Digital Coal Yard Stacker-Reclaimer Operation Path Planning Method and System

By combining 3D laser scanning, microwave measurement, and historical records to construct a material feature grid, the coal yard path planning is optimized, solving the problem that the influence of material physicochemical characteristics was not considered in existing technologies, and realizing safe and efficient unattended operation.

CN123111271APending Publication Date: 2026-11-17HAIMEN POWER PLANT OF HUANENG (GUANGDONG) ENERGY DEV CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611141053.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-30
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing digital coal yard path planning schemes fail to fully consider the impact of the complex physical and chemical characteristics of the material on the terrain evolution, which makes it easy for stacker-reclaimers to cause micro-collapses or local collapses of the coal wall during operation, making it difficult to achieve fully automated unattended operation.

Method used

By acquiring point cloud data from a 3D laser scanner, moisture content values ​​from a microwave moisture meter, and compaction stress and particle size distribution from historical stockpiling records, an initial topological grid is constructed and a spatiotemporal assessment of collapse risk is conducted. A predictive evolution terrain grid is then generated, and path planning is optimized to avoid the risk of slippage and instability.

Benefits of technology

It has enabled improved stacker-reclaimer operating efficiency and trajectory adaptability in unattended operation, significantly reduced equipment overload frequency, and ensured production safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111271A_ABST
    Figure CN123111271A_ABST
Patent Text Reader

Abstract

This invention provides a method and system for planning the operation path of a digital coal yard stacker-reclaimer. By using the physicochemical characteristics of coal yard materials as dynamic constraints, and employing a novel paradigm of first pre-simulating the dynamic morphological evolution under cutting disturbances and then generating spatial paths, it achieves spatiotemporal coupling between operational instructions and collapse risks. Furthermore, by constructing a predictive evolution terrain grid and implementing four-dimensional dynamic mapping, it can effectively avoid the risk of slippage and instability during operation, significantly reduce the frequency of equipment overload, and thus, while ensuring production safety, greatly improve the operating efficiency and trajectory adaptability of the digital coal yard stacker-reclaimer in unattended operation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of intelligent planning technology, specifically to a method and system for planning the operation path of a digital coal yard stacker-reclaimer. Background Technology

[0002] With the deepening of the intelligent transformation of bulk cargo ports and power plants, digital coal yard management has become a core means to improve operational efficiency and ensure safe production. As the most critical production equipment within the coal yard, the operational performance of the stacker-reclaimer directly hinders or promotes the overall stability of material storage and circulation. In order to achieve fully automated unattended operation of the stacker-reclaimer, designing and constructing a digital operation path planning scheme that can respond to environmental characteristics in real time and achieve optimal cutting trajectories is of great practical significance for reducing energy consumption and avoiding potential equipment operation hazards.

[0003] Existing digital coal yard path planning schemes typically utilize 3D laser scanning technology to obtain the surface contour of the coal pile and determine the material reclaiming trajectory based on a pre-defined static geometry. However, in practical applications, these schemes often treat the coal pile as a rigid structure with constant physical properties and no deformation, failing to fully consider the impact of the complex physicochemical characteristics of the material's interior on terrain evolution. During actual operations, the powerful cutting action of the stacker-reclaimer's bucket wheel can instantly disrupt the original mechanical balance of the pile. Due to significant differences in moisture content, particle size, and compaction degree in different areas, the cutting behavior can easily trigger micro-collapses or even localized collapses of the coal wall near the working face. Simultaneously, because the hardware refresh rate of mainstream 3D scanning equipment is usually much slower than the physical dynamic process of a collapse, the system struggles to capture instantaneous changes in the pile's morphology in real time. This path planning, entirely based on static surface geometry, often fails during execution due to sudden changes in the pile's morphology, causing the stacker-reclaimer to plunge into a suddenly slipping coal mass, leading to mechanical jamming or power overload failure. This lack of perception of dynamic evolution and the absence of evolution prediction has become the core technical bottleneck restricting the improvement of the non-interventional continuous operation rate of fully automated stacking and reclaiming systems.

[0004] Therefore, an optimized method for planning the operation path of digital coal yard stacker-reclaimers is needed. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a method and system for planning the operation path of a digital coal yard stacker-reclaimer.

[0006] In a first aspect, embodiments of the present invention provide a method for planning the operation path of a digital coal yard stacker-reclaimer, comprising: Acquire the original laser point cloud coordinate set collected by the 3D laser scanner, the current moisture content value of the material pile output by the microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical material pile records; By voxelization downsampling and spatial coordinate registration, geometric mesh construction and physical property coupling mapping are performed on the original laser point cloud coordinate set to obtain the initial topological mesh and physical physicochemical feature vector. Using the physical physicochemical feature vector as the boundary constraint, a spatiotemporal assessment of the collapse risk of the initial topological mesh is performed to obtain the spatiotemporal tensor of the collapse risk. Based on a preset slip safety threshold, the set of unstable nodes is extracted from the spatiotemporal tensor of collapse risk, and the corresponding nodes in the initial topological grid are subjected to spatial reprojection deformation and surface reconstruction to obtain the predicted evolution terrain grid. Based on the kinematic physical constraints of the stacker-reclaimer, contour line fitting trajectory optimization and curvature smoothing are performed on the outer envelope of the predicted evolution terrain grid to obtain the spatial path spline curve. Based on the spatiotemporal tensor of collapse risk, a four-dimensional dynamic mapping of kinematic command flow is performed on the spatial path spline curve to obtain the terminal execution command flow.

[0007] Secondly, embodiments of the present invention provide a digital coal yard stacker-reclaimer operation path planning system, which includes: The multimodal coal pile state data integration module is used to acquire the original laser point cloud coordinate set collected by the 3D laser scanner, the current pile moisture content value output by the microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical pile records. The 3D mesh model and attribute coupling construction module is used to construct a geometric mesh and couple physical attributes onto the original laser point cloud coordinate set through voxelization downsampling and spatial coordinate registration, so as to obtain the initial topological mesh and physical physicochemical feature vector. The spatiotemporal dynamic assessment module for landslide risk is used to assess the spatiotemporal risk of the initial topological mesh using the physical physicochemical feature vector as the boundary constraint to obtain the spatiotemporal tensor of landslide risk. The instability prediction and terrain evolution simulation module is used to extract the set of unstable nodes from the spatiotemporal tensor of collapse risk based on a preset slip safety threshold, and to perform spatial reprojection deformation and surface reconstruction on the corresponding nodes in the initial topological grid to obtain the predicted evolution terrain grid. The smooth path planning module is used to perform contour line fitting trajectory optimization and curvature smoothing on the outer envelope surface of the predicted evolution terrain grid based on the kinematic physical constraints of the stacker-reclaimer to obtain the spatial path spline curve. The adaptive instruction generation module is used to perform four-dimensional dynamic mapping of kinematic instruction flow on spatial path spline curves based on the spatiotemporal tensor of collapse risk to obtain the terminal execution instruction flow.

[0008] The digital coal yard stacker-reclaimer operation path planning method and system of this invention uses the physicochemical characteristics of coal yard materials as dynamic constraints. It achieves spatiotemporal coupling of operation commands and collapse risks by first pre-simulating the dynamic morphological evolution under cutting disturbances and then generating spatial paths, thus realizing a novel paradigm. Furthermore, by constructing a predictive evolution terrain grid and implementing four-dimensional dynamic mapping, it can effectively avoid the risk of slippage and instability during operation, significantly reduce the frequency of equipment overload, and thereby greatly improve the operating efficiency and trajectory adaptability of the digital coal yard stacker-reclaimer in unattended operation while ensuring production safety. Attached Figure Description

[0009] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0010] Figure 1 A flowchart of a digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention; Figure 2 This is a data flow diagram illustrating the digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention. Figure 3 This is a flowchart of step two in the digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention; Figure 4 This is a block diagram of a digital coal yard stacker-reclaimer operation path planning system according to an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0012] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0013] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0014] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0015] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0016] The present invention proposes a method for planning the operation path of a digital coal yard stacker-reclaimer. Figure 1 This is a flowchart of a digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention. Figure 2 This is a system architecture diagram of a digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention. Figure 1and Figure 2 As shown, the digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention includes the following steps: S1, acquiring the original laser point cloud coordinate set collected by a 3D laser scanner, the current pile moisture content value output by a microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical pile records; S2, through voxel downsampling and spatial coordinate registration, performing geometric mesh construction and physical property coupling mapping on the original laser point cloud coordinate set to obtain an initial topological mesh and a physical physicochemical feature vector; S3, using the physical physicochemical feature vector as boundary constraint conditions, performing a spatiotemporal assessment of collapse risk on the initial topological mesh to obtain a spatiotemporal tensor of collapse risk; S4, based on a preset slip safety threshold, extracting the set of unstable nodes from the spatiotemporal tensor of collapse risk, and performing spatial reprojection deformation and surface reconstruction on the corresponding nodes in the initial topological mesh to obtain a predicted evolution terrain mesh; S5, based on the kinematic physical constraints of the stacker-reclaimer, performing contour line fitting trajectory optimization and curvature smoothing on the outer envelope surface of the predicted evolution terrain mesh to obtain a spatial path spline curve; S6, based on the spatiotemporal tensor of collapse risk, performs four-dimensional dynamic mapping of kinematic command flow on spatial path spline curves to obtain the terminal execution command flow.

