Real-time monitoring method and system for geotechnical engineering slope stability

Abstract

The application discloses a geotechnical engineering slope stability real-time monitoring method and system, which introduces blind source separation technology combining independent component analysis and physical constraints, takes original displacement time series and environmental temperature time series of multiple GNSS measuring points as multi-channel mixed signals, utilizes statistical independence between signal sources, effectively separates periodic environmental noise, instrument random noise and drift from mixed observation signals and real slope deformation signals; subsequently, through correlation verification with physical quantities such as field environmental temperature, the separated source signals are automatically calibrated, temperature effect source signals and long-term creep source signals with physical labels are accurately identified and extracted, thereby realizing accurate removal of noise components and high-fidelity reconstruction of pure deformation signals. In this way, the accuracy of geotechnical engineering slope stability real-time monitoring can be effectively improved, and a solid data foundation and decision basis are provided for early warning and prevention of slope disasters.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring, and more specifically, to a method and system for real-time monitoring of slope stability in geotechnical engineering. Background Technology

[0002] The stability of slopes in geotechnical engineering is directly related to the safety of people's lives and property and the smooth operation of engineering projects. Therefore, real-time and accurate monitoring and early warning are crucial. However, traditional slope monitoring methods face many challenges in practical applications, especially in complex and ever-changing field environments, where the reliability of monitoring data is often compromised.

[0003] Existing high-precision monitoring systems based on GNSS and other technologies are severely affected by various types of noise when acquiring millimeter-level slope deformation data. On the one hand, changes in ambient temperature cause thermal expansion and contraction of monitoring piers, observation instruments, and even the surface soil and rock of the slope, resulting in periodic diurnal fluctuations in displacement data. The amplitude of this periodic noise is much greater than the weak actual creep deformation in the early stages of slope instability, making it difficult for analysts to separate the weak but continuously increasing "weak signal" from the "strong background." On the other hand, the inherent random noise of the instruments and the long-term zero-point drift problem are also common. Even under constant environmental conditions, GNSS calculations still contain high-frequency random "glitch" and tiny, unidirectional zero-point drifts that occur over time (months or years). This makes instantaneous velocity and acceleration calculations extremely unreliable. Moreover, this low-frequency drift is very similar in morphology to the long-term slow creep of slopes, making it easy to confuse and potentially misdiagnose real deformation as an instrument problem or vice versa. In addition, the signal-to-noise ratio of real slope deformation signals is extremely low in the early creep stage, and the signal strength only increases significantly when entering the acceleration stage, but by then the best warning opportunity has often been missed. At the same time, real deformation is not a simple linear growth, but often exhibits a complex nonlinear step growth due to the coupling of external factors such as rainfall and reservoir water level changes. These complex signals are easily smoothed out by traditional filtering methods along with the noise, resulting in delayed or missed warnings.

[0004] Therefore, an optimized method for real-time monitoring of slope stability in geotechnical engineering is needed. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method and system for real-time monitoring of slope stability in geotechnical engineering. By introducing blind source separation technology combining independent component analysis and physical constraints, the original displacement time series and ambient temperature time series of multiple GNSS measuring points are treated as multi-channel mixed signals. Utilizing the statistical independence between signal sources, periodic environmental noise, instrument random noise, drift, and the actual slope deformation signal are effectively separated from the mixed observation signal. Subsequently, through correlation verification with physical quantities such as ambient temperature, the separated source signals are automatically calibrated, accurately identifying and extracting temperature effect source signals and long-term creep source signals with physical labels. This achieves precise removal of noise components and high-fidelity reconstruction of pure deformation signals. Thus, the accuracy and reliability of real-time monitoring of slope stability in geotechnical engineering can be effectively improved, providing a solid data foundation and decision-making basis for early warning and prevention of slope disasters.

[0006] According to one aspect of this application, a method for real-time monitoring of slope stability in geotechnical engineering is provided, comprising:

[0007] Acquire the raw displacement time series and ambient temperature time series of multiple GNSS measuring points;

[0008] The original displacement time series of multiple GNSS measuring points are centered and matrix constructed to obtain the observation signal matrix;

[0009] The observed signal matrix is ​​separated into source signals based on independent component analysis to obtain the estimated independent source signal matrix and the estimated mixture matrix;

[0010] The estimated independent source signal matrix and the on-site ambient temperature time series are used to automatically calibrate the source signals to obtain physically tagged source signals;

[0011] Based on the source signal with physical labels, the estimated mixing matrix and the observed signal matrix are reconstructed using high-fidelity deformed signal to obtain the high-fidelity deformed signal matrix.

[0012] Trend analysis was performed on the high-fidelity deformation signal matrix to obtain the deformation rate and acceleration rate.

