A landslide identification method based on physical constraint machine learning
By embedding the physical constraint equations of landslides into neural networks, the problem of machine learning models lacking physical mechanism information in landslide prediction is solved, and accurate prediction and early warning of landslides are achieved, and the success rate of forecasting is improved.
Patent Information
- Application Number
- CN202310705803.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-14
- Publication Date
- 2024-12-31
- Estimated Expiration
- 2043-06-14
AI Technical Summary
In the existing landslide recognition technology, machine learning models are difficult to achieve accurate landslide prediction and early warning due to lack of sufficient data and physical mechanism information, and traditional methods cannot effectively monitor the rapid deformation process of landslides, resulting in a low success rate of forecasting.
Using a machine learning method based on physical constraints, the mathematical physical equations, boundary conditions and initial conditions related to landslides are embedded into the neural network, and the conservation of mass and momentum equations are learned to build a prediction model to achieve accurate prediction and early warning of landslides.
This method makes neural network physical constraints, avoids the complexity of meshing of traditional numerical methods, improves the accuracy and generalization ability of prediction models, and can achieve accurate prediction and early warning of landslides under a small amount of observation data, improving Predict the success rate of forecasts.
Smart Images
Figure CN116738844B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of landslide identification and early warning, and in particular relates to a landslide identification method based on physical constraint machine learning. Background Art
[0002] Landslides are natural hazards that can cause severe damage to human settlements and infrastructure, resulting in loss of life and economic losses. Understanding and predicting landslides is one of the frontiers of engineering geology and an important task in mitigating their impacts. Machine learning (ML) is a powerful tool that can be used to analyze landslide data, identify landslides, and predict them. In recent years, the majority of applications of machine learning in landslide research published in the literature fall into one of three broad categories: landslide identification, spatial prediction of landslide susceptibility, and prediction of landslides in the temporal dimension. One way in which machine learning is used in landslide research is to predict the occurrence of landslides by analyzing geospatial and geological data, such as slope, elevation, land cover, rainfall patterns, and drainage density to estimate landslide susceptibility and risk. Another way in which machine learning is used in landslide research is to analyze the dynamic processes of landslides, including their initiation, propagation, and stability. For example, machine learning has been used to analyze the deformation and movement of landslides based on satellite imagery and remote sensing data.
[0003] In machine learning, the data collection methods used in landslide research mainly include satellite-based synthetic aperture radar differential interferometry technology, satellite optical images, drone photogrammetry, laser radar measurement technology (LiDAR) global navigation satellite system, etc. In order to form a complete landslide observation system, it is usually necessary to integrate multiple technologies to build a multi-dimensional monitoring system based on satellite platforms, InSAR and high-resolution optical images, drone photogrammetry and airborne LiDAR technology based on aviation platforms, and slope surface and internal monitoring perception based on ground platforms, so as to achieve multi-level, multi-angle, and multi-means all-weather monitoring of major landslide hazards, that is, a sky-air-ground coordinated monitoring system. However, due to the long revisit cycle of satellites or drones and the single monitoring indicators, it is impossible to conduct real-time automatic monitoring of the rapid deformation process of landslides and the related physical quantities that induce landslide deformation in high-frequency time series. It is difficult to achieve short-term and emergency monitoring and early warning of landslides by relying solely on "sky-air" monitoring technology. Therefore, ground and soil monitoring is also required. At the same time, landslides are a natural disaster that occurs infrequently and is often localized, so obtaining sufficient data is a challenge. In addition, due to the lack of data, the trained machine learning model may overfit the existing data, resulting in an inability to accurately predict unknown situations. Moreover, due to the scarcity of data and the complexity of features in landslide research, the generalization ability of machine learning models is often limited, which may cause the performance of the model in new scenarios to decline to a certain extent. Many factors and variables in landslide research are uncertain, which will also bring certain difficulties to the modeling of machine learning algorithms.
[0004] In summary, although landslide identification research has benefited a lot from the progress of machine learning, temporal and spatial prediction still does not seem to have gained obvious advantages from machine learning (ML) algorithms. This is mainly because landslide detection is essentially a computer vision (CV) problem, which does not require much physical understanding. In terms of landslide spatiotemporal prediction, more information on physical mechanisms is needed, because landslide development is the result of long-term evolution of internal and external dynamics in the slope geological body. Scientific understanding of landslide evolution process and physical mechanism is the key to achieving accurate landslide prediction. However, most landslide machine learning algorithms are data-driven and out of touch with landslide evolution process and physical and mechanical mechanisms, which makes these prediction models encounter huge theoretical and technical bottlenecks, resulting in slow progress in theoretical research on landslide prediction and low success rate of landslide prediction. Summary of the invention
[0005] In view of the above shortcomings, the present invention discloses a landslide identification method based on physical constraint machine learning, which adds physical constraints such as mathematical physics equations, boundary conditions, initial conditions and so on related to landslides into the neural network of machine learning, scientifically understands the physical mechanism and evolution process of landslides, realizes accurate prediction and early warning of landslides, and solves the problem that landslides cannot be accurately predicted through the network when the amount of observation data is small.