[0017] Specifically, S1 involves acquiring the original laser point cloud coordinate set collected by the 3D laser scanner, the current pile moisture content value output by the microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical pile records. It should be understood that in the process of planning the operation path of a digital coal yard stacker-reclaimer, the safety and effectiveness of the path highly depend on a comprehensive understanding of the current state of the coal pile. Traditional path planning schemes rely solely on single 3D laser scanning data to obtain the geometric contour of the coal pile, treating it as a rigid structure with constant physical properties, ignoring the spatial heterogeneity of coal materials in terms of moisture content, particle size distribution, and compaction degree. However, in actual reclaiming operations, the mechanical response and collapse risk after the bucket wheel cuts into the coal pile depend not only on the geometric shape of the coal pile but also closely related to the internal physicochemical characteristics of the material. Specifically, the moisture content directly affects the cohesion and internal friction angle between particles, the particle size distribution determines the material's packing stability and shear strength, and the compaction stress reflects the stress history and density of materials at different depths. By simultaneously acquiring three-dimensional spatial geometric data of the coal pile, real-time moisture content data, and compaction stress and particle size distribution data extrapolated from historical records at the initial stage of path planning, a multimodal and multidimensional data foundation can be provided for the entire planning process.

[0018] In practice, the 3D laser scanner is first mounted on a fixed support or the boom structure of a stacker-reclaimer in the coal yard. Its scanning head rotates and scans horizontally and vertically at a preset angular resolution. At each scanning angle, the laser emitter emits a laser pulse, which propagates at the speed of light to the surface of the coal pile and undergoes diffuse reflection. The reflected signal is received by a photodetector inside the scanner. The scanner's ranging module calculates the slant distance between the measurement point and the scanner based on the time difference between the laser pulse emission and reception. Simultaneously, the scanner's angle encoder records the current horizontal rotation angle and vertical pitch angle of the scanning head. Based on the slant distance and angle information, each measurement point is converted from the scanner's body coordinate system to a 3D spatial coordinate system in the global coordinate system of the coal yard through the principle of spherical to rectangular coordinate conversion. In this way, the 3D coordinates of all measurement points are aggregated into an original laser point cloud coordinate set, and the total number of elements in the set represents the total number of valid measurement points in this scanning cycle.

[0019] Meanwhile, a microwave moisture meter is installed at the front end of the stacker-reclaimer boom or at a fixed monitoring position in the coal yard. Its microwave antenna emits a specific frequency microwave signal towards the coal pile surface. After the microwave signal penetrates the surface material of the coal pile, due to the strong absorption characteristics of microwave energy by water molecules, the amplitude and phase of the signal will change in relation to the moisture content of the material. The receiving antenna of the microwave moisture meter captures the transmitted or reflected microwave signal and extracts the amplitude attenuation and phase shift of the received signal through the built-in signal processing unit. Based on a pre-calibrated mapping model between moisture content and microwave attenuation, the microwave moisture meter converts the amplitude attenuation and phase shift into a volumetric moisture content value of the current pile. By outputting a scalar moisture content value in each measurement cycle, this value represents the average moisture content level of the surface material of the coal pile in the current measurement area.

[0020] In addition, historical stockpiling records include parameters such as the timestamp of each stockpiling operation, stockpiling location, coal type information corresponding to each stockpiling batch, single stockpiling thickness, and material bulk density at the time of stockpiling. Based on these historical records, the system first discretizes the spatial area of ​​the coal pile into a grid, dividing the coal pile into several spatial nodes. For each spatial node, the system calculates the compaction stress at that node based on the historical stockpiling sequence and the bulk density and thickness of each layer of material, according to the superposition principle of vertical effective stress in soil mechanics. Simultaneously, based on the coal type information corresponding to each stockpiling batch, the system retrieves the standard particle size distribution curve of that coal type from the coal type database, and weights and mixes the particle size distribution of multiple layers of material according to the weighting ratio of each layer thickness, thereby obtaining the comprehensive particle size distribution characteristic parameters at that node. By arranging the compaction stress values ​​and particle size distribution characteristic parameters of all spatial nodes in rows, the compaction stress and particle size distribution matrix is ​​obtained.

[0021] Specifically, in step S2, through voxelization downsampling and spatial coordinate registration, a geometric mesh is constructed and physical property coupling mapping is performed on the original laser point cloud coordinate set to obtain an initial topological mesh and physical physicochemical feature vectors. It should be understood that although the original laser point cloud coordinate set completely records the spatial geometric information of the coal pile surface, this data exists in the form of discrete, scattered points, without topological connections between points, and cannot be directly used for subsequent mechanical simulation and collapse risk assessment. At the same time, the original point cloud contains noise points generated by suspended coal dust and water mist interference. These noise points deviate from the true surface of the coal pile, and if not removed, will seriously affect the accuracy of subsequent terrain modeling. Furthermore, the current pile moisture content, compaction stress, and particle size distribution matrix obtained in step one are independent physical parameter data, and have not yet established a correspondence with the spatial geometric structure of the coal pile, making it impossible to characterize the distribution differences of physical physicochemical features in the spatial dimension. Therefore, in the technical solution of this invention, firstly, point cloud redundancy and noise interference are eliminated by voxelization downsampling, and the scattered point cloud is transformed into a three-dimensional geometric mesh with a regular topological structure; secondly, physical properties such as material moisture content, compaction stress and particle size distribution are mapped to each vertex of the mesh, so that the geometric structure and physical properties are coupled in the same data carrier; finally, global physical and chemical feature vectors are extracted from the coupled mesh to provide boundary constraints for the spatiotemporal assessment of collapse risk in step three.

[0022] Figure 3 This is a flowchart of step two in the digital coal yard stacker-reclaimer operation path planning method according to an embodiment of the present invention. Figure 3 As shown, S2 includes: S21, performing voxelization spatial division and dust noise removal on the original laser point cloud coordinate set and calculating the geometric center of the point cloud coordinates within each effective voxel to obtain a basic three-dimensional geometric mesh; S22, mapping the current pile moisture content value, compaction stress, and particle size distribution matrix to each vertex of the basic three-dimensional geometric mesh to obtain an initial topological mesh; S23, performing physical feature weighting and global attribute extraction on the initial topological mesh to obtain a physical physicochemical feature vector.

[0023] Specifically, in step S21, the original laser point cloud coordinate set is subjected to voxel-based spatial partitioning and dust noise removal, and the geometric center of the point cloud coordinates within each effective voxel is calculated to obtain a basic three-dimensional geometric mesh. In this process, firstly, the three-dimensional spatial range covered by the original laser point cloud coordinate set is voxel-based spatial partitioning. Based on the three-dimensional coordinates of all points in the original laser point cloud coordinate set, the extreme values ​​of the point cloud coordinates in the horizontal longitudinal, horizontal transverse, and vertical elevation directions are determined, thereby determining the spatial bounding box of the point cloud. Subsequently, using a preset voxel side length as the unit, the spatial bounding box is uniformly divided into regularly arranged cubic voxel meshes. The number of voxels divided in each direction is equal to the difference of the extreme values ​​of the coordinates in that direction divided by the voxel side length and rounded up. Each voxel occupies a cubic region with a fixed side length in three-dimensional space, and all voxels collectively cover the complete spatial range of the original point cloud. After completing the voxel-based spatial partitioning, the system traverses each point in the original laser point cloud coordinate set and determines its voxel index based on the three-dimensional coordinates of that point. Through the above traversal process, all points in the original laser point cloud coordinate set are assigned to their respective voxels.

