Artificial Intelligence-Based Method for Locating and Predicting Quartz Vein-Type Gold Deposits

Abstract

This invention relates to the field of mineral deposit location and prediction technology, specifically to an artificial intelligence-based method for locating and predicting quartz vein-type gold deposits. The method includes a surrounding rock monitoring and acquisition module, a mineralization feature construction module, a negative sample screening module, a negative sample generation module, and a boundary correction analysis module. This invention obtains multi-dimensional geological information of spatial points in the surrounding rock in real time through automatic acquisition and linked analysis. Combining signal segmentation and parameter classification, it unifies the spatial expression of metal signals and strain characteristics. A data system is established using spatial numbering and time labels. By comparing the changing trends of gold elements and various mineral parameters, it achieves accurate characterization of mineralization features in continuous space. Thickness and confidence analysis are introduced to optimize negative sample generation and screening, dynamically adjusting the spatial distribution structure of negative samples. Based on the inference results, it automatically verifies and corrects boundary areas, achieving spatial consistency in ore body distribution and high reliability of prediction results.

Description

Technical Field

[0001] This invention relates to the field of mineral deposit location and prediction technology, and in particular to a method for locating and predicting quartz vein-type gold deposits based on artificial intelligence. Background Technology

[0002] Mineral deposit location prediction involves the scientific inference and spatial positioning of the distribution of metallic mineral resources, especially gold ore bodies, in the Earth's crust. This technical field is based on multi-source data fusion and modeling, combining existing mineral deposit data with regional metallogenic background, and using quantitative and model-based methods to assist exploration personnel in identifying potential mineral deposit target areas over a wide region. Traditional methods for locating and predicting quartz vein-type gold deposits refer to techniques that use geological mapping, geochemical anomaly identification, and remote sensing image interpretation to determine the possible distribution areas of quartz vein-type gold deposits. These methods are typically based on field geological surveys, collecting information on stratigraphy, structure, alteration, and mineralization in the mining area, and combining this with sample analysis data and historical prospecting results to delineate target areas and predict ore bodies.

[0003] Existing technologies for locating and predicting quartz vein-type gold deposits largely rely on field geological surveys and chemical sample analysis. Data acquisition is primarily conducted in a single manner, resulting in limited spatial resolution and data synchronization. The dynamic response of acquired parameters is weak, and changes in the mineralization degree of the surrounding rock are difficult to reflect in real time. Negative sample screening lacks dynamic judgment of geological field changes, and the sample set is prone to mislabeling and distribution imbalance. Ore body boundary inference relies on static data and subjective experience, and the predicted target area is prone to boundary ambiguity or error accumulation. In actual operation, the accuracy of deposit location and the effectiveness of exploration decision-making are affected. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an artificial intelligence-based method for locating and predicting quartz vein-type gold deposits.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for locating and predicting quartz vein-type gold deposits based on artificial intelligence, the method comprising:

[0006] The surrounding rock monitoring and acquisition module is based on the surrounding rock monitoring points in the quartz vein type gold mine area. It analyzes the collected electrical and mechanical signals, performs signal segmentation and feature classification, and maps the gold element concentration and strain characteristics to the borehole number and coordinates to obtain the surrounding rock monitoring parameter set.

[0007] The mineralization feature construction module, based on the surrounding rock monitoring parameter set, compares the changes in gold and iron element signals at monitoring points, analyzes the distribution of gold and silicon signals, combines borehole strain data, determines the changes in mineralization intensity, identifies abnormal zones, and obtains mineralization feature distribution parameters.

[0008] The negative sample screening module selects borehole thickness variation data based on the mineralization characteristic distribution parameters, analyzes the relationship between mineralization indicators and thickness trends, compares the target with the performance of adjacent collection points, determines whether it is in the screening interval, and obtains a high-confidence screening number.

[0009] The negative sample generation module analyzes the confidence performance of the collection points based on the high confidence screening number, maps the confidence results to the sample generation control, and adjusts the sample point generation state in combination with random signals to obtain dynamic negative sample distribution information.

[0010] Based on the dynamic negative sample distribution information, the boundary correction analysis module analyzes the collection point number and time label, compares the predicted boundary coordinates with the collection point distribution, determines whether the boundary has shifted, re-analyzes the associated features, and obtains the boundary shift correction parameters.

[0011] The present invention is improved in that the surrounding rock monitoring parameter set includes the three-dimensional coordinates of the monitoring point, the original geological signal code, and the monitoring data label; the mineralization characteristic distribution parameters include the mineralization zoning code, the mineralization distribution level, and the abnormal mineralization identifier; the high confidence screening number includes the negative sample candidate number, the spatial positioning label, and the confidence screening level; the dynamic negative sample distribution information includes the sample spatial distribution information, the generation time series record, and the confidence label; and the boundary offset correction parameters include the offset detection result, the correction coordinate set, and the correction rule parameters.

[0012] The present invention is improved in that the surrounding rock monitoring and acquisition module includes:

[0013] The data stream receiving submodule is based on the monitoring points of the surrounding rock in the quartz vein gold mine area. It analyzes the electrical and mechanical signals collected by the geological sensors at the monitoring points, classifies and organizes the signal channel numbers and acquisition times, judges the consistency of channel data during the data merging process, and obtains the signal dataset.

[0014] The signal feature segmentation submodule calculates the time series and amplitude changes of each channel signal based on the signal dataset, compares the peak amplitude characteristics, periodic distribution and frequency band spread of the waveform in the signal segment, identifies the category labels of the signal features, adjusts the feature segmentation order, and obtains the signal feature sequence.

[0015] The monitoring point mapping and identification submodule performs attribution processing on the gold element concentration data and the surrounding rock strain change data based on the signal feature sequence, maps the data to the monitoring point borehole number and the three-dimensional coordinates of the ground surface, adjusts the order of the signals in the preprocessing process, and obtains the surrounding rock monitoring parameter set.

[0016] The present invention is improved in that the mineralization feature construction module includes:

[0017] The signal amplitude comparison submodule analyzes the signal amplitudes of gold and iron at each monitoring point based on the surrounding rock monitoring parameter set, calculates the differences between the two types of element signals in the spatial sequence, compares the signal change direction and continuous change between each monitoring point, judges the change of trend turning point in coordinate distribution, and obtains the gold and iron signal change trend factor.

[0018] Based on the gold and iron signal change trend factor, the feature fusion calculation submodule calculates the signal amplitudes of gold and silicon elements at the monitoring points, analyzes the combined characteristics of the two elements in three-dimensional coordinates, filters the surrounding rock strain change data combined with the borehole record, optimizes the process of fusing metal signal features and strain parameters, and obtains a multi-parameter fusion feature group.

[0019] The anomaly identification submodule, based on the multi-parameter fusion feature group, determines the fusion feature change status of each acquisition point, identifies the fusion parameter change range between spatially continuous acquisition points, analyzes data segments whose change amplitude exceeds the normal fluctuation of continuous areas, and obtains mineralization feature distribution parameters.

[0020] The present invention is improved in that the negative sample screening module includes:

[0021] Based on the mineralization characteristic distribution parameters, the profile thickness analysis submodule analyzes the borehole profile thickness variation data, compares the continuous thickness variation of the target acquisition point with that of adjacent acquisition points, determines the direction of change and calculates the thickness difference, identifies the associated acquisition points as the result, and obtains the thickness change rate sequence.

