Railway tunnel portal slope instability risk assessment method and system based on image recognition

Abstract

The invention provides a railway tunnel portal slope instability risk assessment method and system based on image recognition, and relates to the technical field of computers, and the method comprises the steps: obtaining video image data of a target railway tunnel portal slope region, and synchronously obtaining waveform data; constructing a slope surface deformation field according to the video image data to obtain a dynamic differential displacement field; performing tunnel axial dynamic load propagation simulation according to the waveform data, and performing inversion to obtain a dynamic stress tensor field; performing slope-tunnel structure dynamic interaction analysis according to the dynamic differential displacement field and the dynamic stress tensor field to obtain a time-varying stability coefficient sequence; carrying out instability precursor identification according to the time-varying stability coefficient sequence to obtain a slope instability mode; and according to the slope instability mode, carrying out operation risk quantification, and outputting a risk assessment result with out-of-plan driving interruption time as a measure. According to the method, accurate and dynamic evaluation of the tunnel portal slope instability risk from a mechanical mechanism to an operation consequence under the cyclic dynamic load of the train is realized.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a method and system for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition. Background Technology

[0002] In the field of railway transportation safety, monitoring and risk assessment of tunnel entrance slope stability has always been a crucial link in ensuring the safe operation of the line. Especially in mountainous railways with complex terrain, the tunnel entrance, as a transitional zone between the slope and the artificial structure, is affected by the combined influence of train cyclic dynamic loads, geological conditions, and climatic factors, exhibiting significant dynamic time-varying characteristics and structural coupling. Traditional assessment methods rely on periodic manual inspections or isolated sensor data, which are difficult to effectively capture the continuous evolution of slope surface deformation and internal stress state in the space and time, and even more difficult to quantify the potential threat to long-term stability posed by residual deformation accumulation during train passage intervals. Although existing technologies attempt to assess stability through simplified mechanical models or single data sources, they often fail to fully consider the dynamic interaction of the slope-tunnel system and lack a precise description of the dynamic evolution process of instability precursors, resulting in limitations in the timeliness and accuracy of risk assessment results, making it difficult to provide sufficient basis for preventive maintenance decisions.

[0003] Based on the shortcomings of the existing technologies, there is an urgent need for a method and system for assessing the risk of slope instability at railway tunnel entrances based on image recognition. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for assessing the risk of slope instability at railway tunnel entrances based on image recognition, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0005] Firstly, this application provides a method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition, including:

[0006] The video image data of the slope area at the entrance of the target railway tunnel is acquired. The video image data includes visual information of the slope surface, the tunnel entrance lining structure and the connecting area. Simultaneously, waveform data of the train passing through is collected by sensors installed on the sleepers inside the tunnel entrance.

[0007] Based on the video image data, the deformation field of the slope surface is constructed. By introducing the geometric constraints of the tunnel entrance terrain as the manifold boundary condition, the spatiotemporal characteristic manifold of the image is constrainedly expanded and its curvature analyzed to obtain the dynamic differential displacement field of the slope surface.

[0008] Based on the waveform data, the axial dynamic load propagation of the tunnel is simulated. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion.

[0009] Based on the dynamic differential displacement field and the dynamic stress tensor field, the dynamic interaction between the slope and the tunnel structure is analyzed to obtain a time-varying stability coefficient sequence. The time-varying stability coefficient sequence is the sequence of the ratio of the anti-sliding virtual work of the potential sliding body of the slope to the load virtual work as time changes.

[0010] Based on the time-varying stability coefficient sequence, instability precursors are identified. By monitoring the residual cumulative effect and its attenuation characteristics during the train passing interval, and determining whether the convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, the slope instability mode is obtained.

[0011] Operational risks are quantified based on the slope instability mode, and risk assessment results are output with unplanned traffic interruption time as the metric.

[0012] Secondly, this application also provides an image recognition-based system for assessing the risk of slope instability at railway tunnel entrances, including:

[0013] The acquisition module is used to acquire video image data of the slope area at the entrance of the target railway tunnel. The video image data includes visual information of the slope surface, tunnel entrance lining structure and connecting area, and simultaneously acquires waveform data of the train passing through collected by sensors installed on the sleepers inside the tunnel entrance.

[0014] The construction module is used to construct the slope surface deformation field based on the video image data. By introducing the geometric constraints of the tunnel entrance terrain as the manifold boundary condition, the spatiotemporal characteristic manifold of the image is constrainedly expanded and its curvature analyzed to obtain the dynamic differential displacement field of the slope surface.

[0015] The simulation module is used to simulate the propagation of axial dynamic load in the tunnel based on the waveform data. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion.

[0016] The analysis module is used to perform dynamic interaction analysis of the slope-tunnel structure based on the dynamic differential displacement field and the dynamic stress tensor field, and obtain a time-varying stability coefficient sequence. The time-varying stability coefficient sequence is a sequence of the ratio of the anti-sliding virtual work of the potential sliding body of the slope to the load virtual work as time changes.

[0017] The identification module is used to identify instability precursors based on the time-varying stability coefficient sequence. By monitoring the residual cumulative effect and its decay characteristics during the train passing interval, and determining whether the convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, the slope instability mode is obtained.

[0018] The assessment module is used to quantify operational risks based on the slope instability mode and output risk assessment results measured by unplanned traffic interruption time.

[0019] The beneficial effects of this invention are as follows:

[0020] This invention achieves a precise and dynamic assessment of the slope instability risk at the tunnel entrance under cyclic dynamic loads from the mechanical mechanism to the operational consequences by simultaneously acquiring tunnel entrance images and sleeper vibration waveforms, and then sequentially performing the following steps: constructing the slope surface deformation field based on geometric constraints, inverting the dynamic stress field of the tunnel surrounding rock based on the waveguide effect, analyzing the time-varying stability under the dynamic interaction of the slope and tunnel structure, identifying instability precursors based on the residual cumulative effect, and quantifying operational risks. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating a method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition, as described in an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the structure of a railway tunnel entrance slope instability risk assessment system based on image recognition, as described in an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram of the structure of a railway tunnel entrance slope instability risk assessment device based on image recognition, as described in an embodiment of the present invention.

