A coal bunker intelligent exploration method and system based on multi-source data fusion

Abstract

The application discloses a coal bunker intelligent exploration method and system based on multi-source data fusion, which collects point cloud data in the bunker through a 3D laser scanning radar, completes data cleaning in combination with a dust-resistant filtering algorithm, constructs a real-time digital twin model of the coal bunker through a Delaunay triangulation algorithm, performs weighted fusion calculation on the average coal level output by the real-time digital twin model of the coal bunker and the coal level collected by a sensing optical fiber, so that continuous coal level data meeting the accuracy requirement are output, and intelligent identification of arching caused by blocking of the bunker is realized based on the real-time digital twin model of the coal bunker; LoRa wireless temperature measurement data in combination with a dual-spectrum thermal imaging camera realizes self-ignition grading early warning, and intelligent processing and linkage control are completed through an AI edge computing unit in the whole process. The application has high-precision measurement and whole-scene hidden danger early warning capabilities, has strong dust interference resistance, and can be widely applied to intelligent safety management and control scenes of various coal bunkers.

Description

Technical Field

[0001] This invention relates to the field of coal bunker safety management and control technology, and in particular to a method and system for intelligent coal bunker exploration based on multi-source data fusion. Background Technology

[0002] Coal is the primary energy source in my country's energy system. As the core storage and transportation facility in the entire coal production, transportation, and utilization chain, the coal bunker is a crucial storage facility in industrial scenarios such as coal storage and transportation. Its safe and stable operation directly affects the efficiency and safety of the entire coal production and transportation system. Accurate material level monitoring, early warning of blockage faults, and prevention of spontaneous combustion of coal directly impact the stable operation of the production system, on-site operational safety, and enterprise operational efficiency. With the continuous advancement of intelligent mines and smart power plants, the industry has placed higher demands on the measurement accuracy, environmental adaptability, and intelligent control capabilities of coal bunker probing technology.

[0003] Current mainstream coal bunker detection technologies suffer from several core shortcomings. In terms of material level monitoring, contact-based equipment is prone to wear and tear and has a high failure rate; non-contact single-point measurements are easily affected by high-dust environments; existing laser scanning solutions lack specific optimization for dusty conditions, have no redundant verification mechanisms, and suffer from insufficient data reliability. Blockage identification largely relies on manual inspections, feeder current monitoring, or single-vision recognition, resulting in significant delays, high false alarm and false negative rates, and poor adaptability to harsh environments. Coal temperature monitoring solutions generally suffer from complex wiring, blind spots in temperature measurement, susceptibility to dust interference, and a lack of multi-source cross-verification, leading to insufficient early warning accuracy. Overall, existing technologies generally suffer from single data sources, information silos formed by multiple systems, and poor adaptability to harsh conditions, failing to meet the core requirements of intelligent coal bunker management. Chinese invention patent application CN120327976A discloses a coal bunker intelligent inspection system based on three-dimensional laser. It uses three-dimensional laser to explore the inside of the coal bunker. However, three-dimensional laser is easily affected by factors such as dust, so it needs to be equipped with relevant algorithms to identify dust in order to make the point cloud data collected by the three-dimensional laser more accurate. Summary of the Invention

[0004] To address the aforementioned technical shortcomings, the purpose of this invention is to provide a method and system for intelligent coal bunker exploration based on multi-source data fusion. This system achieves full-dimensional data acquisition through the collaboration of multiple types of equipment, constructs a high-precision real-time digital twin model of the coal bunker through anti-dust point cloud processing, and realizes high-precision coal level monitoring, intelligent identification of blockage, and graded early warning of coal temperature through deep fusion of multi-source data. It has strong environmental adaptability and operational reliability and can fully meet the application requirements of intelligent safety management and control of coal bunkers.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: This invention provides an intelligent coal bunker exploration method based on multi-source data fusion, comprising the following steps: S1. Multi-source data acquisition: Multi-source data acquisition is carried out through various sensors deployed in the coal bunker. The specific acquisition methods are as follows: 3D spatial morphology data acquisition: 3D laser scanning radar is deployed at the preset installation points on the top of the coal bunker. Relying on the high-speed 3D ranging and omnidirectional scanning capabilities of the 3D laser scanning radar, the 3D point cloud data of the fully enclosed space inside the coal bunker is fully covered and collected. The core information such as the spatial outline, surface morphology, and 3D coordinates of the coal pile inside the bunker is accurately captured, providing the core spatial data source for subsequent 3D modeling and structural identification of the coal pile. Based on passive fiber optic distributed sensing technology, multiple mining-specific strain-temperature dual-parameter sensing optical fibers are vertically deployed along the inner wall of the coal bunker, and equipped with an intrinsically safe fiber optic demodulator outside the bunker, so that the real-time coal level height can be directly output, providing a stable data source for subsequent coal level data fusion verification and redundant backup. Distributed coal temperature acquisition: Multiple sets of LoRa wireless temperature sensors are distributed and deployed at preset installation points in the vertical and circumferential directions of the coal bunker wall, thereby realizing direct contact measurement between the sensors and the coal inside the bunker. Two dual-spectrum thermal imaging cameras are symmetrically installed at both ends of the top of the coal bunker to achieve non-contact measurement of the surface coal temperature inside the bunker. By continuously collecting real-time temperature data of the coal inside the bunker at different heights and circumferential positions, distributed monitoring of coal temperature across the entire height and cross-section of the bunker is achieved, providing a core temperature data source for early warning of coal spontaneous combustion risks. S2. Digital Twin Modeling: Through the AI ​​edge computing unit, an anti-dust filtering algorithm is executed on the collected raw 3D point cloud data to remove virtual points and noise, so as to obtain effective coal pile point cloud data. Based on the effective coal pile point cloud data, a real-time digital twin model of the coal bunker in the whole scene is constructed and synchronized with the coal bunker site in real time. S3. Multi-source fusion computing and linkage control: The AI ​​edge computing unit calculates coal level parameters based on the real-time digital twin model of the coal bunker, and performs weighted fusion of the coal level parameters with the collected coal level height data to output continuous coal level data that meets the accuracy requirements; S4. Based on the real-time digital twin model of the coal bunker, identify the structural anomalies of the coal pile, and combine the changing trends of the coal feeder's operating status and continuous coal level data to make a comprehensive judgment on the blockage; S5. Receive temperature data collected by the wireless temperature sensor and the dual-spectrum thermal imaging camera, and make a graded early warning judgment on the risk of spontaneous combustion based on the preset temperature grading threshold; S6. The AI ​​edge computing unit outputs corresponding alarm signals and linkage control commands based on the continuous coal level data that meets the accuracy requirements, the comprehensive judgment result of the blockage, and the graded early warning judgment result of the spontaneous combustion risk.

