A landslide hidden point displacement monitoring system using multi-source data fusion

Abstract

The application discloses a landslide hidden point displacement monitoring system using multi-source data fusion, comprising: a multi-source sensing unit, the multi-source sensing unit comprising: a spaceborne radar sensing module, which is used for acquiring large-range ground surface deformation data through synthetic aperture radar; a ground surface deformation sensing module, which is used for monitoring ground surface displacement in real time through a dual-frequency GNSS receiver and a fiber bragg grating sensor; an underground stress sensing module, which is used for acquiring underground rock and soil body stress changes through a microseismic monitoring array and a distributed optical fiber sensor; an edge computing unit, the edge computing unit comprising: a dynamic weight fusion module, which is used for dynamically distributing multi-source sensor weights according to a geological state, and generating high-precision fused displacement field data. The application can break through the bottleneck of traditional monitoring systems in terms of data quality, model accuracy and response efficiency through multi-source data fusion, real-time model parameter updating and closed-loop optimization mechanism, and provide data support for precise early warning and intelligent emergency of landslide disasters.

Description

Technical Field

[0001] The present invention relates to the field of monitoring systems, and in particular to a landslide potential point displacement monitoring system using multi-source data fusion. Background Art

[0002] Landslides are a common geological disaster that poses a serious threat to people's lives, property, and infrastructure. Displacement monitoring of potential landslide sites is a key component of geological disaster monitoring and early warning. Its core is to use technical means to track the deformation dynamics of potential landslide areas in real time, providing a scientific basis for disaster prediction, risk assessment, and emergency response.

[0003] Existing monitoring systems attempt to improve monitoring capabilities through multi-sensor integration, but they still suffer from problems such as data fusion rigidity, fixed model parameters, single early warning mechanism, and inefficient emergency decision-making. Therefore, a landslide hazard point displacement monitoring system using multi-source data fusion is proposed. Summary of the Invention

[0004] The technical problem to be solved by the present invention is: how to solve the problems that the existing monitoring system attempts to improve the monitoring capability through multi-sensor integration, but still has problems such as data fusion rigidity, model parameter fixation, single early warning mechanism and inefficient emergency decision-making. A landslide hazard point displacement monitoring system using multi-source data fusion is provided.

[0005] The present invention solves the above-mentioned technical problems through the following technical solutions, which include:

[0006] Multi-source sensing unit, the multi-source sensing unit includes:

[0007] Spaceborne radar sensing module, used to obtain large-scale surface deformation data through synthetic aperture radar (InSAR);

[0008] A surface deformation sensing module, which is used to monitor surface displacement in real time using a dual-frequency GNSS receiver and a fiber Bragg grating sensor;

[0009] The underground stress sensing module is used to obtain the stress changes of underground rock and soil through microseismic monitoring arrays and distributed optical fiber sensors;

[0010] Edge computing unit, the edge computing unit includes:

[0011] Dynamic weight fusion module, used to dynamically assign multi-source sensor weights according to geological conditions to generate high-precision fused displacement field data;

[0012] Communication protocol adapter module, used to achieve protocol conversion and low-latency transmission of multi-source sensor data;

[0013] Digital twin modeling unit, the digital twin modeling unit includes:

[0014] Geomechanical parameterized model, used to construct a three-dimensional mechanical model of the landslide based on the finite element method;

[0015] Real-time data assimilation module, used to input fused data into the model for inversion and update of geomechanical parameters;

[0016] Model self-correction module, used to optimize model accuracy through parameter perturbation testing;

[0017] Intelligent early warning unit, the intelligent early warning unit includes:

[0018] Multi-level warning trigger module, used to trigger graded warnings based on displacement rate, acceleration and underground parameter anomalies;

[0019] The three-dimensional path planning module is used to dynamically plan safe evacuation paths based on digital twin prediction results.

[0020] Furthermore, the dynamic weight fusion module performs the following operations:

[0021] Acquire real-time measurement data of each sensor module and its confidence parameter, wherein the confidence parameter is determined jointly based on sensor noise variance and historical error statistics;

[0022] Calculation of geological state assessment coefficients , whose expression is:

[0023] ;

[0024] Where, is the inclination change influence function, is the real-time tilt angle change, is the critical dip threshold (related to geological type), is the weight coefficient of the inclination change, is the displacement rate influence function, is the real-time displacement rate, Critical rate threshold, is the weight coefficient of displacement rate

[0025] according to The fusion weight of each sensor module is dynamically allocated based on the value. The weight calculation formula is:

[0026] ;

[0027] Where, is the fusion weight of the i-th sensor module, is the geological status assessment coefficient, is the confidence parameter of the i-th sensor module, is the total number of sensor modules.

