Collaborative monitoring and early warning method and system for landslides in high-risk areas of geological disasters

Abstract

The present invention discloses a method and system for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters, which is applied to the field of intelligent monitoring and early warning technology. The method includes: obtaining landslide surface data and landslide geological structure data of a target landslide; setting multiple monitoring devices; obtaining landslide movement type data and landslide deformation type data; obtaining predicted displacement data; calculating a first correction weight, and correcting the first correction weight to form a second correction weight; updating a second neural network model; judging whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model. The present invention analyzes different landslide movement types and landslide deformation types, and applies the results of the analysis to the updating process of the neural network model, which effectively improves the accuracy of model prediction and the iteration efficiency, so that after various emergencies, the prediction model can be updated to the latest state in a timely manner, effectively improving safety and applicability.

Description

Technical Field

[0001] The present invention relates to intelligent monitoring and early warning technology, and in particular to a collaborative monitoring and early warning method and system for landslides in high-risk areas of geological disasters. Background Art

[0002] The study of landslide early warning models has long been a research focus in the field of geological engineering, attracting extensive attention and in-depth research from many scholars at home and abroad. Some studies focus on the triggering conditions of landslides, especially the critical rainfall analysis of rainfall-type landslides, as shown in the work of Aleotti (2004), Cardinali et al. (2006) and Martelloni et al. (2012). On the other hand, landslide early warning methods based on deformation displacement have also made significant progress. Representative studies include the series of results of Xu Qiang et al. (2008, 2009, 2015). As the main inducing factor of landslides, rainfall plays a vital role in landslide prediction (Guzzetti et al., 2008), and can be subdivided into physical thresholds and empirical thresholds according to methodological differences (Aleotti, 2004). In recent years, the application of synthetic aperture radar interferometry (InSAR) technology in landslide deformation monitoring and early identification has significantly improved the early warning capability and has become an indispensable part of the "three checks" system (Xu Qiang, 2020). This technology has successfully captured the significant acceleration stage before the landslide in cases such as the Xinmo Village landslide (Dai et al., 2020; Dong et al., 2018), the Huangnibazi landslide in Li County (He Chaoyang et al., 2019; Jie Mingli et al., 2020), and the Canadian gold mine slope (Carlà et al., 2019). It is mainly used to monitor the long-term slow deformation process of the landslide body and its trend prediction. The fitting and extrapolation prediction of landslide deformation monitoring curves based on mathematical methods have become a research hotspot. A variety of mathematical models have been applied in this field, including the Verhulst model (Yan Tongzhen, 1988), the grey prediction model (Chen Mingdong and Wang Lansheng, 1988), the golden section prediction model (Zhang Zhuoyuan and Huang Runqiu, 1988), the collaborative prediction model and the nonlinear dynamic model (Qin Siqing, 2005), etc. Xu Qiang et al. (2008, 2009, 2015, 2018, 2020) conducted a comprehensive analysis of the monitoring, early warning and emergency rescue practices of dozens of major landslide disasters, and deeply explored the general laws of the landslide deformation instability process and the stage characteristics of the spatial evolution of slope cracks, revealing the slope and acceleration change characteristics of the creep landslide deformation-time curve. On this basis, they established an improved tangent angle model for landslide process warning and proposed corresponding landslide warning criteria. The model showed good warning effects in practical applications and made important contributions to the development of landslide warning theory and practice.

[0003] Landslide geological disasters are highly concealed and sudden. Every year, some geological disasters still occur outside the plans for geological disaster risk points that have been checked. The existing geological disaster monitoring and early warning platform mainly adopts a single monitoring point, a single monitoring instrument, and a single parameter threshold early warning mode. However, due to the complex geological environment of the landslide geological disaster site, the complexity of the geological structure of the landslide slope, the suddenness of geological disasters and other characteristic factors, there are false alarms. Therefore, it is necessary to conduct further research on high-risk geological disaster risk points and establish a set of collaborative early warning models to improve the accuracy of early warning forecasts. Summary of the invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of the present application is to provide a method or system for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters.

[0005] In a first aspect, the present application provides a method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters, including:

[0006] Acquire the landslide surface data and landslide geological structure data of the target landslide; the landslide surface data is acquired by at least one of space-based satellite observation, unmanned aerial vehicle observation and surface three-dimensional laser observation; the landslide geological structure data is acquired by geophysical exploration method;

[0007] According to the landslide surface data and the landslide geological structure data, a plurality of monitoring devices are set for the target landslide, and the monitoring devices are divided into front edge devices, middle section devices and rear edge devices according to their positions relative to the target landslide;

[0008] Input the monitoring data at the current moment into a preset first neural network model, and obtain the landslide movement type data and landslide deformation type data output by the first neural network model; the monitoring data includes leading edge monitoring data, middle section monitoring data and trailing edge monitoring data; the leading edge monitoring data is the displacement data monitored by the leading edge device, the middle section monitoring data is the displacement data monitored by the middle section device, and the trailing edge monitoring data is the displacement data monitored by the trailing edge device;

[0009] Inputting the monitoring data at the current moment into a preset second neural network model, and obtaining displacement data output by the second neural network model as predicted displacement data;

[0010] Calculating a first correction weight by the difference between the predicted displacement data and the monitoring data at the next moment, and correcting the first correction weight according to the landslide deformation type data and the landslide movement type data to form a second correction weight;

[0011] updating the second neural network model according to the first revised weight and the second revised weight, and repeatedly acquiring the monitoring data to update the second neural network model;

[0012] Whether the target landslide is safe is determined based on the predicted displacement data output by the updated second neural network model.

