A mine exploitation process training method and system based on BIM and VR technology

Abstract

This invention discloses a mining process training method and system based on BIM and VR technologies, belonging to the field of mine safety training technology. It includes applying a finite element-discrete element coupled algorithm to perform BIM modeling and geotechnical mechanics simulation on a coupled geological dataset, outputting a geomechanical digital twin. The geomechanical digital twin and standard employee job vectors are then input into a process decomposition model. The process logic layer performs process chain failure deduction, and the job interaction layer performs role-task matching, outputting a VR interactive process training scheme. This invention enhances the scientific rigor of the training scenario and the realism of disaster early warning education by integrating the finite element-discrete element coupled algorithm and the process decomposition model, thereby improving the operational standardization and emergency response capabilities of mine workers.

Description

Technical Field

[0001] This invention relates to the field of mine safety training technology, and in particular to a mine mining process training method and system based on BIM and VR technologies. Background Technology

[0002] With the development of information technology, Building Information Modeling (BIM) and Virtual Reality (VR) are increasingly being applied to the mining process training industry. BIM provides precise data support for process training through 3D modeling and data management, while VR enables an immersive interactive experience, allowing operators to conduct process training and practical exercises in a safe environment. The synergistic mechanism of BIM and VR lays the technological foundation for the digital transformation of mining process training.

[0003] However, existing BIM and VR-based mine training methods still have significant technical limitations. Firstly, existing BIM models focus on geometric modeling and static information integration, lacking dynamic simulation capabilities for geomechanical behavior. This leads to a disconnect between the virtual environment and the physical response characteristics of real mines, weakening the scientific rigor and early warning education function of process training. Secondly, current training content is mostly pre-set and standardized processes, lacking intelligent assessment and feedback mechanisms based on individual employee behavior. It cannot dynamically adjust the training difficulty and content according to employees' actual performance in the VR environment, making it difficult to achieve personalized and precise skills enhancement. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a mining process training method based on BIM and VR technologies to address the problems of lack of geomechanical simulation and insufficient ability to personalize the training process.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a mining process training method based on BIM and VR technologies, which includes collecting a coupled dataset of mine geology and employee job data in a collaborative task chain. The finite element-discrete element coupled algorithm is applied to perform BIM modeling and geotechnical mechanics simulation on the coupled dataset of mine geology, generating a geomechanical digital twin; feature encoding mapping is performed on employee job data to form a standard employee job vector; The geomechanical digital twin and standard employee job vectors are input into the process decomposition model. The process logic layer performs process chain failure deduction, the job interaction layer performs role task matching, and outputs a VR interactive process training program. The VR interactive process training solution is imported into the Unity simulation engine to build an immersive process training scenario. Based on the immersive process training scenario, employees are guided to conduct safety procedure training and virtual mining operations. Simultaneously, the simulated mining data of employees is monitored and verified against the corresponding BIM process elements, and the employee training logs are integrated and generated. Perform quantitative analysis of operational behaviors in employee training logs to generate operational compliance scores. Based on these scores, dynamically adjust the VR interactive process training plan to obtain personalized supplementary training plans. Then, send the personalized supplementary training plans and operational compliance scores back to the collaborative task chain, register the training process status, and retain approval and signature records.

[0007] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the mining geological coupling dataset includes three-dimensional point cloud of the ground surface, geological attribute data and mining geometric structure data.

[0008] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the generation of the geomechanical digital twin specifically includes the following steps. Spatial registration and topological association are performed on the three-dimensional point cloud of the ground surface, geological attribute data and mine geometric structure data to generate a three-dimensional structure of geological entities. The finite element-discrete element coupling algorithm is applied to perform multi-scale coupling domain partitioning and BIM parametric modeling of the three-dimensional structure of geological entities, and to obtain the geotechnical coupled BIM structure. Multi-field boundary condition loading and geotechnical mechanics simulation are performed on the coupled geotechnical BIM structure to output a geomechanical digital twin.

[0009] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the formation of standard employee job vectors specifically refers to performing feature encoding mapping on employee job data to obtain job feature sequences, and using principal component analysis to perform dimensional compression on the job feature sequences to form standard employee job vectors.

[0010] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the output VR interactive process training scheme specifically includes the following steps. The process logic layer is built using graph convolutional networks, and the job interaction layer is built using a Transformer encoder architecture. Cross-layer feature cascading and residual gating stacking are performed on the process logic layer and the job interaction layer to construct a process decomposition model; The geomechanical digital twin and standard employee job vectors are input into the process decomposition model. The process logic layer applies the fault tree analysis algorithm to perform process chain failure deduction on the geomechanical digital twin and generate a process risk heat map. The job interaction layer matches employees to roles and tasks based on standard employee job vectors and standard operating rules to obtain the standard flow of job operations. By aligning and weighting the process risk heat map and the standard operating procedures for each job position in time and space, a dynamic training logic diagram is formed. The dynamic training logic diagram is adapted for multiple terminals, and a VR interactive process training solution is output.

[0011] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the construction of an immersive process training scenario specifically includes the following steps. The VR interactive process training program was imported into the Unity simulation engine, and the Terrain tool was used to recreate the real-world mine surface to obtain a high-precision mine scene base. By embedding instructional videos on standardized operating procedures into a high-precision mining scenario base, an immersive process training scenario can be constructed.

[0012] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the generation of employee training logs specifically includes the following steps. In immersive process training scenarios, employees are trained in risk scenario identification and mining operation drills. After the training, employees conduct virtual mining operations through a physical simulation console. The simulation data is monitored in real time, and the simulation data is verified against BIM process elements to generate employee training logs.