[0028] Furthermore, the tilt angle change affects the function and displacement rate influence function Defined as:

[0029] ;

[0030] ;

[0031] in, To prevent zero and tiny amounts, its value is positively correlated with the measurement accuracy of the sensor.

[0032] Furthermore, the weight coefficient and Dynamic configuration through geological knowledge base includes the following steps:

[0033] Step 1: Knowledge base construction:

[0034] Based on geotechnical theory and historical landslide data, the correlation between different geological types and weight coefficients is established;

[0035] The geological types include at least typical landslide body types, which include clay, sand, rock and accumulation layers;

[0036] Weight coefficient 、 The allocation principle is:

[0037] Dip geology (e.g. clay): gives higher weight to dip changes ( > );

[0038] Rate-sensitive geology (such as rock): Give higher weight to displacement rate ( > );

[0039] Step 2: Identification of on-site geological types and multi-source data collection:

[0040] Obtaining geophysical parameters (such as moisture content, particle composition, and shear strength) through on-site drilling sampling, geological radar detection, and drone multispectral imaging;

[0041] Input the collected data into a pre-trained geological classification model (such as a random forest or convolutional neural network) to output the geological type label of the target area;

[0042] Step 3: Dynamically load the weight coefficient and search the corresponding geological type label in the knowledge base according to the geological type label output in step 2. 、 value;

[0043] Inject the matching weight coefficient into the dynamic weight fusion module in real time to update the geological state assessment coefficient The calculation logic of

[0044] Step 4: Exception handling mechanism: if the geological type of the site is not included in the mapping table, the default weight is used ( =0.5, =0.5) and trigger the manual verification process;

[0045] The rationality of the weight configuration is then verified by replaying historical data. If the deviation between the fusion results and the measured data exceeds the limit for N consecutive times, the knowledge base update is triggered.

[0046] Furthermore, the real-time data assimilation module performs:

[0047] Step 1: Receive the fused displacement field data sent by the edge computing unit ;

[0048] Step 2: Input geomechanical parameterized model for inversion calculation and update the sliding zone parameter set ,in is the shear modulus, is the pore pressure coefficient, for cohesion;

[0049] Step 3: Calculate the parameter change rate ,when The model self-correction module is triggered when is the preset sensitivity threshold;

[0050] is the rate of change of the sliding band parameter set, is the updated sliding belt parameter set (including shear modulus , pore pressure coefficient , cohesion ), is the sliding belt parameter set before updating, is the preset sensitivity threshold.

[0051] Furthermore, the inversion calculation adopts a regularized adjoint gradient method, which specifically includes:

[0052] Construct the objective function:

[0053]

[0054] in, is the finite element forward operator, is the initial parameter, is the regularization coefficient;

[0055] By solving the adjoint equation:

[0056] ;

[0057] Get the parameter gradient ▽J, where is the stiffness matrix, is the accompanying variable;

[0058] The quasi-Newton method is used to iteratively update the parameters until convergence.

[0059] Furthermore, the model self-correction module performs the following operations:

[0060] Save a snapshot of the current parameters to the historical database;

[0061] Generate parameter perturbation set , =1, 2, 3...m;

[0062] Parallel calculation of the predicted displacement field corresponding to each disturbance parameter ;

[0063] Select the parameter group with the highest matching degree with the actual observation data in the subsequent Δt period as the update benchmark:

[0064] ;

[0065] Where, is the predicted displacement field of the kth group of disturbance parameters at time t;

[0066] is the actual observed displacement data at time t;

[0067] T is the length of the verification time window.

[0068] Furthermore, the dynamic weight fusion module and the real-time data assimilation module form a bidirectional coupling mechanism, which is specifically manifested as follows:

[0069] When the model self-correction module updates the sliding belt parameter set After that, the geological state assessment coefficient is automatically triggered The critical inclination threshold is and critical rate threshold According to the new Dynamic value adjustment:

[0070] ;

[0071] ;

[0072] in, 、 is an empirical function based on geotechnical mechanics, is the rock mass density.