[0025] The diagram is labeled as follows: 800, a railway tunnel entrance slope instability risk assessment device based on image recognition; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component; 901, acquisition module; 902, construction module; 903, simulation module; 904, analysis module; 905, recognition module; 906, evaluation module. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0027] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0028] The railway tunnel entrance is one of the most geologically complex and structurally stress-prone sections on mountain railway lines. The slope and tunnel lining structure are rigidly connected, forming a mutually constrained and dynamically coupled mechanical system. When a train passes at high speed, the vibration load generated by the wheel-rail interaction is transmitted to the tunnel lining through the rails, sleepers, and ballast, and further diffuses to the tunnel entrance slope via the lining-surrounding rock interface. This process is not a simple static loading, but a dynamic process involving multiple physical effects: vibration waves disperse and attenuate as they propagate through the layered medium (rail-sleeper-ballast-lining); the tunnel's tubular structure produces unique waveguide and focusing effects on vibration energy, making the tunnel entrance lining and connected slopes "sensitive areas" and "amplifiers" of dynamic response; more importantly, the stress fluctuations generated by the cyclical train load in the slope undergo complex dynamic coupling with the deformation field of the slope itself due to rainfall, weathering, and other factors. This coupling effect significantly alters the stress state of potential slip surfaces and may trigger nonlinear positive feedback between deformation and stress, resulting in strong time-varying stability of the slope. Traditional methods based on periodic inspections or static parameter monitoring struggle to capture the progressive damage and instability precursors induced by transient dynamic excitations and accumulating at structural interfaces. Conventional numerical simulation methods, on the other hand, struggle to accurately acquire train dynamic load inputs, complex terrain geometric constraints, and dynamic coupling parameters between the structure and the soil, leading to discrepancies between assessment results and actual risks. Particularly in predicting the operational consequences of these risks, there is a lack of quantitative assessment methods that combine mechanical instability modes with railway operation windows and maintenance efficiency. Therefore, there is an urgent need for a technical solution that can integrate multi-source data, reflect the unique geometric and dynamic boundary conditions at tunnel entrances, and map the mechanical state to operational risks in real time.

[0029] Example 1:

[0030] This embodiment provides a method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition.

[0031] See Figure 1 The figure shows that the method includes steps S100 to S600.

[0032] Step S100: Obtain video image data of the slope area at the entrance of the target railway tunnel. The video image data includes visual information of the slope surface, tunnel entrance lining structure and connecting area. Simultaneously, obtain waveform data of the train passing through, collected by sensors installed on the inner side sleepers of the tunnel entrance.

[0033] Understandably, the video image data is collected by multiple surveillance cameras deployed on the sides and in front of the tunnel entrance, fully covering the slope surface, the concrete or masonry lining structure at the tunnel entrance, and the funnel-shaped transition area connecting the two. This provides a continuous sequence of visual information containing topographic and structural features for subsequent analysis. Simultaneously acquired waveform data is obtained by a triaxial accelerometer fixed to a specific sleeper inside the tunnel entrance. This directly records the timing of the original vibration acceleration excited by the wheel-rail interaction and transmitted through the sleeper as the train passes. The timestamps are precisely aligned with the video frame acquisition times to establish a spatiotemporal correlation between dynamic excitation and structural response.

[0034] Step S200: Construct the slope surface deformation field based on video image data. By introducing the geometric constraints of the tunnel entrance terrain as manifold boundary conditions, perform constrained expansion and curvature analysis on the spatiotemporal characteristic manifold of the image to obtain the dynamic differential displacement field of the slope surface.

[0035] It should be noted that this step combines the manifold formed by the continuous image sequence in four-dimensional spatiotemporal space with the rigid geometric constraint formed by the tunnel lining contour, which can be considered as undeformable during the analysis period. By performing Riemannian geometric unfolding and curvature analysis on this constrained manifold, the pixel-level apparent motion can be transformed into a real displacement vector field of slope surface points relative to a fixed reference point in the tunnel. This process is equivalent to mathematically removing the nonlinear interference caused by camera perspective and terrain undulations, thereby extracting the dynamic differential displacement field purely caused by the deformation of the soil and rock mass, providing a basis for quantifying the strain distribution on the slope surface.

[0036] Step S300: Simulate the propagation of axial dynamic load in the tunnel based on waveform data. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion.

[0037] Understandably, the goal of step S300 is to transform the point-like sleeper vibration measurement into a complete dynamic stress state within the surrounding rock and soil mass of the tunnel entrance. The principle is that when the dynamic load of a train propagates in the composite medium composed of rails, sleepers, ballast, and lining, different frequency components have different propagation velocities and attenuation characteristics. By analyzing the dispersion relationships inherent in the waveform data and combining the waveguide and mode selection effects of the tunnel's tubular structure on elastic waves, the propagation path and distribution of vibration energy in the medium can be deduced in reverse. Finally, by solving the wave equations that satisfy these physical constraints, the dynamic stress tensor field at each point within the surrounding rock at the tunnel entrance can be derived, thus revealing the spatial distribution and temporal evolution of the dynamic load in the rock and soil mass, rather than simply knowing the force at the contact point.

[0038] Step S400: Perform dynamic interaction analysis of slope-tunnel structure based on dynamic differential displacement field and dynamic stress tensor field to obtain time-varying stability coefficient sequence. The time-varying stability coefficient sequence is the ratio sequence of the anti-sliding virtual work of potential slope sliding body to the load virtual work as time changes.

[0039] Step S400 places the dynamic differential displacement field, which characterizes the slope deformation trend, and the dynamic stress tensor field, which characterizes the external load, under a unified energy framework for analysis. By constructing a virtual displacement model that conforms to the potential slip surface morphology, and calculating the virtual work done by the internal anti-slip force and the virtual work done by the external dynamic stress under this virtual displacement, the ratio of the two reflects the slope's safety reserve against sliding in real time. This time-varying stability coefficient sequence directly characterizes the dynamic process of slope safety fluctuating and evolving over time under cyclic train dynamic load, essentially extending the static limit equilibrium concept to dynamic time history analysis.

[0040] Step S500: Identify instability precursors based on the time-varying stability coefficient sequence. By monitoring the residual cumulative effect and its attenuation characteristics during the train passing interval, and determining whether the convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, the slope instability mode is obtained.