[0013] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the step of generating an operational compliance score specifically includes the following steps. The isolated forest algorithm was used to analyze abnormal behavior patterns in employee training logs to obtain operational deviation characteristics. Risk probability calculations are performed on operational deviation characteristics to generate operational compliance scores.

[0014] As a preferred embodiment of the mining process training method based on BIM and VR technology described in this invention, the step of obtaining a personalized supplementary training plan specifically includes the following steps. Based on the operational compliance score, the VR interaction technology training program is adjusted in terms of scenario complexity to obtain a supplementary training needs list. The supplementary training needs list is then dynamically weighted and a personalized supplementary training program is output. Personalized retraining plans and operational compliance scores are sent back to the collaborative task chain via API interface; The collaborative task chain uses smart contracts to reach a consensus on personalized supplementary training plans and operational compliance scores, registers the status of the training process, and signs and stores the training process status, retaining approval and signing records.

[0015] Secondly, the present invention provides a mining process training system based on BIM and VR technologies, including a digital twin module for collecting mine geological coupled datasets and employee job data; applying the finite element-discrete element coupled algorithm to perform BIM modeling and geotechnical mechanics simulation on the mine geological coupled datasets, and outputting a geomechanical digital twin; The solution generation module is used to input the geomechanical digital twin and employee job data into the process decomposition model, the process logic layer performs process chain failure simulation, the job interaction layer performs role task matching, and outputs VR interactive process training solutions. The execution feedback module is used to import the VR interactive process training plan into the Unity simulation engine to build an immersive process training scenario; based on the immersive process training scenario, it guides employees to conduct safety procedure training and virtual mining operations, synchronously monitors employees' simulated mining data, and integrates and generates employee training logs. The training optimization module is used to quantitatively analyze the operational behavior of employees' training logs and generate operational compliance scores. Based on the operational compliance scores, the VR interactive process training program is dynamically adjusted and personalized supplementary training programs are output.

[0016] The beneficial effects of this invention are as follows: By integrating the finite element-discrete element coupled algorithm to perform BIM modeling and geotechnical mechanics simulation on a coupled dataset of mine geology, the virtual training environment can realistically reproduce the geomechanical simulation process during mining, significantly improving the scientific rigor of the training scenario and the realism of disaster early warning education. By generating VR interactive training programs through process decomposition models and dynamically adjusting the training content and complexity, the relevance and effectiveness of the training are significantly enhanced, effectively improving the operational standardization and emergency response capabilities of mine workers. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart for a training method for mining processes based on BIM and VR technologies.

[0019] Figure 2 This is a schematic diagram of a mining process training system based on BIM and VR technologies.

[0020] Figure 3 A flowchart for constructing a process decomposition model.

[0021] Figure 4 The flowchart for generating a geomechanical digital twin. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a mining process training method based on BIM and VR technologies, including the following steps: S1. Collect coupled datasets of mine geology and employee job data; apply the finite element-discrete element coupling algorithm to perform BIM modeling and geotechnical mechanics simulation on the coupled dataset of mine geology, and output a geomechanical digital twin.

[0026] Specifically, the steps include the following: S1.1 In the collaborative task chain, collect the coupled dataset of mine geology and employee job data.

[0027] In specific operations, In a BIM collaborative environment (such as the Autodesk BIM 360 platform), mining processes are associated and mapped with resource elements to obtain a set of discretized task nodes. Logical dependency relationships are then modeled on the set of discretized task nodes to generate a collaborative task chain. Smart contracts are used to configure trigger conditions for the collaborative task chain. For example, when the status attribute of the previous process in the collaborative task chain is updated to "completed", the start of the next process task is automatically triggered. It should be noted that logical dependency modeling refers to the process of defining state transitions and synchronizing concurrent processes for a set of discretized task nodes.

[0028] In the collaborative task chain, a coupled dataset of mine geology and employee job data are collected. Furthermore, the coupled dataset of mine geology reflects the spatial morphology and physical characteristics of the mine geological body, including three-dimensional point cloud of the surface, geological attribute data and mine geometric structure data. Among them, the three-dimensional point cloud of the surface is collected by a lidar scanner, the geological attribute data is collected by a geophysical surveying instrument, and the mine geometric structure data is collected by ground-penetrating radar.

[0029] Employee job data refers to the corresponding job positions of employees in the mining operations, including drilling, blasting, loading, transportation, and rock dumping positions; it is collected through enterprise human resource management software (such as Yonyou or Kingdee). The superior collaborative task chain can not only realize the automated flow and status tracking of process tasks, but also ensure the standardization and auditability of the training process through smart contracts. Compared with the existing linear training process, it has significant advantages in dynamic response and high execution efficiency.

[0030] S1.2 Preprocess the coupled dataset of mine geology and employee job data.

[0031] In specific operations, For 3D point clouds on the Earth's surface, spatial reference alignment is achieved through affine transformation to ensure spatial consistency. Noise suppression is performed through Gaussian filtering to improve the clarity of the 3D point clouds. Simultaneously, bilateral filtering is used for surface smoothing to enhance the continuity of the 3D point clouds.

[0032] For geological attribute data, Z-score standardization is used to unify the data structure and achieve consistency in the format of geological attribute data. Kriging interpolation is used to fill missing items to ensure the integrity of geological attribute data. Simultaneously, nonlinear dimensionality reduction is performed to optimize the feature representation of geological attribute data and the efficiency of subsequent calculations.