[0024] Secondly, dust noise is removed from the point cloud data within each voxel. Because dust particles are suspended in the air in the coal yard environment, some laser pulses emitted by the 3D laser scanner are reflected by dust particles before reaching the coal pile surface, resulting in noise points deviating from the actual coal pile surface mixed into the original point cloud. The typical characteristics of dust noise are: the number of point clouds within the voxel is significantly less than the number of point clouds within the voxels on the coal pile surface, because the distribution of dust particles in space is sparse and random, while the point clouds on the coal pile surface exhibit a dense and continuous distribution. Based on this difference, the system counts the number of point clouds contained in each voxel and identifies voxels with a point cloud count below a preset minimum threshold as dust noise voxels and removes them. Specifically, if the number of point cloud data points contained in a voxel is less than the preset minimum threshold, the point cloud data in that voxel is considered to originate from dust reflection rather than coal pile surface reflection, and the system marks that voxel and all point cloud data within it as invalid and excludes it from subsequent calculations. After dust and noise are removed, the remaining voxels are the effective voxels, and the point cloud data inside them are considered to truly reflect the spatial position of the coal pile surface.

[0025] Next, the geometric center of the point cloud coordinates within each effective voxel is calculated. For each effective voxel, the system calculates the arithmetic mean of the three-dimensional coordinates of all point cloud data points within that voxel in the three directions of horizontal longitudinal, horizontal transverse, and vertical elevation. The three average coordinate values ​​constitute the geometric center coordinates of that effective voxel. In this way, each effective voxel is compressed into a single geometric center point, the spatial location of which represents the average spatial location of the coal pile surface within the area covered by that voxel. The geometric center points of all effective voxels together constitute the downsampled, regularized point set.

[0026] Subsequently, based on the 3D indices of each valid voxel in the voxelized spatial partitioning, it is determined whether any two valid voxels are adjacent in 3D space. If the absolute value of the index difference between two valid voxels in the horizontal longitudinal, horizontal transverse, or vertical elevation directions does not exceed one, then the two voxels are considered adjacent in space, and a topological connection edge is established between their corresponding geometric center points. By traversing all valid voxel pairs and establishing adjacency connections, a basic 3D geometric mesh is finally formed with the geometric center points of the valid voxels as vertices and adjacency connections as edges. This mesh has a regular topological structure, and the spatial distribution of its vertices reflects the geometric morphology of the coal pile surface after downsampling and denoising.

[0027] Specifically, in step S22, the current moisture content of the stockpile and the compaction stress and particle size distribution matrix are mapped to the vertices of the basic three-dimensional geometric grid to obtain the initial topological grid. In this process, firstly, spatial coordinate registration is performed. The vertex coordinates of the basic three-dimensional geometric grid originate from the measurement data of the three-dimensional laser scanner, and its coordinate system is the global coordinate system of the coal yard; while the spatial positions of each node in the compaction stress and particle size distribution matrix originate from the historical stockpile records of the coal yard management system, and its coordinate system is the grid coordinate system of the coal yard management system. Since the two coordinate systems may have origin offsets, axial rotations, or scale differences, the system needs to unify them to the same coordinate reference system through spatial coordinate registration. The specific implementation method of spatial coordinate registration is as follows: several fixed reference points with known spatial positions in the coal yard (such as coal yard boundary markers, track reference points, etc.) are selected, and the coordinates of these reference points are obtained in both coordinate systems. Then, the transformation relationship between the two coordinate systems is solved through least-squares fitting of rigid body transformation (including translation vectors and rotation matrices). After the transformation relationship is determined, the coordinates of each node in the compaction stress and particle size distribution matrix are transformed from the grid coordinate system of the coal yard management system to the global coordinate system of the coal yard, so that they are in the same spatial reference frame as the vertex coordinates of the basic three-dimensional geometric grid.

[0028] Secondly, the current moisture content value of the coal pile is mapped to each vertex of the basic three-dimensional geometric mesh. Since the microwave moisture meter outputs a scalar moisture content value in each measurement cycle, representing the average moisture content level of the surface material of the coal pile in the current measurement area, in this step, the system assigns this scalar moisture content value as a uniform attribute to all vertices in the basic three-dimensional geometric mesh.

[0029] Then, the compaction stress and grain size distribution matrix is ​​mapped to each vertex of the basic 3D geometric mesh. Since the spatial distribution of nodes in the compaction stress and grain size distribution matrix does not usually completely coincide with the spatial distribution of vertices in the basic 3D geometric mesh, the system needs to map the physical parameters in the matrix to the mesh vertices through spatial interpolation. For each vertex in the basic 3D geometric mesh, the system searches for the nearest matrix nodes in the coordinate-registered set of nodes of the compaction stress and grain size distribution matrix. Then, using the inverse ratio of distance as a weight, it performs weighted interpolation on the compaction stress values ​​and grain size distribution characteristic parameters of these neighboring matrix nodes. The resulting interpolation value is the compaction stress value and grain size distribution characteristic parameters at that mesh vertex. Through the above interpolation process, the attribute fields of each vertex in the basic 3D geometric mesh are augmented with a compaction stress component and three characteristic grain size components corresponding to cumulative pass rates of 10%, 50%, and 90%.

[0030] After mapping the moisture content and the compaction stress and particle size distribution, each vertex in the basic three-dimensional geometric mesh simultaneously carries three-dimensional spatial coordinates, moisture content value, compaction stress value, and three characteristic particle size values, thus forming an initial topological mesh that couples geometric information and physical properties. The complete attributes of each vertex in the initial topological mesh include: three-dimensional spatial coordinates (horizontal longitudinal coordinates, horizontal transverse coordinates, and vertical elevation coordinates), moisture content component, compaction stress component, and three characteristic particle size components (characteristic particle size values ​​corresponding to cumulative throughput of 10%, 50%, and 90%), for a total of seven attribute dimensions.

[0031] Specifically, in step S23, physical feature weighting and global attribute extraction are performed on the initial topological mesh to obtain a physical physicochemical feature vector. In this process, firstly, the physical attributes of each vertex in the initial topological mesh are weighted. Each vertex in the initial topological mesh carries physical attributes including a moisture content component, a compaction stress component, and three characteristic particle size components. These physical parameters have different dimensions and numerical ranges. To eliminate the influence of dimensional differences on subsequent global attribute extraction, the system first normalizes each physical parameter component. Next, based on prior knowledge of the impact of each physical parameter component on the risk of collapse, corresponding weight coefficients are assigned to each component. The weight coefficient of the moisture content component reflects the influence of moisture content on interparticle cohesion; the weight coefficient of the compaction stress component reflects the influence of compaction degree on material density; and the weight coefficients of each characteristic particle size component reflect the influence of particle size composition on packing stability. The system multiplies the normalized physical parameter components of each vertex by their corresponding weight coefficients to obtain weighted physical feature values.

[0032] Furthermore, global attribute extraction is performed on the weighted physical features to generate a physical physicochemical feature vector. Specifically, the system performs global statistical calculations on each weighted physical parameter component of all vertices in the initial topological mesh, extracting the global mean, global standard deviation, global maximum, and global minimum value for each component. For the five physical parameter components—moisture content, compaction stress, and three characteristic particle sizes—four global statistics (mean, standard deviation, maximum, and minimum) are extracted for each component, resulting in twenty statistical feature values ​​after global attribute extraction. The system then concatenates these twenty statistical feature values ​​into a fixed-dimensional numerical vector according to a preset arrangement; this vector is the physical physicochemical feature vector.

