Strip mine autonomous mining simulation system based on digital twinning

Abstract

The invention relates to the technical field of intelligent mine mining and automatic driving simulation testing, in particular to a digital twinning-based strip mine autonomous mining simulation system, which comprises a global dynamic geological map construction unit for maintaining a dynamically evolved geological database; the follow-up voxel field interaction unit outputs force feedback and deformation signals; the vehicle-road physical attribute bidirectional coupling and memory unit is used for realizing environment remodeling; the virtual-real deviation online self-calibration unit is used for reversely adjusting the element attributes; the predictive ecological scheduling planning unit is used for generating a re-planning instruction; the system further integrates a multi-dimensional high-fidelity twinborn body construction unit, a digital twinborn model unit, an autonomous mining decision-making unit and a virtual-real interaction and control unit. By constructing a real-time interactive digital twinning system, the problems that in traditional mining engineering, simulation neglects the time-varying characteristics of rock and soil media and is lack of a safe trial and error environment are solved, and stride from static design to dynamic intelligent operation of strip mine mining is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent mining and autonomous driving simulation testing technology, specifically to an open-pit mine autonomous mining simulation system based on digital twins. Background Technology

[0002] In the current open-pit mine autonomous mining simulation environment, the simulation system usually needs to build a virtual mining area scenario to test the perception, decision-making and control algorithms of mining and unmanned mining trucks; To achieve this goal, existing solutions generally simplify the terrain environment as a static rigid surface and use a memoryless instantaneous collision model for dynamic calculation. While such solutions can meet basic geometric passability verification, they neglect the multi-source heterogeneous data and time-varying characteristics of soil and rock media during open-pit mining. They cannot dynamically characterize physical properties such as soil water saturation and looseness coefficient, nor can they simulate cumulative physical changes such as rainfall infiltration, water evaporation, and rutting and hardening caused by repeated vehicle traffic. This static approach results in a lack of state memory in the simulation environment, preventing virtual vehicles from receiving realistic mechanical feedback such as getting stuck in mud, slipping, or bumping. This leads to significant discrepancies between simulation results and physical reality, severely hindering the seamless integration and collaborative optimization of long-term mining and short-term dynamic scheduling. Therefore, constructing a dynamic evolution model with physical fidelity and bidirectional interactive capabilities to address simulation distortion caused by neglecting the time-varying characteristics of the medium has become an urgent technical problem to solve. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides an autonomous open-pit mining simulation system based on digital twins. Specifically, the technical solution of this invention includes: The autonomous mining decision-making unit is configured to use reinforcement learning and operations research algorithms in a virtual environment to simulate and compare mining schemes, optimize mining plans, equipment scheduling and blasting parameters, and generate decision schemes to predict production bottlenecks. The virtual-real interaction and control unit is configured to enable real-time interaction between the digital twin model and the real equipment through Internet of Things (IoT) technology, predict and diagnose equipment failures, and provide a testing and training environment for unmanned mining trucks and remotely controlled drilling rigs. The multidimensional high-fidelity twin building blocks are configured to integrate geological exploration data, UAV mapping data and equipment parameters to create a digital twin model of an open-pit mine that includes geological structure, topography, mining equipment and production environment. The digital twin model unit is configured to establish a dynamic model containing ore bodies, equipment and processes based on physical modeling and machine learning algorithms to reflect the physical state and operating rules of open-pit mines, and to support multi-scale collaborative operation simulation from individual equipment to the entire mining area. The predictive ecological scheduling planning unit is configured to call the above units for ultra-real-time simulation based on the mine truck traffic; monitor road surface health indicators; and generate early warning replanning instructions when the road surface health indicators are lower than the passable threshold. The virtual-real deviation online self-calibration unit is configured to collect real vehicle operation data; perform residual analysis on the response of a virtual vehicle at the same location; and when the residual exceeds the fitting tolerance, use an iterative optimization algorithm to reverse adjust the voxel physical properties in the global dynamic geological map construction unit. The global dynamic geological map construction unit is configured to segment the mining area space into voxels based on geological exploration data and UAV scanning data; assign physical attribute labels to voxels; configure read and write interfaces to receive updated data; and maintain a global geological database that evolves dynamically over time. The servo voxel field interaction unit is configured to extract voxel data within a preset range around the geometric center of the virtual vehicle as the anchor point; construct a servo voxel field; convert the physical attribute labels into calculation parameters through a soil mechanics parameter mapping model; perform ground mechanics calculations in conjunction with vehicle state data; and output road reaction force signals and medium deformation signals. The vehicle-road physical property bidirectional coupling and memory unit, which connects the follow-up voxel field interaction unit and the global dynamic geological map construction unit, is configured to receive the medium deformation signal; perform positive environmental reshaping to convert the deformation into voxel property change amount; and force write the voxel property change amount back to the global dynamic geological map construction unit through the read-write interface.

