Slope stability monitoring method and system based on artificial intelligence

Abstract

The invention provides a slope stability monitoring method and system based on artificial intelligence, and relates to the technical field of slope monitoring, and the method comprises the steps: obtaining key position monitoring data; acquiring historical key position monitoring data and historical external environment data; acquiring external environment data; determining a key position stability coefficient according to the historical key position monitoring data, the key position monitoring data and the external environment data; determining the overall stability coefficient of the slope according to the stability coefficient of the key position; processing the key position monitoring data and the external environment data through a trained slope stability prediction model to obtain a predicted key position stability coefficient and a predicted slope overall stability coefficient; and monitoring the slope according to the key position stability coefficient, the slope overall stability coefficient, the predicted key position stability coefficient and the predicted slope overall stability coefficient. According to the invention, the accuracy of slope stability monitoring can be improved.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring technology, and in particular to a slope stability monitoring method and system based on artificial intelligence. Background Technology

[0002] In related technologies, slope stability can be monitored by professionals based on isolated sensor data and empirical thresholds. However, these technologies are difficult to reflect the overall condition of the slope and the evolution of internal damage, and it is also difficult to guarantee the accuracy of the monitoring results.

[0003] The information disclosed in the background section of this application is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0004] This invention provides an artificial intelligence-based slope stability monitoring method and system, which can solve the technical problems that related technologies are unable to reflect the overall state of the slope and the evolution of internal damage, and are unable to guarantee the accuracy of monitoring results.

[0005] According to a first aspect of the present invention, an artificial intelligence-based slope stability monitoring method is provided, comprising: acquiring key location monitoring data at multiple times during a monitoring cycle using a combination of sensors installed at key locations; acquiring historical key location monitoring data and historical external environment data for historical monitoring cycles; acquiring external environment data at multiple times during the monitoring cycle; determining a key location stability coefficient based on the historical key location monitoring data, the key location monitoring data, and the external environment data; determining an overall slope stability coefficient based on the key location stability coefficient; processing the key location monitoring data and the external environment data using a trained slope stability prediction model to obtain a predicted key location stability coefficient and a predicted overall slope stability coefficient; and monitoring the slope based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

[0006] According to the present invention, determining the stability coefficient of a critical location based on the historical critical location monitoring data, the critical location monitoring data, and the external environmental data includes: determining multiple critical location monitoring physical quantities based on the critical location monitoring data, wherein the critical location monitoring physical quantities include: surface horizontal displacement, deep horizontal displacement, pore water pressure, and soil pressure; determining a comprehensive deviation distance and a first standard deviation based on the critical location monitoring physical quantities and the historical critical location monitoring data; determining effective rainfall and vibration velocity data based on the external environmental data; determining historical comprehensive deformation acceleration based on the historical critical location monitoring data; obtaining a slope material damage sensitivity coefficient; and determining the critical location stability coefficient based on the slope material damage sensitivity coefficient, the critical location monitoring physical quantities, the comprehensive deviation distance, the first standard deviation, the effective rainfall, the vibration velocity data, and the historical comprehensive deformation acceleration.

[0007] According to the present invention, determining the comprehensive deviation distance and the first standard deviation based on the key location monitoring physical quantity and the historical key location monitoring data includes: determining the historical key location monitoring physical quantity based on the historical key location monitoring data; determining the physical deviation vector based on the key location monitoring physical quantity and the historical key location monitoring physical quantity; determining the comprehensive deviation distance based on the physical deviation vector; and determining the first standard deviation based on the historical key location monitoring physical quantity.

[0008] According to the present invention, the stability coefficient of a key location is determined based on the slope material damage sensitivity coefficient, the monitored physical quantities at the key locations, the comprehensive deviation distance, the first standard deviation, the effective rainfall, the vibration velocity data, and the historical comprehensive deformation acceleration, including: according to the formula: Determine the key position stability coefficient of the i-th key position at the j-th time of the monitoring period. Where max is the function for finding the maximum value. For mapping functions, and To preset the weights, For the i-th key location, monitor the physical quantity at the k-th key location at the j-th time of the monitoring period. The preset warning threshold for the physical quantity monitored at the k-th key location. Let i be the effective rainfall at the i-th key location at the j-th time point of the monitoring period. To preset the effective rainfall threshold, Let i be the vibration velocity data of the i-th key location at the j-th time of the monitoring cycle. To preset the vibration acceleration data threshold, Let i be the historical comprehensive deformation acceleration of the i-th key location at the j-th moment of the e-th historical monitoring period. Let be the slope material damage sensitivity coefficient at the i-th critical location. The duration between adjacent moments in the historical monitoring cycle. Let be the comprehensive deviation distance of the i-th key location at the j-th time point in the monitoring period. Let K be the first standard deviation, K be the number of physical quantities monitored at key locations, k≤K, E be the number of historical monitoring cycles, e≤E, m be the number of moments in the monitoring cycle, j≤m, and k, K, e, E, j, and m are all positive integers.

[0009] According to the present invention, determining the overall stability coefficient of a slope based on the critical location stability coefficient includes: determining a minimum critical location stability coefficient based on the critical location stability coefficients of multiple critical locations; determining dangerous locations of the slope based on the minimum critical location stability coefficient; determining the danger distance between each critical location and the dangerous location of the slope; and determining the overall stability coefficient of the slope based on the minimum critical location stability coefficient, the critical location stability coefficient, and the danger distance.

[0010] According to the present invention, determining the overall slope stability coefficient based on the minimum critical location stability coefficient, the critical location stability coefficient, and the danger distance includes: according to the formula: Determine the overall slope stability coefficient at time j of the monitoring period. ,in, Let be the minimum critical position stability coefficient at time j of the monitoring period. Let i be the key position stability coefficient at the j-th time of the monitoring period. Let be the critical distance between the i-th critical location and the dangerous location on the slope. This is the preset distance.

