Mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization

Abstract

The present application belongs to the technical field of open-pit metal mine exploitation, and discloses a mine transportation process optimization method based on three-stage coordinated transportation and intelligent optimization. In view of the problems of low efficiency and poor coordination of the traditional open-pit mine transportation system, the present application firstly establishes a short-range transfer connection system to realize rapid collection and transportation of ore and rock, secondly adopts a movable crushing station for in-situ crushing operation, and finally completes large-inclination transportation of ore by a large-inclination continuous lifting device. At the intelligent decision-making level, an intelligent optimization system of multi-algorithm fusion is proposed, that is, a deep neural network (DBO-DNN) improved by the scarab beetle optimization algorithm is used to construct a transportation system process performance prediction model, production scheduling is modeled as a mixed integer programming problem, and a genetic algorithm (GA) is used for solution. The integrated scheme can improve the comprehensive performance of the transportation system and reduce the relocation frequency of the crushing station, thereby providing an intelligent transportation process mode with autonomous decision-making capability for continuous mining of open-pit metal mines.

Description

A Mining Transportation Process Optimization Method Based on Three-Stage Collaborative Transportation and Intelligent Optimization Technical Field

[0001] This invention relates to the field of open-pit metal mining technology, and in particular to a mining transportation process optimization system and method based on three-stage collaborative transportation and intelligent optimization, which is especially suitable for continuous mining operations in deep-pit metal mines. Background Technology

[0002] In open-pit mining, there are several different transportation technologies, including intermittent transportation, continuous transportation, and semi-continuous transportation.

[0003] Currently, over 80% of open-pit mines worldwide use intermittent shovel-truck transport for loading and transportation operations. When using the shovel-truck model, the equipment idle rate reaches 38%-45%, and the transportation cost per ton of ore accounts for over 42% of the total mining cost, resulting in low efficiency, discontinuous transportation, and significant environmental pollution. Continuous transport refers to the process of transporting ore directly via conveyor belts after mining. Coal, with its low hardness and fragility, can be directly transported using continuous mechanical equipment, making this process widely used in open-pit coal mines. In contrast, metallic ores have higher hardness and require complex crushing processes, making this process difficult to apply effectively in open-pit metal mines. Semi-continuous transport reduces the need for traditional trucks by directly crushing and transporting materials within the mining pit. This process combines the advantages of both intermittent and continuous processes, retaining the flexibility and adaptability of the mining process while achieving continuous ore transportation. It is particularly suitable for mining medium-hard rock ores, effectively reducing transportation costs and showing broad development prospects in the field of open-pit metal mining.

[0004] However, semi-continuous transportation technology is less flexible in adapting to changes in the mining area. When the ore deposit layout or mining plan changes, adjusting the production scheduling plan and the optimal relocation plan for the crushing station is very complex and costly. This makes metal open-pit mines face many challenges when using semi-continuous transportation technology, such as discontinuous transportation process, high transportation costs, low mobile crushing capacity, and high energy consumption of mobile crushing. Summary of the Invention

[0005] The purpose of this invention is to overcome the prominent problems faced by current open-pit metal mining, such as discontinuous transportation processes, high transportation costs, low mobile crushing capacity, and high mobile crushing energy consumption. It proposes a mine transportation process optimization system and method based on three-stage collaborative transportation and intelligent optimization, which is a staged mobile crushing continuous transportation process mode with low transportation costs, high mobile crushing capacity, and low mobile crushing energy consumption.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: a mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization, including a three-stage collaborative transportation process architecture and an intelligent decision-making system; the three-stage collaborative transportation process architecture provides a continuous process of ore transportation and crushing, and the intelligent decision-making system provides decision support for the three-stage collaborative transportation process architecture by predicting and optimizing production scheduling and crushing station relocation;

[0007] The three-stage collaborative transportation process architecture is as follows: based on the geometry, geological characteristics, mining technology, equipment selection, and transportation conditions of the ore body, the mining area is divided into a hierarchical structure from top to bottom as "section-block-strip", and a continuous transportation system is constructed consisting of a short-distance transfer module, a mobile crushing module, and a steep-angle lifting module. The short-distance transfer module is equipped with electric mining trucks for ore and rock collection and transportation, the mobile crushing module uses a semi-mobile crushing station for in-situ crushing operations, and the steep-angle lifting module completes steep-angle transportation of ore through a belt conveyor.

[0008] The intelligent decision-making system constructs a DBO-DNN transportation system performance prediction model based on a deep neural network improved by the dung beetle optimization algorithm, and models production scheduling and crushing station relocation as a mixed integer programming problem. By using the output of the DBO-DNN transportation system performance prediction model as data input and combining it with an improved genetic algorithm for solution, the transportation system can be optimized.

[0009] The three-stage collaborative transportation process architecture specifically includes:

[0010] a) Divide the mining area into three levels of mining units: section, block, and strip;

[0011] b) Using electric shovels for ore mining operations in strip units;

[0012] c) Short-distance transportation of ore to the semi-mobile crushing station using mining trucks;

[0013] d) The crushed ore is vertically lifted and transported by a steep-angle belt conveyor; the steep-angle belt conveyor has an inclination range of 35°-55° and a lifting speed of 2.5-4m / s.

[0014] The method for constructing the DBO-DNN transportation system performance prediction model includes: using the dung beetle optimization algorithm to globally optimize the initial hyperparameters of the deep neural network, establishing a three-layer network structure with an adaptive activation function; input feature parameters include 18 indicators in three categories: mining operation parameters, equipment status parameters, and environmental parameters; output parameters are transportation process performance indicators, including three core indicators: ore output, energy consumption per ton of ore, and transportation cost, with a determination coefficient R0. 2 ≤3.2%.

