A copper mine multi-parameter prediction method and evaluation system based on ground features and geochemistry

Abstract

The present application relates to the technical field of mineral resources exploration, in particular to a copper mine multi-parameter prediction method and evaluation system based on geo-geochemistry, the prediction method comprising the following steps: step 1, data acquisition: collecting geophysical data and geochemical data of copper ore prospecting area through field exploration equipment; step 2, data processing; step 3, model training: training the preprocessed geophysical-geochemical multi-parameter data using machine learning algorithm to establish a prospecting prediction model; step 4, visual analysis; step 5, building a three-dimensional mineralization model. The present application integrates the geophysical data and hydrogen-oxygen-carbon-sulfur stable isotope geochemical data of the copper ore prospecting area, realizes multi-parameter feature fusion, intelligent identification of mineralization anomalies and prediction model training with the help of machine learning algorithm, realizes accurate prediction of copper mineralization potential and target area distribution, and improves the efficiency and accuracy of ore prospecting.

Description

Technical Field

[0001] This invention relates to the field of mineral resource exploration technology, specifically to a multi-parameter prediction method and evaluation system for copper deposits based on geomorphology. Background Technology

[0002] The multi-parameter geophysical-geochemical method for copper ore prospecting prediction refers to a comprehensive approach to copper ore exploration that integrates geophysical exploration with stable isotope geochemical analysis. Geophysical exploration detects differences in the density, magnetism, and electrical properties of underground geological bodies, and uses data acquisition and inversion techniques to locate potential mineralization zones. Stable isotope geochemical analysis studies the composition and spatial distribution of isotopes such as hydrogen, oxygen, carbon, and sulfur to trace the origin of ore-forming materials, analyze the evolution paths and mixing ratios of ore-forming fluids, and identify mineralization centers. Finally, through the processing, modeling, and analysis of both types of data, the method assesses copper ore mineralization potential and predicts target area distribution. The corresponding copper ore prospecting evaluation system is a functional system integrating these methods. It includes modules for acquiring and preprocessing geophysical and stable isotope data, modules for modeling and analyzing multi-parameter data to identify mineralization anomalies, and modules for constructing mineralization models and delineating prospecting target areas. The system aims to improve the efficiency and accuracy of copper ore prospecting through systematic operation.

[0003] Current technologies suffer from the following major shortcomings: insufficient synergy between geophysical exploration and stable isotope geochemical analysis. While geophysical exploration can locate subsurface physical anomalies, it cannot explain whether these anomalies are related to mineralization; while stable isotope geochemical analysis can reveal mineralization mechanisms, it struggles to accurately correlate geochemical characteristics with subsurface spatial locations. The two technologies are applied independently, resulting in ineffective integration of mineralization information and hindering the formation of a complete mineral exploration logic based on "spatial location and genetic tracing." Therefore, a multi-parameter prediction method and evaluation system for copper deposits based on geophysical features and geochemistry is urgently needed. Summary of the Invention

[0004] To address the aforementioned issues, this invention provides a multi-parameter prediction method and evaluation system for copper deposits based on geophysical features and geochemical data. By integrating geophysical data and stable isotopic geochemical data of hydrogen, oxygen, carbon, and sulfur from copper exploration areas, and utilizing machine learning algorithms, it achieves multi-parameter feature fusion, intelligent identification of mineralization anomalies, and training of prediction models. This enables accurate prediction of copper mineralization potential and target area distribution, thereby improving exploration efficiency and accuracy.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows: A multi-parameter prediction method for copper deposits based on land features and geochemical analysis, comprising the following steps: Step 1, Data Acquisition: Deploy several field exploration devices in the prospecting area and collect geophysical and geochemical data of the copper prospecting area through the field exploration devices. Geophysical data includes gravity exploration, magnetic exploration and electrical exploration data, and geochemical data includes stable isotope data. Step 2, Data Processing: Preprocess the collected geophysical and geochemical data, including data cleaning, outlier removal, normalization, and interference suppression. Step 3, Model Training: Machine learning algorithms are used to train the preprocessed geophysical-geochemical multi-parameter data to establish a mineral exploration prediction model; Step 4, Visualization Analysis: Based on the mineral exploration prediction model, conduct big data visualization analysis of copper deposits to generate a multi-parameter fusion visualization map; Step 5: Construct a three-dimensional mineralization model: Combine geophysical inversion results and geochemical characteristics to predict copper mineralization potential and target area distribution and output the visualization.

[0006] Furthermore, in step 1, the electrical exploration data is acquired using the high-density resistivity method and the induced polarization method, and the mineralization zone is located using an inversion algorithm. During the inversion process, topographic correction and resistivity-polarizability joint analysis are integrated to reduce the positioning error.

[0007] Furthermore, in step 1, the stable isotope data includes hydrogen, oxygen, carbon, and sulfur isotopes, which are used to analyze the source, evolution path, and mixing ratio of ore-forming fluids, and to identify mineralization centers through the spatial distribution characteristics of isotopes.

[0008] Furthermore, in step 2, interference suppression employs wavelet transform and filtering techniques. When the intensity of the electromagnetic interference signal exceeds twice that of the effective signal, a multi-level decomposition is performed using the db4 wavelet basis, and hard thresholding is applied to the high-frequency coefficients to improve the signal-to-noise ratio.

[0009] Furthermore, in step 3, the multi-parameter data includes the combination of CSAMT and TDIP electrical resistivity data and stable isotope data. Machine learning algorithms are used for feature fusion and anomaly identification. When the multi-parameter correlation coefficient is greater than 0.7, it is determined to be a mineralization target signal.

[0010] Furthermore, in step 5, the three-dimensional mineralization model is constructed using finite element forward simulation and conjugate gradient inversion algorithm, combined with borehole data and geochemical potential relationship. The iteration stops when the root mean square error is less than 2.5%, and the three-dimensional resistivity model and mineralization probability distribution map are output.

