Foundation pit slope intelligent monitoring method, system and device

Abstract

This invention discloses an intelligent monitoring method for foundation pit slopes, belonging to the field of foundation pit monitoring. The method includes: deploying vibration sensors and reference wells in the stable soil area of ​​the slope; dividing the slope into diagnostic sections and soil layer sub-units; and using a reference slip surface to divide the stable area and the area to be diagnosed. A standard excitation device, in conjunction with vibration sensors, is used to calculate the reference wave velocity of each soil layer and the representative wave velocity of each area to be diagnosed. A feature vector containing representative wave velocity, variance, and soil layer coding is constructed and input into a pre-trained soil parameter model. The equivalent cohesion and internal friction angle of each area are output in real time. Based on the reference slip surface, a limit equilibrium equation considering slope geometry, top load, and slope protection structure design anti-sliding force is established, and the overall safety factor of each diagnostic section is iteratively solved. Finally, based on the comparison between the safety factor and a preset threshold, a graded load control decision is automatically generated, realizing quantitative and intelligent proactive control of foundation pit slope stability.

Description

Technical Field

[0001] This invention relates to the field of foundation pit monitoring. More specifically, this invention relates to an intelligent monitoring method, system, and equipment for foundation pit slopes. Background Technology

[0002] Stability monitoring of foundation pit slopes is a core aspect of ensuring project safety. Currently, the industry generally relies on automated monitoring or periodic structural surveys of geometric deformations such as slope surface displacement and settlement. However, this monitoring method is essentially a passive, reactive monitoring based on geometric results. Its core logic is to start from observed displacements, cracks, and other damage phenomena and work backward to deduce the causes of risks.

[0003] Current monitoring methods lack direct consideration of the internal stress state and strength evolution of slopes. Even though some advanced systems achieve real-time data acquisition and automatic alarms for exceeding limits, their alarm thresholds still rely on empirically determined allowable deformation. The alarms respond to the results of mechanical behavior after it has occurred, rather than based on the actual stress and safety reserves within the slope soil and support system. The deformation of slope protection structures is a gradual process from an elastic bearing or non-bearing state to plastic yielding and then to cumulative failure. There is a lag period that may be overlooked by monitoring systems between the stress exceeding its elastic threshold and the appearance of visible displacement sufficient to trigger a traditional alarm. During this lag period, the internal safety reserves of the slope are continuously being depleted, but traditional monitoring data may still show normal values. This leads to a significant delay in early warning, failing to provide valuable lead time for risk intervention.

[0004] Another technical issue lies in the discrepancy in the actual strength performance of soil under complex service environments. Firstly, soil strength values ​​vary depending on its environment. Studies such as "Evolution Model of Shear Strength Index of Unsaturated Soil with Wide Saturation Range" show that the cohesion and internal friction angle of unsaturated soil change with saturation. Spatial differences in moisture content caused by local drainage conditions, surface crack development, vegetation, and temporary loads can lead to significant variations in soil strength across different slope sections. Secondly, soil strength undergoes creep under long-term stress changes during engineering activities, resulting in long-term strength lower than the instantaneous peak strength measured in laboratory tests. Studies such as "Research on Unloading Creep Characteristics of Red Clay and its Impact on Excavated Slope Stability" indicate that under excavation and unloading conditions, the long-term strength of red clay can decrease to 75% to 79% of its original strength. Ignoring this time-dependent strength decay would severely overestimate the long-term safety reserve of the slope. However, obtaining the effective cohesion and internal friction angle of each area of ​​the covered slope in real time and accurately at the construction site relies on high-density, destructive drilling and in-situ testing, which is costly and time-consuming, making it difficult to implement.

[0005] Therefore, breaking through the limitations of traditional monitoring and developing an intelligent monitoring method capable of directly sensing the internal mechanical state of slopes, especially dynamically and quantitatively assessing the impact of surcharges, has become an urgent need in the industry. The ideal technical solution should strive to establish a closed loop from internal state perception to real-time stability quantification and proactive risk control, thereby providing revolutionary technical tools for the balanced management of construction safety and efficiency. Summary of the Invention

[0006] One objective of this invention is to provide an intelligent monitoring method for foundation pit slopes. This method involves calibrating the reference wave velocities of each major soil layer, obtaining representative wave velocities of each soil sub-unit using elastic wave travel time data constructed from matrix excitation on the slope surface and sensor array reception, and then using a pre-trained soil parameter model to predict the equivalent cohesion and equivalent internal friction angle of each region in real time. Furthermore, it integrates the anti-sliding force of the slope protection structure design and the surcharge information at the top of the slope, iteratively solving the limit equilibrium equation to obtain the real-time overall safety factor for each diagnostic section. Finally, based on the comparison between the safety factor and a preset threshold, it automatically generates graded surcharge control decisions, providing a quantitative basis for the refined, intelligent, and proactive management of foundation pit slope construction safety.

[0007] To achieve these objectives and other advantages according to the present invention, according to one aspect of the present invention, an intelligent monitoring method for foundation pit slopes is provided, comprising the following steps: S1. In the stable soil area outside the potential slip zone of the slope, a number of vibration sensors are set up and the three-dimensional coordinates of each vibration sensor are calibrated. A reference well penetrating each main soil layer is set up in the stable soil area. S2. Divide the slope into several continuous diagnostic sections along its length. Within each diagnostic section, divide it into several soil sub-units according to the main soil layers. Using the predetermined reference slip surface as the boundary, divide each soil sub-unit into a stable region located outside the surface and a region to be diagnosed located inside the surface. S3. Use a standard vibration device to strike the reference measuring points corresponding to each major soil layer in the reference well and the vibration points set on the slope of each diagnostic area, and the vibration sensor synchronously collects the vibration signal. S4. Calculate the reference wave velocity of each major soil layer based on the vibration signal obtained from the impact reference point, and calculate the representative wave velocity of each area to be diagnosed through the tomographic inversion algorithm. S5. For each area to be diagnosed, construct an input feature vector containing its representative wave velocity, wave velocity statistical variance, and the encoding of its main soil layer type. Input the input feature vector into a pre-trained soil parameter model. The soil parameter model outputs the equivalent cohesion and equivalent internal friction angle of the area to be diagnosed. S6. For each diagnostic section, based on its reference slip surface, establish a limit equilibrium equation that includes slope geometry parameters, slope top load and slope protection structure design anti-sliding force parameters, and perform iterative calculations to obtain the overall safety factor of the diagnostic section considering the support effect. S7. Based on the overall safety factor of the diagnostic section and the preset load safety threshold, the safety level of the diagnostic section is determined, and a graded load control decision is generated.

[0008] Preferably, the calculation method for the representative wave velocity includes the following steps: A1. For each effective propagation path from any excitation point to any vibration sensor, determine its travel length in the stable region and the region to be diagnosed of each soil sub-unit by ray tracing. A2. Establish the theoretical travel time equations for each effective propagation path, and calculate the formulas as follows: in, i, k, u, n These are the excitation point, main soil layer, soil layer sub-unit, and index marker of the vibration sensor. As the reference wave velocity, To represent wave speed, The travel length within the stable region. The length of travel within the area to be diagnosed. U This represents the total number of soil sub-units traversed by the effective propagation path. The theory is outdated; A3. Combine the travel time equations of all effective propagation paths to form a set of equations with each representative wave velocity as the unknown. Perform iterative inversion to solve the problem with the goal of minimizing the theoretical travel time and the measured travel time. Based on the inversion results, calculate the representative wave velocity of the area to be diagnosed for each soil sub-unit.

[0009] Preferably, the overall safety factor is calculated as follows: Based on the reference slip surface of the diagnostic section, the area to be diagnosed is divided into several vertical blocks. For any vertical block, its bottom slip surface length, slip angle, soil self-weight, top load and bottom pore water pressure are obtained. Based on the equivalent cohesion and equivalent internal friction angle of each soil sub-unit, the comprehensive cohesion and comprehensive internal friction angle of the vertical block are calculated by weighting. The design anti-sliding force provided by the slope protection structure of the diagnostic section and its lever arm about the center of the reference slip surface are obtained. The calculation formula of the limit equilibrium equation is as follows: in, j For the index of the vertical bars, mThis represents the total number of vertical strips. l j The length of the bottom sliding surface. a j For the sliding tilt angle, W j Due to the weight of the soil, ΔW j For top stacking, u j Pore ​​water pressure, C j To enhance cohesion, φ j To take into account the internal friction angle, R The design anti-sliding force provided for the slope protection structure in this diagnostic section R a The radius of the sliding arc of the reference sliding surface. d To design the lever arm of the anti-slip force with respect to the reference slip surface, F s This represents the overall safety factor for the diagnostic section.

