Blasting dust suppressant formula prediction method and intelligent dust fall system

Abstract

The invention relates to the technical field of surface mine blasting dust control, in particular to a blasting dust suppressant formula prediction method and an intelligent dust falling system. The method comprises the following steps: acquiring an environment-material multi-dimensional data set; generating a target data set through abnormal value elimination and feature weighting preprocessing; constructing a random forest hyper-parameter optimization model fused with a grey wolf optimization algorithm (GWO); predicting dust suppressant formula key parameters by using the optimized random forest; and dynamically adjusting a formula and controlling the intelligent dust fall system in combination with real-time environmental data. Artificial experience is replaced by an intelligent algorithm, the environmental adaptability of a dust suppression formula is improved, the risk that the dust concentration exceeds the standard is reduced, and waste of chemical reagents is reduced.

Description

Technical Field

[0001] This invention relates to the field of intelligent dust control in open-pit mine blasting, specifically a method for predicting the formulation of blasting dust suppressants and an intelligent dust reduction system. Background Technology In open-pit mining operations, blasting is an open-space blasting process, characterized by numerous dust-generating points, large dust volumes, and wide diffusion ranges. Dust suppressants play a crucial role in reducing the amount of fumes and dust generated after blasting. They not only significantly reduce the large amounts of smoke and dust produced during blasting, protecting the health of mine workers, but also significantly improve the cleanliness of mining operations, contributing to the construction of green mines. However, existing technologies typically rely on static experimental data and cannot adapt to dynamic environmental changes at the blasting site, resulting in large fluctuations in dust suppression efficiency. Traditional manual adjustments suffer from problems such as response lag and excessive use of chemicals.

[0002] To determine the optimal method for predicting blasting dust suppressant formulations, current methods often employ model experiments and numerical simulations to model the dust suppression effect under different formulation conditions. However, these methods are limited by computational resources and can only simulate the properties of single-component dust suppressants, making it difficult to fully reflect the complex situation of multiple-component dust suppressants acting simultaneously. Furthermore, even with sufficient computational resources, high-precision, large-scale numerical simulations require a significant amount of time, making it difficult to respond promptly to on-site needs and develop flexible dust suppression solutions.

[0003] Existing technologies primarily rely on subjective judgment based on experience and numerical simulations performed by computers. These methods, while effective in determining the optimal dust suppressant formulation for dust suppression, all have limitations, including but not limited to excessive subjectivity, limited simulation accuracy, and excessive time consumption. Therefore, there is an urgent need for more scientific, efficient, and flexible methods for predicting blasting dust suppressant formulations to optimize dust suppression design schemes and improve the overall effectiveness of mining and environmental management. Summary of the Invention

[0004] To address the shortcomings of existing methods and the inadequacies of practical applications, and in order to reduce errors caused by subjective human experience, decrease post-blasting dust pollution, and improve the cleanliness of the blasting environment, this invention aims to achieve accurate prediction and intelligent dust control of blasting dust suppressant formulations by constructing an environment-formulation coupled database and combining random forest algorithms and dynamic spraying control logic. In the first aspect, the present invention constructs a mining environment-formulation coupled database to collect multidimensional data on the mining environment, including databases of environmental parameters such as rock hardness, ambient humidity, and dust particle size distribution. Data is collected through equipment such as laser scattering instruments and tension meters to form a characteristic map of blasting dust.

[0005] Table 1: Environmental Parameter Collection Specifications The collected data is preprocessed, including outlier removal and missing value imputation, to obtain high-quality multi-dimensional target data of the mining environment; an improved optimization algorithm and optimized hyperparameter combination are established; a parameter update model is established based on the improved optimization algorithm; the optimal dust suppressant formulation parameters are obtained through the parameter update model, and an optimal formulation prediction model is obtained based on the optimal formulation parameters; the optimal formulation prediction model is trained and predicted based on the multi-dimensional target data of the mining environment to select the best dust suppressant formulation after blasting in open-pit mines.

[0006] Secondly, this invention adopts a GWO-random forest joint optimization architecture: the hyperparameter combination of the random forest is optimized by the Grey Wolf Optimization Algorithm (GWO) to improve the prediction accuracy of the model.

[0007] Thirdly, the present invention also provides an intelligent dust suppression system after blasting in an open-pit mine, which can efficiently execute a blasting dust suppressant formulation prediction method provided by the present invention. The system uses dynamic spray control logic and can dynamically adjust the flow ratio of dust suppressant components according to the prediction results to achieve intelligent dust suppression.

[0008] This invention processes and analyzes multidimensional data of the mining environment, and combines it with an improved optimization algorithm to find the optimal dust suppressant formulation parameters for the current mining environment conditions. This results in the best dust suppressant formulation, which can reduce dust generation and diffusion after blasting and optimize the overall mining environment.

