A method and related equipment for dynamic big data monitoring of mineral products

Abstract

This application discloses a big data dynamic monitoring system for mineral products, relating to the field of dynamic monitoring. The system includes: a data acquisition module for acquiring monitoring information from mining areas; a data classification and storage module for classifying the monitoring information by feature to obtain reserve prediction correlation information, and uploading the prediction correlation information to a data cache unit; a data prediction module for predicting the total mineral resources of a target region; a data verification module for verifying the total mineral resources of the region; and a user interaction module for acquiring data query requests from target users and authenticating their identities and permissions. Once the target user's identity and permission authentication is successful, the user interaction module sends the total mineral resources of the region to the target user. This application can effectively monitor and predict data changes in the total mineral resources of a target region.

Description

Technical Field

[0001] This application relates to the field of data monitoring, and in particular to a method and related equipment for dynamic monitoring of big data of mineral products. Background Technology

[0002] With the accelerating pace of industrialization in modern society, the demand for and importance placed on mineral products are increasing daily, leading to stricter requirements for mineral price regulation. Mineral reserves are a key factor influencing mineral prices, thus necessitating rigorous monitoring of mineral reserves.

[0003] Currently, monitoring of mineral reserves mainly relies on manual inventory and data upload. This method is susceptible to human error, leading to significant data discrepancies and impacting price control. Furthermore, this method is inefficient, with time lags in data upload, resulting in untimely updates and an inability to capture real-time changes in reserves. This hinders price forecasting and regulation. Therefore, a more efficient and accurate method for monitoring dynamic changes in mineral reserves is needed to ensure precise price control and establish a healthy mineral market. Summary of the Invention

[0004] This application provides a big data dynamic monitoring system for mineral products, which solves the problems of untimely and inaccurate updates of mineral reserves information in the prior art.

[0005] To achieve the above objectives, the embodiments of this application adopt the following technical solutions:

[0006] In a first aspect, a big data dynamic monitoring system for mineral products is provided. The system includes a cloud server and a user registration center, which are communicatively connected. The cloud server includes a data storage module, which comprises a data cache unit and a main data storage unit. The main data storage unit pre-stores historical mineral reserve prediction records, which include historical prediction datasets and historical reserve prediction results. The cloud server further includes:

[0007] The data acquisition module is used to obtain mining area monitoring information of all target mining areas within the target region from the mineral product full-process management system;

[0008] The data classification and storage module is used to classify the mining area monitoring information according to the historical prediction dataset to obtain reserve prediction association information, upload the reserve prediction association information to the data cache unit, and upload the mining area monitoring information after removing the reserve prediction association information to the data main storage unit. The reserve prediction association information includes mining area exploration information and mining area mining information.

[0009] The data prediction module is used to extract the mining area exploration information and mining area mining information from the data cache unit, predict the exploitable reserves of the target mining area based on the mining area exploration information and by constructing a three-dimensional model of the mining area, and predict the total mineral resources of the target area by combining the mining area mining information and exploitable reserves of all the target mining areas.

[0010] The data verification module is used to verify the total mineral resources in the region based on the historical reserve prediction results. If the verification is successful, the total mineral resources in the region are transmitted to the data cache unit.

[0011] The user interaction module is used to obtain the data query request of the target user, obtain the user permission information pre-entered in the user registration center based on the user information in the data query request, and perform identity and permission authentication on the target user based on the user permission information.

[0012] If the target user's identity and permissions are successfully authenticated, the user interaction module is further configured to send the total amount of mineral resources in the region from the data cache unit to the target user according to the data query request.

[0013] Optionally, the data classification and storage module includes:

[0014] The feature extraction submodule is used to extract the monitoring data features of the mining area monitoring information and the prediction data features of all historical prediction data in the historical prediction dataset.

[0015] The data association retrieval submodule is used to construct a monitoring data KD tree based on the monitoring data characteristics, and retrieve the reserve prediction association information that is related to the historical prediction data from the monitoring data KD tree based on the prediction data characteristics and using the nearest neighbor search method.

[0016] The classification storage submodule is used to upload the reserve prediction association information in the mining area monitoring information to the data cache unit, and to upload all information in the mining area monitoring information other than the reserve prediction association information to the main data storage unit.

[0017] Optionally, the data prediction module includes:

[0018] The data preprocessing submodule is used to extract the mining area exploration information and the mining area mining information from the data cache unit, and to perform data cleaning and preprocessing on the mining area exploration information and the mining area mining information. The mining area exploration information includes UAV exploration information and geological exploration information.

[0019] The model construction submodule is used to construct a surface model of the mining area based on the preprocessed UAV exploration information, construct a geological model of the target mining area based on the preprocessed geological exploration information, and obtain a three-dimensional model of the mining area by combining the surface model and the geological model of the mining area and through digital modeling.

[0020] The reserve prediction submodule is used to predict the mine's available reserves for the target mine based on the three-dimensional model of the mine and the pre-processed mining information of the mine.

[0021] The efficiency prediction submodule is used to perform data correlation analysis on the exploitable reserves and mining information of the mining area based on the three-dimensional model of the mining area, and predict the mining efficiency of the target mining area based on the data correlation analysis results.

[0022] The mineral reserves prediction submodule is used to combine the mining efficiency of all the target mineral areas and the predicted available reserves of the mineral areas to obtain the total regional mineral resources of the target area.

[0023] Optionally, the reserve mobilization prediction submodule includes:

[0024] The ore body volume calculation unit is used to calculate the ore body volume of the target mining area based on the three-dimensional model of the mining area and by the integration method.

[0025] The first data extraction unit is used to extract the ore mining weight data and ore mining grade data from the mining information of the mining area. Both the ore mining weight data and the ore mining grade data are time-series data.

[0026] The reserve utilization calculation unit is used to preprocess the ore mining weight data and the ore mining grade data, combine the ore body volume with the preprocessed ore mining weight data and ore mining grade data, and calculate the time series data of the target mining area's reserves utilization using the mining area reserve utilization calculation formula.

[0027] The reserve utilization prediction unit predicts the mine utilization reserves of the target mining area based on the time series data of the reserve utilization and using the exponential smoothing method.

[0028] Optionally, the efficiency prediction submodule includes:

[0029] The data preprocessing unit is used to preprocess the available reserves and mining information of the mining area.

[0030] Mesh partitioning unit, used to divide the three-dimensional model of the mining area into multiple mining area model regions;

[0031] The graph network construction unit is used to construct a heterogeneous graph network of the mining area by combining the preprocessed mining area reserves and mining information, as well as multiple mining area model regions.

[0032] The model building unit is used to build an efficiency prediction model based on a graph neural network model. The efficiency prediction model includes a spatiotemporal graph convolutional layer, a graph attention layer, a cross-layer information fusion module, a global pooling layer, and a fully connected layer. The spatiotemporal graph convolutional layer is used to perform spatiotemporal graph convolution operations on the spatiotemporal layers in the heterogeneous graph network of the mining area. The cross-layer information fusion module is used to fuse information from multiple different layers in the heterogeneous graph network of the mining area.

[0033] The efficiency prediction unit is used to use the mean square error of the efficiency prediction model as the loss function for predicting mining efficiency, input the heterogeneous graph network of the mining area into the efficiency prediction model, and output the mining efficiency corresponding to the target mining area through the efficiency prediction model.

[0034] Optionally, the data preprocessing unit includes:

[0035] The data extraction subunit is used to extract mining equipment deployment data and mining personnel allocation data from the mining information of the mining area;

[0036] The time alignment subunit is used to align the mining area's utilized reserves, the mining equipment deployment data, and the mining personnel allocation data according to a preset time scale and time length.

[0037] The data interpolation subunit is used to process missing data in the mining area's available reserves, the mining equipment deployment data, and the mining personnel allocation data using a multivariate interpolation method.

[0038] Optionally, the graph network building unit includes:

[0039] The first layer construction subunit is used to construct a mining area spatial layer by taking each of the mining area model regions as the first graph node and taking the spatial adjacency relationship of each of the mining area model regions as the first graph node edge of the first graph node.

[0040] The second layer construction subunit is used to construct a mining spatiotemporal layer by taking the mining area status of all the mining area model areas at different time points in the time length as the second graph node and the time interval between different time points as the second graph node edge of the second graph node.

[0041] The third layer construction subunit is used to construct a device collaboration layer by taking the mining equipment in the mining equipment deployment data as the third graph node and the device collaboration relationship of any number of mining equipment at the same time point as the third graph node edge of the third graph node.

[0042] The fourth layer construction subunit is used to construct a personnel collaboration layer by taking the mining personnel in the mining personnel allocation data as the fourth graph node and the personnel collaboration relationship of any number of mining personnel at the same time point as the fourth graph node edge of the fourth graph node.

[0043] The graph network construction subunit is used to construct a heterogeneous graph network of the mining area by combining the spatial layer of the mining area, the spatiotemporal layer of the mining area, the equipment collaboration layer, and the personnel collaboration layer.

[0044] Optionally, the data verification module includes:

[0045] The second data extraction unit is used to extract the historical reserve prediction results from the main data storage unit;

[0046] The prediction model construction unit is used to fit the historical reserve prediction results using the least squares method to obtain the reserve prediction model.

[0047] The data verification unit is used to obtain the predicted reserve value of the target area based on the reserve prediction model, calculate the difference between the predicted reserve value and the total mineral resources of the area, and if the difference is less than a preset difference threshold, the verification is successful and the total mineral resources of the area are transmitted to the data caching unit.

[0048] Secondly, this application provides a method for dynamic monitoring of mineral products using big data, applied to the dynamic monitoring system for mineral products using big data as described in the first aspect, the method comprising the following steps:

[0049] Used to obtain mining area monitoring information for all target mining areas within a target region from the mineral product whole process management system;

[0050] Based on the historical prediction dataset, the monitoring information of the mining area is classified by feature to obtain reserve prediction association information. The reserve prediction association information is uploaded to the data cache unit, and the monitoring information of the mining area after removing the reserve prediction association information is uploaded to the main data storage unit. The reserve prediction association information includes mining area exploration information and mining area mining information.

[0051] The exploration information and mining information of the mining area are extracted from the data cache unit. Based on the exploration information of the mining area, the exploitable reserves of the target mining area are predicted by constructing a three-dimensional model of the mining area. The total mineral resources of the target area are predicted by combining the mining information and exploitable reserves of all the target mining areas.

[0052] The total mineral resources in the region are verified based on the historical reserve prediction results. If the verification is successful, the total mineral resources in the region are transmitted to the data cache unit.

[0053] Obtain the data query request of the target user, obtain the user permission information pre-entered in the user registration center based on the user information in the data query request, and perform identity and permission authentication on the target user based on the user permission information;

[0054] If the target user's identity and permissions are successfully authenticated, the user interaction module will send the total amount of mineral resources in the region in the cache to the target user according to the data query request.

[0055] Thirdly, this application provides a big data dynamic monitoring device for mineral products, characterized in that it includes a big data dynamic monitoring system for mineral products according to any one of the first aspects.