[0041] It should be noted that after a train passes, the deformation and stress of the slope do not immediately return to their initial state; residual deformation or stress remains. This residual effect may gradually accumulate under continuous cyclic loading. This step quantifies this residual accumulation effect by analyzing the decay characteristics of the time-varying stability coefficient sequence during the intervals between train passages. When the accumulated damage causes the residual value of the stability coefficient sequence or its decay rate to reach a convergence threshold defined by the maximum allowable deformation of the tunnel lining structure, it indicates that the deformation has developed to a stage that may endanger the safety of the main tunnel structure. This identifies a slope instability mode that strongly interacts with the tunnel, rather than simply the instability of the slope itself.

[0042] Step S600: Quantify operational risks based on slope instability modes and output risk assessment results measured by unplanned traffic interruption time.

[0043] Understandably, step S600 transforms the determination of the mechanical state into a risk indicator that can guide operational decisions. Based on the instability mode identified in step S500, its spatial orientation and development rate determine the urgency of the hazard. This step simulates the time required for instability to develop to the point of threatening train operation safety without intervention, and compares this with the railway's established maintenance window time to determine the feasibility of maintenance intervention during the maintenance window. When the estimated intervention time exceeds the maintenance window time, it means that planned train operation must be interrupted for emergency repairs. By calculating the probability and estimated duration of such unplanned train operation interruptions, the final output is a risk assessment result measured by the unplanned train operation interruption time, thus directly linking geomechanical analysis with operational safety management.

[0044] Further, step S200 includes steps S210 to S230.

[0045] Step S210: Based on the video image data, perform spatiotemporal manifold construction processing. By identifying the edge contour of the tunnel entrance structure and defining it as an indeformable manifold boundary constraint, the spatiotemporal characteristic manifold of the slope under geometric constraints is obtained.

[0046] Step S220: Based on the spatiotemporal characteristic manifold of the slope, perform constrained manifold expansion processing. By performing Riemannian geometric expansion on the spatiotemporal characteristic manifold under boundary constraints, the high-dimensional manifold is mapped to a measurable low-dimensional space to obtain the deformation field of the slope surface.

[0047] Step S230: Perform differential displacement solution processing based on the surface deformation field of the slope. By calculating the tangent space covariant derivative of the surface deformation field with respect to the fixed reference point of the tunnel lining, the dynamic differential displacement field of the slope surface is obtained.

[0048] Specifically, in step S210, spatiotemporal manifold construction is first performed based on video image data. By identifying the clear edge contours of the tunnel lining structure and defining these contours as indeformable rigid boundary constraints, the continuous image sequence is fused into a high-dimensional data set constrained by the actual terrain and structural geometry in both time and space, namely, the spatiotemporal characteristic manifold of the slope. Then, in step S220, constrained manifold expansion is performed based on this geometrically constrained spatiotemporal characteristic manifold. Using the principle of handling curved space in Riemannian geometry, the high-dimensional manifold is mathematically expanded and reduced under fixed boundary constraints, mapping it to a low-dimensional space that can be directly measured and calculated. This yields a slope surface deformation field that can intuitively describe the displacement distribution at each point on the slope surface. Subsequently, in step S230, differential displacement solution is performed based on the slope surface deformation field. The tangent space covariant derivative of each point in the deformation field relative to a pre-selected fixed reference point on the tunnel lining is calculated. The reason for using the tangent space covariant derivative is that the slope surface is a complex curved surface, and directly calculating the displacement change would include measurement distortion caused by the curvature of the surface itself. This derivative, however, is a differential operation specifically defined for curved spaces, which can automatically deduct the effect of terrain curvature, thereby extracting the local relative displacement change rate caused only by the actual deformation of the soil and rock mass. The entire process gradually peels away the geometric constraints from the image with a complex background and uses mathematical tools adapted to the characteristics of the curved surface to gradually extract accurate deformation information, effectively overcoming the visual distortion and measurement benchmark problems caused by the special terrain at the tunnel entrance, and providing accurate input for subsequent analysis.

[0049] The formula for projection of the deformation field is:

[0050] ;

[0051] ;

[0052] In the formula, : The three-dimensional spatial coordinates of any point on the slope surface; : Fixed reference points on the tunnel lining structure, whose three-dimensional coordinates are pre-calibrated by design drawings or precision measurements, and are regarded as invariants in calculations; :point The three-dimensional displacement vector at the location; Fixed reference point The three-dimensional displacement vector at the location; , They are respectively at point The two orthogonal unit basis vectors of the local tangent plane of the slope surface defined at the location. Take a horizontal direction parallel to the tunnel axis and pointing outwards from the slope. Take a direction perpendicular to the tunnel axis pointing upwards; :point The displacement in the local tangent plane basis vector The projection component in the direction reflects the tangential displacement along the tunnel axis; :point The displacement in the local tangent plane basis vector The projection component in the direction reflects the tangential displacement perpendicular to the tunnel axis.

[0053] The formula for calculating dynamic differential displacement is:

[0054] ;

[0055] In the formula, For local surface coordinates, ; Let tangent vector field be defined on the slope surface. ,in , ; The index indicates the component. The Christofel notation is used to quantify the change of basis vectors when the local tangent space moves across a surface; Tangent vector field of the slope surface Regarding coordinates The covariant derivative; Represents partial derivatives; This represents the same tangential displacement vector field defined on the slope surface. The amount, and They are the same physical quantity.

[0056] Further, step S300 includes steps S310 to S330.

[0057] Step S310: Perform dynamic load component separation processing on the waveform data. By analyzing the dispersion characteristics of the waveform propagation in the rail-sleeper-ballast-lining medium, separate the direct vibration component belonging to the rail-lining path and the vibration component scattered through the ballast-surrounding rock to obtain the separated train-induced vibration field.

[0058] Step S320: Simulate the axial waveguide propagation of the tunnel based on the train-induced vibration field. By modeling the tunnel lining structure as a tubular waveguide, analyze the modal propagation and attenuation law of the train-induced vibration field under waveguide constraint, and obtain the modal dynamic stress distribution of the tunnel axis.

[0059] Step S330: Using the modal dynamic stress distribution as the boundary condition and combining it with the stress release effect of the free surface at the tunnel entrance, solve the wave equation and invert the dynamic stress tensor field inside the surrounding rock at the tunnel entrance.