[0009] Optionally, processing the multidimensional data of the blasting environment to obtain multidimensional target data of the blasting environment includes: performing a first data processing on the multidimensional data of the blasting environment and obtaining a first data processing result, wherein the first data processing includes an outlier removal algorithm; and performing a second data processing on the first data processing result to obtain multidimensional target data of the blasting environment, wherein the second data processing includes a feature weighting algorithm. This invention removes outliers and weights the multidimensional data of the blasting environment, which helps to obtain high-quality multidimensional target data of the blasting environment and improves the accuracy and reliability of data analysis results.

[0010] Optionally, the outlier removal algorithm satisfies the following relationship: in, Represents the characteristic mean. Indicates standard deviation; The feature weighting algorithm satisfies the following relationship: , in, Indicates information gain. This represents the proportion (probability) of the k-th class feature in the dataset. Represents the entropy function. Representation of features A subset that takes the value v; Indicates the proportion of a subset of samples. Describes the entropy of a subset; Representation of features Normalized importance weights.

[0011] Optionally, the GWO-Random Forest joint optimization architecture includes: in the GWO layer, using the Gray Wolf Optimization Algorithm to optimize the hyperparameters of the random forest, including the number of trees Ntrees and the maximum depth Dmax. The GWO algorithm simulates the hunting behavior of gray wolves, finding the optimal solution through group cooperation. The Gray Wolf Optimization Algorithm is improved, and a fitness function is defined; the fitness function satisfies the following relationship; in, It is the root mean square error, calculated using the following formula: here, This is the actual value. It is a predicted value. It refers to the number of samples.

[0012] Optionally, the GWO-Random Forest joint optimization architecture also includes establishing a random forest layer. In this layer, the GWO-optimized hyperparameters Ntrees and Dmax are used to build a random forest model to predict the key parameters of the dust suppressant. The random forest model consists of Ntrees decision trees, each with a maximum depth of Dmax. Prediction Results It is the average of all decision tree predictions: in, It is the first The prediction results of the decision trees for input x.

[0013] Optionally, based on the aforementioned environment-formulation coupling database and improved optimization algorithm, the dynamic spraying control logic dynamically adjusts the flow ratio of dust suppressant components according to the prediction results to achieve intelligent dust suppression. The dynamic spraying control logic satisfies the following relationship: (Wind speed, distance from dust source) in, For the first The flow rate of the component, wi is the weighting factor, and K is the adjustment coefficient, which is related to wind speed and distance from the dust source.

[0014] Beneficial effects of the present invention Obtaining the optimal formulation parameters through the optimized hyperparameter model includes: optimizing the model parameters of the formulation prediction model using the hyperparameter model, and obtaining the optimal formulation parameters. This invention, through the hyperparameter model, can systematically adjust the key parameters of the formulation prediction model, thereby improving the accuracy and reliability of the prediction results.

[0015] This invention utilizes multi-dimensional target data of the open-pit mine blasting environment to train and predict the optimal formulation prediction model for selecting the best dust suppressant formulation after blasting. The process includes: training and predicting the optimal formulation prediction model based on multi-dimensional target data of the mine environment; using prediction evaluation indicators including weighted root mean square error, symmetric mean absolute percentage error, and mean square error; and combining the multi-dimensional target data of the mine environment, the optimal formulation prediction model, and the prediction evaluation indicators to select the best dust suppressant formulation after mine blasting. This invention improves prediction accuracy, scientific rigor, and objectivity by training and predicting the optimal formulation prediction model based on multi-dimensional target data of the mine environment to select the best dust suppressant formulation after open-pit mine blasting.

[0016] This invention provides an intelligent dust suppression system for mine blasting, capable of efficiently executing a dust suppressant formulation prediction method provided by this invention. The system includes a data acquisition device, a processor, an output device, and a spray control device, wherein the data acquisition device, processor, output device, and spray control device are interconnected. The processor integrates an optimization module for storing a computer program, which includes program instructions. The processor is configured to invoke the program instructions. The system provided by this invention has a compact structure, strong applicability, and greatly improves operating efficiency. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for predicting the formulation of blasting dust suppressant according to the present invention; Figure 2 This is a flowchart illustrating the GWO-Random Forest joint optimization architecture for dust suppressant formulation prediction in this invention. Figure 3 This is a schematic diagram of the intelligent dust suppression system after mine blasting according to the present invention. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to specific embodiments.

