, or the absolute value of the coefficient corresponding to the noise exceeding the specified proportion is mostly less than the dynamic threshold T, that is, the absolute value of the valid signal coefficient is mostly greater than T or the absolute value of the coefficient corresponding to the noise is mostly less than the dynamic threshold T, it indicates that the valid signal and the noise can be clearly distinguished in amplitude by the dynamic T, that is, the boundary between the high-frequency noise and the valid signal is clear, and the hard threshold method is used to directly filter out the noise below the threshold; otherwise, if the noise and the valid signal spectrum overlap a lot, the soft threshold method is used to smooth the high-frequency coefficients to retain signal details.

[0050] When using hard thresholding: ; ; in, It is j The high-frequency coefficients (detail coefficients) after layer wavelet decomposition, represents the high-frequency coefficient after denoising, It is a preset dynamic threshold. If the absolute value of the coefficient is greater than the set threshold, the coefficient is retained; if it is less than the threshold, it is set to zero. Dynamic Threshold It can be obtained by setting the following formula: (18) in, For the The standard deviation of the layer's high-frequency coefficients (detail coefficients), For the The length of the high-frequency coefficients of the layer, is the tunnel vibration noise impact factor (for example, it can be taken as 0.1-0.3).

[0051] As shown in formula (18), this embodiment constructs a dynamic threshold based on the extreme value statistical theory of Gaussian noise. , first assume that The noise component of the layer wavelet detail coefficient obeys the zero-mean Gaussian distribution, combined with the coefficient length , the theoretical threshold term of pure noise scene is obtained from the extreme asymptotic characteristics However, the detail coefficient of tunnel stress contains both effective signals (thermal-mechanical-vibration coupling characteristics) and noise. The signal will rise to the extreme value, and the vibration noise needs to be dynamically adapted. Therefore, the signal-noise coupling correction term is introduced. Finally, the theoretical threshold is multiplied by the dynamic correction term to construct a dynamic threshold that integrates the statistical laws of noise, signal energy characteristics, and environmental interference characteristics. , can realize dynamic threshold adjustment by fusing noise intensity and signal mean, thereby further improving the accuracy and robustness of denoising.

[0052] When using the soft threshold method: (19) The soft threshold method is similar to the hard threshold method. The high-frequency coefficient of the layer is greater than the dynamic threshold T ,Right now ,according to Get the high-frequency coefficient after denoising, but when the absolute value of the high-frequency coefficient When it is less than the dynamic threshold T, that is, , the high frequency coefficients of the current layer are reduced and used as the high frequency coefficients after denoising , instead of setting it to zero directly.

[0053] After the above threshold processing, the denoised low-frequency and high-frequency coefficients are reconstructed back to the original signal through the Inverse Wavelet Transform (IWT), assuming that the signal has been decomposed by wavelet to obtain low-frequency and high-frequency coefficients at different levels. ., the transform combines these coefficients to recover an approximate version of the original signal: ).

[0054] The stress signal is subjected to preliminary noise reduction processing through the above wavelet transform, which mainly removes the high-frequency noise in the signal. However, the signal may still contain some low-frequency noise or abnormal fluctuations that have not been completely removed. The signal can be further optimized through Kalman filtering. Figure 4 As shown, through the prediction and update steps of the Kalman filter, an accurate signal is extracted from the noise interference. In this embodiment, the actual stress data under the force-vibration multi-field coupling calculated in step S01 is first used as the initial state estimation, and the Kalman filter is used for prediction. The expression is: (20) (twenty one) in, is the predicted stress state at the current moment, and at the initial moment This is the actual stress data obtained in step S01 under the thermal-mechanical-vibration multi-field coupling. is the covariance matrix of the state estimate, is the process noise covariance matrix.

[0055] After the prediction, the stress signal after the above noise reduction processing is used as the observation data, and the actual observation value is combined with the predicted value to update the current state estimate. The expression for state update using Kalman filtering is: (twenty two) (twenty three) (twenty four) in, is the Kalman gain used to weigh the “confidence” between the predicted value and the observed value, is the measurement noise covariance matrix, is the updated state estimate, that is, the corrected thermal-mechanical-vibration coupled stress data.