[0011] According to the present invention, the training steps of the slope stability prediction model include: determining the historical key location stability coefficient and the historical overall slope stability coefficient at multiple historical prediction times based on historical key location monitoring data and historical external environment data at multiple historical prediction times; processing the historical key location monitoring data and the historical external environment data through the slope stability prediction model to determine the historical sample predicted key location stability coefficient and the historical sample predicted overall slope stability coefficient at multiple historical prediction times; determining the training loss function of the slope stability prediction model based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient; and training the slope stability prediction model according to the training loss function to obtain the trained slope stability prediction model.

[0012] According to the present invention, the training loss function of the slope stability prediction model is determined based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient, including: according to the formula: Determine the training loss function for the slope stability prediction model. Where max is the function for finding the maximum value. These are preset parameters. Let be the historical key position stability coefficient of the i-th key position at the r-th historical prediction time. For the i-th critical location, predict the critical location stability coefficient for historical samples at the r-th historical prediction time. Let be the historical overall slope stability coefficient at the r-th historical prediction time. For the historical sample at the r-th historical prediction time, predict the overall slope stability coefficient. For the i-th key position, predict the key position stability coefficient for the historical sample at the (r+1)-th historical prediction time. R is the number of historical prediction times, r≤R, n is the number of key positions, i≤n, and r, R, i and n are all positive integers.

[0013] According to a second aspect of the present invention, an artificial intelligence-based slope stability monitoring system is provided, comprising: a monitoring data module for acquiring key location monitoring data at multiple times during a monitoring period using a combination of sensors located at key locations; a historical data module for acquiring historical key location monitoring data and historical external environment data for historical monitoring periods; an environmental data module for acquiring external environment data at multiple times during a monitoring period; a location evaluation module for determining a key location stability coefficient based on the historical key location monitoring data, the key location monitoring data, and the external environment data; an overall evaluation module for determining an overall slope stability coefficient based on the key location stability coefficient; a model prediction module for processing the key location monitoring data and the external environment data using a trained slope stability prediction model to obtain a predicted key location stability coefficient and a predicted overall slope stability coefficient; and a slope monitoring module for monitoring the slope based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

[0014] Technical Effects: According to this invention, monitoring data and external environmental data at key locations can be accurately collected. Based on historical monitoring data of key locations, the stability of each key location is assessed, and a key location stability coefficient is determined. Furthermore, based on the key location stability coefficient, the overall stability of the slope is assessed, and the overall slope stability coefficient is determined. Simultaneously, the slope stability prediction model can be used to predict the key location stability coefficient and the overall slope stability coefficient, and slope stability is monitored based on the prediction results, the overall slope stability coefficient, and the key location stability coefficient. When determining the key location stability coefficient, it can be determined based on the slope material damage sensitivity coefficient, key location monitoring physical quantities, comprehensive deviation distance, first standard deviation, effective rainfall, vibration velocity data, and historical comprehensive deformation acceleration. During the calculation process, the stability of key locations can be assessed based on external environmental and internal factors, and the damage accumulation status and abnormal conditions of the key location monitoring physical quantity system are fully analyzed. Furthermore, based on the internal and external key location stability, damage accumulation status, and abnormal conditions of the key location monitoring physical quantity system, the key location stability coefficient is determined, improving the comprehensiveness and accuracy of the key location stability coefficient. When determining the overall slope stability coefficient, it can be based on the minimum critical location stability coefficient, the critical location stability coefficient, and the danger distance. During calculation, the isolation status of dangerous locations and the spatial propagation characteristics of risk can be considered to improve the accuracy of the overall slope stability coefficient. When determining the training loss function for the slope stability prediction model, it can be based on historical critical location stability coefficients, historical overall slope stability coefficients, historical sample predicted critical location stability coefficients, and historical sample predicted overall slope stability coefficients. During calculation, the loss function can be designed based on the comprehensive prediction error loss value and the monotonic physical constraint loss value. This allows the slope stability prediction model to improve its prediction of both critical locations and overall slope stability during training, and ensures that the predictions conform to a fundamental directional law of slope stability evolution, thus more effectively improving the performance of the slope stability prediction model.

[0015] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Other features and aspects of the invention will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0017] Figure 1 An exemplary flowchart of an artificial intelligence-based slope stability monitoring method according to an embodiment of the present invention is shown.

[0018] Figure 2 An exemplary schematic diagram illustrating the determination of the stability coefficient of a critical location according to an embodiment of the present invention is shown;

[0019] Figure 3 An exemplary schematic diagram illustrating the determination of the overall slope stability coefficient according to an embodiment of the present invention is shown;

[0020] Figure 4 A block diagram of an artificial intelligence-based slope stability monitoring system according to an embodiment of the present invention is shown as an example. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0023] Figure 1An exemplary flowchart of an artificial intelligence-based slope stability monitoring method according to an embodiment of the present invention is shown. The method includes: Step S1, acquiring key location monitoring data at multiple times during a monitoring cycle using a combination of sensors set at key locations; Step S2, acquiring historical key location monitoring data and historical external environment data for historical monitoring cycles; Step S3, acquiring external environment data at multiple times during a monitoring cycle; Step S4, determining a key location stability coefficient based on the historical key location monitoring data, the key location monitoring data, and the external environment data; Step S5, determining an overall slope stability coefficient based on the key location stability coefficient; Step S6, processing the key location monitoring data and the external environment data using a trained slope stability prediction model to obtain a predicted key location stability coefficient and a predicted overall slope stability coefficient; Step S7, monitoring the slope based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

[0024] According to the artificial intelligence-based slope stability monitoring method of the present invention, it can accurately collect monitoring data and external environmental data at key locations, and evaluate the stability of each key location based on historical key location monitoring data, key location monitoring data and external environmental data, and determine the key location stability coefficient. Furthermore, based on the key location stability coefficient, it evaluates the overall stability of the slope and determines the overall slope stability coefficient. At the same time, it can predict the key location stability coefficient and the overall slope stability coefficient through a slope stability prediction model, and monitor the slope stability based on the prediction results, the overall slope stability coefficient and the key location stability coefficient.