[0019] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0020] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale. Please see Figure 1 To improve the convenience and efficiency of blasting dust suppression operations and reduce the impact of subjective human judgment on the results, this invention optimizes multi-dimensional data of the blasting environment to ensure data accuracy and comprehensiveness. The data is then input into an improved algorithm to adjust model parameters, enabling the model to more accurately predict the complex relationship between open-pit mine blasting environmental conditions and the dust suppression effect of dust suppressant formulations. This improves the overall quality of post-blasting operations and provides strong support for green and safe mine production. This invention provides a method for predicting blasting dust suppressant formulations, which includes the following steps: S1. Acquire multi-dimensional data of the blasting environment, process the multi-dimensional data to obtain multi-dimensional target data of the blasting environment, and the implementation steps and specific contents are as follows: To comprehensively and accurately collect multidimensional data related to the blasting environment, the multidimensional data of the blasting environment in this embodiment mainly covers geological characteristic parameters, such as rock type (granite / sandstone / shale, etc.), hardness level (Mohs hardness), joint development degree (joint density / m³), and weathering layer thickness (m); borehole layout parameters, including but not limited to borehole diameter, borehole depth, borehole spacing, row spacing, type of packing material, and packing length; explosive parameters, such as explosive type, charge per borehole, charge density, and delay setting; and dust suppressant physicochemical parameters, including but not limited to surfactant type (anionic / cationic), polymer concentration, viscosity, pH value, and environmental temperature / humidity adaptability.

[0021] In an optional embodiment, rock samples can be obtained using rock drilling equipment, including but not limited to hydraulic drilling rigs and rotary drilling rigs. The mineral composition is then determined using a rock analyzer, such as an X-ray diffractometer or electron microscope. A three-dimensional model of joint distribution is obtained using a ground-based three-dimensional laser scanner. Hole coordinates are measured using a total station. A large-scale borehole grid is scanned using a UAV equipped with a lidar system. The surfactant components are quantitatively analyzed using high-performance liquid chromatography (HPLC), and viscosity and shear stress response are measured using a rheometer. The stability of the dust suppressant is measured using a high-low temperature constant-temperature test chamber at temperatures ranging from -20°C to 50°C and humidity.

[0022] In this embodiment, relevant monitoring equipment is deployed to obtain multi-dimensional data on the blasting environment. Different data acquisition equipment needs to be selected or combined depending on the geological conditions and blasting requirements, thereby ensuring and improving the efficiency and accuracy of data acquisition, and providing strong support for the design of dust suppressant formulations.

[0023] Then, the geological multidimensional data is collected and processed to obtain the data processing results. In this embodiment, the data processing mainly includes the removal of outliers, and the specific implementation details are as follows: To ensure the robustness and accuracy of subsequent data analysis, outliers in the multidimensional data of the blasting environment are caused by measurement errors, data entry errors, or extreme natural events.

[0024] First, the text data of the multidimensional data of the blasting environment is converted into numerical data for subsequent statistical analysis or machine learning modeling. This mainly involves text encoding techniques, such as one-hot encoding, tag encoding, bag-of-words model, word frequency-inverse document frequency conversion, etc. Based on this, the formulation method of the present invention can be applied to different scenarios. When the text information is represented digitally, the rock type and its numerical encoding relationship in this embodiment are as follows: [granite, sandstone, shale...] = [1, 2, 3, ...], which refers to granite (1), sandstone (2), shale (3), etc. The surfactant category and its numerical encoding relationship are as follows: [sodium dodecyl sulfate, Tween 80, ...] = [1, 2, ...], which refers to sodium dodecyl sulfate (1), Tween 80 (2), etc. In addition to directly mapping to a numerical sequence, various other methods can be used to express this correspondence to increase readability, understandability, or conform to specific application scenarios.

[0025] Optionally, the outlier removal algorithm satisfies the following relationship: in, Represents the characteristic mean. Standard deviation is used to measure the dispersion of data distribution, while ValidRange represents the range of valid values, i.e., the confidence interval. The normal range of data refers to the normal fluctuation range of multidimensional data or data characteristics of the blasting environment determined by statistical methods. Using the normal range of data as a reference standard is beneficial to improving the quality and accuracy of the collected multidimensional data.

[0026] In this embodiment, the normal distribution theory is used for calculation. It is assumed that 99.7% of the normal data is within the range of ±3σ. A dynamic threshold is set, and the rejection criteria are automatically adjusted according to the real-time data distribution to avoid the subjectivity of the fixed threshold method and adapt to the data characteristics of different environments.

[0027] Next, the data after removing outliers undergoes a second data processing step to obtain multi-dimensional target data for the blasting environment. In this embodiment, the second data processing mainly includes a feature weighting algorithm, the specific implementation of which is as follows: In this embodiment, the first data processing result is subjected to a second data processing, which is to perform feature engineering processing on the multidimensional data of the blasting environment after removing outliers, in order to obtain multidimensional target data of the blasting environment.

