Multi-target intelligent optimization method and system for blasting parameters of strip mine in high-altitude cold region

Abstract

The invention discloses a multi-target intelligent optimization method and system for blasting parameters of a strip mine in a high-altitude cold region. The method comprises the following steps: carrying out data acquisition to obtain a parameter data set; performing data preprocessing on the parameter data set to obtain a feature sample set; constructing an initial blasting parameter model based on a machine learning algorithm, and performing hyper-parameter optimization on the model to obtain a blasting parameter model; a multi-objective optimization function is constructed: based on the multi-objective optimization function and the blasting parameter model, solving is carried out in combination with environmental condition constraints, and a pareto optimal solution set is obtained; according to the pareto optimal solution set, a representative solution is selected, a visual scheme is generated, and blasting parameter optimization of the strip mine in the high-altitude cold region is completed. According to the method, temperature, oxygen and frozen soil constraint conditions of the high-cold and high-altitude environment are introduced, blasting safety, lumpiness uniformity and the explosive utilization rate are considered at the same time through multi-target collaborative optimization, the method can adapt to the extreme environment, meanwhile, the one-sidedness of single-target optimization is avoided, and the intelligent level of blasting design and implementation is greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent mine blasting optimization technology, specifically relating to a multi-objective intelligent optimization method and system for blasting parameters in open-pit mines in high-altitude cold regions. Background Technology

[0002] Blasting technology is a core component of rock stripping and crushing in open-pit mines, directly impacting the mine's safety, economy, and production efficiency. Traditional blasting design methods typically rely on empirical formulas and experimental verification, making it difficult to simultaneously consider multiple target parameters such as blasting vibration velocity, block size distribution, and explosive consumption.

[0003] In high-altitude and frigid regions, environmental factors introduce significant uncertainties into blasting effectiveness: low temperatures and low oxygen levels reduce explosive reaction efficiency, leading to unstable energy release; the presence of permafrost and joints alters energy propagation paths, resulting in difficulties in controlling blasting vibration velocity, increased block size dispersion, and higher explosive consumption per unit area. These issues make it difficult for traditional blasting methods to guarantee operational safety and economic benefits in high-altitude open-pit mines.

[0004] In recent years, artificial intelligence and optimization algorithms have been gradually applied to the engineering field, achieving some progress in blasting parameter prediction and scheme optimization. However, existing research still has the following shortcomings: First, it fails to explicitly introduce high-altitude and cold-climate environmental parameters as constraints, resulting in a lack of applicability of optimization results; second, most studies focus on single-objective optimization, lacking a systematic framework for multi-objective collaborative optimization; and third, it lacks a closed-loop mechanism of monitoring-feedback-relearning, making it impossible to achieve adaptive model updates.

[0005] Therefore, there is an urgent need to propose an intelligent optimization method and system for blasting that can integrate artificial intelligence prediction and multi-objective optimization, and combine the characteristics of high-altitude and cold environments with feedback learning mechanisms, so as to achieve a balance between safety, economy and efficiency in open-pit mine blasting design. Summary of the Invention

[0006] To address the shortcomings of existing technologies, one of the objectives of this invention is to provide a multi-objective intelligent optimization method for blasting parameters in open-pit mines in high-altitude cold regions. This method involves: introducing environmental constraints and corrections (temperature, altitude / oxygen, permafrost, etc.); constructing a multi-objective optimization model with minimum vibration velocity, maximum block size uniformity, and minimum unit consumption as its core objectives; employing a combined strategy of machine learning prediction and multi-objective evolutionary algorithms; and establishing a closed loop of monitoring, error determination, and retraining to achieve dynamic optimization and engineering feasibility of the blasting scheme.

[0007] The second objective of this invention is to provide a system for implementing the multi-objective intelligent optimization method for blasting parameters in open-pit mines in high-altitude cold regions.

[0008] This invention provides a multi-objective intelligent optimization method for blasting parameters in open-pit mines in high-altitude cold regions, comprising the following steps:

[0009] S1. Collect data on the blasting environment and the preset blasting target to obtain a parameter dataset;

[0010] S2. Perform data preprocessing on the parameter dataset to obtain the feature sample set;

[0011] S3. Construct an initial blasting parameter model based on machine learning algorithms, and then perform hyperparameter optimization on the model to obtain the blasting parameter model;

[0012] S4. Construct a multi-objective optimization function with the optimization objectives of minimizing blasting vibration velocity, maximizing blasting block uniformity, and minimizing explosive consumption:

[0013] S5. Based on the multi-objective optimization function and the blasting parameter model, combined with environmental constraints, the Pareto optimal solution set is obtained.