[0022] The mineralization index extraction submodule, based on the thickness change rate sequence, compares the gold element signals of the target sampling point with those of adjacent points, calculates the proportional relationship between the sum of the signals at each point and the thickness change, using the formula:

[0023] ;

[0024] The distribution values ​​of mineralization indicators were obtained, and the amplitude of the gold element signal was integrated with the thickness change rate to obtain the mineralization indicator distribution set, where, This represents the distribution value of mineralization index at the nth sampling point. This represents the amplitude of the gold element signal at the nth sampling point. This represents the amplitude of the gold element signal at the (n-1)th sampling point. This represents the amplitude of the gold element signal at the (n+1)th sampling point. This represents the rate of change of thickness at the nth sampling point. This is to avoid the denominator being zero due to zero thickness disturbance, and at the same time to suppress the excessive amplification of the index by high disturbance;

[0025] The screening interval determination submodule filters the index performance of each collection point based on the mineralization index distribution set, determines whether it meets the screening criteria, and organizes the collection point numbers that pass the screening to obtain high confidence screening numbers.

[0026] The present invention is improved in that the negative sample generation module includes:

[0027] The confidence assessment submodule, based on the high confidence screening number, identifies the spatial region of signal fluctuation by comparing the mineralization characteristics and profile stratigraphic changes of each monitoring point, calculates the changes in mineralization indicators between the monitoring point and its neighboring points, and obtains confidence intensity configuration data.

[0028] Based on the confidence intensity configuration data, the sample generation control submodule analyzes the correlation between the confidence level, spatial number and disturbance factor of each monitoring point, calculates the mineralization difference and profile stratigraphic parameters between the monitoring point and adjacent points, determines the disturbance amplitude distribution of each monitoring point, and obtains the sampling control parameter set.

[0029] The generation state adjustment submodule, based on the sampling control parameter set, filters monitoring points that meet the correlation conditions between disturbance distribution and confidence level, determines the correspondence between the disturbance state and confidence marker of each point, adjusts the sample point generation state, and obtains dynamic negative sample distribution information.

[0030] The present invention is improved in that the boundary correction analysis module includes:

[0031] The boundary comparison and judgment submodule obtains the spatial number and time label of each collection point based on the dynamic negative sample distribution information, analyzes the spatial arrangement of the predicted boundary coordinates, compares the spatial relationship between the actual collection point coordinates and the predicted boundary, judges the distribution segments with inconsistent coverage, and obtains the boundary offset recognition result.

[0032] Based on the boundary offset identification results, the retrospective analysis submodule calls the gold element signal and surrounding rock strain characteristics within the spatial segment to calculate the combined changes of the two types of parameters in the offset area, analyzes the synchronicity between the combined change trend and the boundary prediction results, and obtains the characteristic change sequence of the offset area.

[0033] Based on the feature change sequence of the offset region, the acquisition point adjustment submodule determines the correspondence between the current classification of the acquisition point and the direction of signal feature change, optimizes the label information of the acquisition point in the classification process, adjusts the status identifier of the acquisition point in the sample set, and obtains the boundary offset correction parameters.

[0034] The present invention is improved in that the segmentation identification and feature classification are to divide the collected raw signal data according to time, depth or geographical location, and to classify and organize the data in different intervals through feature extraction. The abnormal zone refers to the spatial region that is different from the surrounding mineralization state after feature trend analysis.

[0035] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0036] In this invention, multidimensional geological information of spatial points of the surrounding rock is obtained in real time through automatic acquisition and linkage analysis. By combining signal segmentation and identification and parameter classification, the spatial expression of metal signals and strain characteristics is unified. A data system is established using spatial numbering and time labels. By comparing the changing trends of gold elements and various mineral parameters, the mineralization characteristics in continuous space are accurately characterized. Thickness and confidence analysis are introduced to optimize the generation and screening of negative samples, dynamically adjust the spatial distribution structure of negative samples, and automatically check and correct boundary areas based on inference results, so as to achieve spatial consistency of ore body distribution and high reliability of prediction results. Attached Figure Description

[0037] Figure 1 This is a flowchart of the method of the present invention;

[0038] Figure 2 This is a flowchart of the surrounding rock monitoring and acquisition module in this invention;

[0039] Figure 3 This is a flowchart of the mineralization feature construction module in this invention;

[0040] Figure 4 This is a flowchart of the negative sample screening module in this invention;

[0041] Figure 5 This is a flowchart of the negative sample generation module in this invention;

[0042] Figure 6 This is a flowchart of the boundary correction analysis module in this invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and 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, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0045] All user-related information involved in this invention (including but not limited to biometric information, identity verification information, behavioral data, device information, and other data that can be used for identity verification and personalized services) is collected and processed with the user's full knowledge and voluntary consent. The collection, storage, and use of all information strictly comply with applicable national and regional laws and regulations, and meet relevant data protection standards and policy requirements. The use of data is limited to purposes necessary for providing the technical services of this invention, and reasonable technical and management measures will be taken to ensure the security and confidentiality of users' personal information in terms of information protection and privacy.

[0046] Example: Please refer to Figure 1 This invention provides a technical solution for the location and prediction method of quartz vein-type gold deposits based on artificial intelligence, comprising:

[0047] The surrounding rock monitoring and acquisition module is based on the surrounding rock monitoring points in the quartz vein type gold mine area. It analyzes the electrical and mechanical signals collected by the geological sensors, identifies the signal waveforms in segments and classifies their features, and assigns the gold element concentration and surrounding rock strain characteristics to the borehole number and surface coordinates, respectively. It adjusts the sequence of signal conversion, identifies each monitoring point, and obtains the surrounding rock monitoring parameter set.

[0048] The mineralization feature construction module is based on the surrounding rock monitoring parameter set. It compares the changing trends of the gold and iron signal amplitudes at each monitoring point, analyzes the characteristic distribution of gold and silicon signals, combines the surrounding rock strain characteristics recorded in the borehole, merges the metal features and strain features into the data of the same collection point, determines the way the mineralization intensity changes between collection points, identifies abnormal change zones, and obtains the mineralization feature distribution parameters.

[0049] The negative sample screening module filters borehole profile thickness variation data based on mineralization characteristic distribution parameters, analyzes the proportional relationship between mineralization index and profile thickness variation trend at each point, compares the performance of the target collection point with the five adjacent collection points, determines whether the screening interval conditions are met, and extracts the collection point numbers that meet the conditions to obtain high confidence screening numbers.

[0050] The negative sample generation module uses high-confidence screening numbers to determine the confidence performance of each collection point, maps the confidence analysis results to the sample generation controller, generates random signals, and adjusts the sample point generation state according to the correspondence between confidence performance and random signals to obtain dynamic negative sample distribution information.

[0051] The boundary correction analysis module analyzes the spatial number and time label of the acquisition points based on the dynamic negative sample distribution information, compares the predicted boundary coordinates with the acquisition point distribution, and determines whether there is a boundary offset within the spatial range. If an offset exists, the metal information and strain characteristics are re-analyzed, the acquisition point judgment steps are adjusted, and the boundary offset correction parameters are obtained.

[0052] The surrounding rock monitoring parameter set includes the three-dimensional coordinates of the monitoring points, the original geological signal code, and the monitoring data label. The mineralization characteristic distribution parameters include the mineralization zoning code, the mineralization distribution level, and the abnormal mineralization identifier. The high confidence screening number includes the negative sample candidate number, the spatial positioning label, and the confidence screening level. The dynamic negative sample distribution information includes the sample spatial distribution information, the generation time series record, and the confidence label. The boundary offset correction parameters include the offset detection results, the correction coordinate set, and the correction rule parameters.