[0004] Preferably, the physical attribute labels include water saturation, initial shear modulus, and looseness coefficient; the water saturation defines the proportion of water inside the soil to characterize soil viscosity; the looseness coefficient defines the compactness of soil particles per unit volume to calculate compaction settlement; and the read / write interface is configured to allow real-time modification of the attribute values ​​of specific voxels.

[0005] Preferably, the soil mechanics parameter mapping model is configured to map the physical property labels into the computational parameters required by the Bekker-Wong model; the computational parameters include cohesion, internal friction angle, and soil deformation modulus; the servo voxel field interaction unit is configured to perform real-time approximate solution of the Bekker-Wong model based on the vehicle's instantaneous axle load, tire tread parameters, and the computational parameters; the pavement reaction force signal includes rolling resistance and shear resistance; the medium deformation signal includes settlement depth and shear displacement.

[0006] Preferably, the specific logic for the vehicle-road physical property bidirectional coupling and memory unit to perform positive environmental reshaping includes: converting the subsidence depth in the medium deformation signal into the reduction amount of voxel height; converting the compaction effect into the reduction amount of the voxel loosening coefficient; after the voxel property change amount is written back, triggering the follow-up voxel field interaction unit to read the updated physical properties to generate a dynamic response that evolves with the road surface state.

[0007] Preferably, the real vehicle operating data includes wheel speed difference and suspension compression; the iterative optimization algorithm is configured to calculate the sensitivity of the vehicle response residual to soil physical properties, generate environmental parameter correction factors with dynamic step size according to the direction of residual decrease, and perform re-simulation verification within the current time step; the fitting tolerance is an allowable error range based on the statistical analysis of historical high-precision test data.

[0008] Preferably, the road surface health index is a comprehensive evaluation parameter derived by weighting the average rut depth and rolling resistance coefficient; the passability threshold is a value set based on the vehicle chassis minimum ground clearance and maximum climbing power; the ultra-real-time simulation is configured to accelerate the simulation process in an independent sandbox copy of the main database to predict the road surface condition within a specific future time window.

[0009] Preferably, the early warning replanning instruction includes a load balancing strategy; the load balancing strategy includes forcing subsequent heavy-load vehicles to detour to protect the road surface, or reducing the load limit of passing vehicles; the early warning replanning instruction is configured to be output to the mine dispatch center.

[0010] Preferably, the digital twin model unit is further configured to simulate hazardous working conditions, including equipment collisions, slope instability, and personnel intrusion; the virtual-real interaction and control unit is further configured to provide early warning and visual prompts for the hazardous working conditions, and to support remote takeover.

[0011] This invention constructs an autonomous open-pit mining simulation system that operates in parallel with and interactively maps to a physical mine. This system enables a shift from experience-driven to data- and model-driven mining activities, and from passive response to proactive control. It provides a novel technological path for safe, efficient, green, and intelligent open-pit mining. Compared with existing technologies, this invention offers the following advantages: 1. Improve mining efficiency and enhance safety; improve equipment operation cycle efficiency and ore recovery rate through autonomous path planning, intelligent collaborative scheduling of multiple equipment, real-time simulation and optimization decision-making; digital twin model can simulate dangerous working conditions, provide early warning and visual prompts for risks such as equipment collision, slope instability and personnel intrusion, and assist in the formulation of safety plans.

[0012] 2. Reduce trial and error costs and improve management decision-making; verify solutions in a virtual environment to avoid trial and error costs in actual production; new equipment or processes can be tested in a digital twin model before being applied to actual production; at the same time, it provides mine managers with an intuitive visualization interface to support multi-dimensional data analysis and decision-making.

