A ground stress direction measuring system based on longitude and latitude strain lines

Abstract

The application discloses a kind of ground stress direction measurement systems based on longitude and latitude strain line, belong to ground stress measurement technical field.It includes: two-dimensional orthogonal strain acquisition array module;Timing synchronous control module, for controlling each group of strain sensors based on uniform time reference synchronous sampling;Interference adaptive filtering module;Ground stress tensor solving module;Structure adaptive compensation module;Dynamic weighted fusion module;Risk trend deduction module;Edge computing module and remote communication module.The system adopts longitude and latitude orthogonal arrangement fiber grating sensing array, constructs symmetric strain tensor and carries out eigenvalue decomposition, to obtain principal stress direction angle and principal strain value, compared with single point strain gauge or traditional sleeve hole method, the application can realize continuous principal direction distribution reconstruction in larger range, with higher spatial resolution and tensor integrity.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ground stress measurement, and particularly relates to a ground stress direction measurement system based on meridian and parallel strain lines. BACKGROUND

[0002] Ground stress is a basic physical quantity existing in rock mass, and is an important control factor affecting the stability of underground engineering, rock mass failure mechanism and tectonic evolution. Accurate determination of the size and principal stress direction of the ground stress tensor has key engineering guiding value for deep engineering such as tunneling, deep well mining and energy storage design.

[0003] With the development of distributed optical fiber sensing technology, it has gradually become a trend to use orthogonally arranged meridian and parallel strain fiber arrays for high-precision strain field measurement. This kind of method can obtain two-dimensional strain tensor information, and derive the principal stress direction through tensor principal value calculation. However, most of the existing methods are only applicable to ideal uniform soil conditions, and do not consider the interference of the ubiquitous fracture structure in rock mass, resulting in insufficient precision of the stress principal direction inversion result in complex geological environment.

[0004] On the other hand, most of the existing optical fiber strain measurement systems only average and fuse the strain values of each sensing node, and do not establish a weighted mechanism based on quality indicators such as node signal-to-noise ratio, stability and layout density. The system lacks the ability to identify the difference in data reliability, and is easily affected by local abnormal points in the overall stress tensor calculation process, resulting in systematic deviation.

[0005] Although existing researches have tried to use geological radar (GPR) and other structure perception methods for rock mass structure discrimination, there is still a lack of fusion path for introducing GPR fracture model data into stress direction correction in the existing ground stress measurement system, that is, the structure heterogeneity cannot be dynamically corrected and compensated in the strain tensor calculation. SUMMARY

[0006] The application scheme is as follows:

[0007] A ground stress direction measurement system based on meridian and parallel strain lines, comprising:

[0008] A two-dimensional orthogonal strain acquisition array module, which is composed of a meridian line strain sensor group arranged along the X-axis direction and a parallel line strain sensor group arranged along the Y-axis direction, and each group is composed of at least two sensor devices at different depth levels, forming a three-dimensional sampling grid;

[0009] A time sequence synchronization control module for controlling each group of strain sensors to perform synchronous sampling based on a unified time reference;

[0010] An interference adaptive filtering module, including a wavelet packet decomposition unit, an interference identification submodule and a filtering reconstruction submodule;

[0011] A ground stress tensor solving module is configured to construct a two-dimensional strain tensor matrix based on the latitude and longitude strain values and solve a principal direction angle thereof;

[0012] A structure adaptive compensation module is configured to construct an anisotropic correction tensor using a crack distribution model obtained by the external GPR and apply the anisotropic correction tensor to the original stress direction result;

[0013] A dynamic weighted fusion module is configured to construct a weight vector based on the signal-to-noise ratio, time stability and spatial distribution density of each measuring point and apply the weight vector to the tensor calculation;

[0014] A risk trend deduction module is configured to generate a stress concentration area risk evolution sequence based on a ground stress direction mutation rate;

[0015] An edge computing module and a remote communication module are integrated in a field node box and have data preprocessing, local direction calculation and 5G transmission capabilities.

