Blasting vibration velocity prediction method and device
By augmenting the initial blasting dataset and training it with the XGBoost algorithm, combined with generative adversarial networks, we achieved accurate modeling of the multidimensional influencing factors of blasting vibration in open-pit mines. This solved the problem of low accuracy in traditional PPV prediction methods and improved prediction accuracy and model robustness.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-04-14
AI Technical Summary
In existing technologies, traditional PPV prediction methods have low accuracy and are difficult to fully characterize the complex nonlinear mechanism of blasting vibration. Furthermore, the high cost of blasting in open-pit mines and the limited sample size of on-site measured data lead to inaccurate prediction results.
Data augmentation was performed on the initial blasting dataset. The XGBoost algorithm and generative adversarial network were used to train a target blasting vibration peak prediction model. Deep fusion modeling was performed by combining multiple influencing factors. The target blasting vibration peak prediction model was then trained using the XGBoost algorithm.
It achieves accurate modeling of multi-dimensional influencing factors of blasting vibration in open-pit mines, improves prediction accuracy and model robustness, overcomes the limitations of small sample data, and reduces costs.
Smart Images
Figure CN121858890A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of engineering blasting safety monitoring technology, and in particular to a method and device for predicting blasting vibration velocity. Background Technology
[0002] Open-pit mine blasting is a critical step in mine production, and the resulting blasting vibrations pose potential threats to surrounding buildings, the ecological environment, and the stability of the mine's own slopes. Therefore, accurate prediction of peak particle velocity (PPV) is necessary for blasting design optimization and vibration hazard assessment and control, thereby ensuring safe production and reducing negative environmental impacts.
[0003] In related technologies, traditional PPV prediction methods (such as empirical formulas based on explosive quantity or distance) consider a small number of macroscopic parameters, have a simple model form, and are difficult to fully characterize the complex nonlinear mechanism of blasting vibration, resulting in low prediction accuracy. In addition, each open-pit mine blasting operation is costly, and the sample size of on-site measured vibration data is very limited, which makes the prediction results obtained by machine learning or deep learning algorithms less accurate. Summary of the Invention
[0004] The present invention aims to at least partially solve one of the technical problems in the related art.
[0005] To achieve the above objectives, the present invention provides a method for predicting blasting vibration velocity, the method comprising: Obtain the initial blasting dataset and perform data augmentation on the initial blasting dataset to obtain the augmented blasting dataset; Based on the enhanced blasting dataset, a target blasting vibration velocity peak prediction model is obtained by training using the XGBoost algorithm. Obtain the blasting input data that needs to be analyzed; The blasting input data is input into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak. If the peak blasting velocity exceeds the blasting vibration threshold, the blasting input data will be adjusted.
[0006] The blasting vibration velocity prediction method of this invention may also have the following additional technical features: In one embodiment of the present invention, the initial blasting dataset includes input data such as row spacing, step height, blockage length, bottom hole depth, borehole diameter, explosive consumption, charge amount, minimum resistance line, and blast center distance for each blasting scheme, as well as output data such as peak blasting vibration velocity.
[0007] In one embodiment of the present invention, the step of performing data augmentation on the initial blasting dataset to obtain an augmented blasting dataset includes: The input data in the initial blasting dataset is normalized to obtain the processed data; The initial generative adversarial network is trained based on the processed data and the output data in the initial explosion dataset to obtain the target generative adversarial network; Based on the target generative adversarial network, a new bombing dataset is generated; The initial blasting dataset and the new blasting dataset are determined as the enhanced blasting dataset.
