Data processing system based on intelligent perception of running posture of coal mining machine

Abstract

The application discloses a data processing system based on intelligent perception of a running posture of a coal mining machine, and belongs to the technical field of intelligent coal mining, and comprises: a dual-vision cooperative perception module, which comprises a machine body posture vision sub-module and a drum cutting infrared vision sub-module; the machine body posture vision sub-module is used for collecting feature target images on the machine body of the coal mining machine and solving six-degree-of-freedom posture of the machine body; the drum cutting infrared vision sub-module is used for collecting infrared thermal radiation images of a cutting drum and a coal-rock interface, and extracting a drum contour line and a coal-rock boundary line; a digital twin modeling module is connected with the dual-vision cooperative perception module, is used for receiving six-degree-of-freedom posture data of the machine body and drum contour line data, constructing and driving a three-dimensional digital twin model based on a kinematic model of the coal mining machine to realize real-time synchronous movement, and outputting full-state parameters of the coal mining machine; high-precision posture perception and advanced control in a harsh environment can be realized, and the safety of fully-mechanized mining operation is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent coal mining technology, specifically relating to a data processing system based on intelligent perception of the operating posture of a coal mining machine. Background Technology

[0002] The coal mining machine is the core equipment in a fully mechanized mining face, and its operating posture directly determines the coal mining efficiency, coal seam flatness, and equipment safety. However, the harsh environment of the underground working face, with high concentrations of coal dust, water mist, insufficient lighting, and severe vibrations, leads to the following technical bottlenecks in the existing coal mining machine posture perception system: In the harsh underground environment, high concentrations of coal dust and water mist render traditional visual perception ineffective, making it difficult to obtain effective images for posture calculation; the drum height adjustment control relies on the coal and rock identification results after cutting, which has a serious lag and cannot achieve forward-looking adjustment, easily causing under-cutting or over-cutting; the changes in machine body posture and drum oscillation are coupled, and traditional single sensors cannot calculate the linkage between the two in real time, making decoupling difficult; at the same time, the massive amount of posture data uploaded to the central control center for processing causes decision-making delays, failing to meet the real-time requirements of cutting control.

[0003] To address the aforementioned issues, this invention proposes an intelligent perception data processing system for the operating posture of a coal mining machine based on digital twins and machine vision, enabling high-precision posture perception and advanced control in harsh environments. Summary of the Invention

[0004] The purpose of this invention is to provide a data processing system based on intelligent perception of the operating posture of a coal mining machine, which can solve the technical problems of visual failure under coal dust interference, lag in cutting control, and difficulty in decoupling the machine body and drum.

[0005] To achieve the above objectives, the present invention provides a data processing system based on intelligent perception of the operating posture of a coal mining machine, comprising:

[0006] The dual-vision collaborative perception module includes a fuselage pose vision submodule and a drum cutting infrared vision submodule. The fuselage pose vision submodule is used to acquire feature target images on the coal mining machine body and calculate the six-degree-of-freedom pose of the fuselage. The drum cutting infrared vision submodule is used to acquire infrared thermal radiation images of the interface between the cutting drum and the coal and rock and extract the drum outline and the coal-rock boundary line.

[0007] The digital twin modeling module is connected to the dual-vision collaborative perception module. It is used to receive the six-degree-of-freedom pose data of the machine body and the contour data of the drum. Based on the kinematic model of the coal mining machine, it constructs and drives the three-dimensional digital twin model to move synchronously in real time and outputs the full state parameters of the coal mining machine.

[0008] The advanced simulation and intervention module is connected to the digital twin modeling module. It is used to predict the roller trajectory within a preset time window based on the current and historical full-state parameters, and to perform collision detection between the predicted trajectory and the coal seam geological model. When the predicted trajectory deviates from the preset threshold or interference is about to occur, control commands are generated and sent to the coal mining machine actuator.

[0009] In one or more embodiments of the present invention, the fuselage pose vision submodule includes:

[0010] Mobile laser pointers deployed on the roadway track or multiple intrinsically safe high-definition cameras deployed on the top beam of the hydraulic support;

[0011] Feature targets set on the body of the coal mining machine;

[0012] The image preprocessing unit is used to perform Retinex multi-scale enhancement processing on the acquired images to suppress interference from strong light or low light conditions downhole.

