Slope monitoring method and system based on multi-source data analysis

Abstract

The invention relates to the field of slope monitoring, in particular to a slope monitoring method and system based on multi-source data analysis, and the method comprises the steps: obtaining slope monitoring data including surface displacement, rainfall and pore water pressure; training a prediction model based on historical data in the monitoring data; monitoring data obtained in real time are input into the prediction model, and the expected surface displacement at the current moment is obtained; calculating the difference between the surface displacement and the expected surface displacement to obtain a deformation residual error; calculating a comprehensive instability index based on a plurality of deformation residual errors in a time window with a set length, wherein the comprehensive instability index comprises an accumulated plasticity index and a time offset; and generating an alarm signal in response to the condition that the comprehensive instability index is greater than a set threshold. The method solves the problem that traditional single-index monitoring is susceptible to environmental interference, so that false alarm and missing alarm are caused.

Description

Technical Field

[0001] This invention relates to the field of slope monitoring, and in particular to a slope monitoring method and system based on multi-source data analysis. Background Technology

[0002] Slope stability monitoring is a crucial aspect of the operational safety of infrastructure such as transportation, water conservancy, and mining. With increasing emphasis on geological disaster risks, traditional point-based, periodic manual inspections are no longer sufficient to meet the demands for real-time and accurate monitoring of high-risk slopes. In recent years, automated monitoring based on multi-source sensors has been applied to slope monitoring. Existing automated slope monitoring systems typically include the following three types of sensor subsystems: surface deformation monitoring systems, employing technologies such as global navigation satellite systems, synthetic aperture radar interferometry, or close-range photogrammetry to acquire the three-dimensional displacement and rate of change of key points on the slope surface; internal condition monitoring systems, including inclinometers, fiber optic strain gauges (FBG), and pore water pressure gauges, used to detect the internal displacement, stress-strain state, and groundwater dynamics of deep soil or rock masses in the slope; and environmental factor monitoring systems, primarily monitoring external load factors such as rainfall, temperature, water level, and soil moisture content.

[0003] In practical engineering applications, existing monitoring and analysis methods have the following problems: current early warning criteria are mostly based on a single indicator threshold triggering mechanism. For example, an alarm is triggered when the displacement rate exceeds a certain fixed value. This method is easily affected by environmental factors. For instance, heavy rainfall can cause slope soil to undergo significant recoverable elastic deformation due to water saturation, which may trigger false alarms and lead to a judgment of abnormal slope condition; while in the dry season without significant rainfall, creep (plastic deformation) caused by internal structural deterioration is ignored because it does not reach the threshold, resulting in missed alarms. Summary of the Invention

[0004] To address the problem that existing technologies are prone to false alarms and missed alarms due to the influence of the external environment, resulting in low accuracy of monitoring results, this invention provides a slope monitoring method and system based on multi-source data analysis.

[0005] In a first aspect, the present invention provides a slope monitoring method based on multi-source data analysis, employing the following technical solution: The slope monitoring method based on multi-source data analysis includes the following steps: Acquire slope monitoring data including surface displacement, rainfall, and pore water pressure; train a prediction model based on historical data from the monitoring data; input the real-time acquired monitoring data into the prediction model to obtain the expected surface displacement at the current moment; calculate the difference between the surface displacement and the expected surface displacement to obtain the deformation residual; calculate the comprehensive instability index based on multiple deformation residuals within a set time window, the comprehensive instability index including cumulative plasticity index and time offset; The cumulative plasticity index is the sum of all deformation residuals within the time window; the time offset is the time lag corresponding to the maximum cross-correlation coefficient between the rainfall sequence and the pore water pressure sequence in the monitoring data. The comprehensive instability index is positively correlated with the cumulative plasticity index and negatively correlated with the time offset; an alarm signal is generated in response to the comprehensive instability index exceeding the set threshold.

[0006] By training a predictive model to predict expected surface displacements caused by environmental factors and calculating the deformation residuals between the actual and expected displacements, environmental interference is effectively eliminated, and irreversible plastic deformation caused by the deterioration of the slope's internal structure is identified. Furthermore, this invention introduces the time offset of rainfall and pore water pressure sequences as another indicator, reflecting changes in the slope's internal permeability. Finally, by constructing a comprehensive instability index that includes cumulative plasticity indicators and time offsets, this method can comprehensively assess slope stability from both external deformation and internal hydrological response dimensions, improving monitoring accuracy and early warning reliability, and effectively avoiding the false alarms and missed alarms of traditional methods.