[0016] Preferably, the latitude strain sensor and the longitude strain sensor are both fiber grating strain gauges, specifically including a Bragg grating area, a mirror surface packaging structure and a flexible mechanical coupling sheet, each strain gauge is transmitted on a single optical fiber in a wavelength division multiplexing manner, and is uniformly arranged on the monitoring surface at an interval of 0.5 meters, each sensor is embedded into the rock surface layer at a depth of 0.8 meters through a depth-fixed drill hole, the drill hole depth tolerance is not more than ±0.1 meters, and a flexible filling fixing agent is used to ensure that the fiber grating sensing head is tightly coupled with the rock mass; the reflection wavelength range of each group of strain gauges is set to 1525-1565 nm, the center wavelength drift sensitivity is greater than 1.1 pm / με, and a double-layer flexible silicone seal layer and a PTFE anti-seepage film layer are arranged outside the optical cable sheath to ensure long-term stable operation in a geothermal environment of -20°C to 80°C.

[0017] Preferably, the interference adaptive filtering module adopts a multi-stage combined filtering structure, including:

[0018] A wavelet packet decomposition unit performs 5-layer full-band signal decomposition based on a Daubechies-8 order wavelet function, and the signal processing frame length is 8192 sampling points;

[0019] An interference identification submodule is internally provided with a feature library including at least 12 types of spectrum templates of seismic disturbance, mechanical disturbance and electromagnetic interference, and uses a sparse matching pursuit combined with Hilbert amplitude analysis to automatically screen abnormal components;

[0020] A filtering and reconstruction submodule uses a Kalman prediction-wavelet packet reconstruction fusion algorithm to compensate and backfill the identified effective strain signals, and superimposes an adaptive Savitzky-Golay window on the reconstruction result to perform second-order smoothing on the edges of the step-type signals.

[0021] Preferably, the in-situ stress tensor solving module is based on the fused strain tensor ε w Constructing the two-dimensional symmetric strain tensor:

[0022]

[0023] Wherein: ε xx , ε yy respectively represent the average linear strain values along the latitudinal and longitudinal directions; ε xy is the shear strain component, which is calculated by the mutual inductance sensor pairs within a distance of 0.2m in orthogonal directions:

[0024]

[0025] The principal stress direction angle θ raw and the principal strain value ε 1,2 are solved by the eigenvalue and eigenvector of the tensor respectively, and the formula is:

[0026]

[0027] Preferably, the structure adaptive compensation module is used for fracture structure coupling correction of the principal stress direction angle θ raw , and the fracture distribution model is acquired by a system integrated ground penetrating radar (GPR) module, the GPR module comprising: a dual-polarized antenna with a working frequency band of 400-800MHz; a walking drag platform or a wall-mounted scanning arm; after the radar scanning data is processed by background removal, gain compensation and Kirchhoff inversion, a two-dimensional B-scan profile is formed.

[0028] The image is subjected to edge detection and Hough transform algorithm to extract the main fracture strike angle φ i and the density D i ; the structure compensation angle Δθ is generated by the following formula:

[0029] The corrected principal stress direction angle θ corr is calculated by the formula: θ corr = θ raw + Δθ.

[0030] Preferably, the generation process of the fused strain tensor ε w is completed by a dynamic weighted fusion module, which synthesizes the strain tensors ε i collected by N distributed sensing nodes in the following weighted manner:

[0031] Wherein the weight coefficient w i is constructed according to the node quality factor, specifically:

[0032]

[0033] wherein: SNR i represents the signal-to-noise ratio of the i-th node; STAB i represents the inverse of the standard deviation of the strain of the node in the last 5 minutes; DENS i represents the average layout density in the sub-region where the node is located; and α, β, and γ are empirical weight coefficients, satisfying α + β + γ = 1.