[0013] The pose calculation unit is used to perform Hough transform on the preprocessed image, extract target features, and calculate the heading and pitch angles of the fuselage.

[0014] In one or more embodiments of the present invention, the drum cutting infrared vision submodule includes:

[0015] The intrinsically safe infrared thermal imager for mining, deployed at the top beam of the hydraulic support, operates in the 8~14μm long-wave infrared band and is used to penetrate coal dust and water mist to collect thermal radiation distribution images of the drum-coal-rock interface.

[0016] The image processing unit is used to extract the outline of the cutting drum from the infrared thermal radiation image and to identify the coal-rock boundary based on temperature difference features.

[0017] In one or more embodiments of the present invention, the technical parameters of the infrared thermal imager meet the following requirements: resolution not less than 384×288, NETD not greater than 50mK@25℃.

[0018] In one or more embodiments of the present invention, the dual-vision collaborative perception module further includes a spatiotemporal registration unit. The spatiotemporal registration unit adopts an FPGA hard synchronization trigger circuit to eliminate the microsecond-level acquisition time delay between the fuselage pose vision submodule and the drum cutting infrared vision submodule, and unifies the two types of vision data into the same spatiotemporal coordinate system.

[0019] In one or more embodiments of the present invention, the digital twin modeling module includes:

[0020] The kinematic modeling unit is used to establish the multi-link kinematic equations of the coal mining machine's traveling mechanism and rocker arm swing mechanism based on the DH model.

[0021] The full-state parameter calculation unit is used to take the six-degree-of-freedom pose of the fuselage as the base coordinate system input and the roller profile as the end effector coordinate system input. Through forward kinematics calculation and inverse kinematics verification, it can invert the full-state parameters, including the rocker arm swing angle, fuselage roll angle, pitch angle, and traction speed, in real time.

[0022] The virtual-real synchronous drive unit is used to input the calculated full-state parameters into the edge computing server, drive the three-dimensional digital twin model to move synchronously in real time, and simulate the interaction force between the cutting drum and the coal wall.

[0023] In one or more embodiments of the present invention, the advanced simulation and intervention module includes:

[0024] The trajectory prediction unit is used to optimize the least squares support vector machine (PSO-LSSVM) using the particle swarm optimization algorithm to build a time series prediction model. It takes the drum position sequence, the rate of change of the fuselage pitch angle, and the cutting current fluctuation within the historical preset time window as input features and outputs the predicted value of the drum height in the next 0.5 seconds.

[0025] The interference detection unit is used to perform spatial collision detection and deviation analysis between the predicted trajectory and the geological model of the coal seam roof and floor. When the predicted trajectory is about to collide with the hydraulic support top beam, cut too deep, or deviate from the coal-rock interface, it generates an early warning signal and calculates the optimal adjustment path.

[0026] The closed-loop control unit is used to automatically generate adjustment commands or traction speed adjustment commands based on the warning level, and send them directly to the coal mining machine PLC through the nearest fiber optic network node to achieve millisecond-level closed-loop "sensing-computing-control".

[0027] In one or more embodiments of the present invention, the warning level includes:

[0028] Level 1 warning corresponds to the predicted trajectory deviating from the coal-rock interface but not reaching the interference threshold, and only sends a prompt message to the operator;

[0029] Level 2 warning corresponds to the prediction that the trajectory is about to interfere or exceed the safety threshold, which automatically triggers the closed-loop control unit to send control commands.

[0030] In one or more embodiments of the present invention, the geological model of the coal seam top and bottom plate is dynamically corrected based on the coal and rock identification results after the previous coal cutting.

[0031] In one or more embodiments of the present invention, the following steps are included:

[0032] Step S1: The machine body pose vision submodule 110 acquires the image of the feature target 113 on the coal mining machine body and calculates the six-degree-of-freedom pose of the machine body; at the same time, the drum cutting infrared vision submodule acquires the infrared thermal radiation image of the interface between the cutting drum and the coal and rock and extracts the drum outline and the coal and rock boundary line.

