A digital twin simulation optimization system and method for mining dilution and loss rate

Abstract

The application discloses a kind of digital twin simulation optimization system and method of mining dilution rate and loss rate, it is related to digital mine technical field, including: by the fusion of geological model and production model in unified platform, the problem of isolated operation of traditional system in geology, production and automation subsystem is solved, and integrated intelligent system of data intercommunication, business cooperation is formed.Dilution rate, loss rate management is changed from traditional post statistics to instantaneous perception and abnormal diagnosis based on real-time data, realizes the online monitoring and active intervention to production process, reduces resource waste.Through simulation engine, a variety of schemes are deduced in parallel and multi-objective optimization, and the scientific decision-making scheme of global optimum under multi-objective can be provided.Through the real-time feedback of execution result, the twin mine model is continuously calibrated and optimized, and the drawbacks of traditional static model accuracy decay over time are overcome, which provides a reliable foundation for long-term optimization.

Description

Technical Field

[0001] This invention relates to the field of digital mining technology, and in particular to a digital twin simulation optimization system and method for mining dilution rate and loss rate. Background Technology

[0002] Traditional mining faces challenges such as deteriorating resource endowments, complex mining conditions, and increasing pressure on safety and environmental protection, making digital transformation an inevitable trend. With the development of technologies such as the Internet of Things, big data, and artificial intelligence, building a virtual-real mapping and real-time interactive digital mining system has become a research hotspot.

[0003] Digital twin technology, as a core path to achieving intelligent mining, aims to comprehensively improve resource recovery rate, production efficiency and safety level through high-precision modeling, real-time perception and intelligent decision-making, leading the mining industry towards a green, safe and efficient direction.

[0004] Current technologies primarily focus on building isolated systems. Geological modeling mainly utilizes software like Micromine for static resource assessment; production management relies on DCS / MES systems for data monitoring and scheduling; and equipment automation achieves single-machine or partial automation. These systems suffer from poor data interoperability, forming "data silos." Analysis and decision-making largely depend on human experience, failing to achieve deep integration and intelligent linkage between geological and production data. Model updates lag behind, hindering real-time optimization and precise control of the production process. Summary of the Invention

[0005] To address some or all of the technical problems existing in the prior art, this invention provides a digital twin simulation optimization system and method for mining dilution rate and loss rate.

[0006] The technical solution of the present invention is as follows:

[0007] A digital twin simulation optimization system and method for mining dilution rate and loss rate are provided, including:

[0008] Model building module: Collect geological and production data of the mining area, build a geological model based on the geological data, build a production model based on the production data, and merge the geological model and the production model to build a twin mining area model;

[0009] Anomaly Diagnosis Module: Based on the twin mining area model and real-time mining area data, the module uses machine learning algorithms to identify the ore-rock mixing state, calculates the instantaneous dilution rate and instantaneous loss rate, performs anomaly diagnosis, and obtains the anomaly diagnosis results.

[0010] Optimal Decision Module: Based on the anomaly diagnosis results, the simulation engine is started to perform parallel simulations of multiple mining schemes, and the simulation results are Pareto optimally screened to obtain the optimal decision scheme;

[0011] Continuous optimization module: It parses the optimal decision scheme into control commands and sends them to the production equipment for execution, and feeds back the execution results to the twin mining area model to continuously optimize the twin mining area model.

[0012] In some optional implementations, the model building module includes:

[0013] Basic processing unit: Performs basic processing on the collected geological and production data;

[0014] Fitting Unit: Establish a structured geological database based on the geological data after basic processing; calculate the variogram for different geographical directions based on the grade data in the geological data, fit the variogram values ​​with the preset theoretical variogram model, and derive the theoretical variogram data model characterizing the spatial correlation of grade based on the fitting results.

[0015] Geological model construction unit: Based on the structured geological database, the orebody wireframe model is determined. The theoretical variation data model is input into the orebody wireframe model to spatially estimate the average grade of each block in the mining area, thereby obtaining the grade distribution attributes of the mining area. Combining the theoretical variation data model with the lithology-mechanics mapping table, the rock mass mechanical properties of each block in the mining area are determined. By integrating the grade distribution attributes and rock mass mechanical properties of each block, the geological model of the mining area is constructed.

[0016] Logical model building unit: Determines the logical model of the production process based on production data;

[0017] Equipment Behavior Model Construction Unit: Extract equipment-related information from the basic processed production data, construct the basic equipment model according to the equipment design documents, bind the real-time pose data and motion state data of each device in the equipment-related information to the basic equipment model, establish the mapping relationship between the spatial pose and motion state of the device and the model animation frames based on the binding results, and obtain the equipment behavior model;

[0018] Production model construction unit: Integrates the logical model and equipment behavior model in a unified 3D platform to obtain the production model;

[0019] Coupling rule determination unit: Import the geological model and the production model into a unified digital twin simulation platform, establish the first coupling rule between the geological model and the production model through the coupling mechanism algorithm, and establish the second coupling rule between the geological model and the production model through the machine learning algorithm;

[0020] Twin mining area model construction unit: Based on the first coupling rule and the second coupling rule, the geological model and the production model are fused to obtain the twin mining area model.

[0021] In some optional implementations, the coupling rule determination unit includes:

[0022] The first coupling block: Based on the digital twin simulation platform, the key physical processes in the geological model that directly affect the production behavior in the production model are identified, and a geological-production association list is generated based on the key physical processes. A corresponding mechanism algorithm table is developed based on the geological-production association list, and the bidirectional triggering conditions are determined based on the mechanism algorithm table to derive the first coupling rule.

[0023] The second coupling block extracts correlation features from the geological model and production model, while simultaneously acquiring the business needs and production pain points of the mining area, extracting the limitations of the first coupling rule, determining the coupling problem based on the business needs, production pain points, and limitations, selecting a machine learning algorithm based on the coupling problem, and determining the second coupling rule by combining the correlation features and the machine learning algorithm.

[0024] In some optional implementations, the anomaly diagnosis module includes:

[0025] Ore-rock mixing state identification unit: calls a pre-trained convolutional neural network model to perform frame-by-frame analysis of the real-time video stream of loading operations in the mining area, identifies the ore-rock mixing state of ore and waste rock in the bucket, and outputs the estimated mixing ratio of the identified ore-rock mixing state.

[0026] Calculation Unit: Based on the estimated mixed ratio, combined with the real-time equipment load data in the real-time data of the mining area and the block theoretical grade data in the twin mining area model, the instantaneous dilution rate and instantaneous loss rate of the current working unit are dynamically calculated according to the mining engineering calculation formula.

[0027] Alarm Information Determination Unit: Based on the preset impoverishment rate threshold and the preset loss rate threshold, the instantaneous impoverishment rate is compared with the preset impoverishment rate threshold for the first time, and the instantaneous loss rate is compared with the preset loss rate threshold for the second time. When either the instantaneous impoverishment rate or the instantaneous loss rate exceeds its corresponding preset threshold, a first-level warning is triggered and a primary alarm information is generated. When both the instantaneous impoverishment rate and the instantaneous loss rate exceed their corresponding preset thresholds simultaneously, or when either the instantaneous impoverishment rate or the instantaneous loss rate continues to exceed the limit for a preset duration, a second-level alarm is triggered and a high-level alarm information is generated.

[0028] Anomaly Diagnosis Unit: Based on the twin mining area model, it initiates anomaly tracing of primary and advanced alarm information, obtains the tracing results, and performs anomaly diagnosis on the mining area by integrating the primary alarm information, advanced alarm information, and tracing results, and obtains the anomaly diagnosis results.