[0010] Preferably, the design anti-sliding force and lever arm are determined according to the slope protection structure type in the following manner: a) If the slope protection structure is a static retaining structure, the design anti-sliding force is the maximum horizontal resistance; b) If the slope protection structure is an anchored slope protection structure, the design anti-sliding force is the sum of the horizontal components of the maximum design tension of all anchor rods or anchor cables; c) If the slope protection structure is a shotcrete layer, the design anti-sliding force is zero; The lever arm is the vertical distance from the line of action of the designed anti-slip force to the center of the corresponding arc of the reference slip surface.

[0011] Preferably, step S7 includes: Set a stacking safety threshold greater than 1.0. F A Safety classification and load control are performed according to the following methods: a) If F s If the value is less than 1.0, the diagnostic section is determined to be in an unstable and dangerous state. A control decision is generated to transfer or remove the top load of the diagnostic section. Based on the changed top load, the limit equilibrium equation is recalculated to obtain the verification safety factor. If the verification safety factor increases to 1.0 or above, the control decision is confirmed to be effective. b) If 1.0 < F s < F A If the diagnostic section is determined to be in a critically stable state, a control decision to maintain the existing load will be generated. c) If F A < F s The system determines that the diagnostic section is in a safe reserve state and generates a control decision that allows for an increase in load at the top of the section according to the design capacity.

[0012] Preferably, the excitation point is set on the concrete surface of the slope protection structure. For each vibration sensor, the total energy and main frequency of the signal transmitted to the excitation point are extracted. If, after a certain impact, the total energy of the monitoring signals received by all sensors is lower than the preset energy threshold and the main frequency is lower than the preset main frequency threshold, it is determined that there is a void in the concrete slope protection layer corresponding to the impact point, and the excitation point is marked as an invalid point in the corresponding void area. The monitoring signal data from the invalid point will be removed in subsequent calculations.

[0013] Preferably, the soil parameter model is a BP neural network model optimized by particle swarm optimization algorithm, and adopts a three-layer feedforward network structure, including: an input layer, at least one hidden layer and an output layer; The input layer has multiple input neuron nodes, and the hidden layer has multiple hidden neuron nodes. Each input neuron node is connected to each of the hidden neuron nodes, and each hidden neuron node is connected to the output layer. The neuron nodes in the input layer are used to receive an input feature vector corresponding to a soil sub-unit. The input feature vector includes the representative wave velocity, wave velocity statistical variance, and the encoding of the main soil layer type of the soil sub-unit. The neuron nodes in the output layer are used to output the current equivalent cohesion and the current equivalent internal friction angle of the soil sub-unit.

[0014] A smart monitoring system for foundation pit slopes is provided to perform the aforementioned smart monitoring method for foundation pit slopes, including a slope information acquisition module, a data acquisition module, a data processing module, and a safety analysis module; The slope information acquisition module is used to divide the slope into diagnostic sections and store the design information of each diagnostic section, as well as to monitor and update the surcharge information at the top of the slope. The data acquisition module includes a standard excitation device, a reference well, and a vibration sensor. The standard excitation device is used to strike and generate elastic waves, and the vibration sensor is used to synchronously acquire the vibration signals generated by the strike. The data processing module is connected to the data acquisition module and is used to process the vibration signal. Its functions include: identifying and eliminating invalid point signals, inverting and calculating the representative wave velocity of each soil sub-unit, and outputting the equivalent cohesion and equivalent internal friction angle of each soil sub-unit using the soil parameter model. The safety analysis module is connected to the slope information acquisition module and the data processing module respectively. By iteratively solving the limit equilibrium equation, it calculates the overall safety factor of each diagnostic section and generates graded safety judgment and load control decision based on the overall safety factor.

[0015] An electronic device is provided, including a memory and a processor. The memory stores computer instructions, and the processor executes the computer instructions to perform the above-described intelligent monitoring method for foundation pit slopes.

[0016] The present invention has at least the following beneficial effects: First, this method establishes a quantitative monitoring system for slope condition based on in-situ physical benchmarks. By deploying benchmark wells and sensor arrays in stable soil areas and using standard excitation devices to obtain benchmark wave velocities for each major soil layer, a traceable and high-precision initial physical field reference is provided for the entire monitoring system. This technical feature ensures that all subsequent dynamic monitoring data can be standardized by comparing them with the stable benchmark, thereby overcoming the interference caused by the natural spatial variability of soil on the monitoring results and improving the consistency and comparability of long-term monitoring data.

[0017] Second, this method achieves intelligent real-time mapping from the internal physical field of the slope to geotechnical parameters. Representative wave velocities of each soil sub-unit are obtained through a tomographic inversion algorithm, and feature vectors containing wave velocity, statistical variance, and soil type are constructed. These vectors are input into a pre-trained soil parameter model, directly outputting the equivalent cohesion and equivalent internal friction angle of each unit. This deeply integrates geophysical exploration with geotechnical analysis, solving the problem that traditional methods cannot obtain spatially distributed mechanical parameters in real-time, continuously, and non-destructively at the engineering site.

[0018] Third, this method enables dynamic and quantitative stability assessment of the impact of surcharge and other time-varying factors. By uniformly substituting real-time acquired equivalent strength parameters, geometric parameters, and continuously updated surcharge information at the slope crest into the limit equilibrium equation considering the anti-sliding force of the slope protection structure for iterative solution, the overall safety factor reflecting the current comprehensive working conditions can be calculated. This allows the assessment results to dynamically respond to construction activities and environmental changes, revealing the direct correlation between surcharge location, magnitude, and slope stability, and providing a quantitative basis for the refined management of construction loads.

[0019] Fourth, this method forms a complete decision-making closed loop from state perception and safety assessment to proactive control. Based on the calculated overall safety factor and preset thresholds, the system can automatically generate clear graded load control decisions. This enables monitoring conclusions to be directly and immediately transformed into specific engineering instructions for controlling slope instability risks, realizing a fundamental shift from passive early warning to proactive intervention based on real-time model prediction.

[0020] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0021] Figure 1 This is a flowchart of one technical solution of this application; Figure 2 This is a schematic diagram of the principle of this application, wherein (a) is a schematic diagram of the component arrangement and (b) is a schematic diagram of the limit equilibrium equation parameters. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0023] It should be noted that, unless otherwise specified, the experimental methods described in the following embodiments are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified. In the description of this invention, the orientation or positional relationship indicated by the terms is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this invention and simplifying the description. It does not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0024] like Figure 1 As shown, this invention provides an intelligent monitoring method for foundation pit slopes, comprising the following steps: S1. Within the stable soil region outside the potential slip zone of the slope, several vibration sensors are deployed and their three-dimensional coordinates are calibrated. Reference wells penetrating all major soil layers are installed within this stable soil region. Specifically, the stable soil region must be determined based on a comprehensive assessment of the detailed geological survey report and slope stability analysis results. The stable soil region should be located a certain safe distance outside the potential slip zone initially delineated through limit equilibrium analysis such as the simplified Bishop method. Geological survey data for this region should show continuous soil layers with relatively homogeneous properties, and no known weak interlayers, tectonic fracture zones, or significant engineering disturbance history. The determination of the stable soil region provides a theoretically and empirically reliable reference system for the entire monitoring system.

[0025] The selection and deployment of vibration sensors must meet the requirements for effective acquisition of elastic wave signals propagating in the soil. Considering that the dominant frequency of the excitation signal is usually in the range of tens to thousands of hertz, and the signal attenuates rapidly during propagation in the soil, vibration sensors suitable for soil and rock media with high sensitivity and appropriate frequency response should be selected. In engineering practice, commercially available sensors include, but are not limited to: the GS-11D low-frequency detector (frequency response range 4.5Hz-1000Hz) manufactured by Geospace Technologies, USA; the MPA series accelerometers from Beijing Shengwang Acoustic & Electrical Technology Co., Ltd.; or similar embeddable three-component vibration sensors. These vibration sensors generally adopt a sealed structure and can be directly buried in shallow holes. Through sand backfill coupling, they can effectively receive and convert mechanical vibration signals generated by impact that propagate in the form of longitudinal waves (P-waves) and transverse waves (S-waves) in the soil. The vibration sensors are deployed in an array to form multi-angle ray coverage of the monitoring area. The number of vibration sensors on each side of the foundation pit slope should not be less than 6. The specific locations should be arranged in a grid or ring pattern in the pre-drilled boreholes, based on the slope geometry and key monitoring areas. The three-dimensional spatial coordinates (X, Y, Z) of each vibration sensor should be determined using a total station to measure the plane coordinates and elevation, and the calibration accuracy should be controlled within ±10 mm.

