Intelligent coal mine system based on three-layer architecture

Abstract

This invention relates to the field of coal mine safety monitoring technology, specifically providing a three-layer intelligent coal mine system. The three-layer architecture includes a perception layer, a network layer, and an application layer. The perception layer collects data to obtain multi-dimensional mine data, which is then preprocessed to obtain multi-dimensional target data. The network layer transmits this multi-dimensional target data to the application layer, which includes at least a data access layer and a business logic layer. The business logic layer includes at least a safety monitoring module, an equipment monitoring module, and a data analysis and simulation module. The safety monitoring module includes an environmental parameter monitoring submodule and a personnel positioning and safety monitoring submodule. The equipment monitoring module is used for equipment fault monitoring and prediction. The data analysis and simulation module monitors the mine's safety status and triggers risk warnings, improving the efficiency and accuracy of safety monitoring in the intelligent coal mine system.

Description

Technical Field

[0001] This invention relates to the field of coal mine safety monitoring technology, specifically providing an intelligent coal mine system based on a three-layer architecture. Background Technology

[0002] With the development of information technologies such as the Internet of Things and artificial intelligence, comprehensive perception, real-time interconnection, and collaborative analysis and decision-making in smart coal mines are becoming increasingly possible, thereby improving coal mine production efficiency and ensuring mine safety. A three-layer architecture-based smart coal mine system comprehensively perceives and collects data in real time through the perception layer, processing the data efficiently and promptly, thus improving the efficiency and accuracy of safety monitoring in smart coal mine systems.

[0003] Traditional coal mine information systems typically exhibit a siloed architecture, with subsystems such as equipment management, production management, personnel positioning, and safety monitoring often built independently. This decentralized architecture can create information silos within the system, making it impossible to monitor the overall operation of the coal mine in real time and comprehensively. For example, when the coal mine system detects an abnormal situation, it is difficult to promptly link production control for emergency response, resulting in incomplete safety monitoring and potentially inadequate monitoring efficiency. Summary of the Invention

[0004] This application provides a three-layer architecture-based intelligent coal mine system to address the problems of low efficiency and low accuracy in traditional coal mine safety monitoring.

[0005] This application provides an intelligent coal mine system based on a three-layer architecture, the system comprising: A smart coal mine system based on a three-layer architecture is constructed, which includes a perception layer, a network layer, and an application layer. Data is collected through the sensing layer to obtain multidimensional data of the mine. The multidimensional data is preprocessed to obtain multidimensional target data. The multidimensional target data includes at least environmental parameters, equipment status data, and personnel location. The environmental parameters include at least gas concentration, oxygen, carbon monoxide, carbon dioxide, smoke, and temperature. The equipment status data includes at least vibration, temperature, and current data. The network layer is used to transmit the multidimensional target data to the application layer, and the application layer includes at least a data access layer and a business logic layer. The business logic layer includes at least a security monitoring module, an equipment monitoring module, and a data analysis and simulation module; The safety monitoring module includes an environmental parameter monitoring submodule and a personnel positioning and safety monitoring submodule. The equipment monitoring module is used for equipment fault monitoring and prediction; The data analysis and simulation module is used to monitor the safety status of the mine and trigger risk warnings.

[0006] In some embodiments, the environmental parameter monitoring submodule includes: Obtain the ventilation parameters of the mine, including fan speed and air volume data; The changes in gas concentration are analyzed in conjunction with the ventilation parameters to determine whether the changes in gas concentration are related to the state of the ventilation parameters. If so, the ventilation parameters are adjusted and environmental parameters are monitored again. If not, an early warning is triggered.

[0007] In some embodiments, the equipment monitoring module is used for equipment fault monitoring and prediction, including: The device status data is processed using a random forest model to output the gear wear probability of the device. When the wear probability is greater than a probability threshold, an early warning is triggered. The isolated forest algorithm is used to quickly detect anomalies in the device status data to identify early anomalies. If an anomaly is detected, the time period of the anomaly is obtained. The features of the abnormal period are input into the XGBoost classification model to determine the fault type and output the fault classification. If the fault is classified as sub-healthy, the RUL prediction model is activated to estimate the remaining lifespan and timely maintenance is carried out; if the fault is classified as a fault, an early warning is triggered directly.

[0008] In some embodiments, the data analysis and simulation module is used to monitor the safety status of the mine and trigger risk warnings, including: The data analysis and simulation module first constructs a digital twin model of the mine, analyzes and simulates the multidimensional target data based on the model, and triggers risk warnings. Based on the digital twin model of the mine, a particle algorithm is used to simulate the gas diffusion phenomenon in the mine. The gas molecule clusters are regarded as several particles, and a gas diffusion heat map is generated based on the distribution of each particle. Based on the gas diffusion heat map, all voxel grids of the heat map are traversed, and grids with gas concentrations greater than or equal to a concentration threshold are selected. The regions corresponding to the grids with gas concentrations greater than or equal to the concentration threshold are determined as high-concentration regions. The volume and time interval of the high-concentration region at two consecutive time points are obtained, and the expansion rate is obtained based on the volume and time interval of the high-concentration region; if the expansion rate of the high-concentration region within the time interval is greater than the rate threshold, a first-level warning is triggered. Traverse each grid in the high-concentration area and calculate the Euclidean distance from the person's location to the center point of each grid using Euclidean distance; if the distance is less than a distance threshold, trigger a level 2 warning; The mine digital twin model is input in real time with oxygen, carbon monoxide, carbon dioxide, smoke, and temperature. It uses an improved particle swarm optimization (IPSO)-RBF neural network algorithm to predict fire warnings. The IPSO optimizes the RBF parameters through a Gaussian function and outputs the optimal RBF parameters. The optimal RBF parameters are used to train the RBF neural network to output the fire risk level. If the fire risk level is greater than or equal to the level threshold, an early warning is triggered.