[0029] In some optional implementations, the optimal decision module includes:

[0030] Simulation Unit: The simulation engine simulates the execution process of each mining scheme according to the first coupling rule and the second coupling rule, obtains simulation results, and determines the simulation output indicators based on the simulation results;

[0031] Constraint determination unit: Determine the optimization objective based on minimizing the dilution rate, minimizing the loss rate, and maximizing the ore production, and determine the constraint conditions based on equipment capacity constraints, resource reserve constraints, and safety boundary constraints;

[0032] Decision-making unit: Based on constraints and optimization objectives, determine a multi-objective optimization algorithm, comprehensively evaluate the simulation output indicators of each mining scheme, solve for the Pareto optimal solution set, and select the optimal decision scheme from the Pareto optimal solution set based on real-time production strategy preferences.

[0033] In some optional implementations, the simulation unit includes:

[0034] Rule simulation block: Calculates the geological-production response of each mining scheme during execution based on the first coupling rule, predicts the block size distribution of each mining scheme, and simulates the intelligent behavior of equipment during execution based on the second coupling rule to generate equipment actions;

[0035] Indicator Determination Block: Data fusion and production process simulation are performed on the block degree distribution and equipment actions to obtain simulation results. The simulation results are then quantitatively analyzed using a predefined indicator calculation model to determine the simulation output indicators.

[0036] In some optional implementations, the continuous optimization module includes:

[0037] Instruction generation unit: Inputs the optimal decision scheme into the instruction parsing engine. The instruction parsing engine integrates a mechanism rule base and a data-driven strategy base. Based on the mechanism rule base, it converts the scheme parameters of the optimal decision scheme into basic control instructions. Based on the data-driven strategy base, it performs local optimization and adjustment on the basic control instructions to generate the final control instruction sequence.

[0038] Model optimization unit: Collects the execution results of the final control commands, optimizes the mechanism rule base and data-driven strategy base based on the execution results, and then optimizes the twin mining area model.

[0039] A digital twin simulation optimization method for mining dilution rate and loss rate is also provided, including:

[0040] Step 1: Collect geological and production data of the mining area, construct a geological model based on the geological data, construct a production model based on the production data, and merge the geological model and the production model to construct a twin mining area model;

[0041] Step 2: Based on the twin mining area model and real-time mining area data, machine learning algorithms are used to identify the ore-rock mixing state, calculate the instantaneous dilution rate and instantaneous loss rate, and perform anomaly diagnosis to obtain the anomaly diagnosis results;

[0042] Step 3: Based on the anomaly diagnosis results, start the simulation engine to perform parallel simulations of multiple mining schemes, and perform Pareto optimal selection on the simulation results to obtain the optimal decision scheme;

[0043] Step 4: Parse the optimal decision scheme into control commands and send them to the production equipment for execution, and feed the execution results back to the twin mining area model for continuous optimization.

[0044] The main advantages of the technical solution of this invention are as follows:

[0045] By constructing a high-fidelity twin model of the mining area, the system can perceive the production status and diagnose anomalies in real time. Simulations are then initiated to optimize decisions under multi-objective constraints. Finally, the optimal solution is deployed and executed, and feedback data is used to continuously optimize the model, forming a closed-loop intelligent control system of "perception-diagnosis-decision-execution-optimization." By integrating the geological and production models on a unified platform, the system completely solves the problem of isolated operation of geological, production, and automation subsystems in traditional systems, forming an integrated intelligent system with data interoperability and business collaboration. The management of dilution and loss rates is transformed from traditional post-event statistics to "instantaneous perception" and anomaly diagnosis based on real-time data, enabling online monitoring and proactive intervention in the production process and significantly reducing resource waste. Through parallel simulation and multi-objective optimization of multiple solutions using a simulation engine, the system changes the decision-making model that relies on human experience, providing globally optimal scientific decision-making solutions under multiple objectives such as resource recovery, production efficiency, cost safety, etc. By providing real-time feedback on the execution results, the twin mining area model is continuously calibrated and optimized, enabling it to evolve dynamically and continuously improve its accuracy with use. This overcomes the drawback of the traditional static model's accuracy decaying over time, providing a reliable foundation for long-term optimization. Attached Figure Description

[0046] The accompanying drawings, which are included to provide a further understanding of embodiments of the invention and constitute a part of this invention, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0047] Figure 1 This is a schematic diagram of the structure of a digital twin simulation optimization system for mining dilution rate and loss rate provided in an embodiment of the present invention;

[0048] Figure 2 This is a flowchart illustrating a digital twin simulation optimization method for mining dilution rate and loss rate, provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0050] The technical solutions provided by the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] refer to Figure 1 This invention provides a digital twin simulation optimization system for mining dilution rate and loss rate, comprising:

[0052] Model building module: Collect geological and production data of the mining area, build a geological model based on the geological data, build a production model based on the production data, and merge the geological model and the production model to build a twin mining area model;

[0053] Anomaly Diagnosis Module: Based on the twin mining area model and real-time mining area data, the module uses machine learning algorithms to identify the ore-rock mixing state, calculates the instantaneous dilution rate and instantaneous loss rate, performs anomaly diagnosis, and obtains the anomaly diagnosis results.

[0054] Optimal Decision Module: Based on the anomaly diagnosis results, the simulation engine is started to perform parallel simulations of multiple mining schemes, and the simulation results are Pareto optimally screened to obtain the optimal decision scheme;

[0055] Continuous optimization module: It parses the optimal decision scheme into control commands and sends them to the production equipment for execution, and feeds back the execution results to the twin mining area model to continuously optimize the twin mining area model.

[0056] In this embodiment of the invention, the geological model is the final output, a high-precision three-dimensional digital ore body that integrates grade and mechanical properties. For example, it can not only display the grade of the ore but also query the rock hardness of the virtual ore body.

[0057] In this embodiment of the invention, the production model is the result of integrating the logical model and the equipment behavior model, and can simulate the operation of the entire production system. For example, in a three-dimensional scene, multiple virtual trucks can be seen transporting ore from the mining area to the unloading point in an orderly manner according to the scheduling logic.

[0058] In this embodiment of the invention, the geological model is input with grade data, variogram model, and lithology-mechanical mapping table derived from geological data, and outputs a three-dimensional block model with grade and mechanical properties. For example, inputting the coordinates and variogram of a certain region outputs the average grade and rock hardness of that region. The production model is input with equipment CAD files, real-time pose data, and production process rules derived from production data, and outputs a dynamically running three-dimensional virtual production system. For example, inputting real-time GPS data of a truck outputs the synchronous movement of a virtual truck in a three-dimensional scene.

[0059] In this embodiment of the invention, the inputs to the twin mining area model are a geological model, a production model, real-time data, and coupling rules, while the outputs are predictions, optimizations, and decision-making schemes. For example, given the current equipment location and geological information, the output might suggest adjusting the shovel loading position to reduce the dilution rate.

[0060] In this embodiment of the invention, the real-time data of the mining area includes real-time video streams of loading operations and real-time load data of the equipment.

[0061] In this embodiment of the invention, the ore-rock mixing state refers to the mixing of ore and waste rock in the material flow during production operations such as loading and transportation. This is a quantitative concept used to describe the mixing ratio of useful minerals and useless surrounding rock in a single production unit. Using a camera mounted above the bucket and an image recognition algorithm, the system analyzes the real-time video stream and outputs a quantitative judgment: currently, the material in the bucket contains approximately 65% ​​ore and approximately 35% waste rock. This 65:35 ratio is a precise description of the ore-rock mixing state.