[0056] The above technical solution addresses the crucial role of regional mineral reserves in influencing mineral prices. By employing a dynamic monitoring system to predict regional mineral reserves, the system first acquires monitoring information for all target mining areas within the target region. Based on historical prediction datasets pre-stored in the main data storage unit, the monitoring information is categorized by feature to obtain reserve prediction correlation information. This information is stored in a data cache unit for quick retrieval during subsequent data predictions. A 3D model of the mining area is constructed using exploration and mining information extracted from the data cache unit. The exploitable reserves of the target mining area are predicted based on this model. The total regional mineral reserves are then predicted based on the exploitable reserves and mining information. This total mineral reserves are verified to improve reliability. Finally, if a target user queries the system, the user interaction module sends the requested total regional mineral reserves of the target region to the user after successful authentication. In summary, this invention can efficiently and accurately predict the total regional mineral reserves of a target region.

[0057] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0058] Figure 1This application provides a schematic diagram of the structure of a big data dynamic monitoring system for mineral products.

[0059] Figure 2 This is a flowchart illustrating a method for dynamic monitoring of mineral products using big data, provided as an embodiment of this application. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for illustration and explanation of the embodiments of this application and are not intended to limit the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0061] It should be noted that if the embodiments of this application involve directional indicators (such as up, down, left, right, front, back, etc.), the directional indicators are only used to explain the relative positional relationship and movement of the components in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicators will also change accordingly.

[0062] Furthermore, if the embodiments of this application involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed in this application.

[0063] Figure 1 The illustration shows a schematic diagram of the structure of a big data dynamic monitoring system for mineral products according to an embodiment of this application. Figure 1 As shown in the embodiment of this application, a big data dynamic monitoring system for mineral products is provided. The system includes a cloud server and a user registration center, which are communicatively connected. The cloud server includes a data storage module, which includes a data cache unit and a main data storage unit. The main data storage unit pre-stores historical mineral reserve prediction records, which include historical prediction datasets and historical reserve prediction results. The cloud server includes:

[0064] The data acquisition module is used to obtain mining area monitoring information of all target mining areas within the target region from the mineral product full-process management system;

[0065] The data classification and storage module is used to classify the mining area monitoring information according to the historical prediction dataset to obtain reserve prediction association information, upload the reserve prediction association information to the data cache unit, and upload the mining area monitoring information after removing the reserve prediction association information to the data main storage unit. The reserve prediction association information includes mining area exploration information and mining area mining information.

[0066] The data prediction module is used to extract the mining area exploration information and mining area mining information from the data cache unit, predict the exploitable reserves of the target mining area based on the mining area exploration information and by constructing a three-dimensional model of the mining area, and predict the total mineral resources of the target area by combining the mining area mining information and exploitable reserves of all the target mining areas.

[0067] The data verification module is used to verify the total mineral resources in the region based on the historical reserve prediction results. If the verification is successful, the total mineral resources in the region are transmitted to the data cache unit.

[0068] The user interaction module is used to obtain the data query request of the target user, obtain the user permission information pre-entered in the user registration center based on the user information in the data query request, and perform identity and permission authentication on the target user based on the user permission information.

[0069] If the target user's identity and permissions are successfully authenticated, the user interaction module is further configured to send the total amount of mineral resources in the region from the data cache unit to the target user according to the data query request.

[0070] In this embodiment, the data acquisition module serves as the data input port for the entire system, responsible for obtaining mining area monitoring information for all target mining areas within the target region from the mineral product end-to-end management system. This process involves several technical details and implementation principles. First, the data acquisition module needs to establish a secure and reliable data connection with the mineral product end-to-end management system. This is typically achieved through API interfaces or dedicated data transmission protocols, such as RESTful API or SOAP-based web services. To ensure data transmission security, the system uses SSL / TLS encryption protocols and implements strict authentication and authorization mechanisms, such as OAuth2.0 or JWT (JSON Web Token). The acquired information includes mining area monitoring information and mining information. Mining information includes UAV exploration information and geological exploration information. Additionally, UAV exploration information includes remote sensing images and oblique images of the mining area; geological exploration information includes ore weight data, ore grade data, ore body dip angle, ore body thickness, and ore body type; and mining information includes mining methods and mining equipment.

[0071] Remote sensing imagery of mining areas utilizes remote sensing equipment mounted on unmanned aerial vehicles (UAVs) to capture images of target mining areas. This equipment includes hyperspectral imagers, thermal infrared sensors, optical cameras, and infrared cameras. The UAV is controlled to acquire images of various areas within the target mining area using its onboard remote sensing equipment. The acquired images are then geometrically corrected and fused to obtain a complete remote sensing image of the target mining area. Information such as the location, mining extent, and mining status of mining areas can be extracted based on color variations and differences in reflectance within the remote sensing images. For example, areas currently being mined will have higher reflectance in the remote sensing image, while areas where mining has ceased will have lower brightness.

[0072] The oblique imagery of the mining area is obtained by drones using oblique photography technology to capture images of the target mining area. The image acquisition equipment used consists of aerial survey cameras mounted on the drone, positioned in five directions: vertical orthogonal, forward-looking, backward-looking, left-looking, and right-looking. The oblique imagery captures complete positional and attitude information of surface features in the target mining area from the drone's cameras at different tilt angles. Geological exploration information is obtained through targeted exploration of the mining area using equipment such as mining borehole depth measuring instruments, mining borehole imaging instruments, and high-precision medium-deep hole acceptance instruments. Exploration points are determined based on the imagery information from the drone exploration. For abnormal areas shown in the images, multiple exploration points need to be identified. Geological exploration can obtain information such as the strike, dip, angle, and thickness of faults and strata in the target mining area.

[0073] In practical implementation, the data acquisition module employs a distributed architecture to improve efficiency and reliability. For example, independent data acquisition nodes can be deployed in mining areas at different geographical locations. These nodes are responsible for collecting and initially processing local data, then aggregating the processed data to a central server. This distributed architecture significantly reduces network transmission pressure and improves system response speed and reliability. The data acquisition module also needs to handle data format issues. Different mining areas use different data formats or standards, therefore the module needs to have data format conversion capabilities. For example, it needs to uniformly convert geological data in different formats (such as GeoJSON, Shapefile, etc.) into the standard format used internally by the system. Furthermore, considering the special nature of the mining industry, the data acquisition module also needs to handle some special cases. For example, some deep mining areas face communication difficulties, leading to delays or interruptions in data transmission. Therefore, the module needs to implement data caching and breakpoint resumption functions to ensure data integrity and timeliness to the greatest extent possible, even under poor network conditions.

[0074] In this embodiment, the main task of the data classification and storage module is to classify the mining area monitoring information based on the historical prediction dataset and store the classified data appropriately. This process involves complex data analysis, machine learning, and database management techniques. First, the module needs to deeply understand the structure and characteristics of the historical prediction dataset. Historical prediction datasets typically contain a large amount of mining area monitoring information and corresponding reserve prediction results. The module uses advanced data mining techniques, such as Principal Component Analysis (PCA) or Factor Analysis, to identify the key features that have the greatest influence on reserve prediction. For example, assuming there are n features, x1, x2, ..., xn, PCA will find a new set of orthogonal basis vectors v1, v2, ..., vn, maximizing the variance of the data's projection onto these new basis vectors. Based on this analysis, the module establishes a feature importance ranking to guide the subsequent feature classification process.

[0075] Next, the module will use a random forest method to build a feature classification model. This model will learn from historical data how to classify new mining area monitoring information into reserve prediction-related information and non-related information. The importance of each feature is assessed by calculating the reduction in Gini impurity, and the final classification result is determined based on feature importance. After classification, the module will upload the reserve prediction-related information to the data cache unit and other information to the main data storage unit. To optimize storage and query efficiency, the module will employ advanced data organization methods. For example, for reserve prediction-related information, columnar storage (such as Apache Parquet format) will be used to improve the performance of aggregation queries. For time series data, a time series database (such as InfluxDB) will be used to optimize time range queries.

[0076] In this embodiment, the data prediction module is responsible for constructing a 3D model based on mining exploration information and predicting the exploitable reserves of the mining area. Simultaneously, it combines mining information to predict the total mineral resources of the target region. First, the module extracts mining exploration and mining information from the data cache unit. This information typically includes geological drilling data, geophysical exploration results, and historical mining records. To handle this heterogeneous data, the module needs to implement an efficient data reading and preprocessing workflow. Next, the module uses this preprocessed exploration data to construct a 3D geological model of the mining area. Geostatistical methods, such as Kriging, are commonly used. Kriging is based on the theory of variograms and can interpolate the values ​​of unknown points in space. Its basic formula is:

[0077]

[0078] Z * (x0) is the value of the point to be estimated, Z(x) i ) represents the value of a known sample point, and λ i These are weighting coefficients. After constructing the 3D model, the module uses machine learning algorithms to predict the exploitable reserves in the mining area.

[0079] In this embodiment, the data verification module is used to verify the total regional mineral resources predicted by the data prediction module based on historical reserve prediction results. Historical reserve prediction results are the total regional mineral resources predicted by the data prediction module using the same type of data and methods before the current time, covering a historical period. Since the historical reserve prediction data is sufficiently large, a reserve prediction model can be built based on these results, and the predictions are relatively accurate. The current reserve prediction value can be directly obtained from the reserve prediction model. The predicted reserve value is compared with the total regional mineral resources predicted by the data prediction module. If the difference is small, it indicates that the total regional mineral resources predicted by the data prediction module are relatively accurate, and the total regional mineral resources are directly uploaded to the data cache unit. If the difference is large, it indicates that a large error occurred during the prediction process, and the total regional mineral resources are directly deleted. The dynamic monitoring system then re-acquires the mining area monitoring information and re-predicts reserves until the difference between the predicted total regional mineral resources and the predicted reserve value is reduced to an acceptable range.

[0080] In this embodiment, the user interaction module is used to obtain the data query request of the target user. When the target user queries the total mineral resources of a certain region within the user's client, the user interaction module can receive the data query request of the target user, and then verify the identity of the target user by comparing and verifying the user information in the data query request with the user permission information pre-entered in the user registration center. This is so that when the target user logs into the client, the client can obtain the target user's identity information. When the target user uses the client to query the total mineral resources of a certain region, the user interaction module can immediately obtain the target user's identity information and the question to be queried.

[0081] In this embodiment, if the target user's identity and permissions are successfully authenticated, the user interaction module will immediately send the total amount of regional mineral resources in the cache to the target user according to the data query request.

[0082] In one embodiment, the data classification and storage module includes:

[0083] The feature extraction submodule is used to extract the monitoring data features of the mining area monitoring information and the prediction data features of all historical prediction data in the historical prediction dataset.

[0084] The data association retrieval submodule is used to construct a monitoring data KD tree based on the monitoring data characteristics, and retrieve the reserve prediction association information that is related to the historical prediction data from the monitoring data KD tree based on the prediction data characteristics and using the nearest neighbor search method.