[0009] This application provides an intelligent coal mine system based on a three-layer architecture, which can solve the information silos in traditional coal mine production. Based on this three-layer architecture, through comprehensive real-time perception, real-time transmission, and intelligent early warning closed-loop management, safety control is transformed from post-event response to pre-event prevention, thus reducing accident risks. AI algorithms are used to perform deep analysis of the perceived data, enabling rapid identification of anomalies and real-time early warnings, and timely and rapid emergency response. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of a three-layer architecture-based intelligent coal mine system provided by the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0013] The following is combined with Figure 1 The following embodiments illustrate the technical solution of the present invention: This application provides an embodiment of an intelligent coal mine system based on a three-tier architecture, referring to... Figure 1 As shown, the intelligent coal mine system based on a three-tier architecture provided in this embodiment includes the following steps: S110: Construct an intelligent coal mine system based on a three-layer architecture. Collect data through the perception layer to obtain multi-dimensional data of the mine, preprocess the multi-dimensional data to obtain multi-dimensional target data.

[0014] In some embodiments, the implementation of the above subsystem S110 (constructing a three-layer intelligent coal mine system, collecting data through the perception layer to obtain multi-dimensional data of the mine, preprocessing the multi-dimensional data to obtain multi-dimensional target data) may include: It should be noted that the perception layer uses sensors to collect data in real time, connects the sensing devices to the system through a smart gateway, performs protocol adaptation, and ensures accurate data transmission.

[0015] Specifically, environmental parameters within the coal mine are collected. Gas sensors are deployed in key areas such as roadway intersections, coal faces, and return airways to collect gas concentrations; temperature sensors collect temperature data; equipment operating status is monitored; mine cameras capture video frames; wind speed sensors collect wind speed data; roof pressure sensors collect roof pressure data; and temperature sensors collect temperature data, enabling real-time monitoring of gas concentration changes at different locations. Vibration sensors and current transformers are embedded in major equipment such as coal mining machines, tunneling machines, pumps, and fans to collect voltage, current, temperature, vibration, and operating status data. UWB technology is used to locate mine personnel in real time and obtain their positions.

[0016] It should be noted that using a multi-sensor fusion method can avoid single-point failures, thereby improving data reliability.

[0017] It should be noted that real-time operating data is collected through sensors, while video and image data are captured through camera video streams, which can be used for video rendering and behavior recognition in digital twins (such as worker violations and coal-rock interface recognition).

[0018] It should be noted that mine pressure sensors are installed at the working face, and UWB base stations are deployed in the underground roadways; working cameras are used to collect data on the conveyor belt area.

[0019] Data is collected using various sensors to obtain multidimensional data about the mine; Environmental parameters: methane concentration, CO concentration, wind speed, temperature, humidity; Equipment status data: vibration, current, voltage, temperature, pressure; Space and personnel: roof displacement, tunnel deformation, personnel location, and equipment movement trajectory.

[0020] Kalman filtering is used to denoise the multidimensional data of the coal mine. The Protocol-Buffers protocol is used to unify the format of multi-source heterogeneous data. The denoised data is then converted into a standardized structure to obtain the multidimensional target data of the coal mine.

[0021] It should be noted that data preprocessing is to ensure the accuracy of subsequent data processing; it unifies the heterogeneous data format of the sensor-collected data to facilitate subsequent data transmission and analysis.

[0022] It should be noted that monitoring the dynamic environment underground is crucial to prevent accidents such as gas explosions; monitoring the vibration, current, voltage, temperature, and pressure of production equipment such as coal mining machines, excavators, tunneling machines, hydraulic supports, conveyor belts, water pumps, and fans is essential for fault monitoring and predictive maintenance.

[0023] S120: The network layer is used to transmit perception layer data to the application layer, which includes the data access layer, business logic layer, and presentation layer.

[0024] In some embodiments, the implementation of the above-mentioned subsystem S120 (the network layer is used to transmit perception layer data to the application layer, the application layer including a data access layer, a business logic layer, and a presentation layer) may include: It should be noted that the intelligent coal mine system is built based on a three-layer architecture: the perception layer, the network layer, and the application layer. The perception layer uses sensors to collect data, the network layer transmits the data collected by the perception layer to the application layer, and the application layer is used to process, analyze, mine, and perform model simulation calculations on the data to provide decision support for coal mine safety monitoring, determine risk warnings, and promptly issue early warning information.

[0025] It should be noted that the constructed intelligent coal mine system can coordinate the work between multiple departments such as production, equipment, safety, and environment. If an abnormal situation is detected, the safety monitoring module will immediately activate the emergency response and take corresponding emergency measures.

