Regional mineral resource prediction analysis method and system

Abstract

The invention relates to the technical field of mineral resource prediction and analysis, and discloses a regional mineral resource prediction and analysis method and system, and the method comprises the steps: carrying out the collection and standardization, and constructing a standardized physical data set; organizing the physical parameters in a spatial dimension, and constructing a multi-dimensional tensor structure; tensor evolution modeling is carried out to generate tensor time sequence data; obtaining an optimal weight combination according to geological unit division; calculating the mineralization probability of each space unit; evaluating the credibility of the mineralization probability result, and outputting a mineralization probability graph and a credibility scoring graph; the system comprises a data acquisition and standardization module, a tensor modeling module, an evolution modeling module, a feature optimization module, a probability prediction module, a credibility evaluation module and a visual output module. According to the method, a space-parameter joint expression mode based on a four-order tensor structure is introduced, various kinds of physical field data are subjected to integrated processing in a unified space framework, and a data model reflecting multi-dimensional geological response characteristics is effectively constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of mineral resource prediction and analysis, and in particular to a regional mineral resource prediction and analysis method and system thereof. Background Art

[0002] The formation of mineral resources is governed by complex geological evolutionary processes, and their spatial distribution is closely related to the characteristics of multiple physical fields, including geophysical and geochemical fields. With the advancement of modern geological survey techniques, the acquisition of multi-source data has become increasingly convenient. High-dimensional physical indicators such as geoelectricity, geomagnetism, geothermal gradients, and geochemical anomalies are widely used to identify and predict prospective mineralization areas. In this context, the effective modeling and integrated analysis of this high-dimensional information has become a key research topic in the field of mineral resource prediction.

[0003] Existing mineral resource prediction and analysis techniques generally employ static analysis models such as grid overlay, weight assignment, and linear weighting. These methods, based on two-dimensional geological maps or single-layer geophysical data, assess the influence of various physical fields on mineralization by establishing empirical rules or statistical correlations. Some studies have also attempted to use machine learning algorithms to model multiple parameters, such as support vector machines and neural networks, to fit high-dimensional data.

[0004] However, the existing mineral resource prediction and analysis technology uses traditional two-dimensional models to deal with three-dimensional and multi-parameter spatial relationships, which makes it difficult to fully express the coupling effect between different parameters, resulting in a decrease in prediction accuracy in some key areas. In addition, existing methods usually regard the mineralization process as a static event, lack of dynamic simulation of the evolution path of geological processes, and it is difficult to truly reflect the time evolution characteristics of the geological system, resulting in a lack of continuity and temporality in mineralization prediction under complex geological backgrounds. On the other hand, existing algorithms generally lack adaptation mechanisms for different geological units. When faced with diverse geological environments, the adjustment of parameter weights is not flexible enough, affecting the stability and generalization ability of the model in different regions. Therefore, the present invention provides a regional mineral resource prediction and analysis method and system thereof to address the shortcomings existing in the prior art. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a regional mineral resource prediction and analysis method and system, which solves the problems of insufficient three-dimensional spatial modeling capabilities, lack of modeling support for geological evolution processes, poor adaptability of parameter weights, and lack of credibility of prediction results in existing mineral resource prediction and analysis technologies.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A regional mineral resource prediction and analysis method comprises the following steps:

[0007] Collect various physical field parameters of the geological target area and the geological unit division information of the corresponding area, and standardize the collected parameters to construct a standardized physical data set;

[0008] Based on the obtained standardized physical data set, various physical parameters are organized in spatial dimensions to construct a multidimensional tensor structure;

[0009] Based on the constructed tensor structure and combined with geological evolution related parameters, tensor evolution modeling is performed to generate tensor time series data reflecting the geological time change process;

[0010] Based on the obtained geological unit division information, the tensor time series data is divided into regions, and the physical parameters in each geological unit are extracted to obtain the optimal parameter weight combination in each geological unit;

[0011] Based on the obtained optimal parameter weight combination and tensor time series data, each spatial unit in the geological target area is calculated to generate the metallogenic probability value of the spatial unit;

[0012] The generated mineralization probability results are evaluated for credibility. By analyzing the volatility of the predicted values in the geological time series and their consistency with historical geological data, a mineralization probability map and a credibility score map are output.