[0062] In this embodiment of the invention, the anomaly diagnosis result is a comprehensive judgment conclusion formed by the system after identifying an abnormal ore-rock mixing state and further analyzing the root cause of the problem. For example, the system calculation found that the instantaneous dilution rate was as high as 40%, far exceeding the 20% threshold. The anomaly diagnosis module initiated source tracing analysis and generated the following anomaly diagnosis result: Conclusion: The dilution rate of the No. 4 electric shovel seriously exceeded the limit when operating in the E5 area of ​​the mining area. Root cause analysis shows that there is a 1.5-meter deviation between the ore body boundary in the geological model and the actual exposed boundary, causing the equipment to actually enter the surrounding rock area when operating within the 'ore body' recognized by the model. Recommendation: Immediate adjustment: Instruct the operator to move the working position back 1.5 meters inside the model. On-site verification: It is recommended that surveyors conduct on-site boundary surveys of the E5 area. Model update: Based on the survey results, correct the ore body outline in the geological model.

[0063] In this embodiment of the invention, Pareto optimal selection is a multi-objective optimization decision-making method. It selects an optimal set of solutions from all feasible options. In this set, improving any objective of any solution inevitably leads to a worsening of at least one other objective. There is no absolute good or bad among these solutions, only different trade-offs. For example, a simulation engine generates 10 mining plans, and after selection, 3 Pareto optimal solutions are obtained: Solution A: Lowest depletion rate of 8%, lowest loss rate of 6%, but also lowest production of 700,000 tons. Solution B: Highest production of 1,000,000 tons, lowest loss rate of 5%, but highest depletion rate of 18%. Solution C: Balanced indicators, depletion rate of 12%, loss rate of 5.5%, and production of 900,000 tons. These three solutions constitute the Pareto optimal solution set. It cannot be said that A is better than B, because reducing the depletion rate comes at the cost of sacrificing production.

[0064] In this embodiment of the invention, the optimal decision-making scheme is an implementation plan ultimately selected from the Pareto optimal solution set based on the current production strategy preferences. It is the product of combining management's value judgment with scientific optimization results. For example, Scenario 1 prefers high-quality mining: the current company strategy is to reduce dilution and improve ore quality; therefore, the optimal decision-making scheme ultimately selected is Scheme A, which has the lowest dilution rate. Scenario 2 prefers to complete production targets: the current company's primary task is to complete the monthly production plan; therefore, the optimal decision-making scheme ultimately selected is Scheme B, which has the highest output. Scenario 3 prefers balanced development: the company hopes to operate steadily and balance efficiency and quality; therefore, the optimal decision-making scheme ultimately selected is Scheme C, which balances all indicators.

[0065] In this embodiment of the invention, the execution result is the actual feedback data collected by sensors after the final control command sequence is executed on the physical device. It reflects the difference between the executed result and the expected result. For example, the command may require drilling to -14.5m, but the sensor feedback indicates an actual depth of -14.7m; the command may expect a block size of d0=25cm, but actual screening data shows d0=28cm. These actual data constitute the execution result.

[0066] In this embodiment of the invention, the specific process of optimizing the twin mining area model involves using execution result feedback data to calibrate and update the parameters of the sub-models within the twin mining area model. The aim is to reduce the discrepancy between simulation predictions and physical reality, enabling the twin mining area model to continuously evolve and maintain high fidelity. For example, the optimization of the mechanism rule base is achieved when execution results show that the actual block size is always 10% larger than the predicted model's output. Therefore, the first step in optimizing the twin mining area model is to calibrate the blasting model and adjust its parameters to make the predicted values ​​closer to the actual values. The optimization of the data-driven strategy base is achieved when execution results show that the recommended speed-down and energy-saving instructions from the strategy base cause task delays. Therefore, the reinforcement learning model is retrained with new data, and its strategy is updated to enable better decisions in the future. The optimization of the geological model is achieved when a systematic deviation exists between the actual borehole depth and the expected depth, indicating inaccurate rock layer interface depths in the geological model. Based on this, relevant layers in the geological model are corrected.

[0067] In this embodiment of the invention, the model building module includes:

[0068] Basic processing unit: Performs basic processing on the collected geological and production data;

[0069] Fitting Unit: Establish a structured geological database based on the geological data after basic processing; calculate the variogram for different geographical directions based on the grade data in the geological data, fit the variogram values ​​with the preset theoretical variogram model, and derive the theoretical variogram data model characterizing the spatial correlation of grade based on the fitting results.

[0070] Geological model construction unit: Based on the structured geological database, the orebody wireframe model is determined. The theoretical variation data model is input into the orebody wireframe model to spatially estimate the average grade of each block in the mining area, thereby obtaining the grade distribution attributes of the mining area. Combining the theoretical variation data model with the lithology-mechanics mapping table, the rock mass mechanical properties of each block in the mining area are determined. By integrating the grade distribution attributes and rock mass mechanical properties of each block, the geological model of the mining area is constructed.

[0071] Logical model building unit: Determines the logical model of the production process based on production data;

[0072] Equipment Behavior Model Construction Unit: Extract equipment-related information from the basic processed production data, construct the basic equipment model according to the equipment design documents, bind the real-time pose data and motion state data of each device in the equipment-related information to the basic equipment model, establish the mapping relationship between the spatial pose and motion state of the device and the model animation frames based on the binding results, and obtain the equipment behavior model;

[0073] Production model construction unit: Integrates the logical model and equipment behavior model in a unified 3D platform to obtain the production model;

[0074] Coupling rule determination unit: Import the geological model and the production model into a unified digital twin simulation platform, establish the first coupling rule between the geological model and the production model through the coupling mechanism algorithm, and establish the second coupling rule between the geological model and the production model through the machine learning algorithm;

[0075] Twin mining area model construction unit: Based on the first coupling rule and the second coupling rule, the geological model and the production model are fused to obtain the twin mining area model.

[0076] In this embodiment of the invention, the basic processing unit is a module that performs preprocessing operations such as cleaning, format standardization, and coordinate unification on the original geological and production data. For example, it removes abnormal grade values ​​from borehole data and unifies the GPS coordinates of different devices to the local coordinate system of the mine. The fitting unit is a module responsible for geostatistical analysis, which quantifies the spatial correlation of grade data by calculating and fitting a variogram. For example, the calculation finds that the correlation of grade in the northeast direction is effective within 100 meters, and changes randomly beyond that, and uses a spherical mathematical model to accurately describe this pattern.

[0077] In this embodiment of the invention, a structured geological database is a database that organizes, associates, and manages geological data according to specific rules. For example, a relational database containing borehole tables, sample tables, and test tables allows querying the grade information of all samples from a borehole by its borehole ID.

[0078] In this embodiment of the invention, the theoretical variation data model is the output of the fitting unit, a mathematical function that defines the spatial variation law of grade. For example, a spherical model formula including parameters such as range and sill value is used to guide subsequent Kriging estimation.

[0079] In this embodiment of the invention, the ore body wireframe model is a triangular mesh surface model representing the ore body boundary, delineated in three-dimensional space by a geological engineer. For example, it could be an irregular, shell-like three-dimensional shape seen in software that encloses all the high-grade ore bodies.

[0080] In this embodiment of the invention, spatial estimation refers to algorithms such as kriging, used to estimate the grade of each unknown block. For example, based on the grades of 20 surrounding boreholes, a weighted average is used to estimate the copper grade of a certain 10m×10m×10m block as 0.58%.

[0081] In this embodiment of the invention, the grade distribution attribute is the result of spatial estimation, that is, the grade value of each block. For example, the entire ore body is divided into millions of blocks, each block having a grade value, forming a three-dimensional grade distribution field.

[0082] In this embodiment of the invention, the lithology-mechanical mapping table is a database table that defines the correspondence between rock types and their mechanical parameters. For example, querying this table shows that the uniaxial compressive strength range for skarn is 150-200 MPa.