[0053] In the surrounding rock monitoring and acquisition module, electrical signals refer to the raw current, voltage, or charge data related to elements such as gold in the geology, measured by geochemical / electrical sensors (such as inductive electrodes, geochemical collectors, etc.) installed in boreholes or on the surface of the mining area. These signals reflect the distribution of metallic elements in the ore body or surrounding rock. Mechanical signals refer to the data obtained by seismic, stress, or deformation monitoring sensors (such as accelerometers, strain gauges, etc.) deployed at geological monitoring points, reflecting changes in the micromotion, vibration, or mechanical state of the strata. These signals directly reflect the response of the surrounding rock structure under underground stress. Segmentation identification and feature classification divide the acquired raw signal data into segments according to time, depth, or geographical location, and extract features (such as waveform analysis, peak value judgment, and statistics). Data from different intervals is categorized and organized (e.g., quantity calculation, etc.) to facilitate subsequent correlation with spatial distribution; gold element concentration refers to the abundance of gold (Au) in soil, rock, or groundwater samples from a specific borehole or sampling point based on geochemical data obtained from the collection point, usually expressed in standard concentration units (e.g., ppm, mg / kg); surrounding rock strain characteristics refer to the deformation or strain behavior parameters of the unmineralized rock strata surrounding the gold ore body obtained by devices such as strain gauges under underground stress, used to characterize the mechanical stability or activity of the surrounding rock; the signal conversion process sequence refers to the technical process of sorting the raw electrical and mechanical signals from the sensor through preprocessing steps such as amplification, analog-to-digital conversion, noise reduction, and feature extraction to ensure data availability and the correspondence of various parameters.

[0054] In the mineralization characteristic construction module, the trend of change describes the continuous change of a certain type of data (such as the amplitude of gold and iron element signals) at different monitoring points with variables such as spatial location and depth, used to determine the enrichment or dilution trend of the ore body; the characteristic distribution refers to the statistical analysis of the distribution pattern of a certain type of signal (such as gold and silicon elements) at all monitoring points in the geological space, revealing the spatial structure and boundary characteristics of different mineralization areas; the metallic characteristics refer to the set of metallic element signals closely related to the location of gold deposits (such as specific combinations or proportional relationships of gold, iron, and silicon), reflecting mineralization or ore-forming processes. The geochemical properties of the collection; strain characteristics refer to the specific manifestations of deformation signals related to geomechanics (such as stress changes, strain rates, etc.) at the borehole monitoring points, which are important bases for judging the structural activity of mineralized and non-mineralized zones; the way mineralization intensity changes describes the variation law of mineralization degree between collection points, which is a basic parameter for ore body enrichment and boundary identification, such as the gradient trend from dense to sparse or from sparse to dense; the anomalous change zone refers to the spatial area that is significantly different from the surrounding mineralization state after analyzing the characteristic trends, which is often a newly discovered mineralization anomaly or gold ore boundary zone.

[0055] In the negative sample screening module, borehole profile thickness variation data refers to the thickness variation records of different strata, ore layers, or surrounding rocks obtained from borehole geological profiles (i.e., drill cores, stratigraphic distribution, etc.), used to analyze the spatial structure of ore body distribution; mineralization index refers to comprehensive quantitative indicators (such as gold content, mineralization intensity ratio, etc.) extracted based on mineralization characteristic distribution parameters, used to reflect the mineralization probability of each collection point; profile thickness variation trend refers to analyzing the variation trend of different rock layer thicknesses along the depth or spatial variation of the borehole, and judging the sedimentary or tectonic background of the mineralization zone; collection point performance refers to the comprehensive reflection of the target collection point under multi-dimensional parameters such as mineralization index and thickness variation, which is the basis for identifying the sample as a negative sample, a high-confidence sample, or an excluded sample; screening interval conditions refer to the parameter range set for screening high-confidence negative samples, used to judge whether the performance of the collection point enters the selected interval, such as mineralization index being below a certain level and thickness variation being within a certain range.

[0056] In the negative sample generation module, the confidence level refers to the reliability and trustworthiness level of each collection point as a negative sample, calculated based on multi-dimensional data such as mineralization characteristics, spatial relationships, and thickness variations. The sample generation controller is a logic control unit used to receive confidence analysis results and allocate negative sample generation tasks; it can be a software logic module or an instruction trigger integrated with hardware. The random signal is a random factor introduced by the controller when allocating sample generation to prevent the sample distribution from becoming mechanically rigid; it is usually implemented through a system random number generator or a physical random circuit. The correspondence refers to the pairing, operation, or judgment mechanism between the confidence level and the random signal, which determines whether a collection point is ultimately included in the negative sample. The sample point generation status specifically describes whether the collection point is currently marked / included as a new round of negative samples, and is used to dynamically adjust the sample set.

[0057] In the boundary correction analysis module, spatial number and time label are unique identifiers for each collection point in the three-dimensional space of the mining area, along with the specific time markers of its collection, generation, and correction operations, used for subsequent tracing and correction. Predicted boundary coordinates refer to spatial location data such as the ore body boundary line and the outer envelope of the mineralized area based on the inference output, often presented in the form of three-dimensional coordinates or GIS vector data. Collection point distribution refers to the specific arrangement and distribution of all monitoring collection points in the mining area, serving as a key spatial basis for boundary correction and model accuracy adjustment. Boundary offset refers to determining whether the model's predicted boundary has shifted or drifted relative to the actual geological boundary by analyzing the spatial overlap between the predicted boundary and the actual collection point distribution. The collection point determination step refers to readjusting and optimizing the original negative / positive sample determination process for collection points based on the offset analysis results, ensuring a high degree of consistency between the model and the actual ore body distribution.

[0058] Please see Figure 2 The surrounding rock monitoring and acquisition module includes:

[0059] The data stream receiving submodule is based on the monitoring points of the surrounding rock in the quartz vein gold mine area. It analyzes the electrical and mechanical signals collected by the geological sensors at the monitoring points, classifies and organizes the signal channel numbers and acquisition times, judges the consistency of channel data during the data merging process, and obtains the signal dataset.

[0060] First, confirm the geographical layout and borehole numbers of the monitoring points. For example, in a gold mine area, boreholes D101 to D120 are deployed. Each borehole is equipped with an electrical sensor and a mechanical stress sensor. The electrical sensor records electrical signals related to metallic elements in the geology, such as voltage changes collected by inductive electrodes. The mechanical sensor records the deformation state of the surrounding rock, such as strain gauges monitoring changes in the strain value of the surrounding rock over a certain period of time. During signal acquisition, all sensors record the sampling time through a unified time synchronization mechanism. When the signal data is written to the database, it is arranged according to the channel number and corresponding timestamp. For example, the electrical signal record of channel CH-001 at 9:32:01:560 AM is recorded as an independent data record. At the same time, the data of each channel is divided into time series in seconds or milliseconds to ensure that the data of different channels under the same time can be compared. In the process of classifying and organizing the signal channel numbers and acquisition times... During the process, each data acquisition point needs to be checked to ensure its channel ID is unique. If two data records have the same channel number and overlap in time, their values ​​are compared to see if they are within the set deviation threshold. For example, if the voltage difference between two signals is greater than 0.05 volts, or the strain value difference is greater than 10 microstrains, the channel data is considered inconsistent and one or both must be discarded. When the proportion of missing or abnormal data exceeds 5%, the entire data segment of that channel is marked as unreliable. The continuity of the time series also needs to be checked during the merging process. If the sampling interval of a certain channel data segment is greater than the specified 2 milliseconds, the data segment is considered to have a time jump and the jump segment needs to be discarded or split. Complete structured data records are formed by combining the time series of each channel data, monitoring point number, signal type, and value to form a signal dataset. This dataset contains continuous records of all electrical and mechanical signals of each monitoring point at different time periods.