[0033] Step S2: Perform spatiotemporal registration of the fuselage six-DOF pose data and the roller profile data to unify them into the same spatiotemporal coordinate system;

[0034] Step S3: Input the registered data into the digital twin modeling module, construct and drive the three-dimensional digital twin model to move synchronously in real time based on the kinematic model of the coal mining machine, and output the full state parameters of the coal mining machine.

[0035] Step S4: Based on the current and historical full-state parameters, use the PSO-LSSVM time series prediction model to predict the roller trajectory within the future preset time window;

[0036] Step S5: Collision detection and deviation analysis are performed between the predicted trajectory and the geological model of the coal seam roof and floor to determine whether an early warning is triggered.

[0037] Step S6: When an early warning is triggered, a corresponding control command is generated according to the warning level and sent directly to the coal mining machine PLC through the nearest fiber optic network node to adjust the drum height or traction speed to achieve closed-loop control.

[0038] Step S7: Monitor the actual response trajectory after the control command is executed, compare it with the predicted trajectory, generate feedback, and optimize the prediction model parameters.

[0039] Compared with existing technologies, the beneficial effects of this invention are as follows: by using an infrared thermal imager to penetrate coal dust and water mist, it solves the failure problem of traditional visible light vision in harsh underground environments, achieving clear imaging of the drum-coal-rock interface; by using FPGA hard synchronization to achieve spatiotemporal registration of dual vision data, eliminating microsecond-level delays; by combining the DH kinematic model to solve the full-state parameters, the attitude perception accuracy reaches the centimeter level; and by using a digital twin-based advanced extrapolation and proximity closed-loop control mechanism, the control delay is compressed from the second level to the millisecond level, meeting the real-time requirements of cutting control. At the same time, through advanced trajectory prediction and collision detection, accidents such as drum collisions with hydraulic supports and excessive cutting depth are effectively avoided, ensuring safe production in the fully mechanized mining face and improving the safety of fully mechanized mining operations. Attached Figure Description

[0040] Figure 1 This is a block diagram of the overall structure of the system of the present invention;

[0041] Figure 2 This is a schematic diagram of the composition of the dual-vision collaborative perception module of the present invention;

[0042] Figure 3 This is a flowchart illustrating the workflow of the digital twin modeling module of the present invention.

[0043] Figure 4 This is a flowchart of the advanced simulation and intervention module of the present invention;

[0044] Figure 5 This is an overall flowchart of the method of the present invention.

[0045] Explanation of key figure labels:

[0046] 100. Dual-vision collaborative perception module; 110. Fuselage pose vision sub-module; 111. Mobile laser pointer; 112. Intrinsically safe high-definition camera for mining; 113. Feature target; 114. Image preprocessing unit; 120. Drum cutting infrared vision sub-module; 121. Intrinsically safe infrared thermal imager for mining; 122. Image processing unit; 130. Spatiotemporal registration unit; 200. Digital twin modeling module; 210. Kinematic modeling unit; 220. Full-state parameter calculation unit; 230. Virtual-real synchronous driving unit; 300. Advanced simulation and intervention module; 310. Trajectory prediction unit; 320. Interference detection unit; 330. Closed-loop control unit. Detailed Implementation

[0047] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0048] Example 1: As Figure 1 As shown, this embodiment provides a data processing system based on intelligent perception of the operating posture of a coal mining machine, including a dual-vision collaborative perception module 100, a digital twin modeling module 200, and an advanced inference and intervention module 300.

[0049] The dual-vision collaborative perception module 100 includes a fuselage pose vision submodule 110 and a drum cutting infrared vision submodule 120. The fuselage pose vision submodule 110 is deployed at the roadway track and the top beam of the hydraulic support to acquire images of the feature targets 113 on the coal mining machine body; the drum cutting infrared vision submodule 120 is deployed at the top beam of the hydraulic support to acquire infrared thermal radiation images of the interface between the cutting drum and the coal and rock.