[0056] Assuming a simple linear model, the state transfer matrix A = 1, that is, the current state is equal to the state at the previous moment (assuming that the state changes linearly), and the observation matrix H = 1, that is, each observation directly reflects the current state. For the first moment (k = 1), there is no state estimate of the previous moment, and the initial estimate can be used directly: ; Forecast error covariance: ; for time, assuming it is known ,as well as ,therefore: ; Update the current state estimate based on the predicted state and Kalman gain: ; That is, the updated state estimate at the first moment is 1.71875, and the updated error covariance is: .

[0057] Furthermore, considering the existence of nonlinear dynamic equations in the tunnel stress system, such as the vibration amplitude of the signal, the elastic limit of the metamaterial, or the existence of an ill-conditioned observation matrix (such as the Jacobian matrix condition number ), or there is non-stationary noise (such as the noise covariance ratio ), etc., use the extended Kalman filter (EKF) for prediction. Based on the core covariance update formula of the Kalman filter, the extended Kalman filter is locally linearized so that it can be applied to nonlinear, pathological, and non-stationary tunnel stress systems.

[0058] Specifically, the calculation expression for prediction using the extended Kalman filter (EKF) in this embodiment is: Status prediction: ; Error covariance prediction: ; Kalman gain: ; Status Update: ; Error covariance update: ; in, is the Kalman gain at the kth moment, is the covariance of the stress state after the update at the kth moment, is the covariance of the predicted stress state at the kth moment, is the observation matrix, is the observed value at the kth moment, that is, the stress signal monitored in real time by the distributed optical fiber sensor at the kth moment is the stress signal after noise reduction obtained by wavelet transform. is the observation noise covariance, To predict the stress state at the kth moment, at the initial moment This is the actual stress data obtained in step S01 under the thermal-mechanical-vibration multi-field coupling. is the identity matrix, is the updated stress state estimate at the kth moment, i.e., the corrected thermal-mechanical-vibration coupled stress data, and is the Jacobian matrix, which is used to linearize nonlinear functions.

[0059] According to the above method, the observation data at the second and third moments are calculated in turn. Through the prediction and update process of the Kalman filter, the stress value at each moment is gradually estimated. As the number of observation data increases, the confidence in the state estimation gradually increases (the error covariance gradually decreases), making the estimated value more reliable than the simple observation value, especially in the case of large noise.

[0060] This embodiment combines multi-level wavelet transform with Kalman filtering for noise reduction processing, which can give full play to the advantages of each and effectively reduce the noise components in the signal. The wavelet transform can decompose the signal into components of different scales, and the threshold denoising method can be used to remove high-frequency noise, retaining the main information of the signal. The Kalman filter is used through the prediction and update steps to further optimize the signal estimation, reduce measurement errors, and improve the accuracy and stability of the signal. It is particularly suitable for high-noise environments and can ensure the accuracy of stress monitoring data.

[0061] The processing of stress data in this embodiment can be divided into the following three stages: ① Theoretical prior: First, in step S01, the multi-field coupling model such as Equation (14) or (17) is used to calculate the actual stress data under the thermal-mechanical-vibration multi-field coupling as the theoretical prior data, and the actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the initial state data of the Kalman filter prediction, that is, the multi-field coupling model , achieving physical constraints; ②Observation purification: The optical signal collected in real time by the distributed optical fiber sensor is subjected to wavelet transform and noise reduction as observation data, that is, the original optical fiber data. , achieving denoised observation and noise suppression; ③ Fusion correction: The actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the initial state data for Kalman filter prediction, and the stress data after noise reduction is used for Kalman filter prediction, that is, , achieve optimal stress estimation and obtain corrected thermal-mechanical-vibration coupled stress data , as input data for subsequent stress prediction models; This embodiment can integrate the "predicted stress state" with the "actual observation value" through the above method, effectively correct the error, and obtain more reliable stress estimation data. .