[0015] The input feature parameters specifically include: mining operation parameters: start and end time of operation, number of trucks, carrying capacity, running distance, and crushing volume; equipment status parameters: truck weight, conveyor layout length, and conveyor belt speed; and environmental parameters: altitude, weather conditions, ambient temperature, snow thickness, energy price, ore grade, and market price.

[0016] The mathematical model establishment of the optimization module includes: establishing a multi-objective optimization function under three-dimensional spatiotemporal constraints with the goal of maximizing total profit.

[0017] (1) Taking the maximization of total profit as the objective function, and with constraints such as mining volume limit, transportation capacity limit, crusher load limit, annual mining rate constraint, maximum beneficiation capacity limit, mining sequence constraint, and crushing station relocation constraint, the optimal production scheduling plan and the optimal relocation plan of the mobile crushing continuous transportation process are expressed as integer programming. Based on the DBO-DNN transportation system performance prediction model, a mathematical model that can simultaneously optimize production scheduling and crushing station relocation is established:

[0018] Objective function:

[0019] In the formula, Z represents the total value, T represents the life cycle of the mine, N represents the total number of strips, and H represents the total number of layers;

[0020] V nt This represents the total value of strip n within time period t: ;

[0021] In the formula, S n V represents the total number of blocks in strip n; Bst This represents the total value of block s over time period t: ;

[0022] In the formula, v Bs Let Gr be the volume of block s. Bs This indicates the grade of block s. Let represent the density of block s, φ represent the depletion rate, η represent the recovery rate, UP represent the price per unit grade, and r represent the discount rate;

[0023] : Indicates whether strip n is mined during time period t, 1 indicates mining, 0 indicates no mining;

[0024] The transportation cost of strip n from the crushing station at layer h during time period t: ;

[0025] : Indicates whether the conveyor used by strip n in time period t is on layer h, 1 indicates yes, 0 indicates no;

[0026] C Rh Indicates the relocation cost at level h: ;

[0027] C R Based on the base cost, α is the unit cost coefficient that increases with the number of levels. As h increases, the relocation cost will gradually increase.

[0028] : Indicates whether the crushing station has been relocated to layer h during time period t, 1 indicates relocation, 0 indicates no relocation;

[0029] λ represents the carbon emission control weight, and e represents the carbon emissions generated per unit of ore transport using the SMIPCC system;

[0030] The total mass of strip n is represented by: ,in, S represents the mass of block s. n The number of blocks in strip n;

[0031] Three types of constraints are set: strip mining volume constraint, crushing station processing capacity constraint, and crushing station relocation constraint. An improved genetic algorithm is used to solve the problem, with chromosomes divided into two gene segments: transportation and relocation decision. The optimization is dynamically adjusted through initialization, fitness evaluation, improved roulette wheel selection, crossover, and mutation operations. Population updates include elite retention and competitive replacement. The termination condition is a maximum of 200 iterations or a fitness improvement of less than 0.1% for 20 consecutive generations. The optimal solution, load distribution map, and convergence curve are output.

[0032] The constraints include:

[0033] Mining volume limit: In each time period t, the total mining volume cannot exceed the maximum mining capacity M. t :

[0034]

[0035] Transportation capacity limit: In each time period t, the total mining volume cannot exceed the truck transportation capacity TS. t :

[0036]

[0037] Crusher load limit: The total mining volume in each time period t cannot exceed the crusher load L. t :

[0038]

[0039] Annual mining rate constraint: The annual mining volume shall not exceed β times the mining volume of the previous year to avoid excessively rapid mining.

[0040]

[0041] In the formula, β is a coefficient that controls the increase in mining output;

[0042] Maximum processing capacity limit for mineral processing: The total amount of ore mined in each time period t cannot exceed the maximum processing capacity D of the mineral processing plant. t :

[0043]

[0044] In the formula, The average ore grade of band n: ;

[0045] In the formula, Let S be the grade of block s. n The number of blocks in stripe n;

[0046] Each strip can only be mined once: ;

[0047] Mining sequence constraint: Ensure that strip n precedes strip n. pr Strip n can only begin mining after it has been completely mined out.

[0048]

[0049] In the formula, PR n Let n be the set of previous stripes;

[0050]

[0051] HP n For the set of previous horizontal stripes of stripe n, VP n Let n be the set of previous stripes perpendicular to strip n;

[0052] Crusher hierarchical deployment constraints: A crusher can only be deployed at one level in each time period t.

[0053]

[0054] Crushing plant relocation constraint: qht is only 1 when the crushing plant level changes.

[0055]

[0056] Relocation sequence constraint: The crushing station can only be relocated from the upper level to the lower level.

[0057]

[0058] The maximum time span T for the relocation of the crushing plant max To avoid multiple relocations in a short period of time;

[0059]

[0060] Each crushing station can only be relocated once within a certain time period:

[0061] .