[0026] A reference well aims to provide a known, repeatable in-situ wave velocity calibration benchmark for each major soil layer. During engineering implementation, existing engineering wells penetrating the target soil layers should be prioritized, such as manholes constructed for dewatering of foundation pits or geological monitoring wells for other purposes. These wells are typically deep enough and located far from the slope, with complete construction records, and can be used directly as reference wells, thus avoiding additional drilling and saving costs and time. If such existing wells are lacking in the slope area, a dedicated reference well must be constructed. This well should be located within a stable soil region near the sensor array, and its depth must penetrate all major soil layers involved in the slope, down to below the potential slip surface. The well wall can be casing-lined, with coupling material backfilled between the casing and the borehole wall to ensure effective excitation and propagation of elastic waves. Along the well depth direction, at the typical depth of each target major soil layer, a clear mark or dedicated pad should be placed on the well wall as a "reference measurement point" for that major soil layer.

[0027] S2. Divide the slope along its length into several continuous diagnostic sections. Within each diagnostic section, divide it into several soil sub-units according to the main soil layers. Using a pre-determined reference slip surface as the boundary, divide each soil sub-unit into a stable region located outside the surface and a region to be diagnosed located inside the surface. Specifically, the division of diagnostic sections is mainly based on the slope's engineering geological conditions, slope protection structure layout, geometric shape, and spatial variability of load distribution.

[0028] The boundaries of diagnostic sections should be set at the design segmentation of the slope protection structure, the inflection point of the slope ratio or height, the boundary of significant changes in the range or properties of the surcharge at the top of the slope, and the location where the distribution or properties of the soil layers revealed by geological surveys show significant changes. The length of a single diagnostic section along the slope length should preferably be controlled between 10 and 20 meters. Excessively wide diagnostic sections may obscure local weak points, making precise location impossible, while excessively narrow divisions will generate too many computational units, increasing the computational load and potentially leading to non-convergence or reduced accuracy of the tomographic inversion results due to insufficient coverage of the vibration sensor ray paths within each diagnostic section.

[0029] Vertically, each diagnostic section needs to be further subdivided into sub-units based on the main soil layers revealed in the geological survey report. Because different soil layers (such as fill, cohesive soil, sand, and weathered rock) have different physical and mechanical properties, their elastic wave velocities and shear strength parameters also differ. If the mixture of different soil layers is analyzed as a homogeneous body, it will not accurately reflect the wave field propagation and the distribution of mechanical parameters in the depth direction. Dividing the soil layers into sub-units based on the main soil layers ensures the relative homogeneity of the medium within each sub-unit.

[0030] The reference slip surface refers to the shear failure surface most likely to occur in the diagnosed section, determined in advance through theoretical calculations. It is typically circular or non-circular in shape. This surface is primarily derived from the stability verification results during the slope engineering design phase. Designers can use professional geotechnical engineering software such as GEO-Slope / W, SLIDE, and Lizheng Geotechnical Calculation Software to calculate the profile of the diagnosed section and search for the critical slip surface corresponding to the minimum safety factor, which serves as the reference slip surface for the monitoring phase. During the monitoring period in the construction and normal use phases, as long as the slope does not exhibit accelerated deformation or obvious signs of progressive failure, its potential failure mode exhibits relative stability. Using this geometric surface determined in the design phase as a fixed analysis boundary throughout the entire monitoring period is reasonable and efficient. This allows the core objective of monitoring to focus on assessing the safety reserve provided by the current actual soil strength along the established potential failure surface, rather than re-searching for the slip surface at each monitoring session, thus achieving stability and efficiency in the calculation process.

[0031] In setting the benchmark slip surface, the calculation model adopts the designed excavation shape of the slope, and the soil parameters are selected from the standard values ​​of strength indicators provided in the geological survey report or lower values ​​based on experience, to simulate an unfavorable and conservative initial state. The existence of the slope protection structure is not considered in this stage of the analysis. This is because, based on the fundamental principles of engineering safety analysis and design, the purpose of determining the benchmark slip surface is to find the inherent and most dangerous potential failure mode of the slope under the most unfavorable working conditions without any intervention. This surface represents the bottom line or intrinsic weak path of slope stability. Only by evaluating this fixed potential failure mode, which is unaffected by external reinforcement measures, and subsequently introducing the current actual soil strength obtained from real-time inversion in step S5, the slope crest load, and the anti-slip force of the support structure, can the calculated dynamic safety factor be comparable.

[0032] Each soil sub-unit within the diagnostic section is a three-dimensional geological body defined by its top, bottom, and side surfaces. The determined reference slip surface, a spatial geometric surface, is then subjected to Boolean operations or spatial cutting with the three-dimensional model of the soil sub-unit. This divides the soil sub-unit into two parts: the portion located outside the reference slip surface (away from the slope and towards the depth of the soil) is defined as the stable region, and the portion located inside the reference slip surface (usually closer to the slope and the side where slippage may occur) is defined as the region to be diagnosed.

[0033] The theoretical basis for dividing soil sub-units into stable and undiagnosed regions stems from the core assumption of limit equilibrium analysis in geotechnical engineering, which considers the reference slip surface as the potential mechanical discontinuity boundary when the slope undergoes overall shear failure. Based on this mechanical mechanism, reasonable inferences can be made by combining the actual field conditions of the foundation pit slope: during the construction period, the stockpiling of materials and engineering disturbances within the foundation pit are all located at the top of the slope, and the most critical mechanical state evolution affecting slope stability is mainly concentrated within the potential slip body, i.e., the undiagnosed region. In the stable region, because the soil is far from the potential slip body, the influence of stress state and environmental changes is relatively small; therefore, its physical and mechanical properties can be considered relatively stable or slowly changing during the monitoring period. For the stable region, its internal wave velocity and other properties can be reliably represented in situ by reference wells located in the same stable soil mass. However, for the undiagnosed region, which is in an active state, high-density wave velocity tomography inversion must be performed to capture its state changes in real time.

[0034] S3. Using a standard vibration device, strike the benchmark measuring points corresponding to each major soil layer within the benchmark well and the vibration points set on the slope surface of each diagnostic area. Vibration sensors synchronously collect vibration signals. Specifically, to achieve stable and repeatable vibration of the slope and vertical measuring points, the standard vibration device can be a power device with controllable impact energy. Specifically, an electric hammer or spring hammer capable of overcoming the limitations of gravity can be used to apply effective instantaneous impact or periodic vibration to vertical or inclined slopes. To achieve automated operation, the selected electric hammer or spring hammer can be integrated into the end of a small robotic arm controlled by a servo motor. This robotic arm can be mounted on a small loader or mobile track, moved, positioned, and the standard vibration device vertically pressed against the measuring point to trigger the impact, thereby replacing manual operation and ensuring the safety and accuracy of the vibration point location on steep slopes.

[0035] The benchmark measuring points are located inside the benchmark well and are used to determine the fixed excitation locations for in-situ wave velocities in each major soil layer. During implementation, benchmark measuring points can be marked on the inner wall of the benchmark well at typical depths corresponding to each major soil layer to ensure effective wave excitation and propagation. Excitation points on the slope are systematically deployed on the slope protection structure or exposed slope surface according to the division of the diagnostic section. The excitation points can be arranged in a grid pattern, and their planar positions and elevations are accurately measured and marked using a total station. The structural positions corresponding to each excitation point and benchmark measuring point can be leveled and small steel shims can be attached to ensure that impact energy is effectively transmitted to the soil while protecting the slope surface.

[0036] When using an electric hammer, the impact force is controlled by adjusting its impact energy level or impact frequency. If using a spring hammer, the impact energy is controlled by calibrating the spring's compression stroke or replacing it with a spring of different stiffness. The control principle is to ensure that the excited elastic wave has sufficient energy to penetrate the target depth and can be clearly received by all or most vibration sensors, without damaging the slope protection structure, while maintaining consistency in the input energy and spectral characteristics of each strike. Before formal monitoring, tests should be conducted at typical locations to determine the optimal impact parameters, and these parameters should be consistently used throughout the monitoring cycle.

[0037] All vibration sensors are connected to a multi-channel high-speed data acquisition unit via cables or a high-precision wireless synchronization module. At the instant each impact is triggered by the standard excitation device, its built-in trigger sensor (such as a force sensor or accelerometer) generates a TTL electrical pulse signal with a steep rising edge. This pulse signal is sent in real-time as the "trigger signal" to the external trigger port of the data acquisition unit. The data acquisition unit uses this pulse signal as the absolute time zero point to synchronously start data recording across all channels. Modern data acquisition units achieve microsecond-level synchronization accuracy between channels, meeting the requirements for elastic wave travel time measurement. In this way, the system can accurately record the vibration waveform sequence originating from the same source and propagating to different vibration sensors, ensuring that all waveforms have a unified time starting point, thus laying the foundation for subsequent accurate calculation of wave propagation time in the soil.