[0034] Preferably, the risk trend deduction module uses the principal stress direction angle θ corr modified by the structural adaptive compensation module as the basis, in combination with the principal strain value sequence in the fused strain tensor ε w output by the dynamic weighted fusion module, to construct a ground stress mutation rate monitoring model, which includes the following processing steps:

[0035] Step one, time series difference is performed on the continuous principal direction angle θ corr (t) in a 1-minute sliding window to calculate the direction mutation rate: Δθ(t) = θ corr (t) - θ corr (t-Δt);

[0036] Step two, the principal strain eigenvalues ε1(t), ε2(t) are extracted synchronously, and a two-dimensional risk feature vector is constructed:

[0037] Step three, the feature vector is input into a support vector machine (SVM) classification model, compared with the training set samples, to determine whether the current region is in a low, medium, or high risk state;

[0038] Step four, the risk level results in the continuous time period are recorded as a risk evolution sequence R(t), and based on the spatial coordinate correspondence, a stress concentration trend map and a warning signal of the current monitoring region are generated;

[0039] The feature vector dimension, sliding window size, and classification threshold of the SVM model can be set by the user or adjusted adaptively by the system.

[0040] Preferably, the edge computing module is constructed using an NVIDIA Jetson Orin Nano platform, which has a direction solving module, a filtering module, and a risk model evaluation module embedded inside, and is configured with a LoRa wireless communication module and a 5G module in a dual-channel configuration, which automatically switches the communication channel when the signal attenuation is greater than 30 dB, ensuring data link redundancy.

[0041] The system is preferably used in a fault zone crossing tunnel, adopts a long axis control network formed by arranging a group of multi-layer orthogonal strain arrays every 20 meters, and forms a main stress change sequence along the axis direction of the tunnel by continuously solving the tensor direction field, so as to determine the specific positions of weak zones, tensile concentrated zones and potential deformation and dislocation zones.

[0042] Compared with the prior art, the application has the advantages that:

[0043] (1) The system adopts fiber grating sensing arrays arranged orthogically along the meridian and the parallel, constructs a symmetric strain tensor and performs eigenvalue decomposition, so as to obtain the principal stress direction angle and the principal strain value. Compared with a single point strain gauge or a traditional hole method, the application can realize continuous principal direction distribution reconstruction in a larger range, has higher spatial resolution and tensor integrity.

[0044] (2) The system quantifies the signal-to-noise ratio, time stability, arrangement density and other parameters of each strain collection point, constructs a fusion weight vector, and realizes multi-point weighted average of the initial strain tensor. Compared with the traditional average method or the filtering method, the application improves the reliability and anti-exceptional interference ability of the tensor solution, and significantly reduces the stress direction misjudgment rate.

[0045] (3) The system has a built-in geological radar (GPR) module, obtains the spatial distribution, trend and density information of the cracks in the monitoring area, and compensates and corrects the original stress principal direction angle through an anisotropic structure correction tensor. This method solves the problem that the existing measurement system cannot perceive the heterogeneity of the medium structure, and has higher principal direction identification accuracy in a complex rock mass environment. BRIEF DESCRIPTION OF DRAWINGS

[0046] Figure 1 It is a block diagram of the ground stress direction measurement system based on the meridian and parallel strain lines. DETAILED DESCRIPTION

[0047] The technical solutions of the embodiments of the application will be explained and described below. The following embodiments are preferred embodiments of the application, but not all. Based on the embodiments in the embodiments, other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0048] Embodiment 1

[0049] A ground stress direction measurement system based on meridian and parallel strain lines comprises:

[0050] A two-dimensional orthogonal strain collection array module is composed of a parallel strain sensor group arranged along the X-axis direction and a meridian strain sensor group arranged along the Y-axis direction, and each group is composed of at least two or more sensor devices at different depth levels, forming a three-dimensional sampling grid.

[0051] a time synchronization control module for controlling each group of strain sensors to perform synchronous sampling based on a unified time reference;

[0052] an interference adaptive filtering module including a wavelet packet decomposition unit, an interference identification submodule, and a filtering reconstruction submodule;

[0053] a ground stress tensor solving module for constructing a two-dimensional strain tensor matrix based on the latitude and longitude strain values and solving the principal direction angle thereof;

[0054] a structure adaptive compensation module for constructing an anisotropic correction tensor using a fracture distribution model obtained by the external GPR and applying the anisotropic correction tensor to the original stress direction result;

[0055] a dynamic weighted fusion module for constructing a weight vector based on the signal-to-noise ratio, time stability, and spatial distribution density of each measuring point and applying the weight vector to the tensor calculation;

[0056] a risk trend deduction module for generating a stress concentration area risk evolution sequence based on the ground stress direction mutation rate;

[0057] an edge computing module and a remote communication module integrated in the field node box and having data preprocessing, local direction calculation, and 5G transmission capabilities.