[0062] The improved genetic algorithm solves the problem as follows:

[0063] 1) Chromosome coding design:

[0064] The chromosomes of the genetic algorithm adopt a segmented coding structure, which includes the following three gene segments: transportation decision gene segment, crushing station location decision gene segment, and relocation decision gene segment;

[0065] Chromosome length: The length of the chromosome is N×T+N×H×T+H×T, where the first N×T positions are the transportation decision segment, the middle N×H×T positions are the crushing station location decision segment, and the last H×T positions are the relocation decision segment;

[0066] 2) Population initialization:

[0067] Generate the initial population:

[0068] Set the population size;

[0069] For each individual's chromosome, gene initialization is performed according to the following rules:

[0070] Each position in the transport decision gene segment is sampled independently, where o nt The probability of =1 is positively correlated with the ore reserves of strip n in time period t. To avoid excessive concentration in certain time periods, the activation probability is adjusted through standardization.

[0071] For the decision gene segment of the crushing station, randomly select a valid level h such that p nht =1, and make it satisfy the following formula: ;

[0072] Randomly generate relocation decision gene segment q nht =1;

[0073] For each individual, the following checks are performed, and if they are not met, they are repaired or removed: annual mining volume constraints; uniqueness and sequence constraints of strip mining; unique deployment constraints of crushing station hierarchy; relocation direction and time interval restrictions;

[0074] 3) Construct the fitness function:

[0075] Fitness calculation includes objective function evaluation and constraint penalty terms, and its mathematical expression is:

[0076]

[0077] Penalty item: Punishment for violations of constraints: ;

[0078] in, The penalty weight for the i-th constraint; violate i The degree of violation of the i-th constraint;

[0079] 4) Select Operation:

[0080] An improved roulette wheel selection strategy is adopted:

[0081] a) Calculate the normalized fitness value of the population: , where: f k f is the initial fitness value of individual k; min It is the minimum fitness value in the population; f max It is the maximum fitness value in the population. The denominator should be a very small positive number to avoid being zero.

[0082] b) Introduce a merit retention mechanism to directly replicate the top 5% of fit individuals to the next generation;

[0083] c) The remaining 95% of individuals use roulette to determine the probability: Selection is conducted to ensure that high-quality individuals have a higher probability of reproduction;

[0084] 5) Crossover operation: Crossover is performed independently on each gene segment. Crossover methods: Transport decision segment (N×T): two-point crossover, retaining continuous intervals and enhancing local structural stability; Crushing station location decision segment (N×H×T): uniform crossover, each position independently inherits 50% probability from the parent generation; Crushing station relocation decision segment (H×T): single-point crossover, only one relocation change is allowed to ensure compliance with constraints.

[0085] 6) Mutation operation: Implement a hierarchical mutation mechanism:

[0086] a) Baseline mutations: Each gene locus has a probability p m =0.01 to flip;

[0087] b) Taboo protection: Mutation of key gene segments is prohibited on the top 10% of individuals in terms of fitness;

[0088] c) Anomaly Repair: If a mutation causes a violation of a hard constraint, dynamic correction is performed to restore the valid solution space. The repair strategy is based on heuristics or rules.

[0089] 7) Population renewal:

[0090] a) Elite Preservation: Directly preserve the top 5% of individuals in terms of fitness in the current generation;

[0091] b) Competitive replacement: The newly generated individuals are merged with the remaining individuals in the original population, and the top K individuals are retained according to their fitness.

[0092] c) Diversity maintenance: Force the retention of at least 3% of individuals with low fitness but high genetic diversity;

[0093] 8) Termination determination and output:

[0094] The algorithm termination condition is set as follows:

[0095] a) The maximum number of iterations reaches G=200;

[0096] b) The improvement in optimal fitness over 20 consecutive generations is less than ε = 0.1%;

[0097] c) Obtain a feasible solution that satisfies all hard constraints;

[0098] The operation terminates when any condition is met, and the output includes: the transportation scheduling scheme and relocation plan corresponding to the optimal chromosome; the equipment load distribution diagram for each time period; and the convergence curve of the objective function.

[0099] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0100] 1. Reduced Transportation Costs: Based on the "section-block-strip" mining model, a new continuous transportation process of "short-distance transfer-mobile crushing-sloping-angle lifting" is implemented, significantly reducing the total transportation cost from the working face to the crusher. Compared with the traditional intermittent transportation process, the truck transportation distance is significantly reduced, transportation efficiency is significantly improved, and transportation costs are significantly reduced.

[0101] 2. Improve the efficiency and capacity of mobile crushing: This invention utilizes advanced optimization algorithms to provide optimal production scheduling and crushing station relocation plans for mobile crushing continuous transportation processes, ensuring the rational allocation and utilization of resources and significantly improving the processing capacity and efficiency of mobile crushing stations.

[0102] 3. Reduced energy consumption: Through accurate performance prediction models and optimization algorithms, energy efficiency is maximized throughout the transportation and crushing process.

[0103] 4. Improve overall system performance: By comprehensively applying advanced technologies such as deep neural networks, metaheuristic algorithms and network flow algorithms, this invention achieves comprehensive optimization of the performance of the entire transportation and crushing system, providing a new, efficient, economical and environmentally friendly solution for the field of open-pit metal mining. Attached Figure Description

[0104] Figure 1 is a schematic diagram of the transportation process of the present invention;

[0105] Figure 2 is a schematic diagram of the "section-block-strip" mining mode;

[0106] Figure 3 is a schematic diagram of the new continuous transportation process of "short-distance transfer-mobile crushing-sharp-angle conveyor"; (a) is short-distance transfer for continuous ore supply; (b) is staged sinking mobile crushing; (c) is "mining-loading-crushing-transporting-discharge" coordinated continuous process; (d) is steep-angle conveyor belt continuous transportation.