[0007] Preferably, the cumulative sum of multiple deformation residuals within the time window is used as the cumulative plasticity index.

[0008] By summing up multiple deformation residuals within a time window, weak but persistent creep deformation trends can be effectively amplified. This method is more sensitive to early, slow plastic deformation, enabling earlier identification of potential slope instability risks.

[0009] Preferably, the time offset is calculated by constructing a pore water pressure sequence using pore water pressure at multiple times. Using rainfall data from multiple times Constructing rainfall series Analysis of rainfall series using cross-correlation function With pore water pressure sequence The time delay relationship between the two sequences is determined by using the offset at which the correlation between the two sequences is at its maximum as the time offset.

[0010] By using cross-correlation function analysis on two time series, rainfall and pore water pressure, the lag time of their responses can be accurately and quantitatively determined, providing a basis for comprehensive judgment.

[0011] The preferred expression for the comprehensive instability index is:

[0012] in, This represents the overall instability index of the slope at time t. This represents the cumulative plasticity index within a time window of length T. This represents the time offset where the correlation between the rainfall sequence and the pore water pressure sequence is greatest. This represents the normalized time offset. This represents the preset first weighting coefficient. This represents the preset second weighting coefficient.

[0013] By weighting and summing the cumulative plasticity index and the normalized time offset, the comprehensive instability index can more comprehensively reflect the overall instability risk of the slope.

[0014] Preferably, the weighting coefficient is calculated as follows: within the time window, the variance of the deformation residual and the variance of the rainfall are calculated, and the ratio of the variance of the deformation residual to the variance of the rainfall is used as the sensitivity. The sensitivity is normalized using a linear normalization algorithm, and the normalized sensitivity is used as the first weighting coefficient. The difference between 1 and the first weighting coefficient is used as the second weighting coefficient.

[0015] By dynamically adjusting the weights by calculating the ratio of the variance of deformation residuals to the variance of rainfall, the slope becomes more sensitive to deformation response to rainfall when slope stability decreases. This method can automatically increase the weight of the cumulative plasticity index, enabling the comprehensive instability index to grow faster and trigger an alarm, thereby achieving adaptive enhancement of the early warning model and improving the early warning sensitivity in near-instability states.

[0016] Preferably, the normalization process for the internal response hysteresis factor includes:

[0017] in, This represents the normalized time offset. This represents the historical average time offset. This represents the time offset at which the correlation between the rainfall sequence and the pore water pressure sequence is greatest.

[0018] By comparing and normalizing with historical average time offsets, the influence of absolute time length is eliminated, allowing the index to be weighted and fused with the cumulative plasticity index on a unified scale, thus enhancing the stability and comparability of the model.

[0019] Preferably, the prediction model is a long short-term memory network model. The long short-term memory network model is trained using mean squared error as the loss function and the Adam optimizer is used for parameter iteration until the validation set error converges.

[0020] LSTM models excel at capturing and learning long-term dependencies and complex nonlinear patterns in time-series data. Therefore, using LSTM can more accurately predict slope surface displacement influenced by multiple historical factors such as rainfall, thereby calculating more accurate deformation residuals and providing a solid foundation for the accuracy of the entire monitoring method.

[0021] Preferably, acquiring slope monitoring data including surface displacement, rainfall, and pore water pressure includes: setting up multiple GNSS receivers on the slope to collect the monitored surface displacement of the slope; installing inclinometers near the sliding surface and configuring pore water pressure sensors to collect the pore water pressure at the monitored slope; and deploying tipping bucket rain gauges to collect rainfall data.

[0022] Preferably, determining the stability state of a slope based on the comprehensive instability index includes: acquiring comprehensive instability indices at multiple historical moments and constructing a comprehensive instability index sequence; calculating the rate of change of the comprehensive instability index sequence to obtain a rate of change sequence; and inputting the comprehensive instability index sequence and the rate of change sequence into a monitoring model to obtain the stability state of the slope.

[0023] It not only uses the instantaneous value of the index to make threshold judgments, but also analyzes its time series and rate of change to determine the stability state of the slope. It can distinguish whether the slope is in a stable high-risk state or a dangerous state that is deteriorating rapidly, thus enabling a variety of early warning levels.