[0062] Due to the introduction of train vibration sensors, this embodiment fuses the vibration signals collected by the vibration sensors with the optical fiber sensor signals before noise reduction processing. Through time synchronization, the data of different types of sensors are aligned in the time dimension, and then a multidimensional data vector containing wavelength change signals and vibration signals is constructed as input for subsequent noise reduction processing.

[0063] Specifically, after obtaining the vibration signal collected by the train vibration sensor and the optical signal collected in real time by the distributed optical fiber sensor and before noise reduction processing, this embodiment also includes a pre-processing step of fusing the vibration signal of the train vibration sensor and the optical signal of the distributed optical fiber sensor, including: The vibration signal of the train vibration sensor is time-synchronized with the optical signal of the distributed optical fiber sensor to construct a multi-dimensional data vector containing the wavelength change signal and the vibration signal; After preliminary denoising processing on the multidimensional data vector, it is input into a pre-trained denoising model to obtain a multi-field coupled signal after collaborative denoising processing. The denoising model is trained based on a deep learning model using a training set including vibration signals from train sensors and optical signals from optical fiber sensors to learn the noise characteristics and true signal characteristics in the signal.

[0064] Specifically, this embodiment processes different types of noise differently. Given the characteristics of low-frequency noise caused by temperature changes and train vibration, adaptive filtering technology (such as an LMS filter) is employed. By continuously adjusting the filter weights, the filter adaptively tracks changes in low-frequency noise, effectively removing the low-frequency noise. When processing high-frequency noise, the wavelet transform optimizes the selection of wavelet bases, selecting a wavelet base with better time-frequency localization characteristics based on the signal's frequency characteristics and noise distribution. Before noise reduction, the fiber optic sensor and vibration sensor signals are fused and preprocessed. Time synchronization is used to ensure temporal consistency between the two sensor data types. A multidimensional data vector is then constructed, and dimensionality reduction is performed on the multidimensional data to extract key data features, reduce data redundancy, and improve noise reduction efficiency. Finally, a joint noise reduction algorithm is employed, leveraging the complementary information from the two sensor data types. The preliminarily noise-reduced fiber optic sensor and vibration sensor signals are then input into a deep learning-based noise reduction model. For example, using a CNN model, multiple convolutional and pooling layers are constructed to automatically learn the noise and true signal characteristics of the signals, achieving collaborative noise reduction of multi-field coupled signals.

[0065] Step S03. Input the corrected thermal-mechanical-vibration coupled stress data into a pre-trained stress prediction model to obtain the final stress prediction result output of the subway tunnel wall. The stress prediction model is obtained by training the LSTM-SVM model using a training data set composed of historical stress data. The LSTM-SVM model uses the LSTM model to obtain preliminary stress prediction results. The output of the LSTM model is used as the input of the SVR model, and the SVR model corrects the output of the LSTM model to obtain the final stress prediction result.

[0066] After signal noise reduction, a relatively clean stress signal is obtained. This embodiment establishes a stress prediction model based on the stress signal, utilizing historical data and current real-time data to accurately predict stress. Stress signal prediction tasks often face challenges due to time-series dependencies and nonlinear characteristics. As a deep learning model, LSTM is highly capable of capturing time-series patterns in data, while SVR excels at processing nonlinear relationships in data. Due to the nonlinearity, time-series dependencies, and noise characteristics of tunnel stress signals, this embodiment combines a long short-term memory network (LSTM) with support vector regression (SVR) (LSTM-SVR) to achieve prediction, leveraging the advantages of both while overcoming their respective limitations. LSTM learns the time-series characteristics of stress changes to accurately predict future stress trends, while SVR performs nonlinear optimization based on the LSTM output, further improving prediction accuracy. The integration of LSTM and SVR enables automatic adjustment and optimization of prediction results based on actual conditions, improving not only the accuracy of stress prediction but also the robustness and adaptability of the system. This enhances the accuracy, robustness, and applicability of the model, enabling it to cope with complex stress variations in tunnel environments.