[0028] The feature weighting algorithm satisfies the following relationship: , in, Information gain measurement features Contribution to the classification results This represents the proportion (probability) of the k-th class feature in the dataset. Represents the total number of features. The entropy function representing dataset S, Representation of features Values A subset of samples; Indicates the proportion of a subset of samples. Representing a subset Entropy; Indicates the first i The weighting factors of each sample, i.e. The normalized information gain value, ; Representation of features The set of all possible values. The index represents the set of all possible values.

[0029] In this embodiment, information gain calculation is considered to quantify the ability of features to distinguish target variables by comparing the entropy changes before and after feature segmentation; weight normalization is adopted to normalize the information gain into a probability distribution, giving higher weights to features with high information gain, thereby improving the model's sensitivity to key features.

[0030] In a multidimensional dataset, blasting environment parameter values ​​are specific to a particular dimension or location. Here, "dimensional" refers to a specific dimension or location within the geological multidimensional dataset. When processing these parameter data, special attention must be paid to the accuracy, completeness, and consistency of the relevant data to avoid biases or errors in the analysis results due to data issues, ensuring that the data meets the needs of analysis or modeling.

[0031] Furthermore, the method for acquiring multidimensional target data in the blasting environment in this embodiment is merely an optional condition of the present invention. In other embodiments, the method for acquiring multidimensional target data can be flexibly adjusted according to the actual geological data and multidimensional data conditions, optimized for the characteristics of the collected data, reducing unnecessary calculation and data conversion steps, thereby improving the overall data processing efficiency and enhancing the accuracy and reliability of the multidimensional target data.

[0032] S2. Establish a coupled model of the Gray Wolf optimization algorithm and random forest, and optimize the hyperparameter combination. The specific steps and implementation content are as follows: In an optional embodiment, the gray wolf optimization algorithm is first introduced, and then the optimization algorithm is improved to obtain an improved optimization algorithm, the details of which are as follows: In this embodiment, the gray wolf optimization algorithm was introduced and improved. The gray wolf optimization algorithm is a metaheuristic optimization algorithm inspired by the social hierarchy and hunting behavior of gray wolf groups, which simulates the cooperative hunting process of gray wolves in nature.

[0033] The gray wolf optimization algorithm mainly consists of three stages: social hierarchy, surrounding the prey, and hunting behavior. Social hierarchy refers to the strict social hierarchy within the gray wolf pack. The alpha wolf is responsible for deciding the group's direction; the next-ranking wolves assist the alpha in decision-making; and the remaining wolves obey the commands of the alpha and the next-ranking wolves, relaying information to the lowest-ranking wolves for execution. Surrounding the prey refers to the gray wolves gradually approaching and surrounding the prey after spotting it. In the algorithm, this process is simulated through a mathematical model, causing individual gray wolves to continuously move closer to the current optimal solution (corresponding to the alpha wolf's position), gradually narrowing the search range. Hunting behavior refers to the gray wolves launching a coordinated attack on the prey. In the algorithm, the next-ranking wolves and the remaining wolves also participate in guiding the search, with multiple gray wolves approaching the prey from different directions to increase the probability of finding the optimal solution.

[0034] To further improve local search capabilities, the improved algorithm employs a non-linear convergence factor, whose rate of change is dynamically adjusted based on the number of iterations and the search situation. In the early stages of iteration, the convergence factor changes slowly, ensuring the algorithm has sufficient time for global search; in the later stages, the convergence factor changes rapidly, prompting the algorithm to converge to the optimal solution more quickly. A restart mechanism is also included. If the algorithm fails to show significant performance improvement after several consecutive iterations, it indicates that it may be trapped in a local optimum. In this case, the positions of some or all the gray wolves are reinitialized, allowing the algorithm to escape the local optimum trap and continue searching for the global optimum in the search space.

[0035] In the improved optimization algorithm described above, the nonlinear convergence factor satisfies the following relationship: in, T represents the current iteration number, and Tmax represents the maximum iteration number. As the iteration number increases, The value of is gradually decreased, thereby controlling the search range of the algorithm.

[0036] The position update formula is: t+ 1 = , in The position of the alpha wolf. =2 r2 is a random vector, where r2 is a random number in the range [0,1]. They represent Wolf, wolves and The wolf's position, that is, the positions of the three solutions: the best, the second best, and the third best; t+ 1 This indicates the gray wolves' position in the next generation. Indicates the current location of the gray wolf; Indicates the current wolf ( wolves or wolves or The direction and distance the wolf moves; in this way, individual gray wolves continuously update their positions based on the position of the alpha wolf, thereby achieving the search for the optimal solution.