[0085] The classification storage submodule is used to upload the reserve prediction association information in the mining area monitoring information to the data cache unit, and to upload all information in the mining area monitoring information other than the reserve prediction association information to the main data storage unit.

[0086] In this embodiment, the core task of the feature extraction submodule is to extract key features from mining area monitoring information and historical prediction datasets. The feature extraction process typically involves various data processing techniques, such as Fourier transform for frequency domain analysis, wavelet transform for multi-scale analysis, and principal component analysis (PCA) for dimensionality reduction. For example, for geophysical data, wavelet transform can be used to separate anomaly information at different scales, calculated as: W(a,b)=(1 / √a)∫f(t)ψ((tb) / a)dt, where a is the scale parameter, b is the translation parameter, and ψ is the wavelet function. For geochemical data, PCA can be used to extract the main element combination features, reducing data dimensionality. The core of PCA is calculating the eigenvalues ​​and eigenvectors of the covariance matrix; the eigenvectors corresponding to the largest eigenvalues ​​are the principal component directions. For historical prediction data, feature extraction focuses on time series features, such as trends, periodicity, and autocorrelation. Time series decomposition techniques can be used to decompose the data into trend, seasonal, and residual components. For example, the additive model can be represented as: Y(t) = T(t) + S(t) + R(t), where T(t) is the trend component, S(t) is the seasonal component, and R(t) is the residual. These feature extraction techniques can transform raw high-dimensional data into more compact and information-rich feature vectors, laying the foundation for subsequent data association and retrieval.

[0087] The data association retrieval submodule utilizes a KD-tree (K-Dimensional Tree) structure and the nearest neighbor search algorithm to efficiently find reserve prediction association information related to historical prediction data. A KD-tree is a data structure that stores points in a k-dimensional space, making it well-suited for fast retrieval of multidimensional data. The process of constructing a KD-tree is recursive: first, a dimension is selected as the splitting dimension (usually the dimension with the largest variance). Then, the median is found along that dimension, dividing the dataset in half. This process is repeated for each half until each leaf node contains only one data point. For example, for a two-dimensional data point set {(3,6),(17,15),(13,15),(6,12),(9,1),(2,7),(10,19)}, the first step is to split along the x-axis, where the median is 9. Then, the left and right subtrees are split along the y-axis, and so on. After constructing the KD-tree, the nearest neighbor search algorithm is used to find the data point closest to the given query point. The search process starts from the root node and recursively traverses the tree downwards. At each non-leaf node, the positional relationship between the query point and the dividing hyperplane is compared to determine whether to enter the left or right subtree. Upon reaching a leaf node, the distance between the query point and the data point stored at that node is calculated. Then, backtracking is performed; if a closer point is found, the search continues in another subtree. This process can be implemented using a priority queue to ensure that the most promising region is always checked first. The average time complexity of nearest neighbor search is O(logn), where n is the number of data points, which is much faster than the O(n) of brute-force search. This method allows for the rapid identification of monitoring data most similar to historical prediction data, thereby establishing correlation information for reserve prediction.

[0088] The primary task of the categorized storage submodule is to store different parts of the mining area monitoring information into the data cache unit and the main data memory unit, based on the nature and purpose of the data. Reserve prediction correlation information is uploaded to the data cache unit, which is typically a high-speed access storage system, such as an in-memory database (e.g., Redis) or a solid-state drive (SSD) array. High-speed caching is chosen because reserve prediction correlation information is frequently accessed during subsequent prediction processes, and storing it in a fast storage device significantly improves system response speed. During data upload, data serialization (e.g., using Protocol Buffers or MessagePack) is involved to optimize storage space and transmission efficiency. Furthermore, to ensure data consistency and reliability, write confirmation mechanisms and data replication technologies are used. For example, Write-Ahead Logging (WAL) technology is employed to ensure data recovery in the event of a system crash. Asynchronous replication technology can also be used to synchronize data across multiple nodes, improving system fault tolerance. All information in the mining area monitoring information, except for reserve prediction correlation information, is uploaded to the main data memory unit. Data storage units are typically large-capacity, persistent storage systems, such as distributed file systems (like HDFS) or relational database clusters. While this data is accessed infrequently, it needs to be stored long-term for future analysis and auditing. During the upload process, data compression (e.g., using Snappy or LZ4 algorithms) is employed to save storage space, and data sharding techniques are used to achieve load balancing and parallel processing. Furthermore, data encryption technologies (such as AES-256) are applied to protect sensitive information. This categorized storage strategy achieves hierarchical optimization of data access, improves overall system performance and resource utilization efficiency, and provides flexibility for long-term data preservation and management.

[0089] In one embodiment, the data prediction module includes:

[0090] The data preprocessing submodule is used to extract the mining area exploration information and the mining area mining information from the data cache unit, and to perform data cleaning and preprocessing on the mining area exploration information and the mining area mining information. The mining area exploration information includes UAV exploration information and geological exploration information.

[0091] The model construction submodule is used to construct a surface model of the mining area based on the preprocessed UAV exploration information, construct a geological model of the target mining area based on the preprocessed geological exploration information, and obtain a three-dimensional model of the mining area by combining the surface model and the geological model of the mining area and through digital modeling.

[0092] The reserve prediction submodule is used to predict the mine's available reserves for the target mine based on the three-dimensional model of the mine and the pre-processed mining information of the mine.

[0093] The efficiency prediction submodule is used to perform data correlation analysis on the exploitable reserves and mining information of the mining area based on the three-dimensional model of the mining area, and predict the mining efficiency of the target mining area based on the data correlation analysis results.

[0094] The mineral reserves prediction submodule is used to combine the mining efficiency of all the target mineral areas and the predicted available reserves of the mineral areas to obtain the total regional mineral resources of the target area.

[0095] In this embodiment, the data extraction submodule extracts mining exploration and mining information from a high-speed data cache unit. Mining exploration information includes two main categories: UAV exploration information and geological exploration information. UAV exploration information is further subdivided into remote sensing images and oblique images of the mining area. For image data, different file formats need to be processed, such as TIFF and JPEG2000. Image processing libraries such as GDAL are used to process geospatial data during data parsing. To optimize performance, memory mapping (mmap) technology is used to directly map files to the process's address space, reducing I / O operations. For structured geological exploration information, stored in a NoSQL database such as MongoDB, appropriate query languages ​​are required for extraction. Mining information includes productivity, equipment usage, and personnel arrangements; this is typically time-series data stored in a time-series database such as InfluxDB. Extracting this data involves complex time-range queries and data downsampling. The effectiveness of data extraction directly affects the accuracy and efficiency of subsequent analysis. Efficient data extraction can significantly reduce data processing latency, supporting real-time decision-making. At the same time, accurate data extraction also ensures the integrity and consistency of data in subsequent analysis.

[0096] The model building submodule is the core of the entire system, responsible for creating an accurate 3D model of the mining area. First, remote sensing and oblique images of the mining area are preprocessed. This process includes several key steps: 1) image enhancement, using histogram equalization or adaptive histogram equalization to improve image contrast; 2) denoising, using Gaussian filtering or median filtering; 3) geometric correction, eliminating distortions caused by terrain undulations and photographic angles. After preprocessing, a surface model of the mining area is constructed using photogrammetry techniques. This typically involves the StructurefromMotion (SfM) algorithm, which reconstructs the 3D structure by identifying and matching feature points (such as SIFT or SURF features) in different images. For example, given corresponding points (x,y) and (x',y') in two images, their relationship can be described using a fundamental matrix F: x'Fx = 0. Then, bundled adjustments are used to optimize camera positions and 3D point coordinates. For geological exploration information, data cleaning is performed first to remove outliers (e.g., using the 3σ rule). The cleaned data is then used to construct the geological model of the mining area. This involves geostatistical methods, such as Kriging interpolation. For example, for a given location x0, its attribute value Z(x0) can be estimated: Z(x0) = ΣλiZ(xi), where λi is the weight and Z(xi) is the value of the known sampling point. The weights are determined by the variogram γ(h), where h is the distance between sampling points: γ(h) = 1 / (2N(h))Σ[Z(xi)-Z(xi+h)]^2.

[0097] Finally, the surface model and geological model are combined to create a 3D model of the mining area using digital modeling techniques such as voxel modeling or tetrahedral mesh modeling. This process uses Markov random fields to integrate data from different sources, ensuring model consistency. The final 3D model is not only visually accurate but also contains rich geological attribute information, providing a solid foundation for subsequent reserve prediction and mining planning. The effect of this integrated modeling method is to generate a highly accurate and information-rich digital twin of the mining area, greatly improving the accuracy of mineral resource assessment and mining decisions.

[0098] In this embodiment, the recoverable reserves prediction submodule utilizes a previously constructed 3D model of the mining area and cleaned geological exploration information to predict the recoverable reserves of the target mining area. This process involves complex geostatistics and machine learning techniques. First, based on the 3D model, geostatistical methods such as conditional simulation can be used to generate multiple orebody distribution scenarios. Conditional simulation considers data from known sampling points and simulates unknown areas while maintaining overall statistical properties. For example, using the Sequential Gaussian Simulation (SGS) method, multiple equally probable implementations can be generated. The basic steps of SGS are: 1) Define a random path to access all unknown grid nodes; 2) At each node, use Kriging interpolation to estimate the local conditional distribution; 3) Randomly sample a value from this distribution as the simulated value for that node; 4) Add this simulated value to the known dataset and continue simulating the next node. Next, the random forest method is used to predict the mineral content of each voxel. The prediction process can consider various factors, such as rock type, fault distance, geochemical indicators, etc. The training data for the model comes from known drilling data and geological exploration information. To quantify uncertainty, Monte Carlo simulation can be used. By running numerous simulations (e.g., 10,000), each using a different random seed to generate orebody distributions, and then applying a trained machine learning model for prediction, a probability distribution for reserve prediction can be obtained, rather than just a point estimate.

[0099] Finally, technical and economic factors need to be considered to determine recoverable reserves. This involves determining the cut-off grade, which is the minimum grade of ore at which mining is economically worthwhile. The cut-off grade can be calculated using the following formula: COG = (MC + PC + OC) / (R * P), where COG is the cut-off grade, MC is the mining cost, PC is the processing cost, OC is other costs, R is the recovery rate, and P is the metal price. By comprehensively considering these factors, an estimated range of recoverable reserves can be obtained, for example: "The recoverable reserves of this mine area are 1 million to 1.2 million tons within a 95% confidence interval, with a maximum value of 1.1 million tons." The advantage of this forecasting method is that it provides a comprehensive reserve assessment that considers both geological uncertainties and economic factors, offering strong support for mining companies' investment decisions and production planning.

[0100] The efficiency prediction submodule's task is to predict the mining efficiency of a target mining area by combining previously predicted reserves and mining information. This process involves a comprehensive analysis of multiple complex factors, including geological conditions, mining technology, equipment performance, and human resources.