[0008] In one embodiment of the present invention, the initial generative adversarial network includes a generator and a discriminator; training the initial generative adversarial network based on the processed data and the output data in the initial bombardment dataset to obtain the target generative adversarial network includes: Fix the generator, and generate a first prediction sample based on the generator; A first real sample is randomly selected from the processed data, and the first real sample is mixed with the first predicted sample to obtain a first target sample. The first target sample is then labeled. The first target sample is input into the discriminator to obtain a first prediction result. Based on the first prediction result and the true result of the label, a first loss value is obtained through a first loss function, wherein the first loss function is:
[0009] in, The first loss value, For H x The expected value of (x), The true result is from the first real sample. The first prediction result generated by the discriminator. H is the first predicted sample generated by the generator. x (x) represents the relationships between the data; The parameters of the discriminator are updated based on the first loss value to obtain the updated discriminator; With the discriminator fixed, a second predicted sample is generated based on the generator; The second predicted sample is input into the updated discriminator to obtain the second prediction result; Based on the second prediction result, a second loss value is obtained through a second loss function, wherein the second loss function is:
[0010] in, This is the second loss value. For H z The mathematical expectation of (z), The second prediction result generated by the discriminator. The second predicted sample generated by the generator; The parameters of the generator are updated based on the second loss value to obtain the updated generator; Repeat the above steps until the preset number of iterations is reached or the network converges to obtain the target generative adversarial network.
[0011] In one embodiment of the present invention, each sample in the enhanced blasting dataset is: ,in, It is the input feature vector. This corresponds to the peak value of the blasting vibration velocity; The target blasting vibration peak prediction model includes The decision tree, the target blasting vibration peak prediction model for the first... Predicted values for each sample Represented as: ,in, This represents the sample mapping relationship calculated by the j-th decision tree; The first objective function during training for: ,in, Let D be the number of samples in the enhanced bombardment dataset, and D be the loss function. This is a regularization term.
[0012] In one embodiment of the present invention, the method further includes: The decision tree is generated iteratively, at the 1st... In the next iteration, generate The peak value of the blasting vibration velocity for the i-th sample after the trees are: Based on the first objective function, the second objective function is obtained as follows:
[0013] use Find the Taylor second-order expansion at the point that makes Minimize the objective function, remove the constant term, and optimize the loss function term. The second objective function is then optimized into a third objective function:
[0014] in, The first derivative, , It is the second derivative. .
[0015] In one embodiment of the present invention, the method further includes: The regularization term The calculation formula is:
[0016] in, The number of leaf nodes in the decision tree. Let be the score of the j-th leaf node. and These are the preset regularization parameters; Based on the regularization term, the fourth objective function is obtained by optimizing the third objective function:
[0017] in, , , This represents the set of sample indices that are assigned to leaf node j; Get the best score of leaf node j Substituting the fourth objective function, we obtain the fifth objective function, which is:
[0018] Wherein, the optimal score of the leaf node j .
[0019] To achieve the above objectives, another aspect of the present invention provides a device for predicting blasting vibration velocity, the device comprising: The data augmentation module is used to acquire the initial blasting dataset and perform data augmentation on the initial blasting dataset to obtain the augmented blasting dataset. The training module is used to train a target blasting vibration velocity peak prediction model based on the enhanced blasting dataset using the XGBoost algorithm. The acquisition module is used to acquire the blasting input data that needs to be analyzed; The prediction module is used to input the blasting input data into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak value; The data adjustment module is used to adjust the blasting input data if the peak blasting vibration velocity exceeds the blasting vibration threshold.
[0020] The blasting vibration velocity prediction method and apparatus of this invention involve acquiring an initial blasting dataset and performing data augmentation on it to obtain an augmented blasting dataset; training a target blasting vibration velocity peak prediction model using the XGBoost algorithm based on the augmented blasting dataset; acquiring the blasting input data to be analyzed; inputting the blasting input data into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak; and adjusting the blasting input data if the blasting vibration velocity peak exceeds a blasting vibration threshold. This invention achieves deep fusion and accurate modeling of multi-dimensional influencing factors of blasting vibration in open-pit mines by training a target blasting vibration velocity peak prediction model using an augmented blasting dataset and the XGBoost algorithm, overcoming the limitations of small sample data and thus improving prediction accuracy and model robustness.