[0033] For mine geometric structure data, the Laplacian operator (called via the cv2.Laplacian function in the OpenCV library) is used for edge enhancement to ensure the clarity of the boundaries of the mine geometric structure data. Voxel downsampling is also performed to reduce data redundancy. Statistical filtering is used simultaneously to remove outliers, thereby improving the accuracy and reliability of the mine geometric structure data.

[0034] For employee job data, a sliding window is used to slide the data over time to ensure the integrity of the data features, and Min-Max scaling is used to standardize the values ​​and eliminate the influence of different units.

[0035] S1.3. Spatial registration and topological association are performed on the preprocessed 3D point cloud of the ground surface, geological attribute data and mine geometric structure data to generate a 3D structure of the geological entity.

[0036] In specific operations, The preprocessed 3D point cloud of the land surface is spatially registered with the mine's geometric structure data using the ICP (Iterative Closest Point) algorithm. Further, spatial benchmark alignment and preliminary data fusion are performed on the preprocessed 3D point cloud of the land surface and the mine's geometric structure data to obtain preliminary fused spatial data. An affine coordinate transformation is then performed on the preliminary fused spatial data to obtain the registration base data. Gaussian filtering and voxel resampling are then applied to the registration base data to improve smoothness and resolution, outputting the registered point cloud data.

[0037] Subsequently, the Delaunay triangulation algorithm was used to perform topological association on the registered point cloud data. Further, vertex connections and edge smoothing were performed on the registered point cloud data to construct a triangular mesh structure. Boundary extraction was performed on the triangular mesh structure to generate a basic topological framework. Holes in the basic topological framework were filled to generate a continuous geometric surface. Surface fitting and mesh refinement were performed on the continuous geometric surface to obtain a high-precision geological interface. Spatial interpolation was then performed on the high-precision geological interface to obtain the three-dimensional structure of the geological entity. It should be noted that void filling refers to the process of filling missing parts and closing surfaces in the basic topological framework; surface fitting refers to the process of approximating and smoothing the shape of a continuous geometric surface using polynomial functions.

[0038] S1.4. Apply the finite element-discrete element coupling algorithm to perform multi-scale coupling domain division and BIM (Building Information Modeling) parametric modeling of the three-dimensional structure of the geological entity, and obtain the geotechnical coupled BIM structure.

[0039] In specific operations, The three-dimensional structure of the geological entity is divided into multi-scale coupling domains. Further, a preset continuity threshold is used to divide the three-dimensional structure of the geological entity into continuous domains. For example, the fracture density parameter of the three-dimensional structure of the geological entity is extracted and spatially interpolated to obtain the fracture density distribution. The fracture density distribution is numerically normalized to generate a fracture density coefficient. When the fracture density coefficient is higher than the continuity threshold, the corresponding region in the three-dimensional structure of the geological entity is defined as a continuous coupling domain (finite element region). When the fracture density coefficient is lower than the continuity threshold, the corresponding region in the three-dimensional structure of the geological entity is defined as a discontinuous coupling domain (discrete element region). The continuous coupling domain is tetrahedral meshed to characterize the properties of intact rock strata, resulting in a finite element mesh. The discontinuous coupling domain is aggregated into polyhedral regions to characterize the properties of fractured rock mass, forming a discrete element set. It should be noted that the continuous threshold is based on the definition of rock mass integrity of the three-dimensional structure of the geological entity, and the exemplary value range is [0.6, 0.8]; tetrahedral meshing refers to the process of connecting nodes and dividing elements in a continuous coupled domain; polyhedral region aggregation refers to the process of combining fragments and reconstructing boundaries in a discontinuous coupled domain.

[0040] Next, BIM parametric modeling is performed on the finite element mesh and discrete element set. Further, the geometric topology of the finite element mesh and discrete element set is reconstructed to generate a finite element-discrete element coupled structure. The finite element-discrete element coupled structure is then extended with IFC standard attributes to obtain BIM attribute parameters. Based on geotechnical engineering classification rules, the mechanical properties and geological properties of the BIM attribute parameters are mapped to obtain a multiphysics BIM structure. Finally, the multiphysics BIM structure is rendered in 3D to output a geotechnical coupled BIM structure. It should be noted that the specific process of geometric topology reconstruction is as follows: vertex fusion and boundary connection are performed on the finite element mesh and discrete element set to obtain a unified geometric framework; topology reconstruction is performed on the unified geometric framework to generate a hybrid mesh structure; interface coupling is performed on the hybrid mesh structure to obtain the finite element-discrete element coupled structure. IFC standard attribute extension refers to the process of adding attribute fields and semantic associations to the finite element-discrete element coupled structure according to IFC standard rules (accuracy level definitions based on BIM modeling). Geotechnical engineering classification rules are defined based on the strength characteristics of rock mechanics and the degree of development of geological structures.

[0041] S1.5 Perform multi-field boundary condition loading and geotechnical mechanics simulation on the geotechnical coupled BIM structure, and output a geomechanical digital twin.

[0042] In specific operations, Multi-field boundary condition loading is applied to the geotechnical coupled BIM structure. Furthermore, according to the engineering geological investigation specifications, load case mapping is performed on the geotechnical coupled BIM structure to obtain a standardized load set. Dynamic parameter values ​​are assigned to the standardized load set. For example, the self-weight of the rock mass is set as a vertical uniformly distributed load, the groundwater pressure is set as a pore pressure field, and the engineering load is set as a surface contact force to obtain initial constraint boundary data. Spatial registration is performed on the initial constraint boundary data to generate multi-field constraint conditions. It should be noted that the engineering geological investigation specification is based on the relationship between the rock mass stability level and the mining depth of the mine; load condition mapping refers to the process of identifying load types and quantifying parameters of the geotechnical coupled BIM structure in accordance with the engineering geological design specification.