[0050] The dual-vision collaborative perception module 100 also includes a spatiotemporal registration unit 130, which uses an FPGA hard synchronization trigger circuit to eliminate the acquisition time delay between the two types of vision sensors and unify the two types of vision data into the same spatiotemporal coordinate system.

[0051] The digital twin modeling module 200 is connected to the dual-vision collaborative perception module 100. It receives the registered six-degree-of-freedom pose data of the machine body and the drum contour data. Based on the DH kinematic model, it constructs a digital twin of the coal mining machine and outputs the full-state parameters of the coal mining machine, including the rocker arm swing angle, the machine body roll angle, pitch angle, traction speed, etc.

[0052] The advanced simulation and intervention module 300 is connected to the digital twin modeling module 200 to receive full-state parameters. It uses the PSO-LSSVM time series prediction model to predict the drum trajectory in the next 0.5 seconds. The predicted trajectory is then compared with the coal seam geological model for collision detection. When the predicted trajectory is about to collide with the hydraulic support top beam, cut too deep, or deviate from the coal-rock interface, a control command is generated and sent directly to the coal mining machine PLC through the nearest fiber optic network node to achieve closed-loop control.

[0053] Example 2: As Figure 2 As shown in the figure, this embodiment provides a detailed description of the dual-vision collaborative perception module 100.

[0054] The fuselage pose vision submodule 110 includes:

[0055] A mobile laser pointer 111, deployed on the roadway track, is used to provide a stable light source reference;

[0056] Multiple high-definition cameras 112, such as the KBA12 intrinsically safe mining camera, are deployed on the top beam of the hydraulic support, with a resolution of 1920×1080 and a frame rate of 30fps.

[0057] The feature target 113, set on the body of the coal mining machine, is made of highly reflective material and has a cross-shaped or circular pattern to facilitate image recognition.

[0058] Image preprocessing unit 114 is used to perform Retinex multi-scale enhancement processing on the acquired image. The specific algorithm is as follows:

[0059] The image is decomposed into illuminance and reflectance components;

[0060] Multi-scale Gaussian filtering is used to extract reflection components at different scales;

[0061] Adaptive gain adjustment of the reflected component is performed to suppress overexposure under strong light and underexposure under low light.

[0062] The pose calculation unit 115 is used to perform Hough transform on the preprocessed image, extract the straight line or circular features of the target, and calculate the heading angle and pitch angle of the fuselage by combining the camera calibration parameters and the PnP (Perspective-n-Point) algorithm.

[0063] The drum cutting infrared vision submodule 120 includes:

[0064] The infrared thermal imager 121 deployed on the top beam of the hydraulic support adopts the YRH500 mining infrared thermal imager 121, with a working band of 8~14μm long-wave infrared, a resolution of 384×288, NETD (noise equivalent temperature difference) ≤50mK@25℃, and a frame rate of 50fps.

[0065] Image processing unit 122 is used to extract the outline of the cutting drum from infrared thermal radiation images and to identify the coal-rock boundary based on temperature difference features. The specific processing flow is as follows:

[0066] Non-uniformity correction is performed on the original infrared image to eliminate inconsistencies in sensor response;

[0067] Median filtering is used to remove salt-and-pepper noise while preserving edge information;

[0068] Extracting the roller contour line based on the Canny edge detection algorithm;

[0069] Identifying coal-rock boundaries based on temperature gradient analysis: The temperature of coal seams is usually lower than that of rocks, and there is a significant temperature gradient at the coal-rock interface. By setting temperature thresholds and gradient thresholds, coal-rock boundaries can be extracted.

[0070] The spatiotemporal registration unit 130 adopts an FPGA hard synchronization trigger circuit, and the specific implementation method is as follows:

[0071] FPGA chips generate synchronous trigger pulse signals; FPGA chips such as the Xilinx Spartan-6 series are used.

[0072] The trigger pulse is simultaneously sent to the high-definition camera 112 and the infrared thermal imager 121, triggering them to acquire data simultaneously.

[0073] The acquired image data is stamped with a uniform timestamp to eliminate microsecond-level acquisition delay;

[0074] The machine's visual and infrared images are unified into the machine's coordinate system, and spatial registration is achieved through a pre-calibrated transformation matrix.