[0101] The task of the mineral reserve prediction submodule is to integrate mining information and exploited reserves from all target mining areas to predict the total mineral output of the entire target region. This is a complex system-level prediction task that needs to consider the interrelationships between multiple mining areas and the impact of the overall geological environment. First, a comprehensive database needs to be established, containing detailed information for each mining area, such as geographical location, geological characteristics, mining history, and exploited reserve predictions. This database uses a distributed storage system such as Hadoop HDFS to process large amounts of structured and unstructured data. Next, spatial statistical methods can be used to analyze the spatial correlations between mining areas. For example, a variogram can be used to describe spatial dependencies.

[0102]

[0103] Where γ(h) is the variogram, h is the spatial lag, Z(x) is the attribute value at location x, and N(h) is the number of sample logs at interval h. Based on variogram analysis, kriging can be used for spatial interpolation to estimate the potential reserves in unexplored areas. For example, the prediction formula for ordinary kriging is:

[0104]

[0105] in, λ_i is the predicted value at position s_0, λ_i is the weight, and Z(s_i) is the observed value at the known position.

[0106] To account for geological uncertainties, stochastic simulation techniques such as Sequential Gaussian Simulation (SGS) or Multipoint Statistical (MPS) can be used. These methods can generate multiple equally probable mineral distribution scenarios, thereby quantifying the uncertainty in our minerals. Considering the impact of mineral resource extraction on the geological conditions of the surrounding area, a dynamic model can be constructed to simulate the time-varying mineral distribution. This can be achieved by combining geological evolution models and mining impact models. For example, the finite element method can be used to simulate the impact of mining on the geological stress field.

[0107]

[0108] Where σ is the stress tensor and F is the body force.

[0109] To integrate information from different sources and handle uncertainty, a Bayesian hierarchical model can be used. This allows for the combination of prior geological knowledge, historical mining data, and current observations. The Bayesian update formula is:

[0110] P(θ|D)proptoP(D|θ)P(θ)

[0111] Where θ is the model parameter, D is the observed data, P(θ|D) is the posterior distribution, P(D|θ) is the likelihood function, and P(θ) is the prior distribution.

[0112] To predict the total mineral resources of the entire region, it is necessary to consider the predicted recoverable reserves of each mining area and its uncertainties. Assume there are n mining areas, and the predicted recoverable reserves for each mining area i are Ri, with a standard error of σi. Monte Carlo simulation can be used to generate the probability distribution of the total reserves. The specific steps are as follows: 1. For each mining area i, from the normal distribution N(Ri,σi) 2 1. Sample a value from the sampled values. 2. Sum all sampled values ​​to obtain a total reserve estimate. 3. Repeat steps 1-2 multiple times (e.g., 10,000 times) to obtain the probability distribution of the total reserves. This method can provide a confidence interval for the total reserves, for example: "Within a 95% confidence interval, the total mineral reserves in this region are between X billion tons and Y billion tons." To improve the accuracy of the prediction, the correlation between mining areas can also be considered. This can be achieved by using a multivariate normal distribution, where the covariance matrix reflects the spatial correlation between mining areas. In addition, the impact of economic and technological factors on recoverable reserves needs to be considered. Sensitivity analysis can be used to assess the impact of different factors (such as metal prices, mining costs, technological advancements, etc.) on the prediction results. For example, using the partial derivative method:

[0113]

[0114] Where Si is the sensitivity index of parameter Xi to output Y.

[0115] This comprehensive forecasting method provides a complete, accurate estimate of the total regional mineral resources while taking into account uncertainties. It considers not only the individual characteristics of each mining area but also the spatial relationships between mining areas and the influence of the overall geological environment. This is of great significance for regional mineral resource planning, investment decisions, and sustainable development strategies.

[0116] In one embodiment, the reserve mobilization prediction submodule includes:

[0117] The ore body volume calculation unit is used to calculate the ore body volume of the target mining area based on the three-dimensional model of the mining area and by the integration method.

[0118] The first data extraction unit is used to extract the ore mining weight data and ore mining grade data from the mining information of the mining area. Both the ore mining weight data and the ore mining grade data are time-series data.

[0119] The reserve utilization calculation unit is used to preprocess the ore mining weight data and the ore mining grade data, combine the ore body volume with the preprocessed ore mining weight data and ore mining grade data, and calculate the time series data of the target mining area's reserves utilization using the mining area reserve utilization calculation formula.

[0120] The reserve utilization prediction unit predicts the mine utilization reserves of the target mining area based on the time series data of the reserve utilization and using the exponential smoothing method.

[0121] In this embodiment, the ore body volume calculation unit uses an integral method to calculate the ore body volume of the target mining area based on a three-dimensional model of the mining area. When calculating the volume using the integral method, the ore body is divided into several tiny volume elements, each of which can be considered as a cube or tetrahedron. The total volume is obtained by numerically integrating these elements. Specifically, Gaussian quadrature or Monte Carlo integration can be used. For example, using Gaussian quadrature, selecting n integration points, the volume calculation formula can be expressed as: V = ∑(wi*f(xi, yi, zi)), where wi is the weight and f(xi, yi, zi) is the function value at each integration point. The advantage of this method is that it can handle ore bodies with complex shapes and improve the accuracy of volume calculation. At the same time, by adjusting the density of integration points, a balance can be achieved between calculation accuracy and efficiency.

[0122] The first data extraction unit is responsible for extracting ore mining weight data and ore grade data from the mining information in the mining area. Both types of data are time-series data. The data extraction process typically involves multiple steps: First, a complete data acquisition system needs to be established, including on-site sensors, weighing equipment, and grade analysis instruments. These devices record weight and grade information in real time during the mining process. Second, a database management system (such as SQL Server or Oracle) is used to store and manage this raw data. During data extraction, SQL queries or specialized data processing software (such as Python's pandas library) are used to retrieve the required time-series data from the database. The extracted data typically includes the date, mining weight, and corresponding grade. To ensure data quality, data cleaning is also required to remove outliers and missing values. Methods such as moving averages or interpolation can be used to handle these issues.

[0123] The recoverable reserves calculation unit first preprocesses the extracted ore mining weight and grade data. Preprocessing steps include: data standardization, unifying data from different scales to the same range; outlier detection and handling, using methods such as Z-score or IQR to identify and process abnormal data points; and time series smoothing, applying techniques such as moving average or exponential smoothing to reduce short-term fluctuations. After preprocessing, combined with the previously calculated ore body volume, the recoverable reserves time series data for the target mining area are calculated using the recoverable reserves calculation formula. A typical recoverable reserves calculation formula is: R(t) = V*ρ*G(t)*(1-D)*E, where R(t) is the recoverable reserves at time t, V is the ore body volume, ρ is the ore density, G(t) is the average grade at time t, D is the dilution rate, and E is the mining recovery rate. This formula considers geological factors (volume, density), mining technology (recovery rate), and economic factors (grade). During the calculation, attention must be paid to the time-varying nature of the grade data; a sliding window method is needed to calculate the dynamic average grade. In terms of implementation effectiveness, this method can provide time-varying estimates of exploitable reserves, reflecting the dynamic process of mining in the mining area, and helps to more accurately assess the utilization and residual value of mineral resources.

[0124] The recoverable reserves prediction unit uses exponential smoothing to predict the future recoverable reserves of the target mining area based on time-series data. Exponential smoothing is a commonly used time-series forecasting method that predicts future values ​​by weighting historical data, with newer data receiving greater weight. The basic formula for single exponential smoothing is: S(t) = αY(t) + (1-α)S(t-1), where S(t) is the smoothed value at time t, Y(t) is the actual observed value, and α is the smoothing coefficient (0 < α < 1). For recoverable reserves prediction, a more complex triple exponential smoothing method (Holt-Winters method) can be considered, which can simultaneously handle trend and seasonal factors. In implementation, appropriate α, β, and γ parameters (corresponding to level, trend, and seasonality, respectively) need to be determined first, and these parameters can be optimized by minimizing the mean square error. Then, the model is trained using historical recoverable reserves data, and finally, the model is applied for future predictions. For example, the prediction formula is in the form: F(t+m)=(S(t)+mB(t))*I(t-L+m), where F(t+m) is the predicted value after m periods, S(t) is the smoothing level, B(t) is the trend factor, I(t) is the seasonality factor, and L is the length of the seasonal cycle. The advantage of this method is that it can capture the long-term trend and cyclical changes in the utilization of reserves, while also having a certain smoothing effect on short-term fluctuations.

[0125] In one embodiment, the efficiency prediction submodule includes:

[0126] The data preprocessing unit is used to preprocess the available reserves and mining information of the mining area.

[0127] Mesh partitioning unit, used to divide the three-dimensional model of the mining area into multiple mining area model regions;

[0128] The graph network construction unit is used to construct a heterogeneous graph network of the mining area by combining the preprocessed mining area reserves and mining information, as well as multiple mining area model regions.

[0129] The model building unit is used to build an efficiency prediction model based on a graph neural network model. The efficiency prediction model includes a spatiotemporal graph convolutional layer, a graph attention layer, a cross-layer information fusion module, a global pooling layer, and a fully connected layer. The spatiotemporal graph convolutional layer is used to perform spatiotemporal graph convolution operations on the spatiotemporal layers in the heterogeneous graph network of the mining area. The cross-layer information fusion module is used to fuse information from multiple different layers in the heterogeneous graph network of the mining area.

[0130] The efficiency prediction unit is used to use the mean square error of the efficiency prediction model as the loss function for predicting mining efficiency, input the heterogeneous graph network of the mining area into the efficiency prediction model, and output the mining efficiency corresponding to the target mining area through the efficiency prediction model.

[0131] In this embodiment, the data preprocessing unit preprocesses the mining area's utilized reserves and mining information, a crucial step to ensure the accuracy of subsequent analysis. First, the utilized reserves data is standardized to a uniform scale, typically using min-max scaling or Z-score standardization. For example, the formula for min-max scaling is: X_norm = (X - X_min) / (X_max - X_min), where X is the original value and X_norm is the standardized value. For mining information, missing and outlier values ​​need to be addressed. Missing values ​​can be filled using interpolation methods such as linear interpolation or more complex time-series interpolation techniques. Outlier detection can use statistical methods such as 3-sigma. Furthermore, feature engineering is required to extract meaningful features from the raw data. For example, derived features such as mining rate and grade change rate can be calculated. Time-series data needs detrending and seasonal adjustment to better capture potential patterns. Finally, correlation analysis is performed on all features to remove highly correlated redundant features, reducing the complexity of subsequent models. The effect of these preprocessing steps is to generate a cleaned, standardized, and information-rich dataset, providing high-quality input for subsequent model building and analysis, thereby improving the overall accuracy and reliability of predictions.

