Intelligent mine mining method and equipment based on industrial cloud platform and medium

Abstract

The invention discloses a smart mine mining method and device based on an industrial cloud platform, and a medium, and relates to the technical field of smart mining, and the method comprises the steps: enabling a digital twinborn analysis module to communicate with a data governance fusion module through a data service bus, and receiving mine comprehensive state data outputted by the data governance fusion module, constructing a three-dimensional geologic model, carrying out abnormal behavior detection through a deep learning algorithm, and outputting an abnormal detection result; and the intelligent decision optimization module is associated with the digital twinborn analysis module through an algorithm cooperation interface, and is used for establishing a disaster prediction model and performing disaster risk assessment based on an abnormal detection result and mine comprehensive state data, and optimizing a mining strategy by using a reinforcement learning algorithm to obtain an optimal mining scheme. And the data is transmitted back to the data acquisition upper cloud module through a feedback control link to guide field acquisition and scheduling. And the overall safety, the operation efficiency and the intelligent level of a mine system are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent mining, in particular to an intelligent mine mining method, equipment and medium based on an industrial cloud platform. BACKGROUND

[0002] With the rapid development of industrial internet and digital technology, intelligent mines, as an important direction of the transformation and upgrading of the mining industry, have received widespread attention and practice at home and abroad in recent years. In order to improve the efficiency and safety of mine operation, mine monitoring systems based on Internet of Things (IoT) and cloud computing have been gradually popularized and applied in recent years. In the prior art, there have been studies that collect mine environment parameters (such as gas concentration, surrounding rock stress and temperature and humidity, etc.) by deploying a sensor network, and transmit the data to a local server for centralized monitoring, to realize preliminary sensing and early warning of the running state of the mine. In addition, some systems introduce geographic information systems (GIS) and three-dimensional modeling technology to visualize the spatial distribution of ore bodies, to assist in mining design and production planning. To some extent, this kind of technical path improves the informatization level of the mine, promotes the transformation from "experience-driven" to "data-driven", and lays a foundation for the in-depth development of intelligent mines.

[0003] The prior art still has obvious limitations in the deep fusion of multi-source heterogeneous data and dynamic decision support. On the one hand, mine geology, equipment status and production progress data are usually collected and managed independently by different subsystems, lacking a unified data integration architecture, resulting in problems such as inconsistent data timing and semantic heterogeneity, making it difficult to form a comprehensive understanding of the overall running state of the mine; on the other hand, most current systems still remain at the "monitoring-alarm" level, lacking intelligent analysis capabilities for the evolution process of complex geological structures, making it difficult to identify potential rock mass instability, water inrush and other disaster precursor characteristics in advance. The above problems restrict the autonomous decision-making level of the mine system under complex and variable working conditions. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an intelligent mine mining equipment based on an industrial cloud platform, which solves the problems of being unable to comprehensively perceive the overall running state of the mine and being unable to effectively identify potential disaster precursor characteristics in the prior art.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides an intelligent mine mining equipment based on an industrial cloud platform, which comprises a data collection and cloud uploading module, a data governance and fusion module, a digital twin analysis module and an intelligent decision optimization module. The data collection cloud uploading module is configured to acquire mine production comprehensive data, and upload the data to the industrial cloud platform after preprocessing and formatting. The data governance fusion module is connected to the data collection cloud uploading module through a cloud data channel, and is configured to receive the processed mine production comprehensive data, and perform storage, cleaning, aggregation, and preliminary analysis. The digital twin analysis module communicates with the data governance fusion module through a data service bus, and is configured to receive mine comprehensive state data output by the data governance fusion module, construct a three-dimensional geological model, and perform abnormal behavior detection through a deep learning algorithm to output an abnormal detection result. The intelligent decision optimization module is associated with the digital twin analysis module through an algorithm coordination interface, and is configured to establish a disaster prediction model based on the abnormal detection result and the mine comprehensive state data, perform disaster risk assessment, and optimize a mining strategy using a reinforcement learning algorithm to obtain an optimal mining scheme, which is fed back to the data collection cloud uploading module through a feedback control link to guide on-site collection and scheduling.

[0007] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the digital twin analysis module includes a three-dimensional modeling submodule, a feature extraction submodule, and an abnormal detection submodule. The three-dimensional modeling submodule is configured to receive mine comprehensive state data, and construct a three-dimensional geological model in combination with real-time monitoring operation data and historical operation records. The feature extraction submodule is configured to perform extraction and standardization processing of multi-dimensional spatiotemporal features of the three-dimensional geological model. The abnormal detection submodule is configured to perform abnormal behavior recognition and analysis based on a deep learning algorithm.

[0008] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the intelligent decision optimization module includes a disaster prediction and evaluation submodule, a strategy generation submodule, and a reinforcement learning optimization submodule. The disaster prediction and evaluation submodule is configured to perform construction and risk assessment of a disaster prediction model. The strategy generation submodule is configured to perform generation of an initial mining strategy. The reinforcement learning optimization submodule is configured to perform reinforcement learning optimization of the initial mining strategy using a reinforcement learning algorithm.

