CN-122022472-A - Disaster area assessment method, system, equipment and medium based on geological disaster investigation

CN122022472ACN 122022472 ACN122022472 ACN 122022472ACN-122022472-A

Abstract

The application relates to a disaster area assessment method, system, equipment and medium based on geological disaster investigation. The method comprises the steps of analyzing multi-source investigation data to obtain unit space-time characteristic tuples and physical incidence matrixes, conducting graphic neural network calculation processing based on the unit space-time characteristic tuples and the physical incidence matrixes to obtain dynamic risk assessment results, multi-level attention weights and physical state deviation amounts, and conducting report generation processing based on the dynamic risk assessment results, the multi-level attention weights and the physical state deviation amounts to obtain a geological disaster dynamic risk assessment report. According to the method, through means of multi-source data analysis, graphic neural network calculation and the like, accuracy, dynamic performance and interpretability of geological disaster assessment are improved, and accurate disaster risk prevention and control and decision support capability is enhanced.

Inventors

DANG CHAO
TANG HU
XU BAIPING

Assignees

四川省第五地质大队

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (9)

1. A disaster area assessment method based on geological disaster investigation, the method comprising: Analyzing the multi-source investigation data to obtain a unit space-time characteristic tuple and a physical association matrix; Based on the unit space-time characteristic tuple and the physical association matrix, performing graph neural network calculation processing to obtain a dynamic risk assessment result, a multi-level attention weight and a physical state deviation amount; And carrying out report generation processing based on the dynamic risk assessment result, the multi-stage attention weight and the physical state deviation amount to obtain a geological disaster dynamic risk assessment report.
2. The method of claim 1, wherein the parsing the multi-source survey data to obtain a unit space-time feature tuple and a physical correlation matrix comprises: Performing space-time gridding alignment processing on remote sensing data, topographic data, geological data, meteorological monitoring data and engineering monitoring data in the multisource investigation data to obtain unified multisource original data, separating static features and dynamic features based on the unified multisource original data, and performing standardized extraction processing on the static features to obtain a static background feature matrix of each space unit; based on a preset seepage mechanics equation, performing physical analysis processing on time sequence rainfall data in the dynamic observation feature tensor to obtain a hydrological physical feature tensor of each space unit; Extracting inter-unit runoff flux from the hydrologic physical feature tensor, combining geological structure similarity information in the topographic data, performing physical association degree calculation on the inter-unit runoff flux to obtain a physical association degree, and constructing the physical association matrix; based on a signal decomposition model, performing physical analysis processing on time sequence deformation data in the dynamic observation feature tensor to obtain deformation physical feature tensors of each space unit, wherein the deformation physical feature tensors comprise deformation residual error components; and performing characteristic dimension splicing processing on the static background characteristic matrix, the hydrologic physical characteristic tensor and the deformation physical characteristic tensor to obtain the unit space-time characteristic tuple.
3. The method of claim 1, wherein the unit spatiotemporal feature tuples include timing information for a plurality of time steps, each of the time steps including a static background feature matrix, a hydrologic physical feature tensor, and a morphing physical feature tensor; and performing graph neural network calculation processing based on the unit space-time characteristic tuple and the physical association matrix to obtain a dynamic risk assessment result, a multi-level attention weight and a physical state deviation amount, wherein the method comprises the following steps of: Traversing each time step of the unit space-time characteristic tuple, constructing an adjacent relation of a graph structure corresponding to each time step based on the physical correlation matrix, and carrying out characteristic initialization processing on each node in the graph structure corresponding to the unit space-time characteristic tuple of the current time step to obtain initialized node characteristics corresponding to each time step and a corresponding initialized graph structure; carrying out multi-layer physical constraint graph convolution processing on the initialized graph structure corresponding to each time step to obtain the optimized node characteristics of the final layer output, wherein the multi-layer physical constraint graph convolution processing corresponds to the following steps: Message aggregation processing is carried out on the current node characteristics and the neighborhood node characteristics in the initialized graph structure in the current layer to obtain aggregation characteristics, wherein the current node characteristics are the initialized node characteristics during the first layer processing; constructing a physical checker based on a preset physical equation, and performing physical consistency check on the intermediate node characteristic through the physical checker to obtain the physical state deviation of the current layer; Performing weighted fusion processing on the physical state deviation and the intermediate node characteristic to obtain an updated node characteristic, and repeatedly performing physical constraint graph convolution processing by taking the updated node characteristic as the current node characteristic of the next layer until each layer performs physical constraint graph convolution processing to obtain an optimized node characteristic corresponding to the final layer output of the time step; performing factor attention weight calculation processing based on the optimized node characteristics output by the final layer in each time step and the physical state deviation amount of each layer to obtain a factor attention weight vector corresponding to each node in each time step; In the process of aggregation processing of each layer of information corresponding to each time step, based on the current node characteristics of the current layer and the deformation physical characteristic tensor corresponding to the time step, carrying out space association degree calculation processing to obtain the space association degree of the current layer; Performing time sequence feature fusion processing on the optimized node features of the final layer corresponding to each time step, extracting a historical hydrologic state from the hydrologic physical feature tensor of each time step in the time sequence feature fusion processing, and performing time attention weight calculation processing on the basis of the historical hydrologic state and the optimized node features output by the final layer corresponding to the time step to obtain time attention weight corresponding to each time step; performing risk probability mapping processing on the optimized node characteristics of the final layer corresponding to each time step, and generating the dynamic risk assessment result corresponding to each time step; And in each time step, integrating the corresponding factor attention weight vector, the time attention weight and the dynamic space attention weight of each layer to obtain the multi-stage attention weight, and summarizing the physical state deviation amounts of each layer corresponding to each time step to form the physical state deviation amounts.