[0043] Next, based on multiple constraints, geotechnical mechanics simulation is performed on the geotechnical coupled BIM structure. Further, the geotechnical coupled BIM structure is input into the ANSYS simulation engine, and multiple constraints are applied to it using an explicit integrator to simulate stress propagation and crack propagation. Simultaneously, residual energy is detected on the geotechnical coupled BIM structure. When the residual energy reaches mechanical equilibrium, the simulation stops, and a digital twin foundation structure is output. The digital twin foundation structure is then rendered in real-time to generate a geomechanical digital twin. It should be noted that the mechanical equilibrium state refers to the state in which the residual energy tends to be stable, based on the definition of the Newton-Raphson iterative convergence criterion.

[0044] S1.6 Perform feature encoding mapping on employee job data to obtain job feature sequences, and use principal component analysis to perform dimensionality compression on the job feature sequences to form standard employee job vectors.

[0045] Entity recognition is performed on employee job data using natural language processing libraries (such as spaCy or NLTK) to extract key job features, such as "operating down-the-hole drill", "performing blasting alert", and "filling out equipment inspection forms". These key job features are then numerically encoded, for example, by using one-hot encoding to hash and map the key job features, transforming them into high-dimensional sparse vectors, i.e., job feature sequences. The job feature sequences completely preserve all the information from the original job descriptions. The StandardScaler from the Scikit-learn library is used to standardize the feature sequences of each job position, eliminating the influence of dimensions and obtaining a standardized feature matrix. The PCA algorithm is then applied to perform a linear transformation on the standardized feature matrix to obtain eigenvalues. The eigenvalues ​​are sorted in descending order, and the top k eigenvalues ​​are selected as principal components for projection dimensionality reduction, resulting in a dense vector representation, namely the standard employee job vector. The standard employee job vector not only significantly reduces the data dimensionality but also preserves the essential differences between different jobs to the greatest extent.

[0046] S2. Input the geomechanical digital twin and standard employee job vectors into the process decomposition model. The process logic layer performs process chain failure deduction, the job interaction layer performs role task matching, and outputs a VR interactive process training program.

[0047] Specifically, the steps include the following: S2.1 Construct a process decomposition model and train it.

[0048] In specific operations, In the PyTorch framework, a graph convolutional network is invoked using the `nn.GCNConv` parameter and initialized, for example, by setting the number of input channels to 128, the number of output channels to 64, and the dropout rate to 0.5. A fault tree analysis algorithm is then applied after the graph convolutional network to perform process chain failure prediction, enhancing the reliability of process dependencies. AyerNorm is used for feature normalization, completing the construction of the process logic layer. The Transformer encoder architecture is invoked using the `nn.Transformer` parameter and initialized, for example, by setting the number of heads to 8, the key dimension to 64, and the hidden layer dimension to 512. Standard operating rules are embedded into the Transformer encoder architecture to constrain the compliance of job interactions, and gradient stabilization is performed through residual connections, completing the construction of the job interaction layer. Cross-layer feature concatenation is performed on the process logic layer and the job interaction layer to obtain a fused feature vector; the fused feature vector is transformed by a fully connected layer to generate higher-order interaction features; linear projection is performed on the higher-order interaction features to generate attention weights; based on the attention weights, residual gating is performed on the process logic layer and the job interaction layer to complete the construction of the process decomposition model. Next, the process decomposition model is trained. Further, the historical geomechanical digital twin and standard employee job vectors are divided into a sample set, a training set, and a validation set. On the sample set, Xavier initialization is used for parameter pre-assignment, and a cosine annealing scheduler is used for learning rate warm-up to form initial process decomposition model parameters. On the training set, stochastic gradient descent is used to backpropagate and optimize the initial process decomposition model parameters, while gradient truncation is applied simultaneously for numerical stabilization to obtain optimized process decomposition model parameters. On the validation set, an early stopping mechanism is used to monitor the loss value fluctuations of the optimized process decomposition model parameters to obtain training stability data. When the training stability data exceeds the convergence threshold for several consecutive rounds (e.g., 10 times), training terminates, and the trained process decomposition model is output simultaneously. It should be noted that the convergence threshold is defined based on the rate of change of loss of the training stability data, and the exemplary range is [1e-5, 1e-4].

[0049] S2.2 The process logic layer applies fault tree analysis algorithm to perform process chain failure simulation on the geomechanical digital twin and generates a process risk heat map.

[0050] In specific operations, Historical geomechanical digital twins and employee job data are input into the trained process decomposition model via the Apache NiFi pipeline; The process logic layer extracts multi-scale features from the geomechanical digital twin through a three-layer graph convolutional network. Furthermore, the first layer extracts the geological stress data of the geomechanical digital twin and performs neighborhood aggregation through graph convolution kernels to obtain stress gradient features. Z-score standardization is performed on the stress gradient features to obtain stress gradient coefficients. The second layer performs spatial uniform sampling on the geomechanical digital twin to generate fracture grid nodes; fracture topology propagation is performed on the fracture grid nodes to obtain fracture connectivity features; time series difference is performed on the fracture connectivity features to obtain the fracture propagation rate. The third layer extracts geological displacement data from the geomechanical digital twin and uses a 3D convolution kernel to perform spatial convolution on the geological displacement data to obtain geological displacement features; then, a sliding window average is performed on the geological displacement features to obtain displacement change parameters. By integrating stress gradient coefficient, crack propagation rate, and displacement change parameters, key mechanical parameters can be obtained.