[0013] Preferably, the normalization process comprises performing a linear normalization operation on the physical field parameters:

[0014] The linear normalization operation is performed according to the following formula:

[0015]

[0016] in, Represents the normalized physical parameter value, x i represents the physical field parameter value, x min with x max are the minimum and maximum values of the parameter respectively;

[0017] The standardized parameters are uniformly input into the subsequent tensor construction model to ensure comparability between different data dimensions;

[0018] The physical field parameters include geoelectric parameters, geomagnetic anomalies, geothermal gradients and geochemical indicators.

[0019] Preferably, the multidimensional tensor structure is a fourth-order tensor, including three spatial dimensions and one physical field parameter dimension:

[0020] The value of each element in the tensor represents the normalized value of a physical parameter at a given spatial coordinate;

[0021] The tensor structure is constructed using sparse tensor storage to support efficient modeling of large areas.

[0022] Preferably, the tensor evolution modeling comprises the following steps:

[0023] Introducing geological evolution factors into the time dimension to form a tensor time series;

[0024] Based on the regional tectonic evolution stage and sedimentary environment changes, a weight adjustment factor is established, and the geological evolution function is used to dynamically adjust the influence intensity of the physical parameters of each time slice;

[0025] The evolution process uses geological ages as time nodes, and the evolution path is constructed through interpolation.

[0026] Preferably, the feature extraction of the physical parameters in each geological unit comprises the following steps:

[0027] Based on the tensor sub-region within each geological unit;

[0028] The extraction process uses the principal component analysis method to perform dimensionality reduction analysis on each unit tensor slice and construct an optimization objective function:

[0029]

[0030] Among them, w u represents the optimal weight vector of the physical parameters of the u-th geological unit, T u represents the tensor data matrix of the u-th geological unit, y is the geological label vector, w is the weight of the physical parameter to be optimized, λ is the regularization factor, ||·||1 represents the L1 norm, ||·|| 2 represents L2 loss;

[0031] LASSO regression is used to solve the problem and obtain the optimal parameter combination with good sparsity.

[0032] Preferably, generating the metallogenic probability value of the spatial unit comprises the following steps:

[0033] The product of the weighted tensor slice of each spatial unit and the corresponding weight combination is calculated using the following formula:

[0034]

[0035] in, It represents the metallogenic probability value of the spatial unit in the tth time slice, The standardized value of the ith physical parameter of the spatial unit (x, y, z) at the tth time slice, w i is the weight coefficient of the physical parameter i, σ(·) is the activation function;

[0036] The generated probability value represents the probability of mineralization of the spatial unit.

[0037] Preferably, the credibility assessment of the generated mineralization probability results comprises the following steps:

[0038] Based on the degree of fluctuation of the predicted probability value in the tensor time series, the volatility index is obtained by calculating the variance of the time series:

[0039]

[0040] Among them, V xyz Indicates the degree of fluctuation of the mineralization probability value of the spatial unit (x, y, z) under T time slices, It represents the average mineralization probability of the location in all time slices, It represents the metallogenic probability value of the spatial unit (x, y, z) in the tth time slice, where T is the total number of time slices;

[0041] Cosine similarity is used to evaluate the similarity between the historical records and the current probability map.

[0042] Preferably, the regularization factor is an adaptive parameter, and its value is selected within a limited interval through cross-validation. A five-fold cross-validation method is adopted to select the regularization factor that minimizes the objective function from multiple candidate values as the final regularization factor.

[0043] Preferably, the cosine similarity adopts the following formula:

[0044]

[0045] Among them, p i represents the predicted probability value of the i-th spatial unit in the current time slice metallogenic probability map, h i represents the historical geological record value of the corresponding location, n is the total number of spatial units involved in the calculation, and S represents the cosine similarity score between the predicted map and the historical record map.

[0046] It also provides a regional mineral resource prediction and analysis system, including:

[0047] Data acquisition and standardization module, used to collect geophysical field parameters and geological unit information, complete normalization processing, and construct standardized physical data sets;

[0048] A tensor modeling module, configured to construct a multidimensional spatial physical parameter tensor based on the standardized data;

[0049] An evolution modeling module, for introducing geological evolution parameters based on the tensor to generate a tensor time series;

[0050] Feature optimization module, used to extract physical features based on tensor time series and geological unit division, and calculate the optimal parameter weight combination;

[0051] A probability prediction module is used to calculate the metallogenic probability value of each spatial unit using the time series and parameter weights;

[0052] Credibility assessment module, used to analyze the fluctuation of mineralization probability in time series and its consistency with historical geological data, and generate a credibility score map;

[0053] The visualization output module is used to output mineralization probability maps and credibility score maps to realize result display and interactive analysis.