[0083] In this embodiment of the invention, the rock mass mechanical properties are mechanical parameter values ​​assigned to each block. For example, using the co-kriging method, a certain block is assigned mechanical properties of UCS=180MPa and internal friction angle=35°.

[0084] In this embodiment of the invention, the logical model is a digital abstraction of the production process, defining the logical order and rules of each step.

[0085] In this embodiment of the invention, the equipment behavior model is a dynamic model that can reproduce the motion, posture, and actions of physical equipment in real time. For example, a virtual loader model moves in real time in a three-dimensional environment, and the lifting and lowering of its bucket is completely synchronized with the real equipment.

[0086] In this embodiment of the invention, the coupling rules act as the brain connecting the geological and production models. The first coupling rule is a model based on physical laws. For example, a blasting dynamics model, given rock strength and blasting parameters, can predict the block size distribution of the rock after blasting. The second coupling rule is a data-driven model. For example, a deep learning model based on camera images can identify the ratio of ore to waste rock in the bucket in real time.

[0087] In this embodiment of the invention, the output of the basic processing unit serves as the input to the fitting unit, the logic model construction unit, and the equipment behavior model construction unit. For example, cleaned borehole data is sent to the fitting unit to calculate the variogram; processed equipment data is sent to the equipment behavior model for driving. The output of the fitting unit is a key input to the geological model construction unit. For example, the Kriging method requires the fitted variogram model for accurate grade estimation. The outputs of the geological model construction unit and the production model construction unit, namely the geological model and the production model, are the inputs to the coupling rule determination unit. For example, blasting simulation requires the geological model to provide rock strength and the production model to provide current borehole equipment parameters. The outputs of the coupling rule determination unit, namely the first coupling rule and the second coupling rule, are the core of the twin mining area model construction unit, used to integrate the static geological and production models into a living twin. For example, embedding the second coupling rule of the ore and rock identification model into the system gives the twin the function of "eyes," enabling it to see the dilution situation in real time.

[0088] The coupling rule determination unit includes:

[0089] The first coupling block: Based on the digital twin simulation platform, the key physical processes in the geological model that directly affect the production behavior in the production model are identified, and a geological-production association list is generated based on the key physical processes. A corresponding mechanism algorithm table is developed based on the geological-production association list, and the bidirectional triggering conditions are determined based on the mechanism algorithm table to derive the first coupling rule.

[0090] The second coupling block extracts correlation features from the geological model and production model, while simultaneously acquiring the business needs and production pain points of the mining area, extracting the limitations of the first coupling rule, determining the coupling problem based on the business needs, production pain points, and limitations, selecting a machine learning algorithm based on the coupling problem, and determining the second coupling rule by combining the correlation features and the machine learning algorithm.

[0091] In this embodiment of the invention, the key physical process is the causal chain of how geological properties directly affect production behavior and results through physical and chemical laws. For example, the geological property of rock hardness directly determines the efficiency of the loading equipment and the size of the blasted blocks by influencing the blasting and crushing effect in the physical process.

[0092] In this embodiment of the invention, the geological-production linkage list is a structured, tabular description of key physical processes. For example, a table may have one column for geological properties, such as rock compressive strength, another column for the affected production stages, and a third column for the influencing mechanism, such as the impact on explosive energy utilization.

[0093] In this embodiment of the invention, the mechanism algorithm table concretizes each item in the correlation list into a computable and executable physical or engineering mathematical model. For example, the correlation between rock hardness and blasting is described above, and the corresponding mechanism algorithm is the Livingston blasting crater model or the Kuz-Ram block size prediction model, whose mathematical formula is block size = f(explosive amount, rock strength, joint density), where f represents a function, and the specific function is given by the selected model.

[0094] In this embodiment of the invention, the bidirectional triggering condition defines the conditional logic that requires the mechanistic algorithm to be activated for calculation under what conditions. This includes: a forward trigger, from geology to production, which is triggered when production activities encounter a new geological environment. For example, when the working trajectory of an electric shovel enters a region of high-hardness rock, the blasting model is triggered to recalculate the required amount of explosives. A reverse trigger, from production to geology, is triggered when production activities change the geological environment. For example, after blasting is completed, the geological model is updated, and the rock mass boundaries and block size distribution are corrected based on the blasting effect.

[0095] In this embodiment of the invention, the first coupling rule is the final, automatically executable IF-THEN rule, which is a combination of mechanism algorithm and triggering condition. For example, IF [device enters high-hardness rock formation] THEN WITH [current rock formation UCS value] EXECUTE [Kuz-Ram model] OUTPUT [new borehole spacing recommendation].

[0096] In this embodiment of the invention, the correlation features are variables that may have hidden relationships, extracted from geological and production models. For example, the skarn content in the geological model and the crusher spindle current in the production model.

[0097] In this embodiment of the invention, business needs and production pain points are goals that are desired to be achieved or problems that urgently need to be solved in mine operations. For example, a business need is to increase the throughput of ball mills; a production pain point is that gyratory crushers frequently experience unplanned downtime.

[0098] In this embodiment of the invention, the limitation of the first coupling rule is that it addresses scenarios that the mechanistic model cannot handle well. Examples include: a knowledge gap where there is currently no precise mechanistic model describing how ore viscosity affects flotation recovery; and excessive complexity where the overall health of the equipment is a complex function of hundreds of parameters, which cannot be described by a single mechanistic equation.

[0099] In this embodiment of the invention, the coupling problem is a specific problem that is suitable for machine learning to solve, defined based on pain points and limitations. For example, by combining the limitation of the first coupling rule—the difficulty in modeling the overall health of the equipment—with the production pain point of unplanned downtime of the crusher, a coupling problem is defined: predicting the health status of the crusher for the next 24 hours based on multi-dimensional sensor data of the equipment.

[0100] In this embodiment of the invention, the second coupling rule is the trained machine learning model and its application logic. For example, for an LSTM neural network model, the rule is: IF[real-time sensor data]THENEXECUTE[LSTM prediction model]OUTPUT[device health score]IF[score < threshold]THEN[trigger warning].

[0101] In this embodiment of the invention, the second coupling block depends on the first coupling block. The limitations in the second coupling block directly stem from problem areas that the first coupling rule cannot adequately address. The two are complementary, together constituting the hybrid-driven coupling rule.

[0102] In this embodiment of the invention, the anomaly diagnosis module includes:

[0103] Ore-rock mixing state identification unit: calls a pre-trained convolutional neural network model to perform frame-by-frame analysis of the real-time video stream of loading operations in the mining area, identifies the ore-rock mixing state of ore and waste rock in the bucket, and outputs the estimated mixing ratio of the identified ore-rock mixing state.

[0104] Calculation Unit: Based on the estimated mixing ratio, combined with the real-time load data of the equipment in the real-time data and the theoretical grade data of the block in the twin mining area model, the instantaneous dilution rate and instantaneous loss rate of the current working unit are dynamically calculated according to the mining engineering calculation formula.

[0105] Alarm Information Determination Unit: Based on the preset impoverishment rate threshold and the preset loss rate threshold, the instantaneous impoverishment rate is compared with the preset impoverishment rate threshold for the first time, and the instantaneous loss rate is compared with the preset loss rate threshold for the second time. When either the instantaneous impoverishment rate or the instantaneous loss rate exceeds its corresponding preset threshold, a first-level warning is triggered and a primary alarm information is generated. When both the instantaneous impoverishment rate and the instantaneous loss rate exceed their corresponding preset thresholds simultaneously, or when either the instantaneous impoverishment rate or the instantaneous loss rate continues to exceed the limit for a preset duration, a second-level alarm is triggered and a high-level alarm information is generated.