[0013] According to another aspect of this application, a real-time monitoring system for slope stability in geotechnical engineering is provided, comprising:

[0014] The data acquisition module is used to acquire the original displacement time series and ambient temperature time series of multiple GNSS measuring points;

[0015] The centralization and matrix construction module is used to centralize and construct matrices from the original displacement time series of multiple GNSS measuring points to obtain the observation signal matrix;

[0016] The source signal separation module is used to perform source signal separation on the observed signal matrix based on independent component analysis to obtain the estimated independent source signal matrix and the estimated mixture matrix;

[0017] The automatic source signal calibration module is used to automatically calibrate the estimated independent source signal matrix and the on-site ambient temperature time series to obtain source signals with physical labels.

[0018] The high-fidelity deformable signal reconstruction module is used to reconstruct the estimated mixing matrix and the observed signal matrix in high fidelity based on the source signal with physical labels to obtain the high-fidelity deformable signal matrix.

[0019] The trend analysis module is used to perform trend analysis on the high-fidelity deformation signal matrix to obtain the deformation rate and acceleration rate.

[0020] Compared with existing technologies, this application provides a real-time monitoring method and system for slope stability in geotechnical engineering. By introducing blind source separation technology combining independent component analysis and physical constraints, it treats the original displacement time series and ambient temperature time series from multiple GNSS measuring points as multi-channel mixed signals. Utilizing the statistical independence between signal sources, it effectively separates periodic environmental noise, instrument random noise, drift, and the actual slope deformation signal from the mixed observation signal. Subsequently, through correlation verification with physical quantities such as ambient temperature, the separated source signals are automatically calibrated, accurately identifying and extracting temperature effect source signals and long-term creep source signals with physical labels. This achieves precise removal of noise components and high-fidelity reconstruction of pure deformation signals. Thus, it effectively improves the accuracy and reliability of real-time monitoring of slope stability in geotechnical engineering, providing a solid data foundation and decision-making basis for early warning and prevention of slope disasters. Attached Figure Description

[0021] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 This is a flowchart of a method for real-time monitoring of slope stability in geotechnical engineering according to an embodiment of this application;

[0023] Figure 2 This is a schematic diagram of the data flow of the real-time monitoring method for slope stability in geotechnical engineering according to an embodiment of this application;

[0024] Figure 3 This is a block diagram of a real-time monitoring system for slope stability in geotechnical engineering according to an embodiment of this application. Detailed Implementation

[0025] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0026] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0027] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0028] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0029] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0030] The technical solution of this application proposes a method for real-time monitoring of slope stability in geotechnical engineering. Figure 1 This is a flowchart of a method for real-time monitoring of slope stability in geotechnical engineering according to an embodiment of this application. Figure 2 This is a system architecture diagram of a real-time monitoring method for slope stability in geotechnical engineering according to an embodiment of this application. Figure 1 and Figure 2As shown, the real-time monitoring method for slope stability in geotechnical engineering according to an embodiment of this application includes the following steps: S1, acquiring the original displacement time series and the ambient temperature time series of multiple GNSS measuring points; S2, performing centering processing and matrix construction on the original displacement time series of multiple GNSS measuring points to obtain an observation signal matrix; S3, performing source signal separation based on independent component analysis on the observation signal matrix to obtain an estimated independent source signal matrix and an estimated mixed matrix; S4, performing automatic source signal calibration on the estimated independent source signal matrix and the ambient temperature time series to obtain source signals with physical labels; S5, based on the source signals with physical labels, performing high-fidelity deformation signal reconstruction on the estimated mixed matrix and the observation signal matrix to obtain a high-fidelity deformation signal matrix; S6, performing trend analysis on the high-fidelity deformation signal matrix to obtain the deformation rate and acceleration rate.

[0031] Specifically, S1 involves acquiring the raw displacement time series and ambient temperature time series from multiple GNSS measuring points. It should be understood that acquiring the raw displacement time series from multiple GNSS measuring points is to directly capture the actual deformation of the slope surface, which is a core indicator for assessing slope stability. However, the true deformation signal of the slope is often obscured by various noises, such as the periodic "day and night" fluctuations in pseudo-displacements caused by changes in ambient temperature, the thermal expansion and contraction of monitoring piers, observation instruments, and even the surface soil and rock, as well as the inherent random noise and long-term zero-point drift of the instruments. The amplitude of these noises is sometimes much greater than the weak actual creep deformation in the early stages of slope instability, making it difficult for analysts to accurately identify the true deformation trend. Therefore, in order to accurately separate the pure deformation signal from these mixed signals and effectively eliminate interference such as temperature effects, it is indispensable to simultaneously acquire high-precision GNSS displacement data and ambient temperature data. Acquiring the temperature time series is crucial for the subsequent automatic calibration of the source signal. As an external physical quantity, it can help distinguish and compensate for the heat conduction hysteresis effect, ensure the accurate identification of the temperature effect component, and thus guarantee the fidelity of the final deformation signal.

[0032] Among them, GNSS measuring points refer to global navigation satellite system receiving points installed on geotechnical engineering slopes to monitor their displacement changes. They accurately determine their own position by receiving satellite signals and record the changes in position over time.