[0009] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the processed mine production comprehensive data is subjected to denoising, outlier elimination and time synchronization processing according to the multi-source original data type and the sampling frequency, and then normalized, unit-converted and coded and mapped according to the preset data standard format and field structure.

[0010] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the mine comprehensive state data is obtained by performing data cleaning on the processed mine production comprehensive data through missing value filling, repeated data elimination and noise smoothing, grouping and aggregating the relevant processed mine production comprehensive data according to the spatial position, the equipment category and the operation link, extracting mine operation characteristic parameters, and performing preliminary analysis on the mine operation characteristic parameters.

[0011] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the three-dimensional geological model is constructed, and the abnormal behavior is detected through the deep learning algorithm, and the abnormal detection result is output, and the specific steps are as follows, The three-dimensional geological model is generated through the spatial interpolation and the voxel reconstruction algorithm based on the spatial coordinates, the rock layer attributes and the equipment operation position in the mine comprehensive state data, the operation surface deformation and the geological change information obtained through real-time monitoring, and the drilling surveying and mapping and the mining track data in the historical operation records; The multi-dimensional characteristic quantities in the spatial voxel are extracted from the three-dimensional geological model, and a multi-dimensional space-time feature matrix is constructed; The multi-dimensional space-time feature matrix is subjected to three-dimensional Gaussian filter smoothing, and a hidden layer representation vector is extracted by using a self-encoding neural network; The hidden layer representation vector is subjected to time series decomposition and spatial gradient mapping, the space-time change characteristics are extracted through multi-scale convolution, and the response intensity distribution of each spatial node in the space-time domain is calculated; Based on the normalized response intensity distribution and the localized anomaly index, the comprehensive judgment result is obtained through comprehensive judgment by using a fusion decision function; The comprehensive judgment result is subjected to classification processing, and isolated abnormal points are eliminated by applying a three-dimensional space-time connectivity constraint, and an abnormal mask and a corresponding confidence map are generated; The abnormal mask and the corresponding confidence map are mapped to the three-dimensional geological model according to the spatial coordinate index, and the abnormal detection result is output.

[0012] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the disaster prediction model is established, and the disaster risk is evaluated, and the specific steps are as follows, The abnormal region, abnormal intensity and space-time evolution characteristics in the abnormal detection result are taken as input characteristic sources, combined with geological structure parameters, operation environment parameters and equipment operation state data in the mine comprehensive state data, and the conditional dependency relationship between each geological environment and operation state characteristic and the disaster triggering causal chain are established in the Bayesian network framework through a multi-source data fusion algorithm to build a disaster prediction model; The abnormal detection result and the mine comprehensive state data are input into the disaster prediction model, and the conditional probability of occurrence of each disaster type is calculated; The conditional probability of occurrence of each disaster type is normalized, and according to the normalized conditional probability, the disaster risk of the mine area is divided into low risk, medium risk and high risk levels, and the disaster risk assessment result is output.

[0013] As a preferred scheme of the intelligent mine mining equipment based on the industrial cloud platform, the mining strategy is optimized by using the reinforcement learning algorithm to obtain an optimal mining scheme, and the specific steps are as follows, Mining operation related features are extracted from historical operation record data, and combined with a preset rule strategy and a three-dimensional geological model, an initial mining strategy is obtained; The disaster risk assessment result and the abnormal detection result are taken as inputs, combined with the mine comprehensive state data, and the initial mining strategy is simulated and optimized through the reinforcement learning algorithm, and through multiple iterations, an optimal mining scheme is obtained.

[0014] In a second aspect, the present application provides an intelligent mine mining method based on an industrial cloud platform, comprising: obtaining mine production comprehensive data and uploading it to an industrial cloud platform; storing, preprocessing, data aggregating and preliminarily analyzing mine geological data, equipment state data and production progress information through the industrial cloud platform to obtain mine comprehensive state data; Based on the mine comprehensive state data, combined with real-time operation monitoring data and historical operation record data, the industrial cloud platform constructs a three-dimensional geological model through a digital twin algorithm, and the cloud platform uses an abnormal behavior detection algorithm based on deep learning to detect the abnormal behavior of the ore body structure in the three-dimensional geological model to obtain an abnormal detection result; Based on the abnormal detection result and the mine comprehensive state data, the industrial cloud platform establishes a disaster prediction model through a multi-source data fusion algorithm to predict and analyze potential mine disasters and obtain a disaster risk assessment result; Based on the historical operation record data, the preset rule strategy and the three-dimensional geological model, an initial mining strategy is generated, and based on the disaster risk assessment result and the abnormal detection result, the industrial cloud platform simulates and optimizes the initial mining strategy by using a reinforcement learning algorithm to obtain an optimal mining strategy.

[0015] In a third aspect, the present application provides a computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements any step of the intelligent mine mining method based on the industrial cloud platform according to the second aspect of the present application.