[0067] In this embodiment, the specific steps of inputting the corrected thermal-mechanical-vibration coupled stress data into a pre-trained stress prediction model to obtain the stress prediction result output of the subway tunnel wall include: Step S301: Input the real-time denoised stress data into an LSTM model pre-trained with historical stress data. The LSTM model predicts stress changes at future moments based on the historical stress data to obtain stress prediction results. Step S302: The output of the LSTM model is used as the input feature of the SVR model, and the SVR model performs nonlinear optimization on the stress prediction value output by the LSTM model to obtain an optimized prediction result; Step S303: Fuse the output of the LSTM model with the output of the SVR model to obtain a fusion prediction result; Step S304: Obtain the final stress prediction result based on the optimized prediction result and the fused prediction result.

[0068] In this embodiment, the output of the LSTM model and the output of the SVR model can be fused using the following fusion strategy: Direct fusion method: The output of the LSTM model is used as the input feature of the SVR model, and the SVR model outputs the fusion prediction result; Multi-stage fusion method: The output of the LSTM model is used as the preliminary prediction result, and the SVR model is used to correct the output of the LSTM model to obtain the fusion prediction result; Weighted fusion method: The output of the LSTM model and the output of the SVR model are weighted with different weights to obtain the fusion prediction result.

[0069] The above fusion strategies can adopt one of them, or a combination of two or more can be adopted according to actual needs. The weight of the model can also be adaptively adjusted according to different stress change patterns to better cope with complex tunnel stress changes.

[0070] In the process of establishing the stress prediction model, this embodiment realizes stress prediction by combining the LSTM and SVR models, which can give full play to the advantages of both in time series dependence and nonlinear regression optimization. LSTM is used to capture the time dependence of tunnel stress data, which can accurately model the change trend of stress. SVR is used to perform nonlinear optimization based on LSTM to further improve the accuracy and robustness of the prediction. The combination of fusion strategy can also effectively combine the advantages of LSTM and SVR, reduce their respective limitations, and thus establish an accurate and robust stress prediction model to achieve more accurate stress prediction and meet the actual needs of subway tunnel stress monitoring.

[0071] In a specific application embodiment, Figure 5 As shown in the figure, the detailed steps of LSTM-SVR model training and prediction are as follows: (1) LSTM model training The basic architecture of LSTM includes: input layer → LSTM layer → fully connected layer → output layer; Specifically, the input layer receives the thermal-mechanical-vibration coupled stress data corrected by the Kalman filter, which is symbolically represented as: ; yes The optimal stress estimate output by the Kalman filter at the moment, i.e., the corrected thermal-mechanical-vibration coupled stress data; The LSTM layer builds a core layer by stacking multiple layers of LSTM units to capture the long-term dependencies of stress time series. The multi-layer expansion strengthens the model's ability to learn non-stationary multi-field coupling characteristics. Since tunnel stress prediction is a continuous value regression problem (non-classification), the output layer preferably uses a linear activation function (to avoid the interval compression error of Sigmoid), and the formula is: (25) in, for Predicted value of thermal-mechanical-vibration coupled stress at each moment, is the hidden state output by the LSTM layer, are the weights and biases of the fully connected layer.

[0072] LSTM training process: Select a loss function suitable for the regression problem, such as mean squared error (MSE) or root mean squared error (RMSE). Use the Adam optimizer to update the model weights through backpropagation. Split the data into training and validation sets. Evaluate the performance of the LSTM model through cross-validation or validation sets, and adjust the model hyperparameters.

[0073] The LSTM model will predict stress changes in the future based on the input historical data.

[0074] (2) SVR optimization and integration The output of the LSTM model is used as the input feature of the SVR. After learning the temporal dependencies, LSTM can provide a preliminary prediction result based on historical information, and SVR further optimizes it. SVR effectively captures nonlinear features through kernel methods and selects appropriate parameters to ensure that SVR has good tolerance to noise and obtains better prediction performance in complex stress patterns. The data is divided into a training set and a test set. The SVR model is trained with the training set, and its performance is evaluated with the test set. SVR generates more accurate stress prediction values ​​based on the output of LSTM and its nonlinear optimization.