[0021] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0022] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for predicting blasting vibration velocity according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating a method for predicting blasting vibration velocity according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a blasting vibration velocity prediction device according to an embodiment of the present invention. Detailed Implementation
[0023] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0025] The following description, with reference to the accompanying drawings, describes the blasting vibration velocity prediction method and apparatus proposed according to embodiments of the present invention.
[0026] Figure 1This is a flowchart illustrating the blasting vibration velocity prediction method according to an embodiment of the present invention.
[0027] like Figure 1 As shown, the method may include the following steps: Step 101: Obtain the initial blasting dataset and perform data augmentation on the initial blasting dataset to obtain the augmented blasting dataset.
[0028] In one embodiment of the present invention, the aforementioned initial blasting dataset may include input data such as row spacing, step height, blockage length, bottom hole depth, borehole diameter, explosive consumption, charge amount, minimum resistance line, and blast center distance for each blasting scheme, as well as output data such as peak blasting vibration velocity.
[0029] In one embodiment of the present invention, the row spacing, step height, plugging length, bottom hole depth, borehole diameter, explosive consumption per unit volume, charge quantity, and minimum resistance line in the aforementioned input data can be obtained from the blasting design scheme. The determining factors depend on the blasting safety regulations of the open-pit mine and the lithology of the mine. Furthermore, in one embodiment of the present invention, since the location of the blasting area changes each time, the blast center distance in the aforementioned input data can be obtained through measurement.
[0030] Furthermore, in one embodiment of the present invention, the peak value of the blasting vibration velocity of the above-mentioned output data can be measured by multiple fixed vibration meters arranged in a row.
[0031] Furthermore, in one embodiment of the present invention, after obtaining the initial blasting dataset through the above steps, the initial blasting dataset can be augmented to obtain an augmented blasting dataset, thereby expanding the initial blasting dataset to overcome the limitations of small sample data.
[0032] Specifically, in one embodiment of the present invention, the method for augmenting the initial blasting dataset to obtain an augmented blasting dataset may include the following steps: Step 1011: Normalize the input data in the initial blasting dataset to obtain the processed data; Step 1012: Train the initial generative adversarial network based on the processed data and the output data in the initial explosion dataset to obtain the target generative adversarial network; Step 1013: Generate a new explosive dataset based on the target generative adversarial network; Step 1014: Determine the initial blasting dataset and the new blasting dataset as the enhanced blasting dataset.
[0033] In one embodiment of the present invention, the input data in the initial blasting dataset can be normalized using a normalization formula to obtain processed data. The normalization formula is as follows:
[0034] Where, x i Original value The value after normalization; x min and x max These are the minimum and maximum values corresponding to each characteristic (row spacing, step height, plugging length, bottom hole depth, borehole diameter, explosive consumption, charge amount, minimum resistance line, and detonation center distance).
[0035] In one embodiment of the present invention, normalizing the initial blasting dataset can eliminate the interference of dimensional differences on distribution learning, ensuring that all features and PPV are in the same numerical range, thus providing a unified basis for subsequent learning of mapping relationships.
[0036] Furthermore, in one embodiment of the present invention, the aforementioned initial generative adversarial network may include a generator and a discriminator. In another embodiment of the present invention, the method for training the initial generative adversarial network based on the processed data and the output data in the initial detonation dataset to obtain the target generative adversarial network may include the following steps: Step 1: Fix the generator and generate the first prediction sample based on the generator; Step 2: Randomly select the first real sample from the processed data, mix the first real sample with the first predicted sample to obtain the first target sample, and label the first target sample. Step 3: Input the first target sample into the discriminator to obtain the first prediction result, and obtain the first loss value based on the first prediction result and the true result of the label through the first loss function, wherein the first loss function is:
[0037] in, The first loss value, For H x The mathematical expectation of (x), H x ( x () represents the relationships between the data. The true result is from the first real sample. The first prediction result generated by the discriminator. The first predicted sample generated by the generator; Step 4: Update the parameters of the discriminator based on the first loss value to obtain the updated discriminator; Step 5: Fix the discriminator and generate a second predicted sample based on the generator; Step 6: Input the second predicted sample into the updated discriminator to obtain the second prediction result; Step 7: Based on the second prediction result, obtain the second loss value using the second loss function, where the second loss function is:
[0038] in, This is the second loss value. For H z The mathematical expectation of (z), The second prediction result generated by the discriminator. The second predicted sample generated by the generator; Step 8: Update the generator parameters based on the second loss value to obtain the updated generator; Step 9: Repeat the above steps until the preset number of iterations is reached or the network converges to obtain the target generative adversarial network.