[0016] The present application has the beneficial effects that: by constructing a multi-dimensional space-time feature matrix in the digital twin analysis module and combining deep learning for anomaly behavior detection, high-precision and low-false alarm identification of potential hidden dangers in the three-dimensional geological model is realized, and the early warning capability of the system is improved; at the same time, by introducing a reinforcement learning algorithm in the intelligent decision optimization module, the anomaly detection result and the disaster risk assessment result are fused to perform multi-round iteration optimization on the mining strategy, and adaptive decision-making considering safety, efficiency and resource utilization rate in a complex dynamic environment is realized. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Fig. 1 It is a schematic diagram of the intelligent mine mining equipment based on the industrial cloud platform.

[0019] Fig. 2 It is a schematic diagram of the digital twin analysis module.

[0020] Fig. 3 It is a schematic diagram of the intelligent decision optimization module.

[0021] Fig. 4 It is a disaster risk assessment flowchart. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0023] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the scope of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0024] Second, the "one embodiment" or "an embodiment" referred to herein can include a particular feature, structure, or characteristic. The various embodiments utilized in the description of the specification are not necessarily all mutually exclusive, but a single embodiment can be employed with a variety of alternatives.

[0025] Referring to Figs. 1-4 For one embodiment of the present application, the embodiment provides a smart mine mining equipment based on an industrial cloud platform, including the following steps: The data acquisition cloud module acquires mine production comprehensive data and performs preprocessing and formatting.

[0026] The mine production comprehensive data includes spatial coordinates, rock layer attributes, device operation positions, operation surface deformation, geological change information, drilling survey data, and mining track data. Further, according to the type and sampling frequency of the multi-source original data, the mine production comprehensive data is denoised, outlier removed, and time synchronized, and then normalized, unit converted, and encoded and mapped output processed according to the preset data standard format and field structure.

[0027] It should be noted that the preset data standard format is determined by analyzing business requirements and determining data exchange rules, formulating data interface protocols, setting database table structures to ensure efficient data storage and query, setting communication message formats to standardize data transmission methods, and setting uniform timestamp precision, value format, and unit identification.

[0028] The field structure is a data field arrangement according to the actual data items involved in the mine production process, such as spatial coordinates, rock layer attributes, device operation positions, operation surface deformation, geological change information, drilling survey data, and mining track data, according to their physical meaning (actual physical properties of data items, such as rock layer depth and device position) and business relevance (interrelationships between data items, such as the relevance of device operation position and operation surface deformation).

[0029] The data governance fusion module receives the processed mine production comprehensive data and performs storage, cleaning, aggregation, and preliminary analysis.

[0030] Specific operation: The processed mine production comprehensive data is cleaned by missing value filling, duplicate data elimination, and noise smoothing, and the relevant processed mine production comprehensive data is grouped and aggregated according to spatial position, device type, and operation link, and mine operation characteristic parameters are extracted, and mine operation characteristic parameters are preliminarily analyzed to obtain mine comprehensive state data.

[0031] Further, the processed mine production comprehensive data is sequentially subjected to missing value filling, duplicate data elimination and noise smoothing processing, the K-neighbor algorithm is used to fill the missing values, the duplicate records are eliminated by comparing the time stamps, equipment numbers and spatial coordinates, and the moving average filter is applied to smooth the vibration signals at the continuous 5 time points; on this basis, the related processed mine production comprehensive data is grouped and aggregated according to the spatial position, equipment type and operation link, the mean value and standard deviation of each group are calculated to form a structured feature set; the mine operation characteristic parameters are extracted from the structured feature set, the trend slope is calculated through a sliding window, and when the slope of the continuous 3 windows is greater than 0.5 and the current value exceeds the historical mean value by 2 times the standard deviation, it is determined as abnormal growth, and the mine comprehensive state data is generated in combination with the analysis result.

[0032] The digital twin analysis module receives the mine comprehensive state data output by the data governance fusion module, constructs a three-dimensional geological model, and performs abnormal behavior detection through a deep learning algorithm to output an abnormal detection result.

[0033] The digital twin analysis module is composed of a three-dimensional modeling submodule, a feature extraction submodule and an abnormal detection submodule.

[0034] The three-dimensional modeling submodule is used to receive the mine comprehensive state data, and combine real-time monitoring operation data and historical operation records to construct a three-dimensional geological model.

[0035] Specific operation: Based on the spatial coordinates, rock layer attributes and equipment operation positions in the mine comprehensive state data, the deformation and geological change information obtained by real-time monitoring are fused, and the drilling surveying and mapping and mining track data in the historical operation records are referred to, a three-dimensional geological model is generated through spatial interpolation and voxel reconstruction algorithm.

[0036] Further, based on the spatial coordinates, rock layer attributes and equipment operation positions in the mine comprehensive state data, combined with the deformation and geological change information obtained by real-time monitoring, the deformation data is mapped to the corresponding position according to the time stamp and spatial coordinates, the rock layer boundary points in the drilling surveying and mapping data and the boundary points in the mining track data are subjected to spatial interpolation, the continuous rock layer interface is generated by using the Kriging method, and then the space is divided into a regular voxel grid by using the voxel reconstruction algorithm, each voxel is assigned with the interpolated rock layer attributes, density and porosity, and a three-dimensional geological model is generated.

