Coal pile compaction device and compaction method

Abstract

The invention relates to the technical field of coal pile compaction data processing, in particular to a coal pile compaction device and method, and the device comprises a data collection module, a federated learning aggregation module, a compaction decision module, a path planning module and an execution control module. A multi-source sensor is used for collecting coal pile humidity distribution, a three-dimensional contour and compaction feedback data in real time, and a coal pile three-dimensional dynamic model is constructed. And the federated learning aggregation module performs multi-target crossover variation on local parameters of multiple devices based on a genetic algorithm to generate a global compaction strategy model, and outputs, fuses and generates a vibration frequency and compaction force regulation and control instruction in combination with the coal hardening risk assessment model. And in combination with block chain encryption storage, compaction quality traceability is realized, the coal pile compaction uniformity and the oxidation reaction inhibition effect are improved, and the hardening risk and the operation energy consumption are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of coal pile compaction data processing, and in particular to a coal pile compaction device and a compaction method. Background Art

[0002] The coal pile compaction device is a special equipment that uses mechanical pressure to increase the bulk density of loose coal. It is mainly composed of a hydraulic drive system, a vibrating compaction roller, and a traveling mechanism. Its working principle is based on the combined action of static pressure and dynamic vibration. It uses high-amplitude, low-frequency vibration waves to destroy the pore structure between coal particles. Combined with the vertical pressure generated by the roller's own weight, it causes the coal particles to rearrange and reduce the gap ratio. This device is widely used in thermal power, coking, and coal storage and transportation. By increasing the compaction density of the coal pile, it can enhance the compressive strength and stability of the pile, reduce the surface weathering rate, and inhibit dust emission and exothermic oxidation reactions, thereby reducing the risk of spontaneous combustion and environmental pollution. Some models are integrated with an intelligent control system that can dynamically adjust the compaction strength according to the coal quality characteristics and pile height parameters to achieve energy efficiency optimization and operation standardization.

[0003] Existing coal pile compaction technology has the following technical pain points: When using crawler-type engineering machinery or excavator buckets to implement surface compaction, the equipment frequently moves up and down the coal pile and back and forth, causing disturbances in the coal structure and posing safety risks of coal pile collapse and equipment overturning; the mechanical reciprocating operation mode leads to prolonged compaction time per unit area, low operating efficiency and high energy consumption; when compacting high-humidity coal, static gravity compaction methods are prone to cause coal hardening under low temperature conditions, making subsequent access difficult; traditional equipment is limited by its mechanical structure and cannot effectively compact the top area of ​​the coal pile, forming an active oxidation reaction area. In addition, existing technologies lack intelligent control capabilities, making it difficult to dynamically adjust the compaction strength according to the characteristics of the coal type and the geometric parameters of the pile. There are problems with insufficient material adaptability and poor compaction uniformity, which accelerates the oxidation process in the middle layer of the coal pile and increases the risk of spontaneous combustion. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a coal pile compaction device and compaction method to solve the problems of low compaction efficiency, high risk of coal compaction, insufficient compaction in the top area due to mechanical structure limitations and lack of intelligent control in the existing coal pile compaction technology, and inability to dynamically adjust the compaction strength according to coal quality and pile parameters, which affects the oxidation reaction inhibition effect and the safety and stability of the pile.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: In a first aspect, the present invention provides a coal pile compacting device, comprising: Data acquisition module, used to obtain moisture distribution data, three-dimensional profile data and compaction feedback data of the coal pile in real time through multi-source sensors; A federated learning aggregation module is used to receive local optimization parameters of multiple compaction devices, perform multi-objective cross-mutation operations on the local optimization parameters based on a genetic algorithm, and generate a global compaction strategy model; A compaction decision module is used to fuse the output data of the global compaction strategy model and the coal compaction risk assessment model to generate a control instruction set including vibration frequency and compaction intensity parameters; a path planning module for constructing an elevation grid map based on the three-dimensional laser scanning data of the coal pile and iteratively optimizing the traversal order of path nodes using a genetic algorithm based on the void ratio distribution characteristics in the global compaction strategy model; The execution control module is used to drive the walking mechanism to move along the track and adjust the vibration mode and contact angle of the compaction roller according to the compaction unit force control instruction.

[0006] Furthermore, in the coal pile compacting device of the present invention, the federated learning aggregation module includes: a parameter encoding unit for mapping the local model weights of each compacting device into a binary chromosome encoding sequence that can be processed by the fitness function; a crossover and mutation unit, configured to perform iterative operations of adaptive crossover probability and dynamic mutation probability on the binary chromosome coding sequence to generate an optimized population that satisfies multi-objective constraints; The strategy generation unit is used to extract the Pareto optimal solution set from the optimized population, generate global compaction strategy model parameters through key decryption, and feed back the updated local model parameters to each compaction device.

[0007] Furthermore, in the coal pile compacting device of the present invention, the path planning module includes: a topology analysis unit for identifying concave areas and slope mutation nodes on the coal pile surface based on laser radar scanning data, and constructing a path taboo search space containing elevation constraints; an iterative optimization unit, configured to use the path taboo search space as an initial solution set, adopt an elite retention strategy in a genetic algorithm to screen path segments with the highest coverage, and generate an optimal path sequence that meets a void ratio threshold; The collaborative correction unit is used to receive the global void ratio heat map sent by the federated learning aggregation module, and dynamically adjust the node traversal weight of the optimal path sequence according to the coordinates of the high void ratio area in the heat map.

[0008] Furthermore, in the coal pile compacting device of the present invention, the compaction decision module includes: a compaction prediction unit, configured to input moisture distribution data into a pre-trained coal quality compaction risk assessment model, optimize the feature weight parameters of the model through a genetic algorithm, and output a predicted value of the compaction risk level; The multi-objective optimization unit is used to use the predicted value of the compaction risk level as a constraint condition, and adopt the multi-objective particle swarm algorithm to simultaneously optimize the compaction strength, energy efficiency and porosity indicators to generate a Pareto front solution set containing vibration parameters; the instruction fusion unit is used to fuse the vibration parameters in the Pareto front solution set with the real-time temperature gradient data collected by the infrared thermal imaging module to generate a composite control instruction containing vibration frequency and pressure gradient value.

[0009] Furthermore, in the coal pile compacting device of the present invention, the execution control module includes: an angle adaptation unit for calculating a real-time deflection angle control value of the universal joint based on the coal pile slope data acquired by the laser scanning module and the contact angle optimization parameters generated by the genetic algorithm; A mode switching unit is used to receive a detection signal from a coal particle hardness sensor and trigger a deformation drive instruction for the rigid protrusion structure on the surface of the compaction drum when the hardness value exceeds a preset threshold; The abnormal response unit is used to call the compaction and spraying collaborative operation instruction set generated by the genetic algorithm in the pre-stored emergency action sequence library when the infrared thermal imaging module detects that the local temperature rise rate exceeds the safety threshold.