[0037] Furthermore, the execution steps of the improved gray wolf optimization algorithm satisfy the following relationship: 1. Initialize the wolf pack position, that is, randomly initialize the values ​​of Ntrees and Dmax, and calculate the fitness of each wolf; then determine the alpha, beta, and delta wolves (i.e., the optimal, second-best, and third-best solutions) based on the fitness, and then update the wolf pack position, moving closer to the alpha, beta, and delta wolves; finally, repeat the above steps until the maximum number of iterations is reached or the fitness requirement is met.

[0038] In an optional embodiment, in addition to introducing the improved gray wolf optimization algorithm, a random forest hyperparameter optimization algorithm is also introduced, the details of which are as follows: In this embodiment, the random forest algorithm is introduced and improved. Random forest is an ensemble learning method whose performance is affected by several hyperparameters. The random forest optimization algorithm mainly optimizes the following three hyperparameters: the number of decision trees Ntrees, the maximum depth Dmax, and the minimum number of split samples Smin. The number of decision trees affects the generalization ability and computational efficiency of the random forest. Appropriately increasing the number of decision trees can improve the accuracy of the model. The maximum depth limits the growth of the decision trees and avoids overfitting. By optimizing the maximum depth, the decision trees can have better generalization ability on complex data. The minimum number of split samples specifies the minimum number of samples required for a decision tree node to split, which helps prevent the decision tree from overgrowing. In this embodiment, the number of decision trees Ntrees∈[50,200], the maximum depth Dmax∈[5,30], and the minimum number of split samples Smin 5.

[0039] Furthermore, a joint optimization architecture of GWO and Random Forest is established, the specific details of which are as follows: The GWO-Random Forest joint optimization architecture consists of a GWO layer and a random forest layer. Through their collaborative work, it achieves optimal prediction of dust suppressant formulation parameters. The first layer is the GWO layer, whose main task is to optimize the hyperparameter combinations of the random forest. A fitness function is used to evaluate the merits of different hyperparameter combinations; the formula for calculating the fitness function is as follows: in, It is the root mean square error, calculated using the following formula: here, This is the actual value. It is a predicted value. It refers to the number of samples.

[0040] The GWO algorithm finds the optimal hyperparameter combination by continuously adjusting the combination of hyperparameters to maximize the value of the fitness function.

[0041] Optionally, based on the hyperparameter combination optimized by the GWO layer, a random forest model is constructed using a random forest layer to predict key parameters of the dust suppressant. The prediction formula satisfies the following relationship: Prediction results It is the average of all decision tree predictions. in, It is the average of all decision tree predictions. It is the p-th decision tree. The input features are used. Random forests improve the accuracy and stability of predictions by averaging the predictions from multiple decision trees.

[0042] Furthermore, the improvement method of the optimization algorithm in this embodiment is only an optional condition of the present invention. In other embodiments, the improvement method of the optimization algorithm can be replaced according to the optimization requirements and model parameter characteristics. By adjusting the improvement method of the optimization algorithm for specific problems and model characteristics, the behavior of the algorithm can be adjusted more precisely, thereby obtaining better performance.

[0043] S3. Obtain the optimal formulation parameters through the optimized hyperparameter model, and obtain the optimal formulation prediction model based on the optimal formulation parameters. The specific implementation steps and related contents are as follows: Based on the aforementioned GWO-Random Forest coupled model, a parameter update model is established. According to the optimized hyperparameters, the random forest model can predict various parameters of the dust suppressant formulation. During training, the model learns the complex relationship between input features (such as rock hardness, ambient humidity, dust particle size distribution, etc.) and output objectives (such as the proportions of different dust suppressant components, dust suppression effect, cost, etc.). The model will make predictions based on the input data and output dust suppression effect evaluation values ​​under different combinations of formulation parameters. We evaluate the prediction results according to the fitness function (defined in S2) and select the formulation parameter combination corresponding to the prediction result with the highest fitness value as the optimal formulation parameter, including but not limited to information such as the type, proportion, and spraying rate of each raw material.

[0044] Furthermore, based on the obtained optimal formulation parameters, an optimal formulation prediction model is established. A dynamic mapping relationship is established between the optimal formulation parameters and the actual dust suppressant formulation, enabling rapid prediction of the optimal formulation under different operating conditions. The optimization process of the parameter update model is as follows; please refer to [link to relevant documentation] for details. Figure 2 .