[0037] The feature extraction submodule is used to extract and standardize the multi-dimensional spatio-temporal features of the three-dimensional geological model.

[0038] Specific operation: The multi-dimensional feature quantities in the spatial voxel are extracted from the three-dimensional geological model to construct a multi-dimensional spatio-temporal feature matrix.

[0039] Further, the stratum attribute, density, porosity, stress state and temperature of each spatial voxel are extracted from the three-dimensional geological model as multi-dimensional feature quantities to form a feature vector of each voxel, and the feature vectors of the same spatial voxel at different time points are arranged in time sequence to construct a multi-dimensional space-time feature matrix containing the number of voxels, time steps and feature quantities, and the matrix dimension is (number of voxels x time steps x number of features).

[0040] The multi-dimensional space-time feature matrix is smoothed by three-dimensional Gaussian filtering, and the hidden layer representation vector is extracted by using the self-encoding neural network.

[0041] Further, the multi-dimensional space-time feature matrix is smoothed by three-dimensional Gaussian filtering, and the hidden layer representation vector is extracted by using the self-encoding neural network.

[0042] It should be noted that the training of the self-encoding neural network uses the smoothed multi-dimensional space-time feature matrix as input data, for example, the input layer dimension is set to 2048, the encoding layer dimension is set to 1024, the hidden layer dimension is set to 512, the decoding layer dimension is set to 1024, the output layer dimension is set to 2048, the mean square error is used as the loss function, the Adam optimizer is used for parameter update, for example, the learning rate is set to 0.001, the batch size is 64, the training rounds are 100, each round traverses all samples to complete forward propagation and back propagation, and the loss function converges until the loss function converges.

[0043] The abnormal detection sub-module is used for executing abnormal behavior recognition and analysis based on a deep learning algorithm.

[0044] Specific operation: The hidden layer representation vector is decomposed in time sequence and mapped in spatial gradient, the space-time change features are extracted by multi-scale convolution, and the response intensity distribution of each spatial node in the space-time domain is calculated.

[0045] Further, the hidden layer representation vectors are arranged in time sequence to form a time series, which is separated into a trend component and a detail component by wavelet decomposition, and a gradient vector of feature values between adjacent voxels is calculated based on the topological relationship of the voxel space and is spatially mapped; the mapping result is input into a three-dimensional convolutional neural network, and multi-scale convolution kernels with sizes of 3x3x3, 5x5x5 and 7x7x7 are used to extract spatio-temporal change features in local, regional and global ranges respectively; the multi-scale spatio-temporal change features are fused by weighted average fusion and normalized by a Softmax function to obtain the response intensity distribution of each spatial node in the time dimension and the spatial dimension.

[0046] It should be noted that the training of the three-dimensional convolutional neural network takes the multi-dimensional spatio-temporal feature data after spatial gradient mapping as input, and the network structure includes 3 three-dimensional convolutional layers (convolution kernel sizes are 3x3x3, 5x5x5 and 7x7x7 respectively, and the number of channels is 32, 64 and 128 respectively), 3 batch normalization layers, 3 ReLU activation function layers and 1 global average pooling layer, the output layer is a 128-dimensional feature vector, a cross-entropy loss function is used, an Adam optimizer is used for parameter update, the learning rate is set to 0.001, the batch size is 32, and the training rounds are 150 rounds. The loss value is calculated by forward propagation and the convolution kernel weight is updated by back propagation before each round, until the loss value converges stably.

[0047] Based on the normalized response intensity distribution and the localized anomaly index, a comprehensive judgment is made by fusion decision function to obtain a comprehensive judgment result.

[0048] Further, the response values of all spatial nodes are subjected to Z-score standardization to generate a standardized response intensity map, eliminating the differences in dimensions or measurement conditions between different nodes, and the standard deviation of the response values in a 3x3x3 neighborhood around each node is calculated as a localized anomaly index to quantify the fluctuation and potential anomaly of the node. The localized anomaly index and the standardized response intensity are combined by weighted combination to form a decision function, which comprehensively evaluates the node state to obtain a comprehensive score value. Based on the comprehensive score value and by setting a dynamic judgment threshold, the spatial nodes exceeding the dynamic judgment threshold are identified as abnormal nodes, and the spatial coordinates, time stamps and response intensities are recorded to obtain a comprehensive judgment result.

[0049] It should be noted that the dynamic judgment threshold is set according to the distribution of the comprehensive score value, which is used to distinguish between normal nodes and abnormal nodes. According to the statistical analysis results of historical data, such as the distribution characteristics of response intensity and the fluctuation of node state, the value range of the dynamic judgment threshold is set to be between 0.75 and 0.85, which is used to distinguish the boundary between severe abnormality and moderate abnormality.

[0050] The comprehensive determination result is classified and processed, and a three-dimensional space-time connectivity constraint is applied to remove isolated abnormal points to generate an anomaly mask and a corresponding confidence map.