[0106] Anomaly Diagnosis Unit: Based on the twin mining area model, it initiates anomaly tracing of primary and advanced alarm information, obtains the tracing results, and performs anomaly diagnosis on the mining area by integrating the primary alarm information, advanced alarm information, and tracing results, and obtains the anomaly diagnosis results.

[0107] In this embodiment of the invention, the core function of the ore-rock mixture state identification unit is to automatically identify whether the material in the bucket is ore or waste rock using computer vision technology. For example, when the electric shovel is performing loading operations, a camera installed in the cab captures real-time video of the material in the bucket. This unit calls a pre-trained AI model to analyze each frame of the image and determines that 70% of the material in the bucket is ore and 30% is waste rock. The calculation unit is responsible for converting the identification information into key management indicators. It receives multi-source data and performs calculations using mining engineering formulas. For example, if the unit obtains a waste rock ratio of 0.3, a bucket load of 15 tons, and that the current loading position comes from a block with a copper grade of 1.5% in the geological model, it can calculate the dilution rate and loss rate. The anomaly diagnosis unit is responsible for automatically analyzing the root cause of the problem after an alarm is triggered. It provides a diagnostic conclusion by integrating various data. For example, after receiving an alarm, the unit automatically analyzes and finds that: the equipment has not crossed the boundary, but the geological model shows that the boundary of the ore body in this area is blurred. The diagnosis result is: there is a deviation between the geological model and the actual situation, and it is recommended to conduct on-site inspection and correction of the model.

[0108] In this embodiment of the invention, the estimated mixing ratio is the output of the aforementioned identification unit, and is a quantified numerical value representing the percentage of ore and waste rock in the bucket. For example, a numerical pair (ore ratio: 0.7, waste rock ratio: 0.3).

[0109] In this embodiment of the invention, the instantaneous dilution rate is one of the outputs of the calculation unit, representing the proportion of waste rock mixed in a single production operation unit, such as a shovel or a truckload. For example, according to the formula: Instantaneous dilution rate = Waste rock weight / Total weight × 100%, the dilution rate of the current shovel is calculated to be 30%.

[0110] In this embodiment of the invention, the alarm information determination unit acts as the system's sentinel, responsible for monitoring indicators and determining whether they are abnormal. It compares real-time calculated values ​​with preset safety thresholds. For example, the unit compares the calculated instantaneous depletion rate with a threshold set by the management system; if a serious exceedance is detected, an alarm is immediately triggered.

[0111] In this embodiment of the invention, the primary alarm message is a warning for a single indicator that exceeds the limit for a short period of time, prompting the operator to pay attention. For example, a prompt sent to the electric shovel operator's screen: Warning: Current depletion rate is 30%, exceeding the 20% threshold.

[0112] In this embodiment of the invention, advanced alarm messages are severe warnings for multiple indicators or persistent exceedances, requiring management intervention. For example, an alarm sent to the production mine manager's mobile phone reads: "Severe Alarm! Electric shovel No. 3 in area E5 of the stope has simultaneously exceeded the dilution rate and loss rate limits, which has persisted for 10 minutes. Please handle immediately!"

[0113] In this embodiment of the invention, the anomaly diagnosis result is the output of the anomaly diagnosis unit, which includes a comprehensive report containing the cause of the anomaly, its location, the responsible party, and handling recommendations. For example, root cause diagnosis: the boundary of the ore and rock in the northwest corner of area E5 is inaccurate; recommended measures: adjust the current position of the electric shovel; arrange for surveyors to verify the boundary; update the geological model.

[0114] In this embodiment of the invention, the optimal decision-making module includes:

[0115] Simulation Unit: The simulation engine simulates the execution process of each mining scheme according to the first coupling rule and the second coupling rule, obtains simulation results, and determines the simulation output indicators based on the simulation results;

[0116] Constraint determination unit: Determine the optimization objective based on minimizing the dilution rate, minimizing the loss rate, and maximizing the ore production, and determine the constraint conditions based on equipment capacity constraints, resource reserve constraints, and safety boundary constraints;

[0117] Decision-making unit: Based on constraints and optimization objectives, determine a multi-objective optimization algorithm, comprehensively evaluate the simulation output indicators of each mining scheme, solve for the Pareto optimal solution set, and select the optimal decision scheme from the Pareto optimal solution set based on real-time production strategy preferences.

[0118] In this embodiment of the invention, the equipment capacity constraint is a specific constraint condition, meaning that the planned workload of the scheme cannot exceed the physical limits of the equipment itself. For example, a scheme requires an electric shovel to load 50,000 tons of material per day, but even if the shovel operates 24 hours a day, its maximum capacity is only 30,000 tons per day. This mining scheme violates the equipment capacity constraint and is not feasible.

[0119] In this embodiment of the invention, the resource reserve constraint is a specific constraint condition, meaning that mining activities must be limited to the known and available resource area. For example, a plan might be to mine ore in a certain area, but a geological model shows that the area is a waste rock zone with no economically valuable ore. This mining plan would violate the resource reserve constraint.

[0120] In this embodiment of the invention, the safety boundary constraint is a specific constraint condition, meaning that any operation must be based on ensuring the safety of personnel, equipment, and the environment. For example, a certain plan, in order to mine more ore, designed an excessively steep slope. Stability analysis revealed that this slope has a risk of collapse. This mining plan violates the safety boundary constraint and must be rejected.

[0121] In this embodiment of the invention, multi-objective optimization algorithms are the core mathematical tools used by the decision-making unit, specifically designed to handle complex decision problems involving conflicting objectives. For example, NSGA-II, by simulating the processes of natural selection and genetic variation, selects the best solutions through generations of computation.

[0122] In this embodiment of the invention, the simulation output indicators are specific data results obtained from the simulation unit and used to evaluate the merits of each mining scheme. For example, for a certain mining scheme, its simulation output indicators might be: dilution rate = 15%, loss rate = 5%, and ore production = 800,000 tons. This is the basic data for the algorithm to evaluate and compare.

[0123] In this embodiment of the invention, the Pareto optimal solution set is the result obtained after calculation by a multi-objective optimization algorithm, representing a set of optimal solutions. Each mining solution in the set is non-dominated, meaning that a more comprehensive solution cannot be found. For example, solution A has a low depletion rate but also low production; solution B has a high production rate but also a high depletion rate; and solution C has moderate performance in both metrics. None of these three mining solutions completely outperforms the others in all metrics, collectively constituting the Pareto optimal solution set.

[0124] In this embodiment of the invention, the real-time production strategy preference is the focus and priority determined based on current actual needs among numerous optimization objectives. This is the key to translating mathematical results into practical decisions. For example, if the current ore market price is high, the company's strategy preference is to maximize output to seize market opportunities; however, next month, the focus may be more on cost savings, and the strategy preference would change to minimizing the depletion rate to reduce waste rock processing costs.

[0125] In this embodiment of the invention, the optimal decision scheme is the final output of the decision-making unit. It is the scheme that best meets the current management intention, selected from the Pareto optimal solution set based on real-time production strategy preferences. For example, under the strategy preference of maximizing output, the optimal decision scheme finally selected from the Pareto optimal solution set {Scheme A, Scheme B, Scheme C} is Scheme B, because although the depletion rate is high, the output is the highest.

[0126] In this embodiment of the invention, the decision-making unit receives and processes the simulation output indicators from the simulation unit, which are the performance data of each mining scheme. It then strictly filters the schemes according to the constraints set in the first step, eliminating all unqualified schemes. For example, if a scheme is found to have a production capacity of up to 1.2 million tons, but violates the constraint that production should not exceed 1 million tons, it is immediately eliminated.