[0054] The present invention provides a regional mineral resource prediction and analysis method and system thereof. It has the following beneficial effects:

[0055] 1. This paper introduces a joint spatial-parametric representation based on a fourth-order tensor structure, integrating and processing multiple physical field data within a unified spatial framework, effectively constructing a data model reflecting multidimensional geological response characteristics. This structural tensor representation replaces previous approaches that rely solely on two-dimensional grids or independent variable processing, significantly improving the model's ability to organize high-dimensional data fusion and avoiding the information loss and correlation loss associated with traditional methods due to inconsistent dimensionality.

[0056] 2. This invention dynamically simulates geological evolution by constructing tensor time series, achieving a natural transition from static to evolutionary modeling of geological information. Constructing multi-temporal tensors based on time slices overcomes the limitations of existing methods that rely on inferences based on a single time slice. This allows mineralization prediction to be integrated with the continuity of geological processes, addressing the inability of static modeling to depict evolutionary trends.

[0057] 3. In the feature extraction stage, the present invention designs a local weight optimization mechanism based on geological unit division and combines regularization to achieve sparse parameter selection. This regional adaptive feature optimization method avoids the extensive processing problems in the traditional unified model and significantly improves the adaptability to complex geological backgrounds. It is particularly suitable for areas with complex structures and uneven physical responses.

[0058] 4. This paper combines the dual indicators of time series volatility and historical consistency to construct a combined credibility scoring system, providing a dual measure of forecast uncertainty and historical consistency. Compared to existing technologies that rely solely on crude credibility assessments of forecast probabilities, this mechanism is more explanatory and practical, and is also conducive to its subsequent application in resource decision-making scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a flow chart of the method steps of the present invention;

[0060] Figure 2 This is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0062] Please see the attached Figure 1 The embodiment of the present invention provides a regional mineral resource prediction and analysis method, comprising the following steps:

[0063] S1. Collect various physical field parameters of the geological target area and the geological unit division information of the corresponding area, and standardize the collected various parameters to construct a standardized physical data set;

[0064] S2. Based on the obtained standardized physical data set, various physical parameters are organized in spatial dimensions to construct a multidimensional tensor structure;

[0065] S3. Based on the constructed tensor structure and in combination with geological evolution related parameters, tensor evolution modeling is performed to generate tensor time series data reflecting the geological time change process;

[0066] S4. Based on the obtained geological unit division information, the tensor time series data is divided into regions, and the physical parameters in each geological unit are feature extracted to obtain the optimal parameter weight combination in each geological unit;

[0067] S5. Based on the obtained optimal parameter weight combination and tensor time series data, each spatial unit in the geological target area is calculated to generate a metallogenic probability value for the spatial unit;

[0068] S6. Conduct credibility assessment on the generated mineralization probability results. By analyzing the volatility of the predicted values in the geological time series and their consistency with historical geological data, the mineralization probability map and credibility score map are output.

[0069] In step S1, in this embodiment, geophysical and geochemical data are collected for the target area, including but not limited to geoelectric parameters (such as apparent resistivity), geomagnetic anomalies, geothermal gradient values, and geochemical indicators (such as the abundance of metal elements such as Cu, Pb, and Zn). These parameters are derived from different measurement systems, and their units and value ranges vary significantly.

[0070] Generally, to improve data fusion capabilities across multiple physical parameters, data interpolation must first be performed within a uniform spatial resolution grid. Interpolation can optionally use bilinear interpolation, kriging interpolation, or inverse distance weighted methods. In one embodiment, a combination of two-dimensional planar interpolation and vertical linear interpolation is used to map physical parameters onto a three-dimensional grid, constructing a standard three-dimensional spatial data model.

[0071] Subsequently, in order to achieve scale uniformity among different physical quantities, the present invention uses a linear normalization method to standardize various physical parameters.