[0014] S6. Based on the Pareto optimal solution set, select representative solutions, generate a visualization scheme, and complete the optimization of blasting parameters in open-pit mines in high-altitude cold regions.

[0015] In step S1, the parameter dataset includes environmental parameters, rock mass parameters, design parameters, and target parameters;

[0016] The environmental parameters include temperature T, altitude H, oxygen volume fraction O, permafrost thickness Ft, wind speed Ws, and humidity Rh.

[0017] The rock mass parameters include uniaxial compressive strength. Elastic modulus E, Poisson's ratio v, joint and fracture density Jv, lithology;

[0018] The design parameters include aperture d, aperture depth L, aperture spacing a, row spacing b, and charge structure / cartridge spacing. Delayed sequence ;

[0019] The target parameters include vibration velocity V and post-explosion particle size distribution. Explosive consumption per unit (W).

[0020] Step S2 specifically involves standardizing the parameter dataset, handling outliers, and eliminating correlation and recursive features to obtain a feature sample set X.

[0021] In step S3, the initial blasting parameter model makes predictions by establishing a nonlinear mapping relationship between blasting parameters and blasting effects; the model is trained based on the feature sample set X, with blasting block size distribution, blasting vibration velocity, and explosive unit consumption as output targets; the machine learning algorithms include XGBoost, random forest, support vector machine, LightGBM, and GBDT.

[0022] The hyperparameter optimization employs grid search or Bayesian optimization; based on coordinated five-fold cross-validation, the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination are used. A weighted fusion approach was used to construct a comprehensive performance evaluation index to evaluate the model's performance. The model with the best comprehensive performance was selected as the final model. The weights of each evaluation index were set according to the on-site blasting accuracy requirements and engineering safety needs. Furthermore, when... When the value is less than 0.9, the model is retrained.

[0023] In step S4, the multi-objective optimization function F is expressed using the following formula: ;in, This is the weighting coefficient for the unit consumption of explosives. This is a weighting coefficient for block size uniformity. is the weighting coefficient for vibration velocity, and V represents the blasting vibration velocity; U represents the block size uniformity index; W represents the explosive consumption per unit volume.

[0024] The blasting vibration velocity V is expressed by the following formula: ;in, The radial component is obtained from the triaxial velocity sensor; The tangential component is obtained from the triaxial velocity sensor; The vertical component is obtained from the triaxial velocity sensor;

[0025] The block size uniformity index is expressed using the following formula: ;in, The coefficient of variation is used to calculate the post-explosion particle size distribution. The mean / variance is obtained; the post-explosion particle size distribution is expressed using the following formula: ; This represents the post-explosion particle size distribution, indicating particle sizes smaller than [value missing]. The cumulative percentage Characteristic particle size, The uniformity coefficient;

[0026] The specific consumption W of the explosive is expressed by the following formula: ;in, The volume of the explosion; This refers to the amount of explosives loaded in a single segment.

[0027] Step S5 includes the following steps:

[0028] Encoding decision variables: setting hole spacing Row spacing b, hole depth L, and cartridge spacing And the delay δ is the decision variable, forming the decision vector. And set its value range and constraints;

[0029] Fitness assessment: For each individual z in the population, the trained detonation parameter model is called. Obtain the predicted value ,in For blasting vibration, For uniformity of block size, The objective functions are minimizing explosive consumption, maximizing the uniformity of blast fragment size, and minimizing blast vibration.

[0030] Imposing high-altitude and cold-weather conditions and safety constraints: Sadovsky-like empirical formulas for environmental modification. Where K is the first empirical constant; It is the second empirical constant; The distance decay exponent; This represents the maximum safe threshold for blasting vibration velocity. This refers to the amount of explosives loaded in a single stage. The distance between the centers of the explosion; For environment correction functions,

[0031] Non-dominated ranking: Based on Pareto dominance, the population is divided into different non-dominated levels. ;

[0032] Selection, crossover, and mutation: A tournament selection strategy is used to select parent individuals from the population; then, offspring individuals are generated through simulated binary crossover (SBX) and polynomial mutation; the crossover probability is set to 0.9, and the mutation probability is set to 0.1.

[0033] Convergence criterion and output: When the maximum number of iterations is reached or the Pareto front tends to stabilize, the iteration stops and the final Pareto optimal solution set is output.