[0039] In one embodiment of the present invention, the labeling of the first target sample may include: labeling the first real sample of the first target sample as 1 and labeling the first predicted sample of the first target sample as 0.
[0040] In one embodiment of the present invention, the above It is the average of the function values corresponding to the data distribution over the sample. For example, the distribution H(x) of the first true sample is averaged over the function values corresponding to the first true sample x.
[0041] In one embodiment of the present invention, the generator parameters are fixed, and the discriminator improves its ability to distinguish between real data and generated data by maximizing the first loss function. The feedback from the discriminator provides direction for the subsequent optimization of the generator, helping the generator to identify feature combinations that do not conform to the real distribution.
[0042] In one embodiment of the present invention, the discriminator parameters are fixed, and the generator adjusts its own weights by minimizing a second objective function, making the generated simulated data more difficult for the discriminator to recognize. During this process, the generator continuously learns the feature combination patterns of real data, causing the mapping relationship to gradually converge towards the real data.
[0043] In one embodiment of the present invention, the target generative adversarial network is obtained by iterating through the above two stages repeatedly until a preset number of iterations is reached or the network converges. At this point, a new bombing dataset can be generated based on the generator in the target generative adversarial network, and the initial bombing dataset and the new bombing dataset are determined as the enhanced bombing dataset, thereby enriching the sample data and overcoming the limitations of small sample data.
[0044] Step 102: Based on the enhanced blasting dataset, a target blasting vibration velocity peak prediction model is obtained by training using the XGBoost algorithm.
[0045] In one embodiment of the present invention, after obtaining the enhanced blasting dataset through the above steps, a target blasting vibration velocity peak prediction model can be obtained by training the XGBoost algorithm based on the enhanced blasting dataset.
[0046] In one embodiment of the present invention, during the training preprocessing stage, all sample values of each input feature in the enhanced blasting dataset can be sorted separately, and the sample index corresponding to each feature value can be recorded; and during the model training process, the sorting results can be directly reused to determine the optimal split point of the decision tree node.
[0047] Furthermore, in one embodiment of the present invention, each sample in the enhanced blasting dataset is... ,in, It is the input feature vector. The corresponding peak blasting vibration velocity; the target blasting vibration velocity peak prediction model includes Decision tree, target blasting vibration peak velocity prediction model for the first Predicted values for each sample Represented as: ,in, Represents the sample mapping relationship calculated by the j-th decision tree; the first objective function during training. for: ,in, Let D be the number of samples in the enhanced bombardment dataset, and D be the loss function. This is a regularization term.
[0048] Furthermore, in one embodiment of the present invention, the above-mentioned decision tree generation process is iterative, and the newly generated tree will fit the residual of the previous prediction of the peak value of the blasting vibration velocity. Based on this, in the first... In the next iteration, generate The peak value of the blasting vibration velocity for the i-th sample after the trees are: Based on the first objective function, the second objective function is obtained as follows:
[0049] Furthermore, in one embodiment of the present invention, utilizing Find the Taylor second-order expansion at the point that makes Minimize the objective function, remove the constant term, and optimize the loss function term. The second objective function described above is optimized into a third objective function as follows:
[0050] in, The first derivative, , It is the second derivative. .