[0033] Specifically, S3 uses the physical physicochemical feature vector as the boundary constraint to perform a spatiotemporal assessment of the collapse risk on the initial topological mesh to obtain the spatiotemporal tensor of the collapse risk. It should be understood that although the initial topological mesh has achieved a coupled expression of the coal pile's geometry and physical properties, this mesh only reflects the static state of the coal pile at the current moment and does not yet contain any information about the dynamic mechanical response and morphological evolution of the coal pile after being subjected to cutting disturbances by the stacker-reclaimer. In actual reclaiming operations, the cutting of the stacker-reclaimer bucket wheel will exert strong mechanical disturbances on the coal pile. This disturbance will induce stress redistribution and breakage of interparticle force chains within the coal pile, which may lead to slippage instability or even collapse in local areas. Due to significant differences in moisture content, compaction degree, and particle size distribution in different regions of the coal pile, the mechanical response triggered by the same cutting action at different locations is also drastically different. Therefore, in the technical solution of this invention, the microscopic dynamic parameters controlling the mechanical behavior between particles are first calculated from the physical and chemical feature vectors of the material. Then, a pre-trained discrete element reduced-order neural network surrogate model is used to simulate the dynamic instability response under cutting disturbance. Finally, the instability response is transformed into a spatiotemporal tensor of collapse risk covering both spatial and temporal dimensions through time-step iterative integration and probability mapping, providing a quantitative risk assessment basis for the extraction of instability nodes and the prediction of terrain evolution in step four.

[0034] In practice, firstly, the internal friction angle and cohesion are calculated for the moisture content and compaction stress components in the physicochemical feature vector to obtain the microscopic dynamic boundary parameters. This process involves first extracting moisture content-related statistical features and compaction stress-related statistical features from the physicochemical feature vector. Secondly, the internal friction angle is calculated based on the global mean of the moisture content and compaction stress components. The calculation of the internal friction angle is based on an empirically modified model of the Mohr-Coulomb strength criterion, which expresses the internal friction angle as a function of compaction stress and moisture content.

[0035] Next, the initial topological mesh is imported into a pre-trained discrete element reduction neural network surrogate model for simulation environment instantiation to obtain a constrained coupled simulation mesh. In this process, firstly, the pre-trained discrete element reduction neural network surrogate model is loaded. This model has been trained offline using a large amount of discrete element simulation data, and its network structure includes an encoder, a dynamic inference core, and a decoder. Next, the data structure of the initial topological mesh is converted into the input format required by the discrete element reduction neural network surrogate model. The seven attribute dimensions of each vertex in the initial topological mesh (three-dimensional spatial coordinates, water content component, compaction stress component, and three characteristic particle size components) are organized into a node feature matrix of the model input tensor, and the topological connections between vertices in the initial topological mesh are organized into an adjacency matrix of the model input tensor. The number of rows in the node feature matrix equals the total number of vertices in the initial topological mesh, and the number of columns equals seven; the adjacency matrix is ​​a sparse matrix, where non-zero elements correspond to vertex pairs with topological connections. Finally, the formatted initial topological mesh data is input into the encoder of the discrete element reduction neural network surrogate model for simulation environment instantiation. The encoder receives the node feature matrix and adjacency matrix, and aggregates and fuses the local attribute information of each vertex with the attribute information of its neighboring vertices through multi-layer graph convolution operations, generating a low-dimensional latent feature vector for each vertex. The computational logic of the graph convolution operation is as follows: for a given vertex, its updated latent feature vector is equal to the weighted aggregation result of the vertex's current feature vector and the feature vectors of all its neighboring vertices, transformed by a nonlinear activation function. The weights are determined by the parameter matrix learned by the model during training. After multi-layer graph convolution operations, the latent feature vector of each vertex not only encodes its own physical attribute information but also integrates the topological structure and attribute distribution information within its spatial neighborhood. The latent feature vectors of all vertices output by the encoder collectively constitute the instantiated representation of the simulation environment. Subsequently, the micro-dynamic boundary parameters output from the first sub-process are injected into the instantiated simulation environment to form a constrained coupled simulation mesh. Specifically, the internal friction angle and cohesion in the micro-dynamic boundary parameters are used as global constraint parameters and appended to the latent feature vector of each vertex output by the encoder.

[0036] Furthermore, based on the preset bucket wheel slewing radius and single feed step size, a virtual cutting displacement vector is applied to the working contact surface of the constrained coupling simulation mesh. Using microscopic dynamic boundary parameters as constraint inputs, the dynamic instability response of the cutting displacement vector is calculated to obtain the mesh node instability response matrix. In this process, firstly, the working contact surface on the constrained coupling simulation mesh is determined. The working contact surface refers to the area where the stacker-reclaimer bucket wheel directly contacts the coal pile surface during the material reclaiming action. Based on the current spatial pose of the stacker-reclaimer (including slewing angle, pitch angle, and cantilever extension length) and the preset bucket wheel slewing radius, the system determines the set of mesh vertices covered by the bucket wheel sweep trajectory on the outer surface of the constrained coupling simulation mesh. The mesh surface area formed by this vertex set is the working contact surface. Specifically, with the slewing center of the stacker-reclaimer bucket wheel as the center and the bucket wheel slewing radius as the radius, an arc-shaped strip area is delineated on the outer surface of the constrained coupling simulation mesh. All mesh vertices within this area are marked as working contact surface vertices. Secondly, a virtual cutting displacement vector is applied to the working contact surface. The virtual cutting displacement vector is used to simulate the mechanical displacement disturbance applied to the working contact surface when the bucket wheel cuts into the coal pile. For each vertex on the working contact surface, the system calculates the virtual cutting displacement vector at that vertex based on the radial direction of that vertex relative to the center of rotation of the bucket wheel and the preset single feed step size. The direction of the virtual cutting displacement vector is a radial unit vector direction from that vertex to the center of rotation of the bucket wheel, and its magnitude is equal to the single feed step size.

[0037] It is worth mentioning that for grid vertices not on the working contact surface, their virtual cutting displacement vectors are set to zero, indicating that these vertices are not directly subjected to the mechanical disturbance of the bucket wheel cutting. Then, the constrained coupled simulation grid with applied virtual cutting displacement vectors is input into the dynamic derivation core of the discrete element reduced-order neural network surrogate model for dynamic instability response calculation. Specifically, the implicit feature vectors of each vertex in the constrained coupled simulation grid and the virtual cutting displacement vectors at each vertex are received as inputs, while microscopic dynamic boundary parameters (internal friction angle and cohesion) are used as global constraints. A multi-layer message-passing neural network simulates the propagation of cutting disturbances from the working contact surface to the surrounding area on the grid's topology. In each layer of message passing, each vertex updates its mechanical response state based on its own implicit feature vector, virtual cutting displacement vector, and messages passed from all its adjacent vertices. For a given vertex, the received message is equal to the aggregated result of the implicit feature vectors of all its adjacent vertices and the corresponding edge features after transformation by the message function. The updated implicit feature vector of this vertex is equal to its current implicit feature vector, the received message, and the output of the virtual cutting displacement vector at this vertex after transformation by the update function. After multiple layers of message passing, the impact of the cutting disturbance propagates layer by layer from the working contact surface to the entire mesh. The latent feature vector of each vertex encodes the mechanical response information of that vertex under the cutting disturbance. Subsequently, the updated latent feature vectors of each vertex output by the dynamics derivation core are input into the decoder of the discrete element reduced-order neural network surrogate model, and decoded into the explicit mechanical response quantities of each vertex. Finally, the mechanical response quantities of all vertices are arranged in rows to form the mesh node instability response matrix. In particular, the number of rows in the mesh node instability response matrix is ​​equal to the total number of vertices in the constrained coupled simulation mesh, and the number of columns is equal to the dimension of the mechanical response quantity (including three components of the transient slip displacement vector, the transient slip velocity scalar, and the transient stress change scalar, for a total of five components). Each row in the mesh node instability response matrix records the transient mechanical response of the corresponding vertex under a single cutting disturbance, where the transient slip displacement vector represents the spatial displacement and direction of the vertex due to the cutting disturbance, the transient slip velocity scalar represents the speed of the vertex's slip motion, and the transient stress change scalar represents the change in the stress state at the vertex relative to before cutting.