[0075] Example 3: As Figure 3 As shown in the figure, this embodiment provides a detailed description of the digital twin modeling module 200.

[0076] Digital twin modeling module 200 includes:

[0077] Kinematic modeling unit 210 is used to establish the kinematic equations of the coal mining machine based on the Denavit-Hartenberg (DH) model. A four-degree-of-freedom kinematic model is established using the traction unit of the coal mining machine as the base, the rocker arm as the connecting rod, and the cutting drum as the end effector. The DH parameter table is as follows:

[0078] Link i Joint angle θi Linkage offset di Link length ai Linkage torsion angle αi 1 θ1 (fuselage roll) d1 (body width) a1 (fuselage length) 90° 2 θ2 (rocker arm swing) 0 a2 (rocker arm length) 0° 3 θ3 (Drum rotation) d3 (roller width) 0 0°

[0079] The full-state parameter calculation unit 220 is used to take the fuselage's six-DOF pose as the base coordinate system input and the drum profile as the end effector coordinate system input. Through forward kinematics calculation and inverse kinematics verification, it inverts the full-state parameters, including the rocker arm swing angle, fuselage roll angle, pitch angle, and traction speed, in real time. Specifically, it is implemented as follows:

[0080] Forward kinematics: Based on the DH parameters and joint variables, calculate the transformation matrix T from the body coordinate system to the roller coordinate system;

[0081] Inverse kinematics: Based on the fuselage pose and the actual position of the drum, solve for joint variables θ1, θ2, d1, etc., and use an iterative method to solve the nonlinear equation system;

[0082] The virtual-real synchronization drive unit 230 is used to input the calculated full-state parameters into the edge computing server, such as the NVIDIA Jetson Xavier NX edge computing platform, to drive the real-time synchronous motion of the 3D digital twin model. The 3D model is built using Unity 3D or Unreal Engine and receives the full-state parameters in real time via the UDP protocol to drive the model's motion. Simultaneously, based on a physics engine such as NVIDIA PhysX, the interaction forces between the cutting drum and the coal wall are simulated, and the deformation of the drum after being subjected to force is calculated for subsequent trajectory correction.

[0083] Example 4: Figure 4 As shown in the figure, this embodiment provides a detailed description of the advanced simulation and intervention module 300.

[0084] The Advanced Prediction and Intervention Module 300 includes:

[0085] The trajectory prediction unit 310 is used to construct a time series prediction model by optimizing a least squares support vector machine (PSO-LSSVM) using the particle swarm optimization algorithm. The specific implementation steps are as follows:

[0086] Data acquisition: Collect historical data such as the drum position sequence (50 data points, sampling rate 10Hz), the rate of change of the machine pitch angle, and the fluctuation of the cutting current within a 5-second period;

[0087] Data preprocessing: Training samples were constructed using the sliding window method, with a window length of 5 seconds and a prediction step size of 0.5 seconds;

[0088] PSO optimization of LSSVM parameters: The penalty factor C and kernel function parameter σ of LSSVM are used as optimization variables, and the root mean square error of prediction is used as the fitness function. The PSO algorithm is used for optimization, with a particle swarm size of 30 and 100 iterations.

[0089] Model training and prediction: The LSSVM model is trained using optimized parameters. The input is the feature data of the previous 5 seconds, and the output is the predicted value of the roller height for the next 0.5 seconds. The prediction result is updated every 0.1 seconds.

[0090] Interference detection unit 320 is used to perform spatial collision detection and deviation analysis between the predicted trajectory and the geological model of the coal seam roof and floor. The geological model of the coal seam roof and floor is dynamically corrected based on the coal and rock identification results after the previous coal cutting and is stored in the database of the roadway control center. The interference detection algorithm is as follows:

[0091] The predicted roller trajectory point sequence was compared with the coal seam roof and floor model;

[0092] Calculate the shortest distance between each prediction point and the top and bottom plates;

[0093] If the distance between the predicted point and the top plate is less than the safety threshold of 0.2m, it is determined that it is about to collide with the top beam;

[0094] If the distance between the predicted point and the bottom plate is less than the safety threshold of 0.1m, it is determined that the bottom is about to be cut.

