A prediction method and system for rock mass blasting characteristic data using artificial intelligence

Abstract

The present invention provides a method and system for predicting rock mass blasting characteristic data using artificial intelligence, which relates to the field of data processing. The method includes: generating and optimizing data based on a true blasting characteristic prediction data set to obtain an optimal generated prediction data set. Among them, the error between the generated block size distribution and the true block size distribution in the generated prediction data is calculated, and error acceptance analysis and verification are carried out in combination with the transportation parameter interval for optimization; using the true blasting characteristic prediction data set and the optimal generated prediction data set to train a rock mass blasting predictor, and inputting actual blasting basic data into the rock mass blasting predictor to output predicted rock mass blasting characteristic data. It aims to solve the technical problem in traditional data prediction methods that due to the insufficient amount of training data for the prediction model, the accuracy and generalization of data prediction results are relatively low. It can significantly improve the diversity and reliability of training data, and greatly enhance the accuracy, generalization and robustness of data prediction.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and particularly to a method and system for predicting rock mass blasting characteristic data using artificial intelligence. Background Art

[0002] Rock mass blasting is one of the indispensable techniques in engineering such as mine exploitation and tunnel construction. Its main purpose is to break rocks through blasting to meet the subsequent requirements of operations such as exploitation, transportation, and crushing. Predicting the fragment size distribution of rock mass blasting, that is, the particle size of the blasted rocks, has important practical significance because it directly affects the efficiency and cost of subsequent operations such as loading, transportation, and crushing.

[0003] Most traditional methods for predicting rock mass blasting rely on historical data and physical models. The core problem lies in the quantity and quality of training data. The samples of rock mass blasting characteristic data collected in actual production are often very limited, resulting in large errors in the prediction of the trained model. Due to the diversity of blasting scenarios and the complexity of external factors (such as environmental changes, transportation parameters, etc.), traditional methods often struggle to comprehensively cover all possible situations, thus leading to poor accuracy and generalization of model prediction. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for predicting rock mass blasting characteristic data using artificial intelligence to solve the technical problem in traditional methods for predicting rock mass blasting characteristic data that due to insufficient training data of the prediction model, the accuracy and generalization of data prediction results are relatively low, including:

[0005] In a first aspect, the present invention provides a method for predicting rock mass blasting characteristic data using artificial intelligence, including: collecting a real blasting characteristic prediction data set for predicting rock mass blasting characteristic data, where each piece of real blasting characteristic prediction data includes real blasting basic data and real fragment size distribution; collecting the actual blasting basic data of the current blasting, and collecting the transportation parameter range of the current blasting transportation; according to the real blasting characteristic prediction data set, performing generation optimization of blasting characteristic prediction data to obtain an optimal generated blasting characteristic prediction data set, where the error between the generated fragment size distribution and the real fragment size distribution in the generated blasting characteristic prediction data is calculated, combined with the transportation parameter range for error acceptance analysis and verification, and optimization is performed; using the real blasting characteristic prediction data set and the optimal generated blasting characteristic prediction data set, training a rock mass blasting predictor using artificial intelligence, inputting the actual blasting basic data into the rock mass blasting predictor, and predicting and outputting to obtain a predicted fragment size distribution as the predicted rock mass blasting characteristic data.

[0006] Preferably, the method for predicting rock mass blasting characteristic data using artificial intelligence further includes: collecting a set of real blasting basic data from the blasting record data within a preset historical time range, where each piece of real blasting basic data includes rock mass characteristic data and blasting characteristic data; collecting the real block size distribution after blasting for each piece of real blasting basic data to obtain a set of real block size distributions, where each real block size distribution includes a plurality of proportion coefficients for a plurality of block size intervals; combining the set of real blasting basic data and the set of real block size distributions to obtain a real blasting characteristic prediction data set.

[0007] Preferably, the method for predicting rock mass blasting characteristic data using artificial intelligence further includes: collecting the actual blasting basic data for the current blasting; collecting the transportation parameter interval for the rock block transportation after the current blasting, where the transportation parameter interval includes the minimum transportation parameter and the maximum transportation parameter.

[0008] Preferably, the method for predicting rock mass blasting characteristic data using artificial intelligence further includes: randomly dividing a preset proportion of the first real blasting characteristic prediction data subset within the real blasting characteristic prediction data set, inputting it into the pre-trained blasting characteristic data generation channel to generate and output the first generated blasting characteristic prediction data set; obtaining the first generated block size distribution set within the first generated blasting characteristic prediction data set; obtaining the real block size distribution set within the real blasting characteristic prediction data set; calculating the first block size distribution error between the first generated block size distribution set and the real block size distribution set, and performing error acceptance analysis and verification in combination with the transportation parameter interval to obtain the first error acceptance; continuing to optimize the generation of blasting characteristic prediction data until convergence, to obtain the optimal generated blasting characteristic prediction data set with the maximum error acceptance.

