Landslide displacement prediction method fusing multi-scale decomposition and physical constraint

Abstract

The invention belongs to the technical field of geological disaster monitoring, and particularly discloses a landslide displacement prediction method fusing multi-scale decomposition and physical constraint, and the method comprises the steps: decomposing an original landslide displacement sequence, and obtaining a low-frequency IMF, an intermediate-frequency IMF and a high-frequency sub-mode of the original landslide displacement sequence; fusing the low-frequency IMF, the intermediate-frequency IMF and the high-frequency sub-mode of the original landslide displacement sequence and the physical characteristics of the original landslide displacement sequence to obtain fused characteristics; and inputting the fused features into a prediction model to obtain a prediction result output by the prediction model. The method can improve the landslide displacement prediction precision.

Description

Technical Field

[0001] This application belongs to the field of geological disaster monitoring technology, and more specifically, relates to a landslide displacement prediction method that integrates multi-scale decomposition and physical constraints. Background Technology

[0002] Existing landslide displacement prediction methods mainly rely on single-mode decomposition models or purely data-driven deep learning models, such as Empirical Mode Decomposition (EMD) or Ensemble Empirical Mode Decomposition (EEMD) combined with Long Short-Term Memory (LSTM) networks, or Variational Mode Decomposition (VMD) combined with traditional machine learning. These methods suffer from the following problems: Single-mode decomposition models are prone to mode aliasing and can only separate single-mode signal features. When predicting landslide displacement based on single-mode signal features, the prediction accuracy is low. Furthermore, pure data-driven deep learning models ignore the physical mechanism of landslide movement, and the prediction results lack interpretability. They are prone to lag or fitting bias in the stage of sudden displacement induced by heavy rainfall, and it is difficult to take into account the consistency between short-term response and long-term trend, resulting in low prediction accuracy. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this application aims to provide a landslide displacement prediction method that integrates multi-scale decomposition and physical constraints, thereby solving the problem of low landslide displacement prediction accuracy caused by existing single-mode decomposition models or purely data-driven deep learning models.

[0004] To achieve the above objectives, in a first aspect, this application provides a landslide displacement prediction method that integrates multi-scale decomposition and physical constraints, comprising: The original landslide displacement sequence is decomposed to obtain the low-frequency intrinsic mode function (IMF), mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence; The low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence, along with the physical characteristics of the original landslide displacement sequence, are fused to obtain the fused features. The fused features are input into the prediction model to obtain the prediction results output by the prediction model.

[0005] This application decomposes the original landslide displacement sequence into low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes, and fuses them with the physical features of the original landslide displacement sequence to obtain multi-scale fusion features. Based on the multi-scale fusion features, a prediction model is used for prediction, which effectively solves the problems of mode aliasing in single-mode decomposition models and low prediction accuracy caused by landslide displacement prediction based on single-mode signal features. Furthermore, since physical features are incorporated into the multi-scale fusion features, the generalization ability of the model is improved, making the prediction results both accurate and interpretable, thus improving the accuracy of landslide displacement prediction.

[0006] According to the landslide displacement prediction method integrating multi-scale decomposition and physical constraints provided in this application, the decomposition of the original landslide displacement sequence to obtain the low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence includes: The original landslide displacement sequence was decomposed by Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to obtain the low-frequency IMF, mid-frequency IMF and high-frequency IMF of the original landslide displacement sequence. The high-frequency IMFs are combined into high-frequency sub-signals; The high-frequency sub-signal is decomposed using the optimized VMD to obtain the high-frequency sub-mode.

[0007] This application decomposes the original landslide displacement sequence using CEEMDAN, introduces an adaptive noise set to suppress mode aliasing, and obtains multimodal signal features. Then, it decomposes the high-frequency sub-signals using an optimized VMD, avoiding the problem that existing VMD is prone to over-decomposition or under-decomposition, which makes it impossible to accurately extract high-frequency disturbance features, and improving the accuracy of high-frequency disturbance feature extraction.

[0008] According to the landslide displacement prediction method integrating multi-scale decomposition and physical constraints provided in this application, the optimization process of the optimized VMD includes: The parameters of VMD were optimized using the Crested Porcupine Optimizer (CPO) algorithm. and Perform adaptive optimization to obtain the optimal parameters. and ; Optimal parameters and Substitute the values ​​into VMD to obtain the optimized VMD.

