Slope seismic intelligent prediction and reinforcement method based on time point safety factor evolution

Abstract

The application discloses a kind of slope anti-seismic intelligent prediction and reinforcement method based on time point safety factor evolution, belong to geological disaster intelligent prediction technical field.This method first constructs slope mechanical analysis model, introduces propagation time delay to represent seismic traveling wave effect, introduces time point safety factor to realize fine identification of local stability;Establish plastic dissipation work driven negative exponential damage evolution model, dynamically reduce shear strength parameter, simulate the whole process of slope progressive failure;Build deep learning model that fuses physical regularization constraint, train with high-fidelity mechanical positive calculation result as sample, finally realize the second-level prediction of slope point safety factor and damage distribution under complex seismic working condition, and automatically identify high-risk area and intelligently optimize anti-slide pile, anchor cable layout parameter, trigger hierarchical warning.The application can accurately identify the internal local progressive instability of slope, and considers both calculation accuracy and efficiency, to provide quantitative technical support for slope anti-seismic stability evaluation and intelligent reinforcement.

Description

Technical Field

[0001] This invention relates to the field of intelligent geological disaster prediction technology, specifically to an intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points. Background Technology

[0002] With the continuous expansion of infrastructure construction in mountainous areas, slope engineering is gradually showing a trend of becoming steeper, larger in scale, and more complex in strata. Under the action of seismic dynamics, the internal stress field of foundation fill / cut slopes exhibits complex spatiotemporal non-uniform distribution characteristics. Local areas first enter the plastic state under the traveling wave effect, and gradually expand to form potential slip channels under reciprocating shear action. This process has significant characteristics of gradualism and sudden coupling.

[0003] Most existing slope stability analysis methods use the overall safety factor as the main evaluation index. In the analysis process, the slope is usually treated as a unified system of overall stability or overall instability. While this overall averaging approach can reflect the overall stability of the slope, it cannot effectively identify stress concentration, plastic development, and strength weakening processes in localized areas within the slope. Due to the lack of dynamic tracking capabilities for localized hazardous areas, existing slope stability analysis methods struggle to capture the initiation location and propagation path of slope instability under seismic loading. This makes it difficult to accurately reveal the progressive failure process of the slope, thus hindering rapid prediction and risk warning of slope stability under complex seismic wave conditions. Summary of the Invention

[0004] To address the challenges of seismic-induced asynchronous input, difficulty in accurately identifying internal localized progressive instability, and challenges in rapidly predicting the overall stability evolution of foundation fill / cut slopes under seismic dynamics, as well as the difficulty in intelligently optimizing reinforcement measures, this invention proposes an intelligent seismic prediction and reinforcement method for slopes based on the evolution of the safety factor at time-history points. This method enables refined identification of localized hazardous areas within foundation fill / cut slopes, accurately captures the initiation location and propagation path of progressive slope instability, improves the physical authenticity and engineering reliability of stability analysis results, achieves rapid prediction and risk warning of the stability state of foundation fill / cut slopes under complex seismic wave conditions, and realizes intelligent optimization of the placement of anti-slide piles, anchor cables, and support structures.

[0005] This invention provides the following technical solution: A method for intelligent prediction and reinforcement of slope seismic resistance based on the evolution of safety factor at time history points, comprising the following steps: S1. Construction of Slope Seismic Dynamic Response Analysis Model Establish a two-dimensional / three-dimensional geometric model of the target foundation fill or cut slope, and assign physical and mechanical parameters to the soil and rock mass according to the actual engineering strata conditions; based on the limit equilibrium theory, the potential sliding body is discretized into n vertical blocks along the sliding direction, the top of the slope is defined as the first block and the bottom of the slope is defined as the nth block, and a recursive calculation order from the top of the slope to the bottom of the slope is established; the horizontal acceleration time history and vertical acceleration time history of seismic waves are used as external dynamic inputs; S2. Dynamic analysis of local stability based on the safety factor at time history points. A time delay model based on shear wave velocity and propagation direction is established to realize the spatiotemporal asynchronous input of seismic excitation; the instantaneous normal force, tangential sliding force, and anti-sliding force on the strip are calculated based on the basic assumptions of the unbalanced thrust method, and the time history point safety factor is introduced to characterize the instantaneous stability state of a single strip; and a dynamic force transmission mechanism between strips based on the time history point safety factor is constructed, the overall slope safety factor is calculated based on the time history point safety factor of each strip, and a quantitative correlation between local instability and overall stability is established. S3, Dynamic Response Physical Mechanism of Coupled Asymptotic Damage Evolution (Positive Calculation) Based on the laws of damage mechanics and thermodynamics of continuous media, a negative exponential asymptotic damage evolution model with plastic dissipation work as the driving variable is constructed. Based on dynamic damage variables, the shear strength parameters of soil and rock mass in the calculation of strip anti-sliding force are reduced and corrected in real time to obtain the dynamic anti-sliding force and dynamic point safety factor with dual nonlinearity of coupled traveling wave delay effect and material damage. The dynamic transmission law of inter-strip force is re-established to simulate the entire process of gradual failure of slope from local yielding to the penetration of the slip surface. S4. Physical experimental reverse calibration of core parameters of damage evolution The softening curve of the soil and rock mass after peak was obtained by indoor constant normal stress direct shear test. The transient macroscopic damage degree and cumulative plastic dissipation work per unit area were calculated. The test data were substituted into the asymptotic damage evolution model. The energy conversion coefficient in the asymptotic damage evolution model was back-calibrated by linear fitting, providing physical parameter basis for dynamic damage assessment. S5. Construction of deep learning models integrating mechanical mechanisms Based on the physical framework of steps S1 to S4, a large number of typical seismic wave acceleration time histories are selected as input to perform batch dynamic time history forward calculation on the target slope. Key mechanical features under each working condition are extracted as training samples, and the spatial distribution matrix of safety factor and cumulative plastic dissipation work distribution matrix at time history points are extracted as output labels. A deep neural network model integrating physical constraints is constructed, and a physical regularization term based on the progressive damage evolution model is introduced into the loss function. The deep neural network model is trained through the backpropagation algorithm to solidify the nonlinear mapping relationship of dynamic response. S6, Sub-second-level risk identification and intelligent hardening decision-making across time-series iterations The new seismic wave signals captured by the monitoring system are processed by extracting features and then input into a trained deep neural network model. The model outputs predicted values ​​of the time history point safety factor envelope and damage variables for the entire slope section within seconds. Based on the output prediction results and high-risk identification criteria, the sliding surface continuity status and high-risk areas are determined. An intelligent reinforcement strategy is then implemented, which includes automatically optimizing the anti-slide pile layout based on the location of the minimum point safety factor, calculating the effective length of the anchor cable based on the depth of damage evolution, and triggering graded early warnings in conjunction with the overall slope safety factor.