[0013] In a possible implementation, the configuration of the monitoring device includes:

[0014] Dividing the target landslide into a front edge area, a middle area and a rear edge area according to the landslide surface data and the landslide geological structure data;

[0015] Selecting a suitable position on the landslide surface to deploy a slope displacement monitoring instrument according to the landslide surface data, and selecting a suitable position inside the landslide to deploy a deep displacement meter monitoring instrument according to the landslide geological structure data;

[0016] The monitoring instrument in the leading edge area is used as the leading edge device, the monitoring instrument in the middle section area is used as the middle section device, and the monitoring instrument in the trailing edge area is used as the trailing edge device.

[0017] In a possible implementation, the construction of the first neural network model includes:

[0018] Constructing finite element models or discrete element models under various working conditions to form numerical simulation models; the constraint conditions corresponding to the working conditions include landslide movement type and landslide deformation type; the landslide movement types include push type, traction type and composite type; the landslide deformation types include creep cracking type, sliding cracking type, bending cracking type, plastic flow cracking type, sliding bending type and sliding compression cracking type;

[0019] Performing deformation and destruction numerical simulation on a plurality of the numerical simulation models to obtain leading edge sample data, middle section sample data and trailing edge sample data; the leading edge sample data is the displacement data of the leading edge of the landslide in the numerical simulation model, the middle section sample data is the displacement data of the middle section of the landslide in the numerical simulation model, and the trailing edge sample data is the displacement data of the trailing edge of the landslide in the numerical simulation model;

[0020] The leading edge sample data, the middle section sample data and the trailing edge sample data are used as input, and the landslide movement type and the landslide deformation type are used as output data to train a neural network model to form the first neural network model.

[0021] In a possible implementation, the construction of the second neural network model includes:

[0022] The leading edge sample data, middle segment sample data and trailing edge sample data at the current moment are used as input data, and the leading edge sample data, middle segment sample data and trailing edge sample data at the next moment are used as output data to train a neural network model to form the second neural network model.

[0023] In a possible implementation, the calculation of the second modified weight includes:

[0024] Calculate the difference between the predicted displacement data and the monitoring data at the next moment as a loss value through a loss function;

[0025] Calculate the gradient of each layer in the second neural network model by back propagation of the loss value, and calculate the weight of the second neural network model by the gradient as the first modified weight;

[0026] Obtaining weights of the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data corresponding to the landslide deformation type data as deformation weights, and obtaining weights of the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data corresponding to the landslide movement type data as movement weights;

[0027] The deformation weight and the motion weight are multiplied by the first correction weight corresponding to the input layer to the hidden layer in the first correction weight to form the second correction weight.

[0028] In a possible implementation, judging whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model includes:

[0029] Acquire the deformation weight and the motion weight of the monitoring device corresponding to the predicted displacement data;

[0030] Performing weighted calculation on the predicted displacement data by using the deformation weight and the motion weight to obtain an early warning value of the target landslide;

[0031] Whether the target landslide is safe is determined according to the warning value.

[0032] In a possible implementation, the landslide movement types include push type, traction type and combined type;

[0033] When the landslide movement type is push-type, the movement weight of the front edge monitoring data is greater than the movement weight of the middle section monitoring data, and the movement weight of the middle section monitoring data is greater than the movement weight of the rear edge monitoring data;

[0034] When the landslide movement type is traction type, the movement weight of the trailing edge monitoring data is greater than the movement weight of the leading edge monitoring data, and the movement weight of the leading edge monitoring data is greater than the movement weight of the middle section monitoring data;

[0035] When the landslide motion type is compound, the motion weight of the middle monitoring data is greater than the motion weight of the trailing edge monitoring data, and the motion weight of the leading edge monitoring data is equal to the motion weight of the trailing edge monitoring data.

[0036] In a possible implementation, the landslide deformation types include creep cracking, sliding cracking, bending cracking, plastic flow cracking, sliding bending and sliding compression cracking;

[0037] When the landslide deformation type is the creep and tensile crack type, the deformation weight of the front edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the rear edge monitoring data;

[0038] When the landslide deformation type is the sliding and pulling type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the leading edge monitoring data, and the deformation weight of the leading edge monitoring data is greater than the deformation weight of the middle section monitoring data;

[0039] When the landslide deformation type is the bending and tensile cracking type, the deformation weight of the front edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the rear edge monitoring data;

[0040] When the landslide deformation type is the plastic flow tensile cracking type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the leading edge monitoring data, and the deformation weight of the leading edge monitoring data is greater than the deformation weight of the middle section monitoring data;

[0041] When the landslide deformation type is the slip bending type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the leading edge monitoring data;

[0042] When the landslide deformation type is the sliding suppression and tensile cracking type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the leading edge monitoring data is equal to the deformation weight of the middle section monitoring data.

[0043] In a possible implementation, updating the second neural network model according to the first revised weight and the second revised weight includes:

[0044] The weights from the input layer to the hidden layer in the second neural network model are updated to the corresponding second revised weights, and the weights from the non-input layer to the hidden layer in the second neural network model are updated to the corresponding first revised weights.