[0034] The Pareto optimal solution set contains several blasting schemes that achieve different balances between safety, block uniformity, and economy.

[0035] In step S6, a representative solution is selected using the weighted sum method, the ideal point method, or the entropy weight method.

[0036] The multi-objective intelligent optimization method for blasting parameters in high-altitude cold-region open-pit mines also includes a feedback learning mechanism. Specifically, it involves: acquiring PPV using a three-dimensional geophone, calculating the post-blast particle size distribution curve using a camera combined with image recognition technology, calculating unit consumption using charge metering and volume, and finally obtaining the measured blasting results, which are then compared with the predicted results output by the model. or relative error , To preset the error threshold, incremental retraining and hyperparameter fine-tuning are triggered, and the feature interval library and constraint library are updated to achieve closed loop.

[0037] The feedback learning mechanism adopts an incremental learning approach, where each new blasting data is used as a sample to update the database.

[0038] The present invention also provides a system for implementing the multi-objective intelligent optimization method for blasting parameters in open-pit mines in high-altitude cold regions, including a data import module, a feature selection and target setting module, a machine learning model training module, a multi-objective optimization module, an optimal solution export module, and a blasting control module;

[0039] The data import module is used to import environmental data from high-altitude and cold open-pit mines (including temperature, altitude, oxygen content, frozen soil thickness, wind speed, and humidity), rock mechanics parameters (uniaxial compressive strength, elastic modulus, Poisson's ratio, and joint and fracture density), blasting design parameters (borehole diameter, hole depth, hole spacing, row spacing, charge structure, and detonation sequence), and blasting monitoring data (blasting vibration velocity, post-blast block size distribution, throwing range, and flyrock distance). The module also standardizes the data, fills in missing values, and removes outliers to form a standardized input sample set.

[0040] The feature selection and target setting module is used to perform correlation analysis, principal component analysis or recursive feature elimination on multi-source data in the sample set, screen key influencing factors, and select multiple blasting effect targets that need to be optimized collaboratively in this case.

[0041] The machine learning model training module is used to build and train prediction models based on XGBoost, random forest, support vector machine, LightGBM or GBDT. Specifically, it includes dataset partitioning, hyperparameter initialization, cross-validation and performance evaluation. The model training results output a prediction blasting effect index, which is used to establish a nonlinear prediction relationship between blasting parameters and blasting effect.

[0042] The multi-objective optimization module sets optimization objectives such as minimizing blasting vibration velocity, maximizing blasting block uniformity, and minimizing explosive consumption. At the same time, it introduces environmental constraints and collaborative optimization intervals for different parameters. It uses the non-dominated sorting genetic algorithm NSGA-II or other multi-objective evolutionary algorithms for optimization. Through non-dominated sorting, crowding calculation and elite retention mechanisms, it obtains the Pareto optimal solution set.

[0043] The optimal solution derivation module is used to select representative solutions from the Pareto optimal solution set, select the optimal combination of blasting parameters through weighted method, ideal point method or entropy weight method, and generate hole mesh parameter table, charge list and detonation network diagram. At the same time, it provides visualization results such as block size distribution curve, vibration prediction curve and three-dimensional stacking morphology to assist designers in scheme evaluation.

[0044] The blasting control module is used to convert the optimal combination of blasting parameters into instructions to be executed and send them to the unmanned drilling rig, automatic charging vehicle and intelligent detonator controller. Through remote communication, it realizes the execution of borehole layout, charge amount control and detonation sequence, and forms a closed-loop control by combining monitoring results, thereby improving the automation level and execution accuracy of blasting design.

[0045] This invention discloses a multi-objective intelligent optimization method and system for blasting parameters in open-pit mines in high-altitude and cold regions. It introduces the constraints of temperature, oxygen, and permafrost in the high-altitude and cold environment, and simultaneously considers blasting safety, block size uniformity, and explosive utilization rate through multi-objective collaborative optimization. This method can not only adapt to extreme environments, but also avoid the one-sidedness of single-objective optimization, and greatly improve the level of intelligence in blasting design and implementation. Attached Figure Description

[0046] Figure 1 This is a schematic flowchart of the method of the present invention;

[0047] Figure 2 This is a schematic diagram of the system structure of the present invention;

[0048] Figure 3 This is a diagram illustrating the actual application of the system of the present invention in the embodiments. Detailed Implementation

[0049] This invention provides a multi-objective intelligent optimization method for blasting parameters in open-pit mines in high-altitude cold regions, the flowchart of which is shown below. Figure 1 As shown, it includes the following steps:

[0050] S1. Collect data on the blasting environment and the preset blasting target to obtain a parameter dataset;

[0051] In step S1, the parameter dataset includes environmental parameters, rock mass parameters, design parameters, and target parameters;

[0052] The environmental parameters include temperature T, altitude H, oxygen volume fraction O, permafrost thickness Ft, wind speed Ws, and humidity Rh.