[0006] Further, in step S2, the delay time model is: ; The safety factor for the time history point is: ,in: ; ; ; ; ; In the formula, For the first The relative propagation time difference of the strip relative to the top of the slope; To go from the top of the slope along the landslide path to the first The cumulative distance between the geometric centers of the strips; Shear wave velocity; The angle between the direction of seismic wave propagation and the positive direction of the landslide; For the first strips in Safety factor at each time point; For the first strips in Anti-slip performance at all times; For the first strips in The tangential glide force at any given moment; For the first strips in The normal force acting on the object at any given moment; For the first The cohesive force of the bottom surface of the strip; For the first The length of the bottom surface of the strip; For the first The internal friction angle of the bottom sliding surface of the strip; For the first The weight of the strip; For the first The angle of inclination of the bottom surface of the strip. For the first Inclination angle of the bottom sliding surface of the strip; For the first The strip comes from the thrust of the previous strip; For the first strips in Horizontal seismic coefficient at time, For the first strips in Vertical seismic coefficient at time; For the first strips in Horizontal acceleration at time t, For the first strips in Vertical acceleration at any given moment; It is the acceleration due to gravity; For seismic waves in Horizontal acceleration at time t, For seismic waves in The vertical acceleration at any given moment.

[0007] Furthermore, in step S2, the dynamic force transmission mechanism between strips is as follows: If This indicates that the block itself is stable and does not require downward thrust, i.e. ;like If the strip is in a state of extreme equilibrium or overload, the remaining sliding force must be borne by the next level strip. The safety factor for this strip point is defined as 1. The expression for the inter-strip force is: ; The overall safety factor of the slope is: ; like This indicates that the slope is generally stable. This indicates that the slope is in a state of limit equilibrium.

[0008] Furthermore, in step S3, the construction process of the asymptotic damage evolution model is as follows: (1) Based on Newmark's rigid body slip theory, calculate the first... strips in The cumulative plastic slip rate is obtained by continuously integrating the relative transient plastic acceleration at time t. ; ; in, For the first strips in The relative transient plastic acceleration at time t; Macauley brackets characterize the plastic response that is elicited only in the presence of unbalanced forces; For the first strips in The cumulative plastic sliding rate at any given time; (2) Define the first Cumulative plastic dissipation work per unit area on the slip surface of the strip: ; (3) Introducing scalar damage variables Characterizing the degree of deterioration of the slip surface strength, in the undamaged state When completely destroyed Based on Kachanov's effective stress equivalence assumption and first-order dynamic attenuation law, an asymptotic damage evolution model is derived: ; in, The energy conversion coefficient is a material constant related to the brittle characteristics of soil and rock media.

[0009] Furthermore, in step S3, the reduction and correction method for the shear strength parameter of the soil and rock mass is as follows: ; ; Among them, subscript and These represent the peak and residual values ​​of the intensity, respectively. For the first Peak cohesion of the bottom surface of the strip, For the first The residual cohesion value of the bottom sliding surface of the strip; For the first The peak value of the internal friction angle of the bottom sliding surface of the strip. For the first The residual internal friction angle of the bottom sliding surface of the strip.