[0006] The present invention is achieved by adopting the following technical solutions:
[0007] A landslide identification method based on physical constraint machine learning includes the following steps:
[0008] S1. Collecting slope data and information and saving them for future use. The slope data and information include but are not limited to the location information of the slope, the rock mass parameters of the slope, the average thickness of the landslide body, the speed of the landslide body, the time for the landslide body to move along the landslide surface, and the distance for the landslide body to move along the landslide surface;
[0009] S2, using the data collected in step S1, and embedding physical constraints into a neural network for learning to obtain a prediction model, wherein the physical constraints include conservation of mass and momentum, boundary conditions, and initial conditions in the erosive landslide flow process;
[0010] The conservation of mass and momentum in the erosional landslide flow process includes the following formulas:
[0011] Formula (1):
[0012] Formula (2):
[0013] Formula (3): Wherein, formula (1) represents the mass conservation equation, and formula (2) represents the momentum conservation equation. Here, h and u represent the average thickness and velocity of the landslide body, respectively; t and x represent the time and distance of the landslide body moving along the landslide surface, respectively; E represents the erosion rate, and γ m is the density ratio between the fluid and the particles, α m is the volume fraction of solid particles, μ m is the friction coefficient, g x and g z They represent the gravitational acceleration along and perpendicular to the landslide surface, respectively, K is the earth pressure coefficient, C DV is the viscous drag coefficient, b represents the surface of the eroded substrate, which is a function of time and space, and λ b is a parameter related to erosion velocity, i.e., erosion drift, and the superscripts m and b denote quantities related to the landslide mass and base, respectively;
[0014] The schematic diagram of the algorithm for embedding the dynamic model of erosive landslide mobility into a neural network is shown in Figure 1 As shown;
[0015] The initial conditions include the following formula:
[0016] Formula (4): u| t＝0 =0
[0017] Formula (5): h| t＝0 =h0
[0018] The boundary conditions include the following formula:
[0019] Formula (6):
[0020] Formula (7): h| m =0
[0021] S3, collecting slope data information of the slope to be measured, and then using the prediction model obtained in step S2 to perform prediction analysis on the slope to be measured, so as to achieve landslide identification and early warning of the slope to be measured.
[0022] For further optimization, in step S2, the conservation of mass and momentum in the erosive landslide flow process is embedded into the neural network, that is, formula (1) to formula (7) are added to the loss function, and the gradient descent method is used to gradually reduce the value of the loss function. Through continuous training, the neural network finally satisfies the equation. Since the speed may not satisfy formula (1) and formula (2) during the random initialization stage of the neural network, the gradient descent method is used to gradually reduce the value of the loss function. Through continuous training, the neural network finally satisfies the equation.
[0023] For further optimization, the gradient descent method specifically selects an appropriate initial value u(0), continuously iterates, updates the value of u, and minimizes the objective function until convergence. The gradient descent method is as follows:
[0024] Formula (8): Where α represents the learning rate and J represents the loss function.
[0025] For further optimization, based on the comparison between experimental data and simulation results, the prediction model is modified and an accurate early warning model is obtained.
[0026] Compared with the prior art, this technical solution has the following beneficial effects:
[0027] This method gives the neural network physical constraints. For traditional numerical methods, such as the finite element method and the finite difference method, it is necessary to mesh the calculation area. A large amount of meshing is time-consuming, and when the structure of the area is complex, the generation of the mesh is also difficult. This method does not require meshing and is a gridless method. In addition, using neural networks, the derivative part of the equation is very convenient because neural networks can perform back propagation or automatic differentiation. This method can also add other physical information to make the physical mechanism of the neural network more complete. As people gain a better understanding of the physical mechanism of landslides, the equations describing landslides will continue to improve, allowing neural networks to keep up with the development of the physical mechanism of landslides. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 It is a schematic diagram of an algorithm for embedding the dynamic model of erosive landslide fluidity into a neural network. DETAILED DESCRIPTION
[0029] The present invention is further described by the following examples, but is not intended to be limiting of the present invention. Specific experimental conditions and methods not specified in the following examples are conventional methods commonly known to those skilled in the art.
