CN-121390922-B - Mineral resource overburden risk assessment method and system

CN121390922BCN 121390922 BCN121390922 BCN 121390922BCN-121390922-B

Abstract

The invention provides a mineral resource overburden risk assessment method and system, which comprise the steps of obtaining multi-source heterogeneous geological data, carrying out deep fusion on the multi-source heterogeneous geological data by adopting a robust neural network training method controlled by game theory gradient to obtain a fusion data set, carrying out deep mining and reasoning on high-dimensional geological data by utilizing the mode identification and association analysis capability of a large model and combining manifold perception regularization technology to identify geological structure characteristics, establishing a dynamic safe mining depth assessment model by utilizing a fractional peak differential equation neural network with efficient accompanying parameter training, integrating geological structure risks, mining disturbance effects and engineering safety thresholds, calculating safe mining depth intervals and risk probabilities under different confidence degrees, and constructing an assessment report generation module comprising core risk point analysis, key technical evidence and prospective risk early warning to generate an intelligent overburden risk assessment report. The invention can improve the accuracy and reliability of the overburden risk assessment.

Inventors

WANG ZHENGJIN
AO XUANFENG
Luo Yongshuang
Yao Dunting
BAI LANG

Assignees

贵州天翼恒盛科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (6)

1. A mineral resource overburden risk assessment method, comprising the steps of: The method comprises the steps of taking multi-source heterogeneous geological data as game participants, adopting Nash equilibrium theory, determining contribution weights of all data sources to a final fusion result through iterative solution to obtain initial weight distribution, constructing a robust neural network architecture comprising residual connection and batch normalization based on the initial weight distribution, introducing Gaussian noise and data enhancement strategies to improve network anti-interference capability, calculating optimal contribution weights of all data sources in Nash equilibrium state to obtain optimal weight distribution state, carrying out weighted linear combination on all data sources of the multi-source heterogeneous geological data according to the optimal weight distribution state, and completing fusion training through forward propagation and reverse propagation of a neural network to obtain a fusion data set; Based on the fusion data set, carrying out deep mining and reasoning on high-dimensional geological data by utilizing the mode recognition and association analysis capability of a large model and combining manifold perception regularization technology, and recognizing geological structure characteristics including stratum interfaces, lithology distribution, fault fracture zones and joint cracks; based on the geologic structure characteristics, a dynamic safety mining depth evaluation model is established by using a fractional order peak differential equation neural network with efficient accompanying parameter training, and the risk, mining disturbance effect and engineering safety threshold are synthesized, so that the safety mining depth interval and risk probability under different confidence degrees are calculated; and constructing an evaluation report generation module comprising core risk point analysis, key technology demonstration and prospective risk early warning according to the safe mining depth interval and the risk probability, and automatically generating an intelligent override risk evaluation report.
2. The method of claim 1, wherein the constructing a robust neural network architecture including residual connection and batch normalization comprises: Based on the initial weight distribution, multi-source heterogeneous geological data are used as a neural network input layer, a residual error connection module is arranged in the neural network, and a batch normalization technology is adopted to perform standardized processing on each layer of output so as to ensure stable training of a depth network; And introducing Gaussian noise disturbance and a data enhancement strategy to the neural network input layer, wherein the Gaussian noise disturbance and the data enhancement strategy comprise random rotation, scaling and translation transformation, and the robustness of the network to the data disturbance is enhanced.
3. The method of claim 1, wherein the depth mining and reasoning of the high-dimensional geological data in combination with manifold-aware regularization techniques identifies geologic formation features including formation interfaces, lithology distributions, fault fracture zones, joint fissures, comprising: Based on the geological feature points in the fusion data set, constructing a k-nearest neighbor graph, calculating k nearest neighbors of each feature point, and establishing an adjacency relation matrix between the feature points to obtain manifold structure representation of the data; Based on manifold structure representation of the data, calculating local density and curvature information of feature points in manifold space, constructing a manifold Laplacian matrix as a regularization term, and maintaining the internal geometry of the data in the training process of the constraint large model to obtain regularization constraint parameters; and adding the regularization constraint parameters into a large model loss function based on a transducer architecture, optimizing model parameters through a multi-head self-attention mechanism and a counter propagation algorithm, and identifying the spatial distribution of stratum interfaces, lithology distribution, fault fracture zones and joint cracks to obtain the geologic structure characteristics.
4. The method of claim 1, wherein the using a fractional order spike differential equation neural network with efficient accompanying parameter training to build a dynamic safe mining depth evaluation model comprises: Describing the memory effect and long-range correlation of the geological material through a fractional differential equation based on the geological structure characteristics, designing a spike neural network by adopting a leakage integration release model, and establishing a dynamic safe mining depth evaluation model; And establishing a multi-constraint optimization model based on the dynamic safety mining depth evaluation model, wherein constraint conditions comprise that the maximum subsidence amount, the maximum inclination and the maximum curvature deformation of the earth surface do not exceed safety thresholds, and solving constraint optimization problems by adopting a particle swarm optimization algorithm to obtain the safety mining depth interval and the risk probability.
5. The method of claim 1, wherein the constructing an assessment report generation module comprising core risk point analysis, key technical demonstration, and prospective risk early warning comprises: Based on the safe exploitation depth interval and the risk probability, dividing an evaluation area into risk areas with different grades by adopting cluster analysis, wherein the risk areas comprise a high risk area, a medium risk area and a low risk area, extracting geological parameters, calculation model parameters and evaluation results of each risk area, and forming a risk source file; And according to the risk source file, adopting a natural language generation technology, combining an expert knowledge base and an engineering case base to match a corresponding analysis method and a corresponding demonstration logic for each risk type, and automatically generating the intelligent override risk assessment report.
6. A mineral resource overburden risk assessment system comprising: The multi-source data fusion module is used for acquiring multi-source heterogeneous geological data, including geological exploration data, drilling core data, geophysical exploration data, historical evaluation cases and peripheral mine exploitation data, carrying out deep fusion on the multi-source heterogeneous geological data by adopting a robust neural network training method controlled by a game theory gradient to obtain a fusion data set, wherein the multi-source heterogeneous geological data is used as a game participant, a Nash equilibrium theory is adopted, each data source reaches an optimal weight distribution state through iterative solution, the contribution weight of each data source to a final fusion result is determined, initial weight distribution is obtained; The geological structure identification module is used for carrying out deep mining and reasoning on the high-dimensional geological data by utilizing the mode identification and association analysis capability of the large model and combining manifold perception regularization technology based on the fusion data set, and identifying geological structure characteristics including stratum interfaces, lithology distribution, fault fracture zones and joint cracks; the safe mining depth calculation module is used for establishing a dynamic safe mining depth evaluation model by using a fractional order peak differential equation neural network with efficient accompanying parameter training, synthesizing geological structure risks, mining disturbance effects and engineering safety thresholds, and calculating safe mining depth intervals and risk probabilities under different confidence degrees; and the report generation module is used for automatically generating an intelligent coverage risk assessment report comprising core risk point analysis, key technical demonstration and prospective risk early warning according to the safe mining depth interval and the risk probability.