[0010] In a second aspect, the present invention provides a coal pile compacting method, which is applied to the above-mentioned coal pile compacting device, comprising: Multi-source sensors collect coal pile moisture gradient data, 3D laser point cloud data, and real-time pressure feedback data to construct a 3D dynamic model of the coal pile. The three-dimensional dynamic model is input into the local decision-making model of multiple compacting equipment, and the coal quality characteristic parameters in the model are optimized by a genetic algorithm to generate a chromosome coding sequence under the constraint of a fitness function; Performing iterative calculations of adaptive crossover probability and mutation operators on the chromosome coding sequence under a federated learning framework to generate an optimized parameter set of a global compaction strategy model; Extracting the coal pile surface topology features based on the three-dimensional laser point cloud data, and optimizing the traversal priority of the compaction path nodes using a genetic algorithm according to the void ratio distribution parameters in the global compaction strategy model; According to the compaction strength control parameters output by the global compaction strategy model, the walking mechanism is driven to move along the planned path, and the vibration frequency of the compaction roller and the angular offset of the universal joint are synchronously adjusted.

[0011] Furthermore, in the coal pile compaction method of the present invention, generating a chromosome population under the guidance of a fitness function includes: constructing a three-dimensional fitness evaluation space based on a risk level value output by a coal compaction risk assessment model, a coal pile volume parameter extracted from laser radar scanning data, and compaction energy consumption history data; The chromosome coding sequence is screened by multi-objectives using a non-dominated sorting algorithm, and the path planning solution with the highest comprehensive score in coverage and energy efficiency is retained in the Pareto front solution set. According to the surface curvature parameters in the three-dimensional dynamic model of coal pile, the action probability distribution of the genetic algorithm mutation operator is dynamically adjusted.

[0012] Furthermore, in the coal pile compaction method of the present invention, the iterative optimization compaction path includes: An elevation grid map containing void ratio distribution characteristics is constructed based on LiDAR scanning data, and areas with void ratios above a preset threshold are marked as high-priority path nodes. The two-point crossover strategy in the genetic algorithm is used to perform topological matching and exchange on the path segments of adjacent generations to generate an optimized path sequence. According to the regional weight values ​​in the global porosity heat map issued by the federated learning aggregation module, the elite path fragments with a coverage rate of more than 95% in the historical optimization path library are inserted into the current path sequence.

[0013] Furthermore, in the coal pile compaction method of the present invention, adjusting the vibration amplitude of the compaction roller includes: Based on the risk level value output by the coal compaction risk assessment model, the spatial boundaries of the high compaction strength zone and the vibration avoidance zone are divided; A low-frequency, high-amplitude vibration mode is used in high compaction intensity areas, and the amplitude parameters are dynamically adjusted based on compaction feedback data; In the vibration avoidance area, it switches to a high-frequency, micro-amplitude vibration mode and is synchronously triggered by the sealant spraying unit to form a continuous insulation layer on the surface of the coal pile.

[0014] Furthermore, the coal pile compaction method of the present invention further includes: when the infrared thermal imaging module detects that the local temperature gradient exceeds a safety threshold, calling a pre-trained genetic algorithm emergency response model to generate a collaborative operation instruction set for the compaction roller and the sealant spraying unit; After each compaction operation is completed, the equipment operation log containing vibration parameters, void ratio distribution and energy consumption data will be encrypted and uploaded to the blockchain node to generate a time-stamped compaction quality traceability record; The newly collected 3D point cloud data is fused with the elite chromosome coding sequences in the historical optimization parameter library through the federated learning framework to update the initial population generation rules of the genetic algorithm.

[0015] Beneficial effects of the present invention: The coal pile compaction device and method provided by the present invention significantly improve operational efficiency and safety through multi-module collaboration and intelligent algorithm integration. The federated learning aggregation module integrates local parameters of multiple devices to generate a global optimization strategy, and combines genetic algorithms to dynamically adjust the compaction path and strength parameters, effectively solving the problem of insufficient compaction of the coal pile top and complex areas with traditional technologies; the path planning module guides the traversal priority of compaction nodes based on lidar three-dimensional modeling and void ratio heat map, improves the processing efficiency of high void ratio areas through elite path fragment insertion strategy, and reduces the activity of oxidation reactions inside the coal pile; the vibration mode switching mechanism of the execution control module and the sealant spraying cooperate to form a temperature-pressure dynamic feedback closed loop, suppressing local temperature rise anomalies while avoiding the risk of hardening caused by excessive compaction in high humidity areas; the data closed-loop management mechanism realizes the simultaneous enhancement of compaction quality traceability and adaptive adjustment capabilities through blockchain encrypted storage and federated learning parameter updates. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0017] Figure 1 The present invention provides a flowchart of a method for adjusting a coal pile compacting device.

[0018] Figure 2 This is a first schematic diagram of an application scenario of a coal pile compacting device provided by an embodiment of the present invention.

[0019] Figure 3 This is a second schematic diagram of an application scenario of a coal pile compacting device provided by an embodiment of the present invention.

[0020] Figure 4 This is a third schematic diagram of an application scenario of a coal pile compacting device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0022] In a first aspect, the present invention provides a coal pile compacting device, comprising: Data acquisition module, used to obtain moisture distribution data, three-dimensional profile data and compaction feedback data of the coal pile in real time through multi-source sensors; A federated learning aggregation module is used to receive local optimization parameters of multiple compaction devices, perform multi-objective cross-mutation operations on the local optimization parameters based on a genetic algorithm, and generate a global compaction strategy model; A compaction decision module is used to fuse the output data of the global compaction strategy model and the coal compaction risk assessment model to generate a control instruction set including vibration frequency and compaction intensity parameters; a path planning module for constructing an elevation grid map based on the three-dimensional laser scanning data of the coal pile and iteratively optimizing the traversal order of path nodes using a genetic algorithm based on the void ratio distribution characteristics in the global compaction strategy model; The execution control module is used to drive the walking mechanism to move along the track and adjust the vibration mode and contact angle of the compaction roller according to the compaction unit force control instruction.

[0023] The present invention provides a coal pile compaction device that realizes intelligent control of the coal pile compaction process through the collaboration of multiple modules. The device first collects the moisture distribution data, three-dimensional contour data and compaction feedback data of the coal pile in real time through multi-source sensors. The multi-source sensor includes a laser radar scanning unit, an infrared humidity sensor and a pressure feedback device, wherein the laser radar scanning unit constructs a three-dimensional point cloud model of the coal pile surface by emitting pulsed lasers and receiving reflected signals, the infrared humidity sensor inverts the humidity gradient distribution by detecting the reflection spectrum characteristics of the coal pile surface, and the pressure feedback device monitors the stress distribution in the contact area between the compaction roller and the coal pile in real time through a piezoelectric sensor array. The data acquisition module performs time synchronization and spatial registration on the above-mentioned multimodal data to generate a three-dimensional dynamic model of the coal pile.

[0024] The federated learning aggregation module receives local optimization parameters from multiple compaction devices, which include model weights obtained by each device based on historical operation data training. The parameter encoding unit in the module converts the model weights into a binary chromosome coding sequence that can be processed by the fitness function. This coding sequence defines the weight relationship between the compaction path coverage, energy efficiency and compaction risk coefficient through multi-objective constraints. The crossover and mutation unit uses adaptive crossover probability and dynamic mutation probability to iteratively optimize the chromosome sequence, where the crossover probability is dynamically adjusted according to the population diversity, and the mutation probability is negatively correlated with the surface curvature parameter of the coal pile. The optimized population extracts the optimal solution set through the Pareto front screening strategy, generates the global compaction strategy model parameters, and sends them to each compaction device through an encrypted communication link.