[0011] The above approach has the following beneficial effects: 1. This scheme achieves deep synergy between geophysical and geochemical data, effectively overcoming the shortcomings of traditional mineral exploration methods that rely on the isolated application of these two types of data. In traditional mineral exploration, while geophysical exploration (such as electrical and gravity methods) can locate underground anomalies, it is difficult to explain their causes. Geochemical analysis (such as stable isotopes) can trace the source, evolution path, and mineralization center of ore-forming fluids, but it cannot accurately correspond to the spatial location of anomalies. This scheme, however, collects geophysical data such as gravity, magnetic, and electrical methods, along with stable isotope geochemical data of hydrogen, oxygen, carbon, and sulfur. By leveraging machine learning algorithms to achieve multi-parameter feature fusion, it can not only use isotope data to clarify the mineralization background of geophysical anomalies (such as determining whether a low resistivity anomaly is a mineralization zone rather than other geological bodies), but also visualize the mineralization centers revealed by isotopes through geophysical data. This allows mineral exploration analysis to have both "genetic explanation" and "spatial location" support, improving the comprehensiveness and reliability of mineral exploration judgment.

[0012] 2. This solution optimizes the data processing workflow and model building logic, significantly reducing errors and uncertainties in traditional mineral exploration. Traditional data processing is susceptible to electromagnetic interference masking effective signals, and topographical undulations leading to mineralization zone location errors. Furthermore, model building often relies on experience and lacks unified accuracy control standards. This solution addresses interference issues by employing db4 wavelet basis multi-layer decomposition and hard thresholding techniques, effectively suppressing strong electromagnetic interference to improve the data signal-to-noise ratio. For location errors, it integrates a high-precision inversion algorithm combining topographic correction and resistivity-polarizability joint analysis. For model accuracy, it utilizes finite element forward simulation and conjugate gradient inversion algorithms, clearly defining an iteration stopping condition of root mean square error below 2.5%. Simultaneously, it automatically identifies multi-parameter anomalies using machine learning algorithms (determining mineralization signals based on a correlation coefficient greater than 0.7), reducing subjective errors from manual experience-based judgments. This makes data processing more targeted, model building more standardized, and significantly improves the accuracy of mineral exploration predictions.

[0013] 3. This solution improves the efficiency and practicality of mineral exploration, and is better suited to the needs of complex copper exploration scenarios. In traditional mineral exploration, data processing, model training, and target area prediction often require separate manual operations, which are time-consuming and difficult to integrate multi-source data to form intuitive results. In contrast, this solution achieves seamless processing from data acquisition and preprocessing to model training through a mineral exploration evaluation system. It accelerates the analysis and feature extraction of multi-parameter data with the help of machine learning algorithms, and can complete mineralization signal identification without a lot of manual intervention. Then, through visualization analysis, it generates multi-parameter fusion maps, constructs three-dimensional metallogenic models, and outputs mineralization potential and target area distribution. This allows geologists to intuitively grasp the underground mineralization characteristics and quickly identify key target areas, avoiding the problems of "data fragmentation" and "unintuitive results" in traditional mineral exploration. It significantly shortens the mineral exploration cycle and is more adaptable to the needs of copper exploration under different geological conditions (such as volcanic basins, fold belts, etc.), providing more efficient technical support for actual mineral exploration work.

[0014] Furthermore, a multi-parameter evaluation system for copper deposits based on land features and geomorphology, based on any of the aforementioned multi-parameter prediction methods for copper deposits based on land features and geomorphology, includes a data acquisition module, a machine learning modeling module, and a model building module, wherein: The data acquisition module receives raw geophysical and geochemical data transmitted from field exploration equipment. It processes the data using data cleaning, outlier removal, and interference suppression algorithms, and outputs a cleaned, standardized multi-parameter dataset to the machine learning modeling module. When electromagnetic interference exceeds twice the effective signal strength or when terrain undulations cause data distortion, the data acquisition module uses wavelet transform and terrain correction algorithms for enhancement. The machine learning modeling module receives a standardized multi-parameter dataset, uses machine learning algorithms for training and feature fusion, establishes a mineral exploration prediction model, and outputs mineralization anomaly identification results and grade prediction data to the model building module. When the multi-parameter correlation coefficient is greater than 0.7, the machine learning modeling module determines it as a high-probability mineralization target signal and strengthens the output. The model building module receives mineralization anomaly identification results and grade prediction data. It uses big data visualization and 3D mesh modeling technology to generate a multi-parameter fusion visualization map and a 3D mineralization potential model. It outputs the target area prediction results and 3D model data to the user interaction terminal for easy viewing by operators. When the root mean square error of the model inversion is less than 2.5%, the model building module determines that the visualization map and the 3D mineralization potential model have converged and locks the final predicted target area.

[0015] Furthermore, the data acquisition module includes: The geophysical data acquisition unit is used to receive exploration commands, perform measurements using the high-density resistivity method and the induced polarization method, and output the raw electrical exploration data to the interference suppression unit. The geochemical data acquisition unit receives sampling commands, performs micro-area analysis of ore samples using LA-ICP-MS or μ-XRF technology, and outputs stable isotope data to the interference suppression unit. When in-situ analysis of mineral composition is required, the geochemical data acquisition unit uses synchrotron X-ray fluorescence technology for sensitivity detection. The interference suppression unit receives raw electrical exploration data and stable isotope data, performs noise reduction using wavelet transform and filtering techniques, and outputs the denoised fused data to the data standardization unit.

[0016] Furthermore, the machine learning modeling module includes: The multi-parameter fusion unit receives standardized multi-parameter datasets, performs feature extraction and fusion using principal component analysis and correlation analysis, and outputs the fused feature set to the model training unit. When CSAMT and TDIP data coexist, impedance tensor decomposition is used to calculate anisotropy coefficients to identify mineralization anomalies. The model training unit receives the fused feature set, trains it using machine learning algorithms, and outputs the trained mineral exploration prediction model to the prediction unit. When the number of mineralized and non-mineralized samples in the training samples is unbalanced, oversampling and cost-sensitive learning algorithms are used to optimize the model performance. The prediction unit receives the trained mineral exploration prediction model and new exploration data, performs mineralization potential and grade prediction, and outputs mineralization anomaly identification results and grade prediction data to the model building module; when the resistivity of the input data is in the range of 30~50Ω·m, the predicted copper grade is 1.0%~2.5%.