[0075] (3) Fusion of LSTM and SVR This embodiment forms an LSTM → SVR structure. The LSTM model first generates time series predictions based on the input historical data. The SVR receives the output of the LSTM. The LSTM-SVR model is used to As training input, learn time series rules and predict future stress prediction values The output result is further optimized and predicted by nonlinear regression. For example, the following decision rule can be used to realize the multi-field coupling stress over-limit warning through dynamic threshold: (26) in, Indicates the judgment indicator. 1 and 0 are binary decision results. For example, 1 indicates that the over-limit warning is triggered, and 0 indicates that the over-limit warning is not triggered. is the stress prediction value output by the LSTM-SVR model, As the reference threshold, it can take any one of the standard value, simulation value, statistical value or use more than two fusion values. is the tunnel environmental sensitivity coefficient (can be 0.5~1.0), is the historical stress standard deviation.

[0076] In this embodiment, LSTM is responsible for capturing time dependencies, SVR is responsible for refining nonlinear patterns, and the LSTM and SVR models are fully integrated to obtain more accurate prediction results.

[0077] Fiber-optic sensing technology is used to monitor tunnel stress throughout the entire process. Multi-level processing, including signal demodulation, noise reduction, and optimization, ensures high-precision stress data. Furthermore, a prediction model combining LSTM and SVR fully accounts for the temporal characteristics of tunnel stress signals while effectively handling nonlinear variations. This multi-level signal processing and prediction mechanism enables a solution based on fiber-optic sensing technology to achieve full coverage of the tunnel structure, significantly improving the comprehensiveness and real-time nature of the data. Furthermore, traditional signal processing methods may not be able to effectively separate noise from signal in noisy environments, resulting in large measurement errors. By combining wavelet transform and Kalman filtering, this solution effectively removes noise and optimizes the signal, improving data accuracy. Finally, existing stress prediction technologies are often limited by model capabilities when handling complex and nonlinear stress variations, resulting in inaccurate prediction results. By combining LSTM and SVR, this solution leverages the strengths of both to enhance prediction accuracy and robustness. This significantly improves stress prediction accuracy, particularly in tunnel environments where stress changes are rapid and complex.

[0078] This embodiment further provides a subway tunnel wall stress adaptive monitoring device, including a processor and a memory, the memory is used to store a computer program, and the processor is used to execute the computer program to perform the above method.

[0079] It is understandable that the above method of this embodiment can be executed by a single device, such as a computer or server, etc., and can also be applied to a distributed scenario and completed by multiple devices cooperating with each other. In the case of a distributed scenario, one of the multiple devices can only execute one or more steps in the above method of this embodiment, and multiple devices interact to complete the above method. The processor can be implemented in the form of a general-purpose CPU, a microprocessor, an application-specific integrated circuit, or one or more integrated circuits, etc., for executing relevant programs to implement the above method of this embodiment. The memory can be implemented in the form of a read-only memory ROM, a random access memory RAM, a static storage device, and a dynamic storage device. The memory can store an operating system and other application programs. When the above method of this embodiment is implemented by software or firmware, the relevant program code is stored in the memory and called and executed by the processor.

[0080] This embodiment further provides a computer-readable storage medium storing a computer program, which implements the above method when executed by a processor.

[0081] Those skilled in the art will appreciate that the above-described embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions described in the processes. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0082] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed above with reference to the preferred embodiment, it is not intended to limit the present invention. Therefore, any simple modifications, equivalent variations, and modifications to the above embodiment that do not depart from the technical solution of the present invention and are based on the technical essence of the present invention shall fall within the scope of protection of the technical solution of the present invention.