[0013] 3. This invention overcomes the limitation of traditional simulations where terrain is a static rigid body by constructing a global dynamic geological map and a two-way coupling unit for vehicle-road physical properties. The system establishes an environmental memory mechanism that can convert the subsidence and compaction effects caused by vehicle operation into voxel property changes and write them back to the database, presenting cumulative physical changes. This symbiotic evolution mechanism of vehicles changing the environment and the environment reacting to vehicles effectively solves the simulation distortion problem caused by ignoring the time-varying characteristics of soil and rock media. 4. This invention sets up an online self-calibration unit for virtual-real deviation and establishes a data-driven parameter adaptive correction mechanism; by collecting real vehicle operation data and virtual response for residual analysis, it uses an iterative optimization algorithm to adjust the voxel physical properties in the geological map in reverse until the error converges; this closed-loop calibration ensures that the simulation environment can be automatically updated with the accumulation of actual operation data, and guarantees the consistency and credibility of the digital twin model and physical reality in mechanical response; 5. This invention utilizes a follow-up voxel field interaction unit to achieve efficient mechanical calculation. This unit abandons the full-map real-time calculation mode, extracts only local voxel data around the virtual vehicle to construct the follow-up field, and maps the physical labels to soil mechanics model calculation parameters in real time. This design significantly reduces the computational load while ensuring the accuracy of ground mechanics calculation, and can output accurate road reaction force and medium deformation signals, meeting the real-time requirements of simulation systems in complex mining scenarios. 6. This invention integrates an environmental meteorological evolution module and a predictive ecological scheduling planning unit, achieving real-time simulation of all elements. The system can not only simulate the impact of rainfall infiltration and evaporation on soil properties, but also predict the evolution of road surface health within a future window; it triggers early warning replanning commands before the road surface degrades to a critical state, implementing load balancing or detour strategies, thus achieving proactive vehicle scheduling guided by road surface protection and overall efficiency, improving the safety and economy of mine operations. Attached Figure Description

[0014] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0016] Example 1: Please see Figure 1 A digital twin-based autonomous open-pit mining simulation system includes: The autonomous mining decision-making unit is configured to use reinforcement learning and operations research algorithms in a virtual environment to simulate and compare mining schemes, optimize mining plans, equipment scheduling and blasting parameters, and generate decision schemes to predict production bottlenecks. The virtual-real interaction and control unit is configured to enable real-time interaction between the digital twin model and the real equipment through Internet of Things (IoT) technology, predict and diagnose equipment failures, and provide a testing and training environment for unmanned mining trucks and remotely controlled drilling rigs. The multidimensional high-fidelity twin building blocks are configured to integrate geological exploration data, UAV mapping data and equipment parameters to create a digital twin model of an open-pit mine that includes geological structure, topography, mining equipment and production environment. The digital twin model unit is configured to establish a dynamic model containing ore bodies, equipment and processes based on physical modeling and machine learning algorithms to reflect the physical state and operating rules of open-pit mines, and to support multi-scale collaborative operation simulation from individual equipment to the entire mining area. The predictive ecological scheduling planning unit is configured to call the above unit to perform ultra-real-time simulation based on the mine truck traffic, monitor the road surface health index, and generate an early warning replanning instruction when the road surface health index is lower than the passable threshold. The virtual-real deviation online self-calibration unit is configured to collect real vehicle operation data and perform residual analysis on the response of virtual vehicles at the same location. When the residual exceeds the fitting tolerance, the voxel physical properties in the global dynamic geological map construction unit are adjusted in reverse using an iterative optimization algorithm. The global dynamic geological map construction unit is configured to segment the mining area space into voxels based on geological exploration data and UAV scanning data, assign physical attribute labels to voxels, and configure read and write interfaces to receive updated data; at the same time, the environmental meteorological evolution module is integrated and configured to access real-time weather station data, calculate rainfall infiltration and evaporation effects based on soil hydrological models, and maintain a global geological database that evolves dynamically over time. The servo voxel field interaction unit is configured to extract voxel data within a preset range around the geometric center of the virtual vehicle as the anchor point; construct a servo voxel field; convert the physical attribute labels into calculation parameters through a soil mechanics parameter mapping model; perform ground mechanics calculations in combination with vehicle state data; and output road reaction force signals and medium deformation signals. The vehicle-road physical property bidirectional coupling and memory unit, which connects the follow-up voxel field interaction unit and the global dynamic geological map construction unit, is configured to receive the medium deformation signal, perform positive environmental reshaping to convert the deformation into voxel property change amount, and force the voxel property change amount to be written back to the global dynamic geological map construction unit through the read-write interface.