[0045] Secondly, this application also provides a landslide collaborative monitoring and early warning system for high-risk areas of geological disasters, including:

[0046] An acquisition unit is configured to acquire landslide surface data and landslide geological structure data of a target landslide; the landslide surface data is acquired by at least one of space-based satellite observation, drone observation, and surface three-dimensional laser observation; the landslide geological structure data is acquired by a geophysical exploration method;

[0047] A monitoring unit is configured to set a plurality of monitoring devices for the target landslide according to the landslide surface data and the landslide geological structure data, and divide the monitoring devices into front edge devices, middle section devices and rear edge devices according to their positions relative to the target landslide;

[0048] The analysis unit is configured to input the monitoring data at the current moment into a preset first neural network model, and obtain the landslide movement type data and the landslide deformation type data output by the first neural network model; the monitoring data includes leading edge monitoring data, middle section monitoring data and trailing edge monitoring data; the leading edge monitoring data is the displacement data monitored by the leading edge device, the middle section monitoring data is the displacement data monitored by the middle section device, and the trailing edge monitoring data is the displacement data monitored by the trailing edge device;

[0049] A calculation unit, configured to input the monitoring data at the current moment into a preset second neural network model, and obtain displacement data output by the second neural network model as predicted displacement data;

[0050] an updating unit, configured to calculate a first correction weight by a difference between the predicted displacement data and the monitoring data at the next moment, and to correct the first correction weight according to the landslide deformation type data and the landslide movement type data to form a second correction weight;

[0051] a circulation unit, configured to update the second neural network model according to the first revised weight and the second revised weight, and repeatedly obtain the monitoring data to update the second neural network model;

[0052] The judgment unit is configured to judge whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model.

[0053] In a possible implementation, the configuration of the monitoring device includes:

[0054] Dividing the target landslide into a front edge area, a middle area and a rear edge area according to the landslide surface data and the landslide geological structure data;

[0055] Selecting a suitable position on the landslide surface to deploy a slope displacement monitoring instrument according to the landslide surface data, and selecting a suitable position inside the landslide to deploy a deep displacement meter monitoring instrument according to the landslide geological structure data;

[0056] The monitoring instrument in the leading edge area is used as the leading edge device, the monitoring instrument in the middle section area is used as the middle section device, and the monitoring instrument in the trailing edge area is used as the trailing edge device.

[0057] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0058] The present invention effectively utilizes the data generated by integrated above-ground and underground monitoring. By analyzing different types of landslide movement and deformation, and applying the results of the analysis to the updating process of the BP neural network model, the accuracy of model prediction is effectively improved, and the iteration efficiency is improved. After various emergencies, the prediction model can be updated to the latest status in time, effectively improving safety and applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0060] Figure 1 This is a schematic diagram of the steps of the method of the embodiment of the present application;

[0061] Figure 2 This is a schematic diagram of updating the second neural network model according to an embodiment of the present application. DETAILED DESCRIPTION

[0062] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the drawings in the embodiment of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the embodiment of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0063] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0064] Please refer to Figure 1 , which is a flow chart of the collaborative monitoring and early warning method for landslides in high-risk areas of geological disasters provided in an embodiment of the present invention. Furthermore, the collaborative monitoring and early warning method for landslides in high-risk areas of geological disasters may specifically include the contents described in the following steps S1 to S7.

[0065] S1: Obtaining landslide surface data and landslide geological structure data of a target landslide; the landslide surface data is obtained by at least one of space-based satellite observation, drone observation and surface three-dimensional laser observation; the landslide geological structure data is obtained by geophysical exploration methods;

[0066] S2: setting a plurality of monitoring devices for the target landslide according to the landslide surface data and the landslide geological structure data, and dividing the monitoring devices into front edge devices, middle section devices and rear edge devices according to their positions relative to the target landslide;

[0067] S3: input the monitoring data at the current moment into a preset first neural network model, and obtain the landslide movement type data and landslide deformation type data output by the first neural network model; the monitoring data includes leading edge monitoring data, middle section monitoring data and trailing edge monitoring data; the leading edge monitoring data is the displacement data monitored by the leading edge device, the middle section monitoring data is the displacement data monitored by the middle section device, and the trailing edge monitoring data is the displacement data monitored by the trailing edge device;

[0068] S4: inputting the monitoring data at the current moment into a preset second neural network model, and obtaining displacement data output by the second neural network model as predicted displacement data;

[0069] S5: calculating a first correction weight by the difference between the predicted displacement data and the monitoring data at the next moment, and correcting the first correction weight according to the landslide deformation type data and the landslide movement type data to form a second correction weight;

[0070] S6: updating the second neural network model according to the first correction weight and the second correction weight, and repeatedly acquiring the monitoring data to update the second neural network model;

[0071] S7: Determine whether the target landslide is safe based on the predicted displacement data output by the updated second neural network model.