[0058] The latitude strain sensor and the longitude strain sensor are both fiber Bragg grating strain gauges, specifically including a Bragg grating region, a reflective mirror packaging structure, and a flexible mechanical coupling sheet. Each strain gauge is transmitted on a single optical fiber in a wavelength division multiplexing manner and is uniformly arranged on the monitoring surface at an interval of every 0.5 meters. Each sensor is embedded into the rock surface at a depth of 0.8 meters through a depth-fixed drill hole. The drill hole depth tolerance is not more than ±0.1 meters, and a flexible filling fixative is used to ensure that the fiber Bragg grating sensing head is tightly coupled with the rock mass. The reflective wavelength range of each group of strain gauges is set to 1525-1565 nm, the center wavelength drift sensitivity is greater than 1.1 pm / με, and a double-layer flexible silicone seal layer and a PTFE impermeable membrane layer are provided outside the optical cable sheath to ensure long-term stable operation in a geothermal environment of -20°C to 80°C.

[0059] The interference adaptive filtering module adopts a multi-stage combined filtering structure, including:

[0060] The wavelet packet decomposition unit performs 5-layer full-band signal decomposition based on a Daubechies-8 order wavelet function, and the signal processing frame length is 8192 sampling points.

[0061] The interference identification submodule has a feature library including at least 12 types of seismic disturbance, mechanical disturbance, and electromagnetic interference spectrum templates, and uses a combination of sparse matching pursuit and Hilbert amplitude analysis to automatically filter abnormal components.

[0062] The filter reconstruction sub-module adopts a Kalman prediction-wavelet packet reconstruction fusion algorithm to compensate and backfill the effective strain signals identified, and superimposes an adaptive Savitzky-Golay window on the reconstruction result to perform second-order smoothing on the edges of the step-type signals.

[0063] The ground stress tensor solving module is based on the fusion strain tensor ε w The two-dimensional symmetric strain tensor is constructed as follows:

[0064]

[0065] Wherein: ε xx , ε yy respectively represent the average linear strain values along the parallel and meridian directions; ε xy is the shear strain component, which is calculated by mutual inductance sensors within a distance of 0.2m in orthogonal directions:

[0066]

[0067] The principal stress direction angle θ raw and the principal strain value ε 1,2 are solved by the eigenvalues and eigenvectors of the tensor respectively, and the formula is:

[0068]

[0069] The structure adaptive compensation module is used to correct the principal stress direction angle θ raw , and the crack distribution model is obtained by the system-integrated ground penetrating radar (GPR) module, which includes: a dual-polarized antenna with a working frequency band of 400-800MHz; a walking drag platform or a wall-mounted scanning arm; after the radar scanning data is processed by background removal, gain compensation and Kirchhoff inversion, a two-dimensional B-scan profile is formed;

[0070] The image is subjected to edge detection and Hough transform algorithm to extract the main crack strike angle φ i and the density D i ; the structure compensation angle Δθ is generated by the following formula:

[0071] The corrected principal stress direction angle θ corr is calculated by the formula: θ corr = θ raw + Δθ.

[0072] The generation process of the fusion strain tensor ε w is completed by the dynamic weighted fusion module, which is composed of strain tensors εi The following weighting method is used for synthesis:

[0073] wherein the weight coefficient w i According to the node quality factor construction, specifically:

[0074]

[0075] wherein: SNR i represents the signal-to-noise ratio of the i-th node; STAB i represents the inverse of the standard deviation of the strain of the node in the last 5 minutes; DENS i represents the average layout density in the sub-region where the node is located; α, β, γ are empirical weight coefficients, satisfying α+β+γ=1.