[0061] The signal feature segmentation submodule calculates the time series and amplitude changes of each channel signal based on the signal dataset, compares the peak amplitude characteristics, periodic distribution and frequency band spread of the waveform in the signal segment, identifies the category labels of signal features, adjusts the feature segmentation order, and obtains the signal feature sequence.

[0062] The complete time series and corresponding voltage or strain value changes for each channel are read. First, the signal of each channel is divided into several continuous segments based on time length, for example, every 30 seconds. The values ​​in each segment are scanned, and the difference between the maximum and minimum values ​​is extracted. If this difference exceeds a set range, such as 0.8 volts or more, the segment is marked as a peak amplitude signal segment. Simultaneously, the regularity of the periodic distribution in the waveform of this segment is determined by calculating the time interval between two adjacent peaks to obtain the main period of the segment. If the variation between periods exceeds a set range of 20%, the segment is marked as a periodic fluctuation segment. Finally, the frequency components of each signal segment are calculated using Fast Fourier Transform. If the proportion of high-frequency signals in a segment exceeds 70%, it is marked as a bandwidth spread segment. The marking result will be mapped to the category label of the signal characteristics, such as peak stable type, periodic unstable type, or bandwidth spread type. When adjusting the division order, signals of the peak stable type are processed first, then periodic unstable type, and finally bandwidth spread type, so as to achieve a standardized arrangement of the division order. During the adjustment of the order, if a signal segment spans multiple classification features, the category that appears first is taken as the main category. The signal feature sequence generated by this channel records the start and end time of each segment, the main feature type, the amplitude of numerical change, the amplitude of periodic fluctuation, and the feature description composed of bandwidth energy characteristics, which are used for subsequent spatial mapping with monitoring point information.

[0063] The monitoring point mapping and identification submodule performs attribution processing on gold element concentration data and surrounding rock strain change data based on signal feature sequence, maps the data to monitoring point borehole number and three-dimensional coordinates on the surface, adjusts the order of signals in the preprocessing process, and obtains the surrounding rock monitoring parameter set.

[0064] The system reads the feature label and channel number of each signal segment in the signal feature sequence, and assigns the signal segment to a designated monitoring point according to the pre-established correspondence between channel and monitoring point number. For example, CH-001 is assigned to borehole D101. For signal segments marked as peak-stable in electrical signals, the average amplitude and standard amplitude range are extracted, and combined with experimental calibration parameters, the gold element concentration value is converted and a timestamp is assigned to the current signal segment. Similarly, for signal segments marked as periodic fluctuation in mechanical signals, the initial and final strain values ​​are extracted, and the strain rate per unit time is calculated as the surrounding rock strain change data. The two are then bound to the monitoring point to which the signal segment belongs, completing the assignment process. Afterwards, the system is adjusted according to the time order of the signals in all assigned signal data of the monitoring point, so that the segments with prominent features in the mechanical signals are prioritized, which facilitates the extraction of high strain areas in subsequent modules. The resulting surrounding rock monitoring parameter set includes the number of each monitoring point, three-dimensional spatial coordinates, calculated gold element concentration, strain rate, and feature classification results, providing corresponding geological data support for subsequent mineralization intensity judgment.

[0065] Please see Figure 3 The mineralization feature construction module includes:

[0066] The signal amplitude comparison submodule is based on the surrounding rock monitoring parameter set. It analyzes the signal amplitudes of gold and iron at each monitoring point, calculates the differences between the two element signals in the spatial sequence, compares the signal change direction and continuous change between each monitoring point, judges the change of trend turning point in coordinate distribution, and obtains the gold and iron signal change trend factor.

[0067] The amplitude values ​​of gold and iron signals at each monitoring point are extracted. When classifying the gold signal, the segment marked as peak-stable in the electrical signal is selected, and the average amplitude of this segment is used as the representative value of the gold signal. For example, at monitoring point D101, the average amplitude of the electrical signal segment is 3.80 volts, and at D102 it is 3.12 volts. Similarly, the average amplitude of the iron signal is obtained from the iron-related segments marked in the electrical signal, such as 1.25 volts at D101 and 1.60 volts at D102. Then, the difference between the gold and iron signal amplitudes at each monitoring point is calculated by directly subtracting the iron signal from the gold signal value. The results are then arranged into a spatial sequence for the corresponding monitoring points. Subsequently, the direction of change of the signal difference in this spatial sequence is compared. For example, if the difference value gradually decreases from D101 to D105, from 2.55 to 1.80, it is determined to be a signal change. When judging continuous changes, the fluctuation range of the difference value is calculated for any three consecutive monitoring points. If the change value is within 0.5 volts and the direction of increase or decrease of the difference value is consistent, it is judged as a stable continuous change area. If the difference value of the intermediate point is opposite in direction, it is judged as a trend turning point. In this process, the minimum threshold of numerical change for turning point judgment is set to 0.3 volts. If the difference value of a monitoring point changes more than the threshold compared with the previous monitoring point and is inconsistent with the direction before and after, it is confirmed as the turning point coordinate of the spatial trend. For example, the difference value of D105 is 1.80, D106 is 2.20, and D107 is 1.90. D106 is the turning point. The three-dimensional coordinates of this point are used as the record of the turning point in the coordinate distribution. After completing the trend judgment for the monitoring points of the entire area, the coordinate information of all change directions, continuous change segments and turning points is extracted to form the trend factor of the gold and iron signal change.

[0068] The feature fusion calculation submodule calculates the signal amplitudes of gold and silicon elements at monitoring points based on the gold and iron signal change trend factor, analyzes the combination characteristics of the two elements in three-dimensional coordinates, filters and combines the surrounding rock strain change data of the borehole record, optimizes the process of fusing metal signal features and strain parameters, and obtains a multi-parameter fusion feature group.

[0069] For each trend-related monitoring point, the amplitude values ​​of gold and silicon signals are extracted. In the electrical signal dataset, the amplitude of the silicon signal is obtained in the same way as that of the gold signal, taking the average value from the stable segment already classified as a silicon-related channel. For example, at monitoring point D108, the average value of the gold signal is 3.30 volts, and the average value of the silicon signal is 4.80 volts. When calculating their combined characteristics, the amplitude values ​​of the two are constructed into a two-dimensional vector group, which is then mapped to the three-dimensional spatial coordinates of the monitoring point. The distribution of gold-silicon combined characteristics of all monitoring points is statistically analyzed under the spatial coordinates. It is determined whether the amplitude difference between gold and silicon in the same spatial region is greater than the set fusion judgment threshold of 1.5 volts. If such a difference exists in three consecutive points, it is marked as a feature abrupt change area. Combined with the surrounding rock strain change data of the borehole record, mechanical signals are called. The time series results of strain rate in the data, such as the strain rate of D108 being 15.2 microstrains per second, form a ternary feature group corresponding to the difference between it and the gold-silicon combination. This feature group is then fused to construct a unified multidimensional structure from the gold, silicon, and strain rate, and normalized to unify the dimensions. Subsequently, in the feature fusion, the influence of each item is assigned according to the weight coefficients. For example, the weight of the gold signal is set to 0.4, the silicon signal to 0.4, and the strain rate to 0.2. These coefficients are set based on the performance of previous historical samples and the suggestions of domain experts. Among them, the gold and silicon signals contribute more to the judgment of mineralization trend, followed by the strain performance. The fused structure is used to output the multi-parameter fused feature group of the monitoring points. The structure includes spatial coordinates, three numerical values, and fusion coefficient results.