[0051] Based on the key mechanical parameters, a fault tree structure is constructed using a fault tree analysis algorithm. Furthermore, the key mechanical parameters are judged to be in abnormal state by using anomaly thresholds. For example, when a key mechanical parameter exceeds the anomaly threshold, the key mechanical parameter is judged as an abnormal data point. The abnormal data points are integrated to obtain a set of process abnormal events. At the same time, the probability distribution of the key mechanical parameters is calculated to generate the probability of occurrence of basic events. A fault tree structure is constructed by using the set of process anomaly events as the base events and the probability of occurrence of the basic events as the input of the logic gates. The mathematical formulas for calculating the probability distribution of key mechanical parameters are as follows: ; in, Indicates the probability of a basic event occurring. This represents the total number of samples for key mechanical parameters. Sample index representing key mechanical parameters. Indicates the state indicator operator, Indicates the first The mechanical parameter values ​​of each sample, Indicates the reference base value; It should be noted that the anomaly threshold is defined based on the statistical quantile of historical key mechanical parameters, with an exemplary value range of [10MPa, 50MPa]; the state indicator operator is used to adjust the anomaly response intensity of key mechanical parameters, based on the Herveside step function definition, with an exemplary value range of [0,1]; the mechanical parameter values ​​are obtained by performing numerical normalization on the key mechanical parameters; the reference benchmark value is defined based on the critical value of the mechanical parameters for rock mass stability.

[0052] Based on the fault tree structure, process chain failure is deduced. Further, logical path decomposition is performed on the fault tree structure to obtain abnormal event combinations. Density clustering is performed on the abnormal event combinations to obtain a set of failure modes. Monte Carlo sampling is performed on the set of failure modes to generate a process chain failure probability distribution. Spatial probability mapping is performed on the process chain failure probability distribution to output a process risk heatmap. It should be noted that density clustering refers to the process of neighborhood search and weighted aggregation of combinations of anomalous events; Monte Carlo sampling refers to the process of random sampling and probability statistics of a set of failure modes.

[0053] S2.3 The job interaction layer matches employees to roles and tasks based on the standard employee job vector and obtains the standard flow of job operations through standard operating rules.

[0054] In specific operations, The Transformer encoder architecture is used to embed features into the standard employee job vector. Furthermore, multi-head attention is applied to the standard employee job vector to obtain weighted job features. Fully connected layer projection is performed on the weighted job features to obtain standardized job representations. Global average pooling is then applied to the standardized job representations to compress the feature dimensions and obtain comprehensive job parameters. Next, role-task matching is performed using standard operating procedures (SOPs). Furthermore, the job's comprehensive parameters and SOPs are semantically compared line by line to obtain matching similarity. When the matching similarity exceeds the standard range, the match is considered successful, and preliminary task allocation data is output. The preliminary task allocation data undergoes permission conflict verification to ensure that mutually exclusive roles such as blasters and safety supervisors are not assigned conflicting tasks, and compliant role task data is output. The compliant role task data is then bound to spatiotemporal coordinates to generate a standard workflow for job operations. It should be noted that the standard operating procedure rules are based on the atomic task definition of business process decomposition; the standard range is based on the confidence interval definition of historical matching similarity, with an exemplary value range of [0.82, 0.95]; permission conflict verification refers to the operation of performing permission cross-checking and conflict item removal on the initial task allocation data.

[0055] S2.4. The process risk heat map and job operation standard flow are spatiotemporally aligned and weighted to form a dynamic training logic diagram, and multi-terminal adaptation is performed to output a VR interactive process training solution.

[0056] In specific operations, The process risk heatmap and job operation standard flow are spatiotemporally aligned. Furthermore, the process risk heatmap and job operation standard flow are mapped to a unified coordinate system and K-nearest neighbor sampling (e.g., K=3) is performed to obtain risk-task related data pairs. Bilinear interpolation is performed on the risk-task related data pairs to obtain spatial correlation feature vectors. Next, the information entropy and weights of the spatial correlation feature vectors are calculated using the entropy weight method to obtain the coupling weights between risk factors and operational complexity. The specific mathematical formula is as follows. ; in, The weight representing the coupling between risk factors and operational complexity. This represents the theoretical maximum value of information entropy. Indicates the first Information entropy of spatially correlated feature vectors This represents the total number of spatially correlated feature vectors. The index representing the spatial correlation feature vector. Indicates the first Information entropy of spatially correlated feature vectors; It should be noted that the theoretical maximum value of information entropy is obtained by performing limit entropy state deduction on spatially correlated feature vectors.

[0057] Based on the coupling weight of risk factors and operational complexity, feature fusion is performed on the process risk heatmap and the standard operation flow of the job to obtain a weighted training feature set; the weighted training feature set is then processed in a process-oriented manner to generate a training process task chain; the training process task chain is then logically encoded to obtain a dynamic training logic diagram; It should be noted that logical relation encoding refers to the process of defining state transitions and embedding constraints in the training process task chain.

[0058] Next, the dynamic training logic diagram is adapted for multiple terminals. Furthermore, the dynamic training logic diagram undergoes resource lightweighting to reduce computational load, resulting in an optimized logic diagram. The rendering parameters of the optimized logic diagram are adjusted; for example, the geometric complexity is adjusted to no more than 500,000 units, and the texture resolution is adjusted to 2048×2048 to ensure smooth operation on mobile VR devices, thus obtaining an adaptation resource package. The adaptation resource package is then packaged for terminal configuration and output as a VR interaction technology training solution.

[0059] S3. Import the VR interactive process training solution into the Unity simulation engine to build an immersive process training scenario; based on the immersive process training scenario, guide employees to conduct safety procedure training and virtual mining operations, simultaneously monitor employee simulated mining data, and integrate and generate employee training logs.