[0045] In this embodiment, the optimal formulation prediction model adopts a hierarchical structure. The bottom layer is the input layer, which receives various blasting condition-related features, such as blasting scale, dust concentration, ambient temperature, and humidity. The middle layer is the feature processing layer, which performs normalization, feature extraction, and transformation operations on the input features to improve the model's robustness and computational efficiency. The upper-middle layer is the random forest prediction layer, which uses an optimized random forest model to predict the optimal formulation parameters based on the input features. The top layer is the output layer, which outputs the prediction results, including information such as the type, ratio, and spraying amount of each raw material in the dust suppressant.

[0046] Furthermore, the method for establishing the parameter update model in this embodiment is merely an optional condition of the present invention. In other embodiments, the method for establishing the parameter update model can be optimized according to the parameter optimization requirements and the model establishment situation, making the framework of the parameter update model more flexible and adaptable, which helps to deal with complex optimization problems in different fields and improves the universality and application value of the algorithm.

[0047] S4. The optimal formula prediction model is trained and predicted based on the multidimensional data of the blasting environment to select the optimal dust suppressant group distribution ratio for the mine blasting environment. The specific steps and related contents are as follows: To ensure the accuracy and generalization ability of the optimal formulation prediction model, a large amount of historical data was used for training and validation. During training, the dataset was first divided proportionally into a training set (70%), a validation set (15%), and a test set (15%). Next, K-fold cross-validation was used to evaluate the model's performance on different data subsets, avoiding overfitting. Finally, the model's prediction accuracy was quantified using metrics such as weighted root mean square error, symmetric mean absolute percentage error, and mean squared error. If the metrics did not meet expectations, model parameters were adjusted or feature engineering was optimized. By using input features and corresponding optimal formulation parameters as training samples, the random forest algorithm was used to iteratively train the model, continuously adjusting the model's parameters to make the model's prediction results as close as possible to the actual optimal formulation.

[0048] In this embodiment, factors such as blasting conditions and dust suppressant raw materials may change over time and with environmental variations. Therefore, the optimal formulation prediction model needs a dynamic update mechanism. By incrementally updating the optimal formulation prediction model with data and retraining it, the model gains online learning capabilities. Incremental data updates refer to periodically collecting new blasting data and dust suppressant usage effect data to expand the training dataset. Model retraining refers to triggering incremental or full training of the model when new data accumulates to a certain scale, ensuring continuous optimization of the formulation prediction model's predictive ability.

[0049] Furthermore, the specific content of the model parameters in this embodiment is merely an optional condition of the present invention. In other embodiments, the specific content of the model parameters can be replaced according to the parameter optimization requirements and the prediction model structure, which can more accurately control the model's learning process and performance, and help improve the model's prediction accuracy and robustness.

[0050] S5. A smart dust suppression system for mine blasting, which performs dynamic dust suppression operations based on the optimal dust suppressant formula generated by an optimal formula prediction model. The implementation steps and related content are as follows: The intelligent dust suppression system for mine blasting includes a data acquisition device, a processor, an output device, and a spraying control device. These components are interconnected. The data acquisition device integrates a data acquisition module for collecting multi-dimensional data such as rock mass parameters, environmental parameters, and dust suppressant-related parameters. The processor integrates an optimization module for storing computer program instructions, including but not limited to the GWO-random Forest coupled model algorithm. The processor is configured to call these computer program instructions to execute the specific steps of the mine blasting dust suppressant formulation prediction method and related embodiments provided by this invention. The output device integrates a prediction module, which mainly predicts key parameters of the dust suppressant based on the optimized random forest model. The spraying control device integrates a control module, which dynamically adjusts the flow ratio of dust suppressant components according to the prediction results to achieve intelligent dust suppression. This intelligent dust suppression system for mine blasting is structurally complete and objectively stable. For details on the intelligent dust suppression system, please refer to [link to relevant documentation]. Figure 3 .

[0051] The optimal dust suppressant formulation prediction model was trained and predicted based on multi-dimensional target data of blasting conditions. The prediction evaluation indicators include weighted root mean square error, symmetric mean absolute percentage error, and mean square error, the relevant contents of which are as follows: In this embodiment, a multi-dimensional error evaluation system is also constructed to comprehensively analyze the performance of the prediction model. The system mainly includes: Weighted Root Mean Square Error (WRMSE), which effectively solves the problem of traditional RMSE's sensitivity to outliers by assigning differentiated penalty mechanisms to different sample weights; Symmetric Mean Absolute Percentage Error (SMAPE), which uses a symmetric error calculation method to avoid numerical distortion of MAPE when the actual value is close to zero, significantly improving the fairness of prediction across concentration ranges; and Mean Square Error (RSE), which intuitively quantifies the model's ability to capture data distribution characteristics by comparing the ratio of the squared error between the predicted value and the mean of the true value. The formula for the weighted root mean square error (WRMSE) is as follows: in, Indicates the number of samples. This represents the predicted value of the i-th sample. This represents the predicted value of the i-th sample. This is the weighting factor for the i-th sample, reflecting its importance in the blasting operation. The weighting factor can be dynamically adjusted based on factors such as dust concentration, ambient humidity, and blasting scale.