[0051] Furthermore, in one embodiment of the present invention, the above-mentioned regularization term... The calculation formula is:
[0052] in, The number of leaf nodes in the decision tree. Let be the score of the j-th leaf node. and These are the preset regularization parameters.
[0053] Furthermore, in one embodiment of the present invention, the fourth objective function is obtained by optimizing the third objective function based on the regularization term:
[0054] in, , , This represents the set of sample indices that are assigned to leaf node j.
[0055] Furthermore, in one embodiment of the present invention, the optimal score of leaf node j is... Substituting into the fourth objective function, we obtain the fifth objective function, which is:
[0056] Among them, the optimal score of leaf node j .
[0057] Furthermore, in one embodiment of the present invention, a target blasting vibration peak prediction model can be obtained by training using the aforementioned fifth objective function via the XGBoost algorithm. The specific training process for this part can be found in existing technologies, and will not be elaborated upon here.
[0058] Step 103: Obtain the blasting input data that needs to be analyzed.
[0059] In one embodiment of the present invention, the blasting input data to be analyzed may include row spacing, step height, plugging length, bottom hole depth, borehole diameter, explosive consumption per unit, charge amount, minimum resistance line, and blast center distance.
[0060] Step 104: Input the blasting input data into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak.
[0061] In one embodiment of the present invention, after obtaining the blasting input data to be analyzed through the above steps, the blasting input data can be input into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak.
[0062] Step 105: If the peak blasting vibration velocity exceeds the blasting vibration threshold, the blasting input data is adjusted.
[0063] In one embodiment of the present invention, after obtaining the peak blasting velocity through the above steps, the peak blasting velocity can be compared with the blasting vibration threshold, and based on the comparison result, it can be determined whether to adjust the blasting input data.
[0064] Specifically, in one embodiment of the present invention, if the peak blasting velocity exceeds the blasting vibration threshold, an early warning can be issued, and personnel can be organized to redesign the blasting input data, such as by arranging damping holes, changing blasting construction parameters, and adjusting the charge quantity, to obtain adjusted blasting input data. Furthermore, in another embodiment of the present invention, the adjusted blasting input data is re-input into the target blasting vibration peak value prediction model until the peak blasting velocity does not exceed the blasting vibration threshold.
[0065] In one embodiment of the present invention, the aforementioned blasting vibration threshold can be a value specified in the blasting vibration safety standard.
[0066] Furthermore, in one embodiment of the present invention, when performing the next blast, the newly obtained data from the fixed vibration meter deployed on site can be compared with the predicted data to obtain the error between the predicted value and the measured value of the blast vibration velocity. The measured value is then imported into the dataset to repeat steps 101 to 102, so that the error value of subsequent blasts will gradually decrease as the amount of real data increases, achieving the effect that the target blast vibration velocity peak prediction model becomes better with each use, thereby improving the accuracy of the blast vibration velocity peak value predicted in subsequent blasts.
[0067] In one embodiment of the present invention, during blasting operations, blasting vibrations may damage surrounding structures. Predicting the peak blasting vibration velocity in advance using the aforementioned method and taking corresponding protective measures will help reduce vibrations, minimize disputes with nearby residents and environmental protection departments, and ensure project progress and efficiency. Furthermore, in one embodiment of the present invention, open-pit mines have long service times, so the dataset gradually expands through repeated iterations. Because it considers multi-dimensional features such as row spacing, bench height, blockage length, bottom hole depth, borehole diameter, explosive consumption per unit volume, charge quantity, minimum resistance line, and blast center distance, the sample size is sufficient, overcoming the limitations of small sample data.