[0024] Secondly, the present invention provides a slope monitoring system based on multi-source data analysis, which adopts the following technical solution: The slope monitoring system based on multi-source data analysis includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the slope monitoring method based on multi-source data analysis described above.

[0025] The aforementioned slope monitoring method based on multi-source data analysis is used to generate a computer program, which is then stored in a memory for loading and execution by a processor. This allows for the creation of a system based on the memory and processor, making it convenient to use.

[0026] The present invention has the following technical effects: This invention proposes a slope instability early warning method based on deformation residuals and time offsets. It effectively separates elastic deformation caused by environmental factors from plastic deformation caused by the deterioration of the slope's own structure using an LSTM prediction model. Furthermore, by combining rainfall pore pressure time lags reflecting changes in internal hydrological characteristics, a dynamically weighted comprehensive instability index is constructed. This solves the problem of false alarms and missed alarms caused by environmental interference in traditional single-indicator monitoring, achieving accurate assessment of slope stability and improving the accuracy and reliability of early warnings. Attached Figure Description

[0027] Figure 1 This is a flowchart of the slope monitoring method based on multi-source data analysis of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0029] This invention discloses a slope monitoring method based on multi-source data analysis, referring to... Figure 1 This includes the following steps: S1: Acquire slope data and perform preprocessing.

[0030] Multiple GNSS receivers were installed on the slope to collect data on the surface displacement of the slope. It is used for continuous monitoring of the three-dimensional displacement of the slope surface; inclinometer tubes are installed near the potential sliding surface, and pore water pressure sensors are configured to collect and monitor the pore water pressure at the slope. At the same time, tipping bucket rain gauges were deployed to collect rainfall data. .

[0031] The acquired data is time-aligned for surface displacement. Differential processing was used to eliminate atmospheric delay and satellite orbit errors. Rainfall data. Moving average smoothing is used to eliminate instantaneous impulse noise; pore water pressure Step outliers caused by sensor malfunctions were removed. Finally, a linear normalization algorithm was used to normalize all slope data, ensuring that their values ​​fell within the range of [0,1].

[0032] S2: Extract slope deformation residuals based on expected surface displacement.

[0033] Historical data from the past two years were selected as the training set, which includes multiple rainy and dry season cycles, to ensure that the prediction model can learn the slope response patterns under different environmental conditions.

[0034] Construct a Long Short-Term Memory (LSTM) network model. The inputs to the LSM model include: the current time t and the rainfall amounts from the previous n times, i.e. The cumulative rainfall in the previous 72 hours and the current temperature and the historical surface displacement of the previous 72 hours The output is the expected surface displacement at the current moment. The Long Short-Term Memory (LSTM) network model was trained using mean squared error (MSE) as the loss function and the Adam optimizer was used for parameter iteration until the validation set error converged. After training, the model achieved a prediction accuracy of ±3 mm on the test set, indicating that it can effectively capture the complex nonlinear relationship between environmental factors and surface displacement.

[0035] In actual monitoring, the latest slope data is input into the long short-term memory network model to obtain the expected surface displacement at the current moment. And calculate the deformation residual, the expression is:

[0036] in, This represents the deformation residual of the slope at the current time t. This represents the expected surface displacement of the slope at the current time t. This indicates the monitored surface displacement of the slope at the current time t.

[0037] The above calculation method decomposes the original monitored surface displacement signal into environmental elastic deformation response and structural change response, reducing the interference of factors such as rainfall on subsequent instability assessment. The environmental elastic deformation response can be understood as the expected surface displacement. For normal deformation, the structural change response can be understood as the deformation residual. Therefore, the deformation residual can initially reflect the changes in the internal structure of the slope, and is an abnormal deformation.

[0038] For example, during a period of heavy rainfall, the actual monitored surface displacement... The value is 8.5 mm, while the expected surface displacement predicted by the long short-term memory network model is... If it is 7.2mm, then the deformation residual mm. This residual value represents the abnormal deformation beyond the influence of the environment, and can preliminarily reflect the changes in the internal structure of the slope.

[0039] S3: Extract multidimensional instability features of slopes.

[0040] S31: Calculate the cumulative plasticity index.