[0017] Furthermore, the model building module includes: The three-dimensional inversion unit receives mineralization anomaly identification results and grade prediction data, uses finite element forward modeling and conjugate gradient inversion algorithms to perform three-dimensional resistivity modeling, and outputs the three-dimensional resistivity model data to the mineralization potential assessment unit; when the number of inversion iterations reaches 30 or the root mean square error is less than 2.5%, the calculation stops and the final model is output. The mineralization potential assessment unit is used to receive three-dimensional resistivity model data, combine resistivity-grade relationship model and fuzzy comprehensive evaluation method to calculate the mineralization probability of each three-dimensional grid cell, and output a mineralization potential distribution map to the target area delineation unit; when the mineralization probability value of a grid cell is greater than 0.7, it is marked as a first-level target area. The target area delineation unit receives the mineralization potential distribution map, uses spatial clustering and boundary recognition algorithms to automatically delineate the spatial location and boundaries of the prospecting target area, and outputs the final target area prediction results and a 3D visualization report.

[0018] The above approach has the following beneficial effects: 1. This solution achieves collaborative acquisition and precise preprocessing of geophysical and geochemical data through a data acquisition module, effectively solving the problems of isolated acquisition and low processing accuracy of the two types of data in traditional copper prospecting. In traditional prospecting, geophysical exploration often suffers from data distortion due to electromagnetic interference and topographic distortion, while geochemical analysis is prone to difficulty in capturing micro-area compositional characteristics due to coarse sample processing. Furthermore, the two types of data lack effective linkage, failing to form a complete metallogenic information chain. In this solution, the geophysical data acquisition unit uses high-density resistivity and induced polarization methods to accurately acquire electrical resistivity data, while the geochemical data acquisition unit achieves high-precision analysis of stable isotopes through micro-area techniques such as LA-ICP-MS and synchrotron X-ray fluorescence. Both types of data undergo wavelet transform denoising processing by the interference suppression unit, ensuring not only the acquisition quality of individual data types but also forming a collaborative data foundation of "spatial positioning + genetic tracing" through data fusion, providing a comprehensive and reliable data source for subsequent modeling.

[0019] 2. This solution, relying on a machine learning modeling module to construct an intelligent analysis system, overcomes the limitations of traditional mineral exploration that depends on human experience for anomaly identification and prediction. In traditional mineral exploration, feature extraction and anomaly judgment of multi-parameter data largely depend on the subjective experience of geologists, which is prone to misjudgment due to insufficient parameter correlation and sample imbalance, and it is difficult to achieve quantitative correlation between grade and mineralization signals. In this solution, the multi-parameter fusion unit automatically extracts the core features of geophysical and geochemical data through techniques such as principal component analysis and impedance tensor decomposition. The model training unit uses oversampling and cost-sensitive learning to solve the sample imbalance problem, and the prediction unit achieves quantitative prediction based on the correlation between resistivity and grade. The entire process can achieve accurate identification of mineralization target signals (judgment based on multi-parameter correlation coefficient greater than 0.7) without much human intervention. This not only improves the objectivity of anomaly identification but also achieves an upgrade from "qualitative judgment" to "quantitative prediction," significantly reducing the error of traditional experience-based judgment.

[0020] 3. This solution utilizes a model building module to achieve high-precision 3D mineralization modeling and automated target area delineation, significantly improving the practicality and applicability of mineral exploration results. In traditional mineral exploration, 3D model construction often deviates significantly from the actual ore body due to algorithm simplification and lack of precision control. Target area delineation relies on manual division, resulting in vague boundaries and unclear priorities, making it difficult to directly guide exploration operations. In this solution, the 3D inversion unit employs finite element forward modeling and conjugate gradient inversion algorithms, using a root mean square error of less than 2.5% or 30 iterations as a clear accuracy standard to ensure the reliability of the 3D resistivity model. The mineralization potential assessment unit quantifies the mineralization probability through fuzzy comprehensive evaluation, while the target area delineation unit automatically identifies target area boundaries using spatial clustering algorithms. The final output 3D model and target area report not only intuitively present the underground mineralization distribution characteristics but also clearly define the spatial range of the primary target area (mineralization probability greater than 0.7), providing clear and practical guidance for subsequent drilling operations and avoiding the problem of "difficult to transform and apply" traditional mineral exploration results. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the method steps of the copper deposit multi-parameter prediction method based on land features and geochemicals of the present invention; Figure 2 This is a schematic diagram of the system framework of an embodiment of the copper mine multi-parameter evaluation system based on land features and geomorphology of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0024] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0025] The following detailed description illustrates the specific implementation method: Example 1:

[0026] In traditional copper prospecting, geophysical exploration and geochemical analysis are often conducted independently, lacking multi-parameter coordination and intelligent integration. While geophysical methods (such as gravity, magnetic, and electrical methods) can quickly identify underground physical anomalies, they are easily affected by complex geological conditions and struggle to accurately distinguish between mineralized bodies and non-mineralized low-resistivity bodies (such as graphitized strata). Geochemical methods (such as isotope analysis) can effectively trace the source and evolution path of ore-forming fluids, but they rely on point sampling, have limited spatial coverage, and data processing largely depends on human experience, making it difficult to achieve accurate predictions over a large area. Furthermore, existing technologies have significant positioning errors for deep ore bodies. For example, in undulating terrain, two-dimensional inversion ignoring topographic effects can lead to mineralization zone positioning errors as high as 20%, and single-parameter inversion results exhibit significant ambiguity, resulting in low drilling verification hit rates, long exploration cycles, and high costs.