[0033] In practice, multiple GNSS (Global Navigation Satellite System) receivers are strategically deployed as monitoring points on the geotechnical slope to be monitored. These GNSS points continuously and in real-time acquire their position information in three-dimensional space, and through high-precision calculations, obtain the displacement data of each point over time, forming their respective original displacement time series. Simultaneously, environmental temperature sensors are deployed at the slope site or near the monitoring points to record the environmental temperature time series in real-time and continuously. It is worth noting that this process is continuous, ensuring high-precision synchronous acquisition of GNSS displacement data and environmental temperature data on the time axis, so that subsequent analysis can accurately correlate displacement changes with temperature fluctuations. This synchronous acquisition mode is key to achieving automatic calibration of subsequent source signals. It allows the system to accurately identify and eliminate displacement interference caused by temperature effects during the data processing stage, based on the optimal correlation between temperature signals and independent source signals over time delay.

[0034] Specifically, in step S2, the original displacement time series of multiple GNSS measuring points are centered and matrix constructed to obtain the observation signal matrix. In the technical solution of this application, in order to enable the subsequent independent component analysis to be performed effectively and to extract physically meaningful source signals, the original GNSS displacement time series needs to undergo necessary preprocessing. Among them, centering is a common data preprocessing technique, the purpose of which is to eliminate the mean (or DC component) in the time series. By subtracting the mean of the series from each data point, the mean of the processed time series is made zero. This helps to highlight the fluctuation characteristics of the data and meets the requirement of zero mean for some signal processing algorithms (such as independent component analysis), thereby improving the accuracy and efficiency of the analysis. Matrix construction refers to organizing multiple independent, preprocessed (e.g., centered) time series data into a two-dimensional or multi-dimensional mathematical matrix according to specific arrangement rules. In this step, the time series of each GNSS measuring point is usually used as a row or column of the matrix, and the observation values ​​of all measuring points at all time points are collected in a matrix to form a unified data structure, which facilitates subsequent batch calculations and multivariate analysis. Specifically, centering is a common prerequisite for independent component analysis (ICA) algorithms. It eliminates the mean component of the data, ensuring that the signal fluctuates near the zero mean. This simplifies model construction and allows the algorithm to focus more on the second and higher-order statistical properties of the data, thus better separating independent source signals. Matrix construction integrates scattered time-series data from different GNSS measurement points into a unified observation signal matrix, providing a standardized input format for multivariate signal processing (such as ICA). This enables simultaneous analysis of displacement changes at multiple measurement points at different time points, capturing the overall deformation pattern of the slope. Through these preprocessing steps, DC bias or fixed offsets in the data can be effectively removed, improving the accuracy and robustness of subsequent analysis and laying the foundation for accurately identifying the driving factors of slope deformation.

[0035] Specifically, S3 involves performing source signal separation based on independent component analysis (ICA) on the observed signal matrix to obtain an estimated independent source signal matrix and an estimated mixture matrix. The original displacement time series observed by GNSS measuring points is often the result of multiple physical processes. These processes include long-term creep of the slope itself, thermal expansion and contraction caused by changes in ambient temperature, soil softening due to rainfall infiltration, tectonic movements, and potential local instability deformation. When these independent physical source signals reach each GNSS measuring point, they linearly superimpose in some unknown way, forming a mixed observation signal. Traditional signal processing methods struggle to effectively distinguish and separate these mixed source signals, making it impossible to accurately identify the key deformation components that truly reflect the slope's stability. Independent component analysis (ICA) provides a powerful blind source separation capability. It assumes that the observed signal is a linear mixture of statistically independent non-Gaussian source signals through an unknown mixture matrix. Through ICA, these independent source signals and their mixing methods can be derived in reverse, thereby decoupling the complex coupling effects in slope deformation data and revealing the underlying single, pure physical driving factors. This separation process is crucial for subsequent identification of different deformation components (such as temperature effects and long-term creep) and for accurate reconstruction of deformation signals, which helps improve the monitoring system's ability to capture weak precursor signals of instability.

[0036] In practice, the observed signal matrix is ​​first centered and whitened to obtain a whitened signal matrix, a whitened matrix, and a dewhitened matrix. This step is a standard preprocessing step in the ICA algorithm, aiming to convert the observed signal into data with unit variance and no correlation, thereby simplifying the subsequent iterative optimization process and improving the algorithm's convergence speed and accuracy. Here, whitening is a linear transformation designed to eliminate the second-order statistical correlation of the observed signal, ensuring that the whitened signal components are uncorrelated and that each component has a variance of 1. This can be achieved by performing eigenvalue decomposition or singular value decomposition on the observed signal matrix. Specifically, the covariance matrix of the observed signal matrix is ​​first calculated; then, eigenvalue decomposition is performed on the covariance matrix: this process is expressed by the formula:

[0037] ,

[0038] in, It is the eigenvector matrix. It is a diagonal matrix composed of eigenvalues;

[0039] The whitening matrix can be constructed using the following formula:

[0040] ,

[0041] in, It is a whitening matrix;

[0042] Then, the observed signal matrix is ​​multiplied by the whitening matrix to obtain the whitened signal matrix. This process can be expressed by the following formula:

[0043] ,

[0044] in, It is the whitened signal matrix. The observed signal matrix is ​​used; the rows (or columns, depending on the matrix organization) of the whitened signal matrix are uncorrelated and have unit variance. Simultaneously, for subsequent reconstruction, a dewhitening matrix is ​​needed, which is typically the inverse or pseudo-inverse of the whitening matrix.