[0051] Further, the comprehensive determination result is classified and processed according to the response intensity level, and the nodes with response intensity higher than the dynamic determination threshold are marked as serious anomalies, the nodes between the medium intensity determination threshold and the dynamic determination threshold are marked as moderate anomalies, and the nodes between the low intensity determination threshold and the medium intensity determination threshold are marked as mild anomalies; on the basis of classification, a three-dimensional space-time connectivity constraint is applied to construct a cubic neighborhood with each spatial voxel as the center, a spatial dimension of 3x3x3, and containing the current time step and each of the previous and subsequent time steps, to check whether there is at least one node marked as mild anomaly, moderate anomaly or serious anomaly in the cubic neighborhood, if there is, it is determined that the current node meets the connectivity condition, if not, it is determined as an isolated abnormal point and removed; based on the result after removing the isolated points, an anomaly mask is generated, each voxel in the anomaly mask is marked as an abnormal or normal state, and a corresponding confidence map is generated according to the weighted output of the standardized response intensity and the local anomaly index, the value of each voxel in the confidence map reflects the credibility of the abnormal judgment.

[0052] It should be noted that the medium intensity determination threshold is obtained by statistically analyzing the historical data distribution of the response intensity in the comprehensive determination result, and the 75th percentile value is taken as the set value, and the value range is set to 0.75 to 0.85 according to the mutation inflection point feature distinguishing moderate anomalies from serious anomalies; The low intensity determination threshold is obtained by analyzing the joint probability density of the standardized response intensity and the local anomaly index, and the value corresponding to the valley between the two peaks of the bimodal distribution is selected as the set value, and the value range is set to 0.40 to 0.50 according to the statistical separation point of the normal state and the mild abnormal state; The three-dimensional space-time connectivity constraint is obtained based on the physical characteristics of local continuity and diffusion of mine geological anomalies in the process of spatial expansion and time evolution.

[0053] The anomaly mask and the corresponding confidence map are mapped to the three-dimensional geological model according to the spatial coordinate index to output the anomaly detection result.

[0054] Further, each voxel in the anomaly mask is matched with the corresponding voxel in the three-dimensional geological model according to the spatial coordinates to realize accurate mapping of the abnormal state, and the confidence value of each voxel in the corresponding confidence map is associated with the corresponding position of the three-dimensional geological model according to the same spatial coordinate index, and the anomaly mark of the anomaly mask and the confidence value of the corresponding confidence map are superimposed for each voxel in the three-dimensional geological model, and the anomaly detection result containing the abnormal region, the abnormal intensity and the space-time evolution characteristics is output.

[0055] An intelligent decision optimization module is configured to establish a disaster prediction model based on the abnormality detection result and the mine comprehensive state data, perform disaster risk assessment, and optimize mining strategies using a reinforcement learning algorithm to obtain an optimal mining scheme, which is fed back to the data acquisition and cloud uploading module through a feedback control link to guide on-site collection and scheduling.

[0056] The intelligent decision optimization module is composed of a disaster prediction and assessment submodule, a strategy generation submodule, and a reinforcement learning optimization submodule.

[0057] The disaster prediction and assessment submodule is configured to perform construction and risk assessment of the disaster prediction model.

[0058] Specific operations: The abnormal region, abnormal intensity, and spatiotemporal evolution characteristics in the abnormality detection result are taken as input feature sources, combined with the geological structure parameters, operation environment parameters, and equipment operation state data in the mine comprehensive state data, and a condition-dependent relationship between each geological environment and operation state characteristic and a disaster trigger causal chain are established under a Bayesian network framework to construct a disaster prediction model through a multi-source data fusion algorithm.

[0059] Further, the abnormal region, abnormal intensity, and spatiotemporal evolution characteristics in the abnormality detection result are taken as input feature sources, the spatial range, response intensity peak, and time change slope of each abnormal voxel are extracted as key features, and the fault distance, rock layer inclination, gas concentration, and support pressure are selected as related variables, such as the fault distance: the distance between each operation region and the known fault is extracted from the mine geological structure data; the shortest distance from each operation region to the fault is calculated using geological survey data, drilling mapping, and spatial coordinate information, and the operation region with a shorter distance is selected to assess its potential earthquake and collapse risk; the rock layer inclination is selected: the inclination value of the rock layer is extracted according to the rock layer attribute data of the mine; the rock layer inclination of different regions is calculated using geological exploration data and drilling records, and those regions with a larger inclination are selected because a larger rock layer inclination may lead to landslide and collapse risk; the gas concentration is selected: the gas concentration information of each operation region is extracted from the real-time monitoring data of the mine; combined with the gas sensor data, the region with a higher gas concentration, especially in the region with a larger mining depth or poor ventilation, is selected to help assess the potential risk of gas explosion; the support pressure is selected: the support pressure data of each operation region is obtained through equipment monitoring and sensor data; regions with low or unstable support pressure are selected as risk points where support structure failure may occur, especially in deep mining or complex geological conditions. The above features are aligned and matched using multi-source data fusion algorithms (such as weighted average method, principal component analysis (PCA) and canonical correlation analysis (CCA)) to construct a feature dataset under a unified spatiotemporal benchmark. A set of nodes is defined under the Bayesian network framework, including anomalous feature nodes, geological environment feature nodes and disaster state nodes. The joint probability distribution between nodes is statistically analyzed based on historical data. The network parameters are learned by the maximum likelihood estimation method. The conditional dependencies between geological environment and operational status features and the causal chain of disaster triggering are established to form a disaster prediction model.