[0026] It should be noted that a converged communication network covering both above-ground and underground areas is constructed, using a hybrid transmission method of 5G + industrial Ethernet to ensure highly reliable and low-latency transmission; the collected data is then transmitted to the data processing center.

[0027] Specifically, mobile areas underground, such as the working face, utilize a mining-grade 5G network for wireless transmission to meet the needs of coal mining equipment control and video feedback. Fixed areas underground, such as roadways and chambers, employ GEPON technology with industrial Ethernet to transmit sensor data, voice, and video.

[0028] It should be noted that the Weighted Fair Queuing (WFQ) algorithm is used to prioritize transmitted data. In the coal mine system, safety data such as gas exceedance and roof pressure are set to the highest priority, while daily operation data of equipment are set to a low priority to ensure the reliability of critical data transmission.

[0029] The application layer architecture design in the intelligent coal mine system consists of a data access layer that provides data processing, a business logic layer that handles processing, and a presentation layer that displays the data. It should be noted that data access is a data interaction unit within the application layer, responsible for connecting to underlying databases such as InfluxDB for storing time-series data, enabling data reading and writing, and providing a unified data interface for the business logic layer; the business logic layer processes the data and monitors security alerts and device status failures; the presentation layer visualizes the processing results, security alerts, and device status of the business logic layer in the form of monitoring dashboards, mobile apps, web interfaces, etc.

[0030] It should be noted that the sensor data and equipment operation data in the mine are read from the InfluxDB time-series database and MySQL (relational database) by the data access layer and then passed to the application layer. Apache-Kafka is used to access massive amounts of real-time data, and InfluxDB real-time database is used to store the data.

[0031] Apache-Kafka is used to access massive amounts of real-time data. The network layer gateway sends all real-time data to the Kafka cluster. Kafka sends data to the application layer, and InfluxDB retrieves data from Kafka. InfluxDB is a time-series database that can store historical data such as gas concentration, equipment vibration data, and other multi-dimensional data, and can support efficient querying and analysis.

[0032] S130: The business logic layer includes at least a security monitoring module, an equipment monitoring module, and a data analysis and simulation module; the security monitoring module includes an environmental parameter monitoring submodule and a personnel positioning and security monitoring submodule.

[0033] In some embodiments, the implementation of the above-mentioned subsystem S130 (the business logic layer includes at least a security monitoring module, an equipment monitoring module, and a data analysis and simulation module; the security monitoring module includes an environmental parameter monitoring submodule and a personnel positioning and security monitoring submodule) may include: The business logic layer includes a safety monitoring module, an equipment fault module, and a data analysis and simulation module; the safety monitoring module includes an environmental parameter monitoring submodule and a personnel positioning and safety monitoring submodule.

[0034] It should be noted that the safety monitoring module ensures the safety of personnel and the production environment, while the equipment monitoring module ensures the normal and reliable operation of equipment. The data analysis and simulation module is used for data analysis, mining, and simulation to achieve optimized decision-making. Through the processing of these modules, intelligent management of coal mine production and safety is achieved, improving management efficiency and safety risk control.

[0035] It should be noted that the safety monitoring module processes safety data such as gas and roof pressure to provide risk warnings. It analyzes and predicts sensor data on mine gas concentration, roof pressure, and ventilation volume to determine the presence of safety risks. If a risk is identified, an alert is triggered.

[0036] It should be noted that environmental characteristics are extracted based on multi-source data; the rate of change of gas concentration (e.g., an increase of 0.3% within 5 minutes); the correlation between wind speed and gas concentration (a gas accumulation warning is triggered when wind speed < 0.2 m / s and gas concentration > 0.6%); wind speed affects gas warnings; and the roof displacement rate (> 2 mm / h indicates potential roof collapse).

[0037] It should be noted that the environmental parameter monitoring submodule can improve the accuracy of monitoring through correlation analysis between environmental parameters.

[0038] It should be noted that the DS evidence theory is used to fuse multi-source data for analysis, comprehensively assessing the safety status of the mine. This multi-source data fusion method avoids the inaccuracies of data from a single sensor. For example, relying solely on gas concentration data is easily affected by environmental factors (sensor malfunction, local airflow disturbances), leading to misjudgments. By fusing gas concentration data with ventilation parameters, it is possible to determine whether changes in gas concentration are related to the state of ventilation parameters. Through multi-source data fusion processing, the correlation between parameters is identified, thereby improving the accuracy of safety monitoring.

[0039] The changes in gas concentration are analyzed in conjunction with ventilation parameters to determine whether the changes in gas concentration are related to the status of ventilation parameters. If so, the ventilation parameters are adjusted and environmental parameters are monitored again. If not, an early warning is triggered.

[0040] It should be noted that if a gas concentration of 0.8% is used to directly determine an abnormality requiring a production shutdown, there may be a misjudgment based on a single data point. It is also necessary to combine ventilation parameters for judgment and use correlation analysis. If the problem is caused by insufficient ventilation, then adjusting the ventilation will solve the problem and there is no need to shut down production.