[0053] The rock mass parameters include uniaxial compressive strength. Elastic modulus E, Poisson's ratio v, joint and fracture density Jv, lithology;

[0054] The design parameters include aperture d, aperture depth L, aperture spacing a, row spacing b, and charge structure / cartridge spacing. Delayed sequence ;

[0055] The target parameters include vibration velocity V and post-explosion particle size distribution. Explosive consumption per unit (W).

[0056] S2. Perform data preprocessing on the parameter dataset to obtain the feature sample set;

[0057] Step S2 specifically involves standardizing the parameter dataset, handling outliers, and eliminating correlation and recursive features to obtain a feature sample set X.

[0058] S3. Construct an initial blasting parameter model based on machine learning algorithms, and then perform hyperparameter optimization on the model to obtain the blasting parameter model;

[0059] In step S3, the initial blasting parameter model predicts by establishing a nonlinear mapping relationship between blasting parameters and blasting effects; the model is trained with blasting block size distribution, blasting vibration velocity, and explosive consumption as output targets; the machine learning algorithms include XGBoost, random forest, support vector machine, LightGBM, and GBDT.

[0060] The hyperparameter optimization employs grid search or Bayesian optimization; based on coordinated five-fold cross-validation, the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination are used. The model performance was evaluated, and a weighted fusion method was used to construct a comprehensive performance evaluation index. The model with the best comprehensive performance was selected as the final model. The weights of each evaluation index were set according to the on-site blasting accuracy requirements and engineering safety needs. Furthermore, when... When the value is less than 0.9, the model is retrained.

[0061] S4. Construct a multi-objective optimization function with the optimization objectives of minimizing blasting vibration velocity, maximizing blasting block uniformity, and minimizing explosive consumption:

[0062] In step S4, the multi-objective optimization function F is expressed using the following formula: ;in, This is the weighting coefficient for the unit consumption of explosives. This is a weighting coefficient for block size uniformity. is the weighting coefficient for vibration velocity, and V represents the blasting vibration velocity; U represents the block size uniformity index; W represents the explosive consumption per unit volume.

[0063] The blasting vibration velocity V is expressed by the following formula: ;in, The radial component is obtained from the triaxial velocity sensor; The tangential component is obtained from the triaxial velocity sensor; The vertical component is obtained from the triaxial velocity sensor;

[0064] The block size uniformity index is expressed using the following formula: ;in, The coefficient of variation is used to calculate the post-explosion particle size distribution. The mean / variance is obtained; the post-explosion particle size distribution is expressed using the following formula: ; This represents the post-explosion particle size distribution, indicating particle sizes smaller than [value missing]. The cumulative percentage Characteristic particle size, The uniformity coefficient;

[0065] The specific consumption W of the explosive is expressed by the following formula: ;in, The volume of the explosion; This refers to the amount of explosives loaded in a single segment.

[0066] S5. Based on the multi-objective optimization function and the blasting parameter model, combined with environmental constraints, the Pareto optimal solution set is obtained.

[0067] Step S5 includes the following steps:

[0068] Encoding decision variables: setting hole spacing Row spacing b, hole depth L, and cartridge spacing And the delay δ is the decision variable, forming the decision vector. And set its value range and constraints;

[0069] Fitness assessment: For each individual z in the population, the trained detonation parameter model is called. Obtain the predicted value ,in For blasting vibration, For uniformity of block size, The objective functions are minimizing explosive consumption, maximizing the uniformity of blast fragment size, and minimizing blast vibration.

[0070] Imposing high-altitude and cold-weather conditions and safety constraints: Sadovsky-like empirical formulas for environmental modification. Where K is the first empirical constant; It is the second empirical constant; The distance decay exponent; This represents the maximum safe threshold for blasting vibration velocity. This refers to the amount of explosives loaded in a single stage. The distance between the centers of the explosion; For environment correction functions,

[0071] Non-dominated ranking: Based on Pareto dominance, the population is divided into different non-dominated levels. ;

[0072] Selection, crossover, and mutation: A tournament selection strategy is used to select parent individuals from the population; then, offspring individuals are generated through simulated binary crossover (SBX) and polynomial mutation; the crossover probability is set to 0.9, and the mutation probability is set to 0.1.