[0060] It should be noted that the multi-source data fusion algorithm is a data integration method based on feature alignment and attribute association. By unifying the spatiotemporal benchmark, it performs coordinate matching and time alignment between the abnormal areas, abnormal intensity and spatiotemporal evolution features in the anomaly detection results and the geological structural parameters, working environment parameters and equipment operation status data in the comprehensive mine status data. It adopts a weighted fusion strategy to integrate redundant information and uses principal component analysis or canonical correlation analysis to extract key components in the joint feature space, thereby realizing the structured fusion of multi-source heterogeneous data.

[0061] The anomaly detection results and comprehensive mine status data are input into the disaster prediction model, and the conditional probability of each disaster type is calculated.

[0062] Furthermore, the abnormal regions, intensity, and spatiotemporal evolution characteristics in the anomaly detection results are aligned with the geological structural parameters, operational environment parameters, and equipment operating status data in the comprehensive mine status data using a unified spatiotemporal index, and then input into the disaster prediction model. In the Bayesian network, anomaly features are used as evidence nodes, and disaster types such as roof falls, water inrushes, and gas outbursts are used as hypothesis nodes. Probabilistic inference is performed based on the learned conditional probability table and network structure. The Gibbs sampling algorithm is used to calculate the posterior probability of each disaster type under the current input conditions, and the conditional probability value of each disaster state is output.

[0063] The conditional probability values ​​for each disaster state are expressed as follows: ; In the formula, In observational evidence Conditions, No. Types of disasters The posterior probability of occurrence is a dimensionless real number; It is the current sampling step number, and its value range is... Unit [1]; It is the total number of Gibbs sampling iterations, which is a positive integer in units [1]; It is the first State variables for different types of disasters; It is the first disaster variable in subsampling state value of the disaster variable in subsampling is an indicator function, when outputs 1 when the state of the subsampling is equal to the target state, otherwise outputs 0, which is a dimensionless quantity.

[0064] It should be noted that the conditional probability table is obtained by statistically analyzing the historical data, calculating the conditional probability of each node given the state of the parent node, and learning using maximum likelihood estimation or Bayesian inference method.

[0065] The conditional probability of occurrence of each disaster type is standardized, and the disaster risk of the mine area is divided into low risk, medium risk and high risk levels according to the standardized conditional probability, and the disaster risk assessment result is output.

[0066] Further, the conditional probability of occurrence of each disaster type is Min-Max standardized, the original conditional probability value is mapped to the interval [0, 1], and the standardized conditional probability is formed; according to the standardized conditional probability, the low risk determination threshold and the medium risk determination threshold are set, the interval [0, low risk determination threshold) is divided into low risk level, the interval [low risk determination threshold, medium risk determination threshold) is divided into medium risk level, and the interval [medium risk determination threshold, 1] is divided into high risk level; the risk level of each disaster type corresponding to each region of the mine is divided, and the disaster risk assessment result containing spatial position, disaster type, risk level and standardized probability value is generated.

[0067] It should be noted that the low risk determination threshold is selected as the 30th percentile value of the conditional probability distribution before the occurrence of the historical disaster as the set value, and the value range is set to 0.25 to 0.35 according to the statistical division characteristics of normal working conditions (for example, the equipment runs in the standard parameter range, and the gas concentration is lower than the safety threshold) and early abnormal state (for example, the equipment has a slight fault, and the gas concentration has a slight fluctuation).

[0068] The medium risk determination threshold is selected as the 70th percentile of the probability value corresponding to the inflection point of the disaster starting to accelerate development in the historical data as the set value, and the value range is set to 0.65 to 0.75 according to the key critical characteristics of the transition from moderate abnormality to severe abnormality.

[0069] The strategy generation submodule is used to execute the generation of the initial mining strategy.

[0070] Specific operation: Mining operation related features are extracted from historical operation record data, and combined with preset rule strategies and a three-dimensional geological model, a preliminary mining strategy is obtained.​

[0071] Further, the drilling survey data, mining track data, equipment operation position and operation link duration and other mining operation related features are extracted from the historical operation record data to form a historical operation feature set; the feature set is subjected to rule matching and feasibility evaluation in combination with the provisions in the preset rule strategy about minimum working flat width, maximum slope angle, safe mining depth and rock stratum avoidance area; the operation mode meeting the rule strategy is spatially aligned with the rock stratum attributes, spatial coordinates and equipment operation position in the three-dimensional geological model to generate a preliminary mining strategy including recommended operation area, operation sequence and operation parameters.