[0072] When the embodiment of the present application is implemented, it is necessary to obtain the landslide surface data and landslide geological structure data through corresponding monitoring means, which is realized through an integrated monitoring scheme for the ground and underground; wherein the space-based satellite can obtain the landslide surface remote sensing data, observe the deformation signs of the landslide, delineate the impact range of the landslide, and obtain the surface deformation data through the satellite-borne SAR; the drone observation can observe the subtle changes of the surface, and the airborne laser radar can obtain high-precision point cloud data + stripping vegetation cover shielding, which is applied to construct a refined three-dimensional model of the surface; the surface three-dimensional laser observation can establish a surface three-dimensional model; the geophysical exploration method can obtain the underground geological structure, and the surface field verification verifies the data interpretation results obtained by the space-based satellite observation and the aviation platform. Through the acquired landslide surface data and landslide geological structure data, the points suitable for the deployment of monitoring equipment can be determined, and these monitoring equipment can be divided into leading edge equipment, middle section equipment and trailing edge equipment, wherein the leading edge equipment is located at the leading edge position of the target landslide, the middle section equipment is located at the middle section position of the target landslide, and the trailing edge equipment is located at the trailing edge position of the target landslide.

[0073] When the embodiment of the present application is implemented, the monitoring data monitored at the current moment can be used to judge the landslide movement type and landslide deformation type of the target landslide based on the first neural network model. For different landslide movement types and landslide deformation types, the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data will be assigned different weight values ​​to characterize the impact of the landslide movement type and the landslide deformation type on different monitoring data. At this time, when the second neural network model is used to predict the data at the next moment, the second neural network model can be corrected according to the weight values ​​corresponding to the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data, so as to accelerate the calculation convergence speed of the model and improve the prediction accuracy. In the embodiment of the present application, the second neural network model is preferably a BP neural network model, and its back propagation can correct the weight in the model by the monitoring data at the next moment and the prediction data at the previous moment. In the process of correction, the weight values ​​corresponding to the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data are introduced again to make the second neural network model have better applicability, and the prediction result is closer to the current target landslide. In the process of continuous iteration and updating, the prediction results of the second neural network model will become more and more accurate, and even if the target landslide has rainfall, earthquake and other events, and its corresponding shape changes, the corresponding weights newly generated by the first neural network model can better iterate the second neural network model to a state that conforms to the current landslide situation, greatly improving the applicability of the embodiment of the present application. By performing the safety evaluation of the target landslide through the predicted displacement output by the second neural network model, the possible future conditions of the target landslide can be obtained one step earlier, so that corresponding measures can be taken in time, effectively improving safety. The embodiment of the present application uses the data generated by the above-ground and underground integrated monitoring through the above-mentioned technical solution, analyzes different types of landslide movement and deformation types of landslide, and applies the results of the analysis in the updating process of the BP neural network model, effectively improving the accuracy of model prediction, improving iteration efficiency, so that after various emergencies, the prediction model can be updated to the latest state in time, effectively improving safety and applicability.

[0074] In a possible implementation, the configuration of the monitoring device includes:

[0075] Dividing the target landslide into a front edge area, a middle area and a rear edge area according to the landslide surface data and the landslide geological structure data;

[0076] Selecting a suitable position on the landslide surface to deploy a slope displacement monitoring instrument according to the landslide surface data, and selecting a suitable position inside the landslide to deploy a deep displacement meter monitoring instrument according to the landslide geological structure data;

[0077] The monitoring instrument in the leading edge area is used as the leading edge device, the monitoring instrument in the middle section area is used as the middle section device, and the monitoring instrument in the trailing edge area is used as the trailing edge device.

[0078] In a possible implementation, the construction of the first neural network model includes:

[0079] Constructing finite element models or discrete element models under various working conditions to form numerical simulation models; the constraint conditions corresponding to the working conditions include landslide movement type and landslide deformation type; the landslide movement types include push type, traction type and composite type; the landslide deformation types include creep cracking type, sliding cracking type, bending cracking type, plastic flow cracking type, sliding bending type and sliding compression cracking type;

[0080] Performing deformation and destruction numerical simulation on a plurality of the numerical simulation models to obtain leading edge sample data, middle section sample data and trailing edge sample data; the leading edge sample data is the displacement data of the leading edge of the landslide in the numerical simulation model, the middle section sample data is the displacement data of the middle section of the landslide in the numerical simulation model, and the trailing edge sample data is the displacement data of the trailing edge of the landslide in the numerical simulation model;

[0081] The leading edge sample data, the middle section sample data and the trailing edge sample data are used as input, and the landslide movement type and the landslide deformation type are used as output data to train a neural network model to form the first neural network model.

[0082] When the embodiment of the present application is implemented, the first neural network model and the second neural network model are both standardized models applicable to different landslide states, so in the training process, the sample data is also calculated based on the standardized landslide state, and its main purpose is to screen out the difference between the leading edge sample data, the middle section sample data and the trailing edge sample data under different working condition combinations. In the embodiment of the present application, the selection of working conditions is carried out in an orthogonal manner, that is, all landslide motion types and all landslide deformation types are orthogonally combined to form a variety of working conditions, wherein the landslide motion types are three, the landslide deformation types are six, and the working conditions after orthogonal combination are eighteen, and these eighteen working conditions are the output data of the first neural network model. The input data of the first neural network model is input in the form of three data groups, which can form the input sample by weighted averaging the corresponding sample data to form a data group, or it can be directly used as the input sample in the form of an array, and the embodiment of the present application is not limited. The first neural network model finally trained can be applicable to various landslides, and its main function is to judge the landslide motion type and landslide deformation type corresponding to the landslide according to the monitoring data.