[0045] ,

[0046] in, It is a whitening matrix;

[0047] Next, the whitened signal matrix is ​​iteratively optimized to obtain the orthogonal rotation matrix. It should be understood that after whitening, although the signal components are no longer correlated, they may still not be statistically independent. The core task of ICA is to find an orthogonal rotation matrix in the whitening space such that, when applied to the whitened signal, it maximizes its non-Gaussianity (i.e., statistical independence). This process is typically implemented using iterative optimization algorithms, such as FastICA and Infomax. Specifically, the ICA algorithm iteratively updates the orthogonal rotation matrix to maximize a non-Gaussianity metric (such as negative entropy or kurtosis) until convergence. Assume the update rule during the iteration process is as follows:

[0048] ,

[0049] in, This represents a specific ICA optimization algorithm. In each iteration, this algorithm adjusts and orthogonalizes the elements of the orthogonal rotation matrix. The iterations continue until the orthogonal rotation matrix reaches a convergence condition: the change in the orthogonal rotation matrix between consecutive iterations is less than a preset threshold, or the maximum number of iterations is reached. The final converged matrix is ​​the orthogonal rotation matrix.

[0050] Furthermore, based on the orthogonal rotation matrix, whitening matrix, and dewhitening matrix, the observed signal matrix is ​​structurally reconstructed to obtain the estimated independent source signal matrix and the estimated mixture matrix. The independent source signal matrix is ​​obtained by multiplying the whitened signal matrix by the final orthogonal rotation matrix; this process is expressed by the following formula:

[0051] ,

[0052] in, It is an independent source signal matrix, matrix Each row (or column) represents an estimated time series of independent source signals that are statistically independent and have specific physical meanings (such as temperature effects, long-term creep, etc.).

[0053] The estimated mixing matrix describes how independent source signals superimpose to form the observed signal. It can be obtained by taking the inverses of the dewhitening matrix and the orthogonal rotation matrix, a process expressed by the formula:

[0054] ,

[0055] in, It is the transpose of an orthogonal rotation matrix. It is the estimated mixing matrix. Each column of the matrix represents the contribution weight of an independent source signal to each observation point, that is, it describes how the source signals are mixed to form the observation signal.

[0056] Specifically, in step S4, the estimated independent source signal matrix and the field ambient temperature time series are automatically calibrated to obtain physically labeled source signals. Although independent component analysis can decompose complex mixed observation signals into statistically independent source signals, these separated source signals themselves do not have direct physical meaning labels. For example, one source signal may be caused by changes in ambient temperature, another by long-term creep of slope soil, some may represent actual precursors to structural instability, while others may simply be system noise. Without identifying and assigning physical labels to these source signals, it is impossible to perform targeted noise removal, deformation reconstruction, and stability assessment.

[0057] In practice, firstly, based on the time series of ambient temperature, the estimated independent source signal matrix is ​​calibrated using strongly correlated external physical quantities to obtain the index of the temperature effect source signal and the set of uncalibrated remaining source signal indices. It should be understood that the conventional method in geotechnical engineering monitoring, which uses the Pearson correlation coefficient to directly compare the separated source signals with the external temperature signal to calibrate the temperature effect components, has a fundamental physical limitation. Its underlying logic assumes that the slope structure's response to changes in ambient temperature is instantaneous and synchronous. However, in actual physical scenarios, the conduction of heat from the slope surface to the internal monitoring pier (or sensor) is a time-consuming process. This process is affected by the thermal capacity and thermal conductivity of the slope material, inevitably leading to a significant time delay in the structure's thermally induced deformation response relative to the ambient temperature—a heat conduction hysteresis effect. Therefore, the source signal waveform representing the temperature effect and the measured temperature signal waveform will exhibit a phase difference on the time axis. Conventional methods for calculating instantaneous correlation fail to account for this time lag. When a phase difference exists, they significantly underestimate the true correlation strength between the two signals, and may even lead to erroneous conclusions that they are uncorrelated, resulting in misjudgment or failure to identify temperature effect components. To overcome these shortcomings, a time-delay optimal correlation calibration method based on cross-correlation spectrum analysis is proposed. This method no longer relies on instantaneous linear correlation, but delves into the frequency domain structure of the signal. By quantifying the phase relationship between signals, it accurately captures and compensates for the time delay, thereby achieving robust calibration of physical correlation.

[0058] In this process, firstly, a time-domain to frequency-domain signal transformation is performed on the ambient temperature time series and the estimated independent source signal matrix to obtain the Fourier spectrum set of the independent source signals and the Fourier spectrum of the temperature signal. It should be understood that direct comparison in the time domain cannot effectively handle the delay problem between signals. Transforming the signal to the frequency domain allows it to be decomposed into a series of superimposed sine waves of different frequencies, and the time delay between signals is intuitively reflected in the frequency domain as the phase difference at different frequency components. Therefore, entering the frequency domain is a prerequisite for analyzing and quantifying this time delay relationship.