[0025] According to one embodiment of the present invention, in step S1, at multiple moments during the monitoring cycle, key location monitoring data is acquired by a combination of sensors set at key locations.

[0026] For example, by using a sensor array (which includes a total station, inclinometer, piezometer, and earth pressure cell) set at key locations (such as the potential area of ​​tensile cracks at the rear edge, the load-bearing area, the key point of the support structure, the geological anomaly zone, and the slope toe uplift zone), monitoring data (such as surface horizontal displacement, deep horizontal displacement, pore water pressure, and earth pressure) can be obtained.

[0027] According to one embodiment of the present invention, in step S2, historical key location monitoring data and historical external environment data of historical monitoring cycles are acquired.

[0028] For example, historical key location monitoring data and historical external environmental data (such as historical rainfall and historical vibration velocity data) can be obtained through historical databases for historical monitoring cycles.

[0029] According to one embodiment of the present invention, in step S3, external environmental data is acquired at multiple moments during the monitoring cycle.

[0030] For example, external environmental data (such as rainfall and vibration velocity data) can be acquired by setting up rainfall and vibration sensors at various key locations.

[0031] According to an embodiment of the present invention, in step S4, a critical location stability coefficient is determined based on the historical critical location monitoring data, the critical location monitoring data, and the external environment data.

[0032] Figure 2 An exemplary schematic diagram illustrating the determination of the stability coefficient at critical locations according to an embodiment of the present invention is shown.

[0033] According to an embodiment of the present invention, step S4 includes: step S41, determining multiple key location monitoring physical quantities based on the key location monitoring data, wherein the key location monitoring physical quantities include: surface horizontal displacement, deep horizontal displacement, pore water pressure, and soil pressure; step S42, determining a comprehensive deviation distance and a first standard deviation based on the key location monitoring physical quantities and the historical key location monitoring data; step S43, determining effective rainfall and vibration velocity data based on the external environmental data; step S44, determining historical comprehensive deformation acceleration based on the historical key location monitoring data; step S45, obtaining a slope material damage sensitivity coefficient; step S46, determining a key location stability coefficient based on the slope material damage sensitivity coefficient, the key location monitoring physical quantities, the comprehensive deviation distance, the first standard deviation, the effective rainfall, the vibration velocity data, and the historical comprehensive deformation acceleration.

[0034] For example, surface horizontal displacement is obtained using a total station, deep horizontal displacement is obtained using a tiltmeter, pore water pressure is obtained using a piezometer, and earth pressure is obtained using an earth pressure cell; based on the physical quantities monitored at key locations and historical monitoring data at key locations, the comprehensive deviation distance and first standard deviation are determined; based on external environmental data, effective rainfall and vibration velocity data are determined. For example, if the current monitoring period is on October 7th, the rainfall on October 6th was 10 mm, the rainfall on October 5th was 20 mm, and the rainfall on October 4th was 30 mm, then... Regarding the lag effect of rainfall on soil, the effective rainfall at the current monitoring time is determined by summing the current rainfall, 0.8 times the rainfall on October 6th, 0.6 times the rainfall on October 5th, and 0.4 times the rainfall on October 4th. Adjacent monitoring times are separated by ten minutes. The vibration velocity data at the second moment of the monitoring period is the maximum vibration velocity value between the first and second moments of the monitoring period. Historical comprehensive deformation acceleration is determined based on historical monitoring data from key locations. Based on historical monitoring data at key locations, historical surface horizontal displacement and historical deep horizontal displacement were determined. Based on these historical surface and deep horizontal displacements, historical surface and deep horizontal displacement accelerations were determined. The acceleration of the historical surface horizontal displacement acceleration in the same direction as the historical deep horizontal displacement acceleration was determined (since the deep horizontal displacement acceleration directly reflects the shear failure state of the slip zone, its direction is used as the reference direction). The historical comprehensive deformation acceleration was determined by summing the accelerations of 0.7 times and 0.3 times the historical surface horizontal displacement acceleration in the same direction as the historical deep horizontal displacement acceleration. Through laboratory calibration, the material damage sensitivity coefficients of similar materials at each key location of the slope were obtained, i.e., the slope material damage sensitivity coefficients. Based on the slope material damage sensitivity coefficients, key location monitoring physical quantities, comprehensive deviation distance, first standard deviation, effective rainfall, vibration velocity data, and historical comprehensive deformation acceleration, the stability status of each location of the slope was assessed, and the stability coefficients of the key locations were determined.

[0035] According to an embodiment of the present invention, step S42 includes: step S421, determining the historical key location monitoring physical quantity based on the historical key location monitoring data; step S422, determining the physical deviation vector based on the key location monitoring physical quantity and the historical key location monitoring physical quantity; step S423, determining the comprehensive deviation distance based on the physical deviation vector; and step S424, determining the first standard deviation based on the historical key location monitoring physical quantity.