[0041] For example, first define all possible conclusions: Θ = {Gas abnormality is caused by insufficient ventilation (A), Gas abnormality is caused by other reasons (B), No gas abnormality (C)}; DS evidence theory uses multi-source data as sources of evidence: Evidence 1: Real-time monitoring of gas concentration data, with a gas threshold of 0.5%. If the gas concentration data is less than the gas threshold, it is considered normal; otherwise, it is considered abnormal.

[0042] Evidence 2: Ventilation parameter data includes fan speed and air volume data.

[0043] Transforming data into trust: Evidence 1: BPA allocation of gas concentration data, quantifying the confidence level of each proposition based on gas concentration values: Gas concentration is 0.8% (abnormal), but the cause is uncertain, therefore: m1(A) = 0.3 (partially due to insufficient ventilation); m1(B) = 0.4 (partially supports other reasons); m1(C) = 0.1 (low support, no anomalies); m1(Θ) = 0.2 (residual confidence level, representing uncertainty).

[0044] Evidence 2: BPA allocation of ventilation parameter wind speed and air volume data, quantifying the confidence level of each proposition based on whether the ventilation parameters meet the standards: The wind speed is 1.2 m / s (lower than the standard of 1.5 m / s) and the air volume is 800 m³ / min (lower than the standard of 1000 m³ / min), indicating obvious insufficient ventilation. Therefore: m2(A) = 0.5 (caused by insufficient high-support ventilation); m2(B) = 0.1 (low support due to other reasons); m2(C) = 0.2 (low support, no anomalies); m2(Θ) = 0.2 (uncertainty).

[0045] Evidence fusion calculation, using Dempster's combination rule to fuse the trust levels of m1 (gas) and m2 (ventilation): Step 1: Calculate the conflict coefficient K: K = Σ[m1(X)・m2(Y)], where X∩Y = ∅ (propositional conflict): The conflicting terms include: m1(A)・m2(B), m1(A)・m2(C), m1(B)・m2(A), m1(B)・m2(C), m1(C)・m2(A), m1(C)・m2(B); K=0.3×0.1+0.3×0.2+0.4×0.5+0.4×0.2+0.1×0.5+0.1×0.1=0.43.

[0046] Step 1: Calculate the normalization factor 1 / (1-K) = 1 / (1-0.43) ≈ 1.754; Step 3: Synthesize the overall trust levels m(A), m(B), and m(C): m(A)=[m1(A)·m2(A)+m1(A)·m2(Θ)+m1(Θ)·m2(A)]×1.754≈0.544; m(B)=[m1(B)·m2(B)+m1(B)·m2(Θ)+m1(Θ)·m2(B)×1.754≈0.246; m(C)=[m1(C)·m2(C)+m1(C)·m2(Θ)+m1(Θ)·m2(C)]×1.754≈0.140; The remaining trust level m(Θ)≈1-0.544-0.246-0.140≈0.07; Analyze and assess correlation and security status: Trustworthiness of each proposition after fusion: The highest value was m(A)≈0.544, therefore, the overall reliability of the gas anomaly being caused by insufficient ventilation was 54.4%. m(B)≈0.246: The reliability of gas anomalies caused by other reasons is 24.6%; m(C)≈0.140: The reliability of no gas anomaly is low.

[0047] In summary, since m(A) is the highest, it indicates that the change in gas concentration is highly correlated with the ventilation parameters. Therefore, the current gas anomaly is more likely caused by insufficient ventilation, i.e., the wind speed and air volume are not up to standard. The next step should prioritize adjusting ventilation parameters, such as increasing the fan speed, rather than ruling out other causes such as a sudden gas surge in the coal seam.

[0048] It should be noted that multi-dimensional data fusion is used for early warning. A safety early warning model is constructed by integrating LSTM neural network and random forest algorithm. Multi-dimensional data such as gas concentration, roof stress and goaf are input. The LSTM algorithm is used to capture the time series change trend, and the random forest algorithm is combined to improve the generalization ability of the early warning model, so that the early warning accuracy rate reaches more than 90%, and safety risks can be warned 30 minutes in advance.

[0049] It should be noted that the roadway network is modeled as a graph structure using a neural network (GNN), with nodes representing sensors and edges representing spatial correlations. This predicts whether gas accumulation in a certain area will spread to adjacent areas, thus improving the spatial correlation of early warning.

[0050] It should be noted that the real-time personnel positioning and safety early warning system synchronizes personnel positioning card data to the model, displaying the distribution of personnel underground in real time. If personnel enter the goaf or the high-risk working face restricted area, an audible and visual early warning will be triggered, and the warning information will be sent to relevant personnel simultaneously to prevent casualties.

[0051] Specifically, the personnel location and security monitoring submodule: The system can locate the position of personnel in the mine in real time, set up electronic fences in high-risk areas such as coal mining areas, goaf areas, and return airways. If personnel cross the boundary and enter the electronic fence, an audible and visual alarm will be triggered, enabling a rapid and timely emergency response. The system can also count the number of personnel in the mine, monitor their situation in real time, and quickly locate the position of trapped personnel in an emergency.

[0052] It should be noted that real-time early warnings are provided for gas exceeding limits, roof displacement exceeding standards, and personnel entering dangerous areas, and management personnel are notified when an early warning is triggered. An AI model is trained based on historical data to predict the risk of gas outbursts and the probability of equipment failure within the next 24 hours, allowing for the development of emergency response measures in advance. Integrating underground video and personnel positioning data, an emergency rescue route map is generated.