[0060] Specifically, the steps include the following: S3.1. Import the VR interactive process training program into the Unity simulation engine, and use the Terrain tool to recreate the real-world mining landscape and obtain a high-precision mining scene base.

[0061] In specific operations, The VR interactive technology training program is imported into the Unity simulation engine through a resource pipeline interface (such as Unity Asset) for data format conversion to obtain Unity-compatible scene data; the coordinate system of the Unity-compatible scene data is aligned to obtain spatial calibration data; scene element orchestration is performed on the spatial calibration data to build a standardized scene framework; It should be noted that scene element arrangement refers to the process of spatial layout planning and functional logic association of spatial calibration data.

[0062] Next, the Heightmap function of the Terrain tool is used to recreate the real-world scene of the mine. Furthermore, the standardized scene framework and the geotechnical coupled BIM structure are spatially overlaid to unify material properties and physical parameters, thereby obtaining a composite geological scene. The terrain mesh of the composite geological scene is refined to obtain interactive geological entities. Channel shading rendering is performed on the interactive geological entities to output a high-precision mine scene base. It should be noted that spatial overlay refers to the process of superimposing coordinate system and spatial coordinates of the standardized scene framework and the geotechnical coupled BIM structure.

[0063] S3.2 Embed instructional videos of standardized job operations into a high-precision mining scenario base to create an immersive process training scenario.

[0064] In specific operations, Using the Video Player component of the Unity simulation engine, pre-recorded videos explaining the standardized operation of the job are imported into a high-precision mine scene base. Then, using the Render Texture mapper, the videos explaining the standardized operation of the job are mapped onto the virtual display screen of the high-precision mine scene base. Next, the video playback and scene event triggering are precisely synchronized using the Timeline tool of the Unity simulation engine. For example, when the trainee's perspective enters the blasting warning zone in the high-precision mine scene base, the "Safe Evacuation Procedure" video is automatically played. The Spatial Audio component is used to add directional audio to the job operation instruction video to ensure that the direction of the instruction sound source is consistent with the position of the virtual display screen, thus completing the construction of an immersive process training scene.

[0065] S3.3. In immersive process training scenarios, employees are trained in risk scenario identification and mining operation drills. After the training, employees conduct virtual mining operations through a physical simulation console. The simulated mining data is monitored synchronously, and the simulated mining data is weighted and aggregated to generate employee training logs.

[0066] In specific operations, Employees wear VR headsets in an immersive process training scenario; they are guided through risk area identification training by spatial audio prompts, and mining operation drills are conducted using force feedback handles; after the training, employees conduct equipment operation simulations and emergency response drills through a physical simulation control panel, completing virtual mining operations.

[0067] After the virtual mining operation is completed, the employee's eye movement trajectory is monitored using an eye tracker, and the operation duration is monitored using a high-precision timer. Simultaneously, motion analysis software (such as Qualisys Track Manager) is used to monitor motion accuracy and integrate and generate simulated mining data. The employee's simulated mining data is then verified against BIM process elements. Furthermore, process parameters (such as borehole design inclination angle and blasting hole depth threshold) in the BIM process elements are extracted and logically verified against the actual parameters (such as virtual drill rod angle and charge amount) in the employee's simulated mining data to obtain an original verification feature set containing deviation type, deviation amount, and compliance status. The original verification feature set is then logically structured and formatted, and the employee training log is output. It should be noted that BIM process elements are defined based on the property set definition of the IFC standard; logical verification rules refer to the operation process of comparing the similarity of the process parameters in the BIM process elements with the actual parameters in the employee simulated mining data and determining the logical conditions.

[0068] S4. Conduct quantitative analysis of employee training logs to generate operational compliance scores; based on these scores, dynamically adjust the initial process training plan and output personalized supplementary training plans.

[0069] Specifically, the steps include the following: S4.1 Use the Isolation Forest algorithm to analyze abnormal behavior patterns in employee training logs, obtain operational deviation characteristics, calculate the risk probability of operational deviation characteristics, and form an operational compliance score.

[0070] In specific operations, The isolated forest algorithm is applied to analyze abnormal behavior patterns in employee training logs. Further, the spatiotemporal operation sequences of the employee training logs are extracted and feature dimensionality reduction is performed to generate low-dimensional feature projections. Density clustering is performed on the low-dimensional feature projections to obtain abnormal operation clusters. Feature inversion is performed on the abnormal operation clusters to obtain abnormal deviation patterns. The abnormal deviation patterns are structured and encoded to output operation deviation features. It should be noted that density clustering refers to the process of density peak detection and cluster center merging on low-dimensional feature projection; the specific process of feature inversion for anomalous operation clusters is as follows: dimensional expansion of the anomalous operation clusters to obtain anomalous operation behavior vectors; trajectory reconstruction of the anomalous operation behavior vectors to generate anomalous behavior sequences; deviation quantification of the anomalous behavior sequences to obtain constant deviation patterns; and structured encoding refers to the process of semantic association and format conversion of anomalous deviation patterns.

[0071] Next, the risk probability of operational deviation characteristics is calculated to form an operational compliance score. The specific mathematical format is as follows. ; in, Indicates the operational compliance score. This represents the total number of operational deviation characteristics. Indicates the operational deviation characteristic index. Indicates the first Weighting coefficients for each operational deviation feature. Indicates the first Compliance rate of each operational deviation characteristic; It should be noted that the weighting coefficient is defined based on the accident correlation degree of historical operational deviation characteristics, and the exemplary value range is [0.1, 1.0]; the compliance rate is obtained by performing frequency statistics and probability normalization on the operational deviation characteristics.