Claims

1. A subway tunnel wall stress adaptive monitoring method, characterized in that the steps include: Real-time monitoring of vibration signals collected by train vibration sensors and optical signals collected by distributed optical fiber sensors installed on subway tunnel walls, demodulation of the monitored optical signals to obtain stress signals, and extraction of temperature changes and Bragg wavelength changes. Calculation of actual stress under thermal-mechanical coupling based on the extracted temperature changes and Bragg wavelength changes. Extraction of dynamic displacement of the tunnel wall caused by train vibration based on the monitored vibration signals, calculation of dynamic stress based on the dynamic displacement of the tunnel wall, and calculation of actual stress data under thermal-mechanical-vibration multi-field coupling based on the actual stress under thermal-mechanical coupling and the dynamic stress calculation. The stress signal monitored in real time by the distributed optical fiber sensor is subjected to noise reduction processing using wavelet transform, the stress signal after noise reduction processing is used as observation data, the actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the initial stress state data, and Kalman filtering is used to perform prediction to obtain corrected thermal-mechanical-vibration coupled stress data; The corrected thermal-mechanical-vibration coupled stress data is input into a pre-trained stress prediction model to obtain the final stress prediction result output of the subway tunnel wall. The stress prediction model is obtained by training an LSTM-SVM model using a training data set composed of historical stress data. The LSTM-SVM model uses the LSTM model to obtain preliminary stress prediction results. The output of the LSTM model is used as the input of the SVR model, and the SVR model corrects the output of the LSTM model to obtain the final stress prediction result.

2. The subway tunnel wall stress adaptive monitoring method according to claim 1, characterized in that: The actual stress data under the thermal-mechanical-vibration multi-field coupling is obtained by calculating the actual stress under the thermal-mechanical coupling and the dynamic stress, including: The dynamic displacement of the tunnel wall is extracted from the collected vibration signal, and the real-time temperature change and Bragg wavelength change are extracted from the demodulated stress signal; Calculate the actual stress under thermal-mechanical coupling based on the change in Bragg wavelength and real-time temperature change; Calculating dynamic strain based on the real-time dynamic displacement of the tunnel wall, and then calculating dynamic stress caused by train vibration based on the calculated dynamic strain; The actual stress under the thermal-mechanical coupling action and the dynamic stress caused by the train vibration are superimposed to obtain the actual stress under the thermal-mechanical-vibration multi-field coupling action.

3. The subway tunnel wall stress adaptive monitoring method according to claim 2, characterized in that: The calculation expression for calculating the actual stress under the thermal-mechanical coupling is: ， in, represents the actual stress under the thermal-mechanical coupling effect, represents the effective elastic-optical coefficient, represents the effective refractive index, represents the elastic modulus of the optical fiber, represents the total change in Bragg wavelength, represents the Bragg wavelength, represents the thermal expansion coefficient of the optical fiber material, represents the thermo-optical coefficient, Indicates the temperature change.

4. The subway tunnel wall stress adaptive monitoring method according to claim 3, characterized in that: The calculation expression for the dynamic stress caused by train vibration is: ， in, represents the dynamic stress caused by train vibration, represents the elastic modulus of the optical fiber, represents dynamic strain; Actual stress under the coupling of thermal, mechanical and vibration fields The calculation expression is: ， When the train vibration induces the elliptical deformation of the tunnel wall, the dynamic stress caused by the train vibration is the mean tangential stress : ， in, is the tunnel radius, is the train vibration amplitude, is the frequency, is the Poisson's ratio of the material, For time.