[0009] This application uses CPO to configure VMD parameters. and Perform adaptive optimization to avoid the parameters of existing VMD and This addresses the problem of over- or under-decomposition due to reliance on human experience in setting parameters, which can lead to inaccurate extraction of high-frequency disturbance features. It aims to improve the accuracy of high-frequency disturbance feature extraction.

[0010] According to the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints provided in this application, the physical characteristics of the original landslide displacement sequence include velocity characteristics, acceleration characteristics, and improved tangential angle characteristics.

[0011] This application uses the velocity and acceleration features of the original landslide displacement sequence, along with the improved tangential angle feature that more accurately reflects the evolution of the landslide from the isokinetic to the acceleration stage, as physical features. These physical features are then fused with other features to obtain multi-scale fusion features. Based on these multi-scale fusion features, a prediction model is used for prediction, effectively solving the problems of mode aliasing in single-mode decomposition models and the low prediction accuracy caused by landslide displacement prediction based on single-mode signal features. Furthermore, since physical features are incorporated into the multi-scale fusion features, the generalization ability of the model is improved, making the prediction results both accurate and interpretable.

[0012] According to the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints provided in this application, the prediction model is a cascaded hybrid structure of LSTM and Transformer.

[0013] According to the landslide displacement prediction method integrating multi-scale decomposition and physical constraints provided in this application, before decomposing the original landslide displacement sequence, the method further includes: The original landslide displacement sequence is preprocessed.

[0014] Secondly, this application provides a landslide displacement prediction device that integrates multi-scale decomposition and physical constraints, comprising: The decomposition module is used to decompose the original landslide displacement sequence to obtain the low-frequency IMF, mid-frequency IMF and high-frequency sub-modes of the original landslide displacement sequence. The fusion module is used to fuse the low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence, as well as the physical characteristics of the original landslide displacement sequence, to obtain the fused features. The prediction module is used to input the fused features into the prediction model and obtain the prediction results output by the prediction model.

[0015] Thirdly, this application provides an electronic device, comprising: at least one memory for storing a program; and at least one processor for executing the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the landslide displacement prediction method integrating multi-scale decomposition and physical constraints described in the first aspect or any possible implementation of the first aspect.

[0016] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the landslide displacement prediction method integrating multi-scale decomposition and physical constraints described in the first aspect or any possible implementation of the first aspect.

[0017] Fifthly, this application provides a computer program product that, when run on a processor, causes the processor to execute the landslide displacement prediction method integrating multi-scale decomposition and physical constraints described in the first aspect or any possible implementation of the first aspect.

[0018] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0019] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art: This application decomposes the original landslide displacement sequence into low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes, and fuses them with the physical features of the original landslide displacement sequence to obtain multi-scale fusion features. Based on the multi-scale fusion features, a prediction model is used for prediction, which effectively solves the problems of mode aliasing in single-mode decomposition models and low prediction accuracy caused by landslide displacement prediction based on single-mode signal features. Furthermore, since physical features are incorporated into the multi-scale fusion features, the generalization ability of the model is improved, making the prediction results both accurate and interpretable, thus improving the accuracy of landslide displacement prediction. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints provided in the embodiments of this application. Figure 2 This is a schematic diagram of landslide displacement prediction provided in an embodiment of this application; Figure 3 This is a flowchart of the LSTM-Transformer algorithm provided in the embodiments of this application; Figure 4 This is a landslide structure and monitoring point distribution map provided in the embodiments of this application; Figure 5 This is a schematic diagram of the combined displacement and rainfall data of GNSS observation points G1-G5 provided in the embodiments of this application; Figure 6 This is a convergence curve of CPO-optimized VMD parameters provided in the embodiments of this application; Figure 7 This is a schematic diagram comparing the spectra of the original signal and the secondary decomposed signal provided in the embodiments of this application; Figure 8 This is a schematic diagram of the final decomposition result provided in the embodiments of this application; Figure 9 This is a schematic diagram of landslide displacement prediction results provided in an embodiment of this application; Figure 10 This is a schematic diagram comparing prediction results with and without rainfall feature input provided in an embodiment of this application; Figure 11 This is a schematic diagram of the landslide displacement prediction device that integrates multi-scale decomposition and physical constraints provided in the embodiments of this application; Figure 12 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0023] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0024] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0025] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0026] Next, combined Figures 1-10 This paper introduces a landslide displacement prediction method that integrates multi-scale decomposition and physical constraints, as provided in the embodiments of this application.