[0127] In this embodiment of the invention, the multi-objective optimization algorithm calculates the remaining feasible solutions, and the decision-making unit initiates the multi-objective optimization algorithm. This algorithm, guided by the optimization objective, comprehensively evaluates and ranks the simulation output indicators of the solutions, ultimately outputting a Pareto optimal solution set. For example, after calculation, the algorithm outputs three optimal solutions: Solution X has a 10% depletion rate and a production of 900,000 tons; Solution Y has a 15% depletion rate and a production of 990,000 tons; and Solution Z has a 20% depletion rate and a production of 1,000,000 tons.

[0128] In this embodiment of the invention, the decision-making unit introduces the value judgment of real-time production strategy preference, and selects the final optimal decision scheme from a mathematically equal set of Pareto optimal solutions. For example, if the current preference is to maintain output, scheme Z is selected; if the preference is to control poverty, scheme X is selected.

[0129] The simulation unit includes:

[0130] Rule simulation block: Calculates the geological-production response of each mining scheme during execution based on the first coupling rule, predicts the block size distribution of each mining scheme, and simulates the intelligent behavior of equipment during execution based on the second coupling rule to generate equipment actions;

[0131] Indicator Determination Block: Data fusion and production process simulation are performed on the block degree distribution and equipment actions to obtain simulation results. The simulation results are then quantitatively analyzed through a predefined indicator calculation model to finally determine the simulation output indicators.

[0132] In this embodiment of the invention, the geological-production response refers to the state change of a geological body under a specific mining scheme due to the action of production activities, and the chain reaction of this change on subsequent production stages. This is a causal process mediated by geological properties. For example, the response process: the production activity is blasting operations acting on a specific rock mass; the geological state change: the rock mass is broken, forming a specific block size distribution, which is the direct result of this response. The impact on production: this block size distribution directly determines the operating efficiency of subsequent loading equipment, the input load of the crusher, and energy consumption. For example, an excessively high proportion of large blocks will lead to an increase in loading cycle time and a decrease in crusher throughput.

[0133] In this embodiment of the invention, block size distribution is a quantitative description of the degree of rock fragmentation after blasting, typically characterized by a grain size distribution function. Its key parameters include characteristic grain size and uniformity index. For example, the predicted block size distribution result for a blasting scheme is: characteristic grain size d = 35 cm, uniformity index n = 1.0. This means that after the blast, approximately 63.2% of the rock blocks will have a grain size smaller than 35 cm. This data can be directly used to evaluate the blasting effect and predict the processing capacity of subsequent equipment.

[0134] In this embodiment of the invention, intelligent device behavior refers to the strategic capability, implemented by the second coupling rule, of perceiving environmental conditions and making autonomous decisions. Device actions are the specific outputs and execution instructions of intelligent device behavior, serving as the underlying commands controlling the virtual device to perform specific operations in the simulation environment. For example, intelligent behavior: A scheduling model for an unmanned loader determines which workface to proceed to for loading based on real-time updated ore grade distribution maps, its own power supply, and the length of the crushing station queue. This is a decision-making process. Device actions: The specific action instruction output by the scheduling model might be: {Device ID: LHD_03, Action Type: Driving, Target Path Point: (x100, y200, z10), Expected Speed: 20 km / h}. The simulation engine will parse and execute this instruction.

[0135] In this embodiment of the invention, the input of the predefined index calculation model is the simulation result, that is, the complete set of system states D(t) with timestamps after data fusion and reconstruction. This is a high-dimensional dataset. The output of the predefined index calculation model is the simulation output index, that is, the low-dimensional key performance indicator vector K extracted from the input data for evaluating the performance of the scheme.

[0136] In this embodiment of the invention, the simulation output indicators are a set of quantitative indicators used to comprehensively evaluate the overall performance of the mining scheme. These indicators are the final results of the simulation operation, providing direct and objective data basis for scheme comparison and optimization decisions, including: technical indicators: cumulative dilution rate, cumulative loss rate, total ore production, average block size; economic indicators: total production cost, cost per ton of ore, expected revenue; efficiency and effectiveness indicators: equipment comprehensive utilization rate, system throughput, task completion rate.

[0137] In this embodiment of the invention, the geological-production response is a causal mechanism, and block size distribution is a key outcome of this mechanism. Intelligent equipment behavior is the strategy for generating decisions, and equipment actions are the specific execution instructions of this strategy. The input to the predefined index calculation model is massive amounts of raw simulation data, and the output is highly condensed simulation output indices. These indices serve as the final scale for evaluating the merits of the scheme and the basis for decision-making.

[0138] The continuous optimization module includes:

[0139] Instruction generation unit: Input the optimal decision scheme into the instruction parsing engine. The instruction parsing engine integrates the mechanism rule base and the data-driven strategy base. Based on the mechanism rule base, it converts the scheme parameters of the optimal decision scheme into basic control instructions. Based on the data-driven strategy base, it performs local optimization and adjustment on the basic control instructions to generate the final control instruction sequence.

[0140] Model optimization unit: Collects the execution results of the final control commands, optimizes the mechanism rule base and data-driven strategy base based on the execution results, and then optimizes the twin mining area model.

[0141] In this embodiment of the invention, the instruction parsing engine is a software module responsible for converting high-level production decision schemes into low-level, executable equipment control instructions. Its core function is to translate, optimize, and serialize instructions. For example, a software component takes as input a blasting scheme using a 5m×5m hole spacing in the E5 area, and outputs a series of instructions for the automatic drilling rig: G01 X100 Y200 Z-15; G01 X105 Y200 Z-15.

[0142] In this embodiment of the invention, the mechanism rule base is a knowledge base that stores conversion rules based on physical laws and engineering principles, ensuring the feasibility and security of instruction generation in the physical world. For example, the rule base includes: IF [rock hardness > 150MPa] THEN [drilling spacing coefficient = 0.8]; IF [equipment is a hydraulic shovel] THEN [maximum allowable digging force = 250kN]. These rules convert scheme parameters into instructions that conform to physical laws.

[0143] In this embodiment of the invention, the data-driven strategy library is a knowledge base storing optimization strategies generated by machine learning models. These strategies are learned from historical data and used to fine-tune basic instructions to improve efficiency, reduce energy consumption, or reduce wear and tear. For example, a strategy trained by a reinforcement learning model: during peak electricity price periods and when the equipment is idle, reduce the driving speed from the default 25 km / h to 20 km / h to reduce energy costs.

[0144] In this embodiment of the invention, the scheme parameters describe the key variables and settings of the optimal decision scheme, and are the input of the instruction parsing engine, existing at the planning and decision-making level. For example, {blasting area: E5, hole spacing: 5m, row spacing: 5m, ultra-deep: 1m, planned output: 5000 tons / shift}.

[0145] In this embodiment of the invention, the basic control commands are standard equipment control commands directly generated from the mechanism rule base, conforming to physical laws but not optimized for operating conditions. For example, based on the scheme parameters of a 5m hole spacing and the mechanism rules, the generated basic command is: the drilling rig drills to the designed depth at all hole coordinate points.

[0146] In this embodiment of the invention, local optimization adjustment is an adaptive fine-tuning performed by the data-driven strategy library on top of the basic control commands. Based on real-time context information such as equipment status, environmental factors, and cost factors, the command parameters are finely optimized. For example, after reading real-time data, the data-driven strategy library performs local optimization adjustment on the aforementioned basic commands: at coordinates (X105, Y200), due to the softness of the rock mass, the drilling depth is adjusted from -15m to -14.5m to save energy and reduce excessive fragmentation.