5. The subway tunnel wall stress adaptive monitoring method according to claim 1, characterized in that: The method of using wavelet transform to perform noise reduction on stress data monitored in real time by the distributed optical fiber sensor includes: A wavelet basis is selected according to the frequency characteristics and noise distribution of the stress data, and the selected wavelet basis is used to perform multi-layer wavelet decomposition on the stress signal monitored in real time by the distributed optical fiber sensor. In each layer of decomposition, the input stress signal is convolved with the scaling function to obtain low-frequency coefficients, and the input stress signal is convolved with the wavelet basis function to obtain low-frequency coefficients. The next level further decomposes the low-frequency coefficients obtained by the previous level of decomposition, and in each layer of decomposition, a preset dynamic threshold is used to perform denoising on the high-frequency coefficients to remove high-frequency noise. The dynamic threshold is dynamically set according to the parameters of the high-frequency coefficients of each layer. Finally, the low-frequency coefficients obtained by the multi-layer wavelet decomposition and the high-frequency coefficients after denoising are subjected to inverse wavelet transform to obtain the stress data after denoising. The stress data after the primary denoising is subjected to an adaptive filter to remove low-frequency noise, and the weight of the filter is continuously adjusted to adaptively track changes in the low-frequency noise, thereby obtaining stress data after the secondary denoising.

6. The subway tunnel wall stress adaptive monitoring method according to claim 5, characterized in that: The denoising process of using a preset dynamic threshold to remove high-frequency noise on high-frequency coefficients in each layer of decomposition includes: If the absolute value of the effective signal corresponding coefficient exceeding the specified ratio in the current high-frequency coefficient is greater than the dynamic threshold or the absolute value of the noise corresponding coefficient exceeding the specified ratio is less than the dynamic threshold, the hard threshold method is used for processing according to the dynamic threshold, otherwise the soft threshold method is used for processing according to the dynamic threshold. If the hard threshold method is used for processing, hour, , , Indicates the j The high-frequency coefficients after layer wavelet decomposition, Represents the high-frequency coefficient after denoising; If the soft threshold method is used for processing, hour, , When the high frequency coefficients of the current layer are After the reduction process, the high-frequency coefficients after the denoising process are obtained. ; Dynamic Threshold The calculation expression is: ， in, For the The standard deviation of the high-frequency coefficients of the layer, For the The length of the high-frequency coefficients of the layer, is the influencing factor of tunnel vibration noise.

7. The subway tunnel wall stress adaptive monitoring method according to any one of claims 1 to 6, characterized in that: The stress signal after noise reduction is used as observation data, and the actual stress data under the thermal-mechanical-vibration multi-field coupling is used as the stress state data at the initial moment, and prediction is performed using an extended Kalman filter, including: according to To predict the status, To predict the error covariance, Calculate the Kalman gain ,according to To update the status, follow Perform error covariance update, where is the Kalman gain at the kth moment, is the covariance of the stress state after the update at the kth moment, is the covariance of the predicted stress state at the kth moment, is the observation matrix, is the observed value at the kth moment, that is, the stress signal monitored in real time by the distributed optical fiber sensor at the kth moment is the stress signal after noise reduction obtained by wavelet transform. is the observation noise covariance, To predict the stress state at the kth moment, at the initial moment is the actual stress data under the thermal-mechanical-vibration multi-field coupling, is the identity matrix, is the updated stress state estimate at the kth moment, i.e., the corrected thermal-mechanical-vibration coupled stress data, and are the Jacobian matrices respectively.

8. The subway tunnel wall stress adaptive monitoring method according to any one of claims 1 to 6, characterized in that: The corrected thermal-mechanical-vibration coupled stress data is input into a pre-trained stress prediction model to obtain a final stress prediction result output of the subway tunnel wall, including: Inputting the corrected thermal-mechanical-vibration coupled stress data into an LSTM model pre-trained using historical stress data, and the LSTM model obtains a stress prediction result by predicting stress changes at future moments based on the historical stress data; The output of the LSTM model is used as the input feature of the SVR model. The SVR model performs nonlinear optimization on the stress prediction value output by the LSTM model to obtain the optimized prediction result. The output of the LSTM model is fused with the output of the SVR model to obtain the fusion prediction result; The final stress prediction result is obtained according to the optimized prediction result and the fused prediction result.

9. A subway tunnel wall stress adaptive monitoring device, comprising a processor and a memory, wherein the memory is used to store a computer program, characterized in that: The processor is configured to execute the computer program to perform the method according to any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images