[0053] It should be noted that the BP neural network algorithm is used to train historical data to predict the gas emission rate; by adjusting the network weights and thresholds, the error between the output value and the actual gas emission rate is minimized.

[0054] S140: Equipment monitoring module is used for equipment fault monitoring and prediction.

[0055] In some embodiments, the implementation of the above subsystem S140 (equipment monitoring module for equipment fault monitoring and prediction) may include: It should be noted that analyzing the operating data of equipment such as coal mining machines, conveyors, and excavators, including current, temperature, and vibration frequency, is crucial for identifying any abnormalities. This helps pinpoint faulty components, such as worn bearings or motor overload, and allows for timely repairs. Real-time comprehensive monitoring of the equipment, along with fault diagnosis and early warning systems, minimizes unplanned downtime.

[0056] The equipment monitoring module includes equipment fault monitoring and equipment fault prediction, enabling early detection of equipment faults and timely maintenance.

[0057] It should be noted that the algorithm enables early warning analysis and fault prediction. The real-time monitoring center dynamically displays the mine's environmental parameters, equipment status, and personnel data, and uses a red flashing alarm when abnormal data is detected.

[0058] It should be noted that equipment can predict potential failures by means of vibration analysis and oil analysis, such as predicting the wear of coal mining machine bearings 15 days in advance.

[0059] It should be noted that equipment fault detection is crucial to ensuring the continuity of coal mine production. The equipment fault module needs to detect faults quickly, as downtime will directly lead to the interruption of coal mine production, so downtime should be minimized.

[0060] It should be noted that, based on the equipment status data, the characteristics of the equipment are extracted. A peak factor of vibration signal greater than 5 indicates bearing failure, harmonic distortion rate of current (>5% indicates motor abnormality), and associated risks such as early equipment failure and unknown environmental risks.

[0061] The random forest (RF) model is used to process equipment operating characteristics such as vibration, temperature, and current, and outputs the failure probability. For example, if the gear wear probability is >80%, 80% is the probability threshold, and an early warning is triggered. It should be noted that algorithms are used to uncover patterns in data, enabling proactive predictive maintenance and optimization of production processes. For critical equipment failure prediction, vibration, temperature, and current data from equipment such as main ventilation fans and elevators are analyzed to anticipate potential malfunctions, allowing for timely maintenance and minimizing unplanned downtime.

[0062] It should be noted that, based on equipment operating data such as vibration, temperature, and current data, the system automatically identifies equipment fault types, such as wear on coal mining machine gears and abnormal noise from motor bearings, predicts potential equipment failures, and performs maintenance in advance to minimize unplanned downtime. An isolated forest algorithm is used for anomaly detection based on equipment vibration, temperature, and current data to identify early anomalies and enable timely repairs. The XGBoost classification model is used to determine the equipment's status as healthy, sub-healthy, or faulty. Remaining lifespan is predicted based on equipment operating data to avoid unplanned downtime.

[0063] Specifically, the Isolation Forest algorithm is used to quickly detect anomalies in the vibration, temperature, and current data of the equipment, identifying early anomalies as healthy or sub-healthy. When the Isolation Forest detects an anomaly, it automatically inputs the features of that anomaly period into the XGBoost classification model to determine whether it is a sub-healthy condition or a minor fault. The XGBoost multi-classification model is trained using labeled / sub-healthy / faulty samples to classify the anomalies detected by the Isolation Forest and determine the severity of the anomalies. If classified as sub-healthy, the RUL prediction model is activated, estimating the remaining lifespan based on temporal degradation features such as the drift of vibration frequency over time; maintenance is performed after 30 days; if the classification result is a fault, an early warning is triggered directly.

[0064] It should be noted that predictive maintenance can detect faults early, plan downtime windows and spare parts in advance, and ultimately reduce unplanned downtime.

[0065] S150: The data analysis and simulation module is used for simulation and prediction, and to trigger risk warnings.

[0066] In some embodiments, the implementation of the above-mentioned subsystem S150 (the data analysis and simulation module is used for simulation prediction and triggering risk warning) may include: It should be noted that by using data analysis and algorithmic prediction, early warning responses can be issued before accidents occur, transforming passive anomaly responses into proactive prevention and minimizing the occurrence of accidents. Data analysis involves predictive and correlational analyses, which are then combined with simulation and anti-accident measures for intelligent decision-making.

[0067] It should be noted that digital twin modeling can be used to construct 3D models of mines, equipment, and personnel, and then linked to real-time data. Risk prediction and early warning, along with digital twins, constitute simulation and decision support. It should be noted that risks such as gas accumulation, roof deformation, and water seepage can be identified in advance by using models to simulate accident processes such as roof collapse; and emergency scenarios such as gas explosions and fires can be simulated to optimize emergency plans, such as the optimal escape route, which can improve the speed of emergency response.

[0068] It should be noted that a digital twin model of a coal mine can be constructed, which can map the physical space underground in real time and associate sensor data with the 3D model, such as superimposing the gas concentration of a certain area onto the corresponding roadway model.