[0052] The formula for the Symmetric Mean Absolute Percentage Error (SMAPE) is as follows: Where m represents the sample size. This represents the true value of the i-th sample. This represents the predicted value of the i-th sample.

[0053] The formula for the mean squared error RSE is as follows: in, Indicates the number of samples. This represents the true value of the i-th sample. This represents the predicted value of the i-th sample. This represents the average of the true values, i.e. = .

[0054] Furthermore, the error analysis method of the present invention is merely an optional condition of the present invention. In one or more other embodiments, the error analysis method can be adjusted according to the optimization objective and prediction results of the prediction model. Different prediction models and optimization objectives have different sensitivities and focuses on errors. Adjusting the error analysis method according to specific circumstances can more accurately evaluate the performance of the model in different aspects, thereby more effectively guiding the optimization and improvement of the model.

[0055] Finally, by combining multi-dimensional target data of the blasting environment, the optimal dust suppressant formulation prediction model, and prediction evaluation indicators, the best formulation for mine blasting is selected.

[0056] In order to solve the problem of selecting the dust suppressant formula after blasting in the process of optimizing mine blasting operations, this invention proposes a prediction method for mine blasting dust suppressant formula. This method deeply integrates the detailed analysis of multi-dimensional target data of blasting environment, the construction of optimal formula prediction model and rigorous prediction evaluation index system, and determines the best dust suppressant formula through scientific calculation rather than relying on human subjective experience.

[0057] This invention first transforms complex textual information into digital data that can be processed by algorithms, ensuring both accuracy and processability. Subsequently, using digitized historical blasting data from mines, the improved algorithm model, with optimized hyperparameters, is deeply trained. The GWO-Random Forest coupled algorithm model, as the core prediction tool, possesses powerful learning capabilities, enabling it to capture the complex relationships between various influencing factors and dust suppressant formulations during blasting operations, and generate accurate prediction results accordingly.

[0058] In this embodiment, to verify the robustness and generalization ability of the prediction model, this study introduces a composite evaluation system centered on the weighted root mean square error (WRMSE), symmetric mean absolute percentage error (SMAPE), and relative squared error (RSE). These indicators not only construct a multi-dimensional error analysis framework but also systematically reveal the model's stability under different operating conditions through techniques such as weight allocation, symmetry correction, and benchmark comparison. This provides quantifiable decision-making basis for optimizing dust suppressant formulations and dynamic control strategies for intelligent dust suppression systems.

[0059] Based on this, not only are errors caused by insufficient or biased human subjective experience effectively reduced, but dust generation and diffusion during blasting are also significantly reduced, thereby protecting the safety of the surrounding environment and facilities. At the same time, more precise selection of dust suppressant formulations further improves the working environment, optimizes blasting effects, and significantly enhances the efficiency and safety of mining operations.

[0060] The present invention will now be described in further detail with reference to specific embodiments.

[0061] Example 1: Dust suppression by blasting in high-dust environments of hard rock Environmental parameters: rock hardness: granite (Mohs hardness 7.2), peak dust concentration ≥800mg / m³, ambient humidity 40%.

[0062] Comparison between predictions and actual measurements: During the prediction model's operation, the random forest model optimized by GWO (hyperparameters Ntrees=182, Dmax=24) compressed the mixing ratio error to within 3%, achieving an actual dust suppression efficiency of 91% (compared to an average of 82% for traditional methods). The system adjusts the flow rate based on real-time dust concentration (monitored by a laser scattering instrument) and dynamically calculates the flow rate using the formula: ΔQi=wi K (wind speed, distance from dust source) reduces reagent waste by 23%.

[0063] Example 2: Dust suppression in high-humidity environments with soft rock Environmental parameters: Lithology: Shale (joint density ≥ 5 joints / m³), ambient humidity 85%, surface tension ≤ 30mN / m Comparison between predictions and actual measurements: During the prediction model's operation, humidity features are weighted (information gain IG=0.62) to avoid misjudgments caused by high humidity interference; accurate pH prediction reduces the excessive use of neutralizers, resulting in a 17% reduction in chemical consumption; response delay error is less than 5 seconds, suppressing dust diffusion range by 40%.

[0064] Example 3: Dust suppression using a complex multi-dust-source blasting network Environmental parameters: borehole grid: 5 rows × 20 holes, blasting delay interval 50ms, dust source distance gradient 20-100m.