[0017] In the system architecture of this embodiment, the global dynamic geological map construction unit constitutes the data foundation of the entire simulation environment. Addressing the technical pain point of traditional simulations where terrain is merely a static rigid surface, this unit executes high-precision voxel segmentation logic during the initialization phase, discretizing the continuous three-dimensional space composed of mining area geological exploration data and UAV scanning data into tiny volumetric units with independent addressing capabilities. Based on this data structure, the unit instantiates a set of physical attribute tags for each voxel. These tags not only include basic material identifiers but also cover a set of key parameters determining soil mechanical behavior, thereby giving the virtual environment physical fidelity. The read / write interface configured in this unit adopts a data interaction protocol with atomic operation characteristics, enabling it to receive dynamic update requests from downstream units and overwrite the attribute values ​​of specific coordinate voxels in real time. This design allows the unit to maintain not just a static map, but a global geological database that continuously evolves with the operation process and possesses state memory capabilities. When the virtual vehicle is in operation, the servo voxel field interaction unit acts as a real-time computing hub connecting the environment and vehicle dynamics, abandoning the high-energy-consuming mode of full-map real-time computing. This unit uses the geometric center of the virtual vehicle as a dynamic anchor point and uses a spatial indexing algorithm to retrieve and extract voxel data within a preset range around the vehicle in real time, constructing a local servo voxel field. Within this local field, the soil mechanics parameter mapping model integrated within the unit immediately intervenes, translating the physical attribute labels carried by the voxels into constitutive parameters required for dynamic calculations. Combined with the vehicle's current instantaneous state, the unit performs ground mechanics calculations and outputs two key signals: one is the road reaction force signal used to drive the vehicle dynamics model and determine the vehicle's attitude change; the other is the medium deformation signal that quantifies the physical damage caused by the vehicle to the road surface. As the logical core of this invention for realizing vehicle-road cooperative evolution, the vehicle-road physical attribute bidirectional coupling and memory unit establishes a closed-loop link from vehicle behavior feedback to environmental attributes. Upon receiving the medium deformation signal, the unit immediately initiates the positive environment reshaping logic, calculates the geometric reduction of voxel height and the change in internal physical attributes based on the deformation data, and forcibly writes these changes back to the corresponding storage address of the global dynamic geological map construction unit through the aforementioned read-write interface. This write-back operation establishes the environmental memory mechanism, enabling the road surface state to record the historical impact of each vehicle passage, thereby presenting cumulative physical changes in subsequent simulations, which is completely different from the memoryless instantaneous collision model. To ensure the reliability of the simulation system's parameters, the online self-calibration unit for virtual-real deviation establishes a data-driven adaptive parameter correction mechanism. This unit continuously collects CAN bus data from the real mining truck running in the physical world and synchronously compares it with the response data of the virtual vehicle at the same location in the digital twin scenario. When the residual between the two exceeds the fitting tolerance determined based on statistics, the system determines that the current environmental parameters are distorted. This unit triggers an iterative optimization algorithm to calculate the sensitivity gradient of the residual relative to soil physical properties and dynamically adjusts the voxel attributes in the global map along the error descent direction. That is, it only locks the spatial coordinate range covered by the GPS trajectory of the real vehicle at the current moment, performs local attribute updates on specific voxels within this range, and keeps the voxel attributes of other non-interactive areas unchanged. If multiple real trajectories cover the same voxel within the same calibration time window, the coverage update is performed based on the vehicle response data with the latest timestamp. This continues until the simulation response returns to the confidence interval. The predictive ecological scheduling planning unit, located at the top level of decision-making, uses the aforementioned high-fidelity physical evolution model for advanced extrapolation. This unit runs the simulation kernel at a rate higher than real-time to predict the evolution trend of road surface conditions within future time windows. During the extrapolation process, this unit monitors the road surface health index, which comprehensively reflects road quality, in real time. Once the index indicates that a certain road section will degrade to a critical state of impassability, it automatically triggers an early warning replanning instruction and outputs it to the mine scheduling center, realizing proactive vehicle scheduling guided by road surface protection and overall efficiency. The multidimensional high-fidelity twin building unit serves as the data foundation, integrating geological, surveying, and equipment parameters; the digital twin model unit, based on a physics engine and artificial intelligence algorithms, achieves real-time mapping from the physical world to the virtual world; the autonomous mining decision-making unit, as the system's decision-making center, conducts high-frequency trial and error and optimization in a zero-risk virtual environment; the virtual-real interaction and control unit breaks down the boundaries between the virtual and real worlds, supporting remote takeover and predictive maintenance; these four units, combined with the aforementioned global dynamic geological map building unit and vehicle-road physical property bidirectional coupling unit, jointly realize intelligent management and control of the entire process from microscopic soil mechanics to macroscopic mine scheduling; This embodiment realizes a symbiotic evolution mechanism in open-pit mining where vehicles change the environment and the environment reacts back to the vehicles through the organic collaboration of the above-mentioned units, thus solving the simulation distortion problem caused by neglecting the time-varying characteristics of the soil and rock media in existing mining technologies.