[0025] The compaction decision module integrates the compaction strength parameters output by the global compaction strategy model with the prediction results of the coal quality compaction risk assessment model. The compaction risk assessment model analyzes the historical compaction case library through a machine learning algorithm to establish a mapping relationship between the moisture gradient and the probability of coal compaction. The multi-objective optimization unit within the module uses the multi-objective particle swarm algorithm to simultaneously optimize the vibration frequency, compaction strength and travel speed parameters based on the compaction risk level as a constraint condition to generate a Pareto front solution set. The instruction fusion unit superimposes the vibration parameters in the solution set with the real-time temperature distribution data collected by the infrared thermal imaging module to generate a composite control instruction set containing vibration frequency gradient and pressure gradient values.

[0026] The path planning module constructs an elevation grid map of the coal pile surface based on LiDAR scan data. This grid map converts discrete point cloud data into a continuous elevation surface using a spatial interpolation algorithm. The module utilizes the elite retention strategy within the genetic algorithm to select the path segments with the highest coverage within each generation of the path population. It also dynamically adjusts node traversal weights based on the global void ratio heat map generated by federated learning. Specifically, high-void ratio areas are marked as high-priority path nodes in the grid map. During the iterative optimization process, a two-point crossover strategy is used to exchange the topology of path segments between adjacent generations to generate an optimal compaction path sequence that meets the void ratio threshold.

[0027] The execution control module drives the travel mechanism to move along the track according to the path sequence output by the path planning module. The travel mechanism uses a servo motor and an encoder to form a closed-loop control system to achieve millimeter-level positioning accuracy. The vibration mode of the compaction roller is adjusted in amplitude and frequency by a piezoelectric ceramic driver, and the deflection angle of the universal joint is controlled by the contact angle parameters optimized by the genetic algorithm. When the coal particle hardness sensor detects that the hardness of a local area exceeds the standard, the mode switching unit triggers the deformation mechanism of the rigid raised structure on the surface of the compaction roller and switches to the elastic contact mode to reduce the risk of compaction. The abnormal response unit monitors the temperature rise rate of the coal pile in real time through the infrared thermal imaging module. When a local oxidation hotspot is detected, the pre-trained genetic algorithm emergency instruction set is called to coordinate the timing of compaction and sealant spraying operations to form a continuous oxidation inhibition layer.

[0028] Specifically, in the coal pile compacting device of the present invention, the federated learning aggregation module includes: a parameter encoding unit for mapping the local model weights of each compacting device into a binary chromosome encoding sequence that can be processed by the fitness function; a crossover and mutation unit, configured to perform iterative operations of adaptive crossover probability and dynamic mutation probability on the binary chromosome coding sequence to generate an optimized population that satisfies multi-objective constraints; The strategy generation unit is used to extract the Pareto optimal solution set from the optimized population, generate global compaction strategy model parameters through key decryption, and feed back the updated local model parameters to each compaction device.

[0029] The federated learning aggregation module achieves distributed parameter fusion through multi-stage optimization. The parameter encoding unit receives the local model weights of each compaction device and converts the weight matrix into a binary chromosome encoding sequence. Each encoding sequence corresponds to a set of compaction intensity, vibration frequency, and path coverage parameter combinations. The Gray code conversion rule is used in the encoding process to minimize the encoding differences between adjacent parameter values, reducing the mutation risk of subsequent crossover mutation operations. The fitness function constructs a three-dimensional evaluation space based on compaction path coverage, unit energy efficiency, and compaction risk coefficient, and assigns a multidimensional fitness score to each chromosome sequence.

[0030] The crossover and mutation unit iteratively optimizes the binary chromosome coding sequence. The adaptive crossover probability is dynamically adjusted based on the population diversity index. When the similarity of individuals in the population exceeds a threshold, the crossover probability is increased to increase diversity. The dynamic mutation probability is negatively correlated with the surface curvature parameter of the coal pile, and the mutation probability of the corresponding gene position in the surface curvature mutation area increases. During each iteration, the roulette wheel selection mechanism selects the top 20% of individuals with the best fitness score to enter the next generation population. At the same time, an elite retention strategy is introduced to prevent the loss of high-quality solutions. The optimized population must meet the multi-objective constraints of a compaction path coverage rate of at least 95%, a unit energy efficiency higher than the baseline value, and a compaction risk coefficient lower than the critical value.

[0031] The strategy generation unit uses a non-dominated sorting algorithm to extract a Pareto-optimal solution set from the optimized population. This solution set contains optimized combinations of vibration frequency, compaction intensity, and travel speed parameters. The solution set is decrypted using an asymmetric encryption algorithm to generate executable parameters for the global compaction strategy model. The decrypted parameters are distributed to each compaction device via a secure communication protocol, triggering parameter updates in the local decision model. A parameter version control mechanism is established during the feedback process to record the update history of each device's model parameters, providing data traceability support for subsequent federated learning iterations. The difference between the global model parameters and the local parameters serves as the initial weight for the next round of parameter encoding, forming a closed-loop optimization chain.

[0032] Specifically, the coal pile compacting device of the present invention includes a path planning module comprising: a topology analysis unit for identifying concave areas and slope mutation nodes on the coal pile surface based on laser radar scanning data, and constructing a path taboo search space containing elevation constraints; an iterative optimization unit, configured to use the path taboo search space as an initial solution set, adopt an elite retention strategy in a genetic algorithm to screen path segments with the highest coverage, and generate an optimal path sequence that meets a void ratio threshold; The collaborative correction unit is used to receive the global void ratio heat map sent by the federated learning aggregation module, and dynamically adjust the node traversal weight of the optimal path sequence according to the coordinates of the high void ratio area in the heat map.

[0033] The path planning module achieves intelligent planning of the coal pile compaction path through multi-stage optimization. The topology analysis unit receives LiDAR scanning data and generates an elevation raster map of the coal pile surface through point cloud registration and denoising. A morphological opening algorithm is used to identify surface depressions, and combined with slope gradient calculations to locate mutation nodes, areas where the elevation change rate exceeds a threshold are marked as path taboo regions. The taboo search space is constructed based on the Delaunay triangulation algorithm, which divides the coal pile surface into several triangular mesh cells. Elevation constraints are imposed on each cell to generate an initial set of feasible paths.

[0034] The iterative optimization unit uses feasible paths in the tabu search space as the initial population and performs multi-generational optimization using a genetic algorithm. A two-point crossover strategy randomly selects segments of two parent paths and exchanges them for topological matching, generating a sequence of child paths. An elite retention strategy selects the top 10% of individuals in each generation for path coverage and retains them directly to the next generation, preventing the loss of high-quality solutions. The fitness function calculates a comprehensive score based on the void fraction distribution of the path traversed area. Iterations terminate when the average void fraction of the path sequence falls below a preset threshold, outputting the optimal path sequence.