[0041] Using the current moment as the endpoint, a time window is constructed along the historical direction. The length of the time window is set manually according to the actual situation. For example, the length of the time window is the past 7 days. The cumulative plasticity index is calculated based on the time window, and the expression is:

[0042] in, This represents the cumulative plasticity index within a time window of length T. This represents the deformation residual of the slope at time m, where T represents the length of the time window and m represents the index of the time. Indicates the initial time within the time window.

[0043] Deformation residual In the short term, the creep may be very weak and fluctuate little. The cumulative plasticity index can amplify this weak but continuous unidirectional trend, thereby effectively identifying early creep deformation. The cumulative plasticity index can quantify the cumulative effect of irreversible deformation.

[0044] S32: Calculate the time offset between pore water pressure and rainfall.

[0045] Constructing a pore water pressure sequence using pore water pressure at multiple time points Using rainfall data from multiple times Constructing rainfall series Analysis of rainfall series using cross-correlation function With pore water pressure sequence The time delay relationship between them is defined as follows:

[0046] in, This represents the time offset at which the correlation between the two sequences is at its maximum. This represents the cross-correlation function.

[0047] Under normal circumstances, rainwater infiltration takes a certain amount of time to cause changes in deep pore water pressure. It takes approximately 12–24 hours for the slope to develop internal fissures and increase permeability. The time will tend to decrease (e.g., to 6–12 hours), indicating that moisture reaches the potential slip surface more quickly, increasing the risk of instability.

[0048] S4: Calculate the sensitivity of the slope.

[0049] When internal damage accumulates and approaches instability, the slope's stability decreases. At this point, even minor environmental disturbances (such as moderate rain) can cause a larger deformation response than during the stable period. The slope's response to the external environment is considered its sensitivity, and the calculation method is as follows: Within the constructed time window, the variance of the deformation residuals and the variance of the rainfall data are calculated. The expression for sensitivity is:

[0050] in, This represents the sensitivity to slope when the time window length is T. This represents the variance of the deformation residuals within the time window. This represents the variance of rainfall data within the time window. This represents a hyperparameter with a value of 0.1 to prevent the denominator from being zero.

[0051] This reflects the response intensity of slope structural deformation caused by a unit amount of rainfall disturbance. During the slope stability phase, even with significant rainfall, Small fluctuations The residuals remain at a low level; however, when slope stability declines and tends towards instability, the same amount of rainfall will cause greater residual fluctuations, leading to... Increase.

[0052] S5: Construct a dynamically weighted comprehensive instability index.

[0053] Using the linear normalization algorithm to analyze the cumulative plasticity index Normalization was performed, and then the cumulative plasticity index was used as the basis. With time offset The comprehensive instability index is calculated using the following expression:

[0054]

[0055] in, This represents the overall instability index of the slope at time t. This represents the cumulative plasticity index within a time window of length T. This represents the time offset where the correlation between the rainfall sequence and the pore water pressure sequence is greatest. This represents the normalized time offset. This is the historical average time offset, for example, the mean of time offsets over the past two days, used for normalization. This represents the preset first weighting coefficient. This represents the preset second weighting coefficient.

[0056] Among them, the weighting coefficient , The calculation method is as follows: Sensitivity using linear normalization algorithm Perform normalization processing, and then normalize the sensitivity. As the first weighting coefficient , and the first weighting coefficient The difference is used as the second weighting coefficient. .

[0057] When the slope stability is good The value is small, that is When the value is small, at this time Even slight fluctuations will be Suppression leads to an overall instability index The increase is slow, at which point the cumulative plasticity index... The practical meaning of slow growth is: the elastic deformation of the slope soil caused by the influence of the external environment, such as: the obvious recoverable elastic deformation of the slope soil due to water saturation.

[0058] When the slope stability is poor The value is relatively large, that is A large value indicates that the slope has entered a sensitive state, at which point the cumulative plasticity index... The growth will be amplified rapidly, causing the overall instability index to... Rapid rise.

[0059] In summary, the overall instability index reflects the stability of the slope. The higher the value, the worse the slope stability, and the lower the value, the better the slope stability.

[0060] S6: Use the comprehensive instability index to monitor slopes.