[0006] Preferably, in step S2, the anti-dust filtering algorithm specifically includes: based on the reflection intensity and spatial coordinate characteristics of the point cloud data, removing discrete noise points that exceed a preset threshold range; using a statistical filtering algorithm, calculating the average distance between each point cloud and its neighboring points, and removing dust virtual points whose average distance exceeds a preset standard deviation range; and sampling the filtered point cloud data to reduce the amount of point cloud data while retaining the coal pile outline characteristics, thereby obtaining effective coal pile point cloud data.

[0007] This step primarily addresses the issues of high noise, numerous virtual points, and high data redundancy in the raw point cloud data caused by the complex working conditions of high dust and high water mist in coal bunkers. It involves multi-stage cleaning and optimization of the point cloud data, and based on the high-quality point cloud data, constructing a full-scene digital twin of the coal bunker that is synchronized with the on-site working conditions in real time. The specific execution process is as follows: First, by using AI edge computing units deployed in the industrial site, a pre-set anti-dust filtering algorithm is run to perform multi-stage cleaning processing on the raw 3D point cloud data acquired by the 3D laser scanning radar: The first step is to remove abnormal and invalid points. Based on the laser reflection intensity characteristics of point cloud data and the spatial coordinate threshold of the coal bunker, discrete noise points and invalid points that exceed the reasonable reflection intensity range or whose spatial distance exceeds the physical boundary of the coal bunker are removed. The second step is to accurately filter dust virtual points. A statistical filtering algorithm is used to calculate the average spatial distance between each point cloud data and its neighboring points within a preset neighborhood range. Virtual point data formed by the reflection of suspended dust particles in the warehouse with an average distance exceeding the preset standard deviation range are removed. The third step is to perform lightweight optimization of the point cloud data. After noise and virtual point removal, the point cloud data is subjected to voxel downsampling. While fully preserving the coal pile outline features and surface morphology features, the volume of the point cloud data is greatly reduced, improving the processing efficiency of subsequent modeling and calculation, and finally obtaining high-quality, high-fidelity effective coal pile point cloud data.

[0008] Subsequently, based on the above effective coal pile point cloud data, the Delaunay triangulation algorithm was used to fit the discrete coal pile point cloud data to a surface, constructing a high-precision three-dimensional surface model of the coal pile, accurately restoring the real spatial shape, surface slope, stacking structure, concave and convex features and other details of the coal pile. By combining the pre-entered structural dimensions of the coal bunker, including its diameter, total height, cone angle, and the installation location of fixed equipment, the three-dimensional surface model of the coal pile is spatially registered and coordinately fused with the benchmark model of the coal bunker. This results in the construction of a real-time digital twin model of the coal bunker that is synchronized with the actual working conditions on site, enabling precise digital and visual mapping of the coal pile status, bunker space, and operating parameters.

[0009] Preferably, in step S3, the formula used for weighted fusion calculation of the average coal level output by the real-time digital twin model of the coal bunker and the coal level height data collected by the sensing fiber is: H=a×H1+b×H2, where H is the final output continuous coal level data that meets the accuracy requirements, H1 is the average coal level output by the real-time digital twin model of the coal bunker, H2 is the coal level data collected by the sensing fiber, a and b are preset weight coefficients, and a+b=1.

[0010] Based on a real-time digital twin model of the coal bunker, the highest point coal level, the average coal level across the entire cross-section, and the real-time coal volume within the bunker are calculated using a spatial volume integral algorithm. Based on deployed strain-temperature dual-parameter sensing optical fibers, and equipped with an intrinsically safe optical fiber demodulator outside the bunker, the real-time coal level height can be directly output. The two are weighted to obtain high-precision continuous coal level data. Combined with pre-entered coal bulk density parameters, the real-time coal weight within the bunker can also be calculated simultaneously, providing core data support for coal bunker production scheduling, material control, and inventory management.

[0011] Preferably, in step S4, the comprehensive determination of blockage is as follows: when the real-time digital twin model of the coal bunker identifies a bridging structure on the surface of the coal pile exceeding a preset size or a cavity inside the coal pile exceeding a preset volume, it is determined to be a structural identification anomaly; when the coal feeder is in operation and the calculated coal level change rate is lower than a preset threshold, it is determined to be a coal flow stagnation; if either the structural identification anomaly or the coal flow stagnation condition is met, it is determined to be a blockage.

[0012] The process is as follows: First, based on the real-time updated digital twin model of the coal bunker, the three-dimensional spatial structural features of the coal pile are extracted. The abnormal structures such as contour changes on the surface of the coal pile and internal spatial cavities are intelligently identified and their features are extracted to accurately capture structural abnormal features of the coal pile that are prone to blockage, such as bridging, coal bridging, and cavities.