[0072] It should be noted that the rule strategy is based on the statistical analysis results of the high-efficiency and low-risk operation modes in the historical operation records, extracts typical operation parameter ranges and spatial layout features, and induces a rule set for guiding mining operations.

[0073] The reinforcement learning optimization submodule is configured to use a reinforcement learning algorithm to perform reinforcement learning optimization on the initial mining strategy.

[0074] Specific operations: The disaster risk assessment results and the anomaly detection results are taken as inputs, combined with the mine comprehensive state data, and the initial mining strategy is simulated and optimized through a reinforcement learning algorithm, and the optimal mining scheme is obtained through multiple iterations of training.

[0075] Further, the risk level and the standardized conditional probability in the disaster risk assessment results and the abnormal area and the confidence map in the anomaly detection results are taken as state inputs, combined with the equipment operation state, operation environment parameters and geological structure parameters in the mine comprehensive state data to construct a state space for reinforcement learning; the operation area, operation sequence and operation parameters in the preliminary mining strategy are taken as an initial action set, and a reward function is defined to maximize mining efficiency, minimize risk exposure and reduce equipment wear and tear; a deep Q network algorithm is used for simulation training, experience is accumulated through interaction with the environment and network parameters are updated, and the optimal mining scheme is obtained after multiple iterations of optimization and convergence.

[0076] The embodiment also provides a smart mine mining method based on an industrial cloud platform, including: obtaining mine production comprehensive data and uploading it to an industrial cloud platform, storing, preprocessing, data aggregating and preliminarily analyzing mine geological data, equipment state data and production progress information through the industrial cloud platform to obtain mine comprehensive state data; Based on the mine comprehensive state data, in combination with real-time operation monitoring data and historical operation record data, the industrial cloud platform constructs a three-dimensional geological model through a digital twin algorithm, and the cloud platform uses an abnormal behavior detection algorithm based on deep learning to detect the abnormal behavior of the ore body structure in the three-dimensional geological model to obtain an anomaly detection result; Based on the abnormality detection result and the mine comprehensive state data, the industrial cloud platform establishes a disaster prediction model through a multi-source data fusion algorithm to predict and analyze potential mine disasters and obtain a disaster risk assessment result; Based on the historical operation record data, the preset rule strategy and the three-dimensional geological model, an initial mining strategy is generated, and based on the disaster risk assessment result and the abnormality detection result, the industrial cloud platform simulates and optimizes the initial mining strategy by using a reinforcement learning algorithm to obtain an optimal mining strategy.

[0077] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to implement the steps of the method for implementing a smart mine mining method based on an industrial cloud platform as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0078] To sum up, the present application realizes high-precision and low-false alarm identification of potential hidden dangers in the three-dimensional geological model by constructing a multi-dimensional space-time feature matrix in the digital twin analysis module and combining deep learning for abnormal behavior detection, and improves the early warning capability of the system; at the same time, the reinforcement learning algorithm is introduced in the intelligent decision optimization module, and the abnormality detection result and the disaster risk assessment result are fused to perform multi-round iteration optimization on the mining strategy, so as to realize adaptive decision-making considering safety, efficiency and resource utilization rate in a complex dynamic environment.

[0079] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application, and all should be covered in the scope of the claims of the present application.

Claims

1. A smart mining equipment based on an industrial cloud platform, characterized in that: This includes a data acquisition and cloud migration module, a data governance and integration module, a digital twin analysis module, and an intelligent decision optimization module; The data acquisition and cloud upload module is used to acquire comprehensive mine production data, and upload it to the industrial cloud platform after preprocessing and formatting. The data governance and fusion module is connected to the data acquisition and cloud-based module via a cloud data channel to receive, process, store, clean, aggregate, and perform preliminary analysis of the comprehensive mining production data. The digital twin analysis module and the data governance fusion module communicate with each other through a data service bus. The digital twin analysis module receives the comprehensive mine status data output by the data governance fusion module, constructs a three-dimensional geological model, and performs abnormal behavior detection through deep learning algorithms, outputting the abnormal detection results. The intelligent decision optimization module is associated with the digital twin analysis module through the algorithm collaboration interface. It is used to establish a disaster prediction model based on the anomaly detection results and the comprehensive mine status data, conduct disaster risk assessment, optimize the mining strategy using reinforcement learning algorithm, obtain the optimal mining plan, and transmit it back to the data acquisition cloud module through the feedback control link to guide on-site collection and scheduling.

2. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The digital twin analysis module includes a 3D modeling submodule, a feature extraction submodule, and an anomaly detection submodule; The three-dimensional modeling submodule is used to receive comprehensive mine status data and, in combination with real-time monitoring operation data and historical operation records, construct a three-dimensional geological model. The feature extraction submodule is used to extract and standardize multidimensional spatiotemporal features from the three-dimensional geological model. The anomaly detection submodule is used to perform anomaly behavior identification and analysis based on deep learning algorithms.

3. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The intelligent decision optimization module includes a disaster prediction and assessment submodule, a strategy generation submodule, and a reinforcement learning optimization submodule. The disaster prediction and assessment submodule is used to construct disaster prediction models and conduct risk assessments. The strategy generation submodule is used to generate the initial mining strategy; The reinforcement learning optimization submodule is used to perform reinforcement learning optimization on the initial mining strategy using a reinforcement learning algorithm.

4. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The processed comprehensive mining production data is based on the multi-source original data types and sampling frequency. The comprehensive mining production data undergoes noise reduction, outlier removal, and time synchronization. Then, it is normalized, unit converted, and encoded according to the preset data standard format and field structure.

5. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The comprehensive mine status data is obtained by cleaning the processed comprehensive mine production data through missing value imputation, duplicate data elimination, and noise smoothing. The relevant processed comprehensive mine production data are then grouped and aggregated according to spatial location, equipment type, and operation process to extract mine operation characteristic parameters, and then the mine operation characteristic parameters are preliminarily analyzed.

6. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The specific steps for constructing a three-dimensional geological model and detecting abnormal behavior using deep learning algorithms, and outputting the anomaly detection results, are as follows. Based on the spatial coordinates, rock strata properties, and equipment operation locations in the comprehensive mine status data, and integrating the deformation and geological change information of the working face obtained from real-time monitoring, and referring to the borehole mapping and mining trajectory data in historical operation records, a three-dimensional geological model is generated through spatial interpolation and voxel reconstruction algorithms. Multidimensional features are extracted from spatial voxels in a three-dimensional geological model to construct a multidimensional spatiotemporal feature matrix. The multidimensional spatiotemporal feature matrix is ​​smoothed by three-dimensional Gaussian filtering, and the hidden layer representation vector is extracted using an autoencoder neural network. The hidden layer representation vector is decomposed into a time series and mapped to a spatial gradient. Spatiotemporal variation features are extracted through multi-scale convolution, and the response intensity distribution of each spatial node in the spatiotemporal domain is calculated. Based on the normalized response intensity distribution and localization anomaly index, a comprehensive judgment is made by fusing a decision function to obtain the comprehensive judgment result; The comprehensive judgment results are classified and processed, and three-dimensional spatiotemporal connectivity constraints are applied to remove isolated outliers, generating anomaly masks and corresponding confidence maps; The anomaly mask and its corresponding confidence map are mapped to the three-dimensional geological model according to the spatial coordinate index, and the anomaly detection results are output.

7. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The specific steps for establishing a disaster prediction model and conducting a disaster risk assessment are as follows. Using the abnormal areas, abnormal intensity, and spatiotemporal evolution characteristics in the anomaly detection results as input feature sources, and combining the geological structural parameters, operational environment parameters, and equipment operation status data in the comprehensive mine status data, a disaster prediction model is constructed by establishing the conditional dependencies between various geological environments and operational status characteristics and the causal chain of disaster triggering under the Bayesian network framework through a multi-source data fusion algorithm. The anomaly detection results and comprehensive mine status data are input into the disaster prediction model, and the conditional probability of each disaster type is calculated. The conditional probability of each type of disaster is standardized, and based on the standardized conditional probability, the disaster risk of the mining area is divided into low risk, medium risk and high risk levels, and the disaster risk assessment results are output.

8. The smart mining equipment based on an industrial cloud platform as described in claim 1, characterized in that: The specific steps for optimizing the mining strategy using reinforcement learning algorithms to obtain the optimal mining solution are as follows. Extract mining operation-related features from historical operation records and combine them with preset rules and strategies and a three-dimensional geological model to obtain a preliminary mining strategy; Using disaster risk assessment results and anomaly detection results as input, combined with comprehensive mine status data, the initial mining strategy is simulated and optimized through reinforcement learning algorithms, and the optimal mining scheme is obtained through multiple iterations of training.

9. A smart mining method based on an industrial cloud platform, using the smart mining equipment based on an industrial cloud platform as described in any one of claims 1 to 8, characterized in that: include, Acquire comprehensive mine production data and upload it to the industrial cloud platform. The industrial cloud platform stores, preprocesses, aggregates, and performs preliminary analysis of mine geological data, equipment status data, and production progress information to obtain comprehensive mine status data. Based on comprehensive mine status data, combined with real-time operation monitoring data and historical operation record data, the industrial cloud platform constructs a three-dimensional geological model through a digital twin algorithm. The cloud platform uses a deep learning-based abnormal behavior detection algorithm to detect abnormal behavior of the ore body structure in the three-dimensional geological model and obtain the abnormal detection results. Based on anomaly detection results and comprehensive mine status data, the industrial cloud platform establishes a disaster prediction model through multi-source data fusion algorithms to predict and analyze potential mine disasters and obtain disaster risk assessment results. Based on historical operation record data, preset rules and strategies, and a three-dimensional geological model, an initial mining strategy is generated. Based on disaster risk assessment results and anomaly detection results, the industrial cloud platform uses reinforcement learning algorithms to simulate and optimize the initial mining strategy to obtain the optimal mining strategy.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the smart mining method based on an industrial cloud platform as described in claim 9.

Images