[0070] The anomaly identification submodule is based on multi-parameter fusion feature groups to determine the fusion feature change status of each collection point, identify the fusion parameter change range between spatially continuous collection points, analyze data segments whose change amplitude exceeds the conventional fluctuation of continuous areas, and obtain mineralization feature distribution parameters.

[0071] Extract the fusion result value of each collection point and determine its numerical change status with adjacent collection points. In implementation, the judgment window is set to five continuous monitoring points. The fusion feature change value is compared in groups of sliding windows. The difference between the fusion values ​​of any two adjacent points is compared. If the difference exceeds the set threshold of 0.8, the point is recorded as a mutation point. During the sliding comparison, if more than three mutation points appear in a five-point window, it is considered that there is a discontinuous change in the area. If the overall fluctuation of the fusion value of the five points does not exceed 0.4, it is considered a normal fluctuation area. The average change amplitude between the mutation area and the normal area is compared to further confirm whether there is an abnormal cluster in space. The spatial continuity of the cluster is analyzed. If there are more than five consecutive mutation points and they fall on the same spatial axis, such as the X-axis, the continuous area is marked as an abnormal zone. At the same time, the three-dimensional coordinates, fusion values ​​and the average difference between the points in the zone and the normal area are recorded. The output mineralization feature distribution parameters include: the start and end coordinates of each abnormal section, the number of participating points, the mutation amplitude of the fusion feature, the rate of change, the spatial arrangement direction and the comparison value with the adjacent normal area, forming a mineralization abnormal area record.

[0072] Please see Figure 4 The negative sample screening module includes:

[0073] The profile thickness analysis submodule analyzes borehole profile thickness variation data based on mineralization characteristic distribution parameters, compares the continuous thickness variation of the target acquisition point with that of adjacent acquisition points, determines the direction of change and calculates the thickness difference, identifies associated acquisition points as results, and obtains a thickness variation rate sequence.

[0074] Extract mineralization anomaly identifiers, distribution levels, and spatial codes from all boreholes within the corresponding area. Locate the spatial position of each borehole in three-dimensional coordinates and its relative relationship within the anomaly area using parameters. Retrieve numerical data on rock or mineral layer thickness from the borehole geological profile records. Extract thickness variation values ​​for target acquisition points by recording layers from the surface downwards, constructing thickness variation curves corresponding to the boreholes. For example, the surrounding rock thicknesses recorded in boreholes D201 to D206 are 12.0 m, 14.3 m, 15.0 m, 13.8 m, 11.2 m, and 10.7 m, respectively. When comparing target acquisition point D204, first extract the thickness values ​​of its two adjacent acquisition points D203 and D205. Compare the differences between the three thicknesses sequentially. The difference between D204 and D203 is calculated to be -1.2 m, and the difference between D205 and D206 is... The difference between 204 and 204 is -2.6 meters, indicating that the thickness change direction is gradually thinning. In the direction identification, the judgment criteria are set as follows: if the thickness difference values ​​of at least three consecutive points have the same sign and the absolute value of the difference is not less than 0.5 meters, the direction is considered valid; otherwise, the point is marked as having no direction attribute. The thickness difference calculation is to subtract the thickness of every two adjacent collection points and take the absolute value to form a set of thickness differences. A preliminary judgment is made on whether there is an abrupt change in the thickness difference value. If the thickness difference exceeds 3.0 meters in two of the three consecutive points, it is marked as an abrupt change interval. A thickness change rate sequence is established among all monitoring points. This sequence forms a spatial change gradient based on the thickness difference and spacing between every two points. The adjacent corresponding borehole number, change direction, and specific thickness difference of each node are marked in the sequence to form continuous structural data that can be used for subsequent screening.

[0075] The mineralization index extraction submodule, based on the thickness change rate sequence, compares the gold element signals of the target sampling point with those of adjacent points, calculates the proportional relationship between the sum of the signals at each point and the thickness change, using the formula:

[0076] ;

[0077] The distribution values ​​of mineralization indicators were obtained, and the amplitude of the gold element signal was integrated with the thickness change rate to obtain the mineralization indicator distribution set, where, This represents the distribution value of mineralization index at the nth sampling point. This represents the amplitude of the gold element signal at the nth sampling point. This represents the amplitude of the gold element signal at the (n-1)th sampling point. This represents the amplitude of the gold element signal at the (n+1)th sampling point. This represents the rate of change of thickness at the nth sampling point. This is to avoid the denominator being zero due to zero thickness disturbance, and at the same time to suppress the excessive amplification of the index by high disturbance;

[0078] The mineralization index distribution value refers to the ratio of the sum of the gold element signal amplitude of the nth sampling point and its adjacent sampling points before and after it to the thickness change rate of the point (absolute value plus 1). It reflects the coupling characteristics of gold element abundance and geological structure change within the local spatial range of the sampling point and is used to measure the relative relationship between mineralization enrichment and stratigraphic variation at the point and its surrounding area.

[0079] The original amplitude data of the gold element signal from the target acquisition point n and its adjacent acquisition points n-1 and n+1 are retrieved, and the original amplitudes are acquired sequentially as follows: , , The original value of the thickness change rate is To ensure uniform dimensions for all participating parameters, minimum-maximum normalization is adopted. The minimum value for the gold element signal amplitude is 420, and the maximum value is 510. The minimum value for the thickness change rate is 0.2, and the maximum value is 1. The normalization calculation is as follows:

[0080] ;

[0081] ;

[0082] ;

[0083] ;

[0084] The normalized amplitude of the gold element signal at the (n-1)th sampling point. The normalized amplitude of the gold element signal at the nth sampling point. The normalized amplitude of the gold element signal at the (n+1)th sampling point. This refers to the normalized thickness change rate at the nth sampling point. Substituting the normalized value into the formula:

[0085] Molecular part: ;

[0086] Denominator part: ;

[0087] Calculate the distribution values ​​of mineralization indicators: ;

[0088] When the distribution value of mineralization index satisfy At that time, it was determined that the sampling point was located in the effective range of the mineralization enrichment zone. This range represents a strong comprehensive performance of the gold element signal amplitude and thickness disturbance, and is suitable for further screening as a candidate mineralization anomaly point.

[0089] when At that time, it was determined that the mineralization at the collection point was weak and it was not included in the subsequent screening cohort;

[0090] when At that time, it was determined that the collection point was an abnormally enriched area, and further analysis of the cause of the anomaly was needed in conjunction with on-site geological investigation.

[0091] satisfy Therefore, the sampling point n was determined to be a valid mineralization enrichment point. This result indicates that the sampling point can enter the subsequent high-confidence screening and numbering process and participate in further spatial distribution analysis as a valid spatial candidate point.