[0013] At the same time, the system acquires real-time operating status data of the coal feeder at the bottom of the coal bunker, including the feeder's start / stop status, operating frequency, load current, and other core operating parameters. Combined with high-precision fusion coal level data, it calculates the rate and trend of coal level change in real time, accurately judging the falling and circulation status of coal flow in the bunker.

[0014] Subsequently, a multi-dimensional comprehensive judgment of the blockage fault is made: when the coal pile surface is identified by the real-time digital twin model of the coal bunker as having a span greater than or equal to a preset threshold and a height greater than or equal to a preset threshold for bridging structure, or when a cavity or void with a volume greater than or equal to a preset threshold is identified inside the coal pile, it is judged as an abnormal coal pile structure identification. When the coal feeder is in normal operation, but the calculated coal level change rate is less than or equal to the preset threshold, that is, when the coal level does not drop significantly for a long time and the coal flow in the bin is stagnant, it is judged as an abnormal coal flow stagnation. When either the abnormal coal pile structure identification or the abnormal coal flow stagnation condition is met, the system will comprehensively determine that a blockage fault has occurred.

[0015] Finally, upon determining a blockage fault, the system immediately outputs a blockage alarm signal, which is uploaded to the central control room via the industrial network. Simultaneously, it triggers the audible and visual alarms in the central control room and pushes real-time warning information to the mobile phones of management personnel using the industrial cloud monitoring APP. At the same time, it automatically outputs the control signal to start the arch-breaking device, completing the automatic handling of the blockage fault. It also records the entire process data, including the location of the blockage, the degree of blockage, the handling process, and the handling result, providing data support for subsequent coal bunker operation optimization and fault tracing.

[0016] Preferably, the specific logic of the material level alarm and linkage control is as follows: preset high-level overflow warning threshold and low-level material shortage warning threshold for the coal bunker; when the continuous coal level data output by the weighted fusion is greater than or equal to the high-level overflow warning threshold, or when the continuous coal level data is less than or equal to the low-level material shortage warning threshold, the AI ​​edge computing unit immediately outputs an alarm signal, which is uploaded to the central control room through the industrial network, simultaneously triggering the audible and visual alarm in the central control room, and using the industrial cloud monitoring APP mobile application to push the warning information to the mobile phones of management personnel in real time; when the high-level overflow alarm is triggered, the coal feeder is stopped in linkage control; when the low-level material shortage alarm is triggered, a coal replenishment prompt instruction is pushed.

[0017] Preferably, the classification and early warning determination of spontaneous combustion risk specifically involves: presetting temperature classification thresholds including a first-level warning and a second-level alarm; the first-level warning corresponds to the initial spontaneous combustion hazard stage of coal oxidation and temperature rise, and the second-level alarm corresponds to the critical danger stage of coal spontaneous combustion. When the real-time temperature of a single temperature measuring point reaches or exceeds the Level 1 warning threshold, a Level 1 warning is triggered. This information is uploaded to the central control room via the industrial network, simultaneously triggering the audible and visual alarms in the control room and pushing real-time warning information to management personnel's mobile phones via the industrial cloud monitoring APP. When the temperatures of multiple temperature measuring points simultaneously reach or exceed the Level 1 warning threshold, or when the real-time temperature of a single temperature measuring point reaches or exceeds the Level 2 alarm threshold, a Level 2 fire alarm is triggered. This simultaneously activates fire extinguishing devices, cuts off the power to the coal feeder, and sends a sprinkler or inert gas injection trigger signal to the fire protection system. The corresponding fire extinguishing devices, inert gas injection devices, and other fire-fighting equipment are activated to quickly address the risk of spontaneous combustion and prevent the accident from escalating.

[0018] This invention also provides a coal bunker intelligent exploration system based on multi-source data fusion, comprising: The multi-source data acquisition module includes a 3D laser scanning radar, a strain-temperature dual-parameter sensing fiber optic cable, multiple wireless temperature sensors, and a dual-spectrum thermal imaging camera, which are used to acquire point cloud data, coal level data, coal body temperature data, and images and temperature data inside the coal bunker, respectively. The AI ​​edge computing unit is communicatively connected to the multi-source data acquisition module and has built-in anti-dust filtering algorithm, digital twin modeling program, multi-source data fusion algorithm and abnormal state recognition program, which are used to realize digital twin modeling and multi-source fusion computing and linkage control. The anti-dust filtering algorithm uses mean filtering, median filtering, Kalman filtering and adaptive sliding filtering basic algorithms to perform unified preprocessing on the original analog and digital signals output by the multi-source data acquisition module, filtering out high-frequency noise, pulse interference and invalid interference data of random data jitter generated during signal transmission, and achieving preliminary noise reduction and smoothing of the original sensor data. The digital twin modeling program relies on 3D modeling software and industrial simulation platform to construct a static digital model of the coal bunker based on the dimensions of the fixed cavity and the static parameters of the monitoring equipment installation points. The multi-source data fusion algorithm includes a three-level basic fusion architecture of data layer, feature layer and decision layer. It is used for basic preprocessing operations such as time alignment, format standardization, data cleaning and dimensional integration of heterogeneous data with multiple dimensions and different formats, including temperature data, coal level height data, environmental parameter data and equipment operation data. The abnormal condition identification program pre-sets the safety threshold range of various monitoring parameters, compares the collected temperature, coal level, environmental and equipment operating parameters in real time, and determines the abnormal working condition and outputs a basic early warning signal when the monitoring data exceeds the preset threshold, instantaneous data change, data disconnection, or continuous parameter deviation occurs. The early warning and linkage control module is communicatively connected to the AI ​​edge computing unit and includes an audible and visual alarm and on-site control equipment. It is used to receive instructions from the AI ​​edge computing unit and execute alarm prompts and device linkage control. The visualization module, communicating with the AI ​​edge computing unit, constructs a static basic 3D model consistent with the physical coal bunker cavity structure and equipment layout through 3D laser scanning and 3D radar point cloud reconstruction. Real-time coal level height, internal temperature field distribution, and equipment operating parameters, after filtering, noise reduction, spatiotemporal alignment, and data fusion, are then dynamically associated with the static basic 3D model through data-driven dynamic mapping, model vertex deformation driving, and thermal map texture projection fusion, driving the static basic 3D model to update dynamically. Coal level height numerical labels, color thermal cloud maps of the temperature field, equipment operating parameter panels, abnormal alarm pop-ups, and flashing indicators are overlaid in the 3D twin scene, achieving an intuitive visualization of the coal bunker's internal morphology, material surface undulations, temperature distribution, and equipment status. This provides a real-time display of the coal bunker's digital twin 3D image, coal level data, temperature data, alarm information, and equipment operating status.