[0073] Convergence criterion and output: When the maximum number of iterations is reached or the Pareto front tends to stabilize, the iteration stops and the final Pareto optimal solution set is output.

[0074] The Pareto optimal solution set contains several blasting schemes that achieve different balances between safety, block uniformity, and economy.

[0075] S6. Based on the Pareto optimal solution set, select representative solutions, generate a visualization scheme, and complete the optimization of blasting parameters in open-pit mines in high-altitude cold regions.

[0076] In step S6, a representative solution is selected using the weighted sum method, the ideal point method, or the entropy weight method.

[0077] The multi-objective intelligent optimization method for blasting parameters in high-altitude cold-region open-pit mines also includes a feedback learning mechanism. Specifically, it involves: acquiring PPV using a three-dimensional geophone, calculating the post-blast particle size distribution curve using a camera combined with image recognition technology, calculating unit consumption using charge metering and volume, and finally obtaining the measured blasting results, which are then compared with the predicted results output by the model. or relative error , To preset the error threshold, incremental retraining and hyperparameter fine-tuning are triggered, and the feature interval library and constraint library are updated to achieve closed loop.

[0078] The feedback learning mechanism adopts an incremental learning approach, where each new blasting data is used as a sample to update the database.

[0079] This invention also provides a system for implementing the multi-objective intelligent optimization method for blasting parameters in open-pit mines in high-altitude cold regions, the structural schematic diagram of which is shown below. Figure 2 As shown, it includes a data import module, a feature selection and target setting module, a machine learning model training module, a multi-objective optimization module, an optimal solution export module, and a blasting control module;

[0080] The actual application diagram of the system is shown below. Figure 3 As shown.

[0081] The data import module is used to import environmental data from high-altitude and cold open-pit mines (including temperature, altitude, oxygen content, frozen soil thickness, wind speed, and humidity), rock mechanics parameters (uniaxial compressive strength, elastic modulus, Poisson's ratio, and joint and fracture density), blasting design parameters (borehole diameter, hole depth, hole spacing, row spacing, charge structure, and detonation sequence), and blasting monitoring data (blasting vibration velocity, post-blast block size distribution, throwing range, and flyrock distance). The module also standardizes the data, fills in missing values, and removes outliers to form a standardized input sample set.

[0082] The feature selection and target setting module is used to perform correlation analysis, principal component analysis or recursive feature elimination on multi-source data in the sample set, screen key influencing factors, and select multiple blasting effect targets that need to be optimized collaboratively in this case.

[0083] The machine learning model training module is used to build and train prediction models based on XGBoost, random forest, support vector machine, LightGBM or GBDT. Specifically, it includes dataset partitioning, hyperparameter initialization, cross-validation and performance evaluation. The model training results output a prediction blasting effect index, which is used to establish a nonlinear prediction relationship between blasting parameters and blasting effect.

[0084] The multi-objective optimization module sets optimization objectives such as minimizing blasting vibration velocity, maximizing blasting block uniformity, and minimizing explosive consumption. At the same time, it introduces environmental constraints and collaborative optimization intervals for different parameters. It uses the non-dominated sorting genetic algorithm NSGA-II or other multi-objective evolutionary algorithms for optimization. Through non-dominated sorting, crowding calculation and elite retention mechanisms, it obtains the Pareto optimal solution set.

[0085] The optimal solution derivation module is used to select representative solutions from the Pareto optimal solution set, select the optimal combination of blasting parameters through weighted method, ideal point method or entropy weight method, and generate hole mesh parameter table, charge list and detonation network diagram. At the same time, it provides visualization results such as block size distribution curve, vibration prediction curve and three-dimensional stacking morphology to assist designers in scheme evaluation.

[0086] The blasting control module is used to convert the optimal combination of blasting parameters into instructions to be executed and send them to the unmanned drilling rig, automatic charging vehicle and intelligent detonator controller. Through remote communication, it realizes the execution of borehole layout, charge amount control and detonation sequence, and forms a closed-loop control by combining monitoring results, thereby improving the automation level and execution accuracy of blasting design.