[0069] It should be noted that LiDAR is used to scan underground mine roadways and working faces to acquire millisecond-level point cloud data (such as roadway outlines and equipment installation locations), generating 3D models. Static equipment data is used to create digital equipment profiles, serving as the geometric basis for the digital twin model. Real-time data is then mapped onto the digital twin, enabling visualized management and control across the entire mine. Ground personnel can view real-time data at any location within the digital twin, and optimal emergency rescue plans can be planned within the digital twin. Equipment managers can click on equipment within the digital twin to view its real-time operational status data.

[0070] Constructing a data twin model: A mobile 3D laser scanner is used to scan the coal mining face and roadways of the mine to generate point cloud data; the point cloud data is converted into a 3D network model using the professional point cloud processing software CloudCompare, and a digital twin model of the mine is constructed based on the 3D laser scan point cloud data.

[0071] It should be noted that, based on the point cloud data of the perception layer, a three-dimensional mesh model of the mine is constructed using Mesh modeling technology, including the geometric shape of entities such as roadways and equipment. The physical twin model can add attributes and rules to key equipment and the environment; the behavioral twin model can simulate the next action of the physical entity by predicting time series data.

[0072] By using 3D modeling BIM+GIS technology, a 3D visualization digital twin corresponding to the physical coal mine at a 1:1 scale is constructed. Real-time data on equipment status, environmental parameters, and personnel location data are mapped onto the twin entity, and the real-time status data of the coal mine can be viewed through the twin entity.

[0073] Specifically, a digital twin model of the mine is constructed based on three-dimensional excitation scanning point cloud data. A digital twin containing physical attributes and logical rules is created using Unity to achieve visual interaction. Real-time monitoring data such as gas concentration, equipment status, and personnel location are connected to drive the corresponding elements in the twin.

[0074] It should be noted that the particle algorithm is used to simulate the gas diffusion phenomenon, and the model is dynamically updated in combination with real-time monitoring data. Managers can view the risk heat map through AR glasses and monitor the safety status of the mine in real time.

[0075] Specifically, a particle algorithm is used to simulate gas diffusion, treating the gas as numerous particles. The gas emission point is determined based on real-time monitoring data, and its location and emission intensity serve as the emission source of the particle system. Physics is defined for each particle to ensure its motion conforms to the laws of gas diffusion. Based on these particles, a risk heat map is generated. The entire space is divided into a fine voxel grid, and the number of particles within each grid is counted in real time; more particles indicate a higher gas concentration in that area. The grid is rendered with different colors based on the concentration value, ultimately generating a dynamic risk heat map overlaid on the tunnel model.

[0076] For example, key safety indicators are calculated based on a gas diffusion heatmap: the heatmap is a voxel grid, where each grid is a cube with a side length of [missing information]. Each grid cell stores a gas concentration value; all voxel grids are traversed, with a concentration threshold of 1%, and grids with a concentration ≥ 1% are selected; the number of all grids meeting the criteria is counted. ; volume is The expansion rate is calculated based on the volume of the high-concentration area and the time interval between two consecutive time points. The volume change is ΔV = V2 - V1, and the expansion rate is v = ΔV / Δt. If the expansion rate of the high-concentration area is greater than 50% within 5 minutes, 50% is the rate threshold, and it is judged as a level 1 warning, i.e., high risk.

[0077] Traverse each grid in the high-concentration area and calculate the Euclidean distance from the person's location to the grid center point using Euclidean distance; if the shortest distance between the person and the high-concentration area is less than 50m, and 50m is the distance threshold, it is judged as a level 2 warning, i.e., medium risk.

[0078] The main passages of the mine, such as the main transport roadway and the main return airway, are marked with their spatial range using a three-dimensional model. If the gas concentration of any continuous 10m length voxel grid in the main passage is ≥0.8%, the simulation shows that the gas will spread to the main passage within 10 minutes, which is a level three warning.

[0079] It should be noted that simulating high-risk situations such as the diffusion path of gas leaks, the evolution of roof collapse, and the spread trend of fires allows for advance prediction of the risk range under different working conditions, optimizing the layout of safety monitoring points and emergency evacuation routes, such as simulating the fastest personnel evacuation routes. Simulating the diffusion path of gas leaks in a certain area under different ventilation conditions provides early warnings for potentially affected work areas and workers.

[0080] It should be noted that the simulation simulates various working conditions such as "gas exceeding the limit" and "equipment failure" in the digital twin scenario, and outputs prediction results such as the gas diffusion range after 10 minutes.

[0081] It should be noted that, based on historical and real-time equipment failure data, the digital twin model can predict the timing and location of equipment failures. For example, predicting that the scraper conveyor reducer may fail due to gear wear in 15 days allows for proactive maintenance, preventing downtime losses caused by sudden failures. The model also simulates the equipment's operating status under different loads and conditions, predicts the equipment's condition, and generates preventative maintenance plans.