[0010] Furthermore, in step S4, the reverse calibration method for the energy conversion coefficient is specifically as follows: (1) Define the transient macroscopic damage degree of the test as: ; in, Indicates the post-peak plastic shear displacement; This represents the peak shear force during the experiment. Residual shear force; This refers to the transient shear stress under post-peak plastic shear displacement. (2) Calculate the cumulative plastic dissipation work: ; in, This represents the initial plastic displacement corresponding to the peak stress; (3) The transient macroscopic damage degree measured by the experiment is compared with the cumulative plastic dissipation work. Substituting into the asymptotic damage evolution model and performing algebraic transformations, we obtain: ; (4) Extract the post-peak discrete data points of the experiment, in order to As the independent variable, with Using the coordinate system of the dependent variable, a linear fit is performed through the origin using the least squares method, and the slope of the fitted line is the energy conversion coefficient.

[0011] Furthermore, in step S5, the feature vector input to the deep neural network model includes the following features: peak ground acceleration of seismic waves, dominant frequency, effective duration of energy, and discrete acceleration time history; The loss function is: ; in, For the total loss, For the true value, For predicted values, To constrain the weights, The physical constraint term for the progressive damage evolution model is expressed as: ; in, The first output of the deep neural network model strips in Predictive damage variables at time points; For the first strips in The cumulative plastic dissipation work over time.

[0012] Further, in step S6, the determination of the sliding surface penetration status and high-risk area based on the output prediction result and according to the high-risk identification criterion includes: The high-risk identification criterion is defined as follows: ; Among them, the The first output of the deep neural network model Predicted value of the safety factor envelope at the time history point of the block; The critical damage threshold determined experimentally in step S4; If the high-risk identification criteria are met, it is determined to be a high-risk unstable node. If consecutive high-risk nodes form a spatial connectivity zone, it is determined to be a potential slip-through area.

[0013] Furthermore, in step S6, the intelligent hardening strategy specifically includes: (1) The predicted value of the safety factor at the time point Global minimum value bar index As the optimal layout axis center for anti-slide piles; (2) The effective length of the prestressed anchor cable is automatically calculated based on the predicted damage evolution depth: ; In the formula, The effective length of the prestressed anchor cable. The reserved length for the anchoring section to extend into the stabilizing slide. The predicted depth of damage evolution; (3) Combine the predicted overall safety factor to trigger graded early warning, so as to achieve a second-level response for emergency rescue of large-scale foundation filling / cutting slopes.

[0014] Compared with the prior art, the present invention has the following advantages: (1) This invention unifies the seismic wave acceleration time history and traveling wave propagation effect into dynamic time history input, which can reflect the real stress changes and phase delay effects of the foundation fill / cut slope during the earthquake process, avoiding the problem of missing evolution process caused by the traditional static analysis method that only calculates a single stage, thereby improving the authenticity of stability evaluation; (2) The present invention introduces the time-history point safety factor as a local stability evaluation index, which can continuously identify the instantaneous stability state of each local unit or block inside the foundation fill / cut slope. Compared with the traditional overall safety factor method, the present invention can not only determine whether the slope is in an overall stable state, but also further identify the location, range and expansion path of the local dangerous area, thereby improving the ability to capture early signs of gradual instability. (3) Since the instability of the foundation fill / cut slope is usually not an instantaneous overall failure, but is gradually developed by local plastic deformation, strength decay and dangerous zone expansion, this invention introduces a damage evolution model driven by plastic dissipation work, which links the accumulation of local dissipation energy with the dynamic reduction of shear strength, and more realistically describes the process of the soil mass degrading from peak strength to residual strength, avoiding the defect of traditional methods that overestimate slope stability due to the use of fixed strength parameters; (4) This invention establishes a dynamic transmission mechanism between local damage and overall stability. When a local area enters the limit state, its remaining sliding force, deformation response and damage effect can continue to be transmitted to the adjacent area, thereby simulating the whole process of the dangerous area from local initiation to spatial expansion and then to the connection of the potential slip surface. It is not only applicable to the judgment of stability results, but also applicable to revealing the progressive failure mechanism of the foundation filling and foundation excavation slope. (5) The deep learning model of the present invention does not rely solely on historical experience data, but uses the mechanical forward calculation results and damage evolution law as the mechanism prior. Its prediction results have stronger physical constraints and engineering interpretability. Compared with the traditional black box machine learning model, the present invention can reduce the prediction bias caused by insufficient samples, changes in working conditions or differences in boundary conditions, and improve the generalization ability in the scenarios of foundation filling and foundation excavation slope. (6) This invention simultaneously outputs the location of local dangerous areas, the degree of damage evolution and the potential sliding penetration area of ​​the slope, providing quantitative basis for reinforcement schemes such as anti-slide piles, anchor cables, retaining structures, slope protection and drainage measures, avoiding the problem of over-reliance on experience judgment in traditional design, and improving the pertinence, economy and safety of reinforcement layout.