[0072] S4.2 Based on the operational compliance score, adjust the scenario complexity of the VR interactive technology training program, obtain a supplementary training requirement list, dynamically assign weights to the supplementary training requirement list, and output a personalized supplementary training program.

[0073] In specific operations, Based on operational compliance score ( The VR interactive technology training program is classified by scenario complexity. For example, the initial virtual training operation scenario in the VR interactive technology training program is extracted, and the range of difficulty level 1 is set as [0, 0.6], the range of difficulty level 2 is [0.6, 0.8], and the range of difficulty level 3 is [0.8, 1]. When S is in the range of difficulty level 1, the initial virtual training operation scenario is defined as a high-risk scenario; when S is in the range of difficulty level 2, the initial virtual training operation scenario is defined as a medium-risk scenario; and when S is in the range of difficulty level 3, the initial virtual training operation scenario is defined as a low-risk scenario. It should be noted that the difficulty range is defined based on the statistical distribution of historical operational compliance scores.

[0074] Subsequently, supplementary training needs are matched according to the complexity of the scenario. For example, supplementary training content to enhance high-risk operations is matched for high-risk scenarios (such as blasting safety procedures and equipment emergency operations); supplementary training content to improve error-prone links is matched for medium-risk scenarios (such as shovel loading positioning accuracy training); and supplementary training content to consolidate standard procedures is matched for low-risk scenarios (such as equipment daily maintenance training). The operation process data of the VR interactive process training plan is extracted and combined with the supplementary training content to construct a supplementary training needs list.

[0075] The supplementary training requirement list is dynamically weighted. Further, the training frequency data of the supplementary training requirement list is extracted and normalized to obtain a training intensity value. When the training intensity value is below the importance threshold, it is determined to be a regular requirement and assigned a lower weight (e.g., 0.3). When the training intensity value is above the importance threshold, it is determined to be an urgent requirement and assigned a higher weight (e.g., 0.7). Based on the assigned weights, the supplementary training requirement list is weighted, fused, and formatted to output a personalized supplementary training plan. It should be noted that the importance threshold is defined based on the operational error rate of the historical supplementary training requirement list, and the exemplary value range is [0.4, 0.6]. Personalized retraining plans and operational compliance scores are fed back to the collaborative task chain via API interfaces (such as RESTful APIs). The collaborative task chain uses smart contracts to perform consensus verification and status updates on the personalized retraining plans and operational compliance scores, obtaining the training process status. The training process status is then recorded in a distributed ledger for easy auditing and traceability. The training process status is hashed using an asymmetric encryption algorithm to obtain a digital fingerprint, and multi-party private key signing is performed on the digital fingerprint to generate an approval signing record, which is stored on the blockchain to prevent the record from being tampered with. It should be noted that smart contracts achieve consensus verification by matching personalized retraining plans and operational compliance scores according to rules; asymmetric encryption algorithms achieve hash conversion by generating message digests and fixed-length outputs of the training process state.

[0076] This embodiment also provides a mining process training system based on BIM and VR technologies, including: The digital twin module is used to collect coupled mining geological datasets and employee job data in a collaborative task chain; apply the finite element-discrete element coupled algorithm to perform BIM modeling and geotechnical mechanics simulation on the coupled mining geological dataset to generate a geomechanical digital twin; and perform feature encoding mapping on the employee job data to form a standard employee job vector. The solution generation module is used to input the geomechanical digital twin and standard employee job vectors into the process decomposition model. The process logic layer performs process chain failure deduction, the job interaction layer performs role task matching, and outputs a VR interactive process training solution. The execution feedback module is used to import the VR interactive process training plan into the Unity simulation engine, build an immersive process training scenario, guide employees to conduct safety procedure training and virtual mining operations based on the immersive process training scenario, synchronously monitor the employees' simulated mining data and verify it with the corresponding BIM process elements, and integrate and generate employee training logs. The training optimization module is used to quantitatively analyze the operational behavior of employees' training logs, generate operational compliance scores, dynamically adjust the VR interactive process training plan based on the operational compliance scores, obtain personalized supplementary training plans, and send the personalized supplementary training plans and operational compliance scores back to the collaborative task chain, register the training process status, and retain approval and signature records.

[0077] In summary, this invention utilizes a finite element-discrete element coupled algorithm to perform BIM modeling and geotechnical mechanics simulation on a coupled dataset of mine geology. This enables the virtual training environment to realistically reproduce the geomechanical simulation process during mining, significantly enhancing the scientific rigor of the training scenario and the realism of disaster early warning education. By generating VR interactive training programs through process decomposition models and dynamically adjusting the training content and complexity, the invention significantly enhances the relevance and effectiveness of the training, effectively improving the operational standardization and emergency response capabilities of mine workers.

[0078] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A training method for mining processes based on BIM and VR technologies, characterized in that: include, In the collaborative task chain, mine geological coupled datasets and employee job data are collected; The finite element-discrete element coupled algorithm is applied to perform BIM modeling and geotechnical mechanics simulation on the coupled dataset of mine geology, generating a geomechanical digital twin; feature encoding mapping is performed on employee job data to form a standard employee job vector; The geomechanical digital twin and standard employee job vectors are input into the process decomposition model. The process logic layer performs process chain failure deduction, the job interaction layer performs role task matching, and outputs a VR interactive process training program. The VR interactive process training solution is imported into the Unity simulation engine to build an immersive process training scenario. Based on the immersive process training scenario, employees are guided to conduct safety procedure training and virtual mining operations. Simultaneously, the simulated mining data of employees is monitored and verified against the corresponding BIM process elements, and the employee training logs are integrated and generated. Perform quantitative analysis of operational behaviors in employee training logs to generate operational compliance scores. Based on these scores, dynamically adjust the VR interactive process training plan to obtain personalized supplementary training plans. Then, send the personalized supplementary training plans and operational compliance scores back to the collaborative task chain, register the training process status, and retain approval and signature records.