[0035] The collaborative correction unit receives the global void ratio heat map from the federated learning aggregation module. This heat map generates a three-dimensional void ratio distribution model of the coal pile through multi-device data fusion. The correction unit weightedly integrates the coordinate information of high-void ratio areas in the heat map with the optimal path sequence to construct a dynamic priority adjustment matrix. For areas where the void ratio exceeds a critical value, a path fragment insertion strategy is implemented, replacing the corresponding nodes of the current path with the elite path fragments of the corresponding area in the historical optimized path library. During the traversal weight adjustment process, a positive feedback mechanism is established between path coverage and void ratio improvement rate, and the access priority ranking table of path nodes is updated in real time.

[0036] Specifically, the coal pile compaction device of the present invention comprises a compaction decision module comprising: a compaction prediction unit for inputting moisture distribution data into a pre-trained coal quality compaction risk assessment model, optimizing the feature weight parameters of the model through a genetic algorithm, and outputting a predicted value of the compaction risk level; A multi-objective optimization unit is used to use the predicted value of the compaction risk level as a constraint condition, adopt a multi-objective particle swarm algorithm to simultaneously optimize the compaction strength, energy efficiency and void ratio indicators, and generate a Pareto front solution set including vibration parameters; The instruction fusion unit is used to fuse the vibration parameters in the Pareto front solution set with the real-time temperature gradient data collected by the infrared thermal imaging module to generate a composite control instruction including vibration frequency and pressure gradient value.

[0037] The compaction decision module achieves intelligent decision-making on compaction parameters through multi-source data fusion. The compaction prediction unit receives moisture distribution data collected by a multispectral humidity sensor and inputs this data into a pre-trained coal compaction risk assessment model for feature extraction. This model is built on a convolutional neural network architecture and trained by mapping the relationship between moisture gradients and compaction incidence rates in a historical compaction case library. A genetic algorithm optimizes the weight parameters of the model's fully connected layer and employs a tournament selection strategy to select individuals with the highest fitness scores. The fitness function integrates compaction prediction accuracy with model complexity indicators to output classification results with three levels of risk: low, medium, and high.

[0038] The multi-objective optimization unit establishes a decision space based on the compaction risk level as a constraint. When the risk level reaches the medium risk threshold, the upper limit constraint on compaction strength is activated. A multi-objective particle swarm algorithm is used to simultaneously optimize the vibration frequency, travel speed, and pressure gradient parameters. The particle position update formula introduces a dynamic inertia weight mechanism, and the weight value decreases linearly with the number of iterations. In each iteration, the algorithm calculates the Pareto frontier of compaction operation coverage, unit energy consumption, and void ratio improvement rate, and selects the solution set with the highest comprehensive score as the candidate solution. The parameter combinations in the solution set are sorted using the weighted sum method, and the weight distribution is dynamically adjusted according to the current operation stage. Initial operations prioritize improving coverage, while final operations focus on optimizing void ratio.

[0039] The command fusion unit performs spatiotemporal registration between the vibration parameters output by the Pareto solution and the real-time temperature field data from the infrared thermal imaging module. A nonuniformity correction algorithm is used to eliminate ambient noise from the thermal imaging data and extract the temperature gradient distribution characteristics of the coal pile surface. The vibration frequency parameters are dynamically compensated based on the temperature gradient, and a frequency attenuation strategy is implemented in high-temperature areas to prevent localized overheating. The pressure gradient value is coupled with the temperature data using a fuzzy control algorithm to establish a temperature-pressure feedback regulation mechanism. The resulting composite control command contains a three-dimensional control matrix for the vibration motor drive frequency, the hydraulic system pressure valve opening, and the travel mechanism speed parameters. This command set is transmitted to the actuator via the CAN bus protocol.

[0040] Specifically, the coal pile compacting device of the present invention includes: an angle adaptation unit for calculating a real-time deflection angle control value of the universal joint based on the coal pile slope data acquired by the laser scanning module and the contact angle optimization parameters generated by the genetic algorithm; A mode switching unit is used to receive a detection signal from a coal particle hardness sensor and trigger a deformation drive instruction for the rigid protrusion structure on the surface of the compaction drum when the hardness value exceeds a preset threshold; The abnormal response unit is used to call the compaction and spraying collaborative operation instruction set generated by the genetic algorithm in the pre-stored emergency action sequence library when the infrared thermal imaging module detects that the local temperature rise rate exceeds the safety threshold.

[0041] The execution control module achieves precise control of the compaction operation through multi-dimensional parameter adaptation. The angle adaptation unit receives the coal pile surface slope data collected by the laser scanning module and calculates the slope gradient value of each area through a triangular meshing algorithm. Combined with the contact angle parameter library optimized by the genetic algorithm, fuzzy control rules are used to match the current slope gradient with the historical optimal contact angle parameters to generate the three-dimensional deflection angle control value of the universal joint. The control value is decomposed into pitch angle, roll angle and yaw angle components through the quaternion conversion algorithm, driving the three sets of servo motors inside the ball joint to operate in coordination, so that the compaction roller maintains a preset contact angle with the coal pile surface.

[0042] The mode switching unit uses the piezoelectric sensor array of the coal particle hardness sensor to collect real-time hardness distribution data in the contact area, employing a sliding window mean filtering algorithm to eliminate transient noise interference. When the hardness value of the local area exceeds a preset threshold for three consecutive sampling periods, a trigger signal conversion circuit sends a pulse command to the deformation drive mechanism. The drive mechanism, constructed of shape memory alloy, receives this command and, upon receiving it, heats the surface of the compaction drum with an electric current, triggering a lattice structure phase transition. This causes the rigid protrusions on the surface of the compaction drum to contract, exposing the elastic matrix layer. This creates a flexible contact mode to prevent coal compaction.

[0043] The abnormal response unit uses an infrared thermal imaging module to obtain surface temperature distribution data for the coal pile and calculates the temperature rise rate in each area using a time series analysis method. If the temperature rise rate in a local area exceeds the safety threshold for two consecutive minutes, the optimized instruction set generated by the genetic algorithm in the pre-stored emergency action sequence library is called. The instruction set includes the attenuation gradient of the compaction drum vibration frequency, the start and stop timing of the sealant spray unit, and the reverse movement path parameters of the travel mechanism. A multi-threaded control mechanism coordinates the execution timing of the compaction and spraying operations. The sealant spraying uses an array of high-pressure atomizing nozzles to form an insulating layer covering the area behind the compaction drum that is synchronized with the vibration trajectory.

[0044] See also Figures 2 to 4 The device structure of the technical solution of the present invention is as follows: a compacting unit 1, a suspension steel beam 2, a control system 3, a traveling mechanism 4, a track 5, a coal pile 6, a steel wire rope 7 and a universal joint 8.

[0045] The specific embodiment of the present invention relates to the structural configuration and operating process of the coal pile compaction device. The main body of the device consists of a compaction unit, a suspension steel beam, a traveling mechanism and a control system, wherein the compaction unit is suspended below the suspension steel beam via a universal joint connected to a steel wire rope, and a double traveling mechanism is configured at the bottom of the suspension steel beam to move along the track. The surface of the compaction unit is provided with an array of spherical protrusions with a height of 5-8mm. The spacing distribution conforms to the coal particle grading characteristics, and generates multi-directional extrusion pressure when contacting the coal pile. The traveling mechanism adopts a servo drive system with a positioning accuracy of ±2mm. The track surface is paved with an anti-slip texture structure with a friction coefficient of not less than 0.6.