[0018] Example 2: The physical attribute labels include water saturation, initial shear modulus, and looseness coefficient; the water saturation defines the proportion of water inside the soil to characterize soil viscosity; the looseness coefficient defines the compactness of soil particles per unit volume to calculate compaction settlement; the read / write interface is configured to allow real-time modification of the attribute values ​​of specific voxels.

[0019] In the detailed configuration of this embodiment, the physical attribute labels managed by the global dynamic geological map construction unit are precisely defined as core parameter vectors describing the mechanical properties of unstructured pavements; water saturation is a value ranging from... The dimensionless parameters between these parameters are used to accurately quantify the proportion of water filling in soil pores. This parameter has a dual evolution mechanism in the model: it increases with rainfall accumulation or decreases with solar evaporation through the environmental meteorological evolution module. Specifically, the system uses the Green-Ampt infiltration model to calculate the rainfall infiltration amount accumulated over time to increase water saturation, and uses the Penman-Monteith formula combined with air temperature, humidity, and wind speed data to estimate the soil moisture evaporation rate to decrease water saturation. This parameter is directly related to the soil viscosity coefficient and shear strength in the model and is a key factor characterizing the difference in dynamics between muddy and dry hardened pavements after rain. The initial shear modulus defines the stiffness characteristics of soil resisting shear deformation in an undisturbed state. The unit is usually megapascals. Its value is initialized based on borehole sampling data from geological exploration and dynamically decreases or increases with disturbance in subsequent simulations. The looseness coefficient is a scalar describing the density of soil particles per unit volume. Its benchmark value comes from in-situ test data. This parameter directly determines the magnitude of plastic deformation of the pavement under the same axle load. The design of the read / write interface not only enables bidirectional data flow but also has microsecond-level data throughput capabilities, allowing real-time modification of the aforementioned attribute values ​​of specific voxels within the simulation step. This dynamic modification mechanism ensures that when a heavy-load mining truck passes by, causing soil compaction or agitation, the digital twin model can update the underlying physical description in real time, thereby presenting a mechanical response consistent with physical reality in the next calculation cycle.

[0020] Example 3: The soil mechanics parameter mapping model is configured to map the physical property labels into the computational parameters required by the Bekker-Wong model; the computational parameters include cohesion, internal friction angle, and soil deformation modulus; the servo voxel field interaction unit is configured to perform real-time approximate solution of the Bekker-Wong model based on the vehicle's instantaneous axle load, tire tread parameters, and the computational parameters; the pavement reaction force signal includes rolling resistance and shear resistance; the medium deformation signal includes settlement depth and shear displacement. In this embodiment, the soil mechanics parameter mapping model integrated within the follower voxel field interaction unit acts as a translator between geological data and the dynamic model. The translation logic library required by this model is constructed during the system initialization phase: First, in-situ shear tests and borehole sampling are conducted in different typical areas of the mining area to obtain measured datasets of soil physical properties, such as water saturation and density, and mechanical constitutive parameters, such as cohesion and internal friction angle. Then, machine learning algorithms or multivariate nonlinear regression analysis are used to establish a mapping relationship library from physical property labels to Bekker-Wong model parameters. During simulation, the model uses a pre-calibrated nonlinear regression function or multidimensional interpolation lookup table to accurately map the physical property labels carried by the voxels to the constitutive parameter set required by the Bekker-Wong ground mechanics model: cohesion, internal friction angle, and soil deformation modulus. Based on the transformed parameters and combined with the vehicle's current instantaneous axle load distribution and tire tread geometry, this unit performs a real-time approximate solution of the Bekker-Wong model. To meet real-time requirements, the solution process employs the discretized contact element method, dividing the tire-soil contact area into several micro-elements for force integration. The final output road reaction force signal is decomposed into a rolling resistance vector that hinders vehicle movement and a shear resistance vector that provides traction or causes sideslip. Simultaneously, the output medium deformation signal contains precisely quantified values ​​of subsidence depth and shear displacement. This process achieves accurate causal deduction from microscopic soil properties to macroscopic vehicle dynamics response.