[0132] Meshing is responsible for dividing the 3D model of the mining area into multiple model regions, a crucial step for subsequent refined analysis. First, an appropriate mesh size needs to be determined, typically depending on the complexity of the ore body, the resolution of the available data, and computational resource limitations. A common approach is adaptive meshing using an octree structure. This method allows for finer meshes in complex or data-intensive areas, and coarser meshes in simpler areas. In practice, it can start with a cube enclosing the entire ore body and recursively divide it into eight sub-cubes until a predetermined precision is achieved or a specific stopping condition is met. For example, the division can stop when the grade variation within a cube is less than a certain threshold. Another approach is to use a regular cubic mesh, but with localized refinement in key areas (such as high-grade areas or areas with complex geological structures). After meshing, each mesh cell is assigned a unique identifier and associated with its spatial coordinates, volume, average grade, and other attributes. Furthermore, the topological relationships between meshes need to be considered, recording the information of adjacent cells for each mesh cell, which is crucial for subsequent spatial analysis. The effect of mesh generation is to discretize a continuous three-dimensional space into computable units, enabling efficient numerical calculations and spatial analysis while preserving geological features.

[0133] The graph network construction unit combines preprocessed mining area reserves, mining information, and multiple mining area model regions to construct a heterogeneous graph network for the mining area. This process first treats each grid cell as a node in the graph, with node attributes including spatial coordinates, volume, average grade, and recoverable reserves. Edges between nodes are defined based on spatial relationships and geological features. For example, adjacent grid cells can be connected, and edge weights can reflect their geological similarity or spatial distance. Furthermore, long-distance connections can be added based on geological structures (such as faults and folds) to capture complex spatial dependencies. To reflect the heterogeneity of the data, multi-layer graph networks can be constructed. For example, one layer represents spatial relationships, another represents time-series relationships, and yet another can represent relationships between geological units. Inter-layer connections can be based on node correspondences or specific mapping rules. In practice, graph data structures such as adjacency lists or adjacency matrices can be used to store the network. For large-scale mining areas, sparse matrix representation is needed to save memory. The graph construction process also includes feature embedding, which converts the attributes of nodes and edges into vector representations suitable for machine learning. This can be achieved through simple feature engineering or more complex graph embedding techniques (such as Node2Vec). The completed heterogeneous graph network not only preserves the spatial structure and attribute information of the mining area, but also captures time-series and multi-scale features, providing rich input for subsequent graph neural network models. The advantage of this representation method is that it can consider both local and global information simultaneously, which helps to more accurately predict mining efficiency and resource distribution.

[0134] The model building unit constructs an efficiency prediction model based on a graph neural network model. This model includes a spatiotemporal graph convolutional layer, a graph attention layer, a cross-layer information fusion module, a global pooling layer, and a fully connected layer. The spatiotemporal graph convolutional layer is responsible for performing convolution operations on the spatiotemporal layers in the heterogeneous graph network of the mining area to capture local spatiotemporal features. Its implementation can be based on diffusing convolution, with the formula: in The normalized adjacency matrix is ​​H^(l), the feature matrix of layer l is H^(l), the learnable weight matrix is ​​W^(l), and σ is the activation function. The graph attention layer learns the importance weights between nodes and can use a multi-head attention mechanism. The calculation formula for each attention head is: α_i j=softmax(LeakyReLU(a^T[Wh_i||Wh_j])), where h_i and h_j are node features, W is the weight matrix, and a is the attention vector. The cross-layer information fusion module integrates information from multiple layers and can use weighted summation or gating mechanisms, such as GRU units. The global pooling layer aggregates graph-level features into a fixed-dimensional vector and can use average pooling or more complex differentiable pooling operations. Finally, the fully connected layer is used for the final prediction output. During model training, the mean squared error is used as the loss function. Where y_i is the actual efficiency. This refers to predicting efficiency. Optimization algorithms can include Adam or SGDwithmomentum. To prevent overfitting, techniques such as dropout and L2 regularization can be used. The model's effectiveness lies in capturing the complex spatiotemporal dependencies of the mining area and making accurate efficiency predictions based on multi-source heterogeneous data.

[0135] The efficiency prediction unit uses the mean squared error (MSE) of the efficiency prediction model as the loss function for predicting mining efficiency. It inputs the heterogeneous graph network of the mining area into the model and outputs the mining efficiency of the target mining area. The specific implementation process is as follows: First, the constructed heterogeneous graph network data of the mining area is converted into an input format acceptable to the model, typically including node feature matrices, adjacency matrices, and edge feature matrices. Then, this data is input into the efficiency prediction model. The model's forward propagation process sequentially passes through a spatiotemporal graph convolutional layer, a graph attention layer, a cross-layer information fusion module, a global pooling layer, and a fully connected layer, finally outputting the predicted mining efficiency value. During the training phase, the model uses the mean squared error (MSE) as the loss function. Where y_i is the actual efficiency. Here, N represents the predicted efficiency, and N is the number of samples. The gradient is calculated using the backpropagation algorithm, and the model parameters are updated using an optimizer (such as Adam). During training, cross-validation can be used to evaluate model performance and adjust hyperparameters. In the prediction phase, new mining area data is input into the trained model to obtain the predicted mining efficiency. To assess the reliability of the prediction, prediction intervals or confidence levels can be calculated. Furthermore, interpretive techniques (such as SHAP values) can be used to analyze which factors contribute most to the efficiency prediction. The effectiveness of this approach is that it can accurately predict the mining efficiency of a mining area, while providing estimates of prediction uncertainties and analysis of influencing factors, thus offering strong support for mining area management and decision-making.

[0136] In one embodiment, the data preprocessing unit includes:

[0137] The data extraction subunit is used to extract mining equipment deployment data and mining personnel allocation data from the mining information of the mining area;

[0138] The time alignment subunit is used to align the mining area's utilized reserves, the mining equipment deployment data, and the mining personnel allocation data according to a preset time scale and time length.

[0139] The data interpolation subunit is used to process missing data in the mining area's available reserves, the mining equipment deployment data, and the mining personnel allocation data using a multivariate interpolation method.

[0140] In this embodiment, the data extraction subunit is responsible for extracting mining equipment deployment data and mining personnel allocation data from the mining information. This process first requires cleaning and structuring the raw data. Mining equipment deployment data typically includes information such as equipment type, quantity, location, and working status. For example, a table containing fields such as "Equipment ID," "Equipment Type," "Deployment Location," "Working Status," and "Timestamp" can be created. Mining personnel allocation data may include information such as the number of workers, job type, shift, and work location. Similarly, a table containing fields such as "Employee ID," "Job Type," "Work Location," "Shift," and "Timestamp" can be created. The data extraction process may require processing raw data in various formats, such as Excel tables, database records, or text files. Therefore, a flexible data parser needs to be developed to recognize different data formats and extract the required information. During extraction, data validation is also required to ensure that the extracted data conforms to the expected format and range. For example, rules can be set to check whether the number of equipment is a positive integer and whether the personnel's working hours are within a reasonable range. In addition, it is necessary to handle any potential duplicate records by merging or removing duplicates by comparing timestamps and unique identifiers. The final result of data extraction should be two structured datasets, one containing detailed information on equipment deployment and the other on personnel allocation, laying the foundation for subsequent time-series alignment and data analysis. The effect of this step is to transform the raw, potentially chaotic data into a clear, consistent format, significantly improving the efficiency and accuracy of subsequent analysis.

[0141] The main task of the time-series alignment subunit is to align data on mine reserves, equipment deployment, and personnel allocation according to a preset time scale and duration. This process first requires determining a unified time scale, such as days, weeks, or months. The choice of time scale must consider the characteristics of the data and the needs of the analysis; for example, if mining activities change daily, days might be more suitable. Next, the duration of the analysis needs to be determined, which typically depends on the project's requirements and could be several months or years. The specific implementation of time-series alignment can follow these steps: First, convert all data to the same time format, such as using the ISO 8601 standard date-time format (YYYY-MM-DDTHH:MM:SS). Then, create a time-series frame containing all time points from the start date to the end date. For each data type (reserves utilized, equipment deployment, personnel allocation), map it to this time-series frame. If a time point lacks corresponding data, it is temporarily marked as missing. Data aggregation or decomposition may be required during the alignment process. For example, if the original data is recorded hourly, but the time scale is set to days, then the daily data needs to be aggregated (e.g., calculating the average or sum). Conversely, if the original data is recorded monthly, but the time scale is set to days, then data decomposition (e.g., linear interpolation) is required. The result of time series alignment is a unified time series dataset containing the values ​​of all variables at each time point or marked as missing. This alignment allows data from different sources and with different frequencies to be compared and analyzed within the same time frame, providing a consistent foundation for subsequent modeling and prediction.

[0142] The data interpolation subunit processes missing data in mining area reserve utilization, mining equipment deployment, and mining personnel allocation data using multivariate interpolation. Multivariate interpolation is an advanced missing data processing technique that considers the relationships between variables, generating more accurate and consistent estimates. The specific implementation process is as follows: First, the missing data is analyzed to determine the missing pattern (completely random missing, random missing, or non-random missing). This can be done through visualization techniques or statistical tests. Then, a suitable multivariate interpolation model is selected; commonly used methods include multiple linear regression, k-nearest neighbor (k-NN) interpolation, and random forest interpolation. Taking multiple linear regression as an example, its basic idea is to use other variables to predict missing values. For each variable Y containing missing values, a regression model is established: Y = β0 + β1X1 + β2X2 + ... + βnXn + ε, where X1, X2, ..., Xn are other complete variables, β0, β1, ..., βn are regression coefficients, and ε is the error term. These coefficients are estimated using a complete subset of the data and then used to predict missing values. To introduce randomness and reflect uncertainty, multiple imputation datasets (e.g., 5-10) are typically generated. Missing values ​​in each dataset are estimated independently, and may differ slightly. In practice, multivariate imputation can be implemented using the `mice` package in R or the `sk learn.impute` module in Python. After imputation, the results need to be diagnosed and validated, such as comparing the data distribution before and after imputation to check for the introduction of unreasonable estimates. The advantage of multivariate imputation is that it preserves the relationships between variables and provides a measure of the uncertainty of missing value estimates. The effect of this method is to generate a complete dataset that retains the statistical characteristics of the original data while filling in missing values, providing a reliable foundation for subsequent analysis and modeling.

[0143] In one embodiment, the mineral reserves prediction submodule includes:

[0144] The mineral reserves prediction unit is used to obtain the mineral storage scale and average consumption rate of all target areas through the main data storage unit.

[0145] If the mining efficiency of the target area is less than or equal to the average consumption rate, the mineral reserve prediction unit combines the mineral reserves used, mining efficiency and average consumption rate to predict the total mineral resources of the entire target area.

[0146] If the mining efficiency of the target area is greater than the average consumption rate, the mineral reserve prediction unit will use the mineral storage scale as a constraint on the regional mineral reserves, and predict the total regional mineral reserves of the entire target area based on the mineral reserves utilized, mining efficiency and average consumption rate.