[0072] Specifically, for each physical parameter type, its original value at the spatial position (x, y, z) is recorded as , and linear normalization is performed using the following normalization formula:

[0073]

[0074] in, Represents the normalized physical parameter value; x i is the original measured value of a physical parameter in the current spatial unit; x min Indicates the minimum value of the physical parameter in the entire target area; x max Indicates the maximum value of the physical parameter in the entire target area.

[0075] The above formula normalizes all parameter values to the interval [0, 1], preserving the relative proportionality of the data while eliminating the effects of the units and scales of the physical quantities themselves. This normalization operation is a necessary data preprocessing step to ensure that the physical quantities of each dimension in the subsequently constructed tensor data structure are comparable, without causing feature distortion or model drift due to differences in absolute values.

[0076] In some embodiments, to further enhance the effectiveness of marginal data points, an extreme value truncation mechanism can be introduced before normalization to prevent extreme outliers from interfering with the overall data distribution. For example, data exceeding a statistical quantile (such as P1 or P99) can be truncated to maintain the stability of the data distribution.

[0077] In one possible implementation, the standardized data generated after normalization will be uniformly formatted into the basic unit of the four-dimensional structure tensor, and each physical parameter will subsequently be embedded in the fourth-dimensional parameter axis to form a unified and standardized data collection.

[0078] After the standardization operation is completed, consistency verification can also be performed on the data to detect the continuity and physical rationality of the standardization results in the spatial dimension. For example, the geothermal gradient should show an increasing trend along the depth direction, and the geomagnetic anomaly should show regional aggregation characteristics.

[0079] Regarding step S2, in this embodiment, after completing the normalization standard processing of the geophysical data, in order to achieve spatial structured expression and parameter correlation modeling, the standardized data set needs to be further organized to construct a spatial expression model that supports multi-dimensional information fusion. Based on the normalized multi-source physical field parameter data set, a set of unified tensor structures is constructed to describe the spatial physical response distribution. The construction method is as follows:

[0080] Specifically, let the spatial grid of the geological target area be X×Y×Z, where X, Y, and Z represent the number of grid divisions in the east-west, north-south, and vertical directions, respectively. Each spatial unit corresponds to a location point (x, y, z). Assume that a total of M types of physical parameters are collected and processed in the area, and the normalized value of each type of parameter at each spatial unit is recorded as T x,y,z,i , all physical parameters can be organized into a fourth-order tensor:

[0081]

[0082] in, Represents the multidimensional physical response tensor of the entire spatial region; X, Y, and Z are the dimensions of the spatial grid on the three spatial axes; M is the total number of physical field parameters involved in the modeling.

[0083] In some embodiments, to improve computational efficiency and memory utilization in large-area scenarios, the tensor structure is constructed using a sparse tensor representation. Specifically, this approach can be based on establishing an index mapping only for non-zero or valid data locations, and storing the data cell contents in a sparse matrix format to compress the entire data domain. Generally, this approach is particularly effective for deep or edge areas with missing data, avoiding large amounts of invalid zero padding.

[0084] As an option, the tensor structure can also support data fusion at different spatial scales. For example, a high-density grid (high-resolution tensor slices) is used in shallow areas, while low-resolution tensor slices are used to cover deep, data-sparse areas, achieving hierarchical multi-scale modeling.

[0085] In one possible implementation, each physical parameter dimension can be further pre-assigned a weight label or attribute label to identify the measurement method, resolution level, or physical field type (e.g., electric field, magnetic field, thermal field, etc.) from which it originates. This labeling mechanism can be used to provide dimension-level prior structure information support in subsequent tensor evolution or feature optimization stages.

[0086] Furthermore, to meet the interface requirements of subsequent dynamic modeling and evolutionary simulation, the constructed tensor data structure can be organized in memory using a linearized layout with dimension-wise expansion, supporting fast access and processing based on tensor slices. For example, a single-parameter 3D volume can be expanded along the fourth dimension, or all parameter layers can be expanded along a specific spatial dimension, enabling a flexible data call interface.

[0087] Regarding step S3, in this embodiment, after completing the standardized preprocessing of the multi-source physical field data and the construction of the tensor structure, the geological evolution information is introduced into the tensor model to construct tensor time series data that can reflect the temporal variation characteristics of the geological process. The tensor evolution model is constructed with the fourth-order tensor As the initial state, the time dimension is expanded based on the geological history process to form a tensor time series in A tensor representing the t-th geological evolution time node, where T represents the total length of the time series.