[0046] During the operation, the compaction process is started when the height of the coal pile reaches 3 meters. The control system divides the compaction area according to the three-dimensional model generated by the lidar scan, and the traveling mechanism drives the suspended steel beam to move along the track at a speed of 0.5m / s. The compaction unit contacts the surface of the coal pile under the action of gravity, and the spherical protrusions exert a combined effect of vertical pressure and horizontal shear force on the coal body. The initial compaction strength is set to 50kPa. After completing the forward compaction, the traveling mechanism moves in the opposite direction to implement secondary compaction, and the overlap rate of the two compaction trajectories is controlled in the range of 30%-40%. After each layer of compaction operation is completed, the coal stacking equipment adds a 0.3-meter-thick coal layer, and cyclically implements layered compaction until the coal pile reaches the designed height.

[0047] If localized heating areas appear on the surface of the coal pile and the infrared thermal imaging module detects an abnormal temperature rise exceeding 5°C / h, the control system initiates emergency response procedures. The sealant spray unit uses a high-pressure atomizing nozzle to spray a flame-retardant sealant with a particle size of 50-100μm at a pressure of 0.2MPa and a spray volume of 2L / m² into the target area. The synchronously controlled compaction unit performs three alternating compaction operations on the area, increasing the vibration frequency to 35Hz and the compaction force to 80kPa. The timing of the compaction and spraying operations is optimized using a genetic algorithm to create an alternating operation rhythm with five-second intervals.

[0048] The data acquisition module captures real-time data on pressure distribution, temperature gradients, and coal displacement during compaction, transmitting it to the federated learning aggregation module via industrial Ethernet. The LiDAR updates the three-dimensional model of the coal pile every five minutes, achieving a porosity detection accuracy of ±0.5%. The abnormal response unit stores ten standard emergency response plans. When detection parameters exceed preset thresholds, it automatically selects the optimal response plan and generates execution instructions. Operational data is encrypted and stored within blockchain nodes, forming an unalterable record containing timestamps, equipment IDs, and operating parameters, providing traceability support for subsequent parameter optimization.

[0049] Second, see Figure 1 The present invention provides a coal pile compacting method, which is applied to the above-mentioned coal pile compacting device, comprising: Step S101, collecting humidity gradient data, three-dimensional laser point cloud data, and real-time pressure feedback data of the coal pile through multi-source sensors to construct a three-dimensional dynamic model of the coal pile; Step S102: inputting the three-dimensional dynamic model into the local decision-making models of multiple compacting equipment, performing multi-objective optimization selection on the coal quality characteristic parameters in the model using a genetic algorithm, and generating a chromosome coding sequence under the constraints of a fitness function; Step S103, performing iterative calculation of adaptive crossover probability and mutation operator on the chromosome coding sequence under the federated learning framework to generate an optimized parameter set of the global compaction strategy model; Step S104: extracting the coal pile surface topology features based on the three-dimensional laser point cloud data, and optimizing the traversal priority of the compaction path nodes using a genetic algorithm according to the void ratio distribution parameters in the global compaction strategy model; Step S105 , according to the compaction strength control parameters output by the global compaction strategy model, the traveling mechanism is driven to move along the planned path, and the vibration frequency of the compaction roller and the angular offset of the universal joint are synchronously adjusted.

[0050] The coal pile compaction method of the present invention achieves intelligent compaction through multi-stage data fusion and optimization. The multi-source sensor system consists of a laser radar scanning unit, an array pressure sensor, and an infrared moisture detector. The laser radar scans the coal pile surface at a 0.1° angular resolution to generate three-dimensional point cloud data. The pressure sensor acquires the dynamic pressure distribution in the compaction roller contact area at a 50Hz sampling frequency. The moisture detector acquires coal moisture gradient data through near-infrared spectroscopy. The data preprocessing unit uses a spatiotemporal calibration algorithm to fuse the multimodal data and constructs a three-dimensional dynamic model of the coal pile through Delaunay triangulation, achieving a resolution of 5mm.

[0051] The local decision-making model utilizes a deep neural network architecture. The input layer receives spatial feature data from a three-dimensional dynamic model, and the hidden layer consists of three fully connected layers that extract coal hardness, particle size distribution, and moisture distribution characteristics. A genetic algorithm optimizes the model's feature weight parameters. A fitness function constructs a three-dimensional evaluation space based on compaction efficiency, energy consumption, and compaction risk. A tournament selection strategy selects the top 15% of individuals with the highest fitness scores to generate the chromosome coding sequence. The coding sequence uses a mixed binary and real number encoding scheme, with the compaction strength parameter encoded in 8-bit binary and the vibration frequency parameter encoded in real numbers.

[0052] The federated learning aggregation center deploys a model parameter optimization algorithm. After receiving chromosome coding sequences uploaded by each compaction device, it uses a dynamic crossover probability mechanism to implement parameter fusion. The crossover probability is dynamically adjusted based on the Hamming distance of the coding sequences, triggering a high-probability crossover operation when population similarity exceeds 70%. The mutation operator incorporates the curvature parameter of the coal pile surface as a guiding factor for mutation direction, performing targeted mutation at the gene loci corresponding to the curvature mutation area. After 20-30 iterations, a globally optimized parameter set is generated, which is hashed and distributed to each local device.

[0053] The path planning engine extracts topological features of the coal pile surface from 3D point cloud data and employs a morphological gradient algorithm to identify depressed areas and nodes with sudden slope changes. The void distribution parameters output by the global compaction strategy model are combined with topological features in a matrix fusion to generate a path node priority score matrix. The genetic algorithm optimization process employs an elite retention strategy, retaining the five path solutions with the highest coverage in each generation and generating a new generation of path populations through a two-point crossover operation. The path node traversal order is dynamically adjusted based on priority scores, increasing the access weight of nodes in high-voidity areas by 30%-50%.

[0054] After the control system receives the compaction intensity control parameters, the travel mechanism servo driver generates pulse control signals based on the path planning coordinate sequence. The movement speed is dynamically adjusted based on the compaction intensity parameters within a range of 0.2-1.5 m / s. The vibration frequency of the compaction roller is adjusted via a piezoelectric ceramic driver. The control signal is positively correlated with the pressure gradient, achieving a frequency adjustment accuracy of ±0.5 Hz. The universal joint angle control system utilizes a quaternion solution algorithm to convert the contact angle parameters optimized by the genetic algorithm into three-axis rotation angles. These drive three orthogonally arranged stepper motors to achieve a ±15° deflection range with an angle control accuracy of ±0.1°.

[0055] Specifically, the coal pile compaction method of the present invention comprises generating a chromosome population under the guidance of a fitness function, comprising: constructing a three-dimensional fitness evaluation space based on a risk level value output by a coal compaction risk assessment model, coal pile volume parameters extracted from laser radar scanning data, and compaction energy consumption history data; The chromosome coding sequence is screened by multi-objectives using a non-dominated sorting algorithm, and the path planning solution with the highest comprehensive score in coverage and energy efficiency is retained in the Pareto front solution set. According to the surface curvature parameters in the three-dimensional dynamic model of coal pile, the action probability distribution of the genetic algorithm mutation operator is dynamically adjusted.