[0068] The blasting vibration velocity prediction method of this invention involves acquiring an initial blasting dataset and performing data augmentation on it to obtain an augmented blasting dataset. Based on the augmented blasting dataset, a target blasting vibration velocity peak prediction model is trained using the XGBoost algorithm. The required blasting input data is then acquired and input into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak value. If the blasting vibration velocity peak value exceeds a blasting vibration threshold, the blasting input data is adjusted. This invention can obtain a target blasting vibration velocity peak prediction model through training with an augmented blasting dataset and the XGBoost algorithm, achieving deep fusion and accurate modeling of multi-dimensional influencing factors of blasting vibration in open-pit mines. This overcomes the limitations of small sample data, thereby improving prediction accuracy and model robustness.
[0069] Figure 2 This is a schematic flowchart of the blasting vibration velocity prediction method according to an embodiment of the present invention. Figure 2 As shown, the method may include: installing a vibration meter and obtaining an initial blasting dataset through data acquisition; using a generator in a target generative adversarial network to augment the initial blasting dataset to obtain an augmented blasting dataset; training a target blasting vibration peak velocity prediction model based on the augmented blasting dataset using the XGBoost algorithm; acquiring the blasting input data to be analyzed and obtaining the blasting vibration peak velocity through the target blasting vibration peak velocity prediction model; determining whether the blasting vibration peak velocity exceeds the blasting vibration safety standard value (the aforementioned blasting vibration threshold); if the blasting vibration peak velocity exceeds the blasting vibration safety standard value, adjusting the blasting input data and inputting the adjusted blasting input data into the target blasting vibration peak velocity prediction model; if the blasting vibration peak velocity does not exceed the blasting vibration safety standard value, determining the predicted and measured values corresponding to the blast; and adding the blasting input data and measured values corresponding to the blast to the initial blasting dataset.
[0070] Figure 3 This is a schematic diagram of the blasting vibration velocity prediction device according to an embodiment of the present invention.
[0071] like Figure 3 As shown, the device may include: The data augmentation module 301 is used to acquire the initial blasting dataset and perform data augmentation on the initial blasting dataset to obtain the augmented blasting dataset; Training module 302 is used to train a target blasting vibration velocity peak prediction model based on the enhanced blasting dataset using the XGBoost algorithm; The acquisition module 303 is used to acquire the blasting input data that needs to be analyzed; Prediction module 304 is used to input blasting input data into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak value; The data adjustment module 305 is used to adjust the blasting input data if the peak blasting vibration velocity exceeds the blasting vibration threshold.
[0072] In one embodiment of the present invention, the aforementioned initial blasting dataset includes input data such as row spacing, step height, blockage length, bottom hole depth, borehole diameter, explosive consumption, charge amount, minimum resistance line, and blast center distance for each blasting scheme, as well as output data such as peak blasting vibration velocity.
[0073] In one embodiment of the present invention, the data enhancement module 301 described above is specifically used for: The input data in the initial blasting dataset is normalized to obtain the processed data; The initial generative adversarial network is trained based on the processed data and the output data in the initial explosion dataset to obtain the target generative adversarial network; Generate a new bombing dataset based on a target-based generative adversarial network; The initial blasting dataset and the new blasting dataset are identified as the enhanced blasting dataset.
[0074] In one embodiment of the present invention, the initial generative adversarial network includes a generator and a discriminator; the data augmentation module 301 is further configured to: A fixed generator is used to generate the first predicted sample. The first real sample is randomly selected from the processed data, and the first real sample is mixed with the first predicted sample to obtain the first target sample. The first target sample is then labeled. The first target sample is input into the discriminator to obtain the first prediction result. Based on the first prediction result and the true result of the label, the first loss value is obtained through the first loss function, whereby the first loss function is:
[0075] in, The first loss value, For H x The expected value of (x), The true result is from the first real sample. The first prediction result generated by the discriminator. H is the first predicted sample generated by the generator. x (x) represents the relationships between the data; The parameters of the discriminator are updated based on the first loss value to obtain the updated discriminator; With a fixed discriminator, a second predicted sample is generated based on the generator. The second predicted sample is input into the updated discriminator to obtain the second prediction result; Based on the second prediction result, a second loss value is obtained through a second loss function, where the second loss function is:
[0076] in, This is the second loss value. For H z The mathematical expectation of (z), The second prediction result generated by the discriminator. The second predicted sample generated by the generator; The generator parameters are updated based on the second loss value to obtain the updated generator; Repeat the above steps until the preset number of iterations is reached or the network converges to obtain the target generative adversarial network.