[0088] Generally, the mineralization process of a regional geological body is highly correlated with its geological evolution stage, with factors such as tectonic activity, sedimentary environment, and heat flow background changing significantly over time. Therefore, the model introduces a time variable t to represent the geological age division node, generating a tensor state at each time slice to reflect the dynamic evolution of the geological response.

[0089] Specifically, the tensor time evolution function is defined as follows:

[0090]

[0091] in, A tensor representing the t-th geological evolution time node; A tensor representing the initial time point; θ t is a set of parameters related to geological evolution, used to describe tectonic activities, sedimentary characteristics or heat flow changes during this period; f t (·) is the time evolution function, which is used to map the variation driven by geological information into the tensor space.

[0092] In some embodiments, the evolution function may be implemented as an adjustment function based on a tensor perturbation model, an interpolation function, a parameterized transformation, or a temporal weight.

[0093] As an option, the evolution factor θ t The following sub-parameter sets may be included:

[0094] Changes in regional tectonic stress coefficients;

[0095] geothermal gradient change factor;

[0096] Sedimentation rate variation function;

[0097] Geochemical transport index;

[0098] Geological stage weight vector.

[0099] In one possible implementation, θ t Each sub-parameter can be obtained through expert geological knowledge, regional historical literature data or simulation inversion results, and is used to control the evolution amplitude and direction of each physical parameter dimension in the tensor in the time dimension. For example, the geothermal gradient term can be modeled through the following transformation form:

[0100]

[0101] in, is the value of the i-th parameter at the spatial position (x, y, z) in the t-th time slice; represents the standardized value of the i-th physical parameter, is the value of the i-th parameter at the spatial position (x, y, z) in the initial state.

[0102] Under this structure, a continuous time evolution path can be generated through interpolation. For example, key geological stages (such as the Yanshanian and Himalayan periods) can be set as nodes, and the rate of change or response function of physical parameters can be constructed at these nodes. The physical quantities at other intermediate moments can be generated through polynomial interpolation or spline interpolation.

[0103] As an extension method, tensor convolution operators or dynamic convolution kernels can be introduced to model local changes of time series tensors at the regional scale, thereby more accurately simulating the heterogeneous evolution behavior of different geological units.

[0104] Regarding step S4, in this embodiment, after the tensor time series is constructed, in order to further extract the key physical characteristics of each geological unit and construct weight combinations suitable for different geological backgrounds, it is necessary to perform local sub-region analysis on the tensor time series data. Based on the constructed tensor time series, the entire spatial tensor is first divided into several sub-tensor structures according to the geological unit layers that have been divided in the region. Each sub-tensor corresponds to a specific geological unit region.

[0105] Generally speaking, since the evolution paths and physical response patterns of different geological units are significantly different, it is necessary to construct the physical parameter weight vector for each unit separately to achieve localized modeling.

[0106] Specifically, the present invention uses principal component analysis (PCA) to reduce the dimensionality of high-dimensional physical parameter data in tensor subregions. The tensor slice of the u-th unit at a certain time slice t is represented as a two-dimensional data matrix Where: Nu is the number of spatial units contained in the u-th geological unit; M is the number of physical parameter dimensions.

[0107] Based on the above dimensionality reduction, the supervised label vector is introduced Each component y i ∈{0,1} indicates whether the location is a known mineralization point. Then the following optimization objective function is constructed for the sparse extraction of physical parameter weights:

[0108]

[0109] Among them, w u represents the optimal weight vector of the physical parameters of the u-th geological unit, T u represents the tensor data matrix of the u-th geological unit, y is the geological label vector, w is the weight of the physical parameter to be optimized, λ is the regularization factor, ||·||1 represents the L1 norm, ||·|| 2 represents the L2 loss.

[0110] In some embodiments, to further enhance the adaptive capability of the optimization process, the value of the regularization factor λ is not a fixed constant, but is automatically selected through a cross-validation method. As an option, a 5-fold cross-validation strategy is used to perform the following screening within a preset interval:

[0111]

[0112] Among them, Λ is the candidate regularization factor set; K is the fold; Loss (k) (λ) represents the loss value on the validation set after training with the regularization factor λ under the k-fold cross validation; * Indicates the best parameter value after optimization.