[0019] Preferably, the 3D laser scanning radar is installed at the center of the top of the coal bunker, and has a built-in rotary servo motor to drive the laser radar antenna to complete a 360° rotational scan of the entire space inside the bunker; the wireless temperature sensor is a LoRa wireless temperature sensor, which is evenly arranged on the upper, middle and lower parts of the inner wall of the coal bunker, and the temperature data collected by it is uploaded to the AI ​​edge computing unit through the LoRa gateway.

[0020] Preferably, multiple strain-temperature dual-parameter sensing optical fibers are vertically arranged on the inner wall of the coal bunker, fixed with special clamps and with reserved relaxation strain margin, and laid out from top to bottom to avoid twisting, bending and excessive stretching of the optical fibers throughout the process; two dual-spectrum thermal imaging cameras are symmetrically deployed on the top of the coal bunker to achieve redundant monitoring of the temperature data inside the bunker.

[0021] Preferably, the LoRa wireless temperature sensor is fixed to the inner wall of the coal bunker, with its sensing end in direct contact with the coal inside the bunker. Multiple LoRa wireless temperature sensors are distributed along the circumference and height of the inner wall of the coal bunker to achieve full-coverage monitoring of the internal temperature. Each LoRa wireless temperature sensor is powered by an independent explosion-proof battery, with a battery life meeting preset long-term endurance requirements, and collects temperature data periodically at a preset sampling frequency. The system also includes a LoRa gateway for receiving temperature data uploaded by each LoRa wireless temperature sensor and forwarding it to the AI ​​edge computing unit.

[0022] Beneficial effects: 1. This invention achieves comprehensive and highly synchronous data acquisition of core parameters such as coal level, temperature, and spatial morphology within a coal bunker by deploying a 3D laser scanning radar, strain-temperature dual-parameter sensing fiber optic cable, LoRa wireless temperature sensor, and dual-spectrum thermal imaging camera. By weighted fusion of the average coal level output from the real-time digital twin model of the coal bunker and the coal level data acquired by the sensing fiber optic cable, and by setting up a redundant alarm mechanism (automatically switching to independent fiber optic measurement in case of radar failure), high-precision continuous coal level output (measured error ≤ ±0.3%) is achieved, completely solving the problems of single sensors being susceptible to dust interference, low measurement accuracy, and poor data reliability due to lack of redundancy backup.

[0023] 2. This invention cleans the original point cloud using an anti-dust filtering algorithm and constructs a real-time digital twin model of the coal bunker across the entire scene using the Delaunay triangulation algorithm, achieving accurate digital mapping of the coal pile's surface morphology and internal structure. Based on this, the real-time digital twin model automatically identifies bridging structures on the coal pile surface or internal voids. Combined with the coal feeder's operating status and coal level change rate, it achieves comprehensive judgment and early warning of blockage faults, upgrading the traditional delayed alarm system relying on manual inspection or single current monitoring to a proactive prediction based on three-dimensional structural anomalies, significantly reducing false alarm and missed alarm rates.

[0024] 3. This invention constructs a coal temperature monitoring network that combines contact and non-contact methods, creating redundancy, by distributing multiple LoRa wireless temperature sensors (with their sensing ends in direct contact with the coal) along the warehouse wall and symmetrically deploying dual-spectrum thermal imaging cameras on the top. This solves the problems of blind spots, difficult wiring, and susceptibility to dust interference inherent in traditional temperature measurement solutions. Based on preset graded early warning thresholds and combined with multi-point joint judgment logic, it achieves graded and accurate early warning of spontaneous combustion risk and can be linked to fire-fighting devices, significantly advancing the fire prevention and control point.

[0025] 4. This invention integrates core algorithms such as anti-dust filtering, digital twin modeling, multi-source data fusion, silo blockage identification, and hierarchical early warning into an AI edge computing unit. Specifically, the anti-dust filtering algorithm uses mean filtering, median filtering, Kalman filtering, and adaptive sliding filtering to reduce noise and smooth sensor signals; the digital twin modeling program constructs a static model based on silo dimensions and equipment static parameters; the multi-source data fusion algorithm adopts a three-level architecture of data layer, feature layer, and decision layer to perform time-series alignment, standardization, cleaning, and integration of heterogeneous data; the abnormal state identification program compares parameters in real time using preset thresholds to determine abnormal operating conditions and output early warnings. By completing real-time processing and decision-making close to the data source, transmission latency is significantly reduced, achieving millisecond-level local linkage control. Based on coal level, silo blockage, and fire early warning results, the system triggers the shutdown of the coal feeder, the activation of arch-breaking mechanisms, and the activation of fire-fighting devices, and uploads the data to the central control room to trigger audible and visual alarms and APP push notifications, forming a closed-loop process of "perception-analysis-decision-control," significantly improving the safety and intelligence level of coal silo operation.