[0056] In the coal pile compaction method described in the present invention, the generation process of the chromosome population realizes optimized decision-making through multi-dimensional parameter fusion. The coal quality compaction risk assessment model receives moisture distribution data and a historical compaction case library, uses a random forest algorithm to analyze the correlation characteristics between moisture gradient and compaction probability, and outputs classification results including three levels of low risk, medium risk, and high risk. The laser radar scanning unit acquires point cloud data on the surface of the coal pile at a scanning interval of 0.5°, calculates the volume parameters of the coal pile through voxel rasterization processing, and the data preprocessing unit uses a sliding average filter to eliminate noise interference. The compaction energy consumption history data is extracted from the equipment operation log, including vibration motor power consumption, walking mechanism energy consumption and sealant spraying parameters, and is matched with the current operation parameters through a time series alignment algorithm.

[0057] The three-dimensional fitness evaluation space is constructed using the Euclidean distance metric, mapping the risk level to the X-axis, the coal pile volume parameter to the Y-axis, and the compaction energy consumption parameter to the Z-axis. Each chromosome coding sequence corresponds to a coordinate point in the space, and the fitness scoring function calculates a comprehensive evaluation value based on the distance between the coordinate point and the ideal solution. The risk level weight coefficient is set to 0.6, the volume parameter weight is set to 0.3, and the energy consumption parameter weight is set to 0.1, reflecting the priority of compaction risk control.

[0058] When the non-dominated sorting algorithm processes a chromosome population, the top 30% of individuals are retained in the initial screening process and are included in the Pareto front candidate set. Each individual's score in terms of path coverage and energy efficiency is calculated by weighted summation, with a coverage weight of 0.7 and an energy efficiency weight of 0.3. Within the candidate set, individuals are sorted in descending order by their overall score, and the top 10% are selected as the elite solution set. Path planning solutions are evaluated using metrics including node traversal rate, path repetition coefficient, and cumulative turn angles. Solutions with an overall score exceeding 85 are included in the final solution set.

[0059] The three-dimensional dynamic model of the coal pile is constructed using the Delaunay triangulation algorithm to construct a surface mesh model, calculating the curvature parameters of each triangular facet. The surface curvature parameters are categorized into three types: low-curvature flat areas, medium-curvature transition areas, and high-curvature mutation areas. The probability of the mutation operator is set to 0.05 in low-curvature areas, increasing to 0.1 in medium-curvature areas, and reaching 0.15 in high-curvature areas. When the fitness improvement rate of the population falls below 2% for two consecutive generations, a dynamic adjustment mechanism is triggered, increasing the mutation probability of the corresponding gene position in the high-curvature area by a gradient of 0.03. This probability distribution adjustment information is synchronized to each compaction equipment through a federated learning framework to maintain consistency in the collaborative optimization of multiple equipment.

[0060] Specifically, in the coal pile compaction method of the present invention, the iterative optimization compaction path includes: An elevation grid map containing void ratio distribution characteristics is constructed based on LiDAR scanning data, and areas with void ratios above a preset threshold are marked as high-priority path nodes. The two-point crossover strategy in the genetic algorithm is used to perform topological matching and exchange on the path segments of adjacent generations to generate an optimized path sequence. According to the regional weight values ​​in the global porosity heat map issued by the federated learning aggregation module, the elite path fragments with a coverage rate of more than 95% in the historical optimization path library are inserted into the current path sequence.

[0061] The iterative path optimization process for the coal pile compaction method is achieved through multi-source data fusion and intelligent algorithms. A LiDAR scanning unit acquires point cloud data from the coal pile surface at a sampling rate of 200,000 points per second. The preprocessing module uses a statistical filtering algorithm to remove outlier noise points and generates a 0.1-meter-resolution elevation grid map using Kriging interpolation. The void fraction calculation module, based on a volume density inversion algorithm, converts the point cloud density values ​​within a grid cell into void fraction distribution parameters. Cells with void fractions exceeding 15% are marked as high-priority path nodes, and the node coordinates are stored in a spatial database.

[0062] When the genetic algorithm optimization engine initializes the path population, it extracts five elite paths from the historically optimized path library as initial solutions. A two-point crossover strategy selects two segments of the parent path with similar topology and swaps their gene sequences at random breakpoints. The swapped offspring paths are validated and added to the new population. The topology matching process uses the Hausdorff distance to measure the similarity of path segments, with a threshold of 0.5 meters. Segments exceeding this threshold trigger mutation. Each generation retains the ten paths with the highest fitness scores. The fitness function integrates path coverage, cumulative turning radius, and energy consumption.

[0063] The federated learning aggregation module periodically collects void distribution data from each compaction device and generates a global void heat map using the Gaussian process regression algorithm. The regional weight values ​​in the heat map are divided into five levels according to the void size, and the weight coefficient of the highest level area is set to 0.8. When the path optimizer accesses the historical path library, it uses a fuzzy matching algorithm to retrieve historical cases with a volume similarity of more than 90% with the current coal pile, and extracts path segments with a coverage rate of more than 95%. The segment insertion operation follows the principle of spatiotemporal continuity. The Euclidean distance between the starting point of the newly inserted segment and the end point of the current path is controlled within 0.3 meters. The path smoothing module uses the B-spline curve algorithm to eliminate angular mutations at turning points.

[0064] The optimized path sequence is converted into control instructions for the travel mechanism by the motion control module. The coordinates of the path nodes are interpolated using cubic spline interpolation to generate a continuous trajectory. The dwell time of the compaction roller at the path nodes is dynamically adjusted based on the local void fraction, with dwell time extended by 30%-50% in high-void fraction areas. The control system monitors path execution deviations in real time. When the positioning error exceeds 0.2 meters, a trajectory correction program is triggered, using the extended Kalman filter algorithm to estimate the optimal compensation path.

[0065] Specifically, in the coal pile compaction method of the present invention, adjusting the vibration amplitude of the compaction roller includes: Based on the risk level value output by the coal compaction risk assessment model, the spatial boundaries of the high compaction strength zone and the vibration avoidance zone are divided; A low-frequency, high-amplitude vibration mode is used in high compaction intensity areas, and the amplitude parameters are dynamically adjusted based on compaction feedback data; In the vibration avoidance area, it switches to a high-frequency, micro-amplitude vibration mode and is synchronously triggered by the sealant spraying unit to form a continuous insulation layer on the surface of the coal pile.

[0066] Vibration amplitude adjustment during coal pile compaction is achieved through multi-source data fusion and closed-loop control. The coal compaction risk assessment model receives moisture gradient data from a multispectral humidity sensor, analyzes historical compaction case characteristics using a random forest algorithm, and outputs a spatial distribution map containing risk levels. When the risk level reaches the medium risk threshold, the spatial partitioning engine initiates boundary calculation, using the Delaunay triangulation algorithm to divide the coal pile surface into a high compaction intensity zone and a vibration avoidance zone. Morphological dilation is used to create a 5-cm-wide transition zone along the boundary.