[0009] Preferably, the method for predicting rock mass blasting characteristic data using artificial intelligence further includes: collecting a sample blasting characteristic prediction data set from the blasting data records within the historical time, and performing adjustment and generation to obtain a sample generated blasting characteristic prediction data set; based on the generative adversarial network, constructing a generator and a discriminator to obtain a blasting characteristic data generation channel; using the sample blasting characteristic prediction data set and the sample generated blasting characteristic prediction data set to perform supervised training on the blasting characteristic data generation channel until both the generator and the discriminator converge, completing the pre-training.

[0010] Preferably, the method for predicting rock mass blasting characteristic data using artificial intelligence further includes: according to the first generated fragment size distribution set and the true fragment size distribution set, extracting the mean value of the proportion coefficients of each fragment size interval in the first generated fragment size distribution set and the true fragment size distribution set, obtaining a plurality of first generated proportion coefficients and a plurality of true proportion coefficients; calculating the ratio of the absolute difference between each first generated proportion coefficient and the true proportion coefficient to the true proportion coefficient, obtaining a plurality of first fragment size error coefficients; calculating the mean value of the plurality of first fragment size error coefficients, obtaining a first average fragment size error coefficient; obtaining the median value within the transportation parameter interval, combining the minimum transportation parameter and the maximum transportation parameter within the transportation parameter interval, calculating to obtain a transportation parameter fluctuation coefficient; using 1 minus the ratio of the first average fragment size error coefficient to the transportation parameter fluctuation coefficient, obtaining a first error acceptance degree.

[0011] Preferably, the method for predicting rock mass blasting characteristic data using artificial intelligence further includes: using artificial intelligence to construct a rock mass blasting predictor; using the true blasting characteristic prediction data set and the optimal generated blasting characteristic prediction data set as supervised training data and test data, performing supervised training on the rock mass blasting predictor until the accuracy meets the prediction requirements; inputting the actual blasting basic data into the rock mass blasting predictor, and predicting and outputting to obtain a predicted fragment size distribution as the predicted rock mass blasting characteristic data.

[0012] In a second aspect, the present invention further provides a system for predicting rock mass blasting characteristic data using artificial intelligence, which is used to execute the method for predicting rock mass blasting characteristic data using artificial intelligence as described in the first aspect, including: a blasting prediction data collection module, which is used to collect a true blasting characteristic prediction data set for predicting rock mass blasting characteristic data, wherein each true blasting characteristic prediction data includes true blasting basic data and a true fragment size distribution; an actual blasting data collection module, which is used to collect the actual blasting basic data of the current blasting and collect the transportation parameter interval of the current blasting transportation; a prediction data generation and optimization module, which is used to perform generation and optimization of blasting characteristic prediction data according to the true blasting characteristic prediction data set, obtain an optimal generated blasting characteristic prediction data set, wherein calculate the error between the generated fragment size distribution and the true fragment size distribution in the generated blasting characteristic prediction data, combine the transportation parameter interval to perform error acceptance degree analysis and verification, and perform optimization; a predicted fragment size distribution output module, which is used to use the true blasting characteristic prediction data set and the optimal generated blasting characteristic prediction data set, use artificial intelligence to train a rock mass blasting predictor, input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain a predicted fragment size distribution as the predicted rock mass blasting characteristic data.

[0013] The embodiments of the present invention have the following advantages:

[0014] A true blasting characteristic prediction data set for predicting rock mass blasting characteristic data through collection, wherein each true blasting characteristic prediction data includes true blasting basic data and true block size distribution; then collect the actual blasting basic data of the current blasting, and collect the transport parameter interval of the current blasting transportation; further, according to the true blasting characteristic prediction data set, optimize the generation of blasting characteristic prediction data to obtain an optimal generated blasting characteristic prediction data set, wherein calculate the error between the generated block size distribution and the true block size distribution in the generated blasting characteristic prediction data, combine the transport parameter interval to perform error acceptance analysis and verification, and perform optimization; then use the true blasting characteristic prediction data set and the optimal generated blasting characteristic prediction data set, and use artificial intelligence to train a rock mass blasting predictor; finally, input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain a predicted block size distribution as the predicted rock mass blasting characteristic data. That is to say, by using a generative adversarial network to generate new training data based on limited true blasting data and combining the current transport parameters for optimization, the diversity and reliability of the training data can be significantly improved, thereby greatly enhancing the accuracy, generalization ability and robustness of the prediction of rock mass blasting characteristic data, and at the same time, the subsequent crushing and transportation costs can be effectively reduced. Description of the Drawings

[0015] Figure 1 It is a step flowchart of a method for predicting rock mass blasting characteristic data using artificial intelligence according to the present invention;

[0016] Figure 2 It is a structural schematic diagram of a system for predicting rock mass blasting characteristic data using artificial intelligence according to the present invention.