[0059] Specifically, Fourier transforms are performed on the time series of each independent source signal separated by the ICA algorithm, as well as the synchronously acquired temperature time series, mapping these signals from the time domain to the frequency domain to obtain their respective Fourier spectra. This process can be expressed by the following formula:

[0060] ,

[0061] in, Represents time; Represents frequency; It is the first A time series function of an independent source signal; It is a natural constant; The imaginary unit; This is the frequency domain representation of the source signal, i.e., the Fourier spectrum. The same operation is performed on the temperature signal to obtain its Fourier spectrum. This shifts the analytical perspective from comparing time-domain waveforms to comparing frequency-domain spectral structures, laying the foundation for subsequent time delay calculations using phase information and serving as the starting point for the entire time delay compensation correlation algorithm. It can be understood that the original time-series data is transformed into a frequency-domain complex spectrum containing amplitude and phase information. This makes the inherent periodicity and potential phase relationships of the signal explicit, facilitating subsequent, more in-depth mining of inter-signal relationships.

[0062] Next, the cross-power spectral density (CPSD) is calculated between the Fourier spectrum of the temperature signal and the Fourier spectra of each independent source signal in the set of Fourier spectra of the independent source signals to obtain the CPSD matrix. It should be understood that after obtaining the Fourier spectra of each signal, a mathematical tool is needed to simultaneously measure the energy correlation strength and phase difference between the two signals at various frequencies. CPSD is designed for this purpose; it fuses the common information and relative phase information of the two signal spectra through complex multiplication.

[0063] Specifically, the Fourier spectrum of each source signal Fourier spectrum of temperature signal complex conjugate Multiply them to obtain the cross-power spectral density between them. The process can be expressed by the following formula:

[0064] ,

[0065] in, It is the cross-power spectral density of the j-th source signal and the temperature signal; It is the Fourier spectrum of the j-th source signal; It is the complex conjugate of the Fourier spectrum of the temperature signal. This generates an intermediate spectrum function. This function fully encodes all the correlation information between the two original time series under the linear system assumption, and is the core bridge connecting the frequency domain phase difference and the time domain time delay;

[0066] This step generates a new complex spectral matrix. . The modulus at a specific frequency The magnitude of the value at that frequency reflects the correlation strength between the source signal and the temperature signal at that frequency component; while its phase angle at that frequency directly corresponds to the phase difference between the two at that frequency component.

[0067] It is understandable that, although the cross-power spectrum itself is complete in information, it is still in the frequency domain and not intuitive enough. In order to obtain a correlation metric that can be directly used for decision-making and has clear physical meaning, it is necessary to convert the phase information in the frequency domain back to the time domain, which is represented by a specific time delay, and find the "optimal" delay that maximizes the correlation.

[0068] Specifically, firstly, the calculated cross-power spectral density... Perform an inverse Fourier transform and normalize to obtain the result with time delay. Cross-correlation function of variables The process can be expressed by the following formula:

[0069] ,

[0070] Here, This indicates that the j-th source signal and the temperature signal have a time delay of... Correlation coefficient at time; Represents the inverse Fourier transform operator; and These are the standard deviations of the corresponding signals, and N is the number of points in the signal;

[0071] Then, based on the cross-power spectral density matrix, time-delay optimal correlation quantization and calibration are performed on the on-site ambient temperature time series and the estimated independent source signal matrix to obtain the index of the temperature effect source signal and the set of uncalibrated remaining source signal indices. That is, firstly, the peak value is searched on this cross-correlation function; this peak value is the time-delay optimal correlation coefficient. The corresponding horizontal axis is the optimal time delay. The process can be expressed by the following formula:

[0072] ,

[0073] Finally, iterate through all source signals, find the signal with the optimal correlation coefficient and the largest time delay, and index it. The component is identified as a temperature-dependent effect and compared with a preset strong correlation threshold to confirm the validity of the calibration; this process is expressed by the formula:

[0074] ,

[0075] In the above formula, It is the maximum correlation coefficient between the j-th source signal and the temperature signal under the optimal time delay; For time delay variables; The optimal time delay is needed to achieve maximum correlation; Indicates all possible time delays Find the maximum value above; It was ultimately identified as the index of the temperature effect source signal.

[0076] This solves the problem of underestimating correlation caused by time delay. The goal is not limited to zero delay, but to systematically find and determine a delay point within a broad range of delays that maximizes the revelation of the physical correlation between the two, and to use this maximum correlation value as the final basis for calibration.

[0077] In summary, this optimization mechanism can accurately identify the temperature effect components "disguised" by time delay and output their index. In addition to the index set U_1 of the remaining uncalibrated signals, an additional parameter with important physical significance—the optimal time delay—is provided. This parameter can be used to verify and analyze the thermal response characteristics of the slope monitoring system.