[0038] Subsequently, within a preset future time window, the transient slip response of each node in the grid node instability response matrix is ​​iteratively integrated over time and mapped with logistic probability to obtain the spatiotemporal tensor of collapse risk. In this process, firstly, the future time window and time step parameters are set. Based on the stacker-reclaimer's operating speed and reclaiming cycle, the system presets a future time window, covering the time range from the current moment to several seconds or minutes in the future. The system evenly divides this time window into several discrete time steps, each with a fixed duration. The total number of time steps within the future time window is equal to the total duration of the time window divided by the time step. Secondly, the transient slip response of each node in the grid node instability response matrix is ​​iteratively integrated over time to accumulate and evolve the transient mechanical response under a single cutting disturbance in the time dimension, simulating the accumulation process of slip displacement of each node within the future time window over time. Specifically, in the first time step, the cumulative slip displacement of each node is equal to the transient slip displacement vector of the corresponding node in the grid node instability response matrix multiplied by the time step. Starting from the second time step, the system updates the cumulative slip state of each node in each time step. During execution, in each time step, the system first recalculates the relative displacement and contact state changes between nodes based on the cumulative slip displacement of each node in the previous time step. Then, these changes are used as new perturbation inputs to the dynamic derivation core of the discrete element reduced-order neural network surrogate model to obtain the incremental slip displacement vector and incremental slip velocity scalar of each node in that time step. The cumulative slip displacement of each node in the current time step is equal to the cumulative slip displacement of the previous time step plus the incremental slip displacement vector of the current time step multiplied by the time step length.

[0039] By repeating the iterative integration process described above at all time steps, the system obtains the cumulative slip displacement vector of each node at the end of each time step. Then, a logistic probability mapping is performed on the cumulative slip displacement of each node at each time step, transforming the continuous physical displacement into a collapse probability value between zero and one. Specifically, when the cumulative slip displacement amplitude is small, the collapse probability is close to zero, indicating that the node is in a stable state; when the cumulative slip displacement amplitude exceeds a critical threshold, the collapse probability rises rapidly and approaches one, indicating that the node is in a highly unstable state. Subsequently, the collapse probability values ​​of all nodes at all time steps, along with the associated slip direction and slip amplitude information, are assembled into a collapse risk spatiotemporal tensor. This tensor quantitatively describes the collapse risk level and slip evolution trend of each spatial node in the initial topological grid at each time point within a future time window.

[0040] Specifically, in step S4, based on a preset slip safety threshold, an unstable node set is extracted from the collapse risk spatiotemporal tensor, and the corresponding nodes in the initial topological grid are spatially reprojected, deformed, and their surfaces reconstructed to obtain a predicted evolutionary terrain grid. The collapse risk spatiotemporal tensor quantitatively describes the collapse probability value and associated slip displacement information of each spatial node in the initial topological grid at each time point within a future time window. However, this tensor itself is only a risk assessment data and has not yet transformed the risk assessment results into the actual morphological changes of the coal pile terrain under cutting disturbance. In actual material handling operations, when the collapse risk in certain areas exceeds the safety limit, the surface of the coal pile in these areas will undergo slip deformation, and the original terrain contour will no longer hold. If the path planning is still based on the uncorrected original terrain, the planned path will deviate from the actual terrain, causing the bucket wheel to cut too deeply or into empty areas, leading to mechanical jamming or idling failures. Therefore, in the technical solution of the present invention, firstly, unstable nodes with a collapse probability exceeding the safety limit are screened from the spatiotemporal tensor of collapse risk based on a preset slip safety threshold. Then, the spatial position of the nodes at the corresponding positions in the initial topological grid is corrected according to the slip displacement information of these unstable nodes. Finally, a predicted evolution terrain grid reflecting the expected shape of the coal pile after collapse is generated through surface reconstruction, providing a terrain basis after collapse correction for path planning in step five.

[0041] In specific implementation, firstly, based on a preset slip safety threshold, the collapse probability values ​​of each node in the collapse risk spatiotemporal tensor are threshold-judged and filtered to obtain the set of unstable node displacements. During this process, the collapse risk spatiotemporal tensor first performs maximum aggregation along the time dimension. The collapse risk spatiotemporal tensor is a three-dimensional array data structure, where the first dimension corresponds to the grid node index, the second dimension to the time step index, and the third dimension to the risk feature channel. The system traverses the collapse probability channel of each node along the time step dimension, extracting the maximum value of the collapse probability value for that node across all time steps within the entire future time window, as the peak collapse probability of that node. Specifically, for the i-th node, its peak collapse probability is equal to the maximum value of all collapse probability values ​​for that node between the first and last time steps. Simultaneously, the system records the time step index corresponding to this maximum value as the peak risk moment for that node.

[0042] Secondly, all grid nodes are traversed, and the peak collapse probability of each node is compared with a preset slip safety threshold. If the peak collapse probability of a node is greater than or equal to the preset slip safety threshold, the node is determined to be an unstable node; if the peak collapse probability of a node is less than the preset slip safety threshold, the node is determined to be a stable node. Furthermore, for all nodes determined to be unstable, the cumulative slip displacement vector corresponding to the node at its peak risk time is extracted from the collapse risk spatiotemporal tensor.

[0043] Next, based on the slip displacement vectors of each node in the unstable node displacement set, spatial reprojection deformation is performed on the corresponding nodes and their neighboring nodes in the initial topological mesh to obtain a spatially deformed scattered point set. In this process, firstly, direct spatial reprojection is performed on each unstable node in the unstable node displacement set. For each unstable node i in the unstable node displacement set, the original spatial coordinates of the node in the initial topological mesh are translated along the direction of its slip displacement vector to obtain the deformed spatial coordinates of the node. Secondly, attenuated spatial reprojection is performed on the neighboring nodes of each unstable node. In actual coal pile collapse processes, slip deformation does not only occur at the location of the unstable node itself, but spreads to the surrounding area. The closer the area is to the unstable node, the greater the deformation impact; the farther away, the smaller the impact until it disappears. To simulate this spatial diffusion effect, the system applies displacement corrections with distance attenuation to the neighboring nodes of each unstable node. For a neighboring node j of unstable node i (neighboring node j itself does not belong to the unstable node in the unstable node displacement set), the system first calculates the Euclidean distance between node j and unstable node i in the initial topology grid, and then calculates the distance attenuation weight based on the distance and the preset influence radius parameter.

[0044] Finally, the deformed spatial coordinates of all nodes are converged into a spatially deformed scattered point set. For unstable nodes, their deformed spatial coordinates are obtained by direct spatial reprojection; for neighboring nodes within the influence radius of unstable nodes, their deformed spatial coordinates are obtained by attenuated spatial reprojection; for stable nodes that are neither unstable nodes nor within the influence radius of any unstable node, their deformed spatial coordinates remain consistent with their original spatial coordinates in the initial topological mesh.

[0045] Furthermore, a two-dimensional projection topological connection and a three-dimensional elevation centroid interpolation reconstruction are performed on the spatially deformed scattered point set to obtain the predicted evolving terrain mesh. In this process, firstly, the three-dimensional spatial coordinates of all points in the spatially deformed scattered point set are projected onto a horizontal plane along the vertical elevation direction to obtain the two-dimensional planar coordinates of each point. Secondly, Delaunay triangulation is performed on the projected two-dimensional scattered point set to establish topological connections. Delaunay triangulation is a classic computational geometry algorithm whose purpose is to connect discrete scattered points on a two-dimensional plane into a set of non-overlapping triangles, such that the circumcircle of each triangle does not contain other scattered points. By performing the Delaunay triangulation algorithm on the projected two-dimensional scattered point set, the topological connections of triangles on the two-dimensional plane can be obtained, including a triangle list (each triangle records the indices of its three vertices) and an edge list (each edge records the indices of its two endpoints). Furthermore, the two-dimensional topological connections established by Delaunay triangulation are directly applied to the three-dimensional coordinates of the spatially deformed scattered point set. That is, the complete three-dimensional coordinates (including horizontal longitudinal, horizontal transverse, and vertical elevation coordinates) of each point in the spatially deformed scattered point set are used as the vertex coordinates of the triangular mesh, and the triangular connections determined by Delaunay triangulation are used as the topological structure of the mesh. This elevates the triangular mesh on the two-dimensional plane into a triangular patch mesh in three-dimensional space. During the elevation process, due to spatial reprojection deformation, local triangular patches may exhibit geometric anomalies such as normal vector flipping, patch intersections, or abrupt elevation changes. The system needs to perform three-dimensional elevation centroid interpolation correction on these anomaly regions. After reconstruction through two-dimensional projection topological connections and three-dimensional elevation centroid interpolation, the spatially deformed scattered point set is reorganized into a three-dimensional mesh with a regular triangular patch topological structure. This mesh is the predicted evolutionary terrain mesh. The geometric morphology of the predicted evolutionary terrain mesh in the unstable region reflects the expected coal pile surface profile after the collapse deformation, while the geometric morphology in the stable region remains consistent with the initial topological mesh.