[0036] For example, based on historical key location monitoring data, determine the historical key location monitoring physical quantities, such as historical surface horizontal displacement, historical deep horizontal displacement, historical pore water pressure, and historical earth pressure; based on the key location monitoring physical quantities and historical key location monitoring physical quantities, determine the physical deviation vector, such as by averaging the long-term historical key location monitoring physical quantities from multiple historical monitoring periods that are recognized as stable and without abnormal deformation (e.g., long-term monitoring period before construction, or continuous data from a certain dry season) to obtain the benchmark key location monitoring physical quantity; based on the deviation values ​​of multiple key location monitoring physical quantities and the corresponding key location monitoring physical quantities, determine the physical deviation vector, such as [surface horizontal displacement - benchmark surface horizontal displacement, deep horizontal displacement - benchmark deep horizontal displacement]; Based on the physical deviation vector, determine the comprehensive deviation distance, such as by calculating the Euclidean length of the physical deviation vector, i.e., the comprehensive deviation distance; based on the historical key location monitoring physical quantities, determine the first standard deviation, such as by calculating the individual standard deviation of a single key location monitoring physical quantity in multiple historical monitoring periods that are recognized as stable and without abnormal deformation. The individual standard deviation represents the amplitude of individual fluctuations of the sensor corresponding to the key location monitoring physical quantity during the stable period due to its own normal noise and fluctuations. The first standard deviation is determined by the sum of the squares of the individual standard deviations of all key location monitoring physical quantities to the power of half. The first standard deviation represents the total system fluctuation radius of the entire sensor system during the stable period due to the combined effect of all normal environmental noise and measurement errors.

[0037] According to an embodiment of the present invention, step S46 includes: determining the key position stability coefficient of the i-th key position at the j-th moment of the monitoring period according to formula (1). ,

[0038] (1)

[0039] Where max is the function for finding the maximum value. For mapping functions, and To preset the weights, For the i-th key location, monitor the physical quantity at the k-th key location at the j-th time of the monitoring period. The preset warning threshold for the physical quantity monitored at the k-th key location. Let i be the effective rainfall at the i-th key location at the j-th time point of the monitoring period. To preset the effective rainfall threshold, Let i be the vibration velocity data of the i-th key location at the j-th time of the monitoring cycle. To preset the vibration acceleration data threshold, Let i be the historical comprehensive deformation acceleration of the i-th key location at the j-th moment of the e-th historical monitoring period. Let be the slope material damage sensitivity coefficient at the i-th critical location. The duration between adjacent moments in the historical monitoring cycle. Let be the comprehensive deviation distance of the i-th key location at the j-th time point in the monitoring period. Let K be the first standard deviation, K be the number of physical quantities monitored at key locations, k≤K, E be the number of historical monitoring cycles, e≤E, m be the number of moments in the monitoring cycle, j≤m, and k, K, e, E, j, and m are all positive integers.

[0040] According to one embodiment of the present invention, This is the ratio of the monitored physical quantity at the k-th key location at the j-th time of the monitoring period for the i-th key location to the preset warning threshold for the monitored physical quantity at the k-th key location. The larger this ratio, the larger the monitored physical quantity at the corresponding key location, and the more dangerous the state reflected by the monitored physical quantity. For example, a larger horizontal displacement of the ground surface indicates that obvious and irreversible deformation has begun to occur on the slope surface, which is the most intuitive visual precursor to a landslide. A larger horizontal displacement of deeper layers indicates an abnormal increase in water pressure inside the slope, which will reduce the shear strength of the soil and may lead to a landslide. It can be determined based on the warning values ​​specified in industry standards. This indicates that when the monitored physical quantity at a critical location exceeds the corresponding warning threshold, the degree of danger will increase non-linearly (due to the destructive characteristics of soil and rock materials, i.e., when approaching the limit state, a small increase in load or deformation may lead to a sharp loss of bearing capacity, i.e., non-linear failure). This indicates the degree of danger reflected by the physical quantity monitored at the k-th critical location at the j-th moment of the monitoring period at the i-th critical location. This indicates the level of security reflected in the physical quantity monitored at the k-th critical location at the j-th moment of the monitoring period for the i-th critical location. This means that the level of security is smoothly mapped to the range of 0 to 1. The closer to 1, the safer it is. This represents the summation and average of the physical quantities monitored at key locations, indicating the level of security reflected by all the physical quantities monitored at key locations, i.e., the internal physical state.

[0041] According to one embodiment of the present invention, This is the ratio of the effective rainfall at the i-th key location at the j-th time of the monitoring period to the preset effective rainfall threshold. The larger this ratio, the greater the effective rainfall. It can be determined based on historical experience. This represents the influence coefficient of effective rainfall in an exponentially decaying form on the i-th critical location, indicating the "saturation effect" or "threshold effect" in engineering. For example, when rainfall exceeds the soil's infiltration capacity, the excess water forms runoff, and its destructive force on slope stability does not increase indefinitely with rainfall. The coefficient representing the influence of exponentially decaying vibration velocity data on the i-th critical location indicates that when the vibration velocity data is lower than the preset vibration acceleration data threshold (the vibration intensity threshold required to cause irreversible deterioration of the soil structure (e.g., liquefaction, structural resonance, or lock-in segment rupture), the soil at the critical location is in an elastic or slightly plastic state, and most of the deformation can be recovered after the vibration stops, with negligible damage. Once the preset vibration acceleration data threshold is exceeded, the soil enters the plastic flow or failure stage, and damage begins to accumulate dramatically. This represents the stability coefficient at key locations determined based on effective rainfall and vibration velocity data, i.e., the state of external influences. The larger it is, the smaller the external influence.