[0027] Based on the above problems, the inventors have proposed the following: Figure 1 This paper presents a multi-parameter prediction method for copper deposits based on land features and geochemistry. Taking a copper exploration area on the southern margin of the North China Platform as an example, this area has large topographic relief (elevation difference of up to 320m), well-developed overburden and fractured zones, strong electromagnetic interference, and small resistivity difference between deep ore bodies and surrounding rocks (ore bodies 10~50Ω·m, surrounding rocks 200~500Ω·m). Traditional methods are difficult to accurately delineate mineralization boundaries. This scheme is implemented according to the following steps: Step 1, Data Acquisition: Several field exploration devices are deployed within the prospecting area to collect geophysical and geochemical data. Geophysical data includes gravity, magnetic, and electrical resistivity data, while geochemical data includes stable isotope data. Electrical resistivity data is acquired using high-density resistivity and induced polarization methods, and an inversion algorithm is used to locate mineralization zones. Topographic correction and resistivity-polarizability joint analysis are integrated during the inversion process to reduce location bias. Stable isotope data, including hydrogen, oxygen, carbon, and sulfur isotopes, are used to analyze the source, evolution path, and mixing ratio of ore-forming fluids, and to identify mineralization centers through isotope spatial distribution characteristics. Twelve survey lines are deployed within the exploration area, and electrical resistivity data are simultaneously acquired using high-density resistivity (Wenner device, electrode spacing 5m) and induced polarization (intermediate gradient device, electrode spacing 1000m). To overcome the influence of topography, high-precision topographic data (sampling interval 2m) was collected simultaneously during the measurement process to generate a digital elevation model for inversion correction. In the data inversion, Res2Dinv software with topographic adjustment was used, with a horizontally layered medium as the initial model, to iteratively calculate the resistivity distribution layer by layer. Simultaneously, quartz and sulfide samples were systematically collected at survey line nodes, and the hydrogen, oxygen, carbon, and sulfur isotope compositions were analyzed using LA-ICP-MS and isotope mass spectrometry. For example, the measured δ¹⁸O value of quartz was 11.7‰±0.88%, and the δ¹⁸O_H₂O of the ore-forming fluid was inferred to be 1.35‰±4.15% using the fractionation equation. Combined with the δD value (-56.66‰±35.12%), it was determined that the fluid originated from magma in its early stages and was mixed with atmospheric precipitation in its middle and late stages. This step, through the spatial coupling of electrical resistivity and isotope data, reduced the mineralization zone location error from the traditional 20% to within 3m.

[0028] Step 2, Data Processing: Preprocessing of the acquired geophysical and geochemical data includes data cleaning, outlier removal, normalization, and interference suppression. Interference suppression employs wavelet transform and filtering techniques. When the electromagnetic interference signal strength exceeds twice the effective signal strength, multi-level decomposition using the db4 wavelet basis is performed, and high-frequency coefficients are hard-thresholded to improve the signal-to-noise ratio. For strong electromagnetic interference within the area (signal strength 2-3 times the effective signal strength), wavelet transform (db4 wavelet basis) is used to perform multi-level decomposition of the original potential data, and hard-thresholding (threshold 0.02V) is applied to the high-frequency coefficients of the 4th and 5th layers, resulting in a 40% improvement in the signal-to-noise ratio after reconstruction. In the joint processing of CSAMT and TDIP data, impedance tensor decomposition and charge rate decay analysis are used. When the correlation coefficient of multiple parameters is greater than 0.7 (e.g., polarizability anomaly > 8% and resistivity < 50 Ω·m), it is identified as a mineralization target signal. For example, on a north-south cross section, by fusing the CSAMT frequency response and TDIP attenuation characteristics, a low-resistivity-high-polarization anomaly at a depth of 300m was identified, increasing the detection depth by 50m compared to traditional methods.

[0029] Step 3, Model Training: Machine learning algorithms are used to train the preprocessed geophysical-geochemical multi-parameter data, which includes a combination of CSAMT and TDIP electrical resistivity data and stable isotope data. Machine learning algorithms are used for feature fusion and anomaly identification. When the correlation coefficient of the multi-parameters is greater than 0.7, it is determined to be a mineralization target signal, and a mineral exploration prediction model is established. The preprocessed electrical resistivity data (CSAMT resistivity, TDIP charge rate) and isotope parameters (δ34S, δ18O) are input into a random forest model for training. The model uses known borehole mineralization sections as labels to learn the nonlinear relationship between multi-parameters and mineralization intensity. During training, when the feature importance ranking shows resistivity, polarizability, and δ34S as key factors, the model automatically weights the contribution of these parameters in the prediction. In the validation phase, the model's predicted mineralization probability for unknown areas achieved an 85% agreement with borehole validation, a 30% improvement compared to a single geophysical method.

[0030] Step 4, Visualization Analysis: Based on the mineral exploration prediction model, a big data visualization analysis of the copper deposit is performed to generate a multi-parameter fused visualization map. Based on the three-dimensional resistivity data collected by the distributed electrode array, a three-dimensional geoelectric model is constructed using finite element forward modeling and conjugate gradient inversion. During the inversion process, the ore body segment exposed by the borehole (burial depth 120~180m, resistivity 30~60Ω·m) is used as a constraint, and the iteration terminates when the root mean square error drops to 2.5%. In the output model, the low resistivity zone (<50Ω·m) accounts for 12% of the volume, and a mineralization probability distribution map is generated by combining isotope-indicated fluid evolution paths (e.g., δ13C values ​​indicate a carbon source of magma and crustal mixing). In the visualization interface, the system integrates gravity, magnetic anomalies, and isotope spatial distribution to delineate the primary target area (probability >0.7), which, after drilling verification, achieved a hit rate of 78%.