[0095] If the predicted point deviates from the coal-rock interface beyond the preset range of the coal-rock boundary line, it is determined to be a cutting deviation;

[0096] The closed-loop control unit 330 is used to automatically generate adjustment commands or traction speed adjustment commands based on the warning level. The warning levels are divided into two levels:

[0097] Level 1 warning: The predicted trajectory deviates from the coal-rock interface but does not reach the interference threshold. Only a prompt message is sent to the operator through the human-machine interface (HMI) of the roadway control center, suggesting manual intervention.

[0098] Level 2 Early Warning: If the predicted trajectory is about to interfere or exceed a safety threshold, the closed-loop control unit 330 is automatically triggered to send control commands. These commands are transmitted directly to the coal mining machine PLC via a fiber optic network through the nearest hydraulic support controller or edge computing node. If an S7-1200 series PLC is used, the command format conforms to the Modbus TCP protocol and includes:

[0099] Command type: 0x01 (drum height adjustment), 0x02 (traction speed adjustment);

[0100] Target value: Target roller height (unit mm, accuracy 1 mm) or target traction speed (unit m / min, accuracy 0.01 m / min).

[0101] Execution timestamp: Specifies the time when the command is executed, ensuring multi-axis synchronization;

[0102] After the control command is executed, the actual response data returned by the coal mining machine PLC, such as drum height and traction speed, is monitored and compared with the predicted trajectory to form a closed-loop feedback, which is used to optimize the parameters of the prediction model for the next round.

[0103] Example 5: Figure 5 As shown, this embodiment provides a data processing method based on intelligent perception of the operating posture of a coal mining machine, applied to the aforementioned data processing system, including the following steps:

[0104] Step S1: The machine body pose vision submodule 110 acquires the image of the feature target 113 on the coal mining machine body and calculates the six-degree-of-freedom pose of the machine body; at the same time, the drum cutting infrared vision submodule 120 acquires the infrared thermal radiation image of the interface between the cutting drum and the coal and rock and extracts the drum outline and the coal and rock boundary line.

[0105] Step S2: The six-degree-of-freedom pose data of the fuselage and the roller profile data are spatiotemporally registered and unified to the same spatiotemporal coordinate system through the FPGA hard synchronization trigger circuit.

[0106] Step S3: Input the registered data into the digital twin modeling module, construct and drive the three-dimensional digital twin model to move synchronously in real time based on the DH kinematic model, and output the full state parameters of the coal mining machine.

[0107] Step S4: Based on the current and historical full-state parameters, use the PSO-LSSVM time series prediction model to predict the roller trajectory in the next 0.5 seconds;

[0108] Step S5: Perform spatial collision detection and deviation analysis between the predicted trajectory and the dynamically corrected geological model of the coal seam roof and floor to determine whether an early warning is triggered.

[0109] Step S6: When an alarm is triggered, generate corresponding control commands based on the alarm level.

[0110] If it is a Level 1 warning, only a notification message will be sent to the operator's HMI;

[0111] If it is a level 2 early warning, control commands are sent directly to the coal mining machine PLC through the nearest fiber optic network node to adjust the drum height or traction speed, achieving millisecond-level closed-loop control;

[0112] Step S7: Monitor the actual response trajectory after the control command is executed, compare it with the predicted trajectory, calculate the prediction error, and optimize the PSO-LSSVM model parameters using an online learning method to form a feedback closed loop.

[0113] Example 6: To achieve the best technical effect, the key parameters of the present invention are preferably as follows:

[0114] Infrared thermal imager 121 parameters:

[0115] Resolution: No less than 384×288, preferably 640×512;

[0116] NETD: Not greater than 50mK, preferably ≤30mK;

[0117] Frame rate: no less than 50fps, preferably 100fps;

[0118] Operating band: 8~14μm long-wave infrared;

[0119] Forecast time window:

[0120] Historical data window length: 3-8 seconds, preferably 5 seconds;

[0121] Prediction step size: 0.3~1.0 seconds, preferably 0.5 seconds;

[0122] Early warning safety threshold:

[0123] Top plate collision threshold: 0.15~0.25m, preferably 0.2m;

[0124] The impact threshold of the base plate is 0.05~0.15m, preferably 0.1m;

[0125] Controlling delay:

[0126] Closed-loop control delay: ≤10ms, preferably ≤5ms.