[0046] Specifically, in step S5, based on the kinematic physical constraints of the stacker-reclaimer, contour line fitting trajectory optimization and curvature smoothing are performed on the outer envelope surface of the predicted evolving terrain grid to obtain a spatial path spline curve. It should be understood that in actual material reclaiming operations, the bucket wheel boom of the stacker-reclaimer is limited by kinematic physical constraints such as the turning radius limit and pitch angle limit, meaning that not all areas on the predicted evolving terrain grid are within the reachable operating range of the stacker-reclaimer. Furthermore, material reclaiming operations are typically carried out using a layered cutting method, where the bucket wheel cuts into the coal pile surface layer by layer at a certain step height, with each layer sweeping horizontally along the contour line direction. Therefore, in the technical solution of this invention, firstly, the effective operating envelope surface within the kinematic reachable range of the stacker-reclaimer is extracted from the predicted evolving terrain grid. Then, contour line trajectory points are extracted layer by layer on this envelope surface according to a preset step height. Finally, B-spline curve fitting and curvature smoothing are used to transform the discrete trajectory points into a continuous and smooth spatial path spline curve, providing a geometric trajectory basis for the four-dimensional dynamic mapping of the kinematic command flow in step six.

[0047] In practice, the system first extracts the outer envelope surface and performs kinematic constraint trimming on the predicted terrain mesh to obtain the effective operational envelope surface. This process begins by extracting the outer topological envelope surface of the predicted terrain mesh. The system determines the orientation vector of the stacker-reclaimer relative to the coal pile based on the stacker-reclaimer's current spatial pose; this orientation vector points from the center of the coal pile to the stacker-reclaimer's rotation center. Subsequently, the system traverses all triangular faces in the predicted terrain mesh and calculates the normal vector of each triangular face. Then, the system performs a dot product operation between the normal vector of each triangular face and the orientation vector of the stacker-reclaimer. If the dot product result is positive, it indicates that the angle between the face's normal vector and the stacker-reclaimer's orientation is less than 90 degrees, meaning the face faces the stacker-reclaimer and belongs to the outer envelope surface. If the dot product result is negative or zero, it indicates that the face is away from the stacker-reclaimer or parallel to the orientation direction and does not belong to the outer envelope surface. All the triangular facets identified as outer edges collectively constitute the outer topological envelope of the predicted evolving terrain grid. Next, based on the stacker-reclaimer's turning radius limit and pitch angle limit, the outer topological envelope is trimmed to cover inaccessible areas to obtain the effective operational envelope.

[0048] Next, based on the preset single-retrievable step height, contour line trajectory points are extracted along the vertical direction of the effective working envelope to obtain a discrete spatial trajectory point set. In this process, firstly, the elevation range and number of cutting layers for contour line cutting are determined. The system iterates through the vertical elevation coordinates of all vertices in the effective working envelope, extracting the maximum and minimum elevation values. The system sets the cutting elevation layer by layer, starting from the maximum elevation value, using the preset single-retrievable step height as the interval. Secondly, contour line trajectory points are extracted for each cutting layer. For the l-th cutting layer, the system constructs a horizontal cutting plane with the elevation value z_l, and performs an intersection operation with the triangular facets in the effective working envelope. For each triangular facet in the effective working envelope, the system checks the relationship between the vertical elevation coordinates of the facet's three vertices and the cutting elevation z_l. If all three vertices' elevation coordinates are higher than or lower than the cutting elevation, the facet does not intersect the cutting plane, and the facet is skipped. If one of the three vertices is above the cutting elevation and the other is below the cutting elevation, then the facet intersects the cutting plane, and the intersection line is a line segment. Specifically, for each edge of the triangular facet that crosses the cutting elevation (i.e., the two endpoints of the edge are located above and below the cutting elevation, respectively), the system calculates the coordinates of the intersection point between the edge and the cutting plane using linear interpolation. Each triangular facet intersecting the cutting plane generates two intersection points, which form an intersection line segment. The system arranges all the intersection line segments between the triangular facets and the cutting plane in a head-to-tail order to form the contour trajectory of the l-th cutting layer. The system samples the contour trajectory at equal intervals along the contour trajectory at a preset trajectory point sampling interval to obtain the contour trajectory point sequence of the l-th cutting layer. Then, the contour trajectory point sequences of all cutting layers are aggregated into a discrete spatial trajectory point set.

[0049] Then, B-spline curve fitting and curvature smoothing are performed on the discrete spatial trajectory point set to obtain the spatial path spline curve. In this process, firstly, B-spline curve fitting is performed on the contour trajectory point sequences of each cutting layer in the discrete spatial trajectory point set. For the contour trajectory point sequence of the l-th cutting layer, the system uses the trajectory points in this sequence as the data points for B-spline curve fitting. The goal of B-spline curve fitting is to find a set of control points and node vectors such that the generated B-spline curve optimally approximates the given data point sequence in the least squares sense. Secondly, inter-layer connections are made between the B-spline curves of each cutting layer to form a complete spatial path. The system connects the ends and starts of the B-spline curves of adjacent cutting layers through vertical transition segments. The start point of the transition segment is the coordinate of the end point of the B-spline curve of the previous layer, and the end point is the coordinate of the start point of the B-spline curve of the next layer. Through inter-layer connections, the B-spline curves of all cutting layers and the inter-layer transition segments together constitute a complete spatial path from the top to the bottom of the coal pile. Furthermore, curvature smoothing is applied to the complete spatial path to eliminate potential local curvature abrupt changes during B-spline curve fitting, ensuring continuous and gentle curvature changes in the spatial path spline curve. This prevents the stacker-reclaimer from experiencing mechanical impacts due to abrupt directional changes as it moves along the path. Specifically, firstly, isoparametric sampling is performed along the complete spatial path with a fixed parameter step size, calculating the curvature value at each sampling point. Secondly, the curvature values ​​at all sampling points are checked to ensure they do not exceed a preset maximum allowable curvature threshold. If the curvature value at a sampling point exceeds the maximum allowable curvature threshold, the system fine-tunes the control points near that sampling point, reducing the curvature value in that region through iterative optimization. After iterative optimization, the curvature values ​​at all sampling points on the complete spatial path do not exceed the maximum allowable curvature threshold, and the curvature changes are continuous and gentle. The complete spatial path after curvature smoothing is the spatial path spline curve.

[0050] Specifically, in step S6, based on the spatiotemporal tensor of collapse risk, a four-dimensional dynamic mapping of the kinematic command flow of the spatial path spline curve is performed to obtain the terminal execution command flow. It should be understood that in actual material handling operations, the speed of the bucket wheel along the path should not be uniform and constant, but should be adaptively adjusted according to the collapse risk level at each location on the path: in areas with high collapse risk, the bucket wheel should reduce its speed to decrease the intensity of cutting disturbance and reduce the probability of inducing collapse; in areas with low collapse risk, the bucket wheel can appropriately increase its speed to improve operational efficiency. Simultaneously, the stacker-reclaimer's underlying controller receives and executes motion commands with a fixed control cycle, and the register bit width of the controller has strict limitations on the numerical precision and format of the command data. Therefore, the spatial path spline curve must be converted into a digital command sequence that conforms to the hardware specifications of the underlying controller before it can be actually executed. In the technical solution of this invention, the spatial path spline curve is first spatiotemporally coupled with the landslide risk spatiotemporal tensor so that each discrete point on the path carries the risk characteristic information of the corresponding location; then, adaptive speed allocation is performed according to the risk characteristics to generate a four-dimensional dynamic control sequence containing spatial coordinates and speed information; finally, the control sequence is translated into a digital terminal execution instruction stream that can be directly executed by the underlying controller.