Description

Mineral resource overburden risk assessment method and system Technical Field The invention relates to the field of mineral resource development, in particular to a mineral resource coverage risk assessment method and a mineral resource coverage risk assessment system, which are used for assessing the degree of interaction between ground construction engineering and underground mineral resources, determining the safety exploitation depth and risk level and providing scientific basis for mineral resource protection and engineering safety. Background The mineral resource overburden risk assessment is an important technical field for ensuring the safety of ground construction engineering and reasonable development of underground mineral resources, and relates to the intersection of multiple disciplines such as geological engineering, mining engineering, geotechnical engineering and the like. With the continuous improvement of national mineral resource protection and construction engineering safety requirements, the scientificity and the accuracy of the overburden risk assessment technology are increasingly emphasized. Traditional overburden risk assessment methods rely primarily on three-dimensional geologic modeling techniques and empirical formula calculations. Typical evaluation procedures include geologic data collection, three-dimensional geologic model construction, safe mining depth calculation, and risk classification. Another common method is finite element analysis based on numerical modeling to evaluate the extent of influence of mining on surface structures by building a geomechanical model. The most mainstream method in the prior art is to construct a three-dimensional geological model by adopting a statistical method such as Kriging interpolation, then calculate the earth surface subsidence by combining a probability integration method or an influence function method, and finally determine the safe mining depth according to the standard. According to the method, discrete drilling data are subjected to spatial interpolation, a continuous geological interface is established, and the ground surface deformation is calculated by applying an empirical formula based on rock mechanical parameters and mining geometric parameters. However, the prior art has two key problems that firstly, the geological risk identification depth is insufficient, complex correlations among multi-source data are difficult to fully mine by the traditional modeling method, the influence mechanism of geological structures such as faults, joints and the like on the extraction stability cannot be deeply revealed, secondly, the safety extraction depth demonstration is excessively dependent on an empirical formula, and comprehensive consideration of geological condition complexity and extraction disturbance dynamics is lacked, so that the accuracy and reliability of an evaluation conclusion are limited. Therefore, there is a need for an intelligent overburden risk assessment method and system that can deeply fuse multi-source heterogeneous geological data, accurately identify complex geologic structure features, and dynamically assess safe mining depth. Disclosure of Invention The invention aims to provide a mineral resource overburden risk assessment method and system, which improve the accuracy and reliability of overburden risk assessment and solve the technical problems of insufficient geological risk identification depth and limited safety mining depth demonstration accuracy in the traditional assessment method by multi-source heterogeneous geological data depth fusion, complex geological structure characteristic intelligent identification and dynamic safety mining depth assessment. In order to achieve the above purpose, the invention provides a mineral resource coverage risk assessment method, which comprises the following steps: Acquiring multi-source heterogeneous geological data, including geological exploration data, drilling core data, geophysical exploration data, historical evaluation cases and peripheral mine exploitation data, and performing deep fusion on the multi-source heterogeneous geological data by adopting a robust neural network training method controlled by game theory gradient to obtain a fusion data set; Based on the fusion data set, carrying out deep mining and reasoning on high-dimensional geological data by utilizing the mode recognition and association analysis capability of a large model and combining manifold perception regularization technology, and recognizing geological structure characteristics including stratum interfaces, lithology distribution, fault fracture zones and joint cracks; based on the geologic structure characteristics, a dynamic safety mining depth evaluation model is established by using a fractional peak differential equation neural network with efficient accompanying parameter training, and the risk, mining disturbance effect and engineering safety threshold of the geologic stru