4. The method of claim 3, wherein the performing report generation processing based on the dynamic risk assessment result, the multi-level attention weight, and the physical state deviation amount to obtain a geological disaster dynamic risk assessment report includes: Acquiring and comparing threshold values of the dynamic risk assessment results corresponding to the time steps according to preset risk threshold values, and identifying high risk units; Extracting the factor attention weight vector corresponding to the time step for each high-risk unit, and carrying out dominant factor contribution quantitative analysis processing by combining the feature dimension of the time sequence information corresponding to the time step in the unit space-time feature tuple to obtain a dominant factor contribution analysis chart; Extracting the dynamic space attention weight corresponding to the time step for each high risk unit, and performing visualization processing on a space influence range and space influence intensity to obtain a space influence associated thermodynamic diagram; extracting the time attention weight of the high risk unit corresponding to the time step, and carrying out key influence moment quantitative analysis processing to obtain a key influence moment graph; extracting the physical state deviation amount corresponding to the time step for each high-risk unit, and carrying out physical consistency deviation analysis processing by combining a preset physical constraint standard to obtain a physical state consistency analysis chart; and carrying out structural integration on the dynamic risk assessment result, the high risk unit, the dominant factor contribution analysis chart, the spatial influence correlation thermodynamic diagram, the key influence moment curve graph and the physical state consistency analysis chart corresponding to each time step to obtain the geological disaster dynamic risk assessment report.
5. The method according to claim 2, wherein the performing physical analysis processing on the time-series rainfall data in the dynamic observation feature tensor based on the seepage mechanics equation to obtain a hydrographic physical feature tensor of each space unit includes: extracting the time sequence rainfall data corresponding to each time step from the dynamic observation characteristic tensor, and performing rainfall intensity standardization processing on the time sequence rainfall data to obtain a rainfall intensity sequence of unit area corresponding to each time step; determining soil hydraulic parameters of each space unit based on soil type data in the geological data and combining rock stratum distribution data and porosity data in the geological data, wherein the soil hydraulic parameters comprise saturated water conductivity, air intake suction and pore size distribution indexes; The preset seepage mechanics equation is used as a core control equation, the rainfall intensity sequence in unit area corresponding to each time step is used as an infiltration boundary condition, the soil hydraulic parameter is used as an equation initial parameter, and a one-dimensional vertical seepage numerical model of each space unit is constructed; Discretizing the one-dimensional vertical seepage numerical model based on a finite volume method to obtain a discretized seepage numerical model, and solving the discretized seepage numerical model through an implicit differential format to obtain soil water content distribution data and pore water pressure distribution data in the space unit corresponding to each time step; Performing feature extraction processing on the soil water content distribution data and the pore water pressure distribution data corresponding to each time step to obtain a soil water content mean value, a pore water pressure peak value and a seepage velocity vector of each space unit; Calculating the hydraulic gradient between adjacent space units based on the seepage velocity vector of each space unit and the pore water pressure distribution data of the adjacent space units, and carrying out inter-unit runoff calculation processing through Darcy's law by combining the saturated water conductivity in the soil hydraulic parameters to obtain inter-unit runoff corresponding to each time step; And carrying out time sequence dimension integration on the soil water content average value, the pore water pressure peak value, the seepage velocity vector and the inter-unit runoff flux corresponding to each time step, and constructing the hydrologic physical characteristic tensor of each space unit.
6. The method of claim 2, wherein the physical association is calculated by the following formula: Wherein, the Is the first Time step space unit The value range of the physical association degree [0,1]; is the first Time step space unit And (3) with Is a radial flux of (2); is the first A maximum value of runoff flux between each of the space units in time steps; is a space unit And (3) with The value range [0,1]; is a space unit And (3) with Is a spatial euclidean distance; is the distance attenuation coefficient; is the first Time step space unit Pore water pressure peaks of (2); 、 And Are all weight coefficients, satisfy And the contribution weights of the runoff flux, the geological similarity and the pore water pressure are respectively corresponding.
7. A disaster area assessment system based on geological disaster investigation, the system comprising: the data analysis module is used for analyzing the multi-source investigation data to obtain a unit space-time characteristic tuple and a physical association matrix; the graphic neural network analysis module is used for carrying out graphic neural network calculation processing based on the unit space-time characteristic tuple and the physical association matrix to obtain a dynamic risk assessment result, a multi-level attention weight and a physical state deviation; and the report generation module is used for carrying out report generation processing based on the dynamic risk assessment result, the multi-stage attention weight and the physical state deviation amount to obtain a geological disaster dynamic risk assessment report.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.