[0078] The improved mechanism significantly enhances the accuracy and robustness of temperature-related source signal identification in complex physical scenarios with thermal conduction hysteresis. By actively searching for and compensating for time delays, this method reveals and quantifies the true physical correlations overlooked by traditional methods, avoiding erroneous calibrations caused by signal phase mismatch. Simultaneously, by more accurately separating temperature noise components, it ensures that subsequent signal reconstruction processes produce purer deformation signals with higher fidelity. This directly impacts the core performance of real-time monitoring systems for slope stability in geotechnical engineering, effectively reducing the false alarm and false negative rates of early warning models and improving the ability to capture weak pre-instability deformation signals, thus providing a solid data foundation for earlier and more reliable disaster warnings.

[0079] Furthermore, internal time-series characteristic analysis and trend component calibration are performed on the unlabeled remaining source signal index set and the estimated independent source signal matrix to obtain the index of long-term creep source signals and the final remaining source signal index set.

[0080] Specifically, in step S5, based on physically labeled source signals, high-fidelity deformation signal reconstruction is performed on the estimated mixture matrix and the observation signal matrix to obtain a high-fidelity deformation signal matrix. It should be understood that although Independent Component Analysis (ICA) can decompose the mixed observation signal into statistically independent source signals, and the subsequent automatic source signal calibration step has already assigned physical labels to these source signals, distinguishing the true deformation signal from noise (such as displacement caused by temperature effects), the observation signal matrix is ​​still the result of a linear mixture of all source signals (including identified noise and true deformation components). To accurately assess slope stability, these identified noise components need to be removed from the total observation signal to obtain a pure deformation signal unaffected by these interference factors. Without this step, noise components will continue to exist in the data, potentially masking weak pre-instability deformation signals, leading to misjudgments or missed reports, severely impacting the reliability of the monitoring system and the timeliness of early warnings. Therefore, selective reconstruction based on physically labeled source signals can effectively improve the fidelity of the deformation signal, making it more accurately reflect the physical deformation process of the slope itself.

[0081] In practice, firstly, the source signals with physical labels and the estimated mixing matrix are selectively reprojected with noise components to obtain a projected noise matrix. That is, using these identified noise source signals and their corresponding mixing methods, the contribution of these noises to the original observed signal is accurately calculated. In this process, firstly, all source signals labeled as noise are extracted from the independent source signal matrix to form a noise source signal submatrix; simultaneously, column vectors corresponding to these noise source signals are extracted from the estimated mixing matrix to form a noise mixing matrix submatrix; then, by multiplying the noise mixing matrix submatrix with the noise source signal submatrix, a projected noise matrix is ​​obtained, which represents the displacement contribution at the observation point caused by all source signals identified as noise. This process is expressed by the formula:

[0082] ,

[0083] in, These are the columns of the estimated mixing matrix A corresponding to the noise source signals. These are the source signal lines identified as noise within the physically tagged source signals. It is the projected noise matrix, which accurately quantifies the displacement caused by noise factors at each GNSS measurement point, providing a basis for subsequent noise removal from the overall observation;

[0084] Furthermore, differential reconstruction and clean deformation signal extraction are performed on the projected noise matrix and the observed signal matrix to obtain a high-fidelity deformation signal matrix. That is, after calculating the projected noise matrix, this information is used to subtract the noise from the original observed signal matrix, thus obtaining a clean, high-fidelity deformation signal. In this process, firstly, the projected noise matrix is ​​subtracted point-by-point from the original observed signal matrix; the original observed signal matrix contains all mixed displacement components, including the actual slope deformation and various noises. By subtracting the quantized noise contribution, the remaining part is the clean deformation signal after removing noise interference, i.e., the high-fidelity deformation signal matrix; this process is expressed by the formula:

[0085] ,

[0086] in, It is a high-fidelity deformation signal matrix that contains the actual displacements of the slope at each measuring point, unaffected by external environmental noise (such as temperature effects). It can more accurately reflect the internal stability state of the slope, providing a clean data foundation for subsequent deformation rate and acceleration analysis, thereby significantly improving the accuracy and reliability of slope instability early warning.

[0087] Specifically, in S6, trend analysis is performed on the high-fidelity deformation signal matrix to obtain the deformation rate and acceleration rate. It should be understood that simply obtaining slope displacement data is insufficient to comprehensively assess its stability. While the slope displacement itself may be within an acceptable range, excessively rapid displacement or an accelerating trend may indicate potential instability. The deformation rate (i.e., how quickly displacement changes over time) directly reflects the current activity of slope deformation, while the deformation acceleration rate (i.e., how quickly deformation rate changes over time) is a key indicator for measuring the trend of slope stability changes, especially when the acceleration rate shows a continuous increasing trend, which is often considered an important precursor to slope instability. By performing trend analysis on the high-fidelity deformation signal matrix, these physically meaningful dynamic parameters can be extracted from the clean deformation data that has been carefully processed and free of noise interference. This allows the monitoring system to obtain dynamic rate and acceleration information from static displacement data, providing a more comprehensive description of the slope's motion state.