[0017] Description of the Reference Numerals:

[0018] Blasting prediction data collection module 11, actual blasting data collection module 12, prediction data generation and optimization module 13, predicted block size distribution output module 14. Detailed Embodiment

[0019] The present invention provides a method and system for predicting rock mass blasting characteristic data using artificial intelligence, which solves the technical problem in the traditional method for predicting rock mass blasting characteristic data that due to the insufficient amount of training data of the prediction model, the accuracy and generalization ability of the data prediction results are relatively low. By using a generative adversarial network to generate new training data based on limited true blasting data and combining the current transport parameters for optimization, the diversity and reliability of the training data can be significantly improved, thereby greatly enhancing the accuracy, generalization ability and robustness of the prediction of rock mass blasting characteristic data, and at the same time, the subsequent crushing and transportation costs can be effectively reduced.

[0020] Next, the technical solutions in the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments of the present invention. It should be understood that the present invention is not limited by the exemplary embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention. Additionally, it should be noted that for the sake of convenience of description, only the parts related to the present invention are shown in the drawings rather than all of them.

[0021] Embodiment 1. Please refer to the attached Figure 1 The present invention provides a method for predicting rock mass blasting characteristic data using artificial intelligence, which is applied to a system for predicting rock mass blasting characteristic data using artificial intelligence. The method specifically includes the following steps:

[0022] S10: Collect a real blasting characteristic prediction data set for predicting rock mass blasting characteristic data. Each piece of real blasting characteristic prediction data includes real blasting basic data and real fragment size distribution.

[0023] Furthermore, step S10 of the present invention further includes:

[0024] S11: In the blasting record data within a preset historical time range, collect a set of real blasting basic data. Each piece of real blasting basic data includes rock mass characteristic data and blasting characteristic data; S12: Collect the real fragment size distribution after blasting for each piece of real blasting basic data to obtain a set of real fragment size distributions. Each real fragment size distribution includes a plurality of proportion coefficients in a plurality of fragment size intervals; S13: Combine the set of real blasting basic data and the set of real fragment size distributions to obtain a real blasting characteristic prediction data set.

[0025] Specifically, first, obtain the blasting record data within a preset historical time range (which can be set according to the data volume, such as within the last year). Then, collect the rock mass characteristic data and the blasting characteristic data as the real blasting basic data. The rock mass characteristic data includes rock type, hardness, bedding structure, porosity, fracture distribution, etc.; the blasting characteristic data includes blasting charge amount, hole spacing, hole depth, length of explosive column, initiation method, etc.; to obtain a set of real blasting basic data by constructing multiple pieces of real blasting basic data.

[0026] Next, collect the real fragment size distribution after each real blasting basic data blasting. Herein, the fragment size distribution refers to the particle size distribution of the rock after blasting. Each real fragment size distribution includes multiple proportion coefficients for multiple fragment size intervals. For example, the fragment size intervals are 0 to 10 cm, 10 to 20 cm, 20 to 30 cm, etc.; the corresponding proportion coefficients are 30%, 25%, and 45%, that is, the proportion of the real fragment size distribution in the range of 0 to 10 cm is 30%, the proportion in the range of 10 to 20 cm is 25%, and the proportion in the range of 20 to 30 cm is 45%, so as to obtain the real fragment size distribution set. Finally, combine the real blasting basic data set and the real fragment size distribution set, that is, combine each real blasting basic data with the corresponding real fragment size distribution as a piece of real blasting feature prediction data, so as to obtain the real blasting feature prediction data set. Each record therein includes the blasting basic data (rock mass characteristic data and blasting feature data) and the corresponding real fragment size distribution (including the proportion coefficients of multiple fragment size intervals).

[0027] S20: Collect the actual blasting basic data of the current blasting, and collect the transportation parameter interval of the current blasting transportation.

[0028] Furthermore, step S20 of the present invention further includes:

[0029] S21: Collect the actual blasting basic data of the current blasting; S22: Collect the transportation parameter interval of the rock block transportation after the current blasting, wherein the transportation parameter interval includes the minimum transportation parameter and the maximum transportation parameter.