[0076] wherein the risk trend deduction module takes the principal stress direction angle θ corr as the basis, and combines the principal strain value sequence in the fused strain tensor ε w output by the dynamic weighted fusion module to construct a ground stress mutation rate monitoring model, which includes the following processing steps:

[0077] Step one, the continuous principal direction angle θ corr (t) is subjected to time series difference according to a 1-minute sliding window, and the direction mutation rate is calculated: Δθ(t)=θ corr (t)-θ corr (t-Δt);

[0078] Step two, the principal strain eigenvalues ε1(t), ε2(t) are extracted synchronously, and a two-dimensional risk feature vector is constructed:

[0079] Step three, the feature vector is input into a support vector machine (SVM) classification model, compared with the training set samples, and it is determined whether the current area is in a low, medium or high risk state;

[0080] Step four, the risk level results in the continuous time period are recorded as a risk evolution sequence R(t), and based on the spatial coordinate correspondence, a stress concentration trend map and a warning signal of the current monitoring area are generated;

[0081] The feature vector dimension, sliding window size and classification threshold of the SVM model can be set by the user or adjusted adaptively by the system.

[0082] The edge computing module is constructed by using an NVIDIA Jetson Orin Nano platform, which internally embeds a direction solving module, a filtering module and a risk model evaluation module, and is internally configured with a LoRa wireless communication module and a 5G module dual channel, which automatically switches the communication channel when the signal attenuation is greater than 30 dB, to ensure data link redundancy.

[0083] The system is used in a fault zone crossing tunnel, and a long axis control network is formed by arranging a set of multi-layer orthogonal strain arrays every 20 meters, and a main stress change sequence along the tunnel axis direction is formed by continuous solving of the tensor direction field, which is used to determine the specific positions of the weak zone, the tensile concentrated zone and the potential deformation and dislocation zone.

[0084] Embodiment 2

[0085] The embodiment provides a ground stress direction measurement system based on longitude and latitude strain lines, which is suitable for initial stress detection and risk zoning identification of a certain underground tunnel excavation engineering, and the installation area is a rock mass monitoring surface in a range of 20 m in front of the tunnel face.

[0086] The system is composed of the following modules:

[0087] (1) Two-dimensional orthogonal strain acquisition array arrangement

[0088] A two-dimensional orthogonal acquisition array composed of fiber Bragg strain sensors is arranged on the surface of the rock mass in the monitoring area, with a spacing of 0.25 m x 0.25 m, a longitude arrangement length of 10 m and a latitude arrangement length of 8 m, and a total of 128 sensing nodes. Each node can collect axial strain and 45° shear strain, with a sampling frequency of 1 Hz and an error of not more than ±5με.

[0089] (2) Original strain data processing and dynamic weighted fusion

[0090] The collected original strain value After low-pass filtering and denoising, the dynamic weighted fusion module is input. According to the signal-to-noise ratio (SNR), strain change stability (STAB) and regional arrangement density (DENS) of each node in the last 5 minutes, the weighted factor w is constructed i , the tensor contribution value of each node is fused:

[0091]

[0092] The fused two-dimensional strain tensor ε is obtained w , the principal strain values ε1, ε2 are solved from the eigenvalues, and the preliminary principal stress direction angle θ is solved from the eigenvectors raw .

[0093] (3) Fracture distribution model acquisition and structural compensation tensor generation

[0094] In the monitoring area, GPR scanning vehicles are deployed synchronously, and 800 MHz dual-polarized antennas are used for walk-by inversion imaging to extract the reflection characteristics of the main fissure surface within a depth range of 5 m underground. After edge detection and Hough transform processing, the strike angle distribution φ i and the corresponding fissure density D i are extracted. The above data are input into the structure compensation module to construct the fissure compensation tensor and calculate the main direction correction angle:

[0095]

[0096] Thus, the corrected principal stress direction angle θ corr = θ raw + Δθ is obtained.