[0077] In one embodiment of the present invention, each sample in the enhanced blasting dataset is: ,in, It is the input feature vector. This corresponds to the peak value of the blasting vibration velocity; The target blasting vibration peak velocity prediction model includes Decision tree, target blasting vibration peak velocity prediction model for the first Predicted values for each sample Represented as: ,in, This represents the sample mapping relationship calculated by the j-th decision tree; The first objective function during training for: ,in, Let D be the number of samples in the enhanced bombardment dataset, and D be the loss function. This is a regularization term.
[0078] In one embodiment of the present invention, the above-described apparatus is further configured to: The decision tree is generated iteratively, and at the 1st... In the next iteration, generate The peak value of the blasting vibration velocity for the i-th sample after the trees are: Based on the first objective function, the second objective function is obtained as follows:
[0079] use Find the Taylor second-order expansion at the point that makes The objective function to be minimized, after removing constant terms and optimizing the loss function, is the third objective function:
[0080] in, The first derivative, , It is the second derivative. .
[0081] In one embodiment of the present invention, the above-described apparatus is further configured to: Regularization term The calculation formula is:
[0082] in, The number of leaf nodes in the decision tree. Let be the score of the j-th leaf node. and These are the preset regularization parameters; Based on the regularization term, the fourth objective function is obtained by optimizing the third objective function:
[0083] in, , , This represents the set of sample indices that are assigned to leaf node j; Get the best score of leaf node j Substituting into the fourth objective function, we obtain the fifth objective function, which is:
[0084] Among them, the optimal score of leaf node j .
[0085] The blasting vibration velocity prediction device of this invention acquires an initial blasting dataset and performs data augmentation on it to obtain an augmented blasting dataset. Based on the augmented blasting dataset, a target blasting vibration velocity peak prediction model is trained using the XGBoost algorithm. The blasting input data to be analyzed is acquired. The blasting input data is input into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak value. If the blasting vibration velocity peak value exceeds the blasting vibration threshold, the blasting input data is adjusted. This invention can obtain a target blasting vibration velocity peak prediction model through training with the augmented blasting dataset and the XGBoost algorithm, achieving deep fusion and accurate modeling of multi-dimensional influencing factors of blasting vibration in open-pit mines. It overcomes the limitations of small sample data, thereby improving prediction accuracy and model robustness.
[0086] In this specification, the use of terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refers to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0087] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
Claims
1. A method for predicting blasting vibration velocity, characterized in that, The method includes: Obtain the initial blasting dataset and perform data augmentation on the initial blasting dataset to obtain the augmented blasting dataset; Based on the enhanced blasting dataset, a target blasting vibration velocity peak prediction model is obtained by training using the XGBoost algorithm. Obtain the blasting input data that needs to be analyzed; The blasting input data is input into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak. If the peak blasting velocity exceeds the blasting vibration threshold, the blasting input data will be adjusted.
2. The method according to claim 1, characterized in that, The initial blasting dataset includes input data such as row spacing, step height, blockage length, bottom hole depth, borehole diameter, explosive consumption, charge amount, minimum resistance line, and blast center distance for each blasting scheme, as well as output data such as peak blasting vibration velocity.
3. The method according to claim 2, characterized in that, The step of augmenting the initial blasting dataset to obtain the augmented blasting dataset includes: The input data in the initial blasting dataset is normalized to obtain the processed data; The initial generative adversarial network is trained based on the processed data and the output data in the initial explosion dataset to obtain the target generative adversarial network; Based on the target generative adversarial network, a new bombing dataset is generated; The initial blasting dataset and the new blasting dataset are determined as the enhanced blasting dataset.