[0038] S4. Based on the vibration signals obtained from striking the benchmark measuring points, the reference wave velocity of each major soil layer is calculated. The representative wave velocity of each area to be diagnosed is then calculated using a tomographic inversion algorithm. Specifically, the synchronously acquired vibration signals are converted into wave velocity parameters that quantitatively describe the soil state, including the reference wave velocity of each major soil layer and the representative wave velocity of each area to be diagnosed. The reference wave velocity is obtained by striking the benchmark measuring points corresponding to specific major soil layers in benchmark wells located within stable soil areas and recording the vibration signals received by the vibration sensors. The principle is that the propagation speed of elastic waves in a homogeneous medium is an inherent property of the medium and is independent of the excitation and reception locations. By measuring the propagation time of the elastic wave generated from the benchmark measuring point to the vibration sensor at known coordinates, the average wave velocity along that path can be directly calculated. By selecting multiple independent propagation paths composed of different vibration sensors for calculation and statistical analysis, random errors can be eliminated, ultimately obtaining a reference wave velocity value that reliably represents the current state of the major soil layer.

[0039] For the area to be diagnosed, the soil condition may change due to variations in structural loads within the foundation pit, excavation unloading, seepage, creep, and other factors, making it impossible to directly use the reference wave velocity from the stable region. Therefore, a tomographic inversion algorithm is needed to obtain the wave velocity in these areas. By constructing numerous intersecting ray paths composed of a large number of excitation points and vibration sensors, the wave propagation time of each ray path can be accurately measured. Combined with a known geometric model, an inversion problem can be constructed to solve for the internal wave velocity distribution of the medium that best matches the theoretically calculated travel time with the actual measured travel time.

[0040] The calculation method for the representative wave velocity includes the following steps: A1. For each effective propagation path from any excitation point to any vibration sensor, its travel length within the stable region and the region to be diagnosed in each soil sub-unit is determined by ray tracing. First, from all excitation records, signals with high signal-to-noise ratio and clear initial and final values ​​are selected to determine each usable propagation path from the excitation point. i To vibration sensor n The wave propagation path was then determined. Subsequently, ray tracing technology was used to determine the specific trajectory of each effective propagation path within the established three-dimensional slope geometry model. Ray tracing accurately calculates the travel length of the wave propagation path within each soil sub-unit, and the travel length data of all wave propagation paths constitute the geometric foundation data required for the inversion.

[0041] A2. Establish the theoretical travel time equations for each effective propagation path, and calculate the formulas as follows: in, i, k, u, n These are the excitation point, main soil layer, soil layer sub-unit, and index marker of the vibration sensor. As the reference wave velocity, To represent wave speed, The travel length within the stable region. The length of travel within the area to be diagnosed. U This represents the total number of soil sub-units traversed by the effective propagation path. The theory is outdated; A3. Simultaneously solve the theoretical travel time equations for all effective propagation paths to form a system of equations with representative wave velocities as unknowns. Perform iterative inversion to minimize the difference between theoretical and measured travel times. Based on the inversion results, calculate the representative wave velocity of the area to be diagnosed for each soil sub-unit. The specific implementation process is as follows: First, a system of linear equations is established. All the travel time equations established in step A2 for each effective propagation path are combined. Since multiple excitation points and sensors are deployed in the monitoring, the number of effective propagation paths formed between them is typically hundreds or even more, while the number of representative wave velocities as unknowns in each region to be diagnosed is usually only a few dozen. Therefore, combining the equations for all these effective propagation paths results in an overdetermined system of linear equations with a number of equations far exceeding the number of unknowns. In this system, each equation describes the total time for a wave to propagate along a specific effective propagation path. The known quantities in all equations include: the measured total propagation time for each effective propagation path, the travel length of the elastic wave in each region to be diagnosed and the stable region, and the known reference wave velocity used by the elastic wave when traveling through the stable region. An iterative inversion algorithm is used to solve this overdetermined system of equations. The goal of the solution is to find a set of representative wave velocity values ​​such that, after substituting these values ​​into all equations, the overall difference between the calculated theoretical total propagation time and the measured total propagation time is minimized. The least squares method is usually used to define this, which requires that the sum of the squares of the differences between the theoretical travel time and the measured travel time of all effective propagation paths be minimized.

[0042] The solution process is completed iteratively by computer. Starting with an initial wave velocity model, in each iteration, the computer calculates the theoretical travel time for all paths based on the current wave velocity model and compares it with the measured travel time to obtain the error. Then, the algorithm adjusts the wave velocity value for each area to be diagnosed according to a specific mathematical rule, generating an updated wave velocity model that is theoretically closer to reality. The next iteration begins with this new model, and the error is calculated and the wave velocity is adjusted again. This process is repeated until the overall error between the theoretical and measured travel times no longer decreases significantly, or the preset maximum number of iterations is reached, at which point the calculation is considered converged. Finally, representative wave velocity results are output. When the iteration converges, the final set of wave velocity values ​​is the optimal representative wave velocity estimate for each soil sub-unit of the area to be diagnosed. This set of wave velocity spatial distribution results quantitatively characterizes the current physical state of potential slip bodies within the slope.

[0043] The iterative inversion calculation process can be implemented programmatically or using mature commercial or professional scientific computing software. For example, MATLAB or Python (with scientific computing libraries such as NumPy and SciPy) can be used for custom development of matrix construction and least-squares iterative solutions. In addition, some professional geophysical data processing software (such as Voxler from Golden Software, and some specialized wave field tomography modules) also provide general frameworks and efficient algorithms for constructing and solving such inversion problems. After appropriate model parameter configuration and input data formatting, these can be used for this step of the calculation.

[0044] S5. For each area to be diagnosed, construct an input feature vector containing its representative wave velocity, wave velocity statistical variance, and the code of the main soil layer to which it belongs. Input the input feature vector into a pre-trained soil parameter model. The soil parameter model outputs the equivalent cohesion and equivalent internal friction angle of the area to be diagnosed. A soil parameter model is a pre-trained mathematical mapping model used to establish the complex relationship between easily measurable physical characteristics and target mechanical parameters. Besides neural network models, soil parameter models can also be trained using support vector machines, random forests, or other mature machine learning methods, depending on the data foundation and engineering requirements. These methods can learn and solidify the mapping rules from input features to output strength parameters by analyzing historical engineering data or numerical simulation sample libraries.

[0045] The wave velocity statistical variance is calculated based on the dispersion of the wave velocity estimates for the area to be diagnosed during the tomographic inversion process. Elastic wave velocity directly reflects the compactness and stiffness of the soil, which are key physical state indicators determining strength; the statistical variance quantifies the confidence level of the estimate and the homogeneity of the soil; and the principal soil layer type coding provides crucial soil type boundaries and prior knowledge. The combination of these three factors enables the soil parameter model to reliably infer the current equivalent cohesion and equivalent internal friction angle of the area to be diagnosed within a comprehensive and physically meaningful feature space.

[0046] In one technical solution, the soil parameter model is a BP neural network model optimized by particle swarm optimization algorithm, which adopts a three-layer feedforward network structure, including: an input layer, at least one hidden layer and an output layer. The input layer has multiple input neurons, and the hidden layer has multiple hidden neurons. Each input neuron is connected to a specific hidden neuron, and all hidden neurons are connected to the output layer. The neurons in the input layer receive an input feature vector corresponding to a soil sub-unit. This input feature vector includes the representative wave velocity, wave velocity statistical variance, and the encoding of the dominant soil layer type for that sub-unit. The neurons in the output layer output the equivalent cohesion and equivalent internal friction angle of that sub-unit. The particle swarm optimization (PSO) algorithm is introduced to find a better set of initial parameters before the neural network undergoes backpropagation training. PSO simulates the social cooperative behavior of a flock of birds, enabling efficient global parallel search in the solution space by sharing individual and group historical best experiences among particles. Using this algorithm to globally optimize the initial weights and thresholds of the neural network provides a starting point closer to the global optimum for subsequent fine-tuning training. This combination retains the nonlinear mapping and learning capabilities of BP neural networks while significantly improving the model's convergence speed, stability, and final generalization performance.

[0047] The following describes the process of establishing a soil parameter model using a particle swarm optimization neural network algorithm: The first step is the construction of the sample dataset. A large number of "feature-intensity" paired samples are systematically collected. The input feature of each sample is a vector containing representations of wave velocity, wave velocity statistical variance, and the encoding of the main soil layer type. The target output is the actual cohesion and actual internal friction angle of the corresponding soil, obtained through standard indoor tests. Data sources include historical engineering survey reports, specialized scientific research experiments, or virtual sample libraries generated through high-fidelity numerical simulations. All samples undergo preprocessing such as normalization and are randomly divided in a certain ratio (e.g., 7:2:1) into a "training set" for training the model, a "validation set" for monitoring the training process and preventing overfitting, and a "test set" for final performance testing.