[0107] Figure 4 is a diagram of the topology of a deep neural network;

[0108] Figure 5 is a schematic diagram of the optimization process of mine transportation technology based on three-stage collaborative transportation and intelligent optimization. Detailed Implementation

[0109] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to Figures 1-5 and specific embodiments.

[0110] A mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization includes the following:

[0111] 1. Three-stage collaborative transportation process architecture: Based on the three-level mining unit division of "section-block-strip" (section size ≤ 500m, block size ≤ 200m, strip size ≤ 50m), a continuous transportation system is constructed, consisting of a short-distance transfer module, a semi-mobile crushing module, and a steep-angle lifting module. Short-distance transfer module: Electric mining trucks (load capacity 60-100t, range ≥ 8h) are used for ore collection and transportation, with a transportation radius ≤ 1.5km; Semi-mobile crushing module: Modular crushing station is configured (processing capacity 2000-3000t / h, relocation time ≤ 72h) to implement in-situ crushing to a particle size ≤ 300mm; Steep-angle lifting module: Vertical transportation to the surface processing station is achieved via high-strength belt conveyors (inclination angle 35°-55°, belt speed 2.5-4m / s).

[0112] 2. The specific steps of prediction in an intelligent decision-making system are as follows:

[0113] Step 1: Optimize key hyperparameters in the deep neural network using the dung beetle optimization algorithm. Specifically, the dung beetle optimization algorithm is used to optimize hyperparameters such as learning rate, training cycle, activation function, regularization parameter, and network structure to improve the training performance of the deep neural network (DNN). By optimizing these parameters, an accurate training model can be constructed, namely the DBO-DNN transportation system performance prediction model.

[0114] Step 2: Collect real production data of the continuous transportation process for mobile crushing in open-pit metal mines, and use it as input and output parameters for the DBO-DNN transportation system performance prediction model. The production data includes the following categories:

[0115] 1. Input features (feature data):

[0116] Mining operation parameters: including start and end times, number of trucks, carrying capacity, travel distance, and crushing volume;

[0117] Equipment status parameters include truck tare weight, conveyor layout length, and conveyor belt speed.

[0118] Environmental parameters include altitude, weather conditions, ambient temperature, snow depth, energy prices, ore grade, and market prices.

[0119] 2. Output labels (target data):

[0120] Ore output: The output from ore crushing and transportation;

[0121] Energy consumption per ton of ore: the energy consumed per unit amount of ore;

[0122] Transportation costs: including logistics, equipment maintenance and other costs.

[0123] These data are used as the training set for the DBO-DNN transportation system performance prediction model, serving as both input and output. During training, the input features are propagated forward through each layer of the neural network, ultimately outputting the predicted process performance indicators. The DBO-DNN model minimizes prediction error by iteratively adjusting weights and biases.

[0124] Step 3: After training, the DBO-DNN transportation system performance prediction model can predict the corresponding process performance indicators based on the new input features and control the prediction error within 3.2%, thereby providing accurate support for the optimization and decision-making of mine transportation processes.

[0125] 3. The specific steps for optimizing mine transportation processes based on three-stage collaborative transportation and intelligent optimization are as follows:

[0126] Step 1: Joint Modeling of Production Scheduling and Crushing Station Relocation: Construct a multi-objective mixed integer programming model with the objective function of maximizing total profit, and set the following constraints: Equipment operating capacity: mining volume constraints of the mining strip, processing capacity constraints of the crushing station; crushing station relocation constraints (if relocated, raw materials cannot be processed during this period).

[0127] Step 2: Solving with an Improved Genetic Algorithm: An improved genetic algorithm (GA) is used to solve the problem. Chromosomes are divided into two gene segments: transport and relocation decisions. Specific optimization strategies include: initialization, fitness evaluation, improved roulette wheel selection, crossover and mutation operations, and dynamic adjustment optimization. Population updates include elite retention and competitive replacement while maintaining diversity. The termination condition is a maximum of 200 iterations or a fitness improvement of less than 0.1% for 20 consecutive generations. Output the optimal solution, load distribution graph, and convergence curve.

[0128] The following is a specific embodiment. The transportation process structure of the present invention is shown in Figure 1. It specifically includes two steps: establishing a three-stage collaborative transportation process architecture of "short-distance transshipment - mobile crushing - steep-angle lifting"; and establishing an intelligent decision-making system.

[0129] Step 1: Establish a three-stage collaborative transportation process architecture for ore flow: "short-distance transshipment - mobile crushing - steep-angle conveying," specifically including:

[0130] (1) Using existing borehole data, a precise three-dimensional geological model is constructed using 3Dmine industrial software, thereby effectively delineating the boundaries of mining units. As shown in Figure 2, the ore body is divided into progressively smaller regional sections according to the following criteria: section (≤250,000m²) → block (≤40,000m²) → strip (≤2,500m²), forming mining units with modular characteristics.

[0131] (2) Short-distance transfer module: As shown in Figure 3(a), after the electric shovel loads the ore within the strip, the electric mining truck transfers the ore to the semi-mobile crushing station along a preset path. The transportation data (load, speed, energy consumption) is uploaded to the central dispatch system in real time. Electric mining truck configuration: load capacity 60-100t, range ≥8h, transportation radius ≤1.5km. It avoids congested sections and shortens the empty journey through the real-time path planning system.

[0132] (3) Semi-mobile crushing module: As shown in Figure 3(b), the ore is crushed to a particle size of ≤300mm by a jaw crusher. The crushing energy consumption is reduced by 15%-20% through frequency conversion drive. The crushing efficiency is monitored in real time and fed back to the optimization model. Modular crushing station: with a processing capacity of 2000-3000t / h, it consists of modules such as crusher, screening machine, and feeder. It adopts hydraulic positioning and quick connection technology, and the relocation time is ≤72h.