[0097] (4) Risk trend deduction and concentration zone identification

[0098] The time series sliding difference is performed on the sequence of corrected direction angles θ corr (t), the mutation rate Δθ(t) is calculated, the first derivative of ε1(t) is calculated synchronously, and the risk feature vector is constructed:

[0099]

[0100] and input into the risk classification model constructed based on support vector machine (SVM), and the risk level label (low, medium, and high) is output, the stress concentration risk area is marked, and the evolution sequence is formed

[0101] (5) Result output and visualization

[0102] The system forms a risk level heat map, a direction field vector diagram, and a stress mutation trend diagram on the server side, and uploads them to a remote platform for reference by engineering technicians. If the risk level determination within a continuous time window is "high", the system will trigger an audible and light alarm and push an alarm to a mobile terminal.

[0103] The system deployment cycle shown in the embodiment is not more than 24 hours, and the continuous 7×24 hour monitoring of key areas can be realized. The system has been applied and verified in multiple rock heterogeneity zones and fracture zone crossing sections, with an identification accuracy of better than ±8° and a direction mutation detection sensitivity of 0.5° / min, and can be effectively used for surrounding rock support adjustment and early prediction of high-risk excavation sections.

[0104] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Various changes and improvements can be made to the present application without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A ground stress direction measuring system based on longitude and latitude strain lines, characterized by, The application relates to a self-adaptive and intelligent in-situ stress field monitoring system. The application comprises: a two-dimensional orthogonal strain acquisition array module, which is composed of a weft strain sensor group arranged along an X-axis direction and a warp strain sensor group arranged along a Y-axis direction, and each group is composed of at least two sensor devices of different depth levels, forming a three-dimensional sampling grid; a time sequence synchronization control module for controlling each group of strain sensors to carry out synchronous sampling based on a unified time reference; an interference adaptive filtering module, which comprises a wavelet packet decomposition unit, an interference identification sub-module and a filtering reconstruction sub-module; a ground stress tensor solving module for constructing a two-dimensional strain tensor matrix based on warp and weft strain values and solving a principal direction angle thereof; a structure adaptive compensation module for constructing an anisotropic correction tensor by using a crack distribution model obtained by external GPR and applying the anisotropic correction tensor to an original stress direction result; a dynamic weighted fusion module for constructing a weight vector based on signal-to-noise ratios, time stability and spatial distribution density of each measuring point and applying the weight vector to tensor calculation; a risk trend deduction module for generating a stress concentration area risk evolution sequence based on a ground stress direction mutation rate; The ground stress tensor solving module is based on fusion strain tensor Constructing two-dimensional symmetric strain tensor: ； wherein: , respectively represent the average linear strain values along the weft and warp directions; is the shear strain component, calculated from the mutual inductance sensor pairs in orthogonal directions with a distance of no more than 0.2 m. ； principal stress direction angle principal strain value The principal stress direction angle and the principal strain value are solved respectively by the eigenvalue and the eigenvector of the tensor, and the formula is: ； ； A structure adaptive compensation module is used to compensate the main stress direction angle The fracture distribution model is obtained by a system-integrated ground penetrating radar (GPR) module, which comprises a dual-polarized antenna with a working frequency band of 400-800 MHz, a walking-dragging platform or a wall-hung scanning arm, and after background removal, gain compensation and Kirchhoff inversion processing of radar scanning data, a two-dimensional B-scan profile is formed. The image is edge detected and the Hough transform algorithm is used to extract the main fissure strike angle With density ; structure compensation angle Generated by the following formula: ; corrected principal stress direction angle The calculation formula is: .