[0113] This cross-validation mechanism can automatically find the regularization term strength that is most suitable for the current geological unit structure, thereby effectively balancing model complexity and predictive adaptability.

[0114] As a possible implementation method, after the optimization is completed, the weight vector w is obtained u Normalization can be further performed to ensure that it does not introduce proportional bias in subsequent weighted calculations. The normalization operation can be performed as follows:

[0115]

[0116] in, Indicates the weight value after normalization; w u,i Indicates w u The original weight of the i-th physical parameter in; M represents the total number of dimensions; w uIndicates the original weight value obtained after optimization is completed.

[0117] In step S5, in this embodiment, after obtaining the optimal physical parameter weight combination corresponding to each geological unit, to further achieve resource prediction expression on a spatial scale, it is necessary to combine this weight combination with the aforementioned tensor time series data, calculate the mineralization probability for each spatial unit in the region, and construct a characteristic weight vector for the unit at each spatial location based on the extracted optimal parameter weight combination of the geological unit. In general, this weight vector can be directly assigned based on the optimization results of the geological unit to which it belongs.

[0118] Then, for each three-dimensional spatial unit, the weighted combination value is extracted based on the tensor data of its corresponding time slice. Let the normalized tensor data of the spatial unit at the tth time slice be expressed as:

[0119]

[0120] in, Indicates the standardized values of all physical parameters of the spatial unit (x, y, z) at the tth time slice; M is the number of physical parameter dimensions; The normalized value of the i-th physical parameter of the spatial unit (x, y, z) at the t-th time slice.

[0121] Specifically, the calculation of mineralization probability adopts the weighted summation method to perform feature fusion and the activation function to perform probability mapping. The calculation formula is as follows:

[0122]

[0123] in, It represents the metallogenic probability value of the spatial unit in the tth time slice, The standardized value of the ith physical parameter of the spatial unit (x, y, z) at the tth time slice, w i is the weight coefficient of the physical parameter i, σ(·) is the activation function, which in some embodiments is the Sigmoid function, and is defined as follows:

[0124]

[0125] Where σ(x) is the output of the Sigmoid activation function; x is the input variable; and e is the base of the natural logarithm.

[0126] The above function can map the original score after weighted summation to the interval [0,1], making it easier to interpret it as a probability value.

[0127] As an option, in order to enhance the expressive power of the calculation, the activation function may also use other forms, such as the Softplus function or the hyperbolic tangent function (Tanh), to adapt to the specific parameter distribution characteristics, but in the present invention, the Sigmoid function is used as the standard form by default.

[0128] In one possible implementation, in order to reduce the impact of occasional outliers on the prediction results, a sliding time window can be introduced to smooth the prediction results of multiple time slices and calculate the average metallogenic probability of the spatial unit in the time series, which can be specifically expressed as:

[0129]

[0130] in, It represents the average mineralization probability of the location in all time slices; T represents the total number of time slices; Represents the mineralization probability value of the spatial unit (x, y, z) in the tth time slice.

[0131] Preserving the trend information of time evolution also helps to stabilize the final prediction results, facilitating subsequent credibility assessment or probability map generation

[0132] For step S6, in this embodiment, after completing the calculation of the mineralization probability of all spatial units in the region, in order to ensure the reliability and practical usability of the prediction results, it is necessary to conduct a credibility assessment on the obtained mineralization probability value to identify possible uncertainty areas or information biases in the model output. Based on the obtained mineralization probability output under the tensor time series, the prediction volatility index of each spatial unit in the time dimension is calculated. This index is used to measure the stability of the model's prediction results for a certain area in the evolution path. The volatility score can be measured by the time series variance, and the calculation formula is as follows:

[0133]

[0134] Among them, V xyz Indicates the degree of fluctuation of the metallogenic probability value of the spatial unit (x, y, z) under T time slices; It represents the average mineralization probability of the location in all time slices; It represents the metallogenic probability value of the spatial unit (x, y, z) in the tth time slice; T is the total number of time slices.

[0135] Generally speaking, if a spatial unit fluctuates greatly throughout the entire evolutionary sequence, it may mean that there are uncertainties or high model sensitivity issues in the region, and the credibility score should be lowered.