[0133] (4) Inclined Angle Lifting Module: As shown in Figure 3(c), the crushed ore is directly lifted to the surface processing station by the conveyor, reducing the energy consumption of traditional trucks climbing slopes and improving transportation efficiency by more than 40%. High-strength belt conveyor: Inclined angle 35°-55°, belt speed 2.5-4m / s, with corrugated sidewalls and cross diaphragms to prevent material slippage; the drive motor is equipped with a frequency converter to dynamically adjust the power according to the load and reduce no-load energy consumption.

[0134] Step 2, establish a performance prediction model for the staged mobile crushing and continuous transportation process, specifically including:

[0135] (1) A deep neural network (DNN) was constructed using the Torch framework in Python. The network structure is shown in Figure 4. The network consists of five layers: one input layer, three hidden layers, and one output layer. The input layer contains 16 neurons, corresponding to 16 input features, including start and end times of operations, number of trucks, load capacity, travel distance, crushing volume, truck tare weight, conveyor layout length, conveyor belt speed, altitude, weather conditions, ambient temperature, snow thickness, energy price, ore grade, and market price. The output layer contains 3 neurons, corresponding to 3 predicted outputs, including ore production, energy consumption per ton of ore, and total transportation cost. The number of neurons in the hidden layer will be optimized using a genetic algorithm to ensure optimal network performance.

[0136] (2) The DBO was used to optimize the hyperparameters of the established DNN, including the learning rate, training cycle, activation function, regularization parameter, and network structure. The population size of the DBO was set to 100, with each dung beetle (individual) representing a combination of hyperparameters. For each individual, the DNN model was trained using its hyperparameter combination, and its performance metric (the coefficient of determination (R²) on the validation set) was calculated. During model training, the fitness of the model was evaluated based on the different performances of these hyperparameter combinations. The dung beetle optimization algorithm searches for the optimal solution by simulating the four core behaviors of dung beetles (rolling, reproduction, foraging, and stealing). The specific steps for each behavior are as follows:

[0137] Rolling: Dung beetles update their positions based on their own fitness and differences. Rolling behavior can be understood as dung beetles exploring the surrounding solution space with a certain probability.

[0138] Reproduction: Dung beetles pass on the hyperparameters of superior individuals to the next generation through reproduction. Superior individuals are given priority in generating new populations.

[0139] Foraging: Dung beetles assess fitness based on their current location (i.e., the current combination of hyperparameters) and select the best individuals. This behavior is the main driving force of the algorithm; dung beetles "search" for the optimal combination of hyperparameters.

[0140] Stealing: Dung beetles also optimize their performance by interacting with other individuals (exchanging some hyperparameters), which can help the algorithm speed up the search for the optimal solution.

[0141] Based on the fitness of individual dung beetles, the best individuals are selected and preserved to replenish the population. Individuals with lower fitness are eliminated, while those with higher fitness continue to reproduce.

[0142] (3) DBO-DNN continuously iterates and updates until it reaches the maximum number of iterations (e.g., 30 times) or the fitness changes very little over several generations. That is, when no better solution is found, the algorithm stops and outputs the optimal combination of hyperparameters and its corresponding performance indicators, thereby accurately predicting transportation process performance indicators such as ore output, energy consumption per ton of ore and total transportation cost.

[0143] (4) Establish an interface with the real-time monitoring system and use the output of the prediction model with the production scheduling plan and the optimal relocation plan of the crushing station.

[0144] Step 3: Establish a performance optimization model for the staged mobile crushing and continuous transportation process, as shown in Figure 5. The specific steps are as follows:

[0145] (1) Taking the maximization of total profit as the objective function, and with constraints such as mining volume limit, transportation capacity limit, crusher load limit, annual mining rate constraint, maximum beneficiation capacity limit, mining sequence constraint, and crushing station relocation constraint, the optimal production scheduling plan and the optimal relocation plan of the mobile crushing continuous transportation process are expressed as integer programming. Based on the transportation system performance prediction model, a mathematical model that can simultaneously optimize production scheduling and crushing station relocation is established:

[0146] Objective function:

[0147] In the formula, T represents the life cycle of the mine, N represents the total number of clusters, and H represents the total number of levels.

[0148] V nt This represents the total value of clustering n within time period t: ;

[0149] In the formula, S n V represents the total number of blocks in cluster n; Bst This represents the total value of block s over time period t: ;

[0150] In the formula, v Bs Let Gr be the volume of block s. Bs This indicates the grade of block s. Let represent the density of block s, φ represent the depletion rate, η represent the recovery rate, UP represent the price per unit grade, and r represent the discount rate.

[0151] : Indicates whether cluster n is mined during time period t, 1 indicates mining, 0 indicates no mining.

[0152] The cost of transporting cluster n from the crushing station at layer h in time period t:

[0153] : Indicates whether the conveyor used by cluster n in time period t is in layer h, 1 indicates yes, 0 indicates no.

[0154] C Rh Indicates the relocation cost at level h:

[0155] C R Based on the base cost, α is the unit cost coefficient that increases with the number of levels. As h increases, the relocation cost will gradually increase.

[0156] : Indicates whether the crushing station has been relocated to layer h during time period t, 1 indicates relocation, 0 indicates no relocation.

[0157] λ represents the carbon emission control weight, and e represents the carbon emissions generated per unit of ore transport using the SMIPCC system.