[0087] The method of the present invention will be further described below with reference to an embodiment:

[0088] The intelligent optimization method for multi-target blasting provided in this embodiment includes the following steps:

[0089] S1 Data Acquisition and Loading: Acquires historical blasting database and on-site monitoring data, including:

[0090] a) Environmental parameters: temperature T (°C), altitude H (m), oxygen volume fraction O (%), and permafrost thickness Ft (m);

[0091] b) Rock mass parameters: uniaxial compressive strength (MPa), elastic modulus E (GPa), Poisson's ratio ν, joint density Jv ( ), lithological category;

[0092] c) Design parameters: aperture d (mm), aperture depth L (m), aperture spacing a (m), row spacing b (m), resistance line W (m), charge structure (cartridge diameter / spacing) / coupling coefficient), delay network (ms) etc.;

[0093] d) Target parameters: triaxial velocity components , , (cm / s), post-explosion particle size distribution Explosive consumption per unit W ( The data was imported using a unified template and then subjected to noise reduction, missing value imputation (mean imputation or multiple imputation), and standardization (Z-score or MinMax).

[0094] S2 Data Preprocessing and Feature Engineering: Perform Spearman correlation test on all inputs X ( (Shortlisted), and key features were screened using recursive feature elimination or principal component analysis (PCA). ;

[0095] S3 Machine Learning Modeling and Prediction: Based on Training the blast effect prediction model Optional models include XGBoost, Random Forest (RF), Support Vector Machine (SVM), LightGBM, and GBDT. Specific steps include:

[0096] Data partitioning: Training / Validation / Test = 8 / 1 / 1, stratified sampling by burst zone or step;

[0097] Hyperparameter optimization intervals are defined as follows: for XGBoost / LightGBM / GBDT, n_estimators∈[100,1000], learning_rate∈[0.01,0.3], max_depth∈[3,12], subsample∈[0.5,1.0], colsample_bytree∈[0.5,1.0]; for RF, n_estimators∈[200,1200], max_depth∈[5,20], max_features∈{sqrt, log2,0.3~0.8}; for SVM, C∈[1,200], γ∈[1e-4,1e-1], kernel function∈{RBF, poly}.

[0098] Optimization strategy: Use five-fold cross-validation combined with grid / Bayesian optimization for hyperparameter optimization;

[0099] Evaluation metrics: RMSE, MAE, The one with the best overall score on the validation set will be the final one. .

[0100] For example, Table 1 shows the comparison results of the prediction performance of different machine learning models.

[0101] Table 1. Comparison of prediction performance of models built with different machine learning algorithms

[0102] As shown in Table 1, the XGBoost model performs best in terms of accuracy and stability, therefore XGBoost is selected as the final prediction model in this embodiment.

[0103] S4 Multi-Objective Cooperative Optimization: A multi-objective optimization model is established using NSGA-II. Specific steps include:

[0104] Decision variable coding: setting hole spacing Row spacing b, hole depth L, and cartridge spacing The decision vector is composed of decision variables such as delay δ. And set its value range and constraints, for example: , , , It is a positive integer.

[0105] Fitness assessment: For each individual z in the population, the trained prediction model is invoked. Obtain the predicted value ,in For blasting vibration, For uniformity of block size, It is a unit consumption.

[0106] With the objective functions of minimizing explosive consumption, maximizing the uniformity of blast fragment size, and minimizing blast vibration, and with the constraints of high-altitude cold environment and safety, the following is a Sadovsky-like empirical formula with environmental modifications: Where K is an empirical constant, which is generally taken as 50~350 depending on the lithology; The dosage effect index ranges from 1.2 to 1.8. The distance attenuation index ranges from 1.3 to 2.0.

[0107] Non-dominated ranking: Based on Pareto dominance, the population is divided into different non-dominated levels. ;

[0108] Selection, crossover, and mutation: A tournament selection strategy is used to select parent individuals from the population; then, offspring individuals are generated through simulated binary crossover (SBX) and polynomial mutation. The crossover probability is set to 0.9, and the mutation probability is set to 0.1.

[0109] Convergence Criteria and Output: When the maximum number of iterations is reached or the Pareto front stabilizes, iteration stops, and the final Pareto optimal solution set is output. This solution set contains multiple blasting schemes that achieve different balances between safety, block uniformity, and economy.

[0110] Representative solutions are selected from the optimal solution set based on production preference using either a weighted ideal point or an entropy-weighted TOPSIS. (Parameters such as hole spacing a, row spacing b, hole depth L, charge spacing αr, delay strategy δ, etc.) are used to generate a hole network parameter table, charge list, and detonation network. Control commands are then sent to the unmanned drilling rig, automatic charge loading vehicle, and intelligent detonator controller.