[0048] Next, the neural network structure and optimization strategy are defined. A classic three-layer feedforward network structure is adopted. The number of nodes in the input layer corresponds to the dimension of the feature vector, and the output layer has two nodes that predict cohesion and internal friction angle, respectively. The number of nodes in the hidden layer is determined through trial and error to balance fitting ability and generalization. To overcome the tendency of traditional neural network training to get trapped in local optima, a particle swarm optimization algorithm is introduced to globally optimize the initial weights and thresholds of the network. This algorithm treats a set of network parameters as a "particle" and searches the solution space for the parameter combination that minimizes the initial prediction error of the network on the training set by simulating the social cooperative behavior of a flock of birds, thus providing a superior starting point for subsequent training.

[0049] Next, a backpropagation (BP) neural network is trained based on the optimized starting point. Using the optimal parameters found by the particle swarm optimization algorithm in the previous step as initial values, the network is trained using supervised error backpropagation on the training set data. During this process, forward propagation generates predicted values; the error between the predicted and true values ​​is calculated, and this error is backpropagated and used to adjust all network weights and thresholds layer by layer according to the gradient descent rule. The training process is monitored synchronously on the validation set; training is stopped early when the validation error no longer decreases, thus preserving the model state with the best generalization performance and avoiding overfitting to the training data.

[0050] Finally, the model is validated and confirmed. A completely independent "test set" unused during the training process is used to perform a final evaluation of the trained model. Evaluation metrics must include root mean square error and mean absolute error. Only when the model's prediction accuracy meets the preset engineering application standards (e.g., cohesion error within ±5 kPa, internal friction angle error within ±2 degrees) can the soil parameter model be confirmed as successfully established and deployed in a real-world monitoring system, enabling intelligent, real-time conversion from elastic wave velocity characteristics to shear strength parameters.

[0051] S6. For each diagnostic section, based on its reference slip surface, establish a limit equilibrium equation that includes slope geometry parameters, slope top load and slope protection structure design anti-sliding force parameters, and perform iterative calculations to obtain the overall safety factor of the diagnostic section considering the support effect. The core of step S6 is to combine the soil strength parameters obtained in the preceding steps, which reflect the current state, with the slope's geometry, load, and support conditions. Using classical limit equilibrium theory, the real-time overall stability safety factor for each diagnostic section is quantitatively calculated. The benchmark slip surface determined during the design phase is used as the fixed analysis object because, within the regular monitoring period, as long as the slope does not undergo significant deformation or failure, the location of its most dangerous potential slip mode exhibits relative stability.

[0052] The overall safety factor can be calculated based on the classical limit equilibrium theory framework. The core idea of ​​this framework is that, for a given slip surface, when the slope is in limit equilibrium, all forces or moments acting on the potential slip body that cause it to slide down should be in balance with all forces or moments resisting its sliding down. Its general mathematical expression can be summarized as: the overall safety factor equals the ratio of the total resisting force to the total sliding force along the entire slip surface. In slice methods such as the simplified Bishop method, this principle is concretized by analyzing the forces acting on each vertical slice and establishing equilibrium equations about the center of the entire slip arc to solve for the safety factor. Since the safety factor itself is a reduction factor for the shear strength parameter, it appears directly in the formula for calculating the resisting force, causing the safety factor, an unknown variable, to appear on both sides of the equation, making direct analytical solution impossible. Therefore, an iterative calculation method must be adopted: first, assume an initial value of the safety factor and substitute it into one side of the equation for calculation, compare the result with the other side, and then adjust the assumed value according to the difference. Repeat this iterative process until the calculated anti-skid force and sliding force reach equilibrium. At this time, the corresponding safety factor value is the desired convergent overall safety factor.

[0053] Classical limit equilibrium theory provides a basic framework for slope stability analysis, but its standard equations have significant limitations in meeting the needs of intelligent real-time monitoring. It cannot integrate dynamically changing field parameters, resulting in static and lagging assessment results. Therefore, adaptive improvements are needed. By reconstructing the input sources of key parameters, real-time sensed soil conditions, dynamic loads, and fixed support parameters are integrated into the same dynamic equilibrium equation, thereby achieving online and quantitative assessment of slope stability. In one technical solution, the overall safety factor is calculated as follows: Based on the reference slip surface of the diagnostic section, the area to be diagnosed is divided into several vertical blocks. For any vertical block, its bottom slip surface length, slip angle, soil self-weight, top load, and bottom pore water pressure are obtained. Based on the equivalent cohesion and equivalent internal friction angle of each soil sub-unit, the comprehensive cohesion and comprehensive internal friction angle of the vertical block are calculated using weighted averages. The design anti-slip force parameters provided by the slope protection structure of the diagnostic section and its lever arm about the center of the reference slip surface are obtained. The formula for calculating the limit equilibrium equation is as follows: in, j For the index of the vertical bars, m This represents the total number of vertical strips. l j The length of the bottom sliding surface. a j For the sliding tilt angle, W j Due to the weight of the soil,ΔW j For top stacking, u j Pore ​​water pressure, C j To enhance cohesion, φ j To take into account the internal friction angle, R The design anti-sliding force provided for the slope protection structure in this diagnostic section R a The radius of the sliding arc of the reference sliding surface. d To design the lever arm of the anti-slip force with respect to the reference slip surface, F s This represents the overall safety factor for the diagnostic section.

[0054] Unlike conventional analyses that use fixed strength parameters and neglect or simplify the role of support, this method integrates real-time soil strength parameters obtained through dynamic inversion with quantitative slope protection structural design resistance into the same equilibrium equation for iterative solution. This results in a final overall safety factor that is a dynamic and integrated stability index that simultaneously reflects the current actual mechanical state of the soil and the true contribution of the support system.

[0055] The sources and acquisition methods of each parameter in the formula are clear and operable, as detailed below: Bottom surface length l j and sliding tilt angle a j The geometry is derived from the reference sliding surface. By discretizing the sliding body into vertical blocks, the bottom sliding surface length and sliding angle of each vertical block can be obtained analytically or numerically based on the mathematical equation of the surface and the boundary coordinates of the blocks.

[0056] Load parameters W j The soil unit weight is calculated based on the volume of each soil sub-unit within the vertical block and its natural unit weight, which is a known geological exploration parameter.

[0057] Top stack ΔW j The data is derived from real-time monitoring of the loading area at the top of the slope and is then distributed to the corresponding vertical blocks below according to the area.

[0058] Pore ​​water pressure u j The value can be obtained through real-time monitoring by installing pore water pressure gauges within the slope, or calculated based on preset hydrogeological conditions and seepage analysis models. When the drainage of the foundation pit slope is good, the water level is lower than the bottom elevation of the pit, and there has been no rainfall for a long time, the value can be taken as 0.

[0059] For each vertical strip, the equivalent cohesion and equivalent internal friction angle of each soil sub-unit it passes through are weighted and averaged according to the proportion of the volume or area occupied by each soil sub-unit within the vertical strip, thus obtaining the comprehensive cohesion representing the overall shear strength of the vertical strip. C j and combined internal friction angle φ j .

[0060] Overall safety factor F s It is the objective of solving the limit equilibrium equations. Because F s Since the result is embedded on both sides of the equation and cannot be obtained directly analytically, an iterative algorithm must be used. During calculation, first assign... F s An initial assumed value is substituted into the equation for iterative calculation, continuously updating the value until torque equilibrium is achieved on both sides of the equation. The convergent solution obtained at this point represents the overall safety factor reflecting the current operating condition. This iterative solution process can be fully automated by a computer program, ensuring computational efficiency and reliability.

[0061] When determining support parameters, their contribution to slope stability must be systematically integrated into the limit equilibrium analysis framework based on mechanical principles. Design anti-sliding force. R The determination of the lever arm can be achieved by equating the active reinforcement mechanism of the anchor bolts or the passive bearing mechanism of the retaining structure to a concentrated resistance acting on the sliding body. The lever arm follows the geometric relationship of moment balance, characterizing the leverage effect of the anti-sliding moment formed by this concentrated resistance on the center of the sliding arc. The coordinated determination of both ensures that the stabilizing effect provided by the support system can be quantitatively and accurately incorporated into the dynamic calculation of the overall safety factor, forming a key theoretical link from "structural design" to "overall stability assessment".