[0092] The screening interval determination submodule is based on the mineralization index distribution set, filters the index performance of each collection point, determines whether it meets the screening criteria, and organizes the collection point numbers that pass the screening to obtain high confidence screening numbers.

[0093] Based on the mineralization index distribution set, key index values ​​for each sampling point are extracted from the fused feature set and thickness change rate sequence output in the previous stage. Key indicators include gold concentration, strain rate, thickness change rate, and mineralization intensity level. First, all indicators are normalized, converting data of different dimensions into relative scores. The gold concentration is set to 100 with the maximum value as a reference, and the others are scaled proportionally. Strain rate and thickness change rate are processed similarly. Then, screening criteria are set. Points with a gold concentration below 30, a strain rate below 8 microstrains per second, and a thickness change rate below 0.3 are considered low-mineralization index points. During the process of determining whether they meet the screening criteria, each point is checked individually. If a sampling point meets all three of the above conditions, it is excluded if any one of them exceeds the upper limit. The specific item exceeding the limit and its corresponding index value are recorded during the judgment. In one example, the gold concentration of point D204 is 22, the strain rate is 6.1, and the thickness change rate is 0.21. All the indicators are lower than the screening standard, so it is judged to meet the screening requirements. When sorting the sampling point numbers that have passed the screening, all points that meet the conditions are sorted by spatial number and a screening number list is constructed. The three-dimensional coordinates of each point and its original and normalized index values ​​are recorded for the confidence allocation of the subsequent negative sample generation module. The set of sampling point numbers that meet the screening interval standard is output as high confidence screening numbers.

[0094] Please see Figure 5 The negative sample generation module includes:

[0095] The confidence assessment submodule is based on high confidence screening numbers. By comparing the mineralization characteristics and profile stratigraphic changes of each monitoring point, it identifies the spatial regions of signal fluctuations, calculates the changes in mineralization indices between the monitoring point and its neighboring points, and obtains confidence intensity configuration data.

[0096] From the sampling points identified as meeting the negative sample characteristics, key parameters such as spatial location, metal concentration, strain rate, and thickness change rate are extracted. A comparative analysis of mineralization characteristics and stratigraphic changes is performed on each monitoring point. The analysis process first superimposes the normalized percentage values ​​of gold concentration and strain rate to form the comprehensive mineralization performance value for that point. Then, the difference between the mineralization performance values ​​of two adjacent sampling points is calculated. For example, if the mineralization performance value of point D301 is 38, and the values ​​of neighboring points D300 and D302 are 33 and 47 respectively, the differences between D301 and the two points to its left and right are calculated to be 5 and 9 respectively. The criterion for determining whether a point falls within the spatial signal fluctuation zone is that the difference on either side is greater than the set fluctuation threshold of 8. If this condition is met, the point is marked as a fluctuation boundary point. Next, all points meeting the fluctuation conditions are located in three-dimensional space, and continuous spatial blocks are constructed based on their coordinate relationships. The thickness changes of points within the spatial region are then extracted. The rate is used to determine whether the direction of thickness change is consistent with the direction of mineralization index change. If both show an increasing or decreasing trend, the signal fluctuations in the region are considered to have a consistent correlation. Based on this, the changes in mineralization index between the monitoring point and its neighboring points are uniformly calculated. That is, the difference in mineralization performance between each point and the two points before and after it is calculated, the direction of change is recorded, and they are classified according to the rising and falling trends. When the continuous change in the same direction exceeds three points and the change value of each point is not less than 5, the sequence is determined to be a continuous change zone. Finally, the mineralization fluctuation characteristics and thickness layer trend of all points in the spatial region are comprehensively scored. The confidence score is based on the maximum value of 100. Points that simultaneously have continuous fluctuation, consistent direction, and thickness trend support are set as high confidence (above 80 points). If only part of the conditions are met, they are set as medium confidence (50-80 points). Those below are low confidence (below 50 points). Confidence intensity configuration data are formed for all points.

[0097] The sample generation control submodule, based on confidence intensity configuration data, analyzes the correlation between the confidence level, spatial number, and disturbance factor of each monitoring point, and calculates the mineralization differences and stratigraphic parameters between the monitoring point and its neighboring points using the following formula:

[0098] ;

[0099] By determining the disturbance amplitude distribution at each monitoring point, a set of sampling control parameters is obtained, in which... Indicates the first Disturbance intensity at each monitoring point Indicates the first The monitoring point and the first Disturbance factors between adjacent monitoring points Indicates the first The monitoring point and the first Differences in mineralization characteristics between adjacent points Indicates the first Stratigraphic parameters of each monitoring point Indicates the first The average value of the stratigraphic parameters of the profile at adjacent monitoring points;

[0100] Disturbance intensity refers to a comprehensive spatial influence index used in the negative sample generation module to dynamically adjust the distribution of negative samples for each monitoring point (the i-th point). The greater the disturbance intensity, the more pronounced the combined fluctuations in confidence level, mineralization characteristics, and stratigraphic changes within the local spatial range of that point. As a key input for regulating the sample generation controller, disturbance intensity can characterize the degree to which monitoring points are affected by random disturbances and spatial heterogeneity. Based on the distribution of disturbance intensity, the participation status of different collection points in the negative sample set can be adjusted, making the generated sample distribution closer to the complexity and randomness of real geological space.

[0101] At the monitoring points After locating the spatial number, determine its five nearest neighbor monitoring points and obtain the following raw parameters from the data:

[0102] The disturbance factors are as follows:

[0103] , , , , Dimensionless parameters do not require normalization.

[0104] The original values ​​of the differences in mineralization characteristics between adjacent monitoring points are as follows:

[0105] , , , , .

[0106] The five groups of differences were processed using the minimum-maximum normalization method (maximum value 4.6, minimum value 3.9), and the normalization results are as follows:

[0107] , , , , .

[0108] The first normalized The monitoring point and the first Differences in mineralization characteristics between adjacent points; subsequently, raw stratigraphic thickness data from the main monitoring point are retrieved. The section thicknesses at the five adjacent points are 28.7, 30, 29.9, 28.5, and 29.4. The average section thickness at the adjacent points is calculated as follows:

[0109] ;

[0110] Substitute all the above parameters into the disturbance calculation formula:

[0111] The disturbance product terms are as follows:

[0112] ;

[0113] ;

[0114] ;

[0115] ;

[0116] .

[0117] Summing the five terms, we get:

[0118] .

[0119] The denominator is:

[0120] .

[0121] Calculate the disturbance intensity value:

[0122] .

[0123] According to the criteria for determining the level of disturbance intensity:

[0124] like If the disturbance is low, the point is determined to be in a low-disturbance zone. This zone indicates that the disturbance fluctuation of the monitoring point in the spatial range is small, the mineralization characteristics between adjacent points are not significantly different, the stratigraphic structure shows a relatively stable state, the spatial differences are not obvious, and the overall environment is in a low-disturbance environment.

[0125] like This is the middle segment of the disturbance. The disturbance amplitude in this segment has a certain degree of spatial discontinuity. The disturbance factor and the mineralization difference have a considerable combined effect, but it has not yet constituted an extreme abrupt change. It belongs to the medium intensity response characteristic area.

[0126] like This is a strongly disturbed section, which usually reflects significant disturbance coupling within the vicinity of the monitoring point, a clear abrupt connection between the stratigraphic distribution and mineralization differences, an enhanced spatial variation trend, and may involve complex geological structures or the interaction of multiple factors.