[0031] Step 5: Construct a three-dimensional mineralization model. The three-dimensional mineralization model is constructed using finite element forward simulation and conjugate gradient inversion algorithm. Combining borehole data and geochemical level relationships, the iteration stops when the root mean square error is less than 2.5%. The three-dimensional resistivity model and mineralization probability distribution map are output. Combining geophysical inversion results and geochemical characteristics, the copper mineralization potential and target area distribution are predicted and visualized.

[0032] Example 2:

[0033] As attached Figure 2As shown, the difference from Example 1 is that the actual application scenario is the copper prospecting area in the northern part of the Zouping volcanic basin. The geographical coordinates of this mining area are 117°40'~117°43' east longitude and 36°52'~36°54' north latitude, covering an area of ​​about 8.4 km². It belongs to the distribution area of ​​Late Yanshanian intermediate-acidic igneous rocks. The ore bodies are mainly hosted in the fault zone at the edge of the breccia pipe and the skarn contact zone, with a burial depth of 50~350m, and associated with chalcopyrite, molybdenite and other sulfides. The surface of the mining area is covered by a 100m thick Quaternary loose layer, and the maximum local topographic relief is 150m. The electromagnetic interference signal intensity can reach 2.5 times that of the effective signal. In traditional exploration, due to the influence of topographic distortion on electrical resistivity data and the disconnect between isotopic data and geophysical anomalies, three shallow mineralization zones were misjudged, and the drilling hit rate of deep target areas was only 62%. Based on the actual conditions of the mining area, a multi-parameter evaluation system for copper deposits based on land features and geochemistry, and based on the multi-parameter prediction method for copper deposits based on land features and geochemistry described in Example 1, includes a data acquisition module, a machine learning modeling module, and a model building module, wherein: When mineral exploration begins in the mining area, equipment deployed in the field, such as gravimeters (1 μGal resolution), proton magnetometers (accuracy ±0.1 nT), high-density resistivity meters, and isotope analyzers, first transmit raw data in real time to the system's data acquisition module. The data acquisition module receives the raw geophysical and geochemical data transmitted from the field exploration equipment, processes it using data cleaning, outlier removal, and interference suppression algorithms, and outputs a cleaned, standardized multi-parameter dataset, which is then transmitted to the machine learning modeling module. When electromagnetic interference exceeds twice the effective signal strength or when terrain undulations cause data distortion, the data acquisition module uses wavelet transform and terrain correction algorithms for enhancement. The data acquisition module includes: The geophysical data acquisition unit receives exploration commands and uses the high-density resistivity method and induced polarization method for measurement, outputting raw electrical exploration data to the interference suppression unit. Combining the topographic relief and ore body depth characteristics of the mining area, a combined acquisition scheme of "high-density resistivity method + induced polarization method" is adopted: the high-density resistivity method uses a Wenner device with an electrode spacing of 5m, deploying electrodes along 12 north-south survey lines (each 2000m long, with a line spacing of 500m) to cover the entire breccia pipe area and collect raw resistivity data; the induced polarization method uses an intermediate gradient device with a power supply electrode spacing of 800m and a measurement electrode spacing of 20m, collecting polarizability data for 3 key control profiles (passing through known mineralization zones), with an acquisition frequency of 2Hz. Considering that the undulating terrain in the mining area can easily lead to data distortion, this unit synchronously records the terrain coordinates (sampling interval 2m) during the acquisition process, generates 1:2000 terrain contour data, and transmits it to the interference suppression unit along with the original electrical resistivity tomography data. Compared with traditional electrical resistivity tomography acquisition (which only records planar coordinates and ignores terrain), this process provides basic data for subsequent terrain correction and can initially reduce the positioning deviation caused by terrain.

[0034] The geochemical data acquisition unit receives sampling commands and performs micro-area analysis of ore samples using LA-ICP-MS or μ-XRF technology, outputting stable isotope data to the interference suppression unit. When in-situ mineral composition analysis is required, the geochemical data acquisition unit uses synchrotron X-ray fluorescence technology for sensitivity detection. After receiving the sampling command, this unit processes it according to the logic of "sample type adaptation technology": for quartz vein samples (used for hydrogen and oxygen isotope analysis) in borehole cores from the mining area, μ-XRF technology is first used to quickly scan the trace element distribution to screen out quartz grains containing fluid inclusions; then, LA-ICP-MS technology is used to accurately determine the δ¹⁸O of quartz. 18 O value (test accuracy ±0.2‰) and δD value of fluid inclusions (test accuracy ±2‰); for hydrothermal calcite samples (for carbon and oxygen isotope analysis), δD was directly determined using LA-ICP-MS. 13 C (accuracy ±0.1‰) and δ 18 O value; For sulfide samples such as chalcopyrite (for sulfur isotope analysis), when it is necessary to analyze the in-situ sulfur distribution of fine-grained sulfides, this unit automatically switches to synchrotron X-ray fluorescence (spatial resolution 5 μm, detection limit 0.001%), avoiding the component mixing problem caused by traditional crushing sampling, and finally outputting δ 34S-value (accuracy ±0.15‰). All stable isotope data are converted into their format and then transmitted to the interference suppression unit. Compared with traditional chemical wet analysis (which takes 24 hours per sample and damages the sample structure), the micro-area technology used in this unit shortens the analysis time to 2 hours per sample, while preserving the in-situ mineral composition information and improving the accuracy of isotope data by 30%.