[0042] According to one embodiment of the present invention, Stability at critical locations determined based on internal physical state and external influences. and It can be determined through expert experience methods (such as the Analytic Hierarchy Process, AHP).

[0043] According to one embodiment of the present invention, , This indicates that only historical aggregate deformation accelerations greater than 0 are considered. It means that only accelerated deformation (e.g., damage occurring) will be accumulated; decelerated or uniform deformation (e.g., damage stagnation) will not generate new damage. This represents the velocity increment of the i-th key location from time j-1 to time j in the e-th historical monitoring period. This represents the summation of the positive velocity increments over all moments in all historical monitoring periods, and represents the total deformation driving energy accumulated from all accelerated deformation events. Let represent the product of the total deformation driving energy accumulated from all accelerated deformation events and the slope material damage sensitivity coefficient at the i-th critical location, and let represent the effective cumulative damage caused by the external deformation history load at the i-th critical location. The product is dimensionless. Let represent the residual integrity coefficient at the i-th critical location. When the effective damage accumulation is 0, the residual integrity coefficient is 1. Since the accumulation of damage is nonlinear, the residual integrity coefficient decreases as the effective damage accumulation increases, and the decrease is not uniform.

[0044] According to one embodiment of the present invention, Let be the ratio of the comprehensive deviation distance of the i-th key location at the j-th time of the monitoring period to the first standard deviation. When the value equals 0, it indicates that there is no deviation between the monitored physical quantities at all key locations and the reference quantity, and the status is normal. When the value equals 1, it indicates that the deviation of all key location monitored physical quantities is within the historical normal fluctuation range. This occurs when the situation is at the borderline between normal and abnormal conditions. When it is large, for example, A value greater than 2 indicates that the combined deviation of all key location monitored physical quantities in the current monitoring cycle has far exceeded the range of historical normal fluctuations. Even if each key location monitored physical quantity individually may not exceed its single-point alarm threshold, their combined state is extremely abnormal. This is often an early and sensitive sign of systemic instability (e.g., the slip surface begins to connect). When equal to 0, A value of 1 indicates that when all key location monitoring physical quantities are in normal condition, the key location stability coefficient will not be reduced. When equal to 1, Approximately 0.135 indicates that when all monitored physical quantities at key locations are in a critical state, the stability coefficient of the key locations decreases sharply. The above processing can respond drastically to abnormal states that have reached the critical point, thereby raising the warning level to a level that cannot be ignored before the risk actually erupts. When it is large, A value close to 0 indicates that the stability coefficient at the critical location is also close to 0, meaning that the critical location is highly unstable.

[0045] According to one embodiment of the present invention, This indicates that the stability coefficient of a critical location is determined based on the stability of internal and external critical locations, the accumulation of damage, and the abnormal status of the physical quantity system monitored at the critical locations.

[0046] In this way, the stability coefficient of a critical location can be determined based on the slope material damage sensitivity coefficient, key location monitoring physical quantities, comprehensive deviation distance, first standard deviation, effective rainfall, vibration velocity data, and historical comprehensive deformation acceleration. During the calculation process, the stability of the critical location can be assessed based on the external environment and internal factors, and the damage accumulation status and abnormal status of the key location monitoring physical quantity system can be fully analyzed. Furthermore, based on the internal and external key location stability, damage accumulation status, and abnormal status of the key location monitoring physical quantity system, the key location stability coefficient is determined, thus improving the comprehensiveness and accuracy of the key location stability coefficient.

[0047] According to one embodiment of the present invention, in step S5, the overall stability coefficient of the slope is determined based on the stability coefficient of the key location.

[0048] Figure 3 A schematic diagram illustrating the determination of the overall stability coefficient of a slope according to an embodiment of the present invention is shown.

[0049] According to an embodiment of the present invention, step S5 includes: step S51, determining a minimum critical position stability coefficient based on critical position stability coefficients of multiple critical positions; step S53, determining dangerous locations of the slope based on the minimum critical position stability coefficient; step S54, determining dangerous distances between each critical position and dangerous locations of the slope; and step S555, determining an overall slope stability coefficient based on the minimum critical position stability coefficient, the critical position stability coefficients, and the dangerous distances.

[0050] For example, the minimum critical location stability coefficient is determined based on the minimum critical location stability coefficient of multiple critical locations; the dangerous location of the slope is determined based on the location corresponding to the minimum critical location stability coefficient; the straight-line distance between each critical location and the dangerous location of the slope is determined based on the coordinates of each critical location and the dangerous location of the slope, i.e., the danger distance; the overall stability of the slope is assessed based on the minimum critical location stability coefficient, the critical location stability coefficient and the danger distance, and the overall stability coefficient of the slope is determined.

[0051] According to an embodiment of the present invention, step S54 includes: determining the overall slope stability coefficient at the j-th moment of the monitoring period according to formula (2). ,

[0052] (2)

[0053] in, Let be the minimum critical position stability coefficient at time j of the monitoring period. Let i be the key position stability coefficient at the j-th time of the monitoring period. Let be the critical distance between the i-th critical location and the dangerous location on the slope. This is the preset distance.

[0054] According to one embodiment of the present invention, if one location on a slope is highly unstable, even if other critical locations are very stable, the entire slope may still be in a dangerous state. Therefore, the overall safety of the slope is controlled by its most dangerous area, and the minimum critical location stability coefficient at the j-th moment of the monitoring period is used as the criterion. As an evaluation benchmark.