[0062] The design anti-sliding force and lever arm are determined according to the slope protection structure type as follows: a) If the slope protection structure is a static retaining structure, the design anti-sliding force is the maximum horizontal resistance, specifically, the maximum horizontal resistance. In engineering practice, the maximum horizontal resistance can be set as the maximum horizontal reaction force that the structure can provide under the conditions of elastic stress of the material or the ultimate limit state of bearing capacity allowed by the design code.

[0063] From a mechanical perspective, static retaining structures such as retaining walls and retaining piles primarily play a passive restraining role in slope stability. When a potential sliding body tends to displace, the retaining structure itself passively resists the horizontal thrust from the soil through its own stiffness and foundation embedment. Therefore, its ultimate contribution to slope resistance directly depends on the structure's own bending, shear, and overturning bearing capacity, as well as its interaction with the foundation.

[0064] While simplified formulas based on Rankine or Coulomb earth pressure theories can be used to estimate the maximum horizontal resistance, finite element horizontal pushover analysis is a superior choice for more realistically reflecting the actual stress on the structure when it reaches its ultimate limit state, especially considering material nonlinearity, plastic hinge development, and soil-structure interaction. This method accurately simulates the entire process from elastic to elastoplastic to failure by gradually increasing the lateral displacement of the structure, thus obtaining a more scientific and reliable maximum horizontal resistance. R .

[0065] In limit equilibrium analysis, the lever arm d Defined as the vertical distance from the line of action of the maximum horizontal resistance R to the center O of the predetermined reference slip surface. Determine the lever arm. d When doing so, it is necessary to first determine the location and direction of the point of application of the resultant force of the retaining structure's resistance based on the design drawings or finite element analysis results, and then obtain the vertical distance through geometric drawing or analytical calculation. d The accurate determination of this ensures that the anti-slip moment generated by the support structure can be correctly included in the overall moment balance equation.

[0066] (b) If the slope protection structure is an anchored slope protection structure, the design anti-sliding force is the sum of the horizontal components of the maximum design tension of all anchor rods or cables. Specifically, from the perspective of construction and force principle, the anchoring ends of the anchored slope protection structure are all located in a stable region outside the potential sliding surface. Prestress is established in the anchor rods or cables through tensioning, thereby causing the capping beam or related structure to actively apply a compressive stress to the anchored soil. Therefore, in the overall moment balance analysis, its contribution can be equivalently simplified to a concentrated horizontal resistance acting on the sliding body.

[0067] The horizontal anti-slip force provided by a single anchor bolt is the cosine of its maximum design tension force multiplied by the anchor bolt inclination angle. The total design anti-slip force per unit width of this section is obtained by arithmetically superimposing these horizontal components of all anchor bolts within the calculation section. R For the lever arm d The determination of this is based on the aforementioned horizontal anti-slip resultant force. R Line of action to the center of the reference slip surface O The vertical distance.

[0068] c) If the slope protection structure is a shotcrete layer with a design anti-sliding force of zero, specifically, the shotcrete layer mainly serves to seal the surface, prevent weathering, prevent erosion, and suppress localized spalling and peeling of the shallow surface. It is thin, has limited stiffness, and lacks strong mechanical anchoring to the deeper soil and rock mass. When the slope experiences overall sliding along the deep reference sliding surface, this thin structure cannot provide effective lateral restraint, and its shear capacity is negligible compared to the overall sliding force. Considering these stress characteristics, to avoid damage to this type of slope protection structure under continuous stress, the design anti-sliding force of the shotcrete layer is conservatively and reasonably set in the limit equilibrium analysis for deep overall stability. R The value is set to zero. This clarifies the position of this type of support in this assessment system, namely, it is not included in the overall anti-sliding force, thereby avoiding overestimation of the safety reserve of slopes using this type of slope protection structure and ensuring that the assessment results are biased towards safety and reliability.

[0069] S7. Based on the overall safety factor of the diagnostic section and the preset surcharge safety threshold, perform a safety classification judgment on the diagnostic section and generate a graded surcharge control decision. The overall safety factor calculated in step S6, reflecting the current comprehensive operating conditions, is then used. F s With a pre-set safety threshold for stacking management F A A comparison is made. Based on the comparison results, an automated safety classification is performed on each diagnostic section, and clear, actionable graded load control decisions are generated accordingly. This realizes a complete process from "condition monitoring" to "risk assessment" and then to "proactive intervention," upgrading traditional monitoring and alarms to intelligent control based on quantitative analysis.

[0070] In one technical solution, a load safety threshold greater than 1.0 is set. F A Safety classification and load control are performed according to the following methods, and the load safety threshold is determined. F A This is a control indicator for surcharge management. Its value needs to be determined comprehensively based on the project's importance level, design safety margin, geotechnical parameter uncertainties, and local engineering experience, and is typically between 1.15 and 1.30. The purpose of setting a surcharge safety threshold is to establish a buffer zone between theoretical stability and the safe operating space allowed by actual engineering, thereby implementing more refined risk management. Based on the surcharge safety threshold, the following three-level judgment and control framework is established.

[0071] a) If F s<1.0, it is determined that the diagnostic section is in a state of instability risk, and a control decision to transfer or remove the surcharge load on the top of this section is generated. Based on the changed top surcharge load parameters, substitute them back into the limit equilibrium equation for calculation to obtain the verification safety factor. If the verification safety factor is increased to 1.0 or above, it is confirmed that the control decision is effective. Specifically, if the calculated overall safety factor F A <1.0, which indicates that under the current working conditions, the sliding force of the slope along the reference slip surface has exceeded the anti-sliding force, and theoretically it has lost balance. The stress on the slope protection structure is beyond its safe bearing stress, and there is an immediate or potential overall instability risk for the slope protection. The system will automatically generate an emergency control decision of "transfer or remove the surcharge load on the top of this section". To ensure the effectiveness of the decision, the system will immediately re-call the limit equilibrium equation for rapid verification based on the changed surcharge load parameters (i.e., the load value after removal or transfer). If the verification safety factor obtained from the verification is increased to 1.0 or above, it proves that this unloading measure can effectively make the slope return to the basic stable state, thus confirming that the control decision is effective and should be immediately executed.

[0072] b) If 1.0 < F s < F A , it is determined that the diagnostic section is in a critical stable state, and a control decision to maintain the existing surcharge load is generated; if the overall safety factor satisfies 1.0 < Fs < FA, it is determined that the diagnostic section is in a critical stable state. In this state, although the slope meets the theoretical stability conditions, the safety reserve is insufficient, and it is extremely sensitive to additional load increases or environmental condition deteriorations (such as drainage infiltration). Therefore, the control decision generated by the system is "maintain the existing surcharge load". This means that it is not allowed to add new loads on the top of this diagnostic section, and at the same time, the monitoring frequency should be strengthened to closely monitor the change trend of the safety factor. This decision aims to maintain the current situation and avoid pushing the slope into a dangerous state due to blindly increasing the load, which is a risk preventive control strategy.

[0073] c) If F A < F s , it is determined that the diagnostic section is in a safety reserve state, and a control decision allowing the surcharge load to be increased according to the design capacity on the top of this section is generated. If the overall safety factor F A < F sIf the condition is met, the diagnostic section is determined to be in a safe reserve state. This indicates that the slope has sufficient safety margin in its current state and can withstand a certain amount of additional load. The system will generate a control decision to "allow additional load at the top of this section according to the design capacity." Here, "design capacity" refers to the maximum allowable load limit for this area determined by the preliminary engineering design or special assessment. This decision allows for the rational use of the slope's bearing capacity for storing materials, equipment, or carrying out other construction activities, provided that it does not exceed the design capacity, thereby improving the efficiency of construction site utilization while ensuring safety. When executing this decision, the system typically suggests an initial increment lower than the design capacity and initiates a new round of monitoring and calculation cycles after the load is increased to verify that the stability remains under control.

[0074] In practical implementation, the technical solution described in this invention fully considers the conditions and cost control of the engineering site. Its overall implementation will not impose a significant additional economic burden on the existing foundation pit monitoring system and possesses good on-site operability. This is mainly reflected in the following aspects: First, vibration sensors and their associated boreholes are conventional geophysical monitoring equipment with low cost per point. The required array size is determined by the slope dimensions, resulting in limited overall investment. As key benchmarks, existing dewatering wells or geological monitoring wells surrounding the foundation pit can be fully utilized at the engineering site. These wells are typically deep enough and structurally sound, allowing them to be used directly as benchmarks for wave velocity calibration. This avoids the high costs and time required for specialized drilling, effectively integrating existing resources.