[0060] Specifically, step S310 first performs train dynamic load component separation processing on the waveform data. By analyzing the dispersion characteristics of different frequency components with different wave velocities when the vibration waveform propagates in the multi-layered medium of rail, sleeper, ballast, and lining, the direct vibration component propagating along the rail and lining structure, as well as the vibration component propagating through the ballast layer into the surrounding rock via complex paths such as scattering and reflection, are separated from the mixed signal. This yields a separated train-induced vibration field that can distinguish different propagation mechanisms and paths. Based on this, step S320 performs tunnel axial waveguide propagation simulation based on the separated train-induced vibration field. The core of this simulation is to abstract the tunnel lining structure as... A tubular waveguide physical model is used to analyze the specific vibration modes excited by vibrations of different frequencies in the waveguide structure and their propagation laws and distance attenuation characteristics along the tunnel axis. This allows for the calculation of the vibration energy distribution along the tunnel's longitudinal direction, i.e., the modal dynamic stress distribution along the tunnel axis. Finally, in step S330, the aforementioned modal dynamic stress distribution is applied as a known condition to the mechanical boundary between the tunnel and the surrounding rock. Considering the stress wave reflection and energy release effects unique to the free face at the tunnel entrance, the wave equation describing the propagation law of stress waves in the soil and rock medium is solved numerically. This leads to the reverse derivation of the dynamic stress tensor field varying with time within the surrounding rock around the entire tunnel entrance. Preferably, the governing equation for this process is an elastic wave equation considering complex boundary conditions, with the following vector form:

[0061] ;

[0062] The dynamic boundary conditions (tunnel lining-surrounding rock interface) are as follows:

[0063] ;

[0064] The stress relief boundary condition for the free face (tunnel entrance free face) is as follows:

[0065] ;

[0066] In the formula, The density of the surrounding rock around the tunnel entrance; This represents the vibrational displacement vector field of the surrounding rock particles. It is a time variable; , Let be the Lamé constant of the surrounding rock, where It is the shear modulus. Related to the bulk modulus, both describe the elastic properties of the surrounding rock and determine the propagation speed of stress waves (especially P-waves and S-waves) within it. For gradient operators, For divergence operators, For the Laplace operator; This is a physical force term, which in this embodiment mainly represents gravity. This is the Cauchy stress tensor, which is the dynamic stress tensor field inside the surrounding rock at the tunnel entrance that needs to be inverted in this step. The unit vector representing the outward normal direction of the boundary; This represents the interface between the tunnel lining and the surrounding rock. To be applied to The known modal dynamic stress distribution on; The free face representing the tunnel entrance, defined in this embodiment as a boundary condition with zero stress, physically reflects the stress release effect where stress waves are reflected when they propagate to the free face and the normal stress is zero. Represents partial derivatives; This indicates that it is valid on the specified boundary (or area).

[0067] Further, step S400 includes steps S410 to S430.

[0068] Step S410: Based on the dynamic differential displacement field, the potential slip surface mode is constructed. By screening the continuous region with non-zero covariant derivative in the dynamic differential displacement field and combining the geometric boundary constraints of the tunnel lining, the slip surface morphology extending to the tunnel entrance is determined, and the virtual displacement mode of the potential slip surface of the slope is obtained.

[0069] Step S420: Based on the virtual displacement mode and dynamic stress tensor field of the potential slip surface of the slope, perform dynamic virtual work calculation. By convolving the effective stress perpendicular to the slip surface, calculate the anti-slip virtual work and the load virtual work that drives the slip, respectively, and obtain the time series dataset of anti-slip virtual work and load virtual work.

[0070] Step S430: Perform time-varying stability analysis based on the time series dataset. By dividing the anti-sliding virtual work at the same moment by the load virtual work and arranging the ratio along the time axis, a time-varying stability coefficient sequence characterizing the instantaneous safety state of the slope is obtained.

[0071] Specifically, step S410 constructs a potential slip surface model based on the dynamic differential displacement field. By screening continuous regions with non-zero covariant derivatives in the dynamic differential displacement field, it identifies the locations of non-coordinated real deformation within the slope. Combined with the geometric constraints formed by the rigid boundary of the tunnel lining in space, it infers the spatial morphology of the potential slip surface most likely to develop from within the slope and extend to the weak point at the tunnel entrance, thus obtaining a virtual displacement model of the potential slip surface describing the possible movement of the hypothetical slip body. Step S420 calculates the dynamic virtual work based on this virtual displacement model and the dynamic stress tensor field. It convolves the effective stress components perpendicular to the constructed slip surface in the dynamic stress tensor field with the virtual displacement model in the spatiotemporal domain, calculating the anti-slip virtual work (resistance to the shear strength of the soil mass) and the load virtual work (driving the slip through the external train dynamic load) during the hypothetical sliding process, thus obtaining time-series datasets of the anti-slip virtual work and the load virtual work evolving over time. Preferably, the convolution integral formula for the dynamic virtual work is as follows:

[0072] Formula for virtual work of load (driving sliding):

[0073] ;

[0074] Formula for anti-slip virtual work (resistance to slippage):

[0075] ;

[0076] Wherein, the anti-slip force distribution function Defined as:

[0077] ;

[0078] In the formula, For the timing of virtual work performed by the load; This is a potential slip surface on the slope; For potential slip surface Position coordinates on; It is a time variable; This represents the dynamic shear stress component in the sliding direction; This represents the component of the virtual displacement mode in the sliding direction; For historical time variables; For anti-slip virtual work timing; The sliding force distribution function is defined by the Mohr-Coulomb criterion, which characterizes the distributed resistance provided by the shear strength inside the rock mass on the slip surface. Location on the slip surface The coefficient of friction at that location; This refers to the normal dynamic stress (normal stress) on the slip surface. This is the differential symbol.

[0079] Finally, step S430 performs time-varying stability analysis based on this time-series dataset. This involves dividing the virtual work done on the anti-slip at each identical calculation moment by the corresponding virtual work done by the load, obtaining a stability coefficient characterizing the ratio of the instantaneous anti-slip capacity to the driving load. This coefficient is then arranged chronologically across all time steps, ultimately forming a time-varying stability coefficient sequence that continuously reflects the instantaneous fluctuations in the slope's safety reserve under cyclic dynamic loading. This process transforms deformation and stress fields into explicit mechanical criteria, quantifying continuous dynamic interactions into a traceable and assessable stability index.

[0080] Further, step S500 includes steps S510 to S530.

[0081] Step S510: Extract residual effects during the interval period based on the time-varying stability coefficient sequence. By performing bandpass filtering and integration on the sequence during the interval period of train load action, the residual stability coefficient decay curve characterizing the fatigue and damage state of the structure is obtained.