[0083] In a possible implementation, the construction of the second neural network model includes:

[0084] The leading edge sample data, middle segment sample data and trailing edge sample data at the current moment are used as input data, and the leading edge sample data, middle segment sample data and trailing edge sample data at the next moment are used as output data to train a neural network model to form the second neural network model.

[0085] When the embodiment of the present application is implemented, a second neural network model can be initialized and trained based on the same working conditions and sample data, which can have wide applicability and can be iteratively updated through subsequent actual monitoring data.

[0086] In a possible implementation, the calculation of the second modified weight includes:

[0087] Calculate the difference between the predicted displacement data and the monitoring data at the next moment as a loss value through a loss function;

[0088] Calculate the gradient of each layer in the second neural network model by back propagation of the loss value, and calculate the weight of the second neural network model by the gradient as the first modified weight;

[0089] Obtaining weights of the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data corresponding to the landslide deformation type data as deformation weights, and obtaining weights of the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data corresponding to the landslide movement type data as movement weights;

[0090] The deformation weight and the motion weight are multiplied by the first correction weight corresponding to the input layer to the hidden layer in the first correction weight to form the second correction weight.

[0091] When implementing the embodiments of this application, please refer to Figure 2 , showing the iterative update process of the second neural network model, wherein the loss value can be calculated first by a loss function, such as a mean square error function, which characterizes the difference between the predicted displacement data and the monitoring data at the next moment. Then, the gradients of different layers in the second neural network model are calculated based on the back propagation of the BP neural network, and the weights and biases are further calculated, wherein the weights are the first correction weights; the first correction weights can be used to correct the weights of other positions except the input layer to the first hidden layer, and the weights from the input layer to the first hidden layer are further corrected by deformation weights and motion weights. The inventors have found that due to the differences in the impact of different working conditions on the leading edge monitoring data, the middle section monitoring data and the trailing edge monitoring data, in order to accelerate the iterative convergence of the second neural network model, the weights from the input layer to the first hidden layer can be further corrected by the corresponding deformation weights and motion weights, thereby more accurately characterizing the impact of the leading edge monitoring data, the middle section monitoring data and the trailing edge monitoring data on the prediction results.

[0092] In a possible implementation, judging whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model includes:

[0093] Acquire the deformation weight and the motion weight of the monitoring device corresponding to the predicted displacement data;

[0094] Performing weighted calculation on the predicted displacement data by using the deformation weight and the motion weight to obtain an early warning value of the target landslide;

[0095] Whether the target landslide is safe is determined according to the warning value.

[0096] When the embodiment of the present application is implemented, the deformation weight and the motion weight can be further applied to the calculation of the warning value. The warning value is used to express the state of the entire target landslide. The corresponding threshold can be set to determine whether the warning value exceeds the limit; different intervals can also be divided to see which interval the warning value falls into, such as blue warning, yellow warning, orange warning and red warning.

[0097] In a possible implementation, the landslide movement types include push type, traction type and combined type;

[0098] When the landslide movement type is push-type, the movement weight of the front edge monitoring data is greater than the movement weight of the middle section monitoring data, and the movement weight of the middle section monitoring data is greater than the movement weight of the rear edge monitoring data;

[0099] When the landslide movement type is traction type, the movement weight of the trailing edge monitoring data is greater than the movement weight of the leading edge monitoring data, and the movement weight of the leading edge monitoring data is greater than the movement weight of the middle section monitoring data;

[0100] When the landslide motion type is compound, the motion weight of the middle monitoring data is greater than the motion weight of the trailing edge monitoring data, and the motion weight of the leading edge monitoring data is equal to the motion weight of the trailing edge monitoring data.

[0101] When the embodiment of the present application is implemented, the front edge movement weight is assigned Front1, the middle section movement weight is assigned Central1, and the rear edge movement weight is assigned Rear1. The inventors have found in scientific practice that for different types of landslide movements: in a push-type landslide, the rear part slides, squeezes and pushes the front part, and the displacement deformation of the front part determines the instability of the entire slope, so the movement weight is Front1>Central1>Rear1; in a traction-type landslide, the front part is first deformed and destroyed to form an air surface, and the rear slope develops and destroys successively under the action of gravity and other factors, and the deformation and destruction of the rear part determines the overall instability of the landslide, so the movement weight is Rea1r>Front1>Central1; a compound landslide has both push-type and traction-type characteristics, and the middle locking section determines whether the landslide is destroyed as a whole, so the movement weight is Central1>Front1=Rear1.

[0102] In a possible implementation, the landslide deformation types include creep cracking, sliding cracking, bending cracking, plastic flow cracking, sliding bending and sliding compression cracking;

[0103] When the landslide deformation type is the creep and tensile crack type, the deformation weight of the front edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the rear edge monitoring data;

[0104] When the landslide deformation type is the sliding and pulling type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the leading edge monitoring data, and the deformation weight of the leading edge monitoring data is greater than the deformation weight of the middle section monitoring data;

[0105] When the landslide deformation type is the bending and tensile cracking type, the deformation weight of the front edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the rear edge monitoring data;

[0106] When the landslide deformation type is the plastic flow tensile cracking type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the leading edge monitoring data, and the deformation weight of the leading edge monitoring data is greater than the deformation weight of the middle section monitoring data;

[0107] When the landslide deformation type is the slip bending type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the leading edge monitoring data;

[0108] When the landslide deformation type is the sliding suppression and tensile cracking type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the leading edge monitoring data is equal to the deformation weight of the middle section monitoring data.