[0088] In practical implementation, the first and second derivatives of the high-fidelity deformation signal matrix are performed to obtain the deformation rate and acceleration rate. In the technical solution of this application, the first derivative aims to calculate the instantaneous rate of change of the displacement time series of each measuring point in the high-fidelity deformation signal matrix, i.e., the deformation rate. For discrete time series data, numerical differentiation methods are typically used to approximate the derivative calculation. Assume the displacement time series of a certain measuring point j in the high-fidelity deformation signal matrix is... ,in This represents the high-fidelity displacement value of measuring point j at time t, with a time interval of t. The deformation rate is calculated using the central difference method (or other suitable numerical differentiation methods, such as forward or backward difference). The process can be expressed by the following formula:

[0089] ,

[0090] For the start and end points of a time series, forward differencing or backward differencing can be used for processing:

[0091] ,

[0092] ,

[0093] By performing the first-order derivative operation on the time series of each measurement point in the high-fidelity deformation signal matrix, a deformation rate matrix with a size similar to the original matrix can be obtained.

[0094] The second-order derivative aims to calculate the second-order rate of change, or deformation acceleration rate, of the displacement time series at each measurement point in the high-fidelity deformation signal matrix. This can be obtained by taking the first-order derivative of the already calculated deformation rate time series, or directly by performing a second-order numerical differential on the original displacement time series. The central difference method is used to calculate the deformation acceleration rate. The process can be expressed by the following formula:

[0095] ,

[0096] Similarly, for the boundary points of the time series, an adaptive difference formula is required. By performing the above second-order derivative operation on the time series of each measurement point in the high-fidelity deformation signal matrix, a deformation acceleration matrix can be obtained.

[0097] In summary, the real-time monitoring method for slope stability in geotechnical engineering according to the embodiments of this application is explained. It introduces a blind source separation technique combining independent component analysis and physical constraints, treating the original displacement time series and ambient temperature time series of multiple GNSS measuring points as multi-channel mixed signals. Utilizing the statistical independence between signal sources, periodic environmental noise, instrument random noise, drift, and the actual slope deformation signal are effectively separated from the mixed observation signal. Subsequently, through correlation verification with physical quantities such as ambient temperature, the separated source signals are automatically calibrated, accurately identifying and extracting temperature effect source signals and long-term creep source signals with physical labels. This achieves precise removal of noise components and high-fidelity reconstruction of pure deformation signals. Thus, it effectively improves the accuracy and reliability of real-time monitoring of slope stability in geotechnical engineering, providing a solid data foundation and decision-making basis for early warning and prevention of slope disasters.

[0098] Furthermore, a real-time monitoring system for slope stability in geotechnical engineering is also provided.

[0099] Figure 3 This is a block diagram of a real-time monitoring system for slope stability in geotechnical engineering according to an embodiment of this application. Figure 3 As shown, the real-time monitoring system 300 for slope stability in geotechnical engineering according to an embodiment of this application includes: a data acquisition module 310, used to acquire the original displacement time series and the ambient temperature time series of multiple GNSS measuring points; a centering processing and matrix construction module 320, used to center the original displacement time series of multiple GNSS measuring points and construct a matrix to obtain an observation signal matrix; a source signal separation module 330, used to perform source signal separation based on independent component analysis on the observation signal matrix to obtain an estimated independent source signal matrix and an estimated mixed matrix; an automatic source signal calibration module 340, used to automatically calibrate the estimated independent source signal matrix and the ambient temperature time series to obtain source signals with physical labels; a high-fidelity deformation signal reconstruction module 350, used to perform high-fidelity deformation signal reconstruction on the estimated mixed matrix and the observation signal matrix based on the source signals with physical labels to obtain a high-fidelity deformation signal matrix; and a trend analysis module 360, used to perform trend analysis on the high-fidelity deformation signal matrix to obtain the deformation rate and acceleration rate.

[0100] As described above, the real-time monitoring system 300 for geotechnical engineering slope stability according to the embodiments of this application can be implemented in various wireless terminals, such as servers with real-time monitoring algorithms for geotechnical engineering slope stability. In one possible implementation, the real-time monitoring system 300 for geotechnical engineering slope stability according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the real-time monitoring system 300 for geotechnical engineering slope stability can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the real-time monitoring system 300 for geotechnical engineering slope stability can also be one of many hardware modules of the wireless terminal.

[0101] Alternatively, in another example, the real-time monitoring system 300 for geotechnical slope stability and the wireless terminal can also be separate devices, and the real-time monitoring system 300 for geotechnical slope stability can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0102] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A geotechnical engineering method for real-time monitoring of slope stability, characterized in that, The method comprises the following steps: obtaining original displacement time series and field environment temperature time series of a plurality of GNSS observation points; centralizing and matrix constructing the original displacement time series of the plurality of GNSS observation points to obtain an observation signal matrix; source signal separation based on independent component analysis is performed on the observation signal matrix to obtain an estimated independent source signal matrix and an estimated mixing matrix; automatic source signal labeling is performed on the estimated independent source signal matrix and the field environment temperature time series to obtain source signals with physical labels; based on the source signals with physical labels, high-fidelity deformation signal reconstruction is performed on the estimated mixing matrix and the observation signal matrix to obtain a high-fidelity deformation signal matrix; trend analysis is performed on the high-fidelity deformation signal matrix to obtain deformation rate and acceleration rate.