[0136] Secondly, to further evaluate the authenticity of the model output results based on actual geological information, this paper introduces a spatial consistency measurement indicator with historical geological records and uses cosine similarity for calculation. This method performs a vectorized comparison between the current time slice prediction probability map and the known historical mineralization record map. The calculation formula is as follows:

[0137]

[0138] Among them, p i represents the predicted probability value of the i-th spatial unit in the metallogenic probability map of the current time slice; h i represents the historical geological record value of the corresponding location; n is the total number of spatial units involved in the calculation; S represents the cosine similarity score between the predicted map and the historical record map.

[0139] Specifically, the closer the value S is to 1, the stronger the spatial distribution consistency between the model output and the historical data, and the higher the credibility.

[0140] Specifically, the closer the value S is to 1, the stronger the spatial distribution consistency between the model output and the historical data, and the higher the credibility.

[0141] In some embodiments, the predicted volatility indicator V x,y,z It can be further weighted and fused with the consistency score S to form a comprehensive credibility score index, which is convenient for unified standard output. As an option, the fusion method can refer to the following expression:

[0142] C x,y,z =η·(1-V x,y,z )+(1-η)·S x,y,z ;

[0143] Among them, C x,y,z Score the final credibility; V xyz It represents the fluctuation degree of the metallogenic probability value of the spatial unit (x, y, z) under T time slices; η∈[0,1] is the adjustment parameter used to control the proportion of the two indicators in the total score; S x,y,z is the local cosine similarity score that can be obtained by regionalization in the history graph.

[0144] In one possible implementation, the optimal η value can be set through expert scoring or model automatic adjustment strategy to adapt to the evaluation requirements of different regions.

[0145] The regional mineral resource prediction and analysis system described below and the regional mineral resource prediction and analysis method described above can be referenced to each other.

[0146] Please see the attached Figure 2 The present invention also provides a regional mineral resource prediction and analysis system, including:

[0147] Data acquisition and standardization module, used to collect geophysical field parameters and geological unit information, complete normalization processing, and construct standardized physical data sets;

[0148] A tensor modeling module, configured to construct a multidimensional spatial physical parameter tensor based on the standardized data;

[0149] An evolution modeling module, for introducing geological evolution parameters based on the tensor to generate a tensor time series;

[0150] Feature optimization module, used to extract physical features based on tensor time series and geological unit division, and calculate the optimal parameter weight combination;

[0151] A probability prediction module is used to calculate the metallogenic probability value of each spatial unit using the time series and parameter weights;

[0152] Credibility assessment module, used to analyze the fluctuation of mineralization probability in time series and its consistency with historical geological data, and generate a credibility score map;

[0153] The visualization output module is used to output mineralization probability maps and credibility score maps to realize result display and interactive analysis.

[0154] The system of this embodiment can be used to execute the above method embodiments, and its principles and technical effects are similar, so they will not be repeated here.

[0155] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Regional mineral resource prediction and analysis method, characterized in that: The following steps are involved: Collect various physical field parameters of the geological target area and the geological unit division information of the corresponding area, and standardize the collected parameters to construct a standardized physical data set; Based on the obtained standardized physical data set, various physical parameters are organized in spatial dimensions to construct a multidimensional tensor structure; Based on the constructed tensor structure, combined with geological evolution related parameters, tensor evolution modeling is carried out to generate tensor time series data reflecting the geological time change process; Based on the obtained geological unit division information, the tensor time series data is divided into regions, and the physical parameters in each geological unit are extracted to obtain the optimal parameter weight combination in each geological unit; Based on the obtained optimal parameter weight combination and tensor time series data, each spatial unit in the geological target area is calculated to generate the metallogenic probability value of the spatial unit; The generated mineralization probability results are evaluated for credibility. By analyzing the volatility of the predicted values in the geological time series and their consistency with historical geological data, a mineralization probability map and a credibility score map are output.

2. The regional mineral resource prediction and analysis method according to claim 1, characterized in that: The normalization process includes performing a linear normalization operation on the physical field parameters: The linear normalization operation is performed according to the following formula: in, Represents the normalized physical parameter value, x i represents the physical field parameter value, x min with x max are the minimum and maximum values of the parameter respectively; The standardized parameters are uniformly input into the subsequent tensor construction model to ensure comparability between different data dimensions; The physical field parameters include geoelectric parameters, geomagnetic anomalies, geothermal gradients and geochemical indicators.