[0035] The interference suppression unit receives raw electrical resistivity tomography (EPT) data and stable isotope data, performs denoising using wavelet transform and filtering techniques, and outputs the denoised fused data to the data standardization unit. This unit receives the raw EPT data and stable isotope data and processes them according to a "denoise first, then fuse" logic: First, for electromagnetic interference in the EPT data (the interference signal intensity caused by the Quaternary loose layer in the mining area is 2.5 times the effective signal), it uses a db4 wavelet basis to perform a five-level decomposition of the data, applies a 0.02V hard threshold to the high-frequency interference coefficients of the 4th and 5th levels, and then recovers the effective signal through wavelet reconstruction; simultaneously, it addresses outliers in the isotope data caused by instrument fluctuations (such as δ). 34 Data deviating from the reasonable range of -6‰ to 3.8‰ (S) were removed using the 3σ criterion. Subsequently, the element combined the denoised electrical resistivity data (resistivity, polarizability) with isotopic data (δ¹⁸O). 18 O, δD, δ 13 C、δ 34 S) Spatial fusion is performed based on the coordinates of the sampling points to generate a "coordinate-multi-parameter" associated dataset, which is then transmitted to the data standardization unit. The data standardization unit uses a min-max normalization algorithm to map all parameters to the 0-1 range, eliminating dimensional differences (e.g., resistivity is measured in Ω·m, and isotopes in ‰), and finally outputs the standardized multi-parameter dataset to the machine learning modeling module. Testing showed that the signal-to-noise ratio of the data processed by this unit was improved by 45%, far exceeding the traditional single-filter technique (which only improved by 20%), and the data integrity reached 98% (the traditional method resulted in a 10% data missing rate due to interference removal).

[0036] The machine learning modeling module receives a standardized multi-parameter dataset, uses machine learning algorithms for training and feature fusion, establishes a mineral exploration prediction model, and outputs mineralization anomaly identification results and grade prediction data to the model building module. When the multi-parameter correlation coefficient is greater than 0.7, the machine learning modeling module determines it as a high-probability mineralization target signal and enhances the output. The machine learning modeling module includes: The multi-parameter fusion unit receives standardized multi-parameter datasets, performs feature extraction and fusion using principal component analysis and correlation analysis, and outputs the fused feature set to the model training unit. When CSAMT and TDIP data coexist, impedance tensor decomposition is used to calculate anisotropy coefficients to identify mineralization anomalies. First, it receives the standardized multi-parameter dataset and uses a combined algorithm of principal component analysis and correlation analysis to extract key features: first, principal component analysis is used to extract resistivity, polarizability, and δ... 18 O, δD, δ 13 C、δ 34 The 12 original parameters, including S, were reduced to 3 principal components (cumulative contribution rate of 89%), retaining core information. Correlation analysis was then performed to calculate the correlation coefficients between the principal components. Specifically, for CSAMT electrical resistivity data (supplementary dipole-dipole device data, frequencies 8Hz, 16Hz, and 32Hz) and TDIP data (charge rate M, decay time τ), impedance tensor decomposition was used to calculate anisotropy coefficients. Coefficients > 0.2 were marked as potential mineralization anomalies. Simultaneously, electrical parameters (e.g., resistivity 30~50Ω·m) and isotopic parameters (e.g., δ¹⁸O⁻¹) were compared. 34 Correlation analysis was performed on S-3.0‰±3.6‰. For example, at a sampling point in the northern part of the mining area, the resistivity was 38Ω·m and δ 34 S-3.2‰, with a correlation coefficient of 0.76 (>0.7 threshold), is identified as a high-probability mineralization feature. This type of feature is integrated into a fusion feature set and transmitted to the model training unit. Compared with traditional single-parameter identification (such as using low resistivity as the anomaly standard and misjudging graphitized rock layers), this process improves the anomaly identification accuracy to 92% by associating cross-type parameters.

[0037] The model training unit receives the fused feature set, trains it using machine learning algorithms, and outputs the trained mineral exploration prediction model to the prediction unit. When the number of mineralized and non-mineralized samples in the training samples is unbalanced, oversampling and cost-sensitive learning algorithms are used to optimize model performance. After receiving the fused feature set, this unit uses 100 known borehole data from the mining area (52 mineralized samples and 48 non-mineralized samples) as the training set and selects the random forest algorithm to construct the model. Considering the similar number of mineralized and non-mineralized samples but the large grade differences within the mineralized samples (0.3%~2.8%), resulting in a latent sample imbalance problem, this unit adopts an optimization strategy of "oversampling + cost-sensitive learning": oversampling is performed on low-grade (0.3%~1.0%) mineralized samples (expanding them to the same number as high-grade samples), while a higher misclassification cost is set for mineralized samples during algorithm training (misclassification cost of 1 for non-mineralized samples and 3 for mineralized samples) to avoid the model biasing towards non-mineralized samples; during training, model parameters are adjusted using 5-fold cross-validation (the number of decision trees is set to 200, and the maximum depth is 10). When the model's accuracy reaches 90% and recall reaches 88% on the validation set, training is stopped, and the trained mineral exploration prediction model is output to the prediction unit. The optimization strategy of the model training unit significantly improves the model's generalization ability, especially enhancing its ability to identify deep low-grade mineralized zones.

[0038] The prediction unit receives a trained mineral exploration prediction model and new exploration data, predicts mineralization potential and grade, and outputs mineralization anomaly identification results and grade prediction data to the model building module. When the resistivity of the input data is in the range of 30~50 Ω·m, the predicted copper grade is 1.0%~2.5%. After receiving the trained mineral exploration prediction model, when new exploration data (such as data from 20 newly added sampling points in the southern part of the mining area) is input, the data is first standardized and then substituted into the model to calculate the mineralization probability. At the same time, based on the "resistivity-grade" relationship model established by the boreholes in the mining area (in Document 2, resistivity of 30~50 Ω·m corresponds to a grade of 1.0%~2.5%, and 50~100 Ω·m corresponds to 0.3%~1.0%), grade prediction is performed for areas with a mineralization probability > 0.7. For example, the standardized resistivity of a certain point in the southern part of the mining area is 0.35 (corresponding to an actual 42 Ω·m), δ 34 With S = 0.4 (corresponding to an actual grade of -3.1‰), the model outputs a mineralization probability of 0.82 and a predicted grade of 1.8% to 2.2%. Subsequent drilling verification showed that the actual grade at this point was 2.0%, with a prediction error of only 0.2%.