[0075] Secondly, the field data collection operation is efficient and flexible, with minimal disruption to normal construction. The electric hammers or spring hammers required for slope vibration are common construction tools and do not require special customization. Although the hammering vibration introduces an additional step, the number of vibration points required on each diagnostic section of the slope is optimized and the total number is controllable. This operation can be flexibly scheduled for nighttime or construction breaks, when background vibration interference is low, and can be completed quickly. Each operation is short and does not affect daytime main construction activities. After a complete monitoring data collection, the obtained wave velocity field and the inverted intensity parameter field have time stability. The data results can support surcharge safety analysis and control decisions for several days or even longer, eliminating the need for daily repeated testing and resulting in high operational efficiency.

[0076] Finally, this solution replaces traditional, costly monitoring and assessment methods with intelligent analysis, offering significant long-term benefits. Traditional regional soil strength verification relies on intensive borehole sampling and laboratory testing, which is costly, time-consuming, and difficult to replicate. This method, through a single-deployment sensor array and periodic vibration, can achieve non-destructive, large-scale rapid inversion of soil strength status and dynamic stability assessment, upgrading traditional discrete, lagging point-based monitoring to continuous, forward-looking area-based intelligent monitoring. This provides project teams with real-time, quantitative, and visualized data for load control, proactively preventing potential landslides caused by improper loading. The comprehensive economic benefits of risk prevention and schedule assurance far outweigh the initial one-time hardware investment, demonstrating excellent technical and economic viability and feasibility for widespread application.

[0077] In another technical solution, the excitation point is set on the concrete surface of the slope protection structure. For each vibration sensor transmitted to the excitation point after each impact, the total energy and the main frequency of the signal are extracted. If, after a certain impact, the total energy of the monitoring signals received by all sensors is lower than the preset energy threshold and the main frequency of the signal is lower than the preset main frequency threshold, it is determined that there is a void in the concrete slope protection layer corresponding to the impact point, and the excitation point is marked as an invalid point in the corresponding void area. The monitoring signal data from the invalid point will be removed in subsequent calculations.

[0078] When the excitation point is set on the concrete surface of the slope protection structure, the vibration signal generated by the impact mainly propagates outward through the contact surface between the concrete and the soil. If the concrete layer is well bonded to the soil behind it, forming a tight contact, the vibration energy can be effectively coupled into the soil and propagated to the sensor in the form of elastic waves within the soil. Conversely, if there is a void behind the concrete layer, an air layer or weak interlayer will be formed, severely hindering the transmission of vibration energy to the soil. In this case, most of the impact energy is confined within the thin concrete layer and propagates in modes such as plate waves, characterized by short propagation distance, rapid attenuation, and changes in frequency composition. Therefore, the signal propagating to the sensor located in the soil will exhibit two observable anomalous features: first, the total signal energy is significantly attenuated due to the obstruction of the energy transmission path; second, the dominant frequency of the signal shifts to lower frequencies due to the change in vibration modes.

[0079] In practice, after each impact vibration, the system analyzes the monitoring signals transmitted to each vibration sensor in real time. Two key indicators are extracted: the total signal energy, which can be quantified by calculating the sum of squares or root mean square value of the signal amplitude within a specific time window; and the dominant frequency, which can be determined by performing a Fast Fourier Transform on the same signal segment to find the frequency corresponding to the highest peak in its amplitude spectrum. The system presets an energy threshold and a dominant frequency threshold. These two thresholds need to be determined during the initial system calibration by conducting multiple experiments in typical areas without gaps and statistically analyzing the signal characteristics.

[0080] The judgment logic is as follows: If, after a certain impact event, the total energy of the signals extracted from all vibration sensors is lower than a preset energy threshold, and the dominant frequency of the signals is lower than a preset dominant frequency threshold, then a strong joint criterion is formed. This indicates that the vibration emitted from the excitation point failed to effectively excite elastic waves propagating in normal dense soil, thus determining that there is a void defect behind the concrete slope protection layer corresponding to the impact point. Once void is determined, the system will immediately mark the excitation point as an "invalid point" of the corresponding void area. In all subsequent calculations, especially in the wave velocity tomography inversion calculation in step S4, all monitoring signal data from these invalid points will be automatically discarded and will not participate in the travel time picking and the construction of the inversion equation set.

[0081] This technical solution ensures the quality of core inversion data. By eliminating invalid and erroneous signals caused by gaps, it avoids data contamination of the tomographic inversion system, thereby significantly improving the reliability and accuracy of the predicted results of representative wave velocities in the soil layer and subsequent soil strength parameters. Furthermore, it enables synchronous and non-destructive diagnosis of the slope protection structure's health status. Using the same excitation and acquisition system, it additionally obtains location information of structural bonding defects, demonstrating the integrated advantage of the intelligent monitoring system's "multi-purpose" capabilities and providing precise technical basis for preventative slope maintenance.

[0082] A method for intelligent monitoring of foundation pit slopes, used in the above-mentioned method, includes: a slope information acquisition module, a data acquisition module, a data processing module, and a safety analysis module; The slope information acquisition module is used to divide the slope into diagnostic sections and store the design information of each diagnostic section, as well as monitor and update the surcharge information at the top of the slope. As the system's engineering data center, the core of the slope information acquisition module is to construct and maintain a digital twin of the slope. This function can be implemented using existing comprehensive geotechnical engineering monitoring platforms or building information modeling platforms. For example, a smart geotechnical engineering monitoring cloud platform like "Smart Measurement Platform" can be used, or BIM design software and its data management environment such as Autodesk Revit and Bentley OpenRoads can be utilized. On these platforms, survey and design documents can be imported to establish a three-dimensional digital model containing information on topography, strata, slope protection structure, and reference slip surface. For dynamically changing surcharge information at the top of the slope, a data interface can be developed for the platform to connect to a weighbridge system deployed on-site, a monitoring video stream with AI recognition capabilities, or receive data reported from handheld terminals of inspection personnel, thereby achieving real-time updates and visualization of surcharge information in the digital model.

[0083] The data acquisition module includes a standard excitation device, a reference well, and a vibration sensor. The standard excitation device is used to strike and generate elastic waves, and the vibration sensor is used to synchronously acquire the vibration signals generated by the strike. The data acquisition module is responsible for the excitation and synchronous acquisition of on-site physical signals, and its automation and standardization can rely on existing hardware control systems. For synchronous acquisition of the vibration sensor network, mature automated data acquisition systems can be used, such as the GeoMonitor system from Solexperts in Switzerland or similar products. These systems can integrate multi-channel sensors through a single cable and achieve remote control, timed or triggered high-precision synchronous data acquisition. For the operation of the standard excitation device (such as an electric hammer) and the calibration of the excitation point coordinates, its dedicated control software and high-precision measurement technology can be combined. For example, a total station can be used to measure the point coordinates, and a mobile measurement control application such as "QimBoX Assistant" can be used to remotely distribute the measurement plan, automatically transmit and input the data, thereby efficiently completing the layout and coordinate archiving of the excitation point array.

[0084] The data processing module is connected to the data acquisition module and is used to process the vibration signals. Its functions include: identifying and removing invalid point signals, calculating the representative wave velocity of each soil sub-unit, and outputting the current equivalent cohesion and current equivalent internal friction angle of each soil sub-unit using the soil parameter model. The preprocessing of the bottom-level vibration signals, travel time extraction, and wave velocity tomography inversion can be performed using professional geophysical data processing software, such as Geosoft's Oasis Montaj platform or similar software, which provide powerful data editing, filtering, visualization, and various inversion algorithm tools, including tomography. The top-level "soil parameter model," which intelligently predicts the equivalent strength parameters of the soil based on wave velocity and other characteristics, can be developed and deployed using mainstream scientific computing and machine learning environments. For example, it can use Python combined with Scikit-learn, TensorFlow, or PyTorch libraries to build and train neural network models, or use MATLAB and its neural network toolbox and optimization toolbox. The trained model can be packaged as an independent computing module and integrated into the system software.

[0085] The safety analysis module is connected to both the slope information acquisition module and the data processing module. It iteratively solves the limit equilibrium equation to calculate the overall safety factor for each diagnostic section, and generates graded safety assessments and load control decisions based on these overall safety factors. The safety analysis module performs the final quantitative stability assessment and decision generation. Its core iterative solution of the limit equilibrium equation can directly call upon or borrow the algorithm kernels of existing mature slope stability analysis software, such as the widely validated algorithms used in the Lizheng Geotechnical series software, GEO-Slope / W software, or 2DSTABFEM developed by Dalian University of Technology. After calculating the overall safety factor, the logic for graded safety assessment and load control decisions based on preset thresholds can be implemented by writing explicit decision rule code or integrating it into a higher-level monitoring and early warning platform with rule engine capabilities, thereby achieving automatic conversion from "data" to "decision" and command issuance.