[0073] Furthermore, the three-dimensional path planning module performs:

[0074] Receive the landslide impact boundary output by the digital twin modeling unit and then construct a three-dimensional terrain cost function:

[0075] ;

[0076] in, is the slope factor, is the surface crack density, is the distance between the current position and the landslide boundary;

[0077] The improved A* algorithm is used to search for the minimum cost path from the personnel location to the safe zone, and the path weight coefficient 、 and Adjust dynamically according to the warning level.

[0078] Compared with the existing technology, the present invention has the following advantages: the landslide potential point displacement monitoring system using multi-source data fusion, multi-source data collaborative monitoring, and comprehensive improvement of monitoring accuracy

[0079] Through multi-dimensional data collection of satellite-borne radar, surface deformation sensors and underground stress sensors, a comprehensive perception of surface deformation and underground stress changes at landslide risk points is achieved, eliminating the limitations of a single data source. Dynamic weight fusion technology adaptively adjusts sensor weights according to geological conditions, effectively suppresses noise interference, generates high-confidence fusion data, and provides reliable input for subsequent analysis. Based on the geological knowledge base, the weight coefficient is dynamically configured, enabling the system to automatically adapt to different geological types and improve monitoring sensitivity and pertinence. The digital twin model updates geomechanical parameters through real-time data inversion to ensure that the model evolves synchronously with actual geological conditions and avoids prediction deviations caused by parameter solidification. The model self-correction module quickly optimizes the model through parameter perturbation testing and parallel calculation. The model accuracy is improved, and the system's adaptability to complex geological environments is enhanced. The multi-level early warning trigger mechanism combines multi-dimensional indicators such as displacement rate, acceleration and underground parameter anomalies to achieve graded early warning, significantly reducing the risk of false alarms and missed alarms. The three-dimensional path planning module dynamically generates safe evacuation paths based on the landslide impact range predicted by the digital twin, comprehensively considering the slope, crack density and real-time danger distance to improve emergency evacuation efficiency and safety. The improved path search algorithm ensures that the planned path is updated in real time to avoid sudden dangerous areas. The two-way coupling mechanism of dynamic weight fusion and model parameter update forms a closed-loop optimization of data and model, continuously improving monitoring and prediction accuracy. The abnormality handling mechanism ensures the robustness of the system under unknown geological conditions and reduces the need for manual intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0081] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0082] like Figure 1 As shown, this embodiment provides a technical solution: a landslide potential point displacement monitoring system using multi-source data fusion, comprising:

[0083] Multi-source sensing unit, the multi-source sensing unit includes:

[0084] Spaceborne radar sensing module, used to obtain large-scale surface deformation data through synthetic aperture radar (InSAR);

[0085] A surface deformation sensing module, which is used to monitor surface displacement in real time using a dual-frequency GNSS receiver and a fiber Bragg grating sensor;

[0086] The underground stress sensing module is used to obtain the stress changes of underground rock and soil through microseismic monitoring arrays and distributed optical fiber sensors;

[0087] Edge computing unit, the edge computing unit includes:

[0088] Dynamic weight fusion module, used to dynamically assign multi-source sensor weights according to geological conditions to generate high-precision fused displacement field data;

[0089] Communication protocol adapter module, used to achieve protocol conversion and low-latency transmission of multi-source sensor data;

[0090] Digital twin modeling unit, the digital twin modeling unit includes:

[0091] Geomechanical parameterized model, used to construct a three-dimensional mechanical model of the landslide based on the finite element method;

[0092] Real-time data assimilation module, used to input fused data into the model for inversion and update of geomechanical parameters;

[0093] Model self-correction module, used to optimize model accuracy through parameter perturbation testing;

[0094] Intelligent early warning unit, the intelligent early warning unit includes:

[0095] Multi-level warning trigger module, used to trigger graded warnings based on displacement rate, acceleration and underground parameter anomalies;

[0096] The three-dimensional path planning module is used to dynamically plan safe evacuation paths based on digital twin prediction results.