[0082] It should be noted that the coal mine ventilation subsystem is the core of ensuring the safety of personnel underground. Traditional ventilation subsystems have fixed airflow distribution, which can easily lead to insufficient airflow in some areas, causing excessive methane levels, or excessive airflow, resulting in energy waste. Digital twins are used to achieve dynamic optimization of the ventilation subsystem. The model recreates the underground roadway network, the operating status of ventilation fans, and the methane emission rate, and combines this with real-time monitoring data (methane concentration and wind speed in each area) to dynamically calculate the airflow distribution and monitor ventilation dead zones or areas with redundant airflow. Dynamic airflow control allows the model to quickly simulate the effects of adjusting the ventilation fan frequency and switching dampers when the methane emission rate in a certain area suddenly increases, and to implement the optimal control strategy (such as increasing the airflow in that area while avoiding insufficient airflow in other areas). This ensures that the methane concentration remains stable within the safe threshold while reducing ventilation energy consumption.

[0083] It should be noted that fire risk early warning analyzes oxygen, carbon monoxide, carbon dioxide, smoke, and temperature to predict fire risks. The IPSO algorithm can improve prediction accuracy by optimizing the particle search strategy (such as introducing adaptive adjustment of inertial weights and mutation mechanisms).

[0084] Specifically, an improved particle swarm optimization (PSO)-RBF neural network algorithm is used to predict fire early warning. Five nodes—oxygen, carbon monoxide, carbon dioxide, smoke, and temperature—are used as the input layer. The number of nodes in the hidden layer is determined by optimizing the RBF parameters using a Gaussian function (IPSO). Calculate the output. For the hidden layer center, For base width, Given a vector containing 5 nodes: x = [{x}_{1}, {x}_{2}, {x}_{3}, {x}_{4}, {x}_{5}] ,in This represents the oxygen concentration value. This represents the carbon monoxide concentration value. This represents the carbon dioxide concentration value. This represents the smoke concentration value. This vector represents the temperature value. The raw feature data is passed from the input layer to the hidden layer, and the Gaussian function is calculated... With the hidden center Euclidean distance (also a 5-dimensional vector) This is used to measure the similarity between the input sample and the center of the hidden layer neuron, and finally output the activation value of the hidden layer node.

[0085] It should be noted that the IPSO algorithm optimizes the hidden layer center. and base width This enables the RBF neural network to more accurately capture the nonlinear relationship between fire characteristics and early warning results, thereby improving the accuracy of fire early warning.

[0086] IPSO optimization of RBF parameters: Initialization of particle swarm (each particle represents a set of RBF parameters: , , Set inertia weights (e.g., linearly decreasing from 0.9 to 0.4 with each iteration) and learning factors ( = =2); Iterative optimization: Calculate the fitness (predicted MSE) of each particle, update the individual extreme value and the global extreme value until the iteration terminates (e.g., 100 iterations), and output the optimal RBF parameters.

[0087] The RBF neural network is trained with optimal parameters, and sensor data is input into the digital twin virtual model in real time to output the risk level. The output layer has one node: fire risk level, such as 0=normal, 1=low risk, 2=medium risk, 3=high risk. If the output risk level is ≥2 (medium risk), and 2 is the level threshold, an early warning is triggered.

[0088] If a fire risk is predicted for a certain area, the fire spread path (such as the spread of high-temperature smoke) in that area is simulated in the virtual model.

[0089] It should be noted that, due to the simple structure of the IPSO-optimized RBF network, its inference speed is fast, making it suitable for real-time early warning situations in digital twins.

[0090] It should be noted that if simulated personnel accidentally enter a dangerous area (such as the no-entry zone of the tunnel head), the twin model will issue a warning 5 seconds in advance and simulate and plan the optimal evacuation route.

[0091] An LSTM model is used to train historical location data (inputting the previous 10 location sequences, outputting the next 3 predicted locations) and learns personnel movement habits (e.g., speed decreasing to 0.5 m / s when turning). If the prediction indicates that the personnel's location will enter the warning buffer within the next 5 seconds, an alert is triggered. The Dijkstra algorithm is used to plan and calculate the optimal evacuation route.

[0092] It should be noted that by simulating accidents such as water inrush and fire in the virtual twin scenario, trainees practice emergency response procedures in a virtual environment, and apply the optimized procedures to emergency plans to improve the speed of emergency response and ensure coal mine safety.

[0093] If the safety warning module triggers a risk (such as excessive gas levels or abnormal roof pressure), the emergency dispatch will activate the preset emergency plan (such as cutting off the power supply to the area exceeding the limit and turning on the backup ventilation fan); and push evacuation instructions to the personnel underground, and link with the dispatch center to generate an emergency resource allocation list (such as arranging rescue personnel and allocating emergency equipment); and record the entire emergency response process (for post-event optimization).

[0094] The presentation layer is responsible for data visualization, presenting the monitoring results in a visual format.

[0095] The intelligent coal mine system displays the mine's operational status through visual interfaces such as PCs and mobile apps. Coal mine safety is paramount, and safety management is the highest priority in decision-making, ensuring timely and immediate emergency response. The presentation layer uses visual displays.

[0096] It should be noted that the mobile app provides managers with access to their mobile phones, allowing them to view real-time alerts, mine status, and emergency notifications.