[0026] 5. The real-time digital twin model of the coal bunker constructed by this invention can be displayed in three dimensions on the host computer in the central control room, presenting the coal pile shape, coal level data, temperature distribution, equipment status, and alarm information in real time, and supporting historical data backtracking and trend analysis. Management personnel can fully grasp the operational status of the coal bunker without entering the site, significantly reducing the frequency of manual inspections and operational risks. At the same time, it provides accurate data support for production scheduling, inventory management, and equipment maintenance, realizing the digitalization, transparency, and intelligence of coal bunker safety management. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a flowchart illustrating the real-time acquisition process of multi-source data provided in this embodiment of the invention. Figure 2 This is a flowchart illustrating the point cloud preprocessing and digital twin modeling workflow provided in an embodiment of the present invention. Figure 3 This is a flowchart illustrating the multi-source fusion coal level calculation and material level alarm process provided in this embodiment of the invention. Figure 4 This is a flowchart illustrating the intelligent identification and early warning process for warehouse congestion provided in an embodiment of the present invention. Figure 5 A flowchart illustrating the coal temperature monitoring and fire classification early warning process provided in this embodiment of the invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] This embodiment provides a coal bunker intelligent exploration system based on multi-source data fusion, including: The multi-source data acquisition module includes a 3D laser scanning radar, a strain-temperature dual-parameter sensing fiber optic cable, multiple wireless temperature sensors, and a dual-spectrum thermal imaging camera, which are used to acquire point cloud data, coal level data, coal body temperature data, and images and temperature data inside the coal bunker, respectively. An AI edge computing unit is in communication connection with said multi-source data acquisition module, and has built-in dust-resistant filtering algorithms, a digital twin modeling program, a multi-source data fusion algorithm and an abnormal state recognition program. It is configured to implement digital twin modeling, multi-source fusion calculation and linkage control; wherein, the dust-resistant filtering algorithms adopt basic algorithms including mean filtering, median filtering, Kalman filtering and adaptive sliding filtering. Such algorithms can perform unified preprocessing on the original analog signals and digital signals output by sensors, effectively filter invalid interference data such as high-frequency noise, pulse interference and random data jitter generated during signal transmission, and realize preliminary noise reduction, smoothing and normalization of raw sensing data; the digital twin modeling program relies on three-dimensional modeling software and an industrial simulation platform, and constructs a static digital model of the coal bunker according to static parameters such as the fixed cavity size of the coal bunker and the installation positions of monitoring devices. Some modeling programs can access a small amount of real-time monitoring data to realize the basic visual display effect of coal bunker working conditions; the multi-source data fusion algorithm comprises a three-level basic fusion architecture of data layer, feature layer and decision layer, and can complete basic preprocessing operations such as time sequence alignment, format standardization, data cleaning and dimension integration for heterogeneous data with multi-dimensional and differentiated formats, including temperature data, coal level height data, environmental parameter data and equipment operation data obtained through monitoring; the abnormal state recognition program compares collected temperature, coal level, environment and equipment operation parameters in real time by presetting safe threshold ranges for various monitoring parameters in advance. When the monitoring data exceeds the preset threshold, or there are situations such as instantaneous data mutation, data disconnection and continuous parameter offset, it is determined as abnormal working condition and a basic early warning signal is output; An early warning and linkage control module is in communication connection with said AI edge computing unit, and comprises an audible and visual alarm and a field control device. It is configured to receive instructions from said AI edge computing unit and execute alarm prompt and equipment linkage control; A visualization module is in communication connection with said AI edge computing unit, and is configured to display the digital twin three-dimensional image, coal level data, temperature data, alarm information and equipment operation status of the coal bunker in real time.

[0031] Said 3D laser scanning radar is installed at the center of the top of the coal bunker, and has a built-in rotating servo motor configured to drive the laser radar antenna to complete 360° rotating scanning of the entire space in the bunker; said wireless temperature sensors are LoRa wireless temperature measurement sensors, which are evenly arranged on the upper, middle and lower parts of the inner wall of the coal bunker, and the temperature data collected thereby are uploaded to said AI edge computing unit through a LoRa gateway.

[0032] A plurality of said strain-temperature dual-parameter sensing optical fibers are vertically arranged on the inner wall of the coal bunker, fixed by special fixtures with loose strain allowance reserved, laid out in a top-down paying-off mode, and twisted, bent and overstretched of the optical fibers are avoided throughout the process; there are two said dual-spectrum thermal imaging cameras, which are symmetrically arranged on the top of the coal bunker, configured to implement redundant monitoring of temperature data in the bunker.

[0033] The LoRa wireless temperature sensor is fixed to the inner wall of the coal bunker, with its sensing end in direct contact with the coal inside. Multiple LoRa wireless temperature sensors are distributed along the circumference and height of the inner wall of the coal bunker to achieve full-coverage monitoring of the internal temperature. Each LoRa wireless temperature sensor is powered by an independent explosion-proof battery with a battery life meeting preset long-term endurance requirements, and collects temperature data periodically at a preset sampling frequency. The system also includes a LoRa gateway for receiving temperature data uploaded by each LoRa wireless temperature sensor and forwarding it to the AI ​​edge computing unit.