[0097] Furthermore, the dynamic weight fusion module performs the following operations:

[0098] Acquire real-time measurement data of each sensor module and its confidence parameter, wherein the confidence parameter is determined jointly based on sensor noise variance and historical error statistics;

[0099] Calculation of geological state assessment coefficients , whose expression is:

[0100] ;

[0101] Where, is the inclination change influence function, is the real-time tilt angle change, is the critical dip threshold (related to geological type), is the weight coefficient of the inclination change, is the displacement rate influence function, is the real-time displacement rate, Critical rate threshold, is the weight coefficient of displacement rate;

[0102] according to The fusion weight of each sensor module is dynamically allocated based on the value. The weight calculation formula is:

[0103] ;

[0104] Where, is the fusion weight of the i-th sensor module, is the geological status assessment coefficient, is the confidence parameter of the i-th sensor module, is the total number of sensor modules;

[0105] The above process calculates confidence parameters based on sensor noise variance and historical error statistics, automatically reduces the weight of high-noise sensors, suppresses the impact of data anomalies on the fusion results, and dynamically adjusts the weight distribution based on the geological state assessment coefficient, making the fusion process more in line with actual geological conditions and avoiding insufficient sensitivity or misjudgment caused by fixed weights. Through the coupled calculation of the inclination change influence function and the displacement rate influence function, the deformation trend (inclination change) and evolution speed (displacement rate) of the landslide body are captured simultaneously, enhancing the ability to identify landslide stages (such as slow creep and accelerated sliding). The weight coefficient is dynamically bound to the geological type (such as clay focuses on inclination and rock focuses on displacement rate), so that the system can automatically adapt to different landslide mechanisms and improve the targeted monitoring.

[0106] For example, monitoring of a rock landslide hazard point

[0107] Sensor data is from spaceborne radar ( =0.9): Monitored monthly average displacement rate =8mm, lower noise.

[0108] Surface GNSS ( =0.7): Tilt change detected =2°, but the noise is higher due to weather interference.

[0109] Underground microseismic =0.8): Capture the rock fracture signal, displacement rate =5mm;

[0110] Dynamic weight allocation, geological type is rock, loading weight =0.7, =0.3.

[0111] Calculation of geological state assessment coefficients :

[0112] =0.35.

[0113] Total weight calculation: , =28%, =32%;

[0114] Effect: The displacement rate data (spaceborne radar, microseismic) has a higher weight (72%), accurately reflecting the rate-sensitive characteristics of rock landslides; the inclination data has a lower weight to avoid noise interference.

[0115] The tilt angle change affects the function and displacement rate influence function Defined as:

[0116] ;

[0117] ;

[0118] in, To prevent zero and tiny amounts, its value is positively correlated with the measurement accuracy of the sensor.

[0119] By introducing a small amount , to avoid the inclination angle changing function in ≈ When the denominator is zero, ensure the calculation is stable. The value of is positively correlated with the sensor accuracy, automatically adapting to the measurement errors of different sensors and reducing the impact of noise on the fusion results.

[0120] Linearly reflects the effect of inclination angle change on stability. Close to critical value When the function value decreases significantly, the potential risk is warned in time. The exponential decay form can sensitively capture high-rate displacement and respond more sensitively to the acceleration stage of the landslide, thus avoiding the warning delay that may be caused by the linear function.

[0121] Function design is based on the principles of geotechnical mechanics. 、 Directly related to geological type, it is convenient for engineers to adjust the threshold parameters according to site conditions.

[0122] For example, during the monitoring of a clay landslide hazard point:

[0123] Critical tilt threshold =4°, sensor accuracy is 0.05°, so ε=0.05.

[0124] Critical Rate Threshold =6mm / day.

[0125] Example 1: Calculation of tilt angle change function, real-time tilt angle change =3.8°:

[0126] ;

[0127] Effect: When the inclination angle approaches the critical value, the function value drops sharply, triggering an early warning.

[0128] Example 2: Displacement Rate Function Calculation

[0129] Real-time displacement rate =5.5mm / day:

[0130] ;

[0131] Effect: When the rate approaches the critical value, the function value decreases significantly, reflecting an accelerating trend.