[0082] Step S520: Perform cumulative damage and recovery analysis based on the residual stability coefficient decay curve. By fitting the exponential decay law of the decay curve and calculating the cumulative increment of residual effect within adjacent train load cycles, the time-varying cumulative damage degree of the slope-lining coupling system is obtained.

[0083] Step S530: Perform instability mode discrimination processing based on time-varying cumulative damage degree. By mapping the time-varying cumulative damage degree to the failure envelope surface with the maximum allowable deformation of the tunnel lining structure as the boundary, determine whether the envelope surface is triggered and determine the dominant mechanism of instability to obtain the slope instability mode.

[0084] Specifically, in step S510, the residual effect extraction process during the interval period is first performed on the time-varying stability coefficient sequence. By applying a bandpass filter of a specific frequency band to the original sequence during the interval period after the train load is applied, random environmental noise and high-frequency fluctuations are removed. Then, the filtered sequence is integrated to quantify the stability loss that has not recovered after one load cycle, and finally a residual stability coefficient decay curve characterizing the progressive development of structural fatigue and irreversible damage is obtained. Based on this curve, step S520 performs cumulative damage and recovery analysis. The core is to determine the specific parameters of the exponential decay law followed by the decay curve through nonlinear fitting. This law describes the damping recovery characteristics of the system itself, and further calculates the net residual stability coefficient base value between two adjacent train load cycles. The increment represents the unrecoverable portion of the damage. By accumulating these increments, a scalar index reflecting the gradual progression of damage over time is obtained, namely the time-varying cumulative damage degree of the slope-lining coupled system. Finally, step S530 performs instability mode discrimination processing based on this time-varying cumulative damage degree. The principle is to establish a mathematical surface, namely the failure envelope, with the maximum allowable deformation of the tunnel lining structure as the critical condition. This envelope defines the mechanical boundary between the safety and failure of the support structure. By mapping the calculated time-varying cumulative damage degree onto this envelope, it can be determined whether the current cumulative damage state has reached or exceeded the safety boundary. Based on the specific location and evolution direction of the damage degree on the envelope, the dominant mechanism leading to failure is identified as bias, shear, or a coupling of both, thus ultimately determining the specific slope instability mode. This process realizes the separation and extraction of the damage accumulation trend from continuous dynamic signals and compares it with a clear structural failure threshold, completing the quantitative diagnosis and mode determination of the precursors of progressive instability.

[0085] Further, step S600 includes steps S610 to S630.

[0086] Step S610: Simulate the instability development process based on the slope instability mode. By combining the spatial orientation and development rate of the slope instability mode, predict the critical time required for the instability to develop to the point of affecting the safety of the tunnel lining structure under no-intervention conditions, and obtain the instability development timeline.

[0087] Step S620: Based on the timeline of instability development, conduct a simulation of maintenance intervention during the skylight period. By combining the length of the predetermined skylight maintenance time with the limitations of the construction process, simulate the operation process and time required to reinforce or support the potential sliding body during the skylight period, and obtain the time required for maintenance intervention.

[0088] Step S630: Based on the instability development timeline and the time required for maintenance intervention, calculate the risk of train interruption. Determine whether the time required for maintenance intervention exceeds the predetermined maintenance window time, and calculate the total unplanned train delay caused by construction occupying the line. Output the risk assessment result measured by the unplanned train interruption time.

[0089] Specifically, step S610 first simulates the instability development process based on the identified slope instability modes. This involves combining the spatial orientation information within the mode with the quantitative parameters of the development rate. Spatial orientation determines the direction and location of deformation and failure on the tunnel lining, while the development rate characterizes how quickly deformation progresses over time. By integrating these two into an evolutionary model describing the progressive failure process of the soil and rock mass, the critical time required for a potential landslide to develop from its current state to a point where its deformation is sufficient to threaten the safety of the tunnel lining structure can be calculated without human intervention. This yields a clear time prediction line, i.e., the instability development timeline. Based on this prediction, step S620 simulates maintenance intervention during track maintenance windows. The key is to use the existing track maintenance window in railway operations as a rigid constraint and consider the inherent time consumption and efficiency limitations of the reinforcement or support construction techniques used for different landslide morphologies in actual engineering. The entire operation process, from personnel and equipment entry and construction to cleanup and exit, is dynamically simulated in a computer to objectively assess the shortest time required to complete effective treatment, i.e., the maintenance intervention time. Finally, step S630 integrates the results of the first two steps to calculate the risk of train disruption. By directly comparing the time required for maintenance intervention with the length of the predetermined maintenance window, the feasibility of completing the maintenance using the existing maintenance window is determined. If the maintenance time exceeds the maintenance window time, it means that the planned train operation must be interrupted. The final output of the risk calculation is a quantitative assessment of the total expected train delay caused by the need for unplanned maintenance. This transforms the slope instability mode determined by mechanical mechanisms into a specific risk assessment result that the operations department can directly use for decision-making. This process completes a closed loop from the analysis of geomechanical phenomena to the quantitative management and control of operational safety risks.

[0090] Further, step S610 includes steps S611 to S613.

[0091] Step S611: Extract parameters based on the slope instability mode. Analyze the spatial orientation of the slip surface in the instability mode, calculate its angle with the tunnel axis and the effective slip surface dip angle, and obtain the set of geometric and mechanical parameters of the potential sliding body.

[0092] Step S612: Quantify the instability evolution law based on the set of geometric and mechanical parameters. By coupling the development rate with the parameter set, fit the nonlinear evolution function describing the relationship between the deformation of the sliding body and time to obtain the progressive deformation index of slope instability.

[0093] Step S613: Perform critical time prediction processing based on the progressive deformation index. By substituting the progressive deformation index into the differential equation with the maximum allowable deformation of the tunnel lining structure as the boundary condition, the time required for the current state to develop to the critical deformation amount is predicted, and the instability development timeline is obtained.