[0127] Example 7: To verify the technical effect of the present invention, an industrial test was conducted at a fully mechanized coal mining face in a coal mine. The working face was 200m long and 3.5m high, using an MG500 / 1130-WD coal mining machine. During the test, six YRH500 infrared thermal imagers 121 were deployed at the top beam of the hydraulic support, with each infrared thermal imager 121 spaced 30m apart. Two KBA12 high-definition cameras 112 were deployed on the roadway track. The edge computing server used was an NVIDIA Jetson Xavier NX, and the coal mining machine PLC used was an S7-1200.

[0128] The test results are as follows:

[0129] Attitude perception accuracy: Compared with manual measurement values, the body posture error is ≤ ±2cm, and the roller height error is ≤ ±3cm;

[0130] Coal dust penetration capability: When the coal dust concentration is ≥1000mg / m³, the infrared image can still clearly identify the outline of the drum, with a recognition success rate of ≥95%;

[0131] Control real-time performance: The average delay from sensing to issuing the control command is 8ms, which meets the requirements for real-time control;

[0132] Cutting effect: Compared with the control coal without this system, the top coal retention rate is reduced by 35%, the bottom cutting accident rate is reduced by 80%, and the number of hydraulic support collision alarms is reduced to zero.

[0133] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0134] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A data processing system based on intelligent perception of the operating posture of a coal mining machine, characterized in that, include: The dual-vision collaborative perception module includes a fuselage pose vision submodule and a drum cutting infrared vision submodule; The fuselage pose vision submodule is used to acquire feature target images on the coal mining machine body and calculate the six-degree-of-freedom pose of the fuselage; the drum cutting infrared vision submodule is used to acquire infrared thermal radiation images of the interface between the cutting drum and the coal and rock, and extract the drum outline and the coal and rock boundary line. The digital twin modeling module is connected to the dual-vision collaborative perception module. It is used to receive the six-degree-of-freedom pose data of the machine body and the contour data of the drum. Based on the kinematic model of the coal mining machine, it constructs and drives the three-dimensional digital twin model to move synchronously in real time and outputs the full state parameters of the coal mining machine. The advanced simulation and intervention module is connected to the digital twin modeling module. It is used to predict the roller trajectory within a preset time window based on the current and historical full-state parameters, and to perform collision detection between the predicted trajectory and the coal seam geological model. When the predicted trajectory deviates from the preset threshold or interference is about to occur, control commands are generated and sent to the coal mining machine actuator.

2. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 1, characterized in that, The fuselage pose vision submodule includes: Mobile laser pointers deployed on the roadway track or multiple intrinsically safe high-definition cameras deployed on the top beam of the hydraulic support; Feature targets set on the body of the coal mining machine; The image preprocessing unit is used to perform Retinex multi-scale enhancement processing on the acquired images to suppress interference from strong light or low light conditions downhole. The pose calculation unit is used to perform Hough transform on the preprocessed image, extract target features, and calculate the heading and pitch angles of the fuselage.

3. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 1, characterized in that, The drum cutting infrared vision submodule includes: The intrinsically safe infrared thermal imager for mining, deployed at the top beam of the hydraulic support, operates in the 8~14μm long-wave infrared band and is used to penetrate coal dust and water mist to collect thermal radiation distribution images of the drum-coal-rock interface. The image processing unit is used to extract the outline of the cutting drum from the infrared thermal radiation image and to identify the coal-rock boundary based on temperature difference features.

4. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 3, characterized in that, The technical parameters of the infrared thermal imager meet the following requirements: resolution not less than 384×288, NETD not greater than 50mK@25℃.

5. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 1, characterized in that, The dual-vision collaborative perception module also includes a spatiotemporal registration unit. The spatiotemporal registration unit uses an FPGA hard synchronization trigger circuit to eliminate the microsecond-level acquisition time delay between the fuselage pose vision submodule and the drum cutting infrared vision submodule, and unifies the two types of visual data into the same spatiotemporal coordinate system.

6. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 1, characterized in that, The digital twin modeling module includes: The kinematic modeling unit is used to establish the multi-link kinematic equations of the coal mining machine's traveling mechanism and rocker arm swing mechanism based on the DH model. The full-state parameter calculation unit is used to take the six-degree-of-freedom pose of the fuselage as the base coordinate system input and the roller profile as the end effector coordinate system input. Through forward kinematics calculation and inverse kinematics verification, it can invert the full-state parameters, including the rocker arm swing angle, fuselage roll angle, pitch angle, and traction speed, in real time. The virtual-real synchronous drive unit is used to input the calculated full-state parameters into the edge computing server, drive the three-dimensional digital twin model to move synchronously in real time, and simulate the interaction force between the cutting drum and the coal wall.

7. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 1, characterized in that, The advanced simulation and intervention module includes: The trajectory prediction unit is used to optimize the least squares support vector machine (PSO-LSSVM) using the particle swarm optimization algorithm to build a time series prediction model. It takes the drum position sequence, the rate of change of the fuselage pitch angle, and the cutting current fluctuation within the historical preset time window as input features and outputs the predicted value of the drum height in the next 0.5 seconds. The interference detection unit is used to perform spatial collision detection and deviation analysis between the predicted trajectory and the geological model of the coal seam roof and floor. When the predicted trajectory is about to collide with the hydraulic support top beam, cut too deep, or deviate from the coal-rock interface, it generates an early warning signal and calculates the optimal adjustment path. The closed-loop control unit is used to automatically generate adjustment commands or traction speed adjustment commands based on the warning level, and send them directly to the coal mining machine PLC through the nearest fiber optic network node to achieve millisecond-level closed-loop "sensing-computing-control".

8. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 7, characterized in that, The warning levels include: Level 1 warning corresponds to the predicted trajectory deviating from the coal-rock interface but not reaching the interference threshold, and only sends a prompt message to the operator; Level 2 warning corresponds to the prediction that the trajectory is about to interfere or exceed the safety threshold, which automatically triggers the closed-loop control unit to send control commands.

9. The data processing system based on intelligent perception of the operating posture of a coal mining machine according to claim 7, characterized in that, The geological model of the top and bottom of the coal seam is dynamically corrected based on the coal and rock identification results after the previous coal cutting.

10. A data processing method based on intelligent perception of the operating posture of a coal mining machine, applied to the system described in any one of claims 1 to 9, characterized in that, Includes the following steps: Step S1: The machine body pose vision submodule 110 acquires the image of the feature target 113 on the coal mining machine body and calculates the six-degree-of-freedom pose of the machine body; at the same time, the drum cutting infrared vision submodule acquires the infrared thermal radiation image of the interface between the cutting drum and the coal and rock and extracts the drum outline and the coal and rock boundary line. Step S2: Perform spatiotemporal registration of the fuselage six-DOF pose data and the roller profile data to unify them into the same spatiotemporal coordinate system; Step S3: Input the registered data into the digital twin modeling module, construct and drive the three-dimensional digital twin model to move synchronously in real time based on the kinematic model of the coal mining machine, and output the full state parameters of the coal mining machine. Step S4: Based on the current and historical full-state parameters, use the PSO-LSSVM time series prediction model to predict the roller trajectory within the future preset time window; Step S5: Collision detection and deviation analysis are performed between the predicted trajectory and the geological model of the coal seam roof and floor to determine whether an early warning is triggered. Step S6: When an early warning is triggered, a corresponding control command is generated according to the warning level and sent directly to the coal mining machine PLC through the nearest fiber optic network node to adjust the drum height or traction speed to achieve closed-loop control. Step S7: Monitor the actual response trajectory after the control command is executed, compare it with the predicted trajectory, generate feedback, and optimize the prediction model parameters.

Images