[0109] When the embodiment of the present application is implemented, the front edge deformation weight is assigned to Front2, the middle section deformation weight is assigned to Central2, and the rear edge deformation weight is assigned to Rear2. The inventors have found in scientific practice that for different landslide deformation types:

[0110] Creep-slip tensile landslides mainly develop in homogeneous or quasi-homogeneous slopes. The rear edge develops tensile deformation from the slope surface to the deep part, and the potential shear surface shears through to form a landslide. The main stages of deformation and destruction development are: ① The slope bends along the potential sliding surface to the free surface, generating tensile stress on the rear part, ② Tensile stress concentration parts crack, ③ Shear disturbance of the potential shear surface, and the front edge rises. The displacement of the front edge and the middle part becomes the controlling condition for the formation of this type of landslide, so the deformation weight Front2>Central2>Rear2.

[0111] Slip-rupture landslides mainly develop on gently inclined outer layered slopes and blocky slopes. The weak surface characteristics of the slip surface determine the formation of the landslide. Once the rear edge rupture surface appears, it slides rapidly, so the deformation weight Rear2>Front2>Central2.

[0112] The bending and tensile cracking landslide mainly develops in the medium-to-extreme steep slope composed of the steep outer layered slope and the steep vertical-steep inner layered slope. The steep plate-like rock mass bends in the direction of the air as a cantilever beam under its own gravity, and develops into the inside of the slope while generating tensile cracks, and then develops into fractures at the part with severe bending effect. The main stages of deformation and damage development are: ① tensile cracking of the steep surface due to unloading rebound, ② the tensile crack surface of the steep rock mass extends to the deep part, and the plate beam bends, ③ the fracture effect occurs at the part with severe bending effect at the bottom of the plate beam, causing the plate beam to collapse. Generally, the deformation of the rear edge of the slope marks the formation of this type of landslide, so the deformation weight is Front2>Central2>Rear2.

[0113] Plastic flow tensile landslides mainly develop on weak base slopes. The underlying soft rock is squeezed out, causing the overlying rock layer to crack, disintegrate and sink unevenly. The main stages of deformation and damage development are: ① Unloading rebound causes steep cracks in the slope, ② Uneven subsidence of the overlying rock mass at the front edge of the slope causes the rock mass to crack, expand, collapse and collapse, ③ Tensile cracks expand to the rear edge of the slope, ④ The tensile crack surface at the rear edge turns to close, the potential slip surface is connected, and the deformation develops into a landslide. Therefore, the deformation weight Rear2>Front2>Central2.

[0114] Slip-bending landslides mainly develop in medium-steep outer layered slopes with a slightly larger inclination angle of the rock layer than the toe of the slope. The overlying rock and soil of this type of slope has the conditions for sliding down along the slip surface, but does not have the conditions for facing the free surface, resulting in obstruction of sliding, causing the toe of the slope to bear longitudinal compressive stress and produce bending deformation. The main stages of deformation and damage development are: ① Bedding tensile crack surfaces begin to appear in the bending part near the toe of the slope, accompanied by local crushing and loosening of the rock mass; ② The bending is significantly enhanced, the profile is dislocated, the sliding surface is developed, and the degree of deep deformation and damage increases; ③ The slip surface penetrates and forms a landslide. Therefore, the deformation weight Rear2>Central2>Front2.

[0115] Slip suppression tensile crack landslides mainly develop on gentle layered slopes with medium to steep slopes. The main stages of deformation and damage development are: ① The front edge of the slope is unloaded, resulting in an unloading rebound effect, and nearly vertical tensile cracks develop in the tensile stress concentration area near the locking point; ② The suppression tensile crack surface extends toward the slope surface and rotates slightly in a direction close to the main direction of the slope surface; ③ The deformation enters the progressive damage stage, and the rotation characteristics of the deformed body are obvious. The mosaic body at the steep corner is shortened, crushed, and expanded, and a through sliding surface is formed, and the slope is destroyed as a whole. Therefore, the deformation weight Rear2>Central2=Front2.

[0116] In a possible implementation, updating the second neural network model according to the first revised weight and the second revised weight includes:

[0117] The weights from the input layer to the hidden layer in the second neural network model are updated to the corresponding second revised weights, and the weights from the non-input layer to the hidden layer in the second neural network model are updated to the corresponding first revised weights.

[0118] When implementing the embodiment of the present application, a specific updating scheme of the second neural network model is provided, please refer to Figure 2 , shows the specific update method, in Figure 2 In the example of two hidden layers, each node in the input layer inputs different monitoring data, and different monitoring data will have corresponding motion weights and deformation weights. The second correction weight formed by correcting the first correction weight by motion weight and deformation weight can be assigned to the weight from the input layer to the first hidden layer, and the first correction weight can be assigned to the weight from the first hidden layer to the second hidden layer and the weight from the second hidden layer to the output layer. Furthermore, if the loss value output by the loss function is less than the preset value, stop correcting the first correction weight by motion weight and deformation weight, and directly assign the first correction weight to the corresponding position. This method can improve the accuracy of the second neural network model when it is close to the optimal solution.