[0015] In summary, the advantages of this invention are that it can more realistically reflect the stability evolution law of foundation fill / cut slopes, more accurately identify local high-risk areas, more reasonably depict the progressive failure process, and realize rapid prediction and intelligent reinforcement decision-making under complex working conditions through deep learning models. Attached Figure Description

[0016] Figure 1 This is an overall flowchart of the intelligent prediction and reinforcement method for slope seismic resistance of the present invention; Figure 2 This is a schematic diagram showing the overall division of the slope mechanical analysis model constructed in this invention. Figure 3 This is a schematic diagram of the force analysis of a single block in the slope mechanics analysis model constructed in this invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] like Figure 1 As shown in the figure, this embodiment provides a method for intelligent prediction and reinforcement of slope seismic resistance based on the evolution of safety factor at time history points. The specific implementation steps are as follows: S1: Construction of Slope Seismic Dynamic Response Analysis Model First, a two-dimensional or three-dimensional geometric model of the foundation fill or foundation cut slope is established, and physical and mechanical parameters of the soil and rock mass are assigned according to the actual engineering geological conditions, including unit weight, cohesion, internal friction angle, elastic modulus, and permeability parameters, etc.; the slope mechanical analysis model of this invention is as follows: Figure 2 and Figure 3 As shown, using the limit equilibrium theory framework, the potential sliding body of the slope is discretized along the sliding direction as follows: There are three vertical blocks, each with independent geometric and mechanical properties; the block at the top of the slope is defined as block 1, and the block at the bottom of the slope is defined as block 2. The strips, from the top of the slope (the geometric center of the first strip) along the landslide direction to the... The cumulative distance between the geometric centers of the strips is A recursive calculation sequence is established from the top of the slope to the bottom. In this invention, the seismic wave acceleration time history is used as the sole external dynamic input, and the horizontal acceleration time history function of the original input seismic wave is defined as follows: The vertical acceleration time history function is .

[0019] S2: Dynamic Analysis of Local Stability Based on Time-History Point Safety Factor In terms of mechanical equilibrium, the basic assumption of the unbalanced thrust method (transfer coefficient method) is adopted, namely, the direction of the inter-strip force is parallel to the sliding surface of the previous strip, and the strip is defined. From the previous block The thrust is Define boundary conditions That is, there is no external force acting on the top of the slope, the blocks meet the static equilibrium condition, and the rotation effect of the blocks is ignored.

[0020] To reasonably approximate the propagation process of seismic waves inside the slope, a system based on shear wave velocity is established. A time delay model based on the propagation direction is used to achieve asynchronous spatiotemporal input of seismic excitation. Let the angle between the seismic wave propagation direction and the positive direction of the landslide (slope crest → slope toe) be θ. , Then the first The relative propagation time difference of the strip relative to the top of the slope can be uniformly expressed as: Based on the above delay time model, at any time , No. The actual horizontal acceleration experienced by the strip and vertical acceleration It can be obtained by time shifting the original seismic wave time history: This model realizes the asynchronous propagation of seismic excitation in space, i.e., the mathematical representation of the traveling wave effect.

[0021] To facilitate mechanical analysis, seismic wave acceleration is converted into dimensionless seismic coefficients, and the first... strips in Horizontal seismic coefficient at time and vertical seismic coefficient They are respectively: In the formula, This is the acceleration due to gravity.

[0022] To accurately characterize the dynamic response of various components within the foundation fill / cut slope during an earthquake, a time-history point safety factor is introduced as an evaluation index, establishing a dynamic stability analysis method based on the instantaneous stable state of the block. Based on the fundamental assumptions of the unbalanced thrust method, the first... strips Normal force at any moment and tangential sliding force They are respectively: in, For the first The weight of the strip, For the first The angle of inclination of the bottom surface of the strip. For the first Inclination angle of the bottom sliding surface of the strip; No. strips Corresponding anti-slip force at all times Given by the Mohr-Coulomb criterion: in, Refers to the first The cohesive force of the bottom surface of the strip; Refers to the first The internal friction angle of the bottom sliding surface of the strip; Refers to the first The length of the bottom sliding surface of the strip (i.e., the arc length or straight line length of the bottom of the strip in contact with the potential sliding surface) is used to convert cohesion into anti-slip force.

[0023] The safety factor at the time point is defined as the first... strips in The ratio of anti-slip force to sliding force at any given moment: This indicator can intuitively reflect the instantaneous stability state of each segment, capturing local instability phenomena and their evolution processes that traditional overall safety factors cannot reveal. Based on this, a system based on... The dynamic transmission mechanism of inter-strip forces: If This indicates that the block itself is stable and does not require downward thrust, i.e. ;like If the load is too high, the strip is considered to be in a state of limit equilibrium or overload, and the remaining sliding force must be borne by the next level strip. In this case, the safety factor for that strip is defined as 1. Therefore, the expression for the inter-strip force can be written as: To establish the correlation between local instability and overall stability, an overall stability evaluation index is constructed based on the safety factor at time history points. The slope is defined as... Overall safety factor at any time This is a weighted average of the safety factors at each time point of the segment, with the weights being the proportion of the glide force of each segment to the total glide force. Based on the time history response of the overall safety factor, a criterion for the overall stability state of the slope is established: This indicates that the slope is generally stable. This indicates that the slope is in a state of limit equilibrium. Furthermore, by combining the spatial distribution of the safety factor at time-history points, local instability zones and their impact on overall stability can be identified.