4. The method according to claim 3, characterized in that, The initial generative adversarial network includes a generator and a discriminator; training the initial generative adversarial network based on the processed data and the output data in the initial bombardment dataset to obtain the target generative adversarial network includes: Fix the generator, and generate a first prediction sample based on the generator; A first real sample is randomly selected from the processed data, and the first real sample is mixed with the first predicted sample to obtain a first target sample. The first target sample is then labeled. The first target sample is input into the discriminator to obtain a first prediction result. Based on the first prediction result and the true result of the label, a first loss value is obtained through a first loss function, wherein the first loss function is: in, The first loss value, For H x ( x The mathematical expectation of ) The true result is from the first real sample. The first prediction result generated by the discriminator. H is the first predicted sample generated by the generator. x (x) represents the relationships between the data; The parameters of the discriminator are updated based on the first loss value to obtain the updated discriminator; With the discriminator fixed, a second predicted sample is generated based on the generator; The second predicted sample is input into the updated discriminator to obtain the second prediction result; Based on the second prediction result, a second loss value is obtained through a second loss function, wherein the second loss function is: in, This is the second loss value. For H z The mathematical expectation of (z), The second prediction result generated by the discriminator. The second predicted sample generated by the generator; The parameters of the generator are updated based on the second loss value to obtain the updated generator; Repeat the above steps until the preset number of iterations is reached or the network converges to obtain the target generative adversarial network.
5. The method according to claim 1, characterized in that, Each sample in the enhanced blasting dataset is ,in, It is the input feature vector. This corresponds to the peak value of the blasting vibration velocity; The target blasting vibration peak prediction model includes The decision tree, the target blasting vibration peak prediction model for the first... Predicted values for each sample Represented as: ,in, This represents the sample mapping relationship calculated by the j-th decision tree; The first objective function during training for: ,in, Let D be the number of samples in the enhanced bombardment dataset, and D be the loss function. This is a regularization term.
6. The method according to claim 5, characterized in that, The method further includes: The decision tree is generated iteratively, at the 1st... In the next iteration, generate The peak value of the blasting vibration velocity for the i-th sample after the trees are: Based on the first objective function, the second objective function is obtained as follows: use Find the Taylor second-order expansion at the location that makes Minimize the objective function, remove the constant term, and optimize the loss function term. The second objective function is then optimized into a third objective function: in, The first derivative, , It is the second derivative. .
7. The method according to claim 6, characterized in that, The method further includes: The regularization term The calculation formula is: in, The number of leaf nodes in the decision tree. Let be the score of the j-th leaf node. and These are the preset regularization parameters; Based on the regularization term, the fourth objective function is obtained by optimizing the third objective function: in, , , This represents the set of sample indices that are assigned to leaf node j; Get the best score of leaf node j Substituting the fourth objective function, we obtain the fifth objective function, which is: Wherein, the optimal score of the leaf node j .
8. A device for predicting blasting vibration velocity, characterized in that, The device includes: The data augmentation module is used to acquire the initial blasting dataset and perform data augmentation on the initial blasting dataset to obtain the augmented blasting dataset. The training module is used to train a target blasting vibration velocity peak prediction model based on the enhanced blasting dataset using the XGBoost algorithm. The acquisition module is used to acquire the blasting input data that needs to be analyzed; The prediction module is used to input the blasting input data into the target blasting vibration velocity peak prediction model to obtain the blasting vibration velocity peak value; The data adjustment module is used to adjust the blasting input data if the peak blasting vibration velocity exceeds the blasting vibration threshold.
9. An electronic device, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.
10. A computer storage medium, wherein, The computer storage medium stores computer-executable instructions; when executed by a processor, the computer-executable instructions can implement the method described in any one of claims 1-7.