3. The regional mineral resource prediction and analysis method according to claim 1, characterized in that: The multidimensional tensor structure is a fourth-order tensor, including three spatial dimensions and one physical field parameter dimension: The value of each element in the tensor represents the normalized value of a physical parameter at a given spatial coordinate; The tensor structure is constructed using sparse tensor storage to support efficient modeling of large areas.

4. The regional mineral resource prediction and analysis method according to claim 1, characterized in that: The tensor evolution modeling includes the following steps: Introducing geological evolution factors into the time dimension to form a tensor time series; Based on the regional tectonic evolution stage and sedimentary environment changes, a weight adjustment factor is established, and the geological evolution function is used to dynamically adjust the influence intensity of the physical parameters of each time slice; The evolution process uses geological ages as time nodes, and the evolution path is constructed through interpolation.

5. The regional mineral resource prediction and analysis method according to claim 1, characterized in that: The feature extraction of the physical parameters in each geological unit comprises the following steps: Based on the tensor sub-region within each geological unit; The extraction process uses the principal component analysis method to perform dimensionality reduction analysis on each unit tensor slice and construct an optimization objective function: Among them, w u represents the optimal weight vector of the physical parameters of the u-th geological unit, T u represents the tensor data matrix of the u-th geological unit, y is the geological label vector, w is the weight of the physical parameter to be optimized, λ is the regularization factor, ||·||1 represents the L1 norm, ||·|| 2 represents L2 loss; LASSO regression is used to solve the problem and obtain the optimal parameter combination with good sparsity.

6. The regional mineral resource prediction and analysis method according to claim 1, characterized in that: Generating the metallogenic probability value of the spatial unit comprises the following steps: The product of the weighted tensor slice of each spatial unit and the corresponding weight combination is calculated using the following formula: in, It represents the metallogenic probability value of the spatial unit in the tth time slice, The standardized value of the ith physical parameter of the spatial unit (x, y, z) at the tth time slice, w i is the weight coefficient of the physical parameter i, σ(·) is the activation function; The generated probability value represents the probability of mineralization of the spatial unit.

7. The regional mineral resource prediction and analysis method according to claim 1, characterized in that: The credibility assessment of the generated mineralization probability results comprises the following steps: Based on the degree of fluctuation of the predicted probability value in the tensor time series, the volatility index is obtained by calculating the variance of the time series: Among them, V xyz Indicates the degree of fluctuation of the metallogenic probability value of the spatial unit (x, y, z) under T time slices, It represents the average mineralization probability of the location in all time slices, It represents the metallogenic probability value of the spatial unit (x, y, z) in the tth time slice, where T is the total number of time slices; Cosine similarity is used to evaluate the similarity between the historical records and the current probability map.

8. The regional mineral resource prediction and analysis method according to claim 5, characterized in that: The regularization factor is an adaptive parameter, and its value is selected within a limited interval through cross-validation. A five-fold cross-validation method is used to select the regularization factor that minimizes the objective function from multiple candidate values as the final regularization factor.

9. The regional mineral resource prediction and analysis method according to claim 7, characterized in that: The cosine similarity uses the following formula: Among them, p i Indicates the predicted probability value of the i-th spatial unit in the current time slice metallogenic probability map, h i represents the historical geological record value of the corresponding location, n is the total number of spatial units involved in the calculation, and S represents the cosine similarity score between the predicted map and the historical record map.

10. A regional mineral resource prediction and analysis system, applied to the regional mineral resource prediction and analysis method according to any one of claims 1 to 9, characterized in that: include: Data acquisition and standardization module, used to collect geophysical field parameters and geological unit information, complete normalization processing, and construct standardized physical data sets; A tensor modeling module, configured to construct a multidimensional spatial physical parameter tensor based on the standardized data; An evolution modeling module, for introducing geological evolution parameters based on the tensor to generate a tensor time series; Feature optimization module, used to extract physical features based on tensor time series and geological unit division, and calculate the optimal parameter weight combination; A probability prediction module is used to calculate the metallogenic probability value of each spatial unit using the time series and parameter weights; Credibility assessment module, used to analyze the fluctuation of mineralization probability in time series and its consistency with historical geological data, and generate a credibility score map; The visualization output module is used to output mineralization probability maps and credibility score maps to realize result display and interactive analysis.

Images