CN-122022037-A - Intelligent prediction method for magma hot liquid type iron polymetallic ore in desert area based on multisource data fusion and machine learning

CN122022037ACN 122022037 ACN122022037 ACN 122022037ACN-122022037-A

Abstract

The invention relates to the technical field of mineral resource prediction and investigation, and particularly discloses an intelligent prediction method for magma hot liquid type iron polymetallic ore in a desert area based on multisource data fusion and machine learning, which comprises the following steps of S1, multisource ore finding data collection and pretreatment; the method comprises the steps of S2, multi-source data vectorization and thematic map generation, S3, multi-source vector data gray level and standardization, S4, multi-feature data set construction and feature optimization, S5, machine learning model training and optimization, S6, ore forming perspective area prediction and geological interpretation, and deep mineral intelligent prediction by comprehensively utilizing geophysics, geology, structure and drilling data and through multi-source data fusion and machine learning model aiming at magma hot liquid type iron polymetallic ores, so as to overcome the defects of insufficient data utilization, low identification precision of magma hot liquid ore forming systems and poor prediction effect in the current desert region deep mining method.

Inventors

ZHONG MINGFENG
WANG JINHAI
WANG LIANG
LI FENGTING
MA XINLIANG
FU YANAN
LIU ZHIGANG
DING XIAOYING
CHEN LIANGYONG

Assignees

青海省第三地质勘查院
湖南五维地质科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260126

Claims (10)

1. A magma hot liquid type iron polymetallic ore intelligent prediction method based on multisource data fusion and machine learning in a desert area is characterized by comprising the following steps: Step S1, multi-source mining data collection and pretreatment, namely collecting various basic data of a target desert prediction area, and extracting abnormal information related to magma hydrothermal activities; s2, multi-source data vectorization and thematic map generation, namely converting the data into a unified vector format thematic map, and highlighting the ore-forming characteristics of magma hot liquid; Step S3, converting each vector format thematic map into a raster image with the same resolution, and carrying out graying and normalization processing to obtain a gray map; S4, constructing and optimizing the multi-feature data set, namely aligning pixel values of all gray level images according to the same geographic coordinates to form a multi-channel feature data set, and drilling holes The map is used as label data to increase derivative features related to magma hydrothermal solution ore formation; s5, training and optimizing a machine learning model, namely converting a multi-channel characteristic data set and corresponding label data thereof into a characteristic array, training a classification model, and screening key prediction factors; And S6, predicting and geology interpreting an ore-forming remote scenic spot, namely predicting the whole research area by using a trained model, generating an ore-forming probability map, and defining a high-probability area related to magma hydrothermal activity by combining an area ore-forming rule.
2. The intelligent prediction method for magma hot liquid type iron polymetallic ore in desert areas based on multi-source data fusion and machine learning according to claim 1, wherein the basic data comprise geophysical data, geological data, construction data, drilling and mineralization data.
3. The intelligent prediction method for magma hot liquid type iron polymetallic ore in desert area based on multi-source data fusion and machine learning as claimed in claim 2, wherein the preprocessing comprises the following steps: calculating poles and prolongation of magnetic method data; calculating residual anomalies for the gravity data; Solving a vertical first derivative; And extracting low-resistance high-polarization anomalies in a specific numerical range from the electrical data.
4. The intelligent prediction method for the magma hot liquid type iron polymetallic ore in the desert area based on multi-source data fusion and machine learning according to claim 1, wherein the specific steps of the step S2 are as follows: s2a, extracting high-magnetic anomalies from magnetic data, and highlighting local high-value anomalies related to mineralization; S2b, creating a buffer area by taking each drilling hole as a center, and generating an AOI graph according to the mineralization type, the alteration strength and the grade data revealed by drilling; And S2c, calculating the distance from each pixel point to the nearest mine control structure, and generating a vector format thematic map of the mine control structure distance.
5. The intelligent prediction method for magma hot liquid type iron polymetallic ore in desert areas based on multi-source data fusion and machine learning according to claim 4, wherein the step S2 further comprises the step S2d of assigning unique integer values to different geological units in geological data as codes when generating a vector format thematic map of ore structure distance.
6. The intelligent prediction method for the magma hot liquid type iron polymetallic ore in the desert area based on multi-source data fusion and machine learning according to claim 1 or 5, wherein in the step S3, nonlinear normalization is adopted to process abnormal geophysical data so as to enhance the display effect of weak anomalies; and giving different weight values to lithology data according to the correlation with magma hydrothermal mineralization.
7. The intelligent prediction method for magma hot liquid type iron polymetallic ore in desert area based on multi-source data fusion and machine learning as claimed in claim 6, wherein in the step S3, the original pixel value is calculated by Mapping to normalized pixel values The formula of (2) is as follows: ; Wherein, the Pixel values in the original image; is the minimum pixel value in the original image; is the maximum pixel value in the original image; Is the minimum of the target range; is the maximum of the target range.
8. The intelligent prediction method for the magma hot liquid type iron polymetallic ore in the desert area based on multi-source data fusion and machine learning according to claim 1, wherein the derived features in the step S4 are used for constructing intersection density, geophysical anomaly combination features and lithology combination indexes.
9. The intelligent prediction method for the magma hot liquid type iron polymetallic ore in the desert area based on multi-source data fusion and machine learning according to claim 1, wherein the step S5 comprises the following steps: the multi-channel characteristic data set and the corresponding label data are converted into a characteristic array; Training a classification model by adopting a random forest algorithm, and screening characteristic variables most relevant to magma hot liquid ore formation through characteristic importance analysis; Aiming at the characteristics of deep ore deposits of the desert coverage area, model super parameters are optimized, and the recognition capability of weak abnormality and complex mineralization modes is improved.
10. The intelligent prediction method for the magma hot liquid type iron polymetallic ore in the desert area based on multi-source data fusion and machine learning according to claim 9 is characterized in that in the step S6, the contribution degree of each ore control factor is quantitatively evaluated according to the feature importance analysis result, and the geological interpretation of the prediction result is enhanced.