[0067] Within the high compaction intensity zone, the vibration control system activates a low-frequency, high-amplitude mode, with a vibration frequency set in the 10-15 Hz range. The initial amplitude parameters are proportional to the thickness of the coal layer, with an amplitude increment of 1 mm for every 10 cm of thickness. A piezoelectric sensor array collects real-time stress distribution data across the contact area and dynamically adjusts the amplitude using fuzzy control rules. When the local pressure feedback exceeds a set threshold, the amplitude decreases at a rate of 0.2 mm / s. A vibration energy controller simultaneously monitors motor current fluctuations, triggering an amplitude reset mechanism when the current fluctuation exceeds 15%.

[0068] In the vibration avoidance zone, the vibration mode switches to a high-frequency, micro-amplitude state, the frequency is increased to 35-40Hz, and the amplitude is reduced to the range of 0.5-1mm. The sealant spraying unit is linked to the vibration mode switching signal, and the high-pressure atomizing nozzle array forms a fan-shaped spray area 20cm behind the vibrating roller. The particle size of the atomized particles is controlled in the range of 50-80μm. The spraying start and stop sequence is optimized by genetic algorithms, and a 0.5-second spray pulse is triggered at the trough stage of the drum vibration cycle to form an insulation layer coverage pattern that matches the vibration trajectory. The insulation layer thickness detection module uses laser triangulation measurement for real-time monitoring, and automatically increases the spraying pressure by 10% when the detected thickness is lower than 0.3mm.

[0069] The data fusion center integrates vibration parameters, pressure feedback, and insulation layer thickness data to generate a three-dimensional control matrix. This matrix data is then timestamped and fed into the federated learning framework to update the parameter library for vibration mode switching rules. The exception handling module monitors the vibration motor's temperature rise rate. When the temperature rise exceeds 5°C / minute, it automatically switches to a safe vibration mode and triggers a fault diagnosis protocol. Historical operation data is encrypted and stored in blockchain nodes, forming a traceable database containing amplitude adjustment records, zoning parameters, and insulation layer quality indicators.

[0070] Specifically, the coal pile compaction method of the present invention further includes: when the infrared thermal imaging module detects that the local temperature gradient exceeds a safety threshold, calling a pre-trained genetic algorithm emergency response model to generate a coordinated operation instruction set for the compaction roller and the sealant spraying unit; After each compaction operation is completed, the equipment operation log containing vibration parameters, void ratio distribution and energy consumption data will be encrypted and uploaded to the blockchain node to generate a time-stamped compaction quality traceability record; The newly collected 3D point cloud data is fused with the elite chromosome coding sequences in the historical optimization parameter library through the federated learning framework to update the initial population generation rules of the genetic algorithm.

[0071] Exception handling and data management for the coal pile compaction method achieve closed-loop control through multi-system collaboration. The infrared thermal imaging module utilizes a 640×480 resolution sensor array, acquiring surface temperature data at a sampling rate of 30 frames per second. The temperature gradient calculation unit uses a spatial difference algorithm to identify areas of abnormal temperature rise. A level 3 warning signal is triggered when a temperature rise rate exceeding 3°C / min for five consecutive minutes is detected. A pre-trained genetic algorithm emergency response model loads a historical emergency case database, matches current temperature distribution characteristics with coal pile geometry, and generates a coordinated instruction set containing parameters for adjusting the vibration frequency attenuation gradient, sealant spray volume, and travel path. This instruction set, using a timestamp synchronization mechanism, controls the compaction roller to perform three alternating compactions in the target area, with each compaction interval lasting five seconds. This simultaneously triggers the high-pressure atomization spray unit to spray 60μm flame retardant particles at a pressure of 0.3MPa, forming a continuous insulation layer with a thickness of 0.5mm.

[0072] After the compaction operation is completed, the data acquisition module extracts the frequency spectrum characteristics of the vibration motor, the peak distribution of the pressure sensor, and the void ratio detection results of the lidar, generating a structured operation log. The log data is processed using an asymmetric encryption algorithm, with the private key stored in the hardware security module. The encrypted data packets are uploaded to the consortium chain nodes via the IPFS protocol. The blockchain consensus mechanism uses the PBFT algorithm to achieve data consistency confirmation among the four verification nodes. The timestamp server uses the Beidou satellite timing system to generate millisecond-accurate time stamps. The generated traceability record includes a data hash value, device ID, and operating condition snapshot. The storage structure uses a Merkle tree organization to support rapid traceability verification.

[0073] The federated learning framework deploys a parameter update engine, and newly collected 3D point cloud data is used to generate a 128-dimensional feature vector through a feature extraction network. Elite chromosome coding sequences in the historical optimization parameter library are screened using a similarity matching algorithm, and historical codes with a cosine similarity exceeding 0.85 are selected for fusion. The fusion process uses a weighted average strategy, with the new data weight coefficient set to 0.7 and the historical data coefficient to 0.3, to generate an updated chromosome coding rule library. The initial population generator introduces an adaptive diversity control mechanism. When the population's genetic diversity falls below 15%, random mutant individuals are automatically injected. The version control system records each rule update log and supports rollback to the three most recent historical versions to maintain the stability of the algorithm iteration. The updated genetic algorithm parameters are distributed to each compaction device via a secure channel, completing the federated learning closed loop.

[0074] The present invention effectively solves the key problems in coal pile compaction technology through the fusion of multi-module collaboration and intelligent algorithms. The device integrates a federated learning aggregation module and a genetic algorithm optimization mechanism, and generates a global compaction strategy model by aggregating the local parameters of multiple compaction equipment, thereby overcoming the shortcomings of insufficient control capabilities of traditional single equipment. The path planning module constructs a three-dimensional dynamic model of the coal pile based on lidar scanning, and uses taboo search space and elite path insertion strategy to optimize the traversal order, so that high-porosity areas are compacted first, thereby improving the operation coverage rate of the top and complex areas. The execution control module realizes dynamic adjustment of compaction strength and contact angle through universal joint angle adaptation and vibration mode switching, combined with the coordinated operation of the sealant spraying unit, to avoid excessive compaction in high-humidity areas causing compaction.

[0075] To address the need to suppress oxidation reactions in coal piles, the compaction decision module integrates infrared thermal imaging data with a compaction risk assessment model to establish a temperature-pressure feedback control mechanism. When a localized temperature rise anomaly is detected, an emergency instruction set generated by a genetic algorithm is invoked to coordinate the vibration attenuation of the compaction roller with the timing of sealant spraying to form a continuous insulation layer in the heated area. This invention utilizes a multi-objective particle swarm algorithm to simultaneously optimize compaction strength, energy efficiency, and void ratio indicators. Control parameters that balance safety and efficiency are selected from the Pareto front solution set to suppress oxidation reactions while maintaining the stability of the pile structure.

[0076] A closed-loop data management mechanism further enhances system adaptability. Blockchain technology enables encrypted storage and traceability of compaction quality data. A federated learning framework integrates historical optimization parameters with newly acquired 3D point cloud data, dynamically updating the initial population rules of the genetic algorithm. By iteratively optimizing the matching relationship between the traversal weights of compaction path nodes and vibration parameters, the system adaptively adjusts operating modes based on coal quality characteristics and stockpile geometry. This addresses the lack of material adaptability associated with traditional technologies' reliance on fixed operating parameters, significantly improving compaction uniformity and stockpile safety.