[0030] Embodiment 1: A landslide identification method based on physical constraint machine learning, comprising the following steps:
[0031] S1. Collecting slope data and information and saving them for future use. The slope data and information include but are not limited to the location information of the slope, the rock mass parameters of the slope, the average thickness of the landslide body, the speed of the landslide body, the time for the landslide body to move along the landslide surface, and the distance for the landslide body to move along the landslide surface;
[0032] S2, using the data collected in step S1, and embedding physical constraints into a neural network for learning to obtain a prediction model, wherein the physical constraints include conservation of mass and momentum, boundary conditions, and initial conditions in the erosive landslide flow process;
[0033] The conservation of mass and momentum in the erosional landslide flow process includes the following formulas:
[0034] Formula (1):
[0035] Formula (2):
[0036] Formula (3): Wherein, formula (1) represents the mass conservation equation, and formula (2) represents the momentum conservation equation. Here, h and u represent the average thickness and velocity of the landslide body, respectively; t and x represent the time and distance of the landslide body moving along the landslide surface, respectively; E represents the erosion rate, and γ m is the density ratio between the fluid and the particles, α m is the volume fraction of solid particles, μ m is the friction coefficient, g x and g z They represent the gravitational acceleration along and perpendicular to the landslide surface, respectively, K is the earth pressure coefficient, C DV is the viscous drag coefficient, b represents the surface of the eroded substrate, which is a function of time and space, and λ b is a parameter related to erosion velocity, i.e., erosion drift, and the superscripts m and b denote quantities related to the landslide mass and base, respectively;
[0037] The schematic diagram of the algorithm for embedding the dynamic model of erosive landslide mobility into a neural network is shown in Figure 1 As shown;
[0038] The initial conditions include the following formula:
[0039] Formula (4): u| t＝0 =0
[0040] Formula (5): g| t＝0 =h0
[0041] The boundary conditions include the following formula:
[0042] Formula (6):
[0043] Formula (7): h| m =0
[0044] S3, collecting slope data information of the slope to be measured, and then using the prediction model obtained in step S2 to perform prediction analysis on the slope to be measured, so as to realize landslide identification and early warning of the slope to be measured; at the same time, according to the comparison between the experimental data and the simulation results, the prediction model can be corrected and an accurate early warning model can be obtained.
[0045] In such Figure 1 The algorithm diagram shown in the figure shows that the NN part represents the forward neural network, x and z represent the spatial coordinates, and u1~u4 can represent the unknown speed. In order to solve these unknowns, training is required and a loss function needs to be constructed. If some clear data information can be collected, such as Figure 1 u 1,target, then we can use the commonly used loss function, such as the root mean square error (MSE). In the random initialization stage of the neural network, the velocity of the landslide may not satisfy formula (1) and formula (2). Formula (1) to formula (7) are added to the loss function, and the gradient descent method is used to gradually reduce the value of the loss function. Through continuous training, the neural network will eventually satisfy the equation; the gradient descent method specifically selects an appropriate initial value u(0), continuously iterates, updates the value of u, and minimizes the objective function until convergence. The gradient descent method is as follows:
[0046] Formula (8): Where α represents the learning rate and J represents the loss function.
[0047] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.
Claims
1. A landslide identification method based on physical constraint machine learning, characterized by: The following steps are involved: S1. Collecting slope data and information and saving them for future use. The slope data and information include but are not limited to the location information of the slope, rock mass parameters of the slope, average thickness of the landslide body, speed of the landslide body, time for the landslide body to move along the landslide surface, distance for the landslide body to move along the landslide surface, density ratio between fluid and particles, volume fraction of solid particles, friction coefficient, earth pressure coefficient and viscous resistance coefficient; S2, using the data collected in step S1, and embedding physical constraints into a neural network for learning to obtain a prediction model, wherein the physical constraints include conservation of mass and momentum, boundary conditions, and initial conditions in the erosive landslide flow process; The conservation of mass and momentum in the erosional landslide flow process includes the following formulas: Formula (1): Formula (2): Formula (3): Wherein, formula (1) represents the mass conservation equation, and formula (2) represents the momentum conservation equation. Here, h and u represent the average thickness and velocity of the landslide body, respectively; t and x represent the time and distance of the landslide body moving along the landslide surface, respectively; E represents the erosion rate, and γ m is the density ratio between the fluid and the particles, α m is the volume fraction of solid particles, μ m is the friction coefficient, g x and g z They represent the gravitational acceleration along and perpendicular to the landslide surface, respectively, K is the earth pressure coefficient, C DV is the viscous drag coefficient, b represents the surface of the eroded substrate, which is a function of time and space, and λ b is a parameter related to erosion velocity, i.e., erosion drift, and the superscripts m and b denote quantities related to the landslide mass and base, respectively; embedding the physical constraints into a neural network; The initial conditions include the following formula: Formula (4): u| t＝0 =0 Formula (5): h| t＝0 =h0 The boundary conditions include the following formula: Formula (6): Formula (7): h| m =0 In step S2, the conservation of mass and momentum in the process of erosive landslide flow are embedded into the neural network, that is, formula (1) to formula (7) are added to the loss function, and the value of the loss function is gradually reduced by using the gradient descent method. Through continuous training, the neural network finally satisfies the equation; the gradient descent method specifically selects an appropriate initial value u(0), continuously iterates, updates the value of u, and minimizes the objective function until convergence. The gradient descent method is as follows: Formula (8): Where α represents the learning rate and J represents the loss function; S3, collecting slope data information of the slope to be measured, and then using the prediction model obtained in step S2 to perform prediction analysis on the slope to be measured, so as to achieve landslide identification and early warning of the slope to be measured.
2. The landslide identification method based on physical constraint machine learning according to claim 1 is characterized in that: According to the comparison between the experimental data and the simulation results, the prediction model is modified and an accurate early warning model is obtained.