[0158] The total mass of cluster n is represented by: ,in, S represents the mass of block s. n This represents the number of blocks in cluster n.

[0159] Constraints:

[0160] 1. Mining volume limit: The total mining volume in each time period t cannot exceed the maximum mining capacity M. t :

[0161]

[0162] 2. Transportation capacity limit: The total mining volume in each time period t cannot exceed the truck transportation capacity TS. t :

[0163]

[0164] 3. Crusher load limit: The total mining volume in each time period t cannot exceed the crusher load L. t :

[0165]

[0166] 4. Annual mining rate constraint: The annual mining volume shall not exceed β times the mining volume of the previous year to avoid excessively rapid mining.

[0167]

[0168] In the formula, β is a coefficient that controls the increase in mining output (usually set to 110%).

[0169] 5. Maximum processing capacity limit for mineral processing: The total amount of ore mined in each time period t cannot exceed the maximum processing capacity D of the mineral processing plant. t :

[0170]

[0171] In the formula, The average ore grade for cluster n:

[0172] In the formula, Let S be the grade of block s. n Let n be the number of blocks in cluster n.

[0173] 6. Each cluster can only be mined once:

[0174] 7. Mining order constraint: Ensure that cluster n precedes cluster n. pr Cluster n can only begin mining after it has been completely mined.

[0175]

[0176] In the formula, PR n Let n be the set of previous clusters of cluster n.

[0177]

[0178] HP n For the level of cluster n, the previous cluster set, VP n Let n be the set of previous clusters that are perpendicular to cluster n.

[0179] 8. Crusher Hierarchical Deployment Constraints: A crusher can only be deployed at one level in each time period t.

[0180]

[0181] 9. Crushing station relocation constraint: qht is only 1 when the crushing station level changes.

[0182]

[0183] 10. Relocation sequence constraint: The crushing station can only be relocated from the upper level to the lower level.

[0184]

[0185] 11. The maximum time span T for the relocation of the crushing station max To avoid multiple relocations in a short period of time.

[0186]

[0187] 12. Each crushing station can only be relocated once within a given time period:

[0188]

[0189] (2) Solution method for integer programming model based on genetic algorithm:

[0190] 1) Chromosome coding design:

[0191] The genetic algorithm uses a segmented coding structure for its chromosomes, characterized by including the following three gene segments: a transportation decision gene segment, a crushing station location decision gene segment, and a relocation decision gene segment.

[0192] Chromosome length: The length of the chromosome is N×T+N×H×T+H×T, where the first N×T positions are the transportation decision segment, the middle N×H×T positions are the crushing station location decision segment, and the last H×T positions are the relocation decision segment.

[0193] 2) Population initialization:

[0194] The initial population is generated using the following steps:

[0195] Set the population size K=500;

[0196] For each individual's chromosome, gene initialization is performed according to the following rules:

[0197] Each bit of the transportation decision gene segment is sampled independently, where the probability of ont=1 is positively correlated with the ore reserves of cluster n in time period t, avoiding excessive concentration in certain time periods, and the activation probability is adjusted through standardization.

[0198] For the decision gene segment of the crushing station, randomly select a valid level h such that pnht=1, and make it satisfy the following formula: .

[0199] Randomly generate the relocation decision gene segment qnht=1.

[0200] For each individual, the following checks are performed, and if they are not met, the individual is repaired or removed: annual mining volume constraints; cluster mining uniqueness and sequence constraints; unique deployment constraints at the crushing station level; and relocation direction and time interval restrictions.

[0201] 3) Fitness function construction:

[0202] Fitness calculation includes objective function evaluation and constraint penalty terms, and its mathematical expression is:

[0203]

[0204] Penalty item: Punishment for violations of constraints:

[0205] in, Let be the penalty weight for the i-th constraint. (violation) i The degree of violation of the i-th constraint.

[0206] 4) Select Operation:

[0207] An improved roulette wheel selection strategy is adopted:

[0208] a) Calculate the normalized fitness value of the population: , where: f k f is the initial fitness value of individual k. min It is the minimum fitness value in the population. max It is the maximum fitness value in the population. The denominator should be a very small positive number to avoid being 0.

[0209] b) Introduce a merit retention mechanism to directly replicate the top 5% of fit individuals to the next generation;

[0210] c) The remaining 95% of individuals use roulette to determine the probability: Selection is conducted to ensure that high-quality individuals have a higher probability of reproduction.

[0211] 5) Crossover operation: Crossover is performed independently on each gene segment. Crossover methods: Transport decision segment (N×T): two-point crossover, retaining continuous intervals and enhancing local structural stability; Crushing station location decision segment (N×H×T): uniform crossover, each position independently inherits 50% probability from the parent generation; Crushing station relocation decision segment (H×T): single-point crossover, only one relocation change is allowed to ensure compliance with constraints.

[0212] 6) Mutation operation: Implement a hierarchical mutation mechanism:

[0213] a) Baseline mutations: Each gene locus has a probability p m =0.01 is flipped (that is, its value is changed from 0 to 1 or from 1 to 0);

[0214] b) Taboo protection: Mutation of key gene segments is prohibited on the top 10% of individuals in terms of fitness;

[0215] c) Anomaly Repair: If a mutation causes a violation of a hard constraint, dynamic correction to the valid solution space is performed. The repair strategy is based on heuristics or rules.