[0027] Figure 1 This is a flowchart illustrating the landslide displacement prediction method integrating multi-scale decomposition and physical constraints provided in the embodiments of this application. Figure 1 As shown, the method includes the following steps: Step S1: Decompose the original landslide displacement sequence to obtain the low-frequency IMF, mid-frequency IMF and high-frequency sub-mode of the original landslide displacement sequence; Step S2: The low-frequency IMF, mid-frequency IMF, and high-frequency sub-mode of the original landslide displacement sequence, as well as the physical characteristics of the original landslide displacement sequence, are fused to obtain the fused characteristics. Step S3: Input the fused features into the prediction model to obtain the prediction results output by the prediction model.

[0028] The landslide displacement prediction method provided in this application, which integrates multi-scale decomposition and physical constraints, decomposes the original landslide displacement sequence into low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes, and fuses them with the physical features of the original landslide displacement sequence to obtain multi-scale fusion features. Based on the multi-scale fusion features, a prediction model is used for prediction, which effectively solves the problems of mode aliasing in single-mode decomposition models and low prediction accuracy caused by landslide displacement prediction based on single-mode signal features. Furthermore, since physical features are incorporated into the multi-scale fusion features, the generalization ability of the model is improved, making the prediction results both accurate and interpretable.

[0029] In some embodiments, step S1 specifically includes: Step S11: Decompose the original landslide displacement sequence using CEEMDAN to obtain the low-frequency IMF, mid-frequency IMF and high-frequency IMF of the original landslide displacement sequence; Step S12: Combine the high-frequency IMFs into high-frequency sub-signals; Step S13: Decompose the high-frequency sub-signal using the optimized VMD to obtain the high-frequency sub-mode.

[0030] Figure 2 This is a schematic diagram of landslide displacement prediction provided in an embodiment of this application, as shown below. Figure 2As shown, in one embodiment of this application, CEEMDAN is used to decompose the original landslide displacement sequence once, and an adaptive noise set is introduced to suppress mode mixing. The specific process is as follows: 1a, Let the original landslide displacement signal be... It includes a trend term, periodic disturbances, and random noise, and can be expressed as:

[0031] in, This represents the true displacement trend. It is noise.

[0032] 2a, displacement signal towards the original landslide Injecting N times Gaussian white noise of different intensities This forms a sample set:

[0033] in, These are the weighting coefficients for Gaussian white noise. For the first Gaussian white noise during the second processing.

[0034] 3a, for all the above samples Perform EMD decomposition to obtain the first IMF component. And take the average of them as the first IMF component of CEEMDAN. .Right now:

[0035]

[0036] in, The first IMF component generated by the CEEMDAN decomposition; For signal The first EMD decomposition obtained One IMF component; This is the residual term obtained after the first CEEMDAN decomposition.

[0037] 4a, in the residual term By adding different noise components, a new set of samples to be decomposed can be formed. Continue with EMD decomposition to obtain the second CEEMDAN mode component. and residuals Repeat the above steps until the residual term cannot be decomposed by EMD. The final signal is decomposed into:

[0038] In the formula, The component generated by the k-th CEEMADN decomposition. These are the residual terms that cannot be decomposed by EMD.

[0039] CEEMDAN can significantly reduce mode aliasing while preserving the intrinsic structure of the signal, and the decomposition results are more stable. After decomposition, the low-frequency IMF and residual terms can be regarded as the long-term deformation trend of the landslide, while the high-frequency IMF contains short-term changes such as rainfall disturbance, microseismic response and monitoring error.

[0040] The high-frequency IMFs obtained from CEEMDAN decomposition are combined into sub-signals. High-frequency sub-signals are processed through optimized VMD. Perform secondary decomposition to obtain high-frequency sub-modes It accurately captures the characteristics of short-term displacement disturbances induced by heavy rainfall.