Description

Disaster area assessment method, system, equipment and medium based on geological disaster investigation Technical Field The invention belongs to the technical field of computers, and particularly relates to a disaster area assessment method, system, equipment and medium based on geological disaster investigation. Background The geological disaster assessment is a key technical means for guaranteeing the safety of geological environment and reducing disaster loss, and along with the development of investigation technology, multi-source investigation data such as remote sensing data, topographic and geological data, meteorological monitoring data and engineering monitoring data are widely applied to the field of geological disaster assessment. However, the existing geological disaster evaluation method generally relies on a single data source or a traditional physical model to perform risk analysis, has the problems of insufficient multi-source data integration and nonsystem space-time feature extraction, is difficult to form a unit space-time feature tuple which comprehensively reflects a disaster evolution rule, and lacks quantitative characterization on physical association relations among multi-source data, so that the physical association matrix is constructed and is lack, and the accuracy of an evaluation result is affected. Furthermore, the conventional evaluation method often adopts a traditional statistical model or a simple data driving model, and can not fully combine the space association modeling capability of the graph neural network with the constraint action of a physical mechanism, so that the dynamic evolution characteristics of geological disaster risks can not be effectively captured, and high-precision dynamic risk evaluation results are difficult to generate. In addition, the existing assessment report only outputs a risk level result, and the deep analysis of risk causes is lacking, so that the interpretation and practicability of the assessment report are insufficient, and the requirements of accurate disaster risk prevention and control and decision support in engineering practice are difficult to meet. Disclosure of Invention Based on the above, it is necessary to provide a disaster area assessment method, system, device and medium based on geological disaster investigation, aiming at improving accuracy of assessment results and dynamic risk assessment precision and enhancing dynamic monitoring and early warning capability of disaster risks. In a first aspect, the present application provides a disaster area assessment method based on geological disaster investigation, comprising: Analyzing the multi-source investigation data to obtain a unit space-time characteristic tuple and a physical association matrix; based on the unit space-time characteristic tuple and the physical association matrix, performing graph neural network calculation processing to obtain a dynamic risk assessment result, a multi-level attention weight and a physical state deviation amount; and carrying out report generation processing based on the dynamic risk assessment result, the multi-level attention weight and the physical state deviation amount to obtain a geological disaster dynamic risk assessment report. In one embodiment, the parsing of the multi-source survey data to obtain unit spatio-temporal feature tuples and physical correlation matrices includes: Performing space-time gridding alignment processing on remote sensing data, topographic data, geological data, meteorological monitoring data and engineering monitoring data in the multisource investigation data to obtain unified multisource original data, separating static features and dynamic features based on the unified multisource original data, and performing standardized extraction processing on the static features to obtain a static background feature matrix of each space unit; based on a preset seepage mechanics equation, performing physical analysis processing on time sequence rainfall data in the dynamic observation feature tensor to obtain a hydrological physical feature tensor of each space unit; Extracting inter-unit runoff flux from the hydrologic physical feature tensor, carrying out physical association degree calculation processing on the inter-unit runoff flux by combining with geological structure similarity information in the topographic data to obtain a physical association degree, and constructing a physical association matrix; Based on the signal decomposition model, performing physical analysis processing on time sequence deformation data in the dynamic observation feature tensor to obtain deformation physical feature tensor of each space unit, wherein the deformation physical feature tensor comprises deformation residual error components; and performing feature dimension splicing processing on the static background feature matrix, the hydrologic physical feature tensor and the deformation physical feature tensor to obtain the un