[0094] Specifically, step S611 first extracts parameters based on the slope instability model. The core of this process is to analyze the spatial orientation of the slip surface in this model, specifically calculating the angle between the potential slip surface's orientation and the tunnel axis, as well as its effective dip angle on a section perpendicular to the tunnel axis. These two key angles define the relative positional relationship between the spatial morphology of the slip body and the tunnel structure, thus obtaining a set of geometric and mechanical parameters for the potential slip body, encompassing its geometric and basic mechanical properties. Based on this, step S612 quantifies the instability evolution law of this parameter set. The development rate obtained through monitoring is mathematically coupled with the aforementioned parameter set. By fitting a nonlinear evolution function that describes the relationship between slip body deformation (such as displacement or strain) and time, the form of which reflects the characteristics of accelerated or decelerated deformation, a progressive deformation index characterizing the speed and nonlinearity of the instability process is finally obtained. The nonlinear evolution function is expressed as:

[0095] ;

[0096] In the formula, It represents the comprehensive deformation of the potential landslide body on the slope, characterizing the cumulative deformation of the landslide body along the main sliding direction; It is a time variable; This is the initial deformation amount; The initial deformation rate; It is a natural constant; It is a progressive deformation index that comprehensively reflects the inherent acceleration trend and nonlinearity of the deformation development of a potential sliding body under geometric, mechanical properties and external disturbances.

[0097] Finally, step S613 performs critical time prediction based on this progressive deformation index. This is done by substituting the index as a control parameter into a differential equation with the maximum allowable deformation of the tunnel lining structure as a defined boundary condition for numerical solution. This equation describes the path of deformation development over time under a given evolution law. By solving for the time corresponding to the deformation reaching the critical boundary condition, the precise time required for the current state to develop into a dangerous state—that is, the instability development timeline—is predicted. Preferably, the differential equation for predicting the critical time is expressed as:

[0098] ;

[0099] Boundary conditions:

[0100] ;

[0101] In the formula, The normalized comprehensive deformation of the potential landslide body on the slope represents the cumulative deformation of the landslide body along the main sliding direction. Its value is between 0 and 1, where 0 represents no deformation and 1 represents reaching the critical state. It is a time variable; The progressive deformation index; It is a nonlinear strengthening coefficient used to reflect the strength degradation process caused by the accumulation of damage and crack propagation in rock and soil. This is the critical approach coefficient; The reference deformation amount represents the threshold at which deformation development transitions from an approximately linear stage to a significantly nonlinear accelerated stage. To strengthen the index; The critical normalized deformation is the boundary condition corresponding to the maximum allowable deformation of the tunnel lining structure. When the slope deformation reaches this value, it is considered that the tunnel lining structure will undergo unacceptable deformation or damage. for Normalized composite deformation of the potential landslide body at any given time; The predicted critical time; This is the differential symbol.

[0102] The differential equation is solved numerically, and the boundary conditions are determined. The moment This allows us to predict the time required for the current state to develop into a dangerous critical state, and then calculate the actual time, thus completing a quantitative prediction of the instability development timeline.

[0103] Example 2:

[0104] like Figure 2 As shown in the figure, this embodiment provides a railway tunnel entrance slope instability risk assessment system based on image recognition. The system includes:

[0105] The acquisition module 901 is used to acquire video image data of the slope area at the entrance of the target railway tunnel. The video image data includes visual information of the slope surface, tunnel lining structure and connecting area, and simultaneously acquires waveform data of the train passing through collected by sensors installed on the sleepers inside the tunnel entrance.

[0106] Module 902 is used to construct the deformation field of the slope surface based on video image data. By introducing the geometric constraints of the tunnel entrance terrain as manifold boundary conditions, the spatiotemporal characteristic manifold of the image is constrainedly expanded and its curvature analyzed to obtain the dynamic differential displacement field of the slope surface.

[0107] The simulation module 903 is used to simulate the propagation of axial dynamic load in the tunnel based on waveform data. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion.

[0108] Analysis module 904 is used to perform dynamic interaction analysis of slope-tunnel structure based on dynamic differential displacement field and dynamic stress tensor field, and obtain time-varying stability coefficient sequence. The time-varying stability coefficient sequence is the ratio sequence of the anti-sliding virtual work of potential slope sliding body to the load virtual work as time changes.

[0109] The identification module 905 is used to identify instability precursors based on the time-varying stability coefficient sequence. By monitoring the residual cumulative effect and its decay characteristics during the train passing interval, and determining whether the convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, the slope instability mode is obtained.

[0110] Assessment module 906 is used to quantify operational risks based on slope instability modes and output risk assessment results measured by unplanned traffic interruption time.

[0111] In one specific embodiment of this application, the construction module 902 includes:

[0112] The first construction unit is used to perform spatiotemporal manifold construction processing based on video image data. By identifying the edge contour of the tunnel entrance structure and defining it as an indeformable manifold boundary constraint, the spatiotemporal characteristic manifold of the slope under geometric constraints is obtained.

[0113] The second building unit is used to perform constrained manifold expansion processing based on the spatiotemporal characteristic manifold of the slope. By performing Riemannian geometric expansion on the spatiotemporal characteristic manifold under boundary constraints, the high-dimensional manifold is mapped to a measurable low-dimensional space to obtain the deformation field of the slope surface.

[0114] The third building unit is used to perform differential displacement calculation based on the surface deformation field of the slope. By calculating the tangent space covariant derivative of the surface deformation field with respect to the fixed reference point of the tunnel lining, the dynamic differential displacement field of the slope surface is obtained.

[0115] In one specific embodiment of this application, the simulation module 903 includes:

[0116] The first simulation unit is used to separate the dynamic load components of the train based on the waveform data. By analyzing the dispersion characteristics of the waveform propagating in the rail-sleeper-ballast-lining medium, the direct vibration components belonging to the rail-lining path and the vibration components scattered through the ballast-surrounding rock are separated to obtain the separated train-induced vibration field.

[0117] The second simulation unit is used to simulate the axial waveguide propagation of the tunnel based on the train-induced vibration field. By modeling the tunnel lining structure as a tubular waveguide, the modal propagation and attenuation law of the train-induced vibration field under waveguide constraint is analyzed, and the modal dynamic stress distribution of the tunnel axis is obtained.

[0118] The third simulation unit is used to solve the wave equation by taking the modal dynamic stress distribution as the boundary condition and combining it with the stress release effect of the free surface at the tunnel entrance, and then inverting the dynamic stress tensor field inside the surrounding rock at the tunnel entrance.

[0119] Example 3:

[0120] Corresponding to the above method embodiments, this embodiment also provides a railway tunnel entrance slope instability risk assessment device based on image recognition. The railway tunnel entrance slope instability risk assessment device based on image recognition described below and the railway tunnel entrance slope instability risk assessment method based on image recognition described above can be referred to each other.