[0111] S6 Monitoring and Feedback Relearning: After detonation, PPV is acquired using a three-dimensional detector (sampling ≥ 5 kHz, range ≥ 50 cm / s), block size curve is calculated using UAV / fixed camera + image recognition technology, and unit consumption is calculated based on charge metering / volume; thus obtaining the measured values. With prediction Comparison: If <0.90 or relative error> (e.g., 5% to 10%), triggering incremental retraining and hyperparameter fine-tuning, and updating the feature interval library and constraint library to achieve closed loop.

[0112] For example, Table 2 shows the comparison results between the optimized scheme of the present invention and the traditional empirical method.

[0113] Table 2 Comparison results between the optimized scheme of this invention and traditional empirical methods

[0114] As can be seen from Table 2, the method of the present invention takes into account blasting safety, block size uniformity and explosive utilization rate through multi-objective collaborative optimization. This makes the method not only adaptable to extreme environments, but also avoids the one-sidedness of single-objective optimization, and greatly improves the intelligence level of blasting design and implementation.

Claims

1. A multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine, characterized in that, The method comprises the following steps: S1. Collecting data of the blasting environment and preset blasting targets to obtain a parameter data set; S2. Preprocessing the parameter data set to obtain a feature sample set; S3. Constructing an initial blasting parameter model based on a machine learning algorithm, and optimizing the hyperparameters of the model to obtain a blasting parameter model; S4. Constructing a multi-objective optimization function with the minimization of blasting vibration velocity, the maximization of blasting fragmentation uniformity, and the minimization of explosive unit consumption as the optimization objectives; S5. Solving the multi-objective optimization function and the blasting parameter model based on the environmental condition constraints to obtain a pareto optimal solution set; S6. Selecting a representative solution from the pareto optimal solution set, generating a visual scheme, and completing the blasting parameter optimization of the high-altitude cold open-pit mine.

2. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, In step S1, the parameter data set includes environmental parameters, rock mass parameters, design parameters, and target parameters. The environmental parameters include temperature T, altitude H, oxygen volume fraction O, frozen soil thickness Ft, wind speed Ws, and humidity Rh. The rock mass parameters include uniaxial compressive strength , elastic modulus E, Poisson's ratio v, joint fissure density Jv, and lithology category. The design parameters include aperture d, hole depth L, hole spacing a, row spacing b, charge structure / charge column spacing , delay sequence ; The target parameters include vibration velocity V, post-blast lump size particle size distribution , and explosive unit consumption W.

3. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, In step S2, the parameter data set is standardized, outlier processed, and subjected to correlation and recursive feature elimination to obtain a feature sample set X.

4. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, In step S3, the initial blasting parameter model is predicted by establishing a nonlinear mapping relationship between the blasting parameters and the blasting effects. The model training is based on the feature sample set X, and the blasting fragmentation distribution, blasting vibration velocity, and explosive unit consumption are used as output targets for training. The hyperparameter optimization adopts grid search or Bayesian optimization; based on five-fold cross-validation, root mean square error RMSE, mean absolute error MAE and determination coefficient And a weighted fusion method is used to construct a comprehensive performance evaluation index evaluation model performance, and select the model with the best comprehensive performance as the final model; when Less than 0.9, trigger model retraining.

5. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, In step S4, the multi-objective optimization function F is expressed using the following equation: ; wherein, is a weight coefficient of explosive unit consumption, is a weight coefficient of block uniformity, is a weight coefficient of vibration velocity, and ; V is a blasting vibration velocity; U is a block uniformity index; and W is an explosive unit consumption.

6. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 5, characterized in that, The blasting vibration velocity V is expressed by the following equation: ; wherein, is a radial component acquired by a three-directional vibration velocity sensor; is a tangential component acquired by a three-directional vibration velocity sensor; is a vertical component acquired by a three-directional vibration velocity sensor; The blockiness uniformity index is expressed using the following formula: ; wherein, is the coefficient of variation, obtained by calculating the mean / variance of the post-blast blockiness particle size distribution; ; the post-blast blockiness particle size distribution is expressed using the following formula: ; is the post-blast blockiness particle size distribution, representing the cumulative proportion of particle sizes smaller than ; is the characteristic particle size, is the uniformity coefficient; The explosive unit consumption W is expressed by the following equation: ; wherein is the blasting volume; is the single segment charge quantity.

7. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, The machine learning algorithm includes XGBoost, random forest, support vector machine, LightGBM, and GBDT. Coding decision variable: set hole distance , row distance b, hole depth L, cartridge spacing and delay δ as decision variables, constitute decision vector ; and set its value range and constraint conditions; Fitness evaluation: for each individual z in the population, call the trained blasting parameter model to obtain the predicted value , where is the blasting vibration, is the block uniformity, is the unit consumption; the objective function is to minimize the unit consumption of explosives, maximize the blasting block uniformity, and minimize the blasting vibration; Applying high-cold environment and safety constraints: Environment-modified Sadaovski-type empirical formula: ; wherein K is a first empirical constant; is a second empirical constant; is a distance attenuation exponent; is a maximum safety threshold of the blasting vibration velocity; is a single-segment charge quantity; is a blast center distance; is an environment-modified function, Non-dominated sorting: The population is divided into different non-dominated levels according to the Pareto dominance relation ; Step S5 comprises the following steps: Selection, crossover, and mutation: adopt a tournament selection strategy to select parent individuals from the population; then generate offspring individuals through simulated binary crossover (SBX) and polynomial mutation; the crossover probability is set to 0.9, and the mutation probability is set to 0.

1.

8. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, Convergence criterion and output: when the maximum iteration number is reached or the Pareto frontier tends to be stable, stop iteration and output the final Pareto optimal solution set. The Pareto optimal solution set contains several groups of blasting schemes that achieve different balances between safety, fragmentation uniformity, and economy.

9. The multi-objective intelligent optimization method for blasting parameters of high-altitude cold open-pit mine according to claim 1, characterized in that, The representative solution is selected from the Pareto optimal solution set by weighted sum method, ideal point method, or entropy weight method. The feedback learning mechanism is specifically: collecting PPV with a three-way detector, calculating the post-blast block size distribution curve with a camera combined with image recognition technology, calculating the unit consumption with charge metering and volume, and finally obtaining the measured blasting results, which are compared with the predicted results output by the model; if or relative error , is a preset error threshold, triggering incremental retraining and hyperparameter fine-tuning, and updating the feature interval library and the constraint library to realize closed loop.

10. A system for implementing the multi-objective intelligent optimization method of the blasting parameters in high-altitude cold open-pit mines according to any one of claims 1 to 9, characterized in that it comprises: The high-altitude cold open-pit mine blasting parameter multi-objective intelligent optimization method further comprises a feedback learning mechanism that adopts incremental learning, and each new blasting data is used as a sample to update the database. The method comprises a data import module, a feature selection and target setting module, a machine learning model training module, a multi-objective optimization module, an optimal solution export module, and a blasting control module. The data import module is used to import high-altitude cold open-pit mine environmental data, rock mass mechanical parameters, blasting design parameters, and blasting monitoring data, and to standardize, fill in missing values, and remove outliers to form a standardized input sample set. The feature selection and target setting module is used for correlation analysis, principal component analysis or recursive feature elimination on the multi-source data in the sample set, screening of key influence factors, and selection of multiple blasting effect targets to be synergistically optimized this time; The machine learning model training module is used for constructing and training a prediction model based on XGBoost, random forest, support vector machine, LightGBM or GBDT, specifically including data set division, hyperparameter initialization, cross-validation and performance evaluation, model training result output prediction blasting effect index, and establishment of a nonlinear prediction relationship between blasting parameters and blasting effect; The multi-objective optimization module sets the optimization objectives of minimizing blasting vibration velocity, maximizing blasting fragmentation uniformity and minimizing explosive unit consumption, simultaneously introduces environmental constraint conditions and the synergistic optimization interval of different parameters, adopts the non-dominated sorting genetic algorithm NSGA-II or other multi-objective evolutionary algorithms for optimization, obtains a Pareto optimal solution set through non-dominated sorting, congestion calculation and elite reservation mechanism; The optimal solution derivation module is used for screening representative solutions from the Pareto optimal solution set, selecting the optimal blasting parameter combination through the weighting method, ideal point method or entropy weight method, and generating a hole pattern parameter table, a charging list and an initiation network diagram, while providing visual results including fragmentation distribution curve, vibration prediction curve and three-dimensional accumulation form to assist designers in scheme evaluation; The blasting control module is used for converting the optimal blasting parameter combination into to-be-executed instructions, issuing the instructions to unmanned drilling machines, automatic charging vehicles and intelligent detonator controllers, realizing drilling arrangement, charge amount control and initiation sequence execution through remote communication, and forming a closed-loop control combined with monitoring results to improve the automation level and execution accuracy of blasting design.

Images