2. The geotechnical engineering slope stability real-time monitoring method according to claim 1, characterized in that, The source signal separation based on independent component analysis is performed on the observation signal matrix to obtain the estimated independent source signal matrix and the estimated mixing matrix, comprising: centralizing and whitening the observation signal matrix to obtain a whitened signal matrix, a whitening matrix and a de-whitening matrix; iterative optimization is performed on the whitened signal matrix to obtain an orthogonal rotation matrix; based on the orthogonal rotation matrix, the whitening matrix and the de-whitening matrix, structure reconstruction is performed on the observation signal matrix to obtain the estimated independent source signal matrix and the estimated mixing matrix.

3. The geotechnical engineering slope stability real-time monitoring method of claim 1, wherein, The automatic source signal labeling is performed on the estimated independent source signal matrix and the field environment temperature time series to obtain the source signals with physical labels, comprising: based on the field environment temperature time series, external physical quantity strongly correlated component labeling is performed on the estimated independent source signal matrix to obtain an index of temperature effect source signals and a remaining source signal index set that is not labeled; internal time sequence characteristic analysis and trend component labeling are performed on the remaining source signal index set that is not labeled and the estimated independent source signal matrix to obtain an index of long-term creep source signals and a final remaining source signal index set.

4. The geotechnical engineering slope stability real-time monitoring method according to claim 3, characterized in that, The external physical quantity strongly correlated component labeling is performed on the estimated independent source signal matrix based on the field environment temperature time series to obtain an index of temperature effect source signals and a remaining source signal index set that is not labeled, comprising: time-domain-frequency-domain signal transformation is performed on the field environment temperature time series and the estimated independent source signal matrix to obtain a Fourier spectrum set of independent source signals and a temperature signal Fourier spectrum; cross power spectral density between the temperature signal Fourier spectrum and each independent source signal Fourier spectrum in the Fourier spectrum set of independent source signals is calculated to obtain a cross power spectral density matrix; based on the cross power spectral density matrix, time lag optimal correlation quantification and labeling are performed on the field environment temperature time series and the estimated independent source signal matrix to obtain the index of temperature effect source signals and the remaining source signal index set that is not labeled.

5. The geotechnical engineering slope stability real-time monitoring method according to claim 4, wherein, The cross power spectral density between the temperature signal Fourier spectrum and each independent source signal Fourier spectrum in the Fourier spectrum set of independent source signals is calculated to obtain a cross power spectral density matrix, comprising: the cross power spectral density between the temperature signal Fourier spectrum and each independent source signal Fourier spectrum in the Fourier spectrum set of independent source signals is calculated according to the following formula: wherein is the cross-power spectral density of the jth source signal and the temperature signal, denotes the cross-power spectral density of the jth source signal and the temperature signal with a time delay of ; represents the inverse Fourier transform operator; and are the standard deviations of the jth source signal and the temperature signal, respectively, and N is the number of points of the signals.

6. The geotechnical engineering slope stability real-time monitoring method of claim 1, wherein, Based on the source signal with physical labels, high-fidelity deformation signal reconstruction is performed on the estimated mixing matrix and the observation signal matrix to obtain a high-fidelity deformation signal matrix, comprising: Noise component selective re-projection is performed on the source signal with physical labels and the estimated mixing matrix to obtain a projection noise matrix; Difference reconstruction and pure deformation signal extraction are performed on the projection noise matrix and the observation signal matrix to obtain a high-fidelity deformation signal matrix.

7. The geotechnical engineering slope stability real-time monitoring method of claim 1, wherein, Trend analysis is performed on the high-fidelity deformation signal matrix to obtain a deformation rate and an acceleration rate, comprising: first-order derivation and second-order derivation are performed on the high-fidelity deformation signal matrix to obtain the deformation rate and the acceleration rate.

8. A geotechnical engineering slope stability real-time monitoring system, characterized in that, Comprise: A data acquisition module is configured to acquire original displacement time series and field environment temperature time series of a plurality of GNSS measuring points; A centralization processing and matrix construction module is configured to perform centralization processing and matrix construction on the original displacement time series of the plurality of GNSS measuring points to obtain an observation signal matrix; A source signal separation module is configured to perform source signal separation on the observation signal matrix based on independent component analysis to obtain an estimated independent source signal matrix and an estimated mixing matrix; A source signal automatic calibration module is configured to perform source signal automatic calibration on the estimated independent source signal matrix and the field environment temperature time series to obtain a source signal with physical labels; A high-fidelity deformation signal reconstruction module is configured to perform high-fidelity deformation signal reconstruction on the estimated mixing matrix and the observation signal matrix based on the source signal with physical labels to obtain a high-fidelity deformation signal matrix; A trend analysis module is configured to perform trend analysis on the high-fidelity deformation signal matrix to obtain a deformation rate and an acceleration rate.

Images