[0021] Example 4: The specific logic of the vehicle-road physical property bidirectional coupling and memory unit for performing positive environmental reshaping includes: converting the subsidence depth in the medium deformation signal into the reduction amount of voxel height; converting the compaction effect into the reduction amount of the voxel loosening coefficient; after the voxel property change amount is written back, triggering the follow-up voxel field interaction unit to read the updated physical properties to generate a dynamic response that evolves with the road surface state.

[0022] To achieve high-fidelity environmental evolution, the vehicle-road physical property bidirectional coupling and memory unit executes a rigorous positive environmental reshaping algorithm. Upon receiving a medium deformation signal, the unit immediately performs dual update calculations at both the geometric and physical levels: at the geometric level, the subsidence depth value is directly converted into the height reduction of the corresponding voxel grid vertices, thereby depicting the realistic rut morphology in both the visual and geometric models; at the physical level, based on soil compaction theory, the compaction effect applied by the vehicle is converted into a reduction in the voxel loosening coefficient. This reduction is non-linearly positively correlated with the subsidence depth, and its calculation logic follows a preset empirical formula. ,in The amount by which the looseness factor decreases. This is the soil quality adjustment coefficient. This is the cumulative subsidence depth relative to the initial undisturbed terrain surface. A hardening index greater than 1, typically between 1.5 and 2.0, signifies increased soil density and enhanced bearing capacity under pressure. Once these voxel property changes are forcibly written back to the global geological database, the system immediately triggers a data synchronization mechanism. When subsequent vehicles enter the same area, or when the rear wheels of the same vehicle follow the trajectory left by the front wheels, the servo voxel field interaction unit reads the updated physical properties. This allows the force calculation of subsequent wheels to be based not on the initial road surface, but on the reshaped road surface, resulting in complex dynamic responses that evolve with the road surface condition, such as reduced tracking resistance or increased difficulty in escaping, accurately reproducing the real physical interaction process.

[0023] Example 5: The real vehicle operation data includes wheel speed difference and suspension compression; the iterative optimization algorithm is configured to calculate the sensitivity of the vehicle response residual to soil physical properties, generate environmental parameter correction factors with dynamic step size according to the direction of residual decrease, and perform re-simulation verification within the current time step; the fitting tolerance is the allowable error range statistically derived from historical high-precision test data.

[0024] In this embodiment, the virtual-real deviation online self-calibration unit eliminates systematic errors in the simulation model by processing multi-dimensional sensor data in real time; the collected real vehicle operation data focuses on core indicators reflecting vehicle-road interaction characteristics: wheel speed difference and suspension compression; the iterative optimization algorithm running inside this unit adopts a gradient-based parameter optimization strategy; the algorithm calculates the response residual between the simulation output and the real data. Since geological parameter mapping involves discrete operations, the algorithm uses the finite difference method to perform small perturbations on the soil physical properties of the current working point, thereby approximately solving the numerical sensitivity matrix of the residual relative to the current soil physical properties; Based on the numerical sensitivity analysis results, the algorithm generates environmental parameter correction factors along the direction of the residual numerical gradient using an adaptively adjusted dynamic step size, and fine-tunes the parameters in the geological map. After correction, the system immediately triggers a re-simulation verification within the current time step until the residual converges to the preset fitting tolerance range. This fitting tolerance is not arbitrarily set, but is determined based on the statistical analysis of massive historical high-precision test data, and is the allowable error range on the sensor noise basis. This mechanism ensures that the simulation environment can be automatically calibrated as actual operation data accumulates, continuously improving the realism of the physical model.