[0121] Figure 3 This is a block diagram illustrating an image recognition-based railway tunnel entrance slope instability risk assessment device 800 according to an exemplary embodiment. Figure 3 As shown, the image recognition-based railway tunnel entrance slope instability risk assessment device 800 may include: a processor 801 and a memory 802. The image recognition-based railway tunnel entrance slope instability risk assessment device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0122] The processor 801 controls the overall operation of the image recognition-based railway tunnel entrance slope instability risk assessment device 800 to complete all or part of the steps in the image recognition-based railway tunnel entrance slope instability risk assessment method described above. The memory 802 stores various types of data to support the operation of the image recognition-based railway tunnel entrance slope instability risk assessment device 800. This data may include, for example, instructions for any application or method operating on the image recognition-based railway tunnel entrance slope instability risk assessment device 800, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the image recognition-based railway tunnel entrance slope instability risk assessment device 800 and other devices. Wireless communication methods include Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, and an NFC module.

[0123] In an exemplary embodiment, a railway tunnel entrance slope instability risk assessment device 800 based on image recognition may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned railway tunnel entrance slope instability risk assessment method based on image recognition.

[0124] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the image recognition-based railway tunnel entrance slope instability risk assessment method described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions described above. These program instructions may be executed by a processor 801 of an image recognition-based railway tunnel entrance slope instability risk assessment device 800 to complete the image recognition-based railway tunnel entrance slope instability risk assessment method described above.

[0125] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition, characterized in that, include: The video image data of the slope area at the entrance of the target railway tunnel is acquired. The video image data includes visual information of the slope surface, the tunnel entrance lining structure and the connecting area. Simultaneously, waveform data of the train passing through is collected by sensors installed on the sleepers inside the tunnel entrance. Based on the video image data, the deformation field of the slope surface is constructed. By introducing the geometric constraints of the tunnel entrance terrain as the manifold boundary condition, the spatiotemporal characteristic manifold of the image is constrainedly expanded and its curvature analyzed to obtain the dynamic differential displacement field of the slope surface. Based on the waveform data, the axial dynamic load propagation of the tunnel is simulated. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion. Based on the dynamic differential displacement field and the dynamic stress tensor field, the dynamic interaction between the slope and the tunnel structure is analyzed to obtain a time-varying stability coefficient sequence. The time-varying stability coefficient sequence is the sequence of the ratio of the anti-sliding virtual work of the potential sliding body of the slope to the load virtual work as time changes. Based on the time-varying stability coefficient sequence, instability precursors are identified. By monitoring the residual cumulative effect and its attenuation characteristics during the train passing interval, and determining whether the convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, the slope instability mode is obtained. Operational risks are quantified based on the slope instability mode, and risk assessment results are output with unplanned traffic interruption time as the metric.

2. The method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition as described in claim 1, characterized in that, Based on the video image data, a slope surface deformation field is constructed. By introducing the geometric constraints of the tunnel entrance terrain as manifold boundary conditions, the spatiotemporal characteristic manifold of the image is constrainedly expanded and its curvature analyzed to obtain the dynamic differential displacement field of the slope surface, including: Based on the video image data, spatiotemporal manifold construction processing is performed. By identifying the edge contour of the tunnel entrance structure and defining it as an indeformable manifold boundary constraint, a geometrically constrained spatiotemporal characteristic manifold of the slope is obtained. Based on the spatiotemporal characteristic manifold of the slope, a constrained manifold expansion process is performed. By performing Riemannian geometric expansion on the spatiotemporal characteristic manifold under boundary constraints, the high-dimensional manifold is mapped to a measurable low-dimensional space to obtain the deformation field of the slope surface. The differential displacement is solved based on the surface deformation field of the slope. By calculating the tangent space covariant derivative of the surface deformation field with respect to the fixed reference point of the tunnel lining, the dynamic differential displacement field of the slope surface is obtained.

3. The method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition as described in claim 1, characterized in that, Based on the waveform data, the axial dynamic load propagation of the tunnel is simulated. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion, including: Based on the waveform data, the train dynamic load components are separated. By analyzing the dispersion characteristics of the waveform propagating in the rail-sleeper-ballast-lining medium, the direct vibration components belonging to the rail-lining path and the vibration components scattered through the ballast-surrounding rock are separated to obtain the separated train-induced vibration field. Based on the train-induced vibration field, the tunnel axial waveguide propagation simulation was carried out. By modeling the tunnel lining structure as a tubular waveguide, the modal propagation and attenuation law of the train-induced vibration field under waveguide constraint was analyzed, and the modal dynamic stress distribution of the tunnel axial direction was obtained. Using the modal dynamic stress distribution as a boundary condition and combining it with the stress release effect of the free surface at the tunnel entrance, the wave equation is solved, and the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion.

4. The method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition as described in claim 1, characterized in that, Based on the dynamic differential displacement field and the dynamic stress tensor field, a dynamic interaction analysis of the slope-tunnel structure is performed to obtain a time-varying stability coefficient sequence. This time-varying stability coefficient sequence is a sequence of the ratios of the virtual work done by the potential sliding body of the slope to the virtual work done by the load over time, including: The potential slip surface mode is constructed based on the dynamic differential displacement field. By screening the continuous region with non-zero covariant derivative in the dynamic differential displacement field and combining it with the geometric boundary constraints of the tunnel lining, the slip surface morphology extending to the tunnel entrance is determined, and the virtual displacement mode of the potential slip surface of the slope is obtained. Based on the virtual displacement mode of the potential slip surface of the slope and the dynamic stress tensor field, the dynamic virtual work is calculated. By convolving the effective stress perpendicular to the slip surface, the anti-slip virtual work and the load virtual work driving the slip are calculated respectively, and the time series datasets of anti-slip virtual work and load virtual work are obtained. Based on the time-series dataset, time-varying stability analysis is performed. By dividing the anti-sliding virtual work at the same moment by the load virtual work and arranging the ratio along the time axis, a time-varying stability coefficient sequence characterizing the instantaneous safety state of the slope is obtained.