[0216] 7) Population renewal:

[0217] a) Elite Preservation: Directly preserve the top 5% of individuals in terms of fitness in the current generation;

[0218] b) Competitive replacement: The newly generated individuals are merged with the remaining individuals in the original population, and the top K individuals are retained according to their fitness.

[0219] c) Diversity maintenance: Force the retention of at least 3% of individuals with low fitness but high genetic diversity.

[0220] 8) Termination determination and output:

[0221] The algorithm termination condition is set as follows:

[0222] a) The maximum number of iterations reaches G=200;

[0223] b) The improvement in optimal fitness over 20 consecutive generations is less than ε = 0.1%;

[0224] c) Obtain a feasible solution that satisfies all hard constraints;

[0225] The operation terminates when any condition is met, and the output is:

[0226] The optimal chromosome corresponds to the transportation scheduling scheme and relocation plan;

[0227] Equipment load distribution diagram for different time periods;

[0228] Convergence curve of the objective function.

Claims

1. A mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization, characterized in that, This includes a three-stage collaborative transportation process architecture and an intelligent decision-making system. The three-stage collaborative transportation process architecture provides a continuous process for ore transportation and crushing, while the intelligent decision-making system provides decision support for the three-stage collaborative transportation process architecture by predicting and optimizing production scheduling and crushing station relocation. The three-stage collaborative transportation process architecture, based on the geometry of the ore body, geological characteristics, mining technology, equipment selection, and transportation conditions, divides the mining area into a hierarchical structure from top to bottom: "section-block-strip," and constructs a continuous transportation system consisting of a short-distance transfer module, a mobile crushing module, and a steep-angle hoisting module. The transfer module is equipped with electric mining trucks for ore collection and transportation; the mobile crushing module uses a semi-mobile crushing station for in-situ crushing operations; and the steep-angle lifting module uses a belt conveyor to complete steep-angle ore transportation. The intelligent decision-making system constructs a DBO-DNN transportation system performance prediction model based on a dung beetle optimization algorithm-improved deep neural network (DNN), modeling production scheduling and crushing station relocation as a mixed-integer programming problem. By using the output of the DBO-DNN transportation system performance prediction model as data input and combining it with an improved genetic algorithm for solution, the transportation system is optimized. The mathematical model establishment includes: establishing a multi-objective optimization function under three-dimensional spatiotemporal constraints with the goal of maximizing total profit: (1) With the goal of maximizing total profit, and with constraints such as mining volume limit, transportation capacity limit, crusher load limit, annual mining rate constraint, maximum processing capacity limit of mineral processing, mining sequence constraint, and crushing station relocation constraint, the optimal production scheduling plan of mobile crushing continuous transportation process and the optimal relocation plan of crushing station are expressed as integer programming. Based on the DBO-DNN transportation system performance prediction model, a mathematical model that can simultaneously perform production scheduling and crushing station relocation optimization is established: Objective function: In the formula, Z represents the total value, T represents the lifespan of the mine, N represents the total number of strips, and H represents the total number of layers; V nt This represents the total value of strip n within time period t: In the formula, S n V represents the total number of blocks in strip n; Bst This represents the total value of block s over time period t: In the formula, v Bs Let Gr be the volume of block s. Bs This indicates the grade of block s. This represents the density of block s. Let η represent the depletion rate, η represent the recovery rate, UP represent the unit price, and r represent the discount rate. : Indicates whether strip n is mined during time period t, 1 indicates mining, 0 indicates no mining; The transportation cost of strip n from the crushing station at layer h during time period t: ； : Indicates whether the conveyor used by strip n in time period t is on layer h, 1 indicates yes, 0 indicates no; C Rh Indicates the relocation cost at level h: C R Based on the base cost, α is the unit cost coefficient that increases with the number of levels. As h increases, the relocation cost will gradually increase. : Indicates whether the crushing station is relocated to layer h during time period t, 1 indicates relocation, 0 indicates no relocation; λ represents the carbon emission control weight, and e represents the carbon emissions generated per unit of ore transportation using the SMIPCC system; The total mass of strip n is represented by: ,in, The mass of block s is represented; three types of constraints are set: strip mining volume constraint, crushing station processing capacity constraint, and crushing station relocation constraint; an improved genetic algorithm is used to solve the problem, with chromosomes divided into two gene segments: transportation and relocation decision; the optimization is dynamically adjusted through initialization, fitness evaluation, improved roulette wheel selection, crossover, and mutation operations; population updates include elite retention and competitive replacement; the termination condition is a maximum number of iterations of 200 generations or a fitness improvement of less than 0.1% for 20 consecutive generations, and the optimal solution, load distribution map, and convergence curve are output.

2. The mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization according to claim 1, characterized in that, The three-stage collaborative transportation process architecture specifically includes: a) dividing the mining area into three levels of mining units: section, block, and strip; b) using electric shovels to carry out ore mining operations in the strip units; c) transporting the ore short-distance to the semi-mobile crushing station using ore trucks; d) vertically lifting and transporting the crushed ore using a steep-angle belt conveyor; the steep-angle belt conveyor has an inclination range of 35°-55° and a lifting speed of 2.5-4 m / s.

3. The mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization according to claim 1, characterized in that, The method for constructing the DBO-DNN transportation system performance prediction model includes: using the dung beetle optimization algorithm to globally optimize the initial hyperparameters of the deep neural network, establishing a three-layer network structure with an adaptive activation function; input feature parameters include 18 indicators in three categories: mining operation parameters, equipment status parameters, and environmental parameters; output parameters are transportation process performance indicators, including three core indicators: ore output, energy consumption per ton of ore, and transportation cost, with a determination coefficient R0. 2 ≤3.2%.