[0024] S3: Positive calculation of the dynamic response physical mechanism of coupled asymptotic damage evolution Although the time-history point safety factor can accurately reflect the stability state of a slope at any instant during a seismic wave, slope instability in the complex spatiotemporal non-uniform stress field induced by the traveling wave effect is not a sudden, transient change, but rather a gradual evolutionary process involving the development of micro-cracks, accumulation of plastic strain, and dynamic decay of strength parameters in the local soil and rock mass under reciprocating shearing. To better describe the physical picture of how local failure propagates, accumulates, and ultimately evolves into overall instability between blocks, this invention introduces a damage evolution control equation with plastic dissipation work as the driving variable, based on the laws of continuous medium damage mechanics and thermodynamics.

[0025] According to the second law of thermodynamics, the irreversible deformation of rock and soil under the action of external forces is inevitably accompanied by energy dissipation, and the deterioration of macroscopic mechanical properties (i.e. damage) is essentially driven by the plastic work consumed by the initiation of micro-cracks and frictional slip.

[0026] When the traveling wave of a seismic wave propagates to a specific spatial location, if the first wave... strips in When the sliding force exceeds the transient anti-slip force, the local region will transition from an elastic state to a plastic limit equilibrium state and undergo relative sliding along the potential sliding surface. Based on Newmark's rigid body slip theory, the first... strips in The relative transient plastic acceleration at time t is expressed as: In the formula, It is the acceleration due to gravity; Macauley brackets characterize the plastic response that is elicited only in the presence of unbalanced forces;

[0027] The cumulative plastic sliding rate of the block is obtained by continuously integrating the transient plastic acceleration over time. : In the process of overcoming friction and cohesion to produce relative slip, the work done by the external force is converted into irreversible thermodynamic dissipation, defined as the first... Cumulative plastic dissipation work per unit area on the slip surface of the strip for: In equations (13) and (14), It represents the integral variable and has no physical meaning. Introducing scalar damage variables Characterizes the degree of degradation of the slip surface strength. In the initial undamaged state, Completely destroyed, just before reaching residual strength. .

[0028] Based on Kachanov's effective stress equivalence assumption, the incremental damage of the micro-element... With the increase of dissipated work They are positively correlated. Meanwhile, as damage accumulates, the effective load-bearing area of ​​the slip surface that can resist shear gradually decreases, and the damage evolution rate is inevitably affected by the current remaining integrity level. Strict constraints are imposed. Here, it is assumed that the microscopic fracture follows a first-order dynamic decay law: In the formula, The energy conversion coefficient is a material constant related to the brittle characteristics of soil and rock media. The larger the value, the more easily the material is prone to brittle fracture after absorbing the same amount of energy. Separate the variables in the above equation and, under the undamaged condition... Perform definite integrals over the time interval of the state: The logarithmic function relationship can be obtained by calculating formula (16): Taking the exponents from both sides of equation (17) and rearranging them, we can obtain the negative exponential asymptotic damage evolution model based on rigorous thermodynamic derivation: The equation conforms to the macroscopic physical law of strain softening of soil and rock in terms of mathematical properties: the initial damage from energy dissipation grows rapidly in an approximately linear manner; as energy accumulates, the material gradually breaks down to the residual friction state, and the damage evolution rate decays exponentially and gradually approaches the theoretical extreme value of 1.

[0029] Based on the aforementioned dynamic damage variables, the shear strength parameters in the Mohr-Coulomb criterion are reduced and corrected in real time according to the mixing law, and a constitutive relation for the transition from peak strength to residual strength is established: In the formula, the subscript and These represent the peak and residual values ​​of the intensity, respectively. For the first Peak cohesion of the bottom surface of the strip, For the first The residual cohesion value of the bottom sliding surface of the strip; For the first The peak value of the internal friction angle of the bottom sliding surface of the strip. For the first The residual internal friction angle of the bottom sliding surface of the strip.

[0030] Substituting the dynamically reduced strength parameters into equation (8), we obtain the dynamic anti-slip force, which exhibits both nonlinearity due to the coupled traveling wave delay effect and material damage. : Accordingly, the dynamic point safety factor considering damage evolution characteristics is corrected as follows: The law governing the transmission of forces between bars thus evolves into a dynamic form: The overall safety factor of the slope is corrected to: Equations (22), (23), and (24) construct a theoretical framework for the progressive failure mechanism of slopes at the mathematical level: the traveling wave effect causes the local area to briefly exceed the limit equilibrium state, triggering plastic power dissipation and irreversible strength decay; even if the seismic excitation in the area weakens subsequently, the permanent decrease in its anti-sliding force will still generate an abnormally increased residual thrust, which is transferred to the downstream blocks as an external load, and can better simulate the entire process of the progressive spatiotemporal penetration of high slopes under the action of traveling waves.