[0025] Example 6: The road surface health index is a comprehensive evaluation parameter derived from the weighted average rut depth and rolling resistance coefficient; the passability threshold is a value set based on the vehicle chassis minimum ground clearance and maximum climbing power; the ultra-real-time simulation is configured to accelerate the simulation process in an independent sandbox copy of the main database to predict the road surface condition within a specific future time window.

[0026] In this embodiment, the predictive ecological scheduling planning unit introduces a quantitative evaluation system to guide production decisions; the road surface health index is defined as a normalized comprehensive score, the calculation logic of which is based on the weighted sum of the average rut depth and the rolling resistance coefficient; the weighting factor is usually configured according to the focus of mine operation; the passable threshold is a hard constraint boundary set according to the physical limits of the vehicle: its lower limit is determined by the minimum ground clearance of the vehicle chassis, and its upper limit is related to the maximum climbing power of the vehicle under full load. The ultra-real-time simulation function utilizes high-performance computing resources to set the simulation clock frequency to a multiple of physical time, thereby enabling the prediction of road surface condition evolution within a specific future time window in an extremely short period. It is important to note that before initiating the ultra-real-time simulation, the system creates an independent sandbox copy of the current global dynamic geological map. Changes in voxel attributes during the simulation process only affect this sandbox copy, and the copy is released after the simulation concludes. This ensures that future predictive calculations do not overwrite or contaminate the currently synchronized master global geological database. This forward-looking simulation capability allows the scheduling system to identify potential road bottlenecks in advance, enabling an upgrade from post-fault repair to a condition-based early warning maintenance model.

[0027] Example 7: The early warning replanning instruction includes a load balancing strategy; the load balancing strategy includes forcing subsequent heavy-load vehicles to detour to protect the road surface, or reducing the load limit of passing vehicles; the early warning replanning instruction is configured to be output to the mine dispatch center.

[0028] When predictions indicate that the pavement health of a certain road segment is about to fall below the safety threshold, the predictive ecological scheduling planning unit generates an early warning replanning instruction that activates a high-level load balancing strategy. This strategy aims to extend the service life of roads by proactively intervening in traffic flow. When early signs of road fatigue appear, it mandates that subsequent heavy-load vehicles detour to other healthy road segments, allowing only empty or lightly loaded vehicles to pass, thereby dispersing concentrated stress damage to vulnerable roadbeds. Alternatively, if detours are not possible, it dynamically issues instructions to reduce the load limit of vehicles passing through the area to reduce the cumulative shear damage from a single passage. After encapsulation, this instruction is directly output to the human-machine interface or automatic scheduling server of the mine dispatch center via industrial communication protocols, ensuring that physical-level predictions can be transformed into production-level control measures in real time, maximizing the overall operational efficiency of the mine.

[0029] Example 8: The digital twin model unit is also configured to simulate hazardous working conditions, including equipment collisions, slope instability, and personnel intrusion; the virtual-real interaction and control unit is also configured to provide early warning and visual prompts for hazardous working conditions and support remote takeover. This embodiment focuses on the system's safety assurance mechanism; the digital twin model unit not only simulates normal operations, but also simulates various dangerous working conditions based on the physics engine, such as the risk of slope instability when heavy rain reduces the soil's shear strength, or the risk of equipment collision when multiple vehicles are working together; the virtual-real interaction and control unit transforms these invisible physical risks into visual prompts and sends them to the mine dispatch center in real time; when the autonomous mining decision unit cannot handle extreme working conditions, the system seamlessly switches to the manual remote takeover mode through the virtual-real interaction and control unit to ensure production safety.

[0030] 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.