[0041] VMD (Vibration Mode Decomposition) achieves adaptive frequency band division of a signal by minimizing the bandwidth of each mode in the frequency domain. Its basic idea is to decompose the signal into... A mode with finite bandwidth This ensures that each mode is at the center frequency The energy is most concentrated nearby. Assuming the signal is to be transmitted... Decomposed into K IMF components The specific process is as follows: 1b, each IMF component Perform a Hilbert transform to obtain its analytic signal:

[0042] in, Let be the Dirac impulse function. The imaginary unit, It is the impulse response of the Hilbert transform.

[0043] 2b, Predict the center frequency of each component of the analytical signal. index term The components are mixed, and the spectrum of each component is modulated to the corresponding baseband:

[0044] 3b, Calculate the square of the gradient of the demodulated signal above. The norm is used to estimate the bandwidth of each modal component. The optimization objective of VMD can be expressed as:

[0045] In the formula, The result of decomposition One IMF component; Indicates the center frequency of each component; For time partial derivative operators (for the signal in parentheses with respect to time) t (Differentiation).

[0046] In some embodiments, the optimization process of the optimized VMD includes: The parameters of VMD were optimized using the choke pig optimization algorithm. and Perform adaptive optimization to obtain the optimal parameters. and ; Optimal parameters and Substitute the values ​​into VMD to obtain the optimized VMD.

[0047] like Figure 2 As shown, in one embodiment of this application, the parameters of VMD are processed using the CPO algorithm. and To achieve adaptive optimization, the CPO algorithm uses spike defense to perform local search and cluster foraging to perform global exploration, thus balancing the diversity of the solution space and the convergence speed.

[0048] Let the objective function be the VMD decomposition quality index. Its definition is as follows:

[0049] In the formula, It is a modal orthogonality index; This represents the reconstruction error; Sample entropy is used to measure signal complexity. , and These are the weighting coefficients.

[0050] The CPO algorithm randomly initializes a set of candidate individuals, each representing a set of parameter pairs. , By continuously updating the position to minimize The rules for updating an individual's location can be summarized as follows:

[0051] In the formula, and For learning factors; and These are random coefficients; Indicates the first t The global optimal solution at the next iteration. For the first t During the nth iteration iThe position of each candidate individual; For the first t During the nth iteration j The position of each candidate individual ( j Is with i Different individuals are used for local search. The optimal parameters are obtained after several iterations. and .

[0052] In some embodiments, the physical characteristics of the original landslide displacement sequence include velocity characteristics, acceleration characteristics, and improved tangential angle characteristics.

[0053] Based on the preprocessed displacement sequence Calculate three types of physical dynamic characteristic quantities: 1c, speed It is obtained through the first-order time difference of the displacement sequence, reflecting the instantaneous trend of displacement change:

[0054] 2c, acceleration The dynamic response during the landslide acceleration phase is characterized by obtaining the second-order time difference of the displacement sequence.

[0055] 3c. To more accurately reflect the evolution characteristics of landslides from the isokinetic to the accelerating stage, an improved tangential angle is introduced. Definition. First, the displacement-time curve ( Converted to constant velocity stage displacement rate Dimensionally normalized coordinates based on:

[0056] In the formula, For monitoring cycle Cumulative displacement within, The displacement rate during the constant velocity phase.

[0057] After this transformation, we obtain ( (Curve, Improved Tangential Angle) Defined as the curve at the monitoring time The slope angle of the tangent at:

[0058] like Figure 2 As shown, in one embodiment of this application, the low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the optimized VMD decomposition from CEEMD are compared with the velocity... acceleration Improved tangential angle Three types of physical features are time-aligned and standardized to construct a comprehensive input matrix:

[0059] This feature matrix retains the multi-scale statistical features after decomposition and embeds the physical constraint information of landslide movement, providing an interpretable feature space for depth prediction.

[0060] In some embodiments, the prediction model is a cascaded hybrid structure of LSTM and Transformer.

[0061] like Figure 2 As shown, in one embodiment of this application, the prediction model adopts a cascaded hybrid structure of LSTM and Transformer: the front-end LSTM is responsible for capturing the short-term and long-term dependencies of the time series, and its output is input into the Transformer encoder after linear mapping, so as to mine the global correlation between different time scales with a multi-head self-attention mechanism.