[0147] In this embodiment of the invention, the final control command sequence is an ordered set of commands that can be directly issued to a specific device controller after being converted from a mechanism rule base and optimized by a data-driven strategy base. For example, a command sequence sent to an automated drilling rig: [Command 1: Move to (X100, Y200); Command 2: Drill to Z-14.5m; Command 3: Pull up the drill; Command 4: Move to (X105, Y200); Command 5: Drill to Z-15.0m;...].

[0148] refer to Figure 2 This invention also provides a digital twin simulation optimization method for mining dilution rate and loss rate, including:

[0149] Step 1: Collect geological and production data of the mining area, construct a geological model based on the geological data, construct a production model based on the production data, and merge the geological model and the production model to construct a twin mining area model;

[0150] Step 2: Based on the twin mining area model and real-time mining area data, machine learning algorithms are used to identify the ore-rock mixing state, calculate the instantaneous dilution rate and instantaneous loss rate, and perform anomaly diagnosis to obtain the anomaly diagnosis results;

[0151] Step 3: Based on the anomaly diagnosis results, start the simulation engine to perform parallel simulations of multiple mining schemes, and perform Pareto optimal selection on the simulation results to obtain the optimal decision scheme;

[0152] Step 4: Parse the optimal decision scheme into control commands and send them to the production equipment for execution, and feed the execution results back to the twin mining area model for continuous optimization.

[0153] The digital twin simulation optimization method for mining dilution rate and loss rate provided in this embodiment of the invention can be implemented using the aforementioned digital twin simulation optimization system for mining dilution rate and loss rate.

[0154] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Additionally, the terms "front," "back," "left," "right," "upper," and "lower" in this document refer to the placement shown in the accompanying drawings.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital twin simulation optimization system for mining dilution rate and loss rate, characterized in that, include: Model building module: Collects geological and production data of the mining area, builds a geological model based on the geological data, builds a production model based on the production data, and merges the geological model and the production model to build a twin mining area model; Anomaly Diagnosis Module: Based on the twin mining area model and real-time mining area data, the module uses machine learning algorithms to identify the ore-rock mixing state, calculates the instantaneous dilution rate and instantaneous loss rate, performs anomaly diagnosis, and obtains the anomaly diagnosis results. Optimal Decision Module: Based on the anomaly diagnosis results, the simulation engine is started to perform parallel simulations of multiple mining schemes, and the simulation results are Pareto optimally screened to obtain the optimal decision scheme; Continuous optimization module: It parses the optimal decision scheme into control commands and issues them to the production equipment for execution, and feeds back the execution results to the twin mining area model to continuously optimize the twin mining area model; The method of integrating geological models and production models to construct a twin mining area model includes: The geological model and the production model are imported into a unified digital twin simulation platform. The first coupling rule between the geological model and the production model is established through the coupling mechanism algorithm, and the second coupling rule between the geological model and the production model is established through the machine learning algorithm. The geological model and the production model are fused based on the first coupling rule and the second coupling rule to obtain a twin mining area model; Among them, the key physical processes that directly affect the production behavior in the production model are identified based on the geological attributes in the geological model by the digital twin simulation platform, and a geological-production association list is generated based on the key physical processes. A corresponding mechanism algorithm table is developed based on the geological-production association list, and the bidirectional triggering conditions are determined based on the mechanism algorithm table to obtain the first coupling rule. Among them, the correlation features are extracted from the geological model and the production model, and the business needs and production pain points of the mining area are obtained. The limitation points of the first coupling rule are extracted, the coupling problem is determined based on the business needs, production pain points and limitation points, the machine learning algorithm is selected based on the coupling problem, and the second coupling rule is determined by combining the correlation features and the machine learning algorithm. The process of using a twin mining area model and real-time mining area data, employing machine learning algorithms to identify the ore-rock mixing state, calculating the instantaneous dilution rate and instantaneous loss rate, and performing anomaly diagnosis to obtain anomaly diagnosis results includes: The pre-trained convolutional neural network model is invoked to perform frame-by-frame analysis of the real-time video stream of loading operations in the mining area, identify the ore-rock mixing state of ore and waste rock in the bucket, and output the estimated mixing ratio of the identified ore-rock mixing state. Based on the estimated mixed ratio, combined with the real-time load data of equipment in the real-time data of the mining area and the theoretical grade data of the blocks in the twin mining area model, the instantaneous dilution rate and instantaneous loss rate of the current working unit are dynamically calculated according to the mining engineering calculation formula. Based on the preset impoverishment rate threshold and the preset loss rate threshold, the instantaneous impoverishment rate is compared with the preset impoverishment rate threshold for the first time, and the instantaneous loss rate is compared with the preset loss rate threshold for the second time. When either the instantaneous impoverishment rate or the instantaneous loss rate exceeds its corresponding preset threshold, a first-level warning is triggered and a primary alarm message is generated. When both the instantaneous impoverishment rate and the instantaneous loss rate exceed their corresponding preset thresholds simultaneously, or when either the instantaneous impoverishment rate or the instantaneous loss rate continues to exceed the limit for a preset duration, a second-level alarm is triggered and a high-level alarm message is generated. Based on the twin mining area model, anomaly tracing is initiated for primary and advanced alarm information to obtain the tracing results. The anomaly diagnosis of the mining area is then performed by combining the primary alarm information, advanced alarm information, and tracing results to obtain the anomaly diagnosis results.

2. The digital twin simulation optimization system for mining dilution rate and loss rate according to claim 1, characterized in that, The model building module includes: Basic processing unit: Performs basic processing on the collected geological and production data; Fitting Unit: Establish a structured geological database based on the geological data after basic processing; calculate the variogram for different geographical directions based on the grade data in the geological data, fit the variogram values ​​with the preset theoretical variogram model, and derive the theoretical variogram data model characterizing the spatial correlation of grade based on the fitting results. Geological model construction unit: Based on the structured geological database, the orebody wireframe model is determined. The theoretical variation data model is input into the orebody wireframe model to spatially estimate the average grade of each block in the mining area, thereby obtaining the grade distribution attributes of the mining area. Combining the theoretical variation data model with the lithology-mechanics mapping table, the rock mass mechanical properties of each block in the mining area are determined. By integrating the grade distribution attributes and rock mass mechanical properties of each block, the geological model of the mining area is constructed. Logical model building unit: Determines the logical model of the production process based on production data; Equipment Behavior Model Construction Unit: Extract equipment-related information from the basic processed production data, construct the basic equipment model according to the equipment design documents, bind the real-time pose data and motion state data of each device in the equipment-related information to the basic equipment model, establish the mapping relationship between the spatial pose and motion state of the device and the model animation frames based on the binding results, and obtain the equipment behavior model; Production model construction unit: Integrates the logical model and equipment behavior model in a unified 3D platform to obtain the production model; Coupling rule determination unit: Import the geological model and the production model into a unified digital twin simulation platform, establish the first coupling rule between the geological model and the production model through the coupling mechanism algorithm, and establish the second coupling rule between the geological model and the production model through the machine learning algorithm; Twin mining area model construction unit: Based on the first coupling rule and the second coupling rule, the geological model and the production model are fused to obtain the twin mining area model.

3. The digital twin simulation optimization system for mining dilution rate and loss rate according to claim 2, characterized in that, The coupling rule determination unit includes: The first coupling block: Based on the digital twin simulation platform, the key physical processes in the geological model that directly affect the production behavior in the production model are identified, and a geological-production association list is generated based on the key physical processes. A corresponding mechanism algorithm table is developed based on the geological-production association list, and the bidirectional triggering conditions are determined based on the mechanism algorithm table to derive the first coupling rule. The second coupling block extracts correlation features from the geological model and production model, while simultaneously acquiring the business needs and production pain points of the mining area, extracting the limitations of the first coupling rule, determining the coupling problem based on the business needs, production pain points, and limitations, selecting a machine learning algorithm based on the coupling problem, and determining the second coupling rule by combining the correlation features and the machine learning algorithm.