[0119] Based on the same inventive concept, the present application also provides a landslide collaborative monitoring and early warning system for high-risk areas of geological disasters, including:

[0120] An acquisition unit is configured to acquire landslide surface data and landslide geological structure data of a target landslide; the landslide surface data is acquired by at least one of space-based satellite observation, drone observation, and surface three-dimensional laser observation; the landslide geological structure data is acquired by a geophysical exploration method;

[0121] A monitoring unit is configured to set a plurality of monitoring devices for the target landslide according to the landslide surface data and the landslide geological structure data, and divide the monitoring devices into front edge devices, middle section devices and rear edge devices according to their positions relative to the target landslide;

[0122] The analysis unit is configured to input the monitoring data at the current moment into a preset first neural network model, and obtain the landslide movement type data and the landslide deformation type data output by the first neural network model; the monitoring data includes leading edge monitoring data, middle section monitoring data and trailing edge monitoring data; the leading edge monitoring data is the displacement data monitored by the leading edge device, the middle section monitoring data is the displacement data monitored by the middle section device, and the trailing edge monitoring data is the displacement data monitored by the trailing edge device;

[0123] A calculation unit, configured to input the monitoring data at the current moment into a preset second neural network model, and obtain displacement data output by the second neural network model as predicted displacement data;

[0124] an updating unit, configured to calculate a first correction weight by a difference between the predicted displacement data and the monitoring data at the next moment, and to correct the first correction weight according to the landslide deformation type data and the landslide movement type data to form a second correction weight;

[0125] a circulation unit, configured to update the second neural network model according to the first revised weight and the second revised weight, and repeatedly obtain the monitoring data to update the second neural network model;

[0126] The judgment unit is configured to judge whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model.

[0127] In a possible implementation, the configuration of the monitoring device includes:

[0128] Dividing the target landslide into a front edge area, a middle area and a rear edge area according to the landslide surface data and the landslide geological structure data;

[0129] Selecting a suitable position on the landslide surface to deploy a slope displacement monitoring instrument according to the landslide surface data, and selecting a suitable position inside the landslide to deploy a deep displacement meter monitoring instrument according to the landslide geological structure data;

[0130] The monitoring instrument in the leading edge area is used as the leading edge device, the monitoring instrument in the middle section area is used as the middle section device, and the monitoring instrument in the trailing edge area is used as the trailing edge device.

[0131] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0132] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.

[0133] The units described as separate components may or may not be physically separated. As units, it is obvious that a person of ordinary skill in the art can realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0134] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0135] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a grid device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0136] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A collaborative landslide monitoring and early warning method for high-risk areas of geological disasters, characterized in that: include: Obtaining the landslide surface data and landslide geological structure data of the target landslide; The landslide surface data is obtained by at least one of space-based satellite observation, drone observation and surface three-dimensional laser observation; the landslide geological structure data is obtained by geophysical exploration method; According to the landslide surface data and the landslide geological structure data, a plurality of monitoring devices are set for the target landslide, and the monitoring devices are divided into front edge devices, middle section devices and rear edge devices according to their positions relative to the target landslide; Input the monitoring data at the current moment into a preset first neural network model, and obtain the landslide movement type data and landslide deformation type data output by the first neural network model; the monitoring data includes leading edge monitoring data, middle section monitoring data and trailing edge monitoring data; the leading edge monitoring data is the displacement data monitored by the leading edge device, the middle section monitoring data is the displacement data monitored by the middle section device, and the trailing edge monitoring data is the displacement data monitored by the trailing edge device; Inputting the monitoring data at the current moment into a preset second neural network model, and obtaining displacement data output by the second neural network model as predicted displacement data; Calculating a first correction weight by the difference between the predicted displacement data and the monitoring data at the next moment, and correcting the first correction weight according to the landslide deformation type data and the landslide movement type data to form a second correction weight; updating the second neural network model according to the first revised weight and the second revised weight, and repeatedly acquiring the monitoring data to update the second neural network model; Determining whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model; The calculation of the first modified weight and the second modified weight includes: Calculate the difference between the predicted displacement data and the monitoring data at the next moment as a loss value through a loss function; Calculate the gradient of each layer in the second neural network model by back propagation of the loss value, and calculate the weight of the second neural network model by the gradient as the first modified weight; Obtaining weights of the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data corresponding to the landslide deformation type data as deformation weights, and obtaining weights of the front edge monitoring data, the middle section monitoring data and the rear edge monitoring data corresponding to the landslide movement type data as movement weights; Multiplying the deformation weight and the motion weight by the first modified weight corresponding to the input layer to the hidden layer in the first modified weight to form the second modified weight; updating the second neural network model according to the first modified weight and the second modified weight comprises: The weights from the input layer to the hidden layer in the second neural network model are updated to the corresponding second revised weights, and the weights from the non-input layer to the hidden layer in the second neural network model are updated to the corresponding first revised weights.

2. The method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters according to claim 1 is characterized in that: The monitoring equipment includes: Dividing the target landslide into a front edge area, a middle area and a rear edge area according to the landslide surface data and the landslide geological structure data; Selecting a suitable position on the landslide surface to deploy a slope displacement monitoring instrument according to the landslide surface data, and selecting a suitable position inside the landslide to deploy a deep displacement meter monitoring instrument according to the landslide geological structure data; The monitoring instrument in the leading edge area is used as the leading edge device, the monitoring instrument in the middle section area is used as the middle section device, and the monitoring instrument in the trailing edge area is used as the trailing edge device.