[0051] In specific implementation, firstly, based on the fixed control cycle of the underlying controller, the spatial path spline curve is discretized and sampled in the same time domain. The coordinates of each discrete path point are then back-mapped to the collapse risk spatiotemporal tensor using a trilinear interpolation algorithm for risk feature retrieval, thus obtaining a set of discrete trajectory points carrying risk features. In this process, firstly, according to the fixed control cycle of the underlying controller and the initial movement speed of the stacker-reclaimer (before adaptive speed allocation, the system uses a preset rated movement speed as the initial movement speed for isotemporal sampling), time-uniform discretization sampling is performed along the spatial path spline curve. Secondly, the coordinates of each discrete path point are back-mapped to the collapse risk spatiotemporal tensor using a trilinear interpolation algorithm for risk feature retrieval. Through trilinear interpolation, the system obtains the risk feature values ​​at the discrete path point, including the collapse probability value, the three spatial components of the cumulative slip displacement vector, and the cumulative slip velocity scalar, totaling the values ​​of five risk feature channels. It is worth mentioning that the landslide risk spatiotemporal tensor contains both spatial and temporal dimensions. When retrieving risk features, the system determines the time step index corresponding to the discrete path point in the time dimension based on the sampling time corresponding to that path point, and then performs trilinear interpolation on the spatial slice corresponding to that time step. Through the above process, the system couples each discrete path point on the spatial path spline curve with the risk features of the corresponding spatiotemporal location in the landslide risk spatiotemporal tensor, obtaining a set of discrete trajectory points carrying risk features.

[0052] Next, adaptive velocity allocation is calculated for the comprehensive risk assessment value of each point in the discrete trajectory point set carrying risk characteristics to obtain a four-dimensional dynamic control sequence. In this process, firstly, the multi-dimensional risk characteristics of each point in the discrete trajectory point set carrying risk characteristics are comprehensively weighted to calculate the comprehensive risk assessment value. Specifically, firstly, the Euclidean norm of the cumulative slip displacement vector is calculated as the slip displacement amplitude characteristic; secondly, the collapse probability value, slip displacement amplitude, and cumulative slip velocity scalar are normalized; then, based on the prior weights of the impact of each risk characteristic component on operational safety, the three normalized risk characteristic components are weighted and summed to obtain the comprehensive risk assessment value of the trajectory point. Subsequently, adaptive velocity allocation is calculated based on the comprehensive risk assessment value. Specifically, trajectory points with higher comprehensive risk assessment values ​​are assigned lower movement speeds, and trajectory points with lower comprehensive risk assessment values ​​are assigned higher movement speeds. Finally, acceleration constraint verification and smoothing are performed on the velocity changes between adjacent trajectory points. Because the stacker-reclaimer's drive mechanism has a maximum acceleration limit, the speed change between two adjacent control cycles cannot exceed the product of the maximum acceleration and the control cycle. The system iterates through all adjacent trajectory point pairs, checking whether the absolute value of the speed change between adjacent points exceeds the product of the maximum allowable acceleration and the fixed control cycle. If the speed change between a pair of adjacent trajectory points exceeds the acceleration constraint, the system corrects the speed value of the trajectory point that exceeds the constraint to the maximum allowable value that meets the constraint. After adaptive speed allocation and acceleration constraint correction, the spatial coordinates of each trajectory point are recalculated based on the corrected speed sequence. Since the change in speed causes a change in the arc length increment of the isochronous sampling, the system uses the corrected speed sequence as a basis to re-perform isochronous discretization sampling along the spatial path spline curve to obtain an updated discrete path point spatial coordinate sequence. Furthermore, the updated three-dimensional spatial coordinates of each trajectory point are combined with the corresponding target motion speed into a four-dimensional data tuple. All four-dimensional data tuples of trajectory points are arranged in chronological order to form a four-dimensional dynamic control sequence. The data structure of each element in the four-dimensional dynamic control sequence includes the target horizontal longitudinal coordinate, horizontal transverse coordinate, vertical elevation coordinate, and target speed that the stacker-reclaimer should reach within the control cycle.

[0053] Furthermore, based on the register bit width limitation of the underlying controller, the four-dimensional dynamic control sequence is digitally translated and serialized to obtain the terminal execution instruction stream. In this process, firstly, the floating-point values ​​in the four-dimensional dynamic control sequence are quantized and encoded using fixed-point quantization. The register bit width of the underlying controller determines the numerical precision it can process; floating-point values ​​must be converted into fixed-point integers matching the register bit width to be correctly parsed by the controller. The system performs fixed-point quantization and encoding on each dimension of the four-dimensional dynamic control sequence. Secondly, the quantized fixed-point integer values ​​are serialized into binary bits. Specifically, the fixed-point integer values ​​of the four dimensions of a trajectory point are converted into binary representations with the same number of bits as the register bit width. Then, according to a preset dimensional arrangement order (horizontal longitudinal coordinate, horizontal transverse coordinate, vertical elevation coordinate, target velocity), the four binary values ​​are concatenated end-to-end to form a single frame of instruction data for that trajectory point. The total bit width of a single frame of instruction data is equal to four times the register bit width. The system adds a frame header identifier, frame sequence number, and timestamp information to the front end of each single-frame instruction data, and a cyclic redundancy check code to the back end, forming a complete single-frame instruction message. Then, the single-frame instruction messages of all trajectory points are concatenated in chronological order to form a terminal execution instruction stream. The terminal execution instruction stream is a continuous binary data stream containing all instruction frames from the first control cycle to the last control cycle. The terminal execution instruction stream can be directly written to the instruction buffer of the underlying controller. The controller reads the instruction messages frame by frame from the buffer according to a fixed control cycle, parses the fixed-point integer values ​​of the target spatial coordinates and target motion velocity, and then restores them to floating-point values ​​through dequantization, driving the stacker-reclaimer's slewing, pitching, and traveling mechanisms to move according to the planned path.

[0054] In summary, the digital coal yard stacker-reclaimer operation path planning method according to embodiments of the present invention is explained. It achieves spatiotemporal coupling of operation commands and collapse risks by using the physicochemical characteristics of coal yard materials as dynamic constraints, and by first pre-simulating the dynamic morphological evolution under cutting disturbances and then generating spatial paths—a novel paradigm. Furthermore, by constructing a predictive evolution terrain grid and implementing four-dimensional dynamic mapping, it can effectively avoid the risk of slippage and instability during operation, significantly reduce the frequency of equipment overload, and thus, while ensuring production safety, greatly improve the operating efficiency and trajectory adaptability of the digital coal yard stacker-reclaimer in unattended operation.

[0055] Furthermore, a digital coal yard stacker-reclaimer operation path planning system is also provided.

[0056] Figure 4 This is a block diagram of a digital coal yard stacker-reclaimer operation path planning system according to an embodiment of the present invention. Figure 4As shown, the digital coal yard stacker-reclaimer operation path planning system 300 according to an embodiment of the present invention includes: a multimodal coal pile state data integration module 310, used to acquire the original laser point cloud coordinate set collected by a 3D laser scanner, the current pile moisture content value output by a microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical pile records; a 3D mesh model and attribute coupling construction module 320, used to perform geometric mesh construction and physical attribute coupling mapping on the original laser point cloud coordinate set through voxel downsampling and spatial coordinate registration to obtain an initial topological mesh and physical physicochemical feature vector; and a spatiotemporal dynamic assessment of collapse risk. Module 330 is used to perform a spatiotemporal assessment of the collapse risk of the initial topological mesh using physical physicochemical feature vectors as boundary constraints to obtain the spatiotemporal tensor of the collapse risk. Module 340 is used to extract the set of unstable nodes from the spatiotemporal tensor of the collapse risk based on a preset slip safety threshold, and to perform spatial reprojection deformation and surface reconstruction on the corresponding nodes in the initial topological mesh to obtain the predicted evolving terrain mesh. Module 350 is used to perform contour line fitting trajectory optimization and curvature smoothing on the outer envelope of the predicted evolving terrain mesh based on the kinematic physical constraints of the stacker-reclaimer to obtain the spatial path spline curve. Module 360 ​​is used to perform four-dimensional dynamic mapping of the kinematic command flow on the spatial path spline curve based on the spatiotemporal tensor of the collapse risk to obtain the terminal execution command flow.