[0097] When the risk is low (Level 3 warning): a system pop-up notification will appear; When the risk level is medium (Level 2 warning): The system will display a pop-up notification and send a text message to relevant staff. In case of high risk (Level 1 warning): immediately cut off the power supply to the danger zone, activate the emergency plan, and send a text message to the person in charge of the mine; the system displays the location and impact range of the danger zone.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart coal mine system based on a three-layer architecture, characterized in that, include: A smart coal mine system based on a three-layer architecture is constructed, which includes a perception layer, a network layer, and an application layer. Data is collected through the sensing layer to obtain multidimensional data of the mine. The multidimensional data is then preprocessed to obtain multidimensional target data. The multidimensional target data includes at least environmental parameters, equipment status data, and personnel location; the environmental parameters include at least gas concentration, oxygen, carbon monoxide, carbon dioxide, smoke, and temperature; the equipment status data includes at least vibration, temperature, and current data. The network layer is used to transmit the multidimensional target data to the application layer, and the application layer includes at least a data access layer and a business logic layer. The business logic layer includes at least a security monitoring module, an equipment monitoring module, and a data analysis and simulation module; The safety monitoring module includes an environmental parameter monitoring submodule and a personnel positioning and safety monitoring submodule. The equipment monitoring module is used for equipment fault monitoring and prediction; The data analysis and simulation module is used to monitor the safety status of the mine and trigger risk warnings.

2. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The preprocessing of the multidimensional data to obtain multidimensional target data includes: The multidimensional data is denoised using Kalman filtering, and then the denoised multidimensional data is standardized to obtain the multidimensional target data.

3. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The data access layer includes: The gateway in the network layer sends the multidimensional target data to the Kafka cluster, Kafka sends the multidimensional target data to the application layer, and the InfluxDB database reads data from the Kafka cluster. The InfluxDB database is used to store the multidimensional target data.

4. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The environmental parameter monitoring submodule includes: Obtain the ventilation parameters of the mine, including fan speed and air volume data; The changes in gas concentration and ventilation parameters are analyzed together to determine whether the changes in gas concentration are related to the state of the ventilation parameters. If so, the ventilation parameters are adjusted and environmental parameters are continuously monitored. If not, an early warning is triggered.

5. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The personnel location and safety monitoring submodule includes: Electronic fences are set up in each high-risk area of ​​the mine, and the high-risk areas include at least the coal mining area, the goaf, and the return airway; If the personnel's location enters the electronic fence, an audible and visual warning will be triggered, enabling a rapid and timely emergency response.

6. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The equipment monitoring module is used for equipment fault monitoring and prediction, including: The device status data is processed using a random forest model to output the gear wear probability of the device. If the wear probability is greater than the probability threshold, an early warning is triggered.

7. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The equipment monitoring module is used for equipment fault monitoring and prediction, and also includes: The isolated forest algorithm is used to quickly detect anomalies in the device status data to identify early anomalies. If an anomaly is detected, the time period of the anomaly is obtained. The features of the abnormal period are input into the XGBoost classification model to determine the fault type and output the fault classification. If the fault is classified as sub-healthy, the RUL prediction model is activated to estimate the remaining lifespan and timely maintenance is carried out; if the fault is classified as a fault, an early warning is triggered directly.

8. The intelligent coal mine system based on a three-layer architecture according to claim 1, characterized in that, The data analysis and simulation module is used to monitor the safety status of the mine and trigger risk warnings, including: The data analysis and simulation module first constructs a digital twin model of the mine, analyzes and simulates the multidimensional target data based on the model, and triggers risk warnings. A 3D laser scanner is used to scan the coal mining face and roadways of the mine to generate point cloud data corresponding to the coal mining face and roadways; then, professional point cloud processing software is used to convert the point cloud data into a 3D network model to construct a digital twin model of the mine.

9. The intelligent coal mine system based on a three-layer architecture according to claim 8, characterized in that, The process of analyzing and simulating the multidimensional target data based on the model, and triggering risk warnings, includes: Based on the digital twin model of the mine, a particle algorithm is used to simulate the gas diffusion phenomenon in the mine. The gas molecule clusters are regarded as several particles, and a gas diffusion heat map is generated based on the distribution of each particle. Based on the gas diffusion heat map, all voxel grids of the heat map are traversed, and grids with gas concentrations greater than or equal to a concentration threshold are selected. The regions corresponding to the grids with gas concentrations greater than or equal to the concentration threshold are determined as high-concentration regions. The volume and time interval of the high-concentration region at two consecutive time points are obtained, and the expansion rate is obtained based on the volume and time interval of the high-concentration region; if the expansion rate of the high-concentration region within the time interval is greater than the rate threshold, a first-level warning is triggered. Traverse each grid in the high-concentration area and calculate the Euclidean distance from the person's location to the center point of each grid using Euclidean distance; if the distance is less than the distance threshold, trigger a level 2 warning.

10. The intelligent coal mine system based on a three-layer architecture according to claim 8, characterized in that, The process of analyzing and simulating the multidimensional target data based on the model and triggering risk warnings also includes: The mine digital twin model is input in real time with oxygen, carbon monoxide, carbon dioxide, smoke, and temperature. It uses an improved particle swarm optimization (IPSO)-RBF neural network algorithm to predict fire warnings. The IPSO optimizes the RBF parameters through a Gaussian function and outputs the optimal RBF parameters. The optimal RBF parameters are used to train the RBF neural network to output the fire risk level. If the fire risk level is greater than or equal to the level threshold, an early warning is triggered.

Images