[0132] The weight coefficient and Dynamic configuration through geological knowledge base includes the following steps:

[0133] Step 1: Knowledge base construction:

[0134] Based on geotechnical theory and historical landslide data, the correlation between different geological types and weight coefficients is established;

[0135] The geological types include at least typical landslide body types, which include clay, sand, rock and accumulation layers;

[0136] Weight coefficient 、 The allocation principle is:

[0137] Dip geology (e.g. clay): gives higher weight to dip changes ( > )

[0138] Rate-sensitive geology (such as rock): Give higher weight to displacement rate ( > );

[0139] Step 2: Identification of on-site geological types and multi-source data collection:

[0140] Obtaining geophysical parameters (such as moisture content, particle composition, and shear strength) through on-site drilling sampling, geological radar detection, and drone multispectral imaging;

[0141] Input the collected data into a pre-trained geological classification model (such as a random forest or convolutional neural network) to output the geological type label of the target area;

[0142] Step 3: Dynamically load the weight coefficient and search the corresponding geological type label in the knowledge base according to the geological type label output in step 2. 、 value;

[0143] Inject the matching weight coefficient into the dynamic weight fusion module in real time to update the geological state assessment coefficient The calculation logic of

[0144] Step 4: Exception handling mechanism: if the geological type of the site is not included in the mapping table, the default weight is used ( =0.5, =0.5) and trigger the manual verification process;

[0145] The rationality of the weight configuration is then verified by replaying historical data. If the deviation between the fusion results and the measured data exceeds the limit for N consecutive times, the knowledge base update is triggered;

[0146] Adaptive optimization of geological types dynamically matches weight coefficients through the geological knowledge base, allowing the data fusion process to automatically adapt to different landslide mechanisms (such as clay landslides focusing on inclination changes and rock landslides focusing on displacement rates), thereby improving the targeted monitoring.

[0147] To avoid insufficient sensitivity or misjudgment caused by manually preset fixed weights, geological classification models (such as random forests) quickly identify on-site geological types based on multi-source exploration data (drilling, radar, and drones), reducing manual survey time. The knowledge base mapping table converts geotechnical mechanics theory into executable parameter rules to ensure the physical rationality of weight allocation.

[0148] Exception handling mechanisms (such as default weights and manual verification) ensure that the system can still operate stably when the geological type is unknown or exploration data is missing. Historical data playback verification automatically triggers knowledge base updates and promotes continuous optimization of the system.

[0149] The real-time data assimilation module performs:

[0150] Step 1: Receive the fused displacement field data sent by the edge computing unit ;

[0151] Step 2: Input geomechanical parameterized model for inversion calculation and update the sliding zone parameter set ,in is the shear modulus, is the pore pressure coefficient, for cohesion;

[0152] Step 3: Calculate the parameter change rate ,when The model self-correction module is triggered when is the preset sensitivity threshold;

[0153] is the rate of change of the sliding band parameter set, is the updated sliding belt parameter set (including shear modulus , pore pressure coefficient , cohesion ), is the sliding belt parameter set before updating, is the preset sensitivity threshold;

[0154] Real-time parameter updates to improve model accuracy

[0155] By fusing displacement field data and inverting to update geomechanical parameters (such as shear modulus μ, pore pressure coefficient λ, and cohesion c), the digital twin model is dynamically synchronized with the actual geological conditions, avoiding prediction bias caused by the rigidity of traditional model parameters. The calculation of the parameter change rate can quantify the degree of model deviation, triggering the self-correction mechanism in a timely manner to prevent error accumulation. The update of the sliding zone parameter set directly reflects the changes in the internal mechanical properties of the landslide body (such as sliding zone softening and increased pore water pressure), providing a physical basis for early warning decisions. The setting of the sensitivity threshold η enables the system to distinguish between normal fluctuations and significant anomalies, balancing response speed and stability.

[0156] The regularized adjoint gradient method is used to accelerate parameter inversion, which reduces computational overhead while ensuring accuracy and is suitable for edge-cloud collaborative deployment.

[0157] For example, monitoring of a potential landslide site

[0158] Initial state:

[0159] Preset sensitivity threshold =5%;

[0160] First receive the fusion data, the edge computing unit sends the fusion displacement field data , showing that the displacement rate in the slope foot area increases abnormally.

[0161] The inversion calculation updates the parameters, and the real-time data assimilation module inverts the new parameters:

[0162] ;

[0163] Calculate the real-time rate of change, ;

[0164] because > , triggering the model self-correction module;

[0165] Model self-calibration, saving current parameter snapshots = ;

[0166] Generate perturbation parameter set ± , parallel calculation of predicted displacement fields;

[0167] The parameter group that best matches the subsequent 2 hours of measured data is selected as the new benchmark model.