2. The ground stress direction measurement system based on latitude and longitude strain lines according to claim 1, characterized in that, an edge computing module and a remote communication module, which are integrated in a field node box and have data preprocessing, local direction solving and 5G transmission capabilities; 3. The ground stress direction measurement system based on latitude and longitude strain lines according to claim 1, characterized in that, the weft strain sensor and the warp strain sensor are both fiber Bragg grating strain gauges, specifically comprising a Bragg grating area, a reflective mirror surface packaging structure and a flexible mechanical coupling sheet, the fiber Bragg grating strain gauges are connected in series on a single optical fiber through wavelength division multiplexing transmission, and are uniformly arranged on a monitoring surface at an interval of 0.5 meters; each sensor is embedded into a rock surface layer with a depth of 0.8 meters through a depth-fixed drill hole, the drill hole depth tolerance is not more than + / - 0.1 meters, and a flexible filling fixing agent is used to ensure that the fiber Bragg grating sensing head is tightly coupled with the rock mass; the reflective wavelength range of each group of strain gauges is set to 1525-1565 nm, the center wavelength drift sensitivity is greater than 1.1 pm / mu, and a double-layer flexible silicone seal layer and a PTFE anti-seepage film layer are arranged outside the optical cable sheath to ensure long-term stable operation in a ground temperature environment of-20 DEG C to 80 DEG C. The interference adaptive filtering module adopts a multi-stage combined filtering structure, which comprises: a wavelet packet decomposition unit, which carries out 5-layer full-band signal decomposition based on a Daubechies-8 order wavelet function, and the signal processing frame length is 8192 sampling points; an interference identification sub-module, which is internally provided with a feature library comprising at least 12 types of spectrum templates of seismic disturbance, mechanical disturbance and electromagnetic interference, and uses a sparse matching pursuit combined with Hilbert amplitude analysis to automatically screen abnormal components; 4. The ground stress direction measurement system based on latitude and longitude strain lines according to claim 1, characterized in that, Fusion strain tensor The generation process is completed by a dynamic weighted fusion module, which is composed of strain tensors collected by distributed sensor nodes are synthesized in the following weighted manner: ; where the weight coefficient According to the node quality factor construction, specifically: ； wherein: represents the signal-to-noise ratio of the node; represents the inverse of the standard deviation of the strain of the node in the last 5 minutes; represents the average density of the node in the sub-region where the node is located; , , is an empirical weight coefficient, satisfying .​ 5. The ground stress direction measurement system based on latitude and longitude strain lines according to claim 1, characterized in that, The risk trend deduction module corrects the principal stress direction angle obtained by the structural self-adaptive compensation module For the basis, the dynamic weighted fusion module outputs the fusion strain tensor The principal strain value sequence in the fusion strain tensor, and constructs a ground stress mutation rate monitoring model, which includes the following processing steps: Step one, the continuous principal direction angle Time series difference with 1 minute sliding window, calculate the direction mutation rate: ; Step two, synchronously extract principal strain eigenvalues , , and construct a two-dimensional risk feature vector: ; a filtering reconstruction sub-module, which uses a Kalman prediction-wavelet packet reconstruction fusion algorithm to compensate and backfill the effective strain signals identified, and superimposes an adaptive Savitzky-Golay window on the reconstruction result to perform second-order smoothing on the edges of step-type signals. Step three: inputting the feature vector into a support vector machine (SVM) classification model, comparing with training set samples, and judging whether the current area is in a low, medium or high risk state; Step four, record the risk level results in a continuous time period as a risk evolution sequence R(t), and generate a stress concentration trend map and early warning signal for the current monitoring area based on the spatial coordinate correspondence; The feature vector dimension, sliding window size and classification threshold of the SVM model can be set by the user or adaptively adjusted by the system.

6. The ground stress direction measurement system based on latitude and longitude strain lines according to claim 1, characterized in that, The edge computing module is built with an NVIDIA Jetson Orin Nano platform, which internally embeds a direction solving module, a filtering module and a risk model evaluation module, and is configured with a LoRa wireless communication module and a 5G module dual channel, which automatically switches the communication channel when the signal attenuation is greater than 30dB, ensuring data link redundancy.

7. The ground stress direction measurement system based on latitude and longitude strain lines according to claim 1, characterized in that, The system is used in a fault zone crossing tunnel, and a long axis control network is formed by arranging a group of multi-layer orthogonal strain arrays every 20 meters, and a principal stress change sequence along the axis direction of the tunnel is formed by continuous tensor direction field solving, which is used to determine the specific position of the soft zone, the tensile concentrated zone and the potential deformation and dislocation zone.

Images