5. The method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition as described in claim 1, characterized in that, Based on the time-varying stability coefficient sequence, instability precursors are identified by monitoring the residual cumulative effect and its attenuation characteristics during train passage intervals, and determining whether a convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, thus obtaining the slope instability mode, including: Based on the time-varying stability coefficient sequence, residual effect extraction processing is performed during the interval period. By performing bandpass filtering and integration on the sequence during the interval period of train load action, the residual stability coefficient decay curve characterizing the fatigue and damage state of the structure is obtained. Based on the residual stability coefficient decay curve, cumulative damage and recovery analysis is performed. By fitting the exponential decay law of the decay curve and calculating the cumulative increment of residual effect within adjacent train load cycles, the time-varying cumulative damage degree of the slope-lining coupling system is obtained. The instability mode is determined by mapping the time-varying cumulative damage degree to a failure envelope surface bounded by the maximum allowable deformation of the tunnel lining structure. This process determines whether the envelope surface is triggered and identifies the dominant instability mechanism, thus obtaining the slope instability mode.

6. The method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition as described in claim 1, characterized in that, Based on the aforementioned slope instability mode, operational risks are quantified, and risk assessment results measured by unplanned traffic interruption time are output, including: Based on the slope instability mode, the instability development process is simulated. By combining the spatial orientation and development rate of the slope instability mode, the critical time required for the instability development to affect the safety of the tunnel lining structure under no-intervention conditions is predicted, and the instability development timeline is obtained. Based on the instability development timeline, a simulation of maintenance intervention during the skylight period is conducted. By combining the length of the predetermined skylight maintenance time with construction process limitations, the operation process and required time for reinforcing or supporting the potential sliding body during the skylight period are simulated to obtain the required time for maintenance intervention. Based on the instability development timeline and the time required for maintenance intervention, the risk of train interruption is calculated. By determining whether the time required for maintenance intervention exceeds the predetermined maintenance window time, and calculating the total unplanned train delay caused by construction occupying the line, the risk assessment result measured by the unplanned train interruption time is output.

7. The method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition as described in claim 6, characterized in that, Based on the slope instability mode, the instability development process is simulated. By combining the spatial orientation and development rate of the slope instability mode, the critical time required for it to develop to the point of affecting the safety of the tunnel lining structure under no-intervention conditions is predicted, resulting in an instability development timeline, including: Based on the slope instability mode, parameter extraction processing is performed. By analyzing the spatial orientation of the slip surface in the instability mode, the angle between the slip surface and the tunnel axis and the effective slip surface dip angle are calculated to obtain the geometric and mechanical parameter set of the potential sliding body. The instability evolution law is quantified based on the set of geometric and mechanical parameters. By coupling the development rate with the set of parameters, a nonlinear evolution function describing the relationship between the deformation of the sliding body and time is fitted to obtain the progressive deformation index of slope instability. The critical time prediction process is performed based on the progressive deformation index. By substituting the progressive deformation index into the differential equation with the maximum allowable deformation of the tunnel lining structure as the boundary condition, the time required for the current state to develop to the critical deformation amount is predicted, thus obtaining the instability development timeline.

8. A method for assessing the risk of slope instability at the entrance of a railway tunnel based on image recognition, characterized in that, include: The acquisition module is used to acquire video image data of the slope area at the entrance of the target railway tunnel. The video image data includes visual information of the slope surface, tunnel entrance lining structure and connecting area, and simultaneously acquires waveform data of the train passing through collected by sensors installed on the sleepers inside the tunnel entrance. The construction module is used to construct the slope surface deformation field based on the video image data. By introducing the geometric constraints of the tunnel entrance terrain as the manifold boundary condition, the spatiotemporal characteristic manifold of the image is constrainedly expanded and its curvature analyzed to obtain the dynamic differential displacement field of the slope surface. The simulation module is used to simulate the propagation of axial dynamic load in the tunnel based on the waveform data. By analyzing the dispersion relationship of the vibration waveform in the rail-sleeper-ballast-lining medium and the waveguide effect of the tunnel tubular structure, the dynamic stress tensor field inside the surrounding rock at the tunnel entrance is obtained by inversion. The analysis module is used to perform dynamic interaction analysis of the slope-tunnel structure based on the dynamic differential displacement field and the dynamic stress tensor field, and obtain a time-varying stability coefficient sequence. The time-varying stability coefficient sequence is a sequence of the ratio of the anti-sliding virtual work of the potential sliding body of the slope to the load virtual work as time changes. The identification module is used to identify instability precursors based on the time-varying stability coefficient sequence. By monitoring the residual cumulative effect and its decay characteristics during the train passing interval, and determining whether the convergence threshold defined by the maximum allowable deformation of the tunnel lining structure is triggered, the slope instability mode is obtained. The assessment module is used to quantify operational risks based on the slope instability mode and output risk assessment results measured by unplanned traffic interruption time.

9. The image recognition-based railway tunnel entrance slope instability risk assessment system according to claim 8, characterized in that, The building module includes: The first construction unit is used to perform spatiotemporal manifold construction processing based on the video image data. By identifying the edge contour of the tunnel entrance structure and defining it as an indeformable manifold boundary constraint, a geometrically constrained spatiotemporal characteristic manifold of the slope is obtained. The second construction unit is used to perform constrained manifold expansion processing based on the spatiotemporal characteristic manifold of the slope. By performing Riemannian geometric expansion on the spatiotemporal characteristic manifold under boundary constraints, the high-dimensional manifold is mapped to a measurable low-dimensional space to obtain the slope surface deformation field. The third construction unit is used to perform differential displacement solution processing based on the surface deformation field of the slope. By calculating the tangent space covariant derivative of the surface deformation field with respect to the fixed reference point of the tunnel lining, the dynamic differential displacement field of the slope surface is obtained.

10. The image recognition-based railway tunnel entrance slope instability risk assessment system according to claim 8, characterized in that, The simulation module includes: The first simulation unit is used to separate the dynamic load components of the train based on the waveform data. By analyzing the dispersion characteristics of the waveform propagating in the rail-sleeper-ballast-lining medium, the direct vibration components belonging to the rail-lining path and the vibration components scattered through the ballast-surrounding rock are separated to obtain the separated train-induced vibration field. The second simulation unit is used to simulate the axial waveguide propagation of the tunnel based on the train-induced vibration field. By modeling the tunnel lining structure as a tubular waveguide, the modal propagation and attenuation law of the train-induced vibration field under waveguide constraint is analyzed to obtain the modal dynamic stress distribution of the tunnel axis. The third simulation unit is used to take the modal dynamic stress distribution as the boundary condition and, in conjunction with the stress release effect of the free surface at the tunnel entrance, solve the wave equation and invert the dynamic stress tensor field inside the surrounding rock at the tunnel entrance.

Images