[0031] S4: Physical experimental reverse calibration of core parameters of asymptotic damage evolution Energy conversion coefficient in asymptotic damage evolution model It is the core phenomenological parameter controlling the strain softening rate. To ensure the physical rigor of this parameter in engineering calculations, this invention proposes a method for inverse calibration of the post-peak softening curve based on indoor constant normal stress direct shear tests.

[0032] In direct shear tests on soil and rock masses, the shear stress increases with increasing shear displacement. Surpassing the peak It then gradually softens and tends to reach its residual strength. Using shear stress degradation as a macroscopic indicator of damage evolution, the experimental transient macroscopic damage degree is defined. for: In the formula, This represents the post-peak plastic shear displacement. This represents the transient shear stress under the corresponding displacement. Simultaneously, the cumulative plastic dissipation work per unit area during the experiment... It can be obtained by path integration over the post-peak curve: In the formula, This represents the initial plastic displacement corresponding to the peak stress; The transient macroscopic damage measured by the experiment Cumulative plastic dissipation work per unit area Substituting into the asymptotic damage evolution model (Equation (18)): By performing an algebraic transformation on equation (27), rearranging terms, and taking the natural logarithm, the nonlinear exponential evolution relationship is transformed into a strictly linear mapping relationship: Equation (28) can be further expanded to obtain the explicit calibration control equation: In practical applications, the post-peak discrete data points of indoor direct shear tests are extracted, and the accumulated plastic dissipation work is calculated and constructed. As the independent variable, with Let be the coordinate system for the dependent variable. A linear fit passing through the origin is performed using the least squares method; the slope of the fitted line is the energy conversion coefficient required in practical engineering. .

[0033] S5: Building Deep Learning Models Integrating Mechanical Mechanisms Based on the physical framework of steps S1, S2, S3, and S4, a massive amount of typical seismic wave acceleration time histories are selected as input. Batch dynamic time history forward calculations are performed on the target foundation fill / cut slopes, extracting key mechanical features under each working condition as training samples. An input feature vector is defined. for: in, Peak ground acceleration (PGA) of seismic waves For superior frequency, For energy to be effectively sustained, This is the time history of discrete acceleration. Output labels. Includes the spatial distribution matrix of time-history safety factors calculated from physical mechanisms. and cumulative plastic dissipation work distribution matrix .

[0034] Constructing a deep neural network model that incorporates physical constraints To ensure the prediction results possess mechanical rigor, the loss function... Introducing a physical regularization term: in, For the true value, For predicted values, To constrain the weights, The physical constraint term of the progressive damage evolution model in step S3 is expressed as: This constraint ensures that the predicted damage variables output by the deep neural network model are accurate. It strictly follows the negative exponential growth pattern driven by energy dissipation.

[0035] Adjusting network weight parameters using the backpropagation algorithm After training, the model solidifies the complex traveling wave propagation phase difference and the nonlinear material strength dynamic reduction mechanism into a nonlinear mapping function, thus enabling it to replace time-consuming mechanical iterative calculations.

[0036] S6: Millisecond-level risk identification and intelligent hardening decision-making across time-series iterations When the monitoring system captures new seismic wave signals, it directly uses their feature vectors. The input is fed into the trained model. The system bypasses time-consuming nonlinear dynamic time-history iteration and thrust transfer calculations. It directly outputs the predicted value of the safety factor envelope at time history points for the entire cross-section of foundation fill / cut slope within seconds. Compared with the predicted values ​​of damage variables ; The system determines the continuity status of the sliding surface by identifying the spatial connectivity of locally unstable nodes based on the predicted spatial distribution matrix. The high-risk identification criterion is defined as follows: in, The critical damage threshold calibrated in step S4 is defined as follows: In the indoor constant normal stress direct shear test in step S4, during the post-peak softening stage, as the shear displacement continues to increase, the material gradually transitions from peak strength to residual strength. When the shear strength shows a significant and rapid decay and enters a significant damage development stage, the transient macroscopic damage degree corresponding to this characteristic state is defined as the critical damage degree under this test condition. For direct shear tests with multiple sets of normal stresses, the critical damage degree corresponding to each condition is obtained, and the critical damage threshold of the soil and rock material is determined by combining the results of each effective test. This critical damage threshold is used to characterize the boundary by which soil and rock materials transition from a general post-peak softening state to a significantly damaged state.

[0037] Based on the predicted locations of high-risk areas, the system automatically executes reinforcement strategies, with the preferred reinforcement strategy being: (1) Predicted value of safety factor of positioning point Global minimum value bar index This is taken as the optimal layout axis center for the anti-slide piles; (2) Based on the predicted depth of damage evolution Automatically calculate the effective length of prestressed anchor cables Ensure the anchoring section enters the stable slide: In the formula, Reserved length for the anchoring section to extend into the stabilizing slide; (3) Combine the predicted overall safety factor Triggering tiered early warning systems enables second-level response for emergency repairs of large-scale foundation fill / cut slopes.