[0034] After the system is deployed and powered on, initial configuration is performed first. Basic configuration data, such as coal bunker dimensions, material level warning thresholds, temperature grading thresholds, blockage judgment parameters, and equipment linkage logic, are pre-entered into the AI ​​edge computing unit to complete system communication debugging and calibration. Data acquisition then proceeds. like Figure 1 As shown, during system operation, the 3D laser scanning radar continuously emits electromagnetic waves into the entire space inside the coal bunker, collecting 300,000 points of spatial point cloud data per second, including X / Y / Z coordinates and reflection intensity, and transmitting it to the AI ​​edge computing unit in real time; the mining-specific strain-temperature dual-parameter sensing fiber continuously collects real-time coal level data inside the bunker and transmits it synchronously to the AI ​​edge computing unit; multiple LoRa wireless temperature sensors (12 in this embodiment) collect coal temperature data at corresponding points once per minute and upload it to the AI ​​edge computing unit through the LoRa gateway; the dual-spectrum thermal imaging camera monitors the surface temperature of the coal inside the bunker and the real-time image inside the bunker in real time and transmits it to the AI ​​edge computing unit. At the same time, the AI ​​edge computing unit obtains the operating status data of the coal feeder in real time.

[0035] After receiving the raw point cloud data from the 3D laser scanning radar, the AI ​​edge computing unit runs a preset anti-dust filtering algorithm to complete data preprocessing, as shown in the example. Figure 2 As shown, firstly, based on the reflection intensity and spatial coordinate characteristics of point cloud data, a reflection intensity threshold and a spatial distance threshold are set to remove discrete noise points that exceed the threshold range, and invalid points and obvious noise points outside the warehouse are filtered out. Then, a statistical filtering algorithm was used to calculate the average distance between each point cloud and its 50 neighboring points. The standard deviation threshold for the distance was set to 1, and dust virtual points with an average distance exceeding 1 standard deviation were removed to solve the point cloud distortion problem in high dust environments. Finally, the filtered point cloud data was subjected to voxel downsampling processing, with the voxel size set to 5cm×5cm×5cm. This reduced the amount of point cloud data while preserving the coal pile outline features, improving the efficiency of subsequent modeling and ultimately obtaining effective coal pile point cloud data.

[0036] Based on the pre-processed effective coal pile point cloud data, the AI ​​edge computing unit fits the three-dimensional surface of the coal pile using the Delaunay triangulation algorithm. Combined with the pre-entered coal bunker size parameters, a real-time digital twin model of the coal bunker is constructed. The model can reflect the three-dimensional shape, spatial position and contour features of the coal pile in real time, and can also display the three-dimensional image inside the bunker in real time in the existing upper computer visualization module.

[0037] like Figure 3 As shown, based on the constructed real-time digital twin model of the coal bunker, the AI ​​edge computing unit automatically calculates the coal level at the highest point of the coal pile, the average coal level, and the coal storage volume. The average coal level output by the real-time digital twin model of the coal bunker is weighted and fused with the coal level collected by the strain-temperature dual-parameter sensor fiber optic, and the specific formula is: H=0.7×H1+0.3×H2, where H is the final output continuous coal level data that meets the accuracy requirements, H1 is the average coal level output by the real-time digital twin model of the coal bunker, and H2 is the coal level data collected by the strain-temperature dual-parameter sensor fiber optic. In this embodiment, the final output continuous coal level data measurement error is ≤ ±0.3%, which can continuously reflect the amount of coal in the silo in real time. Alarm judgment is completed based on preset material level thresholds. In this embodiment, 90% of the total coal silo height is preset as the high-level overflow warning threshold, and 10% of the total coal silo height is preset as the low-level material shortage warning threshold. When the continuous coal level data output by the weighted fusion is greater than or equal to the high-level overflow warning threshold, or when the continuous coal level data is less than or equal to the low-level material shortage warning threshold, the AI ​​edge computing unit immediately outputs an alarm signal, which is then uploaded to the central control room via the industrial network. This simultaneously triggers the audible and visual alarm in the central control room and pushes the warning information to the mobile phones of management personnel in real time using the industrial cloud monitoring APP. When the high-level overflow alarm is triggered, the coal feeder is stopped in conjunction with the control system to prevent overflow accidents. When a low-level material shortage alarm is triggered, a coal replenishment prompt is pushed to prevent subsequent system material shortages. Simultaneously, this embodiment incorporates a redundant alarm mechanism. The AI ​​edge computing unit monitors the operating status of the 3D laser scanning radar in real time. When a 3D laser scanning radar malfunction is detected, it automatically switches to mining-specific strain-temperature dual-parameter fiber optic coal level data for separate material level alarm judgment, preventing missed alarms due to equipment failure and improving system operational safety.

[0038] See Figure 4 The AI ​​edge computing unit, based on a real-time digital twin model of the coal bunker, extracts the contour features of the coal pile, identifies structural anomalies, and combines the operating status of the coal feeder with the coal level change trend to make a comprehensive judgment on blockage. The specific judgment logic is as follows: Based on the real-time digital twin model of the coal bunker, when a bridging structure with a span ≥1m and a height ≥0.8m appears on the surface of the coal pile, or when a cavity with a volume ≥1m³ appears inside the coal pile, it is judged as a structural identification anomaly. The operating status of the coal feeder is acquired in real time. When the coal feeder is in operation, the rate of change of continuous coal level data is calculated. When the coal level change rate is ≤±0.5% / h, it is judged as coal flow stagnation. When either the structural identification anomaly or the coal flow stagnation condition is met, the AI ​​edge computing unit judges it as a blockage, immediately outputs a signal to start the arch-breaking device, and records the location, size and degree of the blockage data, which is then uploaded to the host computer for display and storage.