[0086] An electronic device includes a memory and a processor. The memory stores computer instructions, and the processor executes the computer instructions to perform the above-mentioned intelligent monitoring method for foundation pit slopes. The computer device can be any terminal device, including mobile phones, laptops, desktop computers, tablets, PDAs (Personal Digital Assistants), POS (Point of Sales), vehicle computers, etc.

[0087] Computer instructions are stored in a readable storage medium, such as a computer floppy disk, USB flash drive, portable hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0088] It should be noted that although the steps are described in a specific order above, this does not mean that they must be performed in that order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required functionality is achieved. The number of devices and processing scale described herein are for simplification of the invention; applications, modifications, and variations of this invention will be readily apparent to those skilled in the art.

[0089] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for intelligent monitoring of foundation pit slopes, characterized in that, Includes the following steps: S1. In the stable soil area outside the potential slip zone of the slope, a number of vibration sensors are set up and the three-dimensional coordinates of each vibration sensor are calibrated. A reference well penetrating each main soil layer is set up in the stable soil area. S2. Divide the slope into several continuous diagnostic sections along its length. Within each diagnostic section, divide it into several soil sub-units according to the main soil layers. Using the predetermined reference slip surface as the boundary, divide each soil sub-unit into a stable region located outside the surface and a region to be diagnosed located inside the surface. S3. Use a standard vibration device to strike the reference measuring points corresponding to each major soil layer in the reference well and the vibration points set on the slope of each diagnostic area, and the vibration sensor synchronously collects the vibration signal. S4. Calculate the reference wave velocity of each major soil layer based on the vibration signal obtained from the impact reference point, and calculate the representative wave velocity of each area to be diagnosed through the tomographic inversion algorithm. S5. For each area to be diagnosed, construct an input feature vector containing its representative wave velocity, wave velocity statistical variance, and the encoding of its main soil layer type. Input the input feature vector into a pre-trained soil parameter model. The soil parameter model outputs the equivalent cohesion and equivalent internal friction angle of the area to be diagnosed. S6. For each diagnostic section, based on its reference slip surface, establish a limit equilibrium equation that includes slope geometry parameters, slope top load and slope protection structure design anti-sliding force parameters, and perform iterative calculations to obtain the overall safety factor of the diagnostic section considering the support effect. S7. Based on the overall safety factor of the diagnostic section and the preset load safety threshold, the safety level of the diagnostic section is determined, and a graded load control decision is generated.

2. The intelligent monitoring method for foundation pit slopes as described in claim 1, characterized in that, The calculation method for the representative wave velocity includes the following steps: A1. For each effective propagation path from any excitation point to any vibration sensor, determine its travel length in the stable region and the region to be diagnosed of each soil sub-unit by ray tracing. A2. Establish the theoretical travel time equations for each effective propagation path, and calculate the formulas as follows: in, i, k, u, n These are the excitation point, main soil layer, soil layer sub-unit, and index marker of the vibration sensor. As the reference wave velocity, To represent wave speed, The travel length within the stable region. The length of travel within the area to be diagnosed. U This represents the total number of soil sub-units traversed by the effective propagation path. The theory is outdated; A3. Combine the theoretical travel time equations of all effective propagation paths to form a set of equations with each representative wave velocity as the unknown. Perform iterative inversion to solve the problem with the goal of minimizing the theoretical travel time and the measured travel time. Based on the inversion results, calculate the representative wave velocity of the area to be diagnosed for each soil sub-unit.

3. The intelligent monitoring method for foundation pit slopes as described in claim 1, characterized in that, The method for calculating the overall safety factor is as follows: Based on the reference slip surface of the diagnostic section, the area to be diagnosed is divided into several vertical blocks. For any vertical block, its bottom slip surface length, slip angle, soil self-weight, top load and bottom pore water pressure are obtained. Based on the equivalent cohesion and equivalent internal friction angle of each soil sub-unit, the comprehensive cohesion and comprehensive internal friction angle of the vertical block are calculated by weighting. The design anti-sliding force provided by the slope protection structure of the diagnostic section and its lever arm about the center of the reference slip surface are obtained. The calculation formula of the limit equilibrium equation is as follows: in, j For the index of the vertical bars, m This represents the total number of vertical strips. l j The length of the bottom sliding surface. a j For the sliding tilt angle, W j Due to the weight of the soil, ΔW j For top stacking, u j Pore ​​water pressure, C j To enhance cohesion, φ j To take into account the internal friction angle, R The design anti-sliding force provided for the slope protection structure in this diagnostic section R a The radius of the sliding arc of the reference sliding surface. d To design the lever arm of the anti-slip force with respect to the reference slip surface, F s This represents the overall safety factor for the diagnostic section.

4. The intelligent monitoring method for foundation pit slopes as described in claim 3, characterized in that, The design anti-sliding force and lever arm are determined according to the slope protection structure type as follows: If the slope protection structure is a static retaining structure, the design anti-sliding force is the maximum horizontal resistance; If the slope protection structure is an anchored slope protection structure, the design anti-sliding force is the sum of the horizontal components of the maximum design tension of all anchor rods or anchor cables; If the slope protection structure is a shotcrete layer, the design anti-sliding force is zero. The lever arm is the vertical distance from the line of action of the designed anti-slip force to the center of the corresponding arc of the reference slip surface.

5. The intelligent monitoring method for foundation pit slopes as described in claim 3, characterized in that, Step S7 includes: Set a stacking safety threshold greater than 1.

0. F A Safety classification and load control are determined based on the following methods: like F s If the value is less than 1.0, the diagnostic section is determined to be in an unstable and dangerous state. A control decision is generated to transfer or remove the top load of the diagnostic section. Based on the changed top load, the limit equilibrium equation is recalculated to obtain the verification safety factor. If the verification safety factor increases to 1.0 or above, the control decision is confirmed to be effective. If 1.0 < F s < F A If the diagnostic section is determined to be in a critically stable state, a control decision to maintain the existing load will be generated. like F A < F s The system determines that the diagnostic section is in a safe reserve state and generates a control decision that allows for an increase in load at the top of the section according to the design capacity.

6. The intelligent monitoring method for foundation pit slopes as described in claim 1, characterized in that, The excitation point is set on the concrete surface of the slope protection structure. For each vibration sensor, the total energy and main frequency of the signal transmitted to the excitation point are extracted. If, after a certain impact, the total energy of the monitoring signals received by all sensors is lower than the preset energy threshold and the main frequency is lower than the preset main frequency threshold, it is determined that there is a void in the concrete slope protection layer corresponding to the impact point, and the excitation point is marked as an invalid point in the corresponding void area. The monitoring signal data from the invalid point will be removed in subsequent calculations.

7. The intelligent monitoring method for foundation pit slopes as described in claim 1, characterized in that, The soil parameter model is a BP neural network model optimized by particle swarm optimization algorithm, and adopts a three-layer feedforward network structure, including: an input layer, at least one hidden layer and an output layer; The input layer has multiple input neuron nodes, and the hidden layer has multiple hidden neuron nodes. Each input neuron node is connected to each of the hidden neuron nodes, and each hidden neuron node is connected to the output layer. The neuron nodes in the input layer are used to receive an input feature vector corresponding to a soil sub-unit. The input feature vector includes the representative wave velocity, wave velocity statistical variance, and the encoding of the main soil layer type of the soil sub-unit. The neuron nodes in the output layer are used to output the current equivalent cohesion and the current equivalent internal friction angle of the soil sub-unit.

8. An intelligent monitoring system for foundation pit slopes, used to execute the intelligent monitoring method for foundation pit slopes as described in any one of claims 1 to 7, characterized in that, It includes a slope information acquisition module, a data acquisition module, a data processing module, and a safety analysis module; The slope information acquisition module is used to divide the slope into diagnostic sections and store the design information of each diagnostic section, as well as to monitor and update the surcharge information at the top of the slope. The data acquisition module includes a standard excitation device, a reference well, and a vibration sensor. The standard excitation device is used to strike and generate elastic waves, and the vibration sensor is used to synchronously acquire the vibration signals generated by the strike. The data processing module is connected to the data acquisition module and is used to process the vibration signal; Its functions include: identifying and eliminating invalid point signals, inverting and calculating the representative wave velocity of each soil sub-unit, and outputting the equivalent cohesion and equivalent internal friction angle of each soil sub-unit using the soil parameter model; The safety analysis module is connected to the slope information acquisition module and the data processing module respectively. By iteratively solving the limit equilibrium equation, it calculates the overall safety factor of each diagnostic section and generates graded safety judgment and load control decision based on the overall safety factor.

9. An electronic device, characterized in that, The method includes a memory and a processor, wherein the memory stores computer instructions, and the processor executes the computer instructions to perform the intelligent monitoring method for foundation pit slopes as described in any one of claims 1 to 7.

Images