[0055] According to one embodiment of the present invention, This represents the unstable state of the i-th critical location at the j-th moment of the monitoring period. The preset distance represents the radius of risk propagation, determined by the slope structure (e.g., potential slip surface depth, rock mass integrity). This indicates that the propagation of risk in space decreases with distance; points closer to the dangerous location on the slope contribute more to the propagation of risk. This represents the effective risk contribution value of the i-th key location to the overall risk of the slope at the j-th time point in the monitoring period. This represents the average effective risk contribution value of all critical locations except for the dangerous slope location. This is true when the dangerous slope location is an isolated point, i.e., when all critical locations other than the dangerous slope location are stable. Approximately equal to 1, Approximately equal to This represents the overall stability coefficient of the slope when the dangerous location is an isolated point. Slope dangerous location It is determined that the risk is localized when there are many unstable critical locations around a dangerous location on a slope. A significant increase indicates that the scale and likelihood of slope failure are amplified exponentially by spatial correlation effects.

[0056] In this way, the overall slope stability coefficient can be determined based on the minimum critical location stability coefficient, the critical location stability coefficient, and the danger distance. During the calculation process, the overall slope stability coefficient can be determined based on the isolation status of the dangerous locations on the slope and the spatial propagation characteristics of the risk, thus improving the accuracy of the overall slope stability coefficient.

[0057] According to an embodiment of the present invention, in step S6, the monitoring data of the key locations and the external environment data are processed by the trained slope stability prediction model to obtain the predicted key location stability coefficient and the predicted overall slope stability coefficient.

[0058] For example, the slope stability prediction model is a type of neural network model, which includes: a data preprocessing and input layer, a feature extraction layer, a feature fusion layer, and a decision and output layer. The slope stability prediction model is trained with historical data so that it can output the predicted stability coefficients of key locations and the predicted overall slope stability coefficients. The predicted stability coefficients of key locations and the predicted overall slope stability coefficients are data from three days after the current monitoring period.

[0059] According to an embodiment of the present invention, the training steps of the slope stability prediction model include: determining the historical key location stability coefficient and the historical overall slope stability coefficient at multiple historical prediction times based on historical key location monitoring data and historical external environment data at multiple historical prediction times; processing the historical key location monitoring data and the historical external environment data through the slope stability prediction model to determine the historical sample predicted key location stability coefficient and the historical sample predicted overall slope stability coefficient at multiple historical prediction times; determining the training loss function of the slope stability prediction model based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient; and training the slope stability prediction model according to the training loss function to obtain the trained slope stability prediction model.

[0060] For example, if the historical prediction time is set to 8:00 AM on October 8th, then based on the historical key location monitoring data and historical external environment data at multiple historical prediction times, the historical key location stability coefficient and the historical overall slope stability coefficient at multiple historical prediction times are determined. The determination methods for the historical key location stability coefficient and the historical overall slope stability coefficient are similar to those for the key location stability coefficient and the overall slope stability coefficient, and will not be repeated here. The historical key location monitoring data and historical external environment data are processed through a slope stability prediction model to determine the historical sample predicted key location stability coefficient and the historical sample predicted overall slope stability coefficient. For example, when the historical prediction time is 8:00 AM on October 8th... At 8:00 AM, the slope stability prediction model is used to process historical key location monitoring data and historical external environment data over a 10.5-day historical monitoring period to obtain historical sample predicted key location stability coefficients and historical sample predicted overall slope stability coefficients at historical prediction times. Based on these coefficients, the training loss function of the slope stability prediction model is determined. The model is then trained using this training loss function to improve its prediction accuracy, resulting in a trained slope stability prediction model.

[0061] According to one embodiment of the present invention, the training loss function of the slope stability prediction model is determined based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient, including: determining the training loss function of the slope stability prediction model according to formula (3). ,

[0062] (3)

[0063] Where max is the function for finding the maximum value. These are preset parameters. Let be the historical key position stability coefficient of the i-th key position at the r-th historical prediction time. For the i-th critical location, predict the critical location stability coefficient for historical samples at the r-th historical prediction time. Let be the historical overall slope stability coefficient at the r-th historical prediction time. For the historical sample at the r-th historical prediction time, predict the overall slope stability coefficient. For the i-th key position, predict the key position stability coefficient for the historical sample at the (r+1)-th historical prediction time. R is the number of historical prediction times, r≤R, n is the number of key positions, i≤n, and r, R, i and n are all positive integers.

[0064] According to one embodiment of the present invention, Let be the square of the difference between the historical key position stability coefficient of the i-th key position at the r-th historical prediction time and the historical sample predicted key position stability coefficient. Let be the square of the difference between the historical overall slope stability coefficient at the r-th historical prediction time and the historical sample predicted overall slope stability coefficient. This represents the overall prediction error loss value. Using this as a loss value during training can reduce the differences between historical key location stability coefficients and historical sample predicted key location stability coefficients, as well as the differences between historical overall slope stability coefficients and historical sample predicted overall slope stability coefficients. This can drive the slope stability prediction model to simultaneously improve its predictions of both key locations and the overall slope stability. The model attempts to reduce... During this loss process, the internal network is forced to learn how to extract and integrate global features from the input data.

[0065] According to one embodiment of the present invention, The difference between the stability coefficients of the i-th key position predicted by historical samples at the (r+1)-th and r-th historical prediction times is given. The loss value is the difference between the stability coefficients of the predicted key positions of the historical samples at the (r+1)th and rth historical prediction times, where all values ​​are positive. This represents the monotonic physical constraint loss value. Training based on this value allows the slope stability prediction model to learn that, without considering major engineering interventions, the slope's stability state will not naturally reverse and recover. The overall slope stability coefficient should exhibit a non-increasing monotonic trend, ensuring that the model's predictions conform to a fundamental directional law of slope stability evolution. This is a preset parameter and can be initially set to 0.