[0127] like If the disturbance is strong, it indicates that the monitoring point is highly dispersed under multi-factor disturbances, and the disturbance index far exceeds the threshold of the stable area. It is necessary to combine the on-site geological survey and multi-source data for comprehensive analysis and mark it as a high-disturbance control target, as an anomaly key point for priority analysis.

[0128] The current result is ,satisfy Therefore, the monitoring point was determined to be in a weak disturbance zone. This zone indicates that the monitoring point experiences relatively small spatial disturbances, with limited differences in mineralization characteristics between neighboring points, a relatively stable stratigraphic structure, and weak spatial variability. According to the sampling control standard classification, this point was assigned a parameter label of disturbance level 1. This disturbance level will then be written into the sampling control parameter set for subsequent spatial analysis and screening processes.

[0129] The generation state adjustment submodule, based on the sampling control parameter set, filters monitoring points that meet the correlation conditions between disturbance distribution and confidence level, judges the correspondence between the disturbance state and confidence marker of each point, adjusts the generation state of sample points, and obtains dynamic negative sample distribution information.

[0130] After the confidence level configuration data is output, the sampling disturbance, monitoring time sequence markers, and historical interference tags recorded for each point are combined for screening. In this process, all points with confidence levels no higher than medium confidence are first extracted, and their interference records are counted to see if they exceed the set disturbance baseline value. The disturbance state is defined as the number of times each monitoring point experiences signal interruption, frequency band abrupt change, or strain reversal within a set time period. The disturbance baseline value is set to two or more anomalies. When judging the correspondence between the disturbance state and the confidence marker, points with high confidence levels but severe interference states are marked as contradictory. The rule for determining whether to adjust the generated state is: when a point is at high confidence but the number of interferences is greater than 2, it is converted to non-parameterized. For points with moderate confidence and 0 or 1 disturbance frequency, the generation status is adjusted according to the following criteria: confidence score ≥ 80 and disturbance frequency ≤ 1 is set as status 1 (participating in generation); score between 50 and 80 and disturbance frequency ≤ 1 is set as status 2 (backup generation); and other statuses are set to 0 (not generated). Then, the status field is marked on each adjusted sample point. The spatial number, sample generation status, historical disturbance frequency, confidence score and current control parameter index of each point are recorded in the dynamic negative sample distribution information. The output dynamic negative sample distribution information structure fully reflects whether each point is currently adopted into the new round of sample generation sequence.

[0131] Please see Figure 6 The boundary correction analysis module includes:

[0132] The boundary comparison and judgment submodule obtains the spatial number and time label of each collection point based on the dynamic negative sample distribution information, analyzes the spatial arrangement of the predicted boundary coordinates, compares the spatial relationship between the actual collection point coordinates and the predicted boundary, judges the distribution segments with inconsistent coverage, and obtains the boundary offset recognition result.

[0133] The spatial identifier and time stamp of each sampling point are extracted. The spatial identifier is used to locate the precise position of the monitoring point in the three-dimensional coordinate system, and the time stamp is used to trace the time sequence of data collection or generation for each monitoring point. During the execution process, a list of coordinates of all sampling points in space is first established. For example, sampling points D401, D402, and D403 correspond to coordinates (125.5, 87.3, –140.2), (126.0, 87.3, –140.2), and (126.5, 87.3, –140.2), respectively. Then, the set of coordinates of the predicted ore body boundary for the current stage is obtained. This boundary is represented as a polygon or surface structure composed of multiple boundary nodes. In the spatial arrangement analysis, adjacent points are compared for each set of boundary line segments or surfaces. The system constructs a predicted boundary envelope and compares the actual collected points to see if they are located inside the boundary. If the coordinates of a point are outside the envelope of all boundary segments, it is marked as an external point. During this process, all external points and their time labels are marked as abnormal. Then, the spatial segments where points are concentrated are statistically analyzed to see if the distribution segments are concentrated in a certain direction. For example, if D401 to D408 are all outside the predicted boundary and are arranged along the X-axis, it is determined that the boundary prediction has shifted. The basis for judging the inconsistent coverage relationship is: if the predicted boundary is not covered and the actual point is located in a strongly mineralized area or a negative sample intensity area with low intensity, then the area is marked as an offset segment. All such spatial abnormal segments are numbered and their coordinates are archived, and the output is the boundary offset identification result.

[0134] The retrospective analysis submodule, based on the boundary offset identification results, calls the gold element signal and surrounding rock strain characteristics within the spatial segment to calculate the combined changes of the two types of parameters in the offset area, analyzes the synchronicity between the combined change trend and the boundary prediction results, and obtains the characteristic change sequence of the offset area.

[0135] All monitoring points within the predicted offset area are identified. Within the spatial segment, the amplitude values ​​of the gold element signal and the strain rate data of the surrounding rock are extracted. These are then combined to construct two sets of geological parameters for the area. For each parameter set, a start and end time period is set for each point to ensure the temporal correspondence of the data. For example, the average amplitude of the gold element signal at D403 is 3.52 volts, corresponding to a strain rate of 12.8 microstrains per second; the gold signal at D404 is 3.76 volts, with a strain rate of 14.2. Subsequently, combined change records are established for each point sequentially. The records include the difference and direction of change between the gold signal and strain values ​​at adjacent points within the same segment, and the change direction of the two types of parameters at each point. The data is compared with the target data. If the gold signal increases and the strain increases synchronously, it is marked as a positive correlation change. Otherwise, it is a asynchronous change. Trend calculation is performed on all recorded data to determine whether the combined change trend within the spatial segment is consistent. In the consistency judgment, the minimum number of consecutive points is set to 3, and the two types of parameters change in the same direction between adjacent points are considered to be consistent trends. If two or more such consistent trend segments are formed in the offset segment, they are recorded as structural change sequences and output as the combined response performance of the offset segment. The trend sequence is labeled with its start and end point numbers, signal difference magnitude, direction sign, and spatial arrangement order to form the characteristic change sequence of the offset area.

[0136] The acquisition point adjustment submodule determines the correspondence between the current classification of the acquisition point and the direction of signal feature change based on the feature change sequence of the offset region, optimizes the label information of the acquisition point in the classification process, adjusts the status identifier of the acquisition point in the sample set, and obtains the boundary offset correction parameters.

[0137] The classification label information of the original sampling points in the sample set is retrieved point by point from the offset segment. This information includes whether the point is currently classified as a negative sample and its hierarchical location within the negative sample structure. For example, D405 is currently a state 1 sample point, corresponding to the negative sample master set. When determining the correspondence between the current classification of the sampling point and the direction of signal feature change, its classification status is compared with the direction of change in the feature change sequence. If the current classification is a negative sample and the point shows positive gold signal growth and strain enhancement in the offset segment, then the original classification is considered inconsistent with the current direction of change, and its label is adjusted to negative. In addition to labeling, the sample status is updated to status 0, indicating that the point will no longer participate in negative sample generation. If the classification is undefined or a spare sample and the current change is clear, it is updated to status 1 or status 2. In the process of optimizing label information, the index position and time label of the point in the sample set structure also need to be updated to ensure that it corresponds to the latest data version. The output boundary offset correction parameters record the original classification status, new classification label, spatial location, change direction and sample status change record of each adjusted collection point. The summaries constitute the correction parameter set to replace the old data of this part of the collection points in the original negative sample structure.