[0147] In this embodiment, the mineral reserve prediction unit is used to obtain the mineral storage capacity and average consumption rate of all target areas through the main data storage unit. The mineral storage capacity is the maximum amount of mineral that can be stored in the target area, and the average consumption rate is the average rate at which mineral products are consumed. If the mining efficiency of the target area is less than or equal to the average consumption rate, then theoretically the total mineral resources in the area will not exceed the mineral storage capacity, so there is no need to use the mineral storage capacity as a constraint. If the mining efficiency of the target area is greater than or equal to the average consumption rate, then theoretically, if the mineral reserves used in the mining area are greater than the mineral storage capacity, then the amount of mineral resources mined may exceed the mineral storage capacity. Therefore, when predicting the total mineral resources in the area, the mineral storage capacity should be used as a constraint. The total mineral resources in the area cannot be greater than the predicted total mineral resources in the area. When the total mineral resources in the area are equal to the predicted total mineral resources in the area, an alarm message will be sent and mining will be stopped. Similarly, the predicted total mineral resources in a region cannot exceed the sum of the usable reserves of all target mining areas in the target region. Based on mining efficiency and average consumption rate, the increase in mineral reserves in the current time period can be predicted. Then, the historical total mineral resources in the region are obtained from the main data storage unit. The increase in mineral reserves and the historical total mineral resources in the region are added together to obtain the total mineral resources in the target region. The same method is used to calculate the total mineral resources in other target regions to obtain the total mineral resources in all target regions.

[0148] In one embodiment, the graph network building unit includes:

[0149] The first layer construction subunit is used to construct a mining area spatial layer by taking each of the mining area model regions as the first graph node and taking the spatial adjacency relationship of each of the mining area model regions as the first graph node edge of the first graph node.

[0150] The second layer construction subunit is used to construct a mining spatiotemporal layer by taking the mining area status of all the mining area model areas at different time points in the time length as the second graph node and the time interval between different time points as the second graph node edge of the second graph node.

[0151] The third layer construction subunit is used to construct a device collaboration layer by taking the mining equipment in the mining equipment deployment data as the third graph node and the device collaboration relationship of any number of mining equipment at the same time point as the third graph node edge of the third graph node.

[0152] The fourth layer construction subunit is used to construct a personnel collaboration layer by taking the mining personnel in the mining personnel allocation data as the fourth graph node and the personnel collaboration relationship of any number of mining personnel at the same time point as the fourth graph node edge of the fourth graph node.

[0153] The graph network construction subunit is used to construct a heterogeneous graph network of the mining area by combining the spatial layer of the mining area, the spatiotemporal layer of the mining area, the equipment collaboration layer, and the personnel collaboration layer.

[0154] In this embodiment, the first layer construction subunit is responsible for constructing the mining area spatial layer, which is a graph model representing the spatial structure of the mining area. In this graph, each mining area model region is treated as a node, and the spatial adjacency between regions is represented by edges between nodes. In practice, two regions sharing a boundary are generally considered to be adjacent. This step can be implemented using Geographic Information System (GIS) technology, such as ArcGIS or QGIS software. The mining area can be divided into several polygonal regions, and then spatial analysis tools can be used to determine which polygons are adjacent to each other. For example, the Delaunay triangulation algorithm can be used to identify adjacency relationships. Once the adjacency relationships are determined, the graph structure can be constructed. Each region is assigned a unique identifier as a node in the graph. If two regions are adjacent, an edge is added between their corresponding nodes. This can be represented using an adjacency matrix or an adjacency list. For example, for an adjacency matrix A, if regions i and j are adjacent, then A[i][j] = A[j][i] = 1, otherwise it is 0. This representation method facilitates subsequent graph analysis operations, such as calculating the shortest path or cluster analysis. The effect of constructing this spatial layer is that it provides a mathematical representation of the spatial structure of the mining area, enabling the application of graph theory and network analysis methods to study the spatial characteristics of the mining area, such as identifying key areas and optimizing resource allocation. Furthermore, this layer also lays the foundation for subsequent spatiotemporal analysis and the construction of multi-layer graph networks.

[0155] The second layer construction subunit is responsible for constructing the spatiotemporal layer of the mining area, a complex graph structure that integrates spatial and temporal dimensions. In this graph, nodes represent the state of a mining area model region at different points in time, while edges represent the passage of time. In practice, the time resolution needs to be determined first, for example, by choosing daily, weekly, or monthly as a time unit. For each point in time, a node is created for each mining area. The node's attributes include a timestamp and the region's state at that moment (such as mining output, remaining reserves, etc.). Edges between nodes represent the continuity of time and are usually directed, pointing from earlier time points to later time points. The weight of the edge can be set to the length of the time interval; for example, if it is daily data, the weight of the edge between adjacent nodes is 1. This structure can be represented by a three-dimensional tensor, where two dimensions represent space (corresponding to the nodes in the first layer), and the third dimension represents time. In practical applications, a sparse matrix can be used to store this structure to save memory space. For example, for a time series t1, t2, ..., t... n It is possible to construct n matrices M1, M2, ..., M n Mi Representing time t i The spatiotemporal relationships are represented by the connections between matrices, indicating the passage of time. This representation allows for capturing the evolution of a mining area's state over time; for example, the mining progress of a region can be analyzed by comparing node attributes at different points in time. The effect of constructing this spatiotemporal layer is that it provides a unified framework for analyzing the spatial patterns of a mining area's changes over time. This enables complex spatiotemporal analyses, such as identifying trends in mining hotspots and predicting future resource distribution.

[0156] The third layer construction subunit is responsible for building the equipment collaboration layer, a dynamic graph structure representing the collaborative relationships between mining equipment. In this graph, nodes represent individual mining equipment, while edges represent the collaborative relationships between equipment. In practice, it's first necessary to extract information about each piece of equipment from the mining equipment deployment data, including equipment ID, type, and location. Each piece of equipment acts as a node in the graph, and node attributes can include various characteristics of the equipment. Collaborative relationships between equipment can be defined based on various factors, such as physical distance (e.g., equipment working in the same area), functional complementarity (e.g., the coordination of excavators and transport vehicles), or time synchronization (equipment operating simultaneously). These relationships are represented as edges in the graph, and the edge weights reflect the strength or frequency of collaboration. Since equipment deployment and collaborative relationships change over time, this layer needs to construct a graph structure at each time point. This dynamic graph can be represented using a time-series adjacency matrix. For example, for time point t, the adjacency matrix A... t Element A in t [i][j] represents the cooperation strength between devices i and j at time t. To capture the temporal dynamics of device cooperation, a time decay factor λ can be introduced, causing the influence of past cooperation relationships on the present to weaken over time: A t =λA t-1 +(1-λ)C t C t This represents the collaborative relationships observed at the current point in time. This representation allows for the analysis of the evolution of equipment collaboration patterns, such as identifying key equipment and optimizing equipment configuration. The effect of constructing this equipment collaboration layer is that it provides a tool for analyzing equipment collaboration efficiency. Through this layer, bottlenecks in equipment collaboration can be identified, equipment scheduling can be optimized, and the impact of equipment failures on overall production can be predicted, thereby improving the overall operational efficiency of the mining area.

[0157] The fourth layer construction subunit is responsible for building the personnel collaboration layer, a dynamic social network structure representing the collaborative relationships among mining personnel. In this graph, nodes represent individual mining personnel, while edges represent the collaborative relationships between them. In practice, it's first necessary to extract information for each employee from the mining personnel allocation data, including employee ID, position, work location, shift, etc. Each employee is a node in the graph, and node attributes can include various employee characteristics. Collaborative relationships between personnel can be defined based on various factors, such as: being in the same work group, having the same work location, overlapping work hours, or direct work interaction. These relationships are represented as edges in the graph, and the edge weights reflect the strength or frequency of collaboration. Since personnel allocation and collaborative relationships change over time, this layer needs to construct a graph structure at each time point. This dynamic graph can be represented using a time-series adjacency matrix. For example, for time point t, the adjacency matrix B... t Element B in t [i][j] represents the collaboration intensity between employees i and j at time t. To capture the temporal dynamics and cumulative effects of employee collaboration, the Exponentially Weighted Moving Average (EWMA) method can be introduced: B t =αB t-1 +(1-α)D t D t This represents the collaborative relationships observed at the current point in time, where α is a smoothing factor (0 < α < 1). This representation allows for the analysis of the evolution of personnel collaboration patterns, such as identifying key personnel and optimizing team composition. The effect of constructing this personnel collaboration layer is that it provides a tool for analyzing personnel collaboration efficiency and organizational structure. Through this layer, informal organizational structures can be identified, information flow efficiency can be assessed, personnel scheduling can be optimized, and the impact of personnel changes on production can be predicted, thereby improving the efficiency of human resource management and overall production efficiency in the mining area.

[0158] The graph network construction subunit is responsible for combining the four layers (mine area spatial layer, mine area spatiotemporal layer, equipment collaboration layer, and personnel collaboration layer) built earlier to construct a comprehensive heterogeneous graph network for the mine area. This heterogeneous graph network is a complex, multi-layered, and multi-dimensional structure capable of comprehensively representing the relationships between various aspects of the mine area, including space, time, equipment, and personnel. In practical implementation, it is necessary to define the connection methods between different layers. For example, the connection between the mine area spatial layer and the spatiotemporal layer can be based on the correspondence of spatial locations; the connection between the spatiotemporal layer and the equipment collaboration layer can be based on the location of equipment at a specific point in time; and the connection between the equipment collaboration layer and the personnel collaboration layer can be based on the allocation relationship between operators and equipment. These cross-layer connections can be represented using tensors. For example, a fourth-order tensor T can be defined, where T[i, j, k, l] represents the relationship strength between the k-th equipment in the i-th spatial region at the j-th time point and the l-th personnel's operation. In practical applications, since this tensor is very sparse, sparse representation methods can be used for storage and processing. To analyze this heterogeneous graph network, graph embedding techniques, such as Node2Vec or GraphSAGE, can be used to map nodes to a low-dimensional vector space, thereby capturing the structural features and semantic information of the nodes. For example, for node v, its embedding vector can be represented as: f(v) = σ(W·AGGREGATE(f(u):u∈N(v)))W, where N(υ) is the set of v's neighbors, AGGREGATE is the aggregation function, σ is the nonlinear activation function, and W is a learnable weight matrix. The effect of constructing this comprehensive heterogeneous graph network is that it provides a unified framework for analyzing various aspects of the mining area and their interactions. Through this network, comprehensive data mining and pattern recognition can be performed, such as predicting resource distribution, optimizing production processes, and identifying potential risks, thereby supporting smarter and more efficient mining area management decisions.

[0159] In one embodiment, the data verification module includes:

[0160] The second data extraction unit is used to extract historical reserve prediction results from the main data storage unit;

[0161] The prediction model building unit is used to fit the historical reserve prediction results using the least squares method to obtain the reserve prediction model;

[0162] The data verification unit is used to obtain the predicted reserves of the target area based on the reserve prediction model, calculate the difference between the predicted reserves and the total mineral resources of the area, and if the difference is less than the preset difference threshold, the verification is successful and the total mineral resources of the area are transmitted to the data cache unit.