[0038] It should be noted that, The spatial geometric parameters are determined based on the spatial prediction results of slope damage variables and the identification results of high-risk areas obtained in steps S5 and S6, combined with the slope geometric model established in step S1. After the deep learning model outputs the spatial distribution of damage variables across the entire slope cross section, the deep boundary of the damage-affected area from the slope surface to the interior of the slope can be further determined based on the safety factor at the time history points and the high-risk nodes and their spatial connectivity determined by the damage variables. This refers to the maximum effective depth used to characterize the area affected by the damage. Therefore, its value dynamically changes with different slope geometry, seismic input, and damage evolution status.

[0039] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent prediction and reinforcement of slope seismic resistance based on the evolution of safety factor at time history points, characterized in that, Includes the following steps: S1. Construction of Slope Seismic Dynamic Response Analysis Model: Establish a two-dimensional / three-dimensional geometric model of the target foundation fill / cut slope, and assign physical and mechanical parameters to the soil and rock mass according to the actual engineering strata conditions; based on the limit equilibrium theory, the potential sliding body is discretized into n vertical blocks along the sliding direction, defining the slope top as the 1st block and the slope toe as the nth block, and establishing a recursive calculation order from the slope top to the slope toe; the horizontal acceleration time history and vertical acceleration time history of seismic waves are used as external dynamic inputs; S2. Dynamic analysis of local stability based on time-history point safety factor: Establish a time delay model based on shear wave velocity and propagation direction to realize spatiotemporal asynchronous input of seismic excitation; calculate the instantaneous normal force, tangential sliding force and anti-sliding force on the strip based on the basic assumptions of the unbalanced thrust method, introduce the time-history point safety factor to characterize the instantaneous stability state of the strip; and construct a dynamic force transmission mechanism between strips based on the time-history point safety factor, calculate the overall slope safety factor based on the time-history point safety factor of each strip, and use it to characterize the overall slope stability. S3. Calculation of the physical mechanism of dynamic response of coupled asymptotic damage evolution: Based on the laws of damage mechanics and thermodynamics of continuous medium, a negative exponential asymptotic damage evolution model with plastic dissipation work as the driving variable is constructed; based on the asymptotic damage evolution model, the shear strength parameters of the strip rock and soil are reduced and corrected in real time to obtain the dynamic anti-sliding force and dynamic point safety factor with the dual nonlinearity of coupled traveling wave delay effect and material damage. S4. Physical test reverse calibration of core parameters of damage evolution: The softening curve of soil and rock mass after peak is obtained by indoor constant normal stress direct shear test, the transient macroscopic damage degree and cumulative plastic dissipation work are calculated, the test data are substituted into the asymptotic damage evolution model, and the energy conversion coefficient in the asymptotic damage evolution model is reverse calibrated by linear fitting. S5. Construction of a deep learning model integrating mechanical mechanisms: Based on the physical framework of steps S1 to S4, a large number of typical seismic wave acceleration time histories are selected as inputs to perform batch dynamic time history forward calculations on the target slope. Key mechanical features under various working conditions are extracted as training samples, and the spatial distribution matrix of the safety factor and the cumulative plastic dissipation work distribution matrix at time history points are extracted as output labels. A deep neural network model integrating physical constraints is constructed. A physical regularization term based on the progressive damage evolution model is introduced into the loss function. The deep neural network model is trained through the backpropagation algorithm to solidify the nonlinear mapping relationship of dynamic response. S6. Second-level risk identification and intelligent reinforcement decision-making: After extracting features from new seismic wave signals captured by the monitoring system, the system inputs them into a trained deep neural network model, outputting the predicted values ​​of the time history point safety factor envelope and damage variables for the entire slope section within seconds. Based on the output prediction results and according to the high-risk identification criteria, the system determines the penetration status of the sliding surface and high-risk areas. The system executes intelligent reinforcement strategies, including automatically optimizing the layout of anti-slide piles based on the location of the minimum point safety factor, calculating the effective length of anchor cables based on the depth of damage evolution, and triggering graded early warnings in conjunction with the overall slope safety factor.

2. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 1, characterized in that, In step S2, the delay time model is: ； The safety factor for the time history point is: ,in: ； ； ； ； ； In the formula, For the first The relative propagation time difference of the strip relative to the top of the slope; To go from the top of the slope along the landslide path to the first The cumulative distance between the geometric centers of the strips; Shear wave velocity; The angle between the direction of seismic wave propagation and the positive direction of the landslide; For the first strips in Safety factor at each time point; For the first strips in Anti-slip performance at all times; For the first strips in The tangential glide force at any given moment; For the first strips in The normal force acting on the object at any given moment; For the first The cohesive force of the bottom surface of the strip; For the first The length of the bottom surface of the strip; For the first The internal friction angle of the bottom sliding surface of the strip; For the first The weight of the strip; For the first The angle of inclination of the bottom surface of the strip. For the first Inclination angle of the bottom sliding surface of the strip; For the first The strip comes from the thrust of the previous strip; For the first strips in Horizontal seismic coefficient at time, For the first strips in Vertical seismic coefficient at time; For the first strips in Horizontal acceleration at time t, For the first strips in Vertical acceleration at any given moment; It is the acceleration due to gravity; For seismic waves in Horizontal acceleration at time t, For seismic waves in The vertical acceleration at any given moment.

3. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 2, characterized in that, In step S2, the dynamic force transmission mechanism between strips is as follows: If This indicates that the block itself is stable and does not require downward thrust, i.e. ;like If the strip is in a state of extreme equilibrium or overload, the remaining sliding force must be borne by the next level strip. The safety factor for this strip point is defined as 1. The expression for the inter-strip force is: ； The overall safety factor of the slope is: ； like This indicates that the slope is generally stable at this point. This indicates that the slope is currently in a state of limit equilibrium.

4. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 3, characterized in that, In step S3, the construction process of the asymptotic damage evolution model is as follows: (1) Based on Newmark's rigid body slip theory, calculate the first... strips in The cumulative plastic slip rate is obtained by continuously integrating the relative transient plastic acceleration at time t. ； ； in, For the first strips in The relative transient plastic acceleration at time t; Macauley brackets characterize the plastic response that is elicited only in the presence of unbalanced forces; For the first strips in The cumulative plastic sliding rate at any given time; For integration variables; (2) Define the first The cumulative plastic dissipation work per unit area on the smooth surface of the strip: ； (3) Introducing scalar damage variables Used to characterize the degree of deterioration of the slip surface strength, in the undamaged state. When completely destroyed Based on Kachanov's effective stress equivalence assumption and first-order dynamic attenuation law, an asymptotic damage evolution model is derived: ； in, The energy conversion coefficient is a material constant related to the brittle characteristics of soil and rock media.

5. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 4, characterized in that, In step S3, the reduction and correction method for the shear strength parameters of the soil and rock mass is as follows: ； ； Among them, subscript and These represent the peak and residual values ​​of the intensity, respectively. For the first Peak cohesion of the bottom surface of the strip, For the first The residual cohesion value of the bottom sliding surface of the strip; For the first The peak value of the internal friction angle of the bottom sliding surface of the strip. For the first The residual internal friction angle of the bottom sliding surface of the strip.

6. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 5, characterized in that, In step S4, the reverse calibration method for the energy conversion coefficient is specifically as follows: (1) Define the transient macroscopic damage degree of the test as: ； in, Indicates the post-peak plastic shear displacement; This represents the peak shear force during the experiment. Residual shear force; This refers to the transient shear stress under post-peak plastic shear displacement. (2) Calculate the cumulative plastic dissipation work: ； in, This represents the initial plastic displacement corresponding to the peak stress; (3) The transient macroscopic damage degree measured by the experiment is compared with the cumulative plastic dissipation work. Substituting into the asymptotic damage evolution model and performing algebraic transformations, we obtain: ； (4) Extract the post-peak discrete data points of the experiment, in order to As the independent variable, with Using the coordinate system of the dependent variable, a linear fit is performed through the origin using the least squares method, and the slope of the fitted line is the energy conversion coefficient.

7. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 6, characterized in that, In step S5, the feature vector input to the deep neural network model includes the following features: peak ground acceleration, dominant frequency, effective energy duration, and discrete acceleration time history; The loss function is: ； in, For the total loss, For the true value, For predicted values, To constrain the weights, The physical constraint term for the progressive damage evolution model is expressed as: ； in, The first output of the deep neural network model strips in Predictive damage variables at time points; For the first strips in The cumulative plastic dissipation work over time.

8. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 7, characterized in that, In step S6, determining the sliding surface penetration status and high-risk area based on the output prediction results and high-risk identification criteria includes: The high-risk identification criterion is defined as follows: ； Among them, the The first output of the deep neural network model Predicted value of the safety factor envelope at the time history point of the block; The critical damage threshold determined experimentally in step S4; If the high-risk identification criteria are met, it is determined to be a high-risk unstable node. If consecutive high-risk nodes form a spatial connectivity zone, it is determined to be a potential slip-through area.

9. The intelligent prediction and reinforcement method for slope seismic resistance based on the evolution of safety factor at time history points as described in claim 8, characterized in that, In step S6, the intelligent hardening strategy specifically includes: (1) The predicted value of the safety factor at the time point Global minimum value bar index As the optimal layout axis center for anti-slide piles; (2) The effective length of the prestressed anchor cable is automatically calculated based on the predicted damage evolution depth: ； In the formula, The effective length of the prestressed anchor cable. The reserved length for the anchoring section to extend into the stabilizing slide. The predicted depth of damage evolution; (3) Combine the predicted overall safety factor to trigger graded early warning, so as to achieve a second-level response for emergency rescue of large-scale foundation filling / cutting slopes.

Images