[0039] See Figure 5 The AI ​​edge computing unit receives real-time temperature data from LoRa wireless temperature sensors and the highest temperature collected by a dual-spectrum thermal imaging camera. It analyzes the coal temperature distribution within the silo and determines early warnings based on preset temperature thresholds. In this embodiment, the preset first-level warning threshold is 75℃, and the second-level alarm threshold is 85℃. The specific warning and linkage logic is as follows: When a single temperature measurement point is detected to be ≥75℃, a first-level warning is triggered, uploaded to the central control room via the industrial network, and simultaneously triggered by the central control room's audible and visual alarm. The warning information is also pushed to the management personnel's mobile phones in real-time via the industrial cloud monitoring APP, prompting on-site verification. When three or more temperature measurement points are detected to be ≥75℃, or a single temperature measurement point is detected to be ≥85℃, a second-level fire alarm is triggered, simultaneously cutting off the coal feeder power and sending a spray or inert gas injection trigger signal to the fire protection system, achieving early prevention and emergency response to coal spontaneous combustion accidents. Compared to existing technologies, this invention achieves high-precision continuous measurement and redundant backup of coal level through multi-source heterogeneous data fusion; enables advanced intelligent identification of blockage structures through a real-time digital twin model of the coal bunker; realizes graded early warning and proactive prevention of spontaneous combustion risks through a distributed temperature measurement network combining contact and non-contact methods; and completes localized intelligent linkage control of the entire process through AI edge computing. This invention combines high-precision measurement, strong environmental adaptability (dust resistance), full-scenario digital perception, and closed-loop management capabilities, effectively overcoming the technical shortcomings of traditional coal bunker exploration techniques, such as limited methods, poor anti-interference capabilities, delayed hazard identification, and inability to achieve full-scenario digital management. It can be widely applied to intelligent safety management scenarios in various coal bunkers.

[0040] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for intelligent coal bunker exploration based on multi-source data fusion, characterized in that, Includes the following steps: S1. Multi-source data acquisition: Multi-source data is acquired through various sensors deployed in the coal bunker. The multi-source data includes: point cloud data, coal level data, coal temperature data, and images and temperature data inside the bunker. S2. Digital Twin Modeling: Through the AI ​​edge computing unit, an anti-dust filtering algorithm is executed on the collected raw 3D point cloud data to remove virtual points and noise, so as to obtain effective coal pile point cloud data. Based on the effective coal pile point cloud data, a real-time digital twin model of the coal bunker in the whole scene is constructed and synchronized with the coal bunker site in real time. S3. Multi-source fusion computing and linkage control: The AI ​​edge computing unit calculates coal level parameters based on the real-time digital twin model of the coal bunker, and performs weighted fusion of the coal level parameters with the collected coal level height data to output continuous coal level data that meets the accuracy requirements; S4. Based on the real-time digital twin model of the coal bunker, identify the structural anomalies of the coal pile, and combine the changing trends of the coal feeder's operating status and continuous coal level data to make a comprehensive judgment on the blockage; S5. Based on the collected coal temperature data and silo temperature data, determine the risk of spontaneous combustion by classifying and warning according to the preset temperature classification threshold. S6. The AI ​​edge computing unit outputs corresponding alarm signals and linkage control commands based on the continuous coal level data that meets the accuracy requirements, the comprehensive judgment result of the blockage, and the graded early warning judgment result of the spontaneous combustion risk.

2. The intelligent coal bunker exploration method based on multi-source data fusion according to claim 1, characterized in that, In step S2, the anti-dust filtering algorithm specifically includes: based on the reflection intensity and spatial coordinate characteristics of the point cloud data, removing discrete noise points that exceed a preset threshold range; using a statistical filtering algorithm, calculating the average distance between each point cloud and its neighboring points, and removing dust virtual points whose average distance exceeds a preset standard deviation range; and sampling the filtered point cloud data to reduce the amount of point cloud data while retaining the coal pile outline characteristics, thereby obtaining effective coal pile point cloud data.

3. The intelligent coal bunker exploration method based on multi-source data fusion according to claim 1, characterized in that, In step S3, the average coal level output by the real-time digital twin model of the coal bunker is weighted and fused with the collected coal level height data. The formula used is: H=a×H1+b×H2, where H is the final output continuous coal level data that meets the accuracy requirements, H1 is the average coal level output by the real-time digital twin model of the coal bunker, H2 is the coal level data collected by the sensing fiber, and a and b are preset weight coefficients, and a+b=1.

4. The intelligent coal bunker exploration method based on multi-source data fusion according to claim 1, characterized in that, In step S4, the comprehensive determination of blockage is specifically as follows: when the real-time digital twin model of the coal bunker identifies a bridging structure on the surface of the coal pile that exceeds a preset size or a cavity inside the coal pile that exceeds a preset volume, it is determined to be an abnormal structure identification. When the coal feeder is in operation and the calculated coal level change rate is lower than the preset threshold, it is determined that the coal flow is stagnant; if either the structural identification anomaly or the coal flow stagnant condition is met, it is determined that the silo is blocked.

5. The intelligent coal bunker exploration method based on multi-source data fusion according to claim 1, characterized in that, In step S5, the classification and early warning determination of spontaneous combustion risk specifically involves: presetting temperature classification thresholds including Level 1 and Level 2 alarms; when the real-time temperature of a single temperature measuring point reaches or exceeds the Level 1 early warning threshold, a Level 1 early warning is triggered, and the data is uploaded to the monitoring platform via the industrial network. The platform calls the message push interface and uses the industrial cloud monitoring APP mobile application to push the early warning information to the mobile phones of management personnel in real time, realizing mobile notification of over-limit alarms; when the temperature of multiple temperature measuring points reaches or exceeds the Level 1 early warning threshold simultaneously, or when the real-time temperature of a single temperature measuring point reaches or exceeds the Level 2 alarm threshold, a Level 2 fire alarm is triggered, and the fire extinguishing device is activated simultaneously, the power supply to the coal feeder is cut off, and a spray or inert gas injection trigger signal is sent to the fire protection system.