[0066] According to one embodiment of the present invention, the training loss function of the slope stability prediction model can be determined by summing the above two items.

[0067] In this way, the training loss function of the slope stability prediction model can be determined based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient. During the calculation process, the loss function can be designed based on the comprehensive prediction error loss value and the monotonic physical constraint loss value, so that the slope stability prediction model can improve the prediction of the stability of key locations and the overall slope during the training process, and make the prediction of the slope stability prediction model conform to a fundamental directional law of slope stability evolution, thereby improving the performance of the slope stability prediction model in a more targeted manner.

[0068] According to an embodiment of the present invention, in step S7, the slope is monitored based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

[0069] For example, a BIM model of a slope can be established based on the stability coefficient of key locations, the overall stability coefficient of the slope, the predicted stability coefficient of key locations, and the predicted overall stability coefficient of the slope. A mapping engine between real-time data and BIM model parameters can be established to drive the stability monitoring of the slope through the BIM model. When the stability coefficient of key locations and the predicted stability coefficient of key locations are less than 0.8, the BIM model issues an early warning message. When the overall stability coefficient of the slope and the predicted overall stability coefficient of the slope are less than 0.9, the BIM model issues an early warning message.

[0070] According to the artificial intelligence-based slope stability monitoring method of the present invention, it can accurately collect monitoring data and external environmental data at key locations, and evaluate the stability of each key location based on historical key location monitoring data, key location monitoring data and external environmental data, and determine the key location stability coefficient. Furthermore, based on the key location stability coefficient, it evaluates the overall stability of the slope and determines the overall slope stability coefficient. At the same time, it can predict the key location stability coefficient and the overall slope stability coefficient through a slope stability prediction model, and monitor the slope stability based on the prediction results, the overall slope stability coefficient and the key location stability coefficient. When determining the stability coefficient of critical locations, the coefficient can be determined based on the slope material damage sensitivity coefficient, monitored physical quantities at critical locations, comprehensive deviation distance, first standard deviation, effective rainfall, vibration velocity data, and historical comprehensive deformation acceleration. During the calculation process, the stability of critical locations can be assessed based on external environmental and internal factors, and the damage accumulation and anomalies in the monitored physical quantity system at critical locations can be fully analyzed. Furthermore, the stability coefficient of critical locations is determined based on the internal and external stability of critical locations, the damage accumulation, and the anomalies in the monitored physical quantity system, thus improving the comprehensiveness and accuracy of the stability coefficient. When determining the overall slope stability coefficient, the coefficient can be determined based on the minimum critical location stability coefficient, the critical location stability coefficient, and the danger distance. During the calculation process, the isolation status of dangerous locations on the slope and the spatial propagation characteristics of risk can be considered to determine the overall slope stability coefficient, thus improving its accuracy. When determining the training loss function of the slope stability prediction model, it can be based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient. During the calculation process, the loss function can be designed based on the comprehensive prediction error loss value and the monotonic physical constraint loss value, so that the slope stability prediction model can improve the prediction of the stability of key locations and the overall slope during the training process, and make the prediction of the slope stability prediction model conform to a fundamental directional law of slope stability evolution, thereby more effectively improving the performance of the slope stability prediction model.

[0071] Figure 4An exemplary block diagram of an artificial intelligence-based slope stability monitoring system according to an embodiment of the present invention is shown. The system includes: a monitoring data module for acquiring key location monitoring data at multiple times during a monitoring cycle using a combination of sensors located at key locations; a historical data module for acquiring historical key location monitoring data and historical external environment data for historical monitoring cycles; an environmental data module for acquiring external environment data at multiple times during a monitoring cycle; a location evaluation module for determining a key location stability coefficient based on the historical key location monitoring data, the key location monitoring data, and the external environment data; an overall evaluation module for determining an overall slope stability coefficient based on the key location stability coefficient; a model prediction module for processing the key location monitoring data and the external environment data using a trained slope stability prediction model to obtain a predicted key location stability coefficient and a predicted overall slope stability coefficient; and a slope monitoring module for monitoring the slope based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

[0072] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0073] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and any variations or modifications may be made to the implementation of the present invention without departing from the stated principles.

Claims

1. A slope stability monitoring method based on artificial intelligence, characterized in that, include: At multiple points during the monitoring cycle, key location monitoring data is acquired using a combination of sensors positioned at key locations; historical key location monitoring data and historical external environment data from previous monitoring cycles are acquired; external environment data is acquired at multiple points during the monitoring cycle; a key location stability coefficient is determined based on the historical key location monitoring data, the key location monitoring data, and the external environment data; an overall slope stability coefficient is determined based on the key location stability coefficient; the key location monitoring data and the external environment data are processed using a trained slope stability prediction model to obtain a predicted key location stability coefficient and a predicted overall slope stability coefficient; the slope is monitored based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

2. The slope stability monitoring method based on artificial intelligence according to claim 1, characterized in that, Based on the historical key location monitoring data, the key location monitoring data, and the external environmental data, the key location stability coefficient is determined, including: determining multiple key location monitoring physical quantities based on the key location monitoring data, wherein the key location monitoring physical quantities include: surface horizontal displacement, deep horizontal displacement, pore water pressure, and soil pressure; determining the comprehensive deviation distance and the first standard deviation based on the key location monitoring physical quantities and the historical key location monitoring data; determining the effective rainfall and vibration velocity data based on the external environmental data; determining the historical comprehensive deformation acceleration based on the historical key location monitoring data; obtaining the slope material damage sensitivity coefficient; and determining the key location stability coefficient based on the slope material damage sensitivity coefficient, the key location monitoring physical quantities, the comprehensive deviation distance, the first standard deviation, the effective rainfall, the vibration velocity data, and the historical comprehensive deformation acceleration.