4. The mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization according to claim 3, characterized in that, The input feature parameters specifically include: mining operation parameters: start and end time of operation, number of trucks, carrying capacity, running distance, and crushing volume; equipment status parameters: truck weight, conveyor layout length, and conveyor belt speed; and environmental parameters: altitude, weather conditions, ambient temperature, snow thickness, energy price, ore grade, and market price.

5. The mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization according to claim 1, characterized in that, The constraints include: extraction volume limit: the total extraction volume in each time period t cannot exceed the maximum extraction capacity M. t : Transportation capacity limit: In each time period t, the total mining volume cannot exceed the truck transportation capacity TS. t : Crusher load limit: The total mining volume in each time period t cannot exceed the crusher load L. t : Annual mining rate constraint: The annual mining volume shall not exceed β times the mining volume of the previous year to avoid excessively rapid mining. In the formula, β is a coefficient controlling the increase in mining volume; the maximum processing capacity limit for ore beneficiation is: the total amount of ore mined in each time period t cannot exceed the maximum processing capacity D. t : In the formula, The average ore grade of band n: In the formula, Let S be the grade of block s. n Let n be the number of blocks in strip n; each strip can only be mined once. Mining sequence constraint: Ensure that strip n precedes strip n. pr Strip n can only begin mining after it has been completely mined out. In the formula, PR n Let n be the set of previous stripes; HP n For the set of previous horizontal stripes of stripe n, VP n Let n be the set of previous stripes perpendicular to stripe n; Crusher hierarchy deployment constraint: a crusher can only be deployed in one hierarchy at any given time interval t. Crushing station relocation constraint: qht is only 1 when the crushing station level changes. Relocation sequence constraint: The crushing station can only be relocated from the upper level to the lower level. The maximum time span T for the relocation of the crushing plant max To avoid multiple relocations in a short period of time; Each crushing station can only be relocated once within a certain time period: 。 6. The mine transportation process optimization method based on three-stage collaborative transportation and intelligent optimization according to claim 1, characterized in that, The improved genetic algorithm solves the problem as follows: 1) Chromosome encoding design: The chromosome of the genetic algorithm adopts a segmented encoding structure, containing the following three gene segments: transportation decision gene segment, crushing station location decision gene segment, and relocation decision gene segment; Chromosome length: The length of the chromosome is: N×T+N×H×T+H×T, where the first N×T positions are the transportation decision segment, the middle N×H×T positions are the crushing station location decision gene segment, and the last H×T positions are the relocation decision segment; 2) Population initialization: Generate the initial population: Set the population size; For each individual chromosome, perform gene initialization according to the following rules: Each position of the transportation decision gene segment is sampled independently, where o nt The probability of p=1 is positively correlated with the ore reserves of strip n in time period t. To avoid excessive concentration in certain time periods, the activation probability is adjusted through standardization. For the decision gene segment of the crushing station, a legal level h is randomly selected such that p nht =1, and make it satisfy the following formula: ； Randomly generate relocation decision gene segment q nht =1; Perform the following checks on each individual, and repair or remove if they are not satisfied: annual mining volume constraint; uniqueness and sequence constraint of strip mining; unique deployment constraint of crushing station level; relocation direction and time interval restriction; 3) Construct fitness function: fitness calculation includes objective function evaluation and constraint penalty term, and its mathematical expression is: Penalty item: Punishment for violations of constraints: ;in, The penalty weight for the i-th constraint; violate i 4) Selection operation: An improved roulette wheel selection strategy is adopted: a) Calculate the normalized fitness value of the population: , where: f k f is the initial fitness value of individual k; min It is the minimum fitness value in the population; f max It is the maximum fitness value in the population. a) Use extremely small positive numbers to avoid denominators of 0; b) Introduce a quality retention mechanism, directly replicating the top 5% of individuals by fitness to the next generation; c) Use a roulette wheel to determine the probability of the remaining 95% of individuals. 5) Crossover operation: Crossover is performed independently on each gene segment. Crossover methods: Transport decision segment (N×T): two-point crossover, retaining continuous intervals and enhancing local structural stability; Crushing station location decision segment (N×H×T): uniform crossover, each position independently inherits 50% probability from the parent; Crushing station relocation decision segment (H×T): single-point crossover, only one relocation change is allowed to ensure compliance with constraints; 6) Mutation operation: Implement a hierarchical mutation mechanism: a) Basic position mutation: Each gene position is mutated with probability p m =0.01 for flipping; b) Tabu protection: prohibit mutation operations on key gene segments of individuals with the top 10% fitness; c) Anomaly repair: if mutation causes a violation of hard constraints, dynamically correct to the legal solution space repair strategy based on heuristics or rules; 7) Population update: a) Elite retention: directly retain individuals with the top 5% fitness of the current generation; b) Competitive replacement: merge newly generated individuals with the remaining individuals of the original population, and retain the top K individuals according to fitness; c) Diversity maintenance: forcibly retain at least 3% of individuals with low fitness but high genetic diversity; 8) Termination judgment and output: the algorithm termination conditions are set as follows: a) the maximum number of iterations reaches G=200; b) the improvement of the optimal fitness for 20 consecutive generations is less than ε=0.1%; c) obtain a feasible solution that satisfies all hard constraints; the operation terminates when any condition is met, and the output is: the transportation scheduling scheme and relocation plan corresponding to the optimal chromosome; the equipment load distribution map of each time period; the convergence curve of the objective function.

Images