6. The intelligent coal bunker exploration method based on multi-source data fusion according to claim 1, characterized in that, In step S6, the specific logic of the material level alarm and linkage control is as follows: preset the high-level overflow warning threshold and the low-level material shortage warning threshold of the coal bunker; when the continuous coal level data output by the weighted fusion is greater than or equal to the high-level overflow warning threshold, or when the continuous coal level data is less than or equal to the low-level material shortage warning threshold, the AI ​​edge computing unit immediately outputs an alarm signal, which is uploaded to the central control room through the industrial network, and the central control room's audible and visual alarm is triggered simultaneously. The alarm information is also pushed to the mobile phones of management personnel in real time using the industrial cloud monitoring APP; when the high-level overflow alarm is triggered, the coal feeder is stopped in linkage control; when the low-level material shortage alarm is triggered, a coal replenishment prompt instruction is pushed.

7. A coal bunker intelligent detection system based on multi-source data fusion, used to execute the coal bunker intelligent detection method based on multi-source data fusion as described in claim 1, characterized in that, include: The multi-source data acquisition module includes a 3D laser scanning radar, a strain-temperature dual-parameter sensing fiber optic cable, multiple wireless temperature sensors, and a dual-spectrum thermal imaging camera, which are used to acquire point cloud data, coal level data, coal body temperature data, and images and temperature data inside the coal bunker, respectively. The AI ​​edge computing unit is communicatively connected to the multi-source data acquisition module and has built-in anti-dust filtering algorithm, digital twin modeling program, multi-source data fusion algorithm and abnormal state recognition program, which are used to realize digital twin modeling and multi-source fusion computing and linkage control. The anti-dust filtering algorithm uses mean filtering, median filtering, Kalman filtering and adaptive sliding filtering basic algorithms to perform unified preprocessing on the original analog and digital signals output by the multi-source data acquisition module, filtering out high-frequency noise, pulse interference and invalid interference data of random data jitter generated during signal transmission, and achieving preliminary noise reduction and smoothing of the original sensor data. The digital twin modeling program relies on 3D modeling software and industrial simulation platform to construct a static digital model of the coal bunker based on the dimensions of the fixed cavity and the static parameters of the monitoring equipment installation points. The multi-source data fusion algorithm includes a three-level basic fusion architecture of data layer, feature layer and decision layer. It is used for basic preprocessing operations such as time alignment, format standardization, data cleaning and dimensional integration of heterogeneous data with multiple dimensions and different formats, including temperature data, coal level height data, environmental parameter data and equipment operation data. The abnormal condition identification program pre-sets the safety threshold range of various monitoring parameters, compares the collected temperature, coal level, environmental and equipment operating parameters in real time, and determines the abnormal working condition and outputs a basic early warning signal when the monitoring data exceeds the preset threshold, instantaneous data change, data disconnection, or continuous parameter deviation occurs. The early warning and linkage control module is communicatively connected to the AI ​​edge computing unit and includes an audible and visual alarm and on-site control equipment. It is used to receive instructions from the AI ​​edge computing unit and execute alarm prompts and device linkage control. The visualization module, communicating with the AI ​​edge computing unit, constructs a static basic 3D model consistent with the physical coal bunker cavity structure and equipment layout through 3D laser scanning and 3D radar point cloud reconstruction. Real-time coal level height, internal temperature field distribution, and equipment operating parameters, after filtering, noise reduction, spatiotemporal alignment, and data fusion, are then dynamically associated with the static basic 3D model through data-driven dynamic mapping, model vertex deformation driving, and thermal map texture projection fusion, driving the static basic 3D model to update dynamically. Coal level height numerical labels, color thermal cloud maps of the temperature field, equipment operating parameter panels, abnormal alarm pop-ups, and flashing indicators are overlaid in the 3D twin scene, achieving an intuitive visualization of the coal bunker's internal morphology, material surface undulations, temperature distribution, and equipment status. This provides a real-time display of the coal bunker's digital twin 3D image, coal level data, temperature data, alarm information, and equipment operating status.

8. The intelligent coal bunker exploration system based on multi-source data fusion according to claim 7, characterized in that, The 3D laser scanning radar is installed at the center of the top of the coal bunker. It has a built-in rotary servo motor to drive the laser radar antenna to complete a 360° rotational scan of the entire space inside the bunker. The wireless temperature sensor is a LoRa wireless temperature sensor, which is evenly arranged on the upper, middle and lower parts of the inner wall of the coal bunker. The temperature data collected by the sensor is uploaded to the AI ​​edge computing unit through the LoRa gateway.

9. The intelligent coal bunker exploration system based on multi-source data fusion according to claim 7, characterized in that, Multiple strain-temperature dual-parameter sensing optical fibers are vertically deployed on the inner wall of the coal bunker, fixed with special clamps and with reserved relaxation strain margin, and laid out from top to bottom to avoid fiber twisting, bending and excessive stretching throughout the process; two dual-spectrum thermal imaging cameras are symmetrically deployed on the top of the coal bunker to achieve redundant monitoring of the temperature data inside the bunker.

10. The intelligent coal bunker exploration system based on multi-source data fusion according to claim 7, characterized in that, The LoRa wireless temperature sensor is fixed to the inner wall of the coal bunker, with its sensing end in direct contact with the coal inside. Multiple LoRa wireless temperature sensors are distributed along the circumference and height of the inner wall of the coal bunker to achieve full-coverage monitoring of the internal temperature. Each LoRa wireless temperature sensor is powered by an independent explosion-proof battery with a battery life meeting preset long-term endurance requirements, and collects temperature data periodically at a preset sampling frequency. The system also includes a LoRa gateway for receiving temperature data uploaded by each LoRa wireless temperature sensor and forwarding it to the AI ​​edge computing unit.

Images