Description

Intelligent prediction method for magma hot liquid type iron polymetallic ore in desert area based on multisource data fusion and machine learning Technical Field The application relates to the technical field of mineral resource prediction and investigation, and particularly discloses an intelligent prediction method for magma hot liquid type iron polymetallic ore in a desert area based on multisource data fusion and machine learning. Background The desert coverage magma hot liquid type iron polymetallic ore is an important mineral resource type, and the formation of the hot liquid polymetallic ore is generally closely related to the activity of medium acid magma and the hot liquid ore forming effect. With the continuous increase of mineral resource exploration in desert areas, deep mining works face a plurality of technical challenges that wind-blown sand, flood deposit and the like are widely covered, matrix is exposed and scarce, effects of traditional methods such as geological map filling and water system deposit measurement are obviously reduced, geochemical anomaly responses are weak and dispersed, deep mineralization information is difficult to effectively indicate, and conventional geophysical methods are high in polynomiance and difficult to accurately identify deep geological structures and mineralization anomalies related to magma hydrothermal activities. At present, the following method is mainly adopted in deep mining research in desert areas: Geophysical methods combine to detect deep geologic structures, typically by high-precision magnetic, gravitational, electromagnetic methods, and the like. However, different geophysical prospecting methods are often interpreted independently, and lack effective data fusion and quantitative analysis, so that the overall recognition capability of a magma hydrothermal system is insufficient. Geological-geochemical method, namely establishing regional ore formation rule through small outcrop observation and drill core research. But has limited application effect in coverage area, and is difficult to accurately predict in a large area. Remote sensing and structural analysis, namely identifying shallow alteration information by utilizing multispectral and hyperspectral remote sensing, and deducing an ore-forming beneficial part by combining structural analysis. However, the method is greatly influenced by the covering layer, and deep information acquisition is insufficient. Machine learning preliminary application there have been research attempts in recent years to apply machine learning algorithms to mineral predictions. However, the original data are mostly directly adopted, and systematic data preprocessing and feature optimization methods aiming at the characteristics of desert coverage areas and magma hydrothermal type ore deposits are lacked Therefore, the existing mining method for deep mining in the desert area has the defects of insufficient data utilization, low recognition precision of a magma hot-liquid mining system and poor prediction effect, and the mining method capable of carrying out multi-source data fusion, mining abnormality recognition and accurate positioning in the prediction of the magma hot-liquid type deposit in the desert coverage area is not available. Disclosure of Invention The invention aims to provide a method for intelligently predicting deep mineral products by comprehensively utilizing geophysical, geological, structural and drilling data and through multi-source data fusion and a machine learning model aiming at magma hot-liquid type iron polymetallic ores, so as to overcome the defects of insufficient data utilization, low identification precision of magma hot-liquid ore forming systems and poor prediction effect in the current desert region deep ore finding method. In order to achieve the above purpose, the basic scheme of the invention provides an intelligent prediction method for magma hot liquid type iron polymetallic ore in a desert area based on multi-source data fusion and machine learning, which comprises the following steps: Step S1, multi-source mining data collection and pretreatment, namely collecting various basic data of a target desert prediction area, and extracting abnormal information related to magma hydrothermal activities; s2, multi-source data vectorization and thematic map generation, namely converting the data into a unified vector format thematic map, and highlighting the ore-forming characteristics of magma hot liquid; Step S3, converting each vector format thematic map into a raster image with the same resolution, and carrying out graying and normalization processing to obtain a gray map; S4, constructing and optimizing the multi-feature data set, namely aligning pixel values of all gray images according to the same geographic coordinates to form a multi-channel feature data set, taking a drilling AOI image as tag data, and adding derivative features related to magma hydrothermal mine formatio