[0138] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for locating and predicting quartz vein-type gold deposits based on artificial intelligence, characterized in that, The method includes: The surrounding rock monitoring and acquisition module is based on the surrounding rock monitoring points in the quartz vein type gold mine area. It analyzes the collected electrical and mechanical signals, performs signal segmentation and feature classification, and maps the gold element concentration and strain characteristics to the borehole number and coordinates to obtain the surrounding rock monitoring parameter set. The mineralization feature construction module, based on the surrounding rock monitoring parameter set, compares the changes in gold and iron element signals at monitoring points, analyzes the distribution of gold and silicon signals, combines borehole strain data, determines the changes in mineralization intensity, identifies abnormal zones, and obtains mineralization feature distribution parameters. The negative sample screening module selects borehole thickness variation data based on the mineralization characteristic distribution parameters, analyzes the relationship between mineralization indicators and thickness trends, compares the target with the performance of adjacent collection points, determines whether it is in the screening interval, and obtains a high-confidence screening number. The negative sample generation module analyzes the confidence performance of the collection points based on the high confidence screening number, maps the confidence results to the sample generation control, and adjusts the sample point generation state in combination with random signals to obtain dynamic negative sample distribution information. Based on the dynamic negative sample distribution information, the boundary correction analysis module analyzes the collection point number and time label, compares the predicted boundary coordinates with the collection point distribution, determines whether the boundary has shifted, re-analyzes the associated features, and obtains the boundary shift correction parameters.

2. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The surrounding rock monitoring parameter set includes the three-dimensional coordinates of the monitoring points, the original geological signal encoding, and the monitoring data label. The mineralization characteristic distribution parameters include the mineralization zoning code, the mineralization distribution level, and the abnormal mineralization identifier. The high confidence screening number includes the negative sample candidate number, the spatial positioning label, and the confidence screening level. The dynamic negative sample distribution information includes the sample spatial distribution information, the generation time series record, and the confidence label. The boundary offset correction parameters include the offset detection result, the correction coordinate set, and the correction rule parameters.

3. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The surrounding rock monitoring and acquisition module includes: The data stream receiving submodule is based on the monitoring points of the surrounding rock in the quartz vein gold mine area. It analyzes the electrical and mechanical signals collected by the geological sensors at the monitoring points, classifies and organizes the signal channel numbers and acquisition times, judges the consistency of channel data during the data merging process, and obtains the signal dataset. The signal feature segmentation submodule calculates the time series and amplitude changes of each channel signal based on the signal dataset, compares the peak amplitude characteristics, periodic distribution and frequency band spread of the waveform in the signal segment, identifies the category labels of the signal features, adjusts the feature segmentation order, and obtains the signal feature sequence. The monitoring point mapping and identification submodule performs attribution processing on the gold element concentration data and the surrounding rock strain change data based on the signal feature sequence, maps the data to the monitoring point borehole number and the three-dimensional coordinates of the ground surface, adjusts the order of the signals in the preprocessing process, and obtains the surrounding rock monitoring parameter set.

4. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The mineralization feature construction module includes: The signal amplitude comparison submodule analyzes the signal amplitudes of gold and iron at each monitoring point based on the surrounding rock monitoring parameter set, calculates the differences between the two types of element signals in the spatial sequence, compares the signal change direction and continuous change between each monitoring point, judges the change of trend turning point in coordinate distribution, and obtains the gold and iron signal change trend factor. Based on the gold and iron signal change trend factor, the feature fusion calculation submodule calculates the signal amplitudes of gold and silicon elements at the monitoring points, analyzes the combined characteristics of the two elements in three-dimensional coordinates, filters the surrounding rock strain change data combined with the borehole record, optimizes the process of fusing metal signal features and strain parameters, and obtains a multi-parameter fusion feature group. The anomaly identification submodule, based on the multi-parameter fusion feature group, determines the fusion feature change status of each acquisition point, identifies the fusion parameter change range between spatially continuous acquisition points, analyzes data segments whose change amplitude exceeds the normal fluctuation of continuous areas, and obtains mineralization feature distribution parameters.

5. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The negative sample screening module includes: Based on the mineralization characteristic distribution parameters, the profile thickness analysis submodule analyzes the borehole profile thickness variation data, compares the continuous thickness variation of the target acquisition point with that of adjacent acquisition points, determines the direction of change and calculates the thickness difference, identifies the associated acquisition points as the result, and obtains the thickness change rate sequence. The mineralization index extraction submodule, based on the thickness change rate sequence, compares the gold element signals of the target sampling point with those of adjacent points, calculates the proportional relationship between the sum of the signals at each point and the thickness change, using the formula: ； The distribution values ​​of mineralization indicators were obtained, and the amplitude of the gold element signal was integrated with the thickness change rate to obtain the mineralization indicator distribution set, where, This represents the distribution value of mineralization index at the nth sampling point. This represents the amplitude of the gold element signal at the nth sampling point. This represents the amplitude of the gold element signal at the (n-1)th sampling point. This represents the amplitude of the gold element signal at the (n+1)th sampling point. This represents the thickness change rate at the nth sampling point; The screening interval determination submodule filters the index performance of each collection point based on the mineralization index distribution set, determines whether it meets the screening criteria, and organizes the collection point numbers that pass the screening to obtain high confidence screening numbers.

6. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The negative sample generation module includes: The confidence assessment submodule, based on the high confidence screening number, identifies the spatial region of signal fluctuation by comparing the mineralization characteristics and profile stratigraphic changes of each monitoring point, calculates the changes in mineralization indicators between the monitoring point and its neighboring points, and obtains confidence intensity configuration data. Based on the confidence intensity configuration data, the sample generation control submodule analyzes the correlation between the confidence level, spatial number and disturbance factor of each monitoring point, calculates the mineralization difference and profile stratigraphic parameters between the monitoring point and adjacent points, determines the disturbance amplitude distribution of each monitoring point, and obtains the sampling control parameter set. The generation state adjustment submodule, based on the sampling control parameter set, filters monitoring points that meet the correlation conditions between disturbance distribution and confidence level, determines the correspondence between the disturbance state and confidence marker of each point, adjusts the sample point generation state, and obtains dynamic negative sample distribution information.

7. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The boundary correction analysis module includes: The boundary comparison and judgment submodule obtains the spatial number and time label of each collection point based on the dynamic negative sample distribution information, analyzes the spatial arrangement of the predicted boundary coordinates, compares the spatial relationship between the actual collection point coordinates and the predicted boundary, judges the distribution segments with inconsistent coverage, and obtains the boundary offset recognition result. Based on the boundary offset identification results, the retrospective analysis submodule calls the gold element signal and surrounding rock strain characteristics within the spatial segment, calculates the combined changes of the gold element signal and surrounding rock strain characteristic parameters in the offset area, analyzes the synchronicity between the combined change trend and the boundary prediction results, and obtains the characteristic change sequence of the offset area. Based on the feature change sequence of the offset region, the acquisition point adjustment submodule determines the correspondence between the current classification of the acquisition point and the direction of signal feature change, optimizes the label information of the acquisition point in the classification process, adjusts the status identifier of the acquisition point in the sample set, and obtains the boundary offset correction parameters.

8. The method for locating and predicting quartz vein-type gold deposits based on artificial intelligence according to claim 1, characterized in that, The segmentation identification and feature classification are to divide the collected raw signal data according to time, depth or geographical location, and to classify and organize the data in different intervals through feature extraction. The abnormal zone refers to the spatial region that is different from the surrounding mineralization state after feature trend analysis.

Images