3. The method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters according to claim 1 is characterized in that: The construction of the first neural network model includes: Constructing finite element models or discrete element models under various working conditions to form numerical simulation models; the constraint conditions corresponding to the working conditions include landslide movement type and landslide deformation type; the landslide movement types include push type, traction type and composite type; the landslide deformation types include creep cracking type, sliding cracking type, bending cracking type, plastic flow cracking type, sliding bending type and sliding compression cracking type; Performing deformation and destruction numerical simulation on a plurality of the numerical simulation models to obtain leading edge sample data, middle section sample data and trailing edge sample data; the leading edge sample data is the displacement data of the leading edge of the landslide in the numerical simulation model, the middle section sample data is the displacement data of the middle section of the landslide in the numerical simulation model, and the trailing edge sample data is the displacement data of the trailing edge of the landslide in the numerical simulation model; The leading edge sample data, the middle section sample data and the trailing edge sample data are used as input, and the landslide movement type and the landslide deformation type are used as output data to train a neural network model to form the first neural network model.

4. The method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters according to claim 3 is characterized in that: The construction of the second neural network model includes: The leading edge sample data, middle segment sample data and trailing edge sample data at the current moment are used as input data, and the leading edge sample data, middle segment sample data and trailing edge sample data at the next moment are used as output data to train a neural network model to form the second neural network model.

5. The method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters according to claim 1 is characterized in that: Judging whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model includes: Acquire the deformation weight and the motion weight of the monitoring device corresponding to the predicted displacement data; Performing weighted calculation on the predicted displacement data by using the deformation weight and the motion weight to obtain an early warning value of the target landslide; Whether the target landslide is safe is determined according to the warning value.

6. The method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters according to claim 1 is characterized in that: The types of landslide movement include push type, traction type and combined type; When the landslide movement type is push-type, the movement weight of the front edge monitoring data is greater than the movement weight of the middle section monitoring data, and the movement weight of the middle section monitoring data is greater than the movement weight of the rear edge monitoring data; When the landslide movement type is traction type, the movement weight of the trailing edge monitoring data is greater than the movement weight of the leading edge monitoring data, and the movement weight of the leading edge monitoring data is greater than the movement weight of the middle section monitoring data; When the landslide motion type is compound, the motion weight of the middle monitoring data is greater than the motion weight of the trailing edge monitoring data, and the motion weight of the leading edge monitoring data is equal to the motion weight of the trailing edge monitoring data.

7. The method for collaborative monitoring and early warning of landslides in high-risk areas of geological disasters according to claim 1 is characterized in that: The landslide deformation types include creep cracking, sliding cracking, bending cracking, plastic flow cracking, sliding bending and sliding compression cracking; When the landslide deformation type is the creep and tensile crack type, the deformation weight of the front edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the rear edge monitoring data; When the landslide deformation type is the sliding and pulling type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the leading edge monitoring data, and the deformation weight of the leading edge monitoring data is greater than the deformation weight of the middle section monitoring data; When the landslide deformation type is the bending and pulling type, the deformation weight of the front edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the rear edge monitoring data; When the landslide deformation type is the plastic flow tensile cracking type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the leading edge monitoring data, and the deformation weight of the leading edge monitoring data is greater than the deformation weight of the middle section monitoring data; When the landslide deformation type is the slip bending type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the middle section monitoring data is greater than the deformation weight of the leading edge monitoring data; When the landslide deformation type is the sliding suppression and tensile cracking type, the deformation weight of the trailing edge monitoring data is greater than the deformation weight of the middle section monitoring data, and the deformation weight of the leading edge monitoring data is equal to the deformation weight of the middle section monitoring data.

8. The landslide collaborative monitoring and early warning system in high-risk areas of geological disasters is characterized by: Using the method according to any one of claims 1 to 7, the system comprises: An acquisition unit is configured to acquire landslide surface data and landslide geological structure data of a target landslide; the landslide surface data is acquired by at least one of space-based satellite observation, drone observation, and surface three-dimensional laser observation; the landslide geological structure data is acquired by a geophysical exploration method; A monitoring unit is configured to set a plurality of monitoring devices for the target landslide according to the landslide surface data and the landslide geological structure data, and divide the monitoring devices into front edge devices, middle section devices and rear edge devices according to their positions relative to the target landslide; The analysis unit is configured to input the monitoring data at the current moment into a preset first neural network model, and obtain the landslide movement type data and the landslide deformation type data output by the first neural network model; the monitoring data includes leading edge monitoring data, middle section monitoring data and trailing edge monitoring data; the leading edge monitoring data is the displacement data monitored by the leading edge device, the middle section monitoring data is the displacement data monitored by the middle section device, and the trailing edge monitoring data is the displacement data monitored by the trailing edge device; A calculation unit, configured to input the monitoring data at the current moment into a preset second neural network model, and obtain the displacement data output by the second neural network model as the predicted displacement data; an updating unit, configured to calculate a first correction weight by a difference between the predicted displacement data and the monitoring data at the next moment, and to correct the first correction weight according to the landslide deformation type data and the landslide movement type data to form a second correction weight; a circulation unit, configured to update the second neural network model according to the first revised weight and the second revised weight, and repeatedly obtain the monitoring data to update the second neural network model; The judgment unit is configured to judge whether the target landslide is safe according to the predicted displacement data output by the updated second neural network model.

Images