Search

CN-122021850-A - Mining geological exploration data analysis method and system

CN122021850ACN 122021850 ACN122021850 ACN 122021850ACN-122021850-A

Abstract

The invention discloses a mining geological exploration data analysis method and system, which comprises the steps of constructing a multi-mode geological knowledge map based on an ore forming rule, fusing space coordinates and geological features and realizing dynamic increment updating, adopting a three-dimensional convolution neural network to conduct space gridding main reserve prediction, combining an abnormal perception agent model based on knowledge distillation and contrast learning to quantify abnormal features of geological space, automatically judging and locating abnormal areas through multi-mode feature fusion and a dynamic threshold mechanism, further executing semantic alignment and parameter compensation based on the knowledge map, carrying out local correction on a grade prediction result, and establishing closed loop feedback to continuously optimize knowledge rules.

Inventors

  • CHEN RENTONG
  • Chen Fuwan

Assignees

  • 海南睿通工程咨询有限公司

Dates

Publication Date
20260512
Application Date
20260116

Claims (10)

  1. 1. The mining geological exploration data analysis method is characterized by comprising the following steps of: S1, constructing a geological knowledge map based on geological mineralization rules, and establishing a quantized mapping relation between space coordinates and geological features; S2, constructing a main reserve estimation model, dividing an exploration area into regular space grids, inputting geophysical data, drilling core data and chemical exploration data, and outputting initial resource grade predicted values and confidence distribution of each grid unit; S3, constructing an abnormal perception agent model, wherein the abnormal perception agent model extracts high-order semantic rules from the geological knowledge graph, generates an abnormal scoring function and establishes quantitative mapping of geological anomaly and space coordinates; S4, synchronously operating the main reserve estimation model and the abnormal perception agent model, and generating two groups of data aiming at the same space grid unit, namely a grade prediction tensor output by the main model and a geological abnormality tensor output by the agent model; s5, carrying out binarization segmentation on the geological abnormal degree tensor based on a dynamic threshold judgment mechanism, and triggering a knowledge intervention flow when the abnormal degree of a local area exceeds an adaptive threshold; s6, performing geological semantic alignment operation, and performing graph matching on the space feature vector triggering the intervention region and a similar construction scene in the knowledge graph to generate a semantic alignment parameter set; and S7, carrying out local correction on the main model prediction result based on the semantic alignment parameter set, and carrying out nonlinear combination on the knowledge guiding parameters and the original prediction value by adopting a weighted fusion strategy to generate a corrected resource grade distribution model.
  2. 2. The method for analyzing mine geological exploration data according to claim 1, wherein said step S7 further comprises: And S8, establishing a closed-loop feedback mechanism, inputting the corrected prediction result and the corresponding geological feature label which are verified by drilling into a knowledge graph updating module, and optimizing geological constraint rules and abnormal scoring function parameters in the knowledge graph through an incremental learning strategy.
  3. 3. The method for analyzing mine geological exploration data according to claim 1, wherein said step S1 specifically comprises: Acquiring multi-source heterogeneous geological data and constructing an original knowledge base of a geological knowledge graph; Performing entity identification and semantic annotation on the multi-source heterogeneous geological data based on geological mineralization rules, and extracting core geological elements as node representations in a knowledge graph; Carrying out relation extraction on the marked geological entity by utilizing a graph neural network and knowledge extraction technology, identifying the spatial association of faults and ore bodies, the causal relation of lithology combination and mineralization effect and the synergistic effect of alteration superposition and mineralization enrichment, and constructing the edge relation in the multi-mode graph structure; Performing node vectorization representation on the multi-modal graph structure based on a graph embedding algorithm, mapping geological entities and relations thereof to a low-dimensional continuous vector space, and generating an embedding model of a geological knowledge graph; Establishing a mapping function between the space coordinates and the geological features, and associating the three-dimensional exploration grid with nodes in the knowledge graph through a space indexing mechanism; and executing an increment updating mechanism on the geological knowledge graph, and carrying out online optimization on node representation and edge weight in the knowledge graph by adopting a graph attention network based on geological feature labels in field verification data and corrected reserves prediction results.
  4. 4. A method of analyzing mining geological survey data according to claim 3, wherein the multi-source heterogeneous geological data includes regional structure maps, borehole lithology logs, geochemical anomaly maps, ore body boundary data, and zone distribution information.
  5. 5. The method for analyzing mine geological exploration data according to claim 1, wherein said step S2 specifically comprises: performing space discretization on a mine exploration area, dividing a three-dimensional geological space into equally-spaced cube grid units based on a geological coordinate system, and establishing a uniform space reference frame; Obtaining geophysical data, drilling core data and chemical exploration data, and carrying out standardized preprocessing on the geophysical data, the drilling core data and the chemical exploration data to generate a structured data cube; Constructing a main reserve estimation model based on a three-dimensional convolutional neural network; performing supervision training on the main reserve estimation model based on a marked training sample set, wherein the training sample comprises the structured data cube and a corresponding real grade value label, and optimizing model parameters by using a mean square error loss function; And inputting the structured data cube into a trained main reserve estimation model, performing resource grade prediction on a grid unit by grid unit, generating a three-dimensional grade distribution tensor, and calculating a prediction confidence index of each unit.
  6. 6. The mining geological exploration data analysis method according to claim 5, wherein the standardized preprocessing comprises missing value interpolation, outlier rejection and characteristic normalization.
  7. 7. The method for analyzing mine geological exploration data according to claim 1, wherein said step S3 specifically comprises: Based on multi-mode geological structure rules stored in a geological knowledge map, extracting high-order semantic representation of fault intersection characteristics, lithology combination modes and alteration superposition effects, and constructing a knowledge guiding template of geological abnormal characteristics; carrying out graph embedding processing on the knowledge guiding template of the geological abnormal feature, and mapping the knowledge guiding template into geological knowledge embedding representation in a low-dimensional semantic vector space by utilizing a graph neural network to form a knowledge representation vector set; Constructing a contrast learning training frame based on the knowledge representation vector set, and carrying out feature contrast on a normal ore forming mode sample and an abnormal structure sample by adopting a twin network structure; deploying the trained agent model into a lightweight neural network structure, inputting the space coordinate feature vector and the local geological attribute data of the current exploration area, and outputting geological anomaly scores of corresponding space units; And constructing a geological anomaly tensor based on the geological anomaly score.
  8. 8. The method for analyzing mine geological exploration data according to claim 1, wherein said step S4 specifically comprises: Performing three-dimensional convolution operation on the divided regular space grids based on the main reserve estimation model, inputting geophysical data, drilling core data and chemical exploration data, and outputting initial resource grade predicted values and predicted confidence tensors of each grid unit to form grade predicted tensors; performing graph matching and semantic rule reasoning operation on the same space grid based on the anomaly perception agent model, inputting a high-order mineralization mode rule extracted from a geological knowledge graph and a current space feature vector, and outputting a geological anomaly degree scoring tensor; performing space-time alignment processing on the grade prediction tensor and the geological anomaly degree scoring tensor, and aligning the prediction result of each grid unit and the anomaly score based on a space coordinate mapping relation to generate a synchronous aligned multi-modal characteristic tensor; Splicing and fusing the aligned grade prediction tensor and the geological abnormality tensor to generate a combined feature space, and constructing a multi-mode feature fused decision space tensor; And performing data format standardization and normalization processing based on the decision space tensor, and unifying the grade predicted value and the abnormality degree score to the same numerical scale.
  9. 9. The method of claim 8, wherein the joint feature space comprises a predicted grade, a predicted confidence and an anomaly score.
  10. 10. A mining geological exploration data analysis system is characterized in that the mining geological exploration data analysis method is adopted to conduct mining geological exploration data analysis.