[0168] The inversion calculation adopts the regularized adjoint gradient method, which specifically includes:

[0169] Construct the objective function:

[0170]

[0171] in, is the finite element forward operator, is the initial parameter, is the regularization coefficient;

[0172] By solving the adjoint equation:

[0173] ;

[0174] Get the parameter gradient ▽J, where is the stiffness matrix, is the accompanying variable;

[0175] The quasi-Newton method is used to iteratively update the parameters until convergence;

[0176] Efficient gradient calculation, significantly reducing computational costs,

[0177] The adjoint method can obtain the complete gradient of the objective function through one forward calculation (solving the finite element equation) and one adjoint calculation (solving the adjoint equation), avoiding the huge amount of calculation required by the traditional finite difference method for multiple forward calculations. Regularization ensures the stability of the inversion. The regularization term is introduced into the objective function to constrain the parameter update direction and prevent parameter drift caused by observation data noise or model errors (such as physically unreasonable values ​​such as negative shear modulus). By balancing the data fitting term and the parameter prior information, it suppresses overfitting and improves the model generalization ability.

[0178] The quasi-Newton method uses gradient information for efficient iterative optimization, accelerating convergence to the global optimal solution and ensuring the physical rationality of the inversion parameters. It accurately calculates the gradient along with the equation, avoids the truncation error of the numerical difference method, improves the accuracy of the parameter update direction, adapts to complex nonlinear problems, and can handle strong nonlinear relationships in geomechanical models (such as the coupling effect of pore pressure and displacement rate). It is suitable for parameter inversion requirements of different types of landslides, such as soil and rock.

[0179] For example, the shear modulus of a landslide Inversion

[0180] Initial parameters =50MPa, regularization coefficient γ=0.1.

[0181] Observational data: fused displacement field Shows abnormal local displacement of the slope.

[0182] Then construct the objective function, ;

[0183] Forward calculation of displacement field = , solve the adjoint equation, ;

[0184] Calculating gradients ;

[0185] The quasi-Newton method iterates along the gradient direction and eventually converges to =42 MPa, which is highly consistent with the drilling test result (41 MPa).

[0186] The model self-correction module performs the following operations:

[0187] Save a snapshot of the current parameters to the historical database;

[0188] Generate parameter perturbation set , =1, 2, 3...m;

[0189] Parallel calculation of the predicted displacement field corresponding to each disturbance parameter ;

[0190] Select the parameter group with the highest matching degree with the actual observation data in the subsequent Δt period as the update benchmark:

[0191] ;

[0192] Where, is the predicted displacement field of the kth group of disturbance parameters at time t;

[0193] is the actual observed displacement data at time t;

[0194] T is the length of the verification time window;

[0195] By generating multiple sets of parameter perturbation combinations and exploring different areas of the parameter space, we can avoid traditional optimization algorithms from falling into local optimal solutions, ensure the global optimality of model parameter updates, and simultaneously calculate the predicted displacement fields of multiple sets of parameters, significantly shortening the parameter screening time and meeting real-time monitoring needs. The prediction results of the perturbation parameters are verified by subsequent observation data, and the parameter group that best matches the actual landslide evolution is screened out to ensure the dynamic synchronization of model parameters and real geological conditions. Parameter snapshots are saved to form a historical database, providing data support for long-term landslide mechanism research. Parameter diversity coverage: Through parallel testing of multiple sets of perturbation parameters, the system can adapt to sudden changes in the mechanical properties of the landslide body (such as sliding zone fractures and sudden increases in pore water pressure), enhancing the model's adaptability to complex geological conditions. The parameter group with the highest matching degree is directly fed back to the digital twin model to optimize subsequent prediction accuracy, forming a closed loop from prediction, verification to optimization.

[0196] The dynamic weight fusion module and the real-time data assimilation module form a bidirectional coupling mechanism, which is specifically manifested as follows:

[0197] When the model self-correction module updates the sliding belt parameter set After that, the geological state assessment coefficient is automatically triggered The critical inclination threshold is and critical rate threshold According to the new Dynamic value adjustment:

[0198] ;

[0199] ;

[0200] in, 、 is an empirical function based on geotechnical mechanics, is the rock mass density;

[0201] Dynamic closed-loop optimization improves system adaptability. After the model self-correction model updates the sliding zone parameter set, it automatically triggers the recalculation of the geological state assessment coefficient, realizing a closed-loop process from data fusion to model update to parameter feedback to re-optimization. The critical inclination threshold and critical rate threshold are dynamically adjusted according to the updated parameters, so that the warning conditions evolve synchronously with the actual geological state, avoiding misjudgment caused by static thresholds, and recalculating through the updated Ψ value. , dynamically adjust sensor weight distribution, give priority to key parameters (such as displacement rate or tilt angle change), empirical function 、 The design is based on geotechnical mechanics theory to ensure the physical rationality of threshold adjustment and improve the response speed to the acceleration stage of landslide. High-precision model parameters provide physical constraints for data fusion. High-quality fused data feeds back to model optimization to form a positive cycle. The automated closed-loop mechanism reduces dependence on manual parameter adjustment and adapts to long-term monitoring needs in complex geological environments.