[0062] Figure 3 This is a flowchart of the LSTM-Transformer algorithm provided in the embodiments of this application, as follows: Figure 3 As shown, let the input be... The model outputs a prediction of future displacement. The loss function is defined as the weighted mean square error with physical constraints:

[0063] In the formula, the first term measures the prediction error; the second term... This is a physical consistency term used to penalize situations where the predicted velocity and acceleration directions do not match the actual physical evolution. This is the weighting factor.

[0064] In one embodiment of this application, the proposed model framework mainly consists of three core modules: a multi-scale signal analysis module: addressing the non-stationary and nonlinear nature of landslide displacement time series, it employs Complete Ensemble Empirical Mode Decomposition (CEEMDAN) as a first-level coarse filter to remove long-term gravity creep trends and high-frequency dynamic components, and then introduces parameter adaptive optimization variational mode decomposition (CPO-VMD) as a second-level fine filter to refine the high-frequency rainfall response signal, solving the mode mixing and over / under-decomposition problems in traditional decomposition methods; and a physical feature enhancement module: based on the kinematic principles of landslides, it constructs a model including velocity, acceleration, and improved tangential... The multidimensional physical feature space of the angle, especially the introduction of the improved tangential angle, establishes a quantitative mapping relationship between displacement rate and landslide evolution stages (creep, uniform, acceleration), providing the model with clear physical criteria for landslide hazard. The physical constraint deep learning module designs an LSTM-Transformer cascaded hybrid network structure, taking into account both the local dependency characteristics of time series data and the global correlation mechanism across long periods. More importantly, a physical consistency constraint term is introduced into the model loss function, forcing the parameter update of the neural network to not only follow the statistical law of data distribution, but also meet the physical directionality constraint of kinematic evolution, thereby achieving a dual improvement in accuracy and interpretability.

[0065] In some embodiments, prior to step S1, the method further includes: The original landslide displacement sequence was preprocessed.

[0066] Optionally, outlier removal, cubic spline interpolation filling for short-term missing measurement points, and zero-mean normalization can be performed on the original landslide displacement sequence obtained from monitoring to eliminate the influence of dimensions and data noise interference. By performing cross-correlation analysis on the rainfall and displacement increment sequences, the lag time of rainfall on displacement can be determined, achieving time alignment of the two types of signals.

[0067] Figure 4 This is a landslide structure and monitoring point distribution map provided in the embodiments of this application, such as... Figure 4 As shown in one embodiment of this application, a landslide in a certain area is used as a research example. This landslide is a medium-sized precipitation-induced landslide with a main sliding direction of 207°, a length of 340m, a width of 45-81m, an average thickness of 12.0m, and a landslide area of ​​2.26×10⁻⁶ m². 4 m², volume 20.34 × 10 4 m³. The landslide body is composed of fragmented rock and soil, with the slip zone consisting of gravelly soil interbedded with silty clay (0.2m thick), and underlying sericite quartz schist. Under long-term rainfall infiltration, it exhibits a "creep-acceleration-restabilization" deformation characteristic. Five GNSS displacement monitoring points (G1-G5) are set up on the main profile of the landslide body, with a spacing of 50-80m between the points. The monitoring cycle is 1 day, and the observation period is 30 months. An automatic rain gauge is set up in the leading platform area to record the daily rainfall.

[0068] Figure 5 This is a schematic diagram of the combined displacement and rainfall data of GNSS observation points G1-G5 provided in this application embodiment. The displacement and rainfall data of monitoring points G1-G5 are as follows: Figure 5 As shown.

[0069] During modeling, the data from G2 to G5 and the first 70% of the data from G1 were used as the training set, and the last 30% of the data from G1 were used as the test set. Samples were constructed using a 3-day sliding window.

[0070] After one decomposition by CEEMDAN, the G1 monitoring point sequence was decomposed into 9 IMF components and 1 trend term. The high-frequency component accounted for 12%-15% of the energy, while the mid-to-low frequency component accounted for over 80%, effectively separating deformation characteristics at different scales.