[0163] In this embodiment, the second data extraction unit is responsible for extracting historical reserve prediction results from the main data storage unit. This is a crucial foundational step for subsequent analysis and modeling. In practice, this process involves multiple stages, including database querying, data cleaning, and formatting. First, the data range to be extracted needs to be determined, including the time span and geographical scope. For example, it may be necessary to extract reserve prediction data for a specific mining area over the past five years. Next, relevant data is retrieved from the main data storage unit (usually a large relational database or data warehouse) using SQL or a similar query language. After acquiring the data, it is formatted into a structure suitable for subsequent analysis, typically a two-dimensional array or data frame, where one column is a timestamp and the other column is the corresponding reserve prediction value. The effect of this process is to generate a clean, structured historical reserve prediction dataset, providing reliable input for subsequent model building.

[0164] The prediction model building unit utilizes historical reserve prediction results obtained from the second data extraction unit and fits them using the least squares method to construct a reserve prediction model. This process essentially involves finding a mathematical function that best describes the changing trends of historical data and can be used for future predictions. Least squares is a commonly used regression analysis method; its core idea is to minimize the sum of squared errors between the predicted and actual values. In practice, a suitable functional form needs to be chosen as the model. Common choices include linear functions, polynomial functions, or exponential functions. For example, if a quadratic polynomial model is chosen, its form is y = ax^2 + bx + c, where y is the predicted reserve and x is time (which could be the number of days from a certain base date). Then, the least squares method is used to determine the values ​​of parameters a, b, and c. Specifically, for n historical data points (x_i, y_i), the following expression needs to be minimized: This can be solved by taking the derivative and setting it to zero, or by using numerical optimization methods such as gradient descent. In practical applications, statistical software packages (such as R's `lm` function or Python's `numpy.polyfit`) can be used to automate this process. After the model is fitted, its performance needs to be evaluated, typically using metrics such as the coefficient of determination (R^2) and root mean square error (RMSE).

[0165] The data validation unit is responsible for validating the constructed reserve prediction model to ensure that the model's predictions match the actual situation. This process involves multiple steps, including model application, error calculation, and threshold comparison. First, a target region needs to be selected for validation. This region should have reliable actual reserve data but was not used during model training. The previously constructed reserve prediction model is used to predict the reserves in this region. The assumed model is y = ax^2 + bx + c, where x is the number of days from a certain base date. The x value corresponding to the target date is substituted to calculate the predicted reserve y. Next, this predicted value is compared with the actual total mineral reserves in the region. The calculation formula is as follows: Difference = |Predicted Reserves - Actual Reserves|. Then, this difference is compared with a pre-set difference threshold. The setting of this threshold needs to consider several factors, including the acceptable error range and the degree of fluctuation in historical data. For example, if the threshold is set to 5% of the actual reserves, then for a mining area with actual reserves of 10 million tons, the maximum allowable error is 500,000 tons. If the calculated difference is less than this threshold, the validation is considered successful, and the model's prediction results are acceptable. In this scenario, the total regional mineral resources are transferred to a data cache unit for later use. If the difference exceeds a threshold, it indicates a problem with the model, requiring a review of the model building process, collection of more data, or experimentation with different model formats.

[0166] This application also discloses a big data dynamic monitoring method for mineral products, applied to the big data dynamic monitoring system for mineral products described in any of the above embodiments. The system includes a cloud server and a user registration center, which are communicatively connected. The cloud server includes a data storage module, which includes a data cache unit and a main data storage unit. The main data storage unit pre-stores historical mineral reserve prediction records, which include historical prediction datasets and historical reserve prediction results. (Refer to...) Figure 2 The method includes the following steps:

[0167] S101. Obtain mining area monitoring information for all target mining areas within the target region from the mineral product whole process management system;

[0168] S102. Based on the historical prediction dataset, the monitoring information of the mining area is classified by feature to obtain the reserve prediction association information. The reserve prediction association information is uploaded to the data cache unit, and the monitoring information of the mining area after removing the reserve prediction association information is uploaded to the main data storage unit. The reserve prediction association information includes mining area exploration information and mining area mining information.

[0169] S103. Extract mining exploration information and mining mining information from the data cache unit. Based on the mining exploration information and by constructing a three-dimensional model of the mining area, predict the exploitable reserves of the target mining area. Combine the mining mining information and exploitable reserves of all target mining areas to predict the total mineral resources of the target region.

[0170] S104. Verify the total mineral resources in the region based on historical reserve prediction results. If the verification is successful, the total mineral resources in the region will be transmitted to the data cache unit.

[0171] S105. Obtain the target user's data query request, obtain the user permission information pre-entered in the user registration center based on the user information in the data query request, and perform identity and permission authentication on the target user based on the user permission information.

[0172] S106. If the target user's identity and authorization are successfully authenticated, the user interaction module will send the total amount of regional mineral resources in the cache to the target user according to the data query request.

[0173] In this embodiment, the first step is to define the scope of the target area, which is typically defined by geographic coordinates or administrative divisions. For example, the focus might be on all mining areas within a province, or mining areas within a specific latitude and longitude range. Next, it's necessary to connect to the database of a mineral product end-to-end management system. This system is usually a comprehensive information management platform containing data from exploration and mining to sales. The connection requires the use of a specific API or database query language. The query language returns information on all mining areas within the target region. Mining area monitoring information typically includes several aspects: 1) geological information, such as rock type and ore body distribution; 2) mining information, such as mining methods, mining depth, and daily output; 3) environmental monitoring data, such as dust concentration and noise levels; 4) equipment operating status, such as excavator working hours and transport vehicle travel distances; and 5) safety information, such as accident records and safety inspection results. This data comes from various sensors, equipment logs, and manual records. After acquiring the data, preliminary data cleaning and formatting are required. This includes handling missing values ​​(using interpolation or direct deletion), standardizing data formats (e.g., ensuring all dates are in the same format), and removing obvious outliers. Finally, the processed data is integrated into a unified data structure.

[0174] Next, it's necessary to determine which features are most relevant to reserve predictions based on historical forecast datasets. This involves statistical analysis and machine learning techniques, such as correlation analysis, principal component analysis (PCA), or feature importance assessment. For example, it might be found that the geological structure, mining depth, and annual output of a mining area are highly correlated with reserve predictions. Based on these analyses, mining area monitoring information is divided into two categories: reserve prediction-related information and non-related information. Reserve prediction-related information mainly includes mining exploration information and mining information. Mining exploration information includes geological drilling data, geophysical survey results, ore body distribution maps, etc. Mining information includes daily output, cumulative output, mining methods, equipment utilization rates, etc. Next, this information needs to be structured. For example, a data frame might be created where each row represents a mining area, and columns include various exploration and mining-related features. For complex information, such as geological maps or 3D models, feature extraction is required to convert it into numerical features. For example, numerical features such as the area and depth distribution of ore bodies can be extracted from geological maps. After processing, the reserve prediction correlation information is uploaded to the data cache unit, which is typically a high-speed access storage system, such as an in-memory database or a file system on a solid-state drive. Simultaneously, the mine area monitoring information, after removing the reserve prediction correlation information, is uploaded to the main data storage unit, a large-capacity disk storage system. This step transforms the raw mine area monitoring information into more targeted and structured data, providing high-quality input for subsequent reserve prediction models. Furthermore, by storing the data separately, system efficiency is improved, allowing subsequent prediction calculations to quickly access the most relevant data.

[0175] Next, the previously stored exploration and mining information needs to be read from the data cache unit. This information includes drilling data, geophysical survey results, historical mining records, etc. Next, a 3D geological model of the mining area is constructed using the exploration information. This is typically done using specialized geological modeling software such as Leapfrog or GOCAD. The modeling process includes the following steps: 1) data import and cleaning; 2) stratigraphic interpretation and orebody boundary definition; 3) 3D mesh generation; 4) attribute interpolation (such as grade distribution). For example, kriging interpolation is used to estimate the grade distribution of the entire orebody. After the model is built, the recoverable reserves of the mining area can be calculated. Recoverable reserves are usually defined as the amount of economically recoverable mineral resources, and the calculation formula is as follows: Recoverable reserves = Σ(volume i * density i * grade i * recoverability i), where i represents different orebody segments. Next, the total mineral resources of the target area need to be predicted by combining the mining information of all target mining areas and the calculated recoverable reserves. This involves time series analysis and forecasting techniques. For example, an ARIMA (Autoregressive Integrated Moving Average) model can be used to predict future mining output, and then the predicted output can be combined with current utilized reserves to obtain a predicted total mineral output for the region. The prediction formula is as follows: Total mineral output for the region = Σ(utilized reserves of mining area i) - Σ(predicted mining output of mining area i) + new reserves, where new reserves can be estimated based on historical exploration data and geological inferences.

[0176] Next, historical reserve forecast results need to be extracted from the database. This historical data typically includes forecast and actual values ​​from the past few years, obtained using SQL queries. Then, the accuracy of the historical forecasts needs to be calculated. A common method is to calculate the Mean Absolute Percentage Error (MAPE): MAPE = (1 / n) * ∑(|actual value - forecast value| / actual value) * 100%, where n is the number of historical data points. For example, if the MAPE over the past 5 years is 5%, a similar threshold would be set to validate the new forecast. Then, the newly predicted total regional mineral reserves are compared with the latest actual total reserves. Assuming the forecast is 10 million tons and the actual value is 9.8 million tons, the difference rate is (1000-980) / 980*100% ≈ 2.04%. If this difference rate is less than the set threshold (e.g., 5%), the validation is considered successful. After successful validation, the validated total regional mineral reserves need to be transferred to the data cache unit. If validation fails, an alert or notification needs to be triggered, prompting relevant personnel to review the forecast model or input data.

[0177] The system receives a data query request from a user. This request is typically submitted via API call or web interface, containing the user's identity information (such as user ID or token) and specific query parameters. Next, the system parses the request to extract the user information. Then, the system connects to the user registry center, which is usually a dedicated database or identity management system. Permission information includes multiple levels, such as "read-only," "edit," and "administrator." The system compares the obtained user permission information with the data type of the request. For example, if the user requests sensitive mineral data, the system requires the user to have at least "senior analyst" level permissions. This comparison process involves a complex rule engine, considering multiple factors such as user role, data sensitivity, and access time. If the user's permission level meets the requirements, the system marks the request as "authenticated." If permissions are insufficient, the system generates an "access denied" response. Furthermore, the system logs all access attempts, both successful and failed, for subsequent auditing and security analysis. This step establishes a security barrier, ensuring that only authorized users can access sensitive mineral data.