[0065] Comparison between predictions and actual measurements: During the operation of the prediction model, formula wi K (wind speed, distance from dust source) ΔQs=wi In K (wind speed, dust source distance), the weight wi decreases exponentially with the dust source distance gradient (20-100m). The dust suppression efficiency is expected to reach 92% (10% improvement over traditional methods), reagent waste is reduced by 25%, and costs are reduced by 23%.

[0066] In summary, this invention proposes a method for predicting the formulation of dust suppressants for mine blasting, which integrates blasting environment data analysis, intelligent prediction model construction, and comprehensive error assessment. This provides technical support for dust suppression operations in mine blasting and enables the intelligent selection of the optimal dust suppressant formulation for mine blasting.

[0067] Please see Figure 3 In an optional embodiment, the present invention also provides a delay selection system between blasting holes in a mine. This system includes a processor, an input device, an output device, and a memory, all interconnected. The memory stores a computer program comprising program instructions. The processor is configured to invoke the program instructions to execute the specific steps of the delay selection method between blasting holes and related embodiments provided by the present invention. The delay selection system between blasting holes of the present invention is structurally complete and objectively stable.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for predicting the formulation of blasting dust suppressant, characterized in that, Includes the following steps: Acquire multidimensional data of the blasting environment and material properties of dust suppressants; The data undergoes outlier removal and feature weighting to generate a dust suppression target dataset. Establish a coupled model of the improved gray wolf optimization algorithm and random forest, and optimize the hyperparameter combination; The optimal formulation parameters are obtained by optimizing the hyperparameter model, and the optimal formulation prediction model is obtained based on the optimal formulation parameters. The optimal formulation prediction model is trained and predicted based on the multidimensional data of the blasting environment to select the optimal dust suppressant group distribution ratio for the mining blasting environment.

2. The method for predicting the formulation of blasting dust suppressant according to claim 1, characterized in that, The process of processing the multidimensional data of the blasting environment to obtain multidimensional target data of the blasting environment includes: The first data processing is performed on the multidimensional data of the blasting environment to obtain the first data processing result, wherein the first data processing includes an outlier removal algorithm; The first data processing result is subjected to a second data processing to obtain multi-dimensional target data of the blasting environment. The second data processing includes a feature weighting algorithm.

3. The method for predicting the formulation of blasting dust suppressant according to claim 2, characterized in that, The outlier removal algorithm satisfies the following relationship: in, Represents the characteristic mean. Indicates standard deviation, The feature weighting algorithm satisfies the following relationship: ， in, Indicates information gain. This represents the proportion of the k-th class feature in the dataset. Represents the total number of features. The entropy function representing dataset S, Representation of features Values A subset of samples; Indicates the proportion of a subset of samples. Representing a subset Entropy; Indicates the first i Weighting factors for each sample; Representation of features The set of all possible values. The index represents the set of all possible values.

4. The method for predicting the formulation of blasting dust suppressant according to claim 1, characterized in that, The establishment of the improved gray wolf optimization algorithm includes: Introducing a nonlinear convergence factor: in, Tmax is the current iteration number, and Tmax is the maximum iteration number. The position update formula is: t+ 1 = ， in The position of the alpha wolf. =2 r2 is a random vector, where r2 is a random number in the range [0,1]. They represent Wolf, wolves and The wolf's location; t+ 1 This indicates the gray wolves' position in the next generation. Indicates the current location of the gray wolf; This indicates the current direction and distance the wolf is moving.

5. The method for predicting the formulation of blasting dust suppressant according to claim 1, characterized in that, The establishment of the coupled model of the improved gray wolf optimization algorithm and random forest, and the optimization of hyperparameter combinations, includes: The number of decision trees, Ntrees∈[50,200], the maximum depth, Dmax∈[5,30], and the minimum number of split samples, Smin 5.

6. The method for predicting the formulation of blasting dust suppressant according to claim 1, characterized in that, The optimal formula prediction model is trained and predicted based on the multidimensional data of the blasting environment to select the optimal dust suppressant distribution ratio in the mine blasting environment. The evaluation indicators for the prediction include weighted root mean square error, symmetric mean absolute percentage error, and mean square error. By combining the multidimensional data of the blasting environment, the optimal formula prediction model, and the evaluation index of the prediction, the optimal blasting delay time between blasting holes in the mine is selected.

7. An intelligent dust suppression system after blasting in an open-pit mine, characterized in that, The intelligent dust suppression system after blasting in an open-pit mine is used to execute the blasting dust suppressant formulation prediction method as described in any one of claims 1-6; The system includes a processor, an input device, an output device, and a memory, which are interconnected. The memory is used to store a computer program, which includes program instructions, and the processor is configured to invoke the program instructions.

Images