Description

Mining geological exploration data analysis method and system Technical Field The invention relates to the technical field of geological resource exploration and reserve prediction, in particular to a mining geological exploration data analysis method and system. Background In the field of mine geological exploration data analysis and resource reserve estimation, the main stream technical scheme mainly takes a machine learning model driven by data as a core, and is assisted by multi-source geological data (such as geophysics, drilling cores, geochemistry and the like) to carry out space feature modeling, so that the prediction of mineral body space distribution and resource grade is realized. In recent years, with the development of artificial intelligence and deep learning technologies, three-dimensional convolutional neural networks (3D-CNN), U-Net variants, graph convolution networks and the like are gradually applied to reserve estimation tasks, and prediction efficiency and automation level are remarkably improved through end-to-end automatic feature extraction and spatial relationship modeling. Meanwhile, the industry also starts focusing on the role of domain knowledge in modeling flow, attempts to incorporate expert experience such as geological formation rules, priori modes and the like into a machine learning model, and partial researches have introduced geological knowledge patterns as auxiliary information sources to realize feature enhancement or priori constraint; However, existing data-driven based reserve estimation methods generally have significant limitations. Firstly, in a region with complex geological structure or sparse effective exploration data, as the main model mainly depends on the existing observation data, the high-order geological process and the space abnormality are difficult to accurately represent, and the reserve prediction result is easy to deviate in local part. Secondly, although the industry has preliminary attempts to encode knowledge maps into model feature inputs to realize 'knowledge pre-injection', such static injection methods fail to form dynamic linkage with the model reasoning process. The knowledge information only participates in feature enhancement at the initial stage of the model, the reasoning stage can not realize targeted correction according to the actual local abnormality, the knowledge and data drive model is fractured, and the timely guiding value of the domain knowledge is difficult to develop. Especially in the face of geological anomalies or complex construction areas such as fault intersection, fold structures, heterogeneous changes and the like, the main flow model often has the problems of insufficient generalization capability, insensitivity to anomaly detection, unreliable prediction and the like. Disclosure of Invention The invention provides a mining geological exploration data analysis method and system for solving the technical problems. The technical scheme of the invention is realized in such a way that the mining geological exploration data analysis method comprises the following steps: S1, constructing a domain knowledge graph based on a geological mineralization rule, wherein the knowledge graph adopts a multi-mode graph structure to store geological constraint rules comprising constructional features, lithology combination, mineralization alteration and other dimensions, and establishes a mapping relation between space coordinates and geological features; S2, constructing a main reserve estimation model based on a three-dimensional convolutional neural network, dividing an exploration area into regular space grids, inputting geophysical data, drilling core data and chemical exploration data, and outputting initial resource grade predicted values and confidence distribution of each grid unit; s3, constructing an abnormal perception agent model based on knowledge distillation, extracting high-order semantic rules from the geological knowledge map by the agent model through a comparison learning strategy, generating an abnormal scoring function comprising fault intersection characteristics, a lithology combination mode and an alteration superposition effect, and establishing quantitative mapping of geological anomaly and space coordinates; S4, synchronously operating a main reserve estimation model and an abnormal perception agent model, and generating two groups of output data aiming at the same space grid unit, wherein the grade prediction tensor output by the main model and the geological abnormality tensor output by the agent model form a multi-mode feature fused decision space; S5, carrying out binary segmentation on the geological anomaly tensor based on a dynamic threshold judgment mechanism, and triggering a knowledge intervention flow when the anomaly degree of a local area exceeds an adaptive threshold, wherein the adaptive threshold is dynamically adjusted according to the exploration data density of the area and the pred