[0039] The model building module receives mineralization anomaly identification results and grade prediction data. Using big data visualization and 3D mesh modeling techniques, it generates a multi-parameter fusion visualization map and a 3D mineralization potential model. It outputs the target area prediction results and 3D model data to the user interface terminal for easy viewing by operators. When the root mean square error of the model inversion is less than 2.5%, the model building module determines that the visualization map and 3D mineralization potential model have converged and locks the final predicted target area. The model building module includes: The 3D inversion unit receives mineralization anomaly identification results and grade prediction data, and uses finite element forward modeling and conjugate gradient inversion algorithms to perform 3D resistivity modeling. It then outputs the 3D resistivity model data to the mineralization potential assessment unit. The calculation stops and the final model is output when the inversion iteration count reaches 30 or the root mean square error is below 2.5%. By receiving mineralization anomaly identification results (such as the resistivity anomaly range of the northern mineralization zone) and grade prediction data, a combined algorithm of "finite element forward modeling + conjugate gradient inversion" is used to construct a 3D resistivity model: first, the 3D space of the mining area is divided into 50m×50m×20m grids (total 100×100×...). The model consists of 50 elements. Known borehole data (e.g., resistivity of ore body at 120-180m, 30-60Ω·m) is used as constraints. The surface is set as a free surface, and the bottom and sides as insulating boundaries. The theoretical electric field distribution is calculated through finite element forward simulation and compared with actual measured resistivity data. A conjugate gradient inversion algorithm is used to iteratively optimize the model parameters. The stopping condition is set at "30 iterations" or "root mean square error < 2.5%"—in the Zouping mining area modeling, the root mean square error decreased to 2.1% (< 2.5%) after 28 iterations, at which point the calculation stopped, and a three-dimensional resistivity model was output (the volume percentage of the low-resistivity region < 50Ω·m is 13%). Compared to traditional two-dimensional modeling (which cannot represent the three-dimensional morphology of the ore body, with a deviation of 20%), the three-dimensional model of this element has an 85% match with the ore body morphology revealed by the actual borehole, clearly presenting the lenticular ore body distribution at a depth of 300m.

[0040] The mineralization potential assessment unit receives three-dimensional resistivity model data, combines the resistivity-grade relationship model with fuzzy comprehensive evaluation method to calculate the mineralization probability of each three-dimensional grid cell, and outputs a mineralization potential distribution map to the target area delineation unit. When the mineralization probability value of a grid cell is greater than 0.7, it is marked as a first-level target area. After receiving the three-dimensional resistivity model data, the mineralization probability is calculated using the "resistivity-grade relationship" and "fuzzy comprehensive evaluation" methods: first, the resistivity value of each three-dimensional grid cell is substituted into the grade prediction model to obtain the theoretical grade; then, based on the mineralization regularity of the mining area (such as proximity to the center of the breccia pipe, δ...), the mineralization probability is calculated... 34Regions near mantle-derived sulfur have a higher mineralization probability. An evaluation system was established with a weighting of resistivity (0.4), isotopic characteristic (0.3), and tectonic location (0.3). A fuzzy comprehensive evaluation method was used to calculate the mineralization probability of each grid. Probability values ​​> 0.7 were marked as primary target areas, 0.5–0.7 as secondary target areas, and < 0.5 as non-target areas. In the Zouping mining area, this unit identified four primary target areas (total area 1.2 km²). 2 There are 3 secondary target areas, among which the mineralization probability of the No. 1 primary target area in the north reaches 0.85, which completely overlaps with the known mineralization zone. The quantitative assessment of this unit makes the target area selection more objective.

[0041] The target area delineation unit receives the mineralization potential distribution map and uses spatial clustering and boundary recognition algorithms to automatically delineate the spatial location and boundaries of the prospecting target area, outputting the final target area prediction results and a 3D visualization report. After receiving the mineralization potential distribution map, the target area delineation unit uses a "spatial clustering + boundary recognition" algorithm to automatically delineate the target area: first, the K-means clustering algorithm is used to aggregate the first-level target area grid into a continuous region, then the edge detection algorithm (Canny operator) is used to identify the target area boundary, and the surface projection range is corrected by combining topographic data, finally generating a "target area spatial location map + 3D visualization report", which is output to the user interactive terminal (such as a geological engineer's computer or mobile device). For example, after processing by this unit, the No. 1 target area in the northern part of the mining area was delineated with a planar range of 800m east-west and 600m north-south, and a three-dimensional range with a burial depth of 80~220m. Subsequent drilling in this target area revealed three copper mineralization zones (350~420m in length and 4.2~5.5m in average thickness), with a target area matching rate of 78%. Compared with the traditional manual delineation of target areas (error ±100m), the automatic identification of this unit controlled the target area boundary error within 30m, significantly reducing the cost of ineffective drilling.

[0042] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A multi-parameter prediction method for copper deposits based on land cover-geochemistry, characterized in that, Includes the following steps: Step 1, Data Acquisition: Deploy several field exploration devices in the prospecting area and collect geophysical and geochemical data of the copper prospecting area through the field exploration devices. Geophysical data includes gravity exploration, magnetic exploration and electrical exploration data, and geochemical data includes stable isotope data. Step 2, Data Processing: Preprocess the collected geophysical and geochemical data, including data cleaning, outlier removal, normalization, and interference suppression. Step 3, Model Training: Machine learning algorithms are used to train the preprocessed geophysical-geochemical multi-parameter data to establish a mineral exploration prediction model; Step 4, Visualization Analysis: Based on the mineral exploration prediction model, conduct big data visualization analysis of copper deposits to generate a multi-parameter fusion visualization map; Step 5: Construct a three-dimensional mineralization model: Combine geophysical inversion results and geochemical characteristics to predict copper mineralization potential and target area distribution and output the visualization.

2. The multi-parameter prediction method for copper deposits based on land cover-geochemistry according to claim 1, characterized in that, In step 1, electrical exploration data is collected using the high-density resistivity method and the induced polarization method, and an inversion algorithm is used to locate the mineralized zone. During the inversion process, topographic correction and resistivity-polarizability joint analysis are integrated to reduce the positioning error.