[0061] A comprehensive instability index sequence is constructed using the combined instability index at the current moment and multiple adjacent historical moments. The rate of change of this sequence is then calculated to obtain a rate of change sequence. Finally, the comprehensive instability index sequence and the rate of change sequence are input into a monitoring model to obtain the slope's stability state, thus enabling slope monitoring. For example, the stability state includes: good, fair, and poor. The monitoring model is a support vector machine or a recurrent neural network.

[0062] This invention also discloses a slope monitoring system based on multi-source data analysis, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the slope monitoring method based on multi-source data analysis according to this invention.

[0063] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0064] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A slope monitoring method based on multi-source data analysis, characterized in that, The method comprises the steps of: obtaining slope monitoring data containing surface displacement, rainfall and pore water pressure; training a prediction model based on historical data in the monitoring data; inputting real-time monitoring data into the prediction model to obtain expected surface displacement at the current time; calculating the difference between the surface displacement and the expected surface displacement to obtain a deformation residual; calculating a comprehensive instability index based on a plurality of deformation residuals in a time window of a set length, the comprehensive instability index comprising a cumulative plasticity index and a time offset; the cumulative plasticity index is the sum of all deformation residuals in the time window; the time offset is the time lag corresponding to the maximum cross-correlation coefficient of the rainfall sequence and the pore water pressure sequence in the monitoring data; the comprehensive instability index is positively correlated with the cumulative plasticity index and negatively correlated with the time offset; in response to the comprehensive instability index being greater than a set threshold, an alarm signal is generated.

2. The method of claim 1, wherein, The cumulative sum of the plurality of deformation residuals in the time window is taken as the cumulative plasticity index. 3.The slope monitoring method based on multi-source data analysis of claim 1, wherein, The time offset is calculated by constructing a pore water pressure sequence using pore water pressure at multiple moments. Using rainfall data from multiple times Constructing rainfall series Analysis of rainfall series using cross-correlation function With pore water pressure sequence The time delay relationship between the two sequences is determined by using the offset at which the correlation between the two sequences is at its maximum as the time offset.

4. The slope monitoring method based on multi-source data analysis according to claim 1 or 3, characterized in that, The expression of the comprehensive instability index is: wherein, represents a comprehensive instability index of the slope at the current time t, represents a cumulative plasticity index in a time window with a length of T, represents a time offset when the correlation between the rainfall sequence and the pore water pressure sequence is the largest, represents a normalized time offset, represents a preset first weight coefficient, represents a preset second weight coefficient.

5. The method of claim 4, wherein, The calculation method of the weight coefficient is: in the time window, the variance of the deformation residual and the variance of the rainfall are calculated, the ratio of the variance of the deformation residual to the variance of the rainfall is taken as the sensitivity, the sensitivity is normalized by using a linear normalization algorithm, the normalized sensitivity is taken as the first weight coefficient, and the difference between 1 and the first weight coefficient is taken as the second weight coefficient.

6. The method of claim 4, wherein, The normalization processing of the internal response lag factor comprises: wherein, denotes the normalized time shift, is the historical average time shift, denotes the time shift at which the correlation between the rainfall series and the pore water pressure series is maximal.

7. The method of claim 1, wherein, The prediction model is a long short-term memory network model, the long short-term memory network model training uses mean square error as a loss function, and an Adam optimizer is used for parameter iteration until the validation set error converges.

8. The method of claim 1, wherein, The slope monitoring data containing surface displacement, rainfall and pore water pressure are obtained, including: setting multiple GNSS receivers on the slope to collect monitoring surface displacement of the slope; installing an inclinometer tube near the sliding surface and configuring a pore water pressure sensor to collect pore water pressure of the monitoring slope; and laying a tipping bucket rain gauge to collect rainfall data. 9.The slope monitoring method based on multi-source data analysis of claim 1, wherein, The stability state of the slope is determined based on the comprehensive instability index, including: obtaining a plurality of historical time comprehensive instability indexes and forming a comprehensive instability index sequence; calculating the change rate of the comprehensive instability index sequence to obtain a change rate sequence; inputting the comprehensive instability index sequence and the change rate sequence into a monitoring model to obtain the stability state of the slope.

10. A slope monitoring system based on multi-source data analysis, characterized in that, The method comprises: a processor and a memory, the memory storing computer program instructions, when the computer program instructions are executed by the processor, the method for monitoring a slope based on multi-source data analysis according to any one of claims 1-9 is realized.

Images