2. The mining process training method based on BIM and VR technology as described in claim 1, characterized in that: The coupled mining geological dataset includes three-dimensional point clouds of the surface, geological attribute data, and mining geometric structure data.

3. The mining process training method based on BIM and VR technology as described in claim 2, characterized in that: The generation of the geomechanical digital twin specifically includes the following steps. Spatial registration and topological association are performed on the three-dimensional point cloud of the ground surface, geological attribute data and mine geometric structure data to generate a three-dimensional structure of geological entities. The finite element-discrete element coupling algorithm is applied to perform multi-scale coupling domain partitioning and BIM parametric modeling of the three-dimensional structure of geological entities, and to obtain the geotechnical coupled BIM structure. Multi-field boundary condition loading and geotechnical mechanics simulation are performed on the coupled geotechnical BIM structure to output a geomechanical digital twin.

4. The mining process training method based on BIM and VR technology as described in claim 1, characterized in that: The formation of the standard employee job vector specifically refers to performing feature encoding mapping on employee job data to obtain job feature sequences, and then using principal component analysis to perform dimensionality compression on the job feature sequences to form the standard employee job vector.

5. The mining process training method based on BIM and VR technology as described in claim 4, characterized in that: The output VR interactive technology training solution specifically includes the following steps. The process logic layer is built using graph convolutional networks, and the job interaction layer is built using a Transformer encoder architecture. Cross-layer feature cascading and residual gating stacking are performed on the process logic layer and the job interaction layer to construct a process decomposition model; The geomechanical digital twin and standard employee job vectors are input into the process decomposition model. The process logic layer applies the fault tree analysis algorithm to perform process chain failure deduction on the geomechanical digital twin and generate a process risk heat map. The job interaction layer matches employees to roles and tasks based on standard employee job vectors and standard operating rules to obtain the standard flow of job operations. By aligning and weighting the process risk heat map and the standard operating procedures for each job position in time and space, a dynamic training logic diagram is formed. The dynamic training logic diagram is adapted for multiple terminals, and a VR interactive process training solution is output.

6. The mining process training method based on BIM and VR technology as described in claim 1, characterized in that: The construction of the immersive craft training scenario specifically includes the following steps. The VR interactive process training program was imported into the Unity simulation engine, and the Terrain tool was used to recreate the real-world mine surface to obtain a high-precision mine scene base. By embedding instructional videos on standardized operating procedures into a high-precision mining scenario base, an immersive process training scenario can be constructed.

7. The mining process training method based on BIM and VR technology as described in claim 6, characterized in that: The generation of employee training logs specifically includes the following steps. In immersive process training scenarios, employees are trained in risk scenario identification and mining operation drills. After the training, employees conduct virtual mining operations through a physical simulation console. The simulation data is monitored in real time, and the simulation data is verified against BIM process elements to generate employee training logs.

8. The mining process training method based on BIM and VR technology as described in claim 1, characterized in that: The process of generating an operational compliance score specifically includes the following steps. The isolated forest algorithm was used to analyze abnormal behavior patterns in employee training logs to obtain operational deviation characteristics. Risk probability calculations are performed on operational deviation characteristics to generate operational compliance scores.

9. The mining process training method based on BIM and VR technology as described in claim 1, characterized in that: The process of obtaining a personalized retraining plan includes the following steps: Based on the operational compliance score, the VR interaction technology training program is adjusted in terms of scenario complexity to obtain a supplementary training needs list. The supplementary training needs list is then dynamically weighted and a personalized supplementary training program is output. Personalized retraining plans and operational compliance scores are sent back to the collaborative task chain via API interface; The collaborative task chain uses smart contracts to reach a consensus on personalized supplementary training plans and operational compliance scores, registers the status of the training process, and signs and stores the training process status, retaining approval and signing records.

10. A mining process training system based on BIM and VR technology, based on the mining process training method based on BIM and VR technology as described in any one of claims 1 to 9, characterized in that: include, The digital twin module is used to collect coupled datasets of mine geology and employee job data in a collaborative task chain; The finite element-discrete element coupled algorithm is applied to perform BIM modeling and geotechnical mechanics simulation on the coupled dataset of mine geology, generating a geomechanical digital twin; feature encoding mapping is performed on employee job data to form a standard employee job vector; The solution generation module is used to input the geomechanical digital twin and standard employee job vectors into the process decomposition model. The process logic layer performs process chain failure deduction, the job interaction layer performs role task matching, and outputs a VR interactive process training solution. The execution feedback module is used to import the VR interactive process training plan into the Unity simulation engine, build an immersive process training scenario, guide employees to conduct safety procedure training and virtual mining operations based on the immersive process training scenario, synchronously monitor the employees' simulated mining data and verify it with the corresponding BIM process elements, and integrate and generate employee training logs. The training optimization module is used to quantitatively analyze the operational behavior of employees' training logs, generate operational compliance scores, dynamically adjust the VR interactive process training plan based on the operational compliance scores, obtain personalized supplementary training plans, and send the personalized supplementary training plans and operational compliance scores back to the collaborative task chain, register the training process status, and retain approval and signature records.

Images