Claims

1. A coal pile compacting device, characterized in that: include: Data acquisition module, used to obtain moisture distribution data, three-dimensional profile data and compaction feedback data of the coal pile in real time through multi-source sensors; A federated learning aggregation module is used to receive local optimization parameters of multiple compaction devices, perform multi-objective cross-mutation operations on the local optimization parameters based on a genetic algorithm, and generate a global compaction strategy model; A compaction decision module is used to fuse the output data of the global compaction strategy model and the coal compaction risk assessment model to generate a control instruction set including vibration frequency and compaction intensity parameters; a path planning module for constructing an elevation grid map based on the three-dimensional laser scanning data of the coal pile and iteratively optimizing the traversal order of path nodes using a genetic algorithm based on the void ratio distribution characteristics in the global compaction strategy model; The execution control module is used to drive the walking mechanism to move along the track and adjust the vibration mode and contact angle of the compaction roller according to the compaction unit force control instruction.

2. The coal pile compacting device according to claim 1, characterized in that: The federated learning aggregation module includes: a parameter encoding unit for mapping the local model weights of each compaction device into a binary chromosome encoding sequence that can be processed by the fitness function; a crossover and mutation unit, configured to perform iterative operations of adaptive crossover probability and dynamic mutation probability on the binary chromosome coding sequence to generate an optimized population that satisfies multi-objective constraints; The strategy generation unit is used to extract the Pareto optimal solution set from the optimized population, generate global compaction strategy model parameters through key decryption, and feed back the updated local model parameters to each compaction device.

3. The coal pile compacting device according to claim 1, characterized in that: The path planning module includes: a topology analysis unit for identifying concave areas and slope mutation nodes on the coal pile surface based on laser radar scanning data, and constructing a path taboo search space containing elevation constraints; an iterative optimization unit, configured to use the path taboo search space as an initial solution set, adopt an elite retention strategy in a genetic algorithm to screen path segments with the highest coverage, and generate an optimal path sequence that meets a void ratio threshold; The collaborative correction unit is used to receive the global void ratio heat map sent by the federated learning aggregation module, and dynamically adjust the node traversal weight of the optimal path sequence according to the coordinates of the high void ratio area in the heat map.

4. The coal pile compacting device according to claim 1, characterized in that: The compaction decision module includes: a compaction prediction unit for inputting moisture distribution data into a pre-trained coal compaction risk assessment model, optimizing the characteristic weight parameters of the model through a genetic algorithm, and outputting a compaction risk level prediction value; A multi-objective optimization unit is used to use the predicted value of the compaction risk level as a constraint condition, adopt a multi-objective particle swarm algorithm to simultaneously optimize the compaction strength, energy efficiency and void ratio indicators, and generate a Pareto front solution set including vibration parameters; The instruction fusion unit is used to fuse the vibration parameters in the Pareto front solution set with the real-time temperature gradient data collected by the infrared thermal imaging module to generate a composite control instruction including vibration frequency and pressure gradient value.

5. The coal pile compacting device according to claim 1, characterized in that: The execution control module includes: an angle adaptation unit for calculating the real-time deflection angle control value of the universal joint based on the coal pile slope data obtained by the laser scanning module and the contact angle optimization parameters generated by the genetic algorithm; A mode switching unit is used to receive a detection signal from a coal particle hardness sensor and trigger a deformation drive instruction of the rigid protrusion structure on the surface of the compaction drum when the hardness value exceeds a preset threshold; The abnormal response unit is used to call the compaction and spraying collaborative operation instruction set generated by the genetic algorithm in the pre-stored emergency action sequence library when the infrared thermal imaging module detects that the local temperature rise rate exceeds the safety threshold.

6. A coal pile compacting method, applied to the coal pile compacting device according to any one of claims 1 to 5, characterized in that: include: Multi-source sensors collect coal pile moisture gradient data, 3D laser point cloud data, and real-time pressure feedback data to construct a 3D dynamic model of the coal pile. The three-dimensional dynamic model is input into the local decision-making model of multiple compacting equipment, and the coal quality characteristic parameters in the model are optimized by a genetic algorithm to generate a chromosome coding sequence under the constraint of a fitness function; Performing iterative calculations of adaptive crossover probability and mutation operators on the chromosome coding sequence under a federated learning framework to generate an optimized parameter set of a global compaction strategy model; Extracting the coal pile surface topology features based on the three-dimensional laser point cloud data, and optimizing the traversal priority of the compaction path nodes using a genetic algorithm according to the void ratio distribution parameters in the global compaction strategy model; According to the compaction strength control parameters output by the global compaction strategy model, the walking mechanism is driven to move along the planned path, and the vibration frequency of the compaction roller and the angular offset of the universal joint are synchronously adjusted.

7. The coal pile compaction method according to claim 6, characterized in that: Generating a chromosome population under the guidance of a fitness function includes: constructing a three-dimensional fitness evaluation space based on the risk level value output by a coal compaction risk assessment model, the coal pile volume parameter extracted from the laser radar scanning data, and the compaction energy consumption history data; The chromosome coding sequence is screened by multi-objectives through the non-dominated sorting algorithm, and the path planning solution with the highest comprehensive score of coverage and energy efficiency in the Pareto front solution set is retained; According to the surface curvature parameters in the three-dimensional dynamic model of coal pile, the action probability distribution of the genetic algorithm mutation operator is dynamically adjusted.

8. The coal pile compacting method according to claim 6, characterized in that: The iterative optimization compaction path includes: An elevation grid map containing void ratio distribution characteristics is constructed based on LiDAR scanning data, and areas with void ratios above a preset threshold are marked as high-priority path nodes. The two-point crossover strategy in the genetic algorithm is used to perform topological matching and exchange on the path segments of adjacent generations to generate an optimized path sequence. According to the regional weight values ​​in the global porosity heat map issued by the federated learning aggregation module, the elite path fragments with a coverage rate of more than 95% in the historical optimization path library are inserted into the current path sequence.

9. The coal pile compacting method according to claim 6, characterized in that: The adjusting of the vibration amplitude of the compacting roller comprises: Based on the risk level value output by the coal compaction risk assessment model, the spatial boundaries of the high compaction strength zone and the vibration avoidance zone are divided; A low-frequency, high-amplitude vibration mode is used in high compaction intensity areas, and the amplitude parameters are dynamically adjusted based on compaction feedback data; In the vibration avoidance area, it switches to a high-frequency, micro-amplitude vibration mode and is synchronously triggered by the sealant spraying unit to form a continuous insulation layer on the surface of the coal pile.

10. The coal pile compacting method according to claim 6, characterized in that: The method further includes: when the infrared thermal imaging module detects that the local temperature gradient exceeds a safety threshold, calling a pre-trained genetic algorithm emergency response model to generate a collaborative operation instruction set for the compaction roller and the sealant spraying unit; After each compaction operation is completed, the equipment operation log containing vibration parameters, void ratio distribution and energy consumption data will be encrypted and uploaded to the blockchain node to generate a time-stamped compaction quality traceability record; The newly collected 3D point cloud data is fused with the elite chromosome coding sequences in the historical optimization parameter library through the federated learning framework to update the initial population generation rules of the genetic algorithm.

Images