3. The slope stability monitoring method based on artificial intelligence according to claim 2, characterized in that, Determining the comprehensive deviation distance and the first standard deviation based on the key location monitoring physical quantities and the historical key location monitoring data includes: determining the historical key location monitoring physical quantities based on the historical key location monitoring data; determining the physical deviation vector based on the key location monitoring physical quantities and the historical key location monitoring physical quantities; determining the comprehensive deviation distance based on the physical deviation vector; and determining the first standard deviation based on the historical key location monitoring physical quantities.

4. The slope stability monitoring method based on artificial intelligence according to claim 2, characterized in that, Based on the slope material damage sensitivity coefficient, the monitored physical quantities at key locations, the comprehensive deviation distance, the first standard deviation, the effective rainfall, the vibration velocity data, and the historical comprehensive deformation acceleration, the stability coefficient at key locations is determined, including: according to the formula: Determine the key position stability coefficient of the i-th key position at the j-th time of the monitoring period. Where max is the function for finding the maximum value. For mapping functions, and To preset the weights, For the i-th key location, monitor the physical quantity at the k-th key location at the j-th time of the monitoring period. The preset warning threshold for the physical quantity monitored at the k-th key location. Let i be the effective rainfall at the i-th key location at the j-th time point of the monitoring period. To preset the effective rainfall threshold, Let i be the vibration velocity data of the i-th key location at the j-th time of the monitoring cycle. To preset the vibration acceleration data threshold, Let i be the historical comprehensive deformation acceleration of the i-th key location at the j-th moment of the e-th historical monitoring period. Let be the slope material damage sensitivity coefficient at the i-th critical location. The duration between adjacent moments in the historical monitoring cycle. Let be the comprehensive deviation distance of the i-th key location at the j-th time point in the monitoring period. Let K be the first standard deviation, K be the number of physical quantities monitored at key locations, k≤K, E be the number of historical monitoring cycles, e≤E, m be the number of moments in the monitoring cycle, j≤m, and k, K, e, E, j, and m are all positive integers.

5. The slope stability monitoring method based on artificial intelligence according to claim 1, characterized in that, The overall slope stability coefficient is determined based on the critical location stability coefficient, including: determining the minimum critical location stability coefficient based on the critical location stability coefficients of multiple critical locations; determining the dangerous locations of the slope based on the minimum critical location stability coefficient; determining the danger distance between each critical location and the dangerous locations of the slope; and determining the overall slope stability coefficient based on the minimum critical location stability coefficient, the critical location stability coefficients, and the danger distances.

6. The slope stability monitoring method based on artificial intelligence according to claim 5, characterized in that, The overall slope stability coefficient is determined based on the minimum critical location stability coefficient, the critical location stability coefficient, and the danger distance, including: according to the formula: Determine the overall slope stability coefficient at time j of the monitoring period. ,in, Let be the minimum critical position stability coefficient at time j of the monitoring period. Let i be the key position stability coefficient at the j-th time of the monitoring period. Let be the critical distance between the i-th critical location and the dangerous location on the slope. This is the preset distance.

7. The slope stability monitoring method based on artificial intelligence according to claim 1, characterized in that, The training steps of the slope stability prediction model include: determining the historical key location stability coefficient and the historical overall slope stability coefficient based on historical key location monitoring data and historical external environment data at multiple historical prediction times; processing the historical key location monitoring data and the historical external environment data through the slope stability prediction model to determine the historical sample predicted key location stability coefficient and the historical sample predicted overall slope stability coefficient at multiple historical prediction times; determining the training loss function of the slope stability prediction model based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient; and training the slope stability prediction model based on the training loss function to obtain the trained slope stability prediction model.

8. The slope stability monitoring method based on artificial intelligence according to claim 7, characterized in that, Based on the historical key location stability coefficient, the historical overall slope stability coefficient, the historical sample predicted key location stability coefficient, and the historical sample predicted overall slope stability coefficient, the training loss function of the slope stability prediction model is determined, including: according to the formula: Determine the training loss function for the slope stability prediction model. Where max is the function for finding the maximum value. These are preset parameters. Let be the historical key position stability coefficient of the i-th key position at the r-th historical prediction time. For the i-th critical location, predict the critical location stability coefficient for historical samples at the r-th historical prediction time. Let be the historical overall slope stability coefficient at the r-th historical prediction time. For the historical sample at the r-th historical prediction time, predict the overall slope stability coefficient. For the i-th key position, predict the key position stability coefficient for the historical sample at the (r+1)-th historical prediction time. R is the number of historical prediction times, r≤R, n is the number of key positions, i≤n, and r, R, i and n are all positive integers.

9. An artificial intelligence-based slope stability monitoring system, characterized in that, A method for performing any one of claims 1-8, comprising: a monitoring data module for acquiring key location monitoring data at multiple times during a monitoring period using a combination of sensors positioned at key locations; a historical data module for acquiring historical key location monitoring data and historical external environment data for historical monitoring periods; an environmental data module for acquiring external environment data at multiple times during a monitoring period; a location assessment module for determining a key location stability coefficient based on the historical key location monitoring data, the key location monitoring data, and the external environment data; an overall assessment module for determining an overall slope stability coefficient based on the key location stability coefficient; a model prediction module for processing the key location monitoring data and the external environment data using a trained slope stability prediction model to obtain a predicted key location stability coefficient and a predicted overall slope stability coefficient; and a slope monitoring module for monitoring the slope based on the key location stability coefficient, the overall slope stability coefficient, the predicted key location stability coefficient, and the predicted overall slope stability coefficient.

Images