[0071] In the process of optimizing VMD parameters by CPO, IMF1 was used as the input signal, with 30 individuals and 30 iterations. The fitness value tended to stabilize after the 25th iteration, and the optimal parameters K=2 and α=939.1 were finally obtained.

[0072] Figure 6 This is a convergence curve diagram of the CPO algorithm for optimizing VMD parameters provided in the embodiments of this application. The fitness convergence curve of the CPO algorithm in the VMD parameter optimization process is shown below. Figure 6 As shown.

[0073] Figure 7 This is a schematic diagram comparing the spectra of the original signal and the secondary decomposed signal provided in this application embodiment. After secondary decomposition, two sub-modes, VMD1 (high-frequency short-period perturbation) and VMD2 (post-rainfall hysteresis response), are obtained. The spectrum separation is good, and the reconstruction error RMSE = 0.8409 mm. The spectrum comparison between the original signal and the secondary decomposed signal is as follows: Figure 7 As shown.

[0074] Figure 8 This is a schematic diagram of the final decomposition result provided in the embodiments of this application, such as... Figure 8 As shown, the high-frequency part consists of VMD1 and VMD2, which are outputs of the optimized VMD, while the mid-frequency and low-frequency parts still retain the IMF4–IMF9 modes of CEEMDAN.

[0075] Figure 9 This is a schematic diagram of the landslide displacement prediction results provided in this application embodiment. The velocity, acceleration, and improved tangential angle features of the G1 monitoring point are extracted and fused with multi-scale decomposition components to construct the input matrix. LSTM-Transformer model training parameters: learning rate 0.001, batch size 64, maximum iterations 100, Dropout layer to suppress overfitting. The prediction results are as follows... Figure 9 As shown.

[0076] Test set validation results show that the model performs excellently in all-time forecasting: RMSE=2.975mm, MAE=2.026mm, R²=0.9849. During the heavy rainfall period in September 2021, the predicted curve almost perfectly matched the measured curve, with a peak time difference of less than 1 day and an amplitude error of less than 5%. The long-term trend forecast showed no significant drift and maintained good consistency.

[0077] Figure 10 This is a schematic diagram comparing prediction results with and without rainfall feature input provided in the embodiments of this application. After introducing rainfall features, the model's RMSE decreases by 47.1% and MAE decreases by 49.7%, significantly improving rainfall response sensitivity. The comparison results are as follows. Figure 10 As shown.

[0078] Comparative experiments show that after introducing rainfall features, the model's RMSE decreased by 47.1% and MAE decreased by 49.7%, significantly improving the sensitivity to rainfall response. Compared with the single CEEMDAN decomposition model and the model without decomposition, the secondary decomposition strategy reduced RMSE by 38% and 53%, respectively. After fusing physical features, the model's RMSE in the mutation phase decreased from 8.164 mm to 2.975 mm, and R² increased from 0.879 to 0.9849. Compared with models such as LSTM, Transformer, GRU, and CNN, the LSTM-Transformer structure proposed in this application has the best overall performance in both short-term mutation capture and long-term trend fitting.

[0079] This embodiment fully verifies the effectiveness and superiority of the method in predicting displacement of complex rainfall-induced landslides. Those skilled in the art can adapt the method to landslide monitoring data in different regions by adjusting the decomposition parameters and model training hyperparameters.

[0080] The landslide displacement prediction device integrating multi-scale decomposition and physical constraints provided in this application is described below. The landslide displacement prediction device integrating multi-scale decomposition and physical constraints described below can be referred to in correspondence with the landslide displacement prediction method integrating multi-scale decomposition and physical constraints described above.

[0081] Figure 11 This is a schematic diagram of a landslide displacement prediction device integrating multi-scale decomposition and physical constraints provided in an embodiment of this application. Figure 11 As shown, the device 1100 includes: The decomposition module 1110 is used to decompose the original landslide displacement sequence to obtain the low-frequency IMF, mid-frequency IMF and high-frequency sub-mode of the original landslide displacement sequence. The fusion module 1120 is used to fuse the low-frequency IMF, mid-frequency IMF and high-frequency sub-mode of the original landslide displacement sequence, as well as the physical characteristics of the original landslide displacement sequence, to obtain the fused characteristics. The prediction module 1130 is used to input the fused features into the prediction model and obtain the prediction results output by the prediction model.