[0202] The 3D path planning module performs:

[0203] Receive the landslide impact boundary output by the digital twin modeling unit and then construct a three-dimensional terrain cost function:

[0204] ;

[0205] in, is the slope factor, is the surface crack density, is the distance between the current position and the landslide boundary;

[0206] The improved A* algorithm is used to search for the minimum cost path from the personnel location to the safe zone, and the path weight coefficient 、 and Dynamic adjustment according to the warning level;

[0207] Real-time landslide boundary prediction: Dynamically update evacuation routes based on the landslide impact boundary output by the digital twin modeling unit to avoid real-time evolving danger zones and ensure the timeliness and safety of escape routes.

[0208] Multi-factor comprehensive assessment: The cost function integrates slope, surface crack density, and distance to the landslide boundary to avoid path risks caused by a single factor.

[0209] Adaptive warning level response, dynamic weight adjustment: path weight coefficient 、 and Automatically adjust according to the warning level. For example:

[0210] During a red alert, avoid the landslide front first (increase weight) ;

[0211] When yellow warning, focus on selecting low-gradient path (increase weight).

[0212] Hierarchical safety strategy: different warning levels match different path optimization goals, balance safety and traffic efficiency.

[0213] Based on the traditional A* algorithm, the cost function dynamic updating mechanism is introduced, which quickly responds to the changes of the terrain, reduces the calculation delay, meets the real-time demand of emergency scenes, and supports path search under complex terrain (such as cliffs and valleys) based on high-precision three-dimensional terrain data of digital twinning, avoiding the limitations of two-dimensional plane planning.

[0214] In addition, the terms "first", "second", "third", etc. are used only for descriptive purposes and should not be construed as indicating or implying relative importance or an indicated number of technical features. Therefore, the features defined with "first", "second", etc. can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, such as two, three, etc., unless otherwise specifically limited.

[0215] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In the present application, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the skilled in the art can combine and combine the different embodiments or examples described in the present application and the features of different embodiments or examples without contradiction.

[0216] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A landslide potential point displacement monitoring system using multi-source data fusion, characterized by: include Multi-source sensing unit, the multi-source sensing unit includes: Spaceborne radar sensing module, used to obtain large-scale surface deformation data through synthetic aperture radar; A surface deformation sensing module, which is used to monitor surface displacement in real time using a dual-frequency GNSS receiver and a fiber Bragg grating sensor; The underground stress sensing module is used to obtain the stress changes of underground rock and soil through microseismic monitoring arrays and distributed optical fiber sensors; Edge computing unit, the edge computing unit includes: Dynamic weight fusion module, used to dynamically assign multi-source sensor weights according to geological conditions to generate high-precision fused displacement field data; Communication protocol adapter module, used to achieve protocol conversion and low-latency transmission of multi-source sensor data; Digital twin modeling unit, the digital twin modeling unit includes: Geomechanical parameterized model, used to construct a three-dimensional mechanical model of the landslide based on the finite element method; Real-time data assimilation module, used to input fused data into the model for inversion and update of geomechanical parameters; Model self-correction module, used to optimize model accuracy through parameter perturbation testing; Intelligent early warning unit, the intelligent early warning unit includes: Multi-level warning trigger module, used to trigger graded warnings based on displacement rate, acceleration and underground parameter anomalies; The three-dimensional path planning module is used to dynamically plan safe evacuation paths based on digital twin prediction results.

2. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 1 is characterized by: The dynamic weight fusion module performs the following operations: Acquire real-time measurement data of each sensor module and its confidence parameter, wherein the confidence parameter is determined jointly based on sensor noise variance and historical error statistics; Calculate the geological state assessment coefficient α, its expression is: α=f1(Δθ,θ c )·w θ +f2(v,v c )·w v ; Where f1 is the inclination change influence function, Δθ is the real-time inclination change, and θ c is the critical tilt angle threshold, w θ is the weight coefficient of the inclination change, f2 is the displacement rate influence function, v is the real-time displacement rate, v c Critical rate threshold, w v is the weight coefficient of displacement rate The fusion weight of each sensor module is dynamically allocated according to the α value. The weight calculation formula is: Where w i is the fusion weight of the i-th sensor module, α is the geological state assessment coefficient, β i is the confidence parameter of the i-th sensor module, and n is the total number of sensor modules.

3. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 2 is characterized by: The inclination change influence function f1 and displacement rate influence function f2 are defined as: Among them, ε is the tiny amount of zero elimination, and its value is positively correlated with the measurement accuracy of the sensor.

4. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 2 or 3 is characterized in that: The weight coefficient w θ With w v Dynamic configuration through geological knowledge base includes the following steps: Step 1: Knowledge base construction: Based on geotechnical theory and historical landslide data, the correlation between different geological types and weight coefficients is established; The geological types include at least typical landslide body types, which include clay, sand, rock and accumulation layers; Weight coefficient w θ 、w v The allocation principle is: Dip geology: gives higher weight to dip changes; Rate-sensitive geology: assigns higher weight to displacement rate; Step 2: Identification of on-site geological types and multi-source data collection: Obtaining rock and soil physical parameters through on-site drilling sampling, geological radar detection, and drone multispectral imaging; Input the collected data into the pre-trained geological classification model to output the geological type label of the target area; Step 3: Dynamically load the weight coefficient. According to the geological type label output in step 2, search the corresponding w in the knowledge base. θ 、w v value; Inject the matching weight coefficient into the dynamic weight fusion module in real time to update the calculation logic of the geological state assessment coefficient α; Step 4: Exception handling mechanism: if the on-site geological type is not included in the mapping table, the default weight is enabled and the manual verification process is triggered; The rationality of the weight configuration is then verified by replaying historical data. If the deviation between the fusion results and the measured data exceeds the limit for N consecutive times, the knowledge base update is triggered.

5. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 1 is characterized by: The real-time data assimilation module performs: Step 1: Receive the fused displacement field data D sent by the edge computing unit f ; Step 2: D f Input the geomechanical parameterized model for inversion calculation and update the sliding zone parameter set Ψ = (μ, λ, c), where μ is the shear modulus, λ is the pore pressure coefficient, and c is the cohesion; Step 3: Calculate the parameter change rate When δ Ψ >η, triggering the model self-correction module, where η is the preset sensitivity threshold; δ Ψ is the rate of change of the sliding band parameter set, Ψ new is the updated sliding belt parameter set, Ψ old is the sliding band parameter set before updating, and η is the preset sensitivity threshold.

6. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 5 is characterized by: The inversion calculation adopts the regularized adjoint gradient method, which specifically includes: Construct the objective function: K(Ψ)=||FEM(Ψ)-D f || 2 +γ||Ψ-Ψ0|| 2 Where FEM is the finite element forward operator, Ψ0 is the initial parameter, and γ is the regularization coefficient; By solving the adjoint equation: Get parameter gradients Where K is the stiffness matrix, u adj is the accompanying variable; The quasi-Newton method is used to iteratively update the parameters until convergence.

7. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 1 is characterized by: The model self-correction module performs the following operations: Save the current parameter snapshot Ψ current to the historical database; Generate parameter perturbation set {Ψ current ±ΔΨ k }, k = 1, 2, 3, ... m; Parallel calculation of the predicted displacement field corresponding to each disturbance parameter Select the parameter group with the highest matching degree with the actual observation data in the subsequent Δt period as the update benchmark: Where, is the predicted displacement field of the kth group of disturbance parameters at time t; D o (t) is the actual observed displacement data at time t; T is the length of the verification time window.

8. The landslide potential point displacement monitoring system using multi-source data fusion according to claim 2 or 5, characterized in that: The dynamic weight fusion module and the real-time data assimilation module form a bidirectional coupling mechanism, which is specifically manifested as follows: When the model self-correction module updates the sliding zone parameter set Ψ, it automatically triggers the recalculation of the geological state assessment coefficient α, where the critical inclination threshold θ c and the critical rate threshold v c Dynamic adjustment based on the new Ψ value: Among them, g1 and g2 are empirical functions based on rock and soil mechanics, and ρ is the rock density.

Images