Claims

1. A digital twin-based autonomous open-pit mining simulation system, characterized in that, include: The autonomous mining decision-making unit is configured to use reinforcement learning and operations research algorithms in a virtual environment to simulate and compare mining schemes, optimize mining plans, equipment scheduling and blasting parameters, and generate decision schemes to predict production bottlenecks. The virtual-real interaction and control unit is configured to enable real-time interaction between the digital twin model and the real equipment through Internet of Things (IoT) technology, predict and diagnose equipment failures, and provide a testing and training environment for unmanned mining trucks and remotely controlled drilling rigs. The multidimensional high-fidelity twin building blocks are configured to integrate geological exploration data, UAV mapping data and equipment parameters to create a digital twin model of an open-pit mine that includes geological structure, topography, mining equipment and production environment. The digital twin model unit is configured to establish a dynamic model containing ore bodies, equipment and processes based on physical modeling and machine learning algorithms to reflect the physical state and operating rules of open-pit mines, and to support multi-scale collaborative operation simulation from individual equipment to the entire mining area. The predictive ecological scheduling planning unit is configured to call the above units for ultra-real-time simulation based on the mine truck traffic. Monitor road surface health indicators; when the road surface health indicators are lower than the passable threshold, generate an early warning replanning instruction; The virtual-real deviation online self-calibration unit is configured to collect real vehicle operation data; perform residual analysis on the response of a virtual vehicle at the same location; and when the residual exceeds the fitting tolerance, use an iterative optimization algorithm to reverse adjust the voxel physical properties in the global dynamic geological map construction unit. The global dynamic geological map construction unit is configured to segment the mining area space into voxels based on geological exploration data and UAV scanning data; assign physical attribute labels to voxels; and configure read / write interfaces to receive updated data. Maintain a global geological database that evolves dynamically over time; The servo voxel field interaction unit is configured to extract voxel data within a preset range around the geometric center of the virtual vehicle as the anchor point; construct a servo voxel field; convert the physical attribute labels into calculation parameters through a soil mechanics parameter mapping model; perform ground mechanics calculations in conjunction with vehicle state data; and output road reaction force signals and medium deformation signals. The vehicle-road physical property bidirectional coupling and memory unit, which connects the follow-up voxel field interaction unit and the global dynamic geological map construction unit, is configured to receive the medium deformation signal; perform positive environmental reshaping to convert the deformation into voxel property change amount; and force write the voxel property change amount back to the global dynamic geological map construction unit through the read-write interface.

2. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The physical property labels include water saturation, initial shear modulus, and looseness coefficient; the water saturation defines the proportion of water inside the soil to characterize soil viscosity. The looseness coefficient defines the compactness of soil particles per unit volume for calculating compaction settlement; the read / write interface is configured to allow real-time modification of the attribute values ​​of specific voxels.

3. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The soil mechanics parameter mapping model is configured to map the physical property labels into the computational parameters required by the Bekker-Wong model; the computational parameters include cohesion, internal friction angle, and soil deformation modulus; the servo voxel field interaction unit is configured to perform real-time approximate solution of the Bekker-Wong model based on the vehicle's instantaneous axle load, tire tread parameters, and the computational parameters; the pavement reaction force signal includes rolling resistance and shear resistance; the medium deformation signal includes settlement depth and shear displacement.

4. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The specific logic of the vehicle-road physical property bidirectional coupling and memory unit for performing positive environmental reshaping includes: converting the subsidence depth in the medium deformation signal into the reduction amount of voxel height; converting the compaction effect into the reduction amount of the voxel loosening coefficient; after the voxel property change amount is written back, triggering the follow-up voxel field interaction unit to read the updated physical properties to generate a dynamic response that evolves with the road surface state.

5. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The real vehicle operation data includes wheel speed difference and suspension compression; the iterative optimization algorithm is configured to calculate the sensitivity of the vehicle response residual with respect to soil physical properties, generate environmental parameter correction factors with dynamic step size according to the direction of residual decrease, and perform re-simulation verification within the current time step; The fitting tolerance is the allowable error range derived from statistical analysis of historical high-precision test data.

6. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The road surface health index is a comprehensive evaluation parameter derived by weighting the average rut depth and rolling resistance coefficient; the passability threshold is a value set based on the vehicle chassis minimum ground clearance and maximum climbing power; the ultra-real-time simulation is configured to accelerate the simulation process in an independent sandbox copy of the main database to predict the road surface condition within a specific future time window.

7. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The early warning replanning instruction includes a load balancing strategy; the load balancing strategy includes forcing subsequent heavy-load vehicles to detour to protect the road surface, or reducing the load limit of passing vehicles. The early warning replanning instruction is configured to be output to the mine dispatch center.

8. The autonomous open-pit mining simulation system based on digital twins according to claim 1, characterized in that, The digital twin model unit is also configured to simulate hazardous working conditions, including equipment collisions, slope instability, and personnel intrusion; the virtual-real interaction and control unit is also configured to provide early warning and visual prompts for the hazardous working conditions and support remote takeover.

Images