[0178] After user authentication is successful, the system retrieves relevant regional mineral resource totals information from the data cache unit based on the user's data query request and sends this information to the target user. Specifically, the system needs to parse the user's data query request and extract key parameters, such as the target region and time range. For example, if the user requests "total mineral resources in XX region in 2023," the system will translate this request into specific data query conditions. Next, the system connects to the data cache unit, which is typically a high-speed storage system such as Redis or Memcached. Using the previously parsed query conditions, the system executes a data retrieval operation, such as: GET "minera l_reserves:XX:2023". If the requested data exists in the cache, the system can quickly obtain the corresponding regional mineral resource totals information. If the data is not in the cache, the system needs to fall back to the main data storage for querying and then update the cache. After obtaining the data, the system needs to format the data to meet the requirements of the user interface. This includes unit conversion (e.g., from tons to tens of thousands of tons), data rounding, and adding metadata (e.g., data update time, data source, etc.). Formatted data is typically converted into standard formats such as JSON or XML for easy transmission over the network and parsing on the client side. Finally, the system sends this data to the target user via an HTTP response or other agreed-upon communication protocol. During transmission, encryption measures are applied to protect the security of data transmission.

[0179] This application also discloses a big data dynamic monitoring device for mineral products, including a big data dynamic monitoring system for mineral products according to any of the above embodiments.

[0180] The present invention also discloses a robotic device, which may include the artificial intelligence-based voice recognition device described in the above embodiments, at least one network interface, user interface, and communication bus.

[0181] The communication bus is used to enable communication between these components.

[0182] The user interface may include a display interface and a camera interface. Optional user interfaces may also include standard wired interfaces and wireless interfaces.

[0183] The network interface may include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0184] In robotic devices, the user interface primarily provides an input interface for users and acquires user input data. The processor can call an application program stored in memory for simulating communication interference training. When executed by one or more processors, this causes the robotic device to perform one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0185] This application also provides a machine-readable storage medium storing instructions that cause a machine to execute the above-described method for dynamic monitoring of big data for mineral products.

[0186] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0187] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0188] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0190] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0191] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0192] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media do not include transient computer-readable media, such as modulated data signals and carrier waves.

[0193] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0194] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A big data dynamic monitoring system for mineral products, characterized in that, The system includes a cloud server and a user registration center, which are communicatively connected. The cloud server includes a data storage module, which comprises a data cache unit and a main data storage unit. The main data storage unit pre-stores historical mineral reserve prediction records, which include historical prediction datasets and historical reserve prediction results. The cloud server also includes: The data acquisition module is used to obtain mining area monitoring information of all target mining areas within the target region from the mineral product full-process management system; The data classification and storage module is used to classify the mining area monitoring information by feature based on the historical prediction dataset, obtain the reserve prediction association information, upload the reserve prediction association information to the data cache unit, and upload the mining area monitoring information after removing the reserve prediction association information to the main data storage unit. The reserve prediction association information includes mining area exploration information and mining area mining information. The data prediction module is used to extract mining exploration and mining information from the data cache unit. Based on the mining exploration information, it predicts the exploitable reserves of the target mining area by constructing a 3D model of the mining area. It is also used to preprocess the exploitable reserves and mining information of the mining area. The 3D model of the mining area is divided into multiple mining area model regions. Each mining area model region is used as the first graph node, and the spatial adjacency relationship of each mining area model region is used as the first graph node edge of the first graph node to construct a mining area spatial layer. The mining area region status at different time points in the time length is used as the second graph node, and the time interval between different time points is used as the second graph node edge of the second graph node to construct a mining area spatiotemporal layer. The mining equipment in the mining equipment deployment data is used as the third graph node, and the equipment cooperation relationship of any number of mining equipment at the same time point is used as the third graph node edge of the third graph node to construct an equipment cooperation layer. The mining personnel in the mining personnel allocation data is used as the fourth graph node, and the personnel cooperation relationship of any number of mining personnel at the same time point is used as the fourth graph node edge of the fourth graph node to construct a personnel cooperation layer. The data prediction module is also used to construct a heterogeneous graph network of the mining area by combining the spatial layer, spatiotemporal layer, equipment collaboration layer, and personnel collaboration layer of the mining area. Based on the graph neural network model, an efficiency prediction model is constructed. The mean square error of the efficiency prediction model is used as the loss function for predicting mining efficiency. The heterogeneous graph network of the mining area is input into the efficiency prediction model. The mining efficiency of the corresponding target mining area is output through the efficiency prediction model. By combining the mining efficiency of all target mining areas and the mining area's utilized reserves, the total mineral resources of the target area are predicted. The data verification module is used to verify the total mineral resources in the region based on historical reserve prediction results. If the verification is successful, the total mineral resources in the region will be transmitted to the data cache unit. The user interaction module is used to obtain the data query request of the target user, obtain the user permission information pre-entered in the user registration center based on the user information in the data query request, and perform identity and permission authentication of the target user based on the user permission information. If the target user's identity and permissions are successfully authenticated, the user interaction module is also used to send the total amount of regional mineral resources in the data cache unit to the target user according to the data query request.

2. The system according to claim 1, characterized in that, The data classification and storage module includes: The feature extraction submodule is used to extract the monitoring data features of the mining area monitoring information and the prediction data features of all historical prediction data in the historical prediction dataset. The data association retrieval submodule is used to construct a monitoring data KD tree based on the characteristics of the monitoring data, and to retrieve the reserve prediction association information that is related to historical prediction data from the monitoring data KD tree based on the characteristics of the prediction data and using the nearest neighbor search method. The classification storage submodule is used to upload the reserve prediction correlation information in the mining area monitoring information to the data cache unit, and to upload all information in the mining area monitoring information except for the reserve prediction correlation information to the main data storage unit.

3. The system according to claim 1, characterized in that, The data prediction module includes: The data preprocessing submodule is used to extract mining exploration information and mining information from the data cache unit, and to perform data cleaning and preprocessing on the mining exploration information and mining information. The mining exploration information includes UAV exploration information and geological exploration information. The model building submodule is used to build a surface model of the mining area based on the preprocessed UAV exploration information, build a geological model of the target mining area based on the preprocessed geological exploration information, and obtain a three-dimensional model of the mining area by combining the surface model and the geological model of the mining area and through digital modeling. The reserve prediction submodule is used to predict the available reserves of the corresponding target mining area based on the three-dimensional model of the mining area and the preprocessed mining information.

4. The system according to claim 3, characterized in that, The reserve mobilization prediction submodule includes: The ore body volume calculation unit is used to calculate the ore body volume of the target mining area based on a three-dimensional model of the mining area and by using the integral method. The first data extraction unit is used to extract ore mining weight data and ore mining grade data from the mining information of the mining area. Both ore mining weight data and ore mining grade data are time series data. The reserve utilization calculation unit is used to preprocess ore mining weight data and ore mining grade data, combine the ore body volume with the preprocessed ore mining weight data and ore mining grade data, and calculate the time series data of the target mining area's reserves using the mining area reserve utilization calculation formula. The reserve utilization prediction unit predicts the mine utilization reserves of the target mining area based on the time series data of the reserves utilization and using the exponential smoothing method.

5. The system according to claim 1, characterized in that, The efficiency prediction model includes a spatiotemporal graph convolutional layer, a graph attention layer, a cross-layer information fusion module, a global pooling layer, and a fully connected layer. The spatiotemporal graph convolutional layer is used to perform spatiotemporal graph convolution operations on the spatiotemporal layers in the heterogeneous graph network of the mining area, and the cross-layer information fusion module is used to fuse information from multiple different layers in the heterogeneous graph network of the mining area.

6. The system according to claim 1, characterized in that, The data preprocessing unit includes: The data extraction subunit is used to extract mining equipment deployment data and mining personnel allocation data from the mining information of the mining area; The time-series alignment subunit is used to align the data on the utilization of reserves in the mining area, the deployment of mining equipment, and the allocation of mining personnel according to a preset time scale and time length. The data interpolation subunit is used to process missing data in mining area utilization reserves, mining equipment deployment data, and mining personnel allocation data using multivariate interpolation.

7. The system according to claim 1, characterized in that, The data verification module includes: The second data extraction unit is used to extract historical reserve prediction results from the main data storage unit; The prediction model building unit is used to fit the historical reserve prediction results using the least squares method to obtain the reserve prediction model; The data verification unit is used to obtain the predicted reserves of the target area based on the reserve prediction model, calculate the difference between the predicted reserves and the total mineral resources of the area, and if the difference is less than the preset difference threshold, the verification is successful and the total mineral resources of the area are transmitted to the data cache unit.

8. A method for dynamic monitoring of mineral products using big data, characterized in that, The method, applied to the big data dynamic monitoring system for mineral products according to any one of claims 1 to 7, comprises the following steps: Obtain mining area monitoring information for all target mining areas within the target region from the mineral product end-to-end management system; Based on historical prediction datasets, the monitoring information of the mining area is classified by features to obtain the reserve prediction association information. The reserve prediction association information is uploaded to the data cache unit, and the monitoring information of the mining area after removing the reserve prediction association information is uploaded to the main data storage unit. The reserve prediction association information includes mining area exploration information and mining area mining information. Mining exploration and mining information are extracted from the data cache unit. Based on the mining exploration information, the exploitable reserves of the target mining area are predicted by constructing a 3D model of the mining area. The exploitable reserves and mining information of the mining area are preprocessed. The 3D model of the mining area is divided into multiple mining area model regions. Each mining area model region is used as the first graph node, and the spatial adjacency relationship of each mining area model region is used as the first graph node edge of the first graph node to construct a mining area spatial layer. The mining area region status at different time points in the time length of all mining area model regions is used as the second graph node, and the time interval between different time points is used as the second graph node edge of the second graph node to construct a mining area spatiotemporal layer. The mining equipment in the mining equipment deployment data is used as the third graph node, and the equipment cooperation relationship of any number of mining equipment at the same time point is used as the third graph node edge of the third graph node to construct an equipment cooperation layer. The mining personnel in the mining personnel allocation data is used as the fourth graph node, and the personnel cooperation relationship of any number of mining personnel at the same time point is used as the fourth graph node edge of the fourth graph node to construct a personnel cooperation layer. A heterogeneous graph network of the mining area is constructed by combining the spatial layer, spatiotemporal layer, equipment collaboration layer, and personnel collaboration layer of the mining area. An efficiency prediction model is built based on a graph neural network model. The mean square error of the efficiency prediction model is used as the loss function for predicting mining efficiency. The heterogeneous graph network of the mining area is input into the efficiency prediction model. The mining efficiency of the corresponding target mining area is output by the efficiency prediction model. The total mineral resources of the target area are predicted by combining the mining efficiency of all target mining areas and the utilized reserves of the mining area. The total mineral resources in the region are verified based on historical reserve prediction results. If the verification is successful, the total mineral resources in the region are transmitted to the data cache unit. Obtain the target user's data query request, retrieve the user permission information pre-entered in the user registration center based on the user information in the data query request, and perform identity and permission authentication on the target user based on the user permission information; If the target user's identity and permissions are successfully authenticated, the user interaction module will send the total amount of regional mineral resources in the cache to the target user based on the data query request.

9. A big data dynamic monitoring device for mineral products, characterized in that, Including a big data dynamic monitoring system for mineral products according to any one of claims 1 to 7.

Images