[0057] As described above, the digital coal yard stacker-reclaimer operation path planning system 300 according to embodiments of the present invention can be implemented in various wireless terminals, such as servers with digital coal yard stacker-reclaimer operation path planning algorithms. In one possible implementation, the digital coal yard stacker-reclaimer operation path planning system 300 according to embodiments of the present invention can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the digital coal yard stacker-reclaimer operation path planning system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the digital coal yard stacker-reclaimer operation path planning system 300 can also be one of many hardware modules of the wireless terminal.

[0058] Alternatively, in another example, the digital coal yard stacker-reclaimer operation path planning system 300 and the wireless terminal can also be separate devices, and the digital coal yard stacker-reclaimer operation path planning system 300 can be connected to the wireless terminal via wired and / or wireless networks, and transmit interactive information in accordance with an agreed data format.

[0059] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for planning the operation path of a digital coal yard stacker-reclaimer, characterized in that, include: Acquire the original laser point cloud coordinate set collected by the 3D laser scanner, the current moisture content value of the material pile output by the microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical material pile records; By voxelization downsampling and spatial coordinate registration, geometric mesh construction and physical property coupling mapping are performed on the original laser point cloud coordinate set to obtain the initial topological mesh and physical physicochemical feature vector. Using the physical physicochemical feature vector as the boundary constraint, a spatiotemporal assessment of the collapse risk of the initial topological mesh is performed to obtain the spatiotemporal tensor of the collapse risk. Based on a preset slip safety threshold, the set of unstable nodes is extracted from the spatiotemporal tensor of collapse risk, and the corresponding nodes in the initial topological grid are subjected to spatial reprojection deformation and surface reconstruction to obtain the predicted evolution terrain grid. Based on the kinematic physical constraints of the stacker-reclaimer, contour line fitting trajectory optimization and curvature smoothing are performed on the outer envelope of the predicted evolution terrain grid to obtain the spatial path spline curve. Based on the spatiotemporal tensor of collapse risk, a four-dimensional dynamic mapping of kinematic command flow is performed on the spatial path spline curve to obtain the terminal execution command flow.

2. The digital coal yard stacker-reclaimer operation path planning method according to claim 1, characterized in that, By voxelization downsampling and spatial coordinate registration, the original laser point cloud coordinate set is subjected to geometric mesh construction and physical property coupling mapping to obtain the initial topological mesh and physical physicochemical feature vectors, including: The original laser point cloud coordinate set is divided into voxel spaces and dust noise is removed. The geometric center of the point cloud coordinates within each effective voxel is calculated to obtain the basic three-dimensional geometric mesh. The current moisture content of the stockpile, along with the compaction stress and particle size distribution matrix, are mapped to each vertex of the basic three-dimensional geometric mesh to obtain the initial topological mesh. Physical feature weighting and global attribute extraction are performed on the initial topological mesh to obtain the physical physicochemical feature vector.

3. The digital coal yard stacker-reclaimer operation path planning method according to claim 1, characterized in that, Using the physical physicochemical feature vectors as boundary constraints, a spatiotemporal assessment of the collapse risk is performed on the initial topological mesh to obtain the spatiotemporal tensor of collapse risk, including: The internal friction angle and cohesion of the moisture content component and compaction stress component in the physicochemical characteristic vector of the material are calculated to obtain the microscopic dynamic boundary parameters. The initial topological mesh is imported into a pre-trained discrete element reduced-order neural network surrogate model to instantiate the simulation environment and obtain a constrained coupled simulation mesh. Based on the preset bucket wheel slewing radius and single feed step size, a virtual cutting displacement vector is applied to the working contact surface of the constrained coupled simulation mesh, and the dynamic instability response of the cutting displacement vector is calculated using microscopic dynamic boundary parameters as constraint input to obtain the mesh node instability response matrix. Within a preset future time window, the transient slip response of each node in the grid node instability response matrix is ​​iteratively integrated over time steps and mapped with logistic probability to obtain the spatiotemporal tensor of collapse risk.

4. The digital coal yard stacker-reclaimer operation path planning method according to claim 1, characterized in that, Based on a preset slip safety threshold, an unstable node set is extracted from the spatiotemporal tensor of collapse risk. Spatial reprojection deformation and surface reconstruction are then performed on the corresponding nodes in the initial topological mesh to obtain a predicted evolving terrain mesh, including: Based on a preset slip safety threshold, the collapse probability values ​​of each node in the collapse risk spatiotemporal tensor are threshold-based and filtered to obtain the set of unstable node displacements. Based on the slip displacement vector of each node in the unstable node displacement set, spatial reprojection deformation is performed on the corresponding node and its neighboring nodes in the initial topological mesh to obtain a spatially deformed scattered point set. Two-dimensional projection topological connection and three-dimensional elevation centroid interpolation reconstruction are performed on the spatially deformed scattered point set to obtain the predicted evolution terrain mesh.

5. The digital coal yard stacker-reclaimer operation path planning method according to claim 1, characterized in that, Based on the kinematic physical constraints of the stacker-reclaimer, contour line fitting trajectory optimization and curvature smoothing are performed on the outer envelope surface of the predicted terrain mesh to obtain spatial path spline curves, including: The outer envelope surface of the predicted terrain mesh is extracted and kinematic constraint clipping is performed to obtain an effective operational envelope surface; Based on the preset single material picking step height, contour line trajectory points are extracted along the vertical direction on the effective working envelope to obtain a discrete spatial trajectory point set; B-spline curve fitting and curvature smoothing are performed on the discrete spatial trajectory point set to obtain the spatial path spline curve.

6. The digital coal yard stacker-reclaimer operation path planning method according to claim 5, characterized in that, The outer envelope surface of the predicted terrain mesh is extracted and kinematically constrained to obtain an effective operational envelope surface, including: Extract the outer topological envelope of the predicted evolving terrain mesh; Based on the gyration radius limit and pitch angle limit of the stacker-reclaimer, the boundary trimming of the inaccessible region of the outer topological envelope is performed to obtain the effective working envelope.

7. The digital coal yard stacker-reclaimer operation path planning method according to claim 1, characterized in that, Based on the spatiotemporal tensor of collapse risk, a four-dimensional dynamic mapping of the kinematic command flow is performed on the spatial path spline curve to obtain the terminal execution command flow, including: Based on the fixed control cycle of the underlying controller, the spatial path spline curve is discretized in the same time domain, and the coordinates of each discrete path point are back-mapped to the landslide risk spatiotemporal tensor through a trilinear interpolation algorithm to retrieve risk features, so as to obtain a set of discrete trajectory points carrying risk features. An adaptive velocity allocation calculation is performed on the comprehensive risk assessment value of each point in the discrete trajectory point set carrying risk characteristics to obtain a four-dimensional dynamic control sequence; Based on the register bit width limitation of the underlying controller, the four-dimensional dynamic control sequence is digitally translated and serialized to obtain the terminal execution instruction stream.

8. A digital coal yard stacker-reclaimer operation path planning system, characterized in that, include: The multimodal coal pile state data integration module is used to acquire the original laser point cloud coordinate set collected by the 3D laser scanner, the current pile moisture content value output by the microwave moisture meter, and the compaction stress and particle size distribution matrix generated based on historical pile records. The 3D mesh model and attribute coupling construction module is used to construct a geometric mesh and couple physical attributes onto the original laser point cloud coordinate set through voxelization downsampling and spatial coordinate registration, so as to obtain the initial topological mesh and physical physicochemical feature vector. The spatiotemporal dynamic assessment module for landslide risk is used to assess the spatiotemporal risk of the initial topological mesh using the physical physicochemical feature vector as the boundary constraint to obtain the spatiotemporal tensor of landslide risk. The instability prediction and terrain evolution simulation module is used to extract the set of unstable nodes from the spatiotemporal tensor of collapse risk based on a preset slip safety threshold, and to perform spatial reprojection deformation and surface reconstruction on the corresponding nodes in the initial topological grid to obtain the predicted evolution terrain grid. The smooth path planning module is used to perform contour line fitting trajectory optimization and curvature smoothing on the outer envelope surface of the predicted evolution terrain grid based on the kinematic physical constraints of the stacker-reclaimer to obtain the spatial path spline curve. The adaptive instruction generation module is used to perform four-dimensional dynamic mapping of kinematic instruction flow on spatial path spline curves based on the spatiotemporal tensor of collapse risk to obtain the terminal execution instruction flow.