3. The multi-parameter prediction method for copper deposits based on land cover-geochemistry according to claim 2, characterized in that, In step 1, stable isotope data, including hydrogen, oxygen, carbon, and sulfur isotopes, are used to analyze the source, evolution path, and mixing ratio of ore-forming fluids, and to identify mineralization centers through the spatial distribution characteristics of isotopes.

4. The multi-parameter prediction method for copper deposits based on land cover-geochemistry according to claim 3, characterized in that, In step 2, interference suppression employs wavelet transform and filtering techniques. When the electromagnetic interference signal strength exceeds twice the effective signal strength, a db4 wavelet basis is used for multi-level decomposition, and high-frequency coefficients are subjected to hard thresholding to improve the signal-to-noise ratio.

5. The multi-parameter prediction method for copper deposits based on land cover-geochemistry according to claim 4, characterized in that, In step 3, the multi-parameter data includes the combination of CSAMT and TDIP electrical resistivity data and stable isotope data. Machine learning algorithms are used for feature fusion and anomaly identification. When the multi-parameter correlation coefficient is greater than 0.7, it is determined to be a mineralization target signal.

6. The multi-parameter prediction method for copper deposits based on land cover-geochemistry according to claim 5, characterized in that, In step 5, the three-dimensional mineralization model is constructed using finite element forward simulation and conjugate gradient inversion algorithm, combined with borehole data and geochemical potential relationship. The iteration stops when the root mean square error is less than 2.5%, and the three-dimensional resistivity model and mineralization probability distribution map are output.

7. A multi-parameter evaluation system for copper deposits based on land cover-geochemistry, based on the multi-parameter prediction method for copper deposits based on land cover-geochemistry as described in any one of claims 1-6, characterized in that, It includes a data acquisition module, a machine learning modeling module, and a model building module, among which: The data acquisition module receives raw geophysical and geochemical data transmitted from field exploration equipment. It processes the data using data cleaning, outlier removal, and interference suppression algorithms, and outputs a cleaned, standardized multi-parameter dataset to the machine learning modeling module. When electromagnetic interference exceeds twice the effective signal strength or when terrain undulations cause data distortion, the data acquisition module uses wavelet transform and terrain correction algorithms for enhancement. The machine learning modeling module receives a standardized multi-parameter dataset, uses machine learning algorithms for training and feature fusion, establishes a mineral exploration prediction model, and outputs mineralization anomaly identification results and grade prediction data to the model building module. When the multi-parameter correlation coefficient is greater than 0.7, the machine learning modeling module determines it as a high-probability mineralization target signal and strengthens the output. The model building module receives mineralization anomaly identification results and grade prediction data. It uses big data visualization and 3D mesh modeling technology to generate a multi-parameter fusion visualization map and a 3D mineralization potential model. It outputs the target area prediction results and 3D model data to the user interaction terminal for easy viewing by operators. When the root mean square error of the model inversion is less than 2.5%, the model building module determines that the visualization map and the 3D mineralization potential model have converged and locks the final predicted target area.

8. The multi-parameter evaluation system for copper deposits based on land features and geochemical analysis according to claim 7, characterized in that, The data acquisition module includes: The geophysical data acquisition unit is used to receive exploration commands, perform measurements using the high-density resistivity method and the induced polarization method, and output the raw electrical exploration data to the interference suppression unit. The geochemical data acquisition unit receives sampling commands, performs micro-area analysis of ore samples using LA-ICP-MS or μ-XRF technology, and outputs stable isotope data to the interference suppression unit. When in-situ analysis of mineral composition is required, the geochemical data acquisition unit uses synchrotron X-ray fluorescence technology for sensitivity detection. The interference suppression unit receives raw electrical exploration data and stable isotope data, performs noise reduction using wavelet transform and filtering techniques, and outputs the denoised fused data to the data standardization unit.

9. The multi-parameter evaluation system for copper deposits based on land features and geochemical analysis according to claim 8, characterized in that, The machine learning modeling module includes: The multi-parameter fusion unit receives standardized multi-parameter datasets, performs feature extraction and fusion using principal component analysis and correlation analysis, and outputs the fused feature set to the model training unit. When CSAMT and TDIP data coexist, impedance tensor decomposition is used to calculate anisotropy coefficients to identify mineralization anomalies. The model training unit receives the fused feature set, trains it using machine learning algorithms, and outputs the trained mineral exploration prediction model to the prediction unit. When the number of mineralized and non-mineralized samples in the training samples is unbalanced, oversampling and cost-sensitive learning algorithms are used to optimize the model performance. The prediction unit receives the trained mineral exploration prediction model and new exploration data, performs mineralization potential and grade prediction, and outputs mineralization anomaly identification results and grade prediction data to the model building module; when the resistivity of the input data is in the range of 30~50Ω·m, the predicted copper grade is 1.0%~2.5%.

10. The multi-parameter evaluation system for copper deposits based on land features and geochemical analysis according to claim 9, characterized in that, The model building module includes: The three-dimensional inversion unit receives mineralization anomaly identification results and grade prediction data, uses finite element forward modeling and conjugate gradient inversion algorithms to perform three-dimensional resistivity modeling, and outputs the three-dimensional resistivity model data to the mineralization potential assessment unit; when the number of inversion iterations reaches 30 or the root mean square error is less than 2.5%, the calculation stops and the final model is output. The mineralization potential assessment unit is used to receive three-dimensional resistivity model data, combine resistivity-grade relationship model and fuzzy comprehensive evaluation method to calculate the mineralization probability of each three-dimensional grid cell, and output a mineralization potential distribution map to the target area delineation unit; when the mineralization probability value of a grid cell is greater than 0.7, it is marked as a first-level target area. The target area delineation unit receives the mineralization potential distribution map, uses spatial clustering and boundary recognition algorithms to automatically delineate the spatial location and boundaries of the prospecting target area, and outputs the final target area prediction results and a 3D visualization report.

Images