[0082] It should be understood that the above-described device is used to execute the methods in the above embodiments. The implementation principle and technical effect of the corresponding program modules in the device are similar to those described in the above methods. The working process of the device can be referred to the corresponding process in the above methods, and will not be repeated here.

[0083] Based on the methods in the above embodiments, Figure 12 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 12 As shown in the illustration, this application provides an electronic device that may include a processor 1210, a communications interface 1220, a memory 1230, and a communication bus 1240. The processor 1210, communications interface 1220, and memory 1230 communicate with each other via the communication bus 1240. The processor 1210 can call logical instructions from the memory 1230 to execute the landslide displacement prediction method integrating multi-scale decomposition and physical constraints described in the above embodiment.

[0084] Furthermore, the logical instructions in the aforementioned memory 1230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the landslide displacement prediction method integrating multi-scale decomposition and physical constraints described in the various embodiments of this application.

[0085] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program. When the computer program runs on a processor, it causes the processor to execute the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints as described in the above embodiments.

[0086] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints as described in the above embodiments.

[0087] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0088] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0089] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0090] It is understood that the various numerical designations used in the embodiments of this application are merely for the convenience of description and are not intended to limit the scope of the embodiments of this application.

[0091] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A landslide displacement prediction method integrating multi-scale decomposition and physical constraints, characterized in that, include: The original landslide displacement sequence is decomposed to obtain the low-frequency IMF, mid-frequency IMF and high-frequency sub-modes of the original landslide displacement sequence; The low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence, along with the physical characteristics of the original landslide displacement sequence, are fused to obtain the fused features. The fused features are input into the prediction model to obtain the prediction results output by the prediction model.

2. The landslide displacement prediction method integrating multi-scale decomposition and physical constraints according to claim 1, characterized in that, The process of decomposing the original landslide displacement sequence to obtain the low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence includes: The original landslide displacement sequence was decomposed using CEEMDAN to obtain the low-frequency IMF, mid-frequency IMF, and high-frequency IMF of the original landslide displacement sequence. The high-frequency IMFs are combined into high-frequency sub-signals; The high-frequency sub-signal is decomposed using the optimized VMD to obtain the high-frequency sub-mode.

3. The landslide displacement prediction method integrating multi-scale decomposition and physical constraints according to claim 2, characterized in that, The optimization process of the optimized VMD includes: The parameters of VMD were optimized using the choke pig optimization algorithm. and Perform adaptive optimization to obtain the optimal parameters. and ; Optimal parameters and Substitute the values ​​into VMD to obtain the optimized VMD.

4. The landslide displacement prediction method integrating multi-scale decomposition and physical constraints according to claim 1, characterized in that, The physical characteristics of the original landslide displacement sequence include velocity characteristics, acceleration characteristics, and improved tangential angle characteristics.

5. The landslide displacement prediction method integrating multi-scale decomposition and physical constraints according to claim 1, characterized in that, The prediction model is a cascaded hybrid structure of LSTM and Transformer.

6. The landslide displacement prediction method integrating multi-scale decomposition and physical constraints according to claim 1, characterized in that, Before decomposing the original landslide displacement sequence, the method further includes: The original landslide displacement sequence is preprocessed.

7. A landslide displacement prediction device integrating multi-scale decomposition and physical constraints, characterized in that, include: The decomposition module is used to decompose the original landslide displacement sequence to obtain the low-frequency IMF, mid-frequency IMF and high-frequency sub-modes of the original landslide displacement sequence. The fusion module is used to fuse the low-frequency IMF, mid-frequency IMF, and high-frequency sub-modes of the original landslide displacement sequence, as well as the physical characteristics of the original landslide displacement sequence, to obtain the fused features. The prediction module is used to input the fused features into the prediction model and obtain the prediction results output by the prediction model.

8. An electronic device, characterized in that, include: At least one memory for storing computer programs; At least one processor is configured to execute a program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to execute the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is run on a processor, the processor executes the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints as described in any one of claims 1-6.

10. A computer program product, characterized in that, When the computer program product is run on a processor, the processor executes the landslide displacement prediction method that integrates multi-scale decomposition and physical constraints as described in any one of claims 1-6.

Images