4. The digital twin simulation optimization system for mining dilution rate and loss rate according to claim 1, characterized in that, The anomaly diagnosis module includes: Ore-rock mixing state identification unit: calls a pre-trained convolutional neural network model to perform frame-by-frame analysis of the real-time video stream of loading operations in the mining area, identifies the ore-rock mixing state of ore and waste rock in the bucket, and outputs the estimated mixing ratio of the identified ore-rock mixing state. Calculation Unit: Based on the estimated mixed ratio, combined with the real-time equipment load data in the real-time data of the mining area and the block theoretical grade data in the twin mining area model, the instantaneous dilution rate and instantaneous loss rate of the current working unit are dynamically calculated according to the mining engineering calculation formula. Alarm Information Determination Unit: Based on the preset impoverishment rate threshold and the preset loss rate threshold, the instantaneous impoverishment rate is compared with the preset impoverishment rate threshold for the first time, and the instantaneous loss rate is compared with the preset loss rate threshold for the second time. When either the instantaneous impoverishment rate or the instantaneous loss rate exceeds its corresponding preset threshold, a first-level warning is triggered and a primary alarm information is generated. When both the instantaneous impoverishment rate and the instantaneous loss rate exceed their corresponding preset thresholds simultaneously, or when either the instantaneous impoverishment rate or the instantaneous loss rate continues to exceed the limit for a preset duration, a second-level alarm is triggered and a high-level alarm information is generated. Anomaly Diagnosis Unit: Based on the twin mining area model, it initiates anomaly tracing of primary and advanced alarm information, obtains the tracing results, and performs anomaly diagnosis on the mining area by integrating the primary alarm information, advanced alarm information, and tracing results, and obtains the anomaly diagnosis results.

5. The digital twin simulation optimization system for mining dilution rate and loss rate according to claim 2, characterized in that, The optimal decision module includes: Simulation Unit: The simulation engine simulates the execution process of each mining scheme according to the first coupling rule and the second coupling rule, obtains simulation results, and determines the simulation output indicators based on the simulation results; Constraint determination unit: Determine the optimization objective based on minimizing the dilution rate, minimizing the loss rate, and maximizing the ore production, and determine the constraint conditions based on equipment capacity constraints, resource reserve constraints, and safety boundary constraints; Decision-making unit: Based on constraints and optimization objectives, determine a multi-objective optimization algorithm, comprehensively evaluate the simulation output indicators of each mining scheme, solve for the Pareto optimal solution set, and select the optimal decision scheme from the Pareto optimal solution set based on real-time production strategy preferences.

6. The digital twin simulation optimization system for mining dilution rate and loss rate according to claim 5, characterized in that, The simulation unit includes: Rule simulation block: Calculates the geological-production response of each mining scheme during execution based on the first coupling rule, predicts the block size distribution of each mining scheme, and simulates the intelligent behavior of equipment during execution based on the second coupling rule to generate equipment actions; Indicator Determination Block: Data fusion and production process simulation are performed on the block degree distribution and equipment actions to obtain simulation results. The simulation results are then quantitatively analyzed using a predefined indicator calculation model to determine the simulation output indicators.

7. The digital twin simulation optimization system for mining dilution rate and loss rate according to claim 1, characterized in that, The continuous optimization module includes: Instruction generation unit: Inputs the optimal decision scheme into the instruction parsing engine. The instruction parsing engine integrates a mechanism rule base and a data-driven strategy base. Based on the mechanism rule base, it converts the scheme parameters of the optimal decision scheme into basic control instructions. Based on the data-driven strategy base, it performs local optimization and adjustment on the basic control instructions to generate the final control instruction sequence. Model optimization unit: Collects the execution results of the final control commands, optimizes the mechanism rule base and data-driven strategy base based on the execution results, and then optimizes the twin mining area model.

8. A digital twin simulation optimization method for mining dilution rate and loss rate, characterized in that, include: Step 1: Collect geological and production data of the mining area, construct a geological model based on the geological data, construct a production model based on the production data, and merge the geological model and the production model to construct a twin mining area model; Step 2: Based on the twin mining area model and real-time mining area data, machine learning algorithms are used to identify the ore-rock mixing state, calculate the instantaneous dilution rate and instantaneous loss rate, and perform anomaly diagnosis to obtain the anomaly diagnosis results; Step 3: Based on the anomaly diagnosis results, start the simulation engine to perform parallel simulations of multiple mining schemes, and perform Pareto optimal selection on the simulation results to obtain the optimal decision scheme; Step 4: Parse the optimal decision scheme into control commands and send them to the production equipment for execution, and feed the execution results back to the twin mining area model for continuous optimization; The method of integrating geological models and production models to construct a twin mining area model includes: The geological model and the production model are imported into a unified digital twin simulation platform. The first coupling rule between the geological model and the production model is established through the coupling mechanism algorithm, and the second coupling rule between the geological model and the production model is established through the machine learning algorithm. The geological model and the production model are fused based on the first coupling rule and the second coupling rule to obtain a twin mining area model; Among them, the key physical processes that directly affect the production behavior in the production model are identified based on the geological attributes in the geological model by the digital twin simulation platform, and a geological-production association list is generated based on the key physical processes. A corresponding mechanism algorithm table is developed based on the geological-production association list, and the bidirectional triggering conditions are determined based on the mechanism algorithm table to obtain the first coupling rule. Among them, the correlation features are extracted from the geological model and the production model, and the business needs and production pain points of the mining area are obtained. The limitation points of the first coupling rule are extracted, the coupling problem is determined based on the business needs, production pain points and limitation points, the machine learning algorithm is selected based on the coupling problem, and the second coupling rule is determined by combining the correlation features and the machine learning algorithm. The process of using a twin mining area model and real-time mining area data, employing machine learning algorithms to identify the ore-rock mixing state, calculating the instantaneous dilution rate and instantaneous loss rate, and performing anomaly diagnosis to obtain anomaly diagnosis results includes: The pre-trained convolutional neural network model is invoked to perform frame-by-frame analysis of the real-time video stream of loading operations in the mining area, identify the ore-rock mixing state of ore and waste rock in the bucket, and output the estimated mixing ratio of the identified ore-rock mixing state. Based on the estimated mixed ratio, combined with the real-time load data of equipment in the real-time data of the mining area and the theoretical grade data of the blocks in the twin mining area model, the instantaneous dilution rate and instantaneous loss rate of the current working unit are dynamically calculated according to the mining engineering calculation formula. Based on the preset impoverishment rate threshold and the preset loss rate threshold, the instantaneous impoverishment rate is compared with the preset impoverishment rate threshold for the first time, and the instantaneous loss rate is compared with the preset loss rate threshold for the second time. When either the instantaneous impoverishment rate or the instantaneous loss rate exceeds its corresponding preset threshold, a first-level warning is triggered and a primary alarm message is generated. When both the instantaneous impoverishment rate and the instantaneous loss rate exceed their corresponding preset thresholds simultaneously, or when either the instantaneous impoverishment rate or the instantaneous loss rate continues to exceed the limit for a preset duration, a second-level alarm is triggered and a high-level alarm message is generated. Based on the twin mining area model, anomaly tracing is initiated for primary and advanced alarm information to obtain the tracing results. The anomaly diagnosis of the mining area is then performed by combining the primary alarm information, advanced alarm information, and tracing results to obtain the anomaly diagnosis results.

Images