[0030] Specifically, first, collect the actual blasting basic data of the current blasting, including the actual rock mass characteristic data and the actual blasting feature data. The rock mass characteristic data includes the rock type, hardness, bedding structure, porosity, fracture distribution, compressive strength, etc. These data can help analyze the properties of the rock mass and its influence on the blasting effect; the blasting feature data includes the blasting charge amount, hole spacing, hole depth, charge column length, initiation method, blasting agent type, etc. These factors directly affect the blasting effect and the subsequent fragment size distribution.

[0031] Next, collect the transportation parameter range for the transportation of rock blocks after the current blasting. Here, the transportation parameter refers to the carrying capacity of the transportation vehicle, and the carrying capacity refers to the maximum mass or volume of rock blocks that each transportation vehicle can carry. To optimize the efficiency and cost of transportation operations, it is necessary to accurately grasp the maximum carrying capacity of each transportation vehicle. Collecting this parameter is of great significance for optimizing the subsequent blasting operation cost and transportation plan. The transportation parameter range includes the minimum transportation parameter (minimum carrying capacity) and the maximum transportation parameter (maximum carrying capacity). For example, the minimum carrying capacity is 15 tons and the maximum carrying capacity is 20 tons. The purpose of collecting the transportation parameter range is to provide a reference basis for subsequent fragment size distribution prediction, transportation efficiency evaluation, and cost control. Through these parameters, the impact of different transportation plans on cost, time, and efficiency can be evaluated under specific transportation conditions of rock blocks after blasting.

[0032] By collecting the data of the carrying capacity range of transportation vehicles after the current blasting, accurate operation parameters can be provided for the transportation of rock blocks after blasting, which helps to optimize the allocation of transportation resources, and can also reduce transportation costs and improve operation efficiency.

[0033] S30: According to the true blasting feature prediction data set, perform optimization of the generation of blasting feature prediction data to obtain the optimal generated blasting feature prediction data set. Among them, calculate the error between the generated fragment size distribution and the true fragment size distribution in the generated blasting feature prediction data, and combine the transportation parameter range to perform error acceptance analysis and verification for optimization.

[0034] Furthermore, step S30 of the present invention further includes:

[0035] S31: Randomly divide a preset proportion of the first true blasting feature prediction data subset within the true blasting feature prediction data set, and input it into the pre-trained blasting feature data generation channel to generate and output the first generated blasting feature prediction data set.

[0036] Furthermore, step S31 of the present invention further includes:

[0037] S311: Collect the sample blasting feature prediction data set from the blasting data records within the historical time, and perform adjustment and generation to obtain the sample generated blasting feature prediction data set; S312: Based on the generative adversarial network, construct a generator and a discriminator to obtain the blasting feature data generation channel; S313: Use the sample blasting feature prediction data set and the sample generated blasting feature prediction data set to perform supervised training on the blasting feature data generation channel until both the generator and the discriminator converge to complete the pre-training.

[0038] Specifically, first, set the data division ratio, which can be set according to the data volume and actual requirements, for example, 10%. Then, within the true blasting feature prediction data set, randomly select multiple true blasting feature prediction data according to the data division ratio, that is, randomly select 10% of the data to obtain the first true blasting feature prediction data subset.

[0039] Next, construct a blasting feature data generation channel. First, within the blasting data records in the historical time (such as in the recent one year), collect a sample blasting feature prediction data set. These data include the basic parameters of blasting (such as blasting charge amount, explosive type, hole diameter, depth, etc.), the crushed stone characteristics after blasting (such as block size distribution, fracture degree, blasting effect, etc.), and so on. Then, adjust and generate the sample blasting feature prediction data set, that is, combine historical data and theoretical models to generate and adjust the data to simulate different blasting scenarios and generate more samples. For example, generate blasting characteristics under different conditions through a physical simulation model (such as blasting simulation software), such as changing parameters such as explosive type, hole layout, and rock mass type to generate new blasting data; or increase the diversity of samples by introducing appropriate noise (such as random perturbations) to simulate blasting characteristics under different environments or operating conditions; obtain the sample generated blasting feature prediction data set.

[0040] Then, construct a blasting feature data generation channel based on the generative adversarial network, aiming to generate high-quality blasting feature data for the training of the blasting feature prediction model. Among them, the blasting feature data generation channel includes a generator and a discriminator. The generator generates data through structures such as multi-layer fully connected layers and convolutional layers. The goal is to generate blasting feature data similar to the real data from random noise (usually a vector). Its input is a random vector, and the generator will generate a fake data sample according to this input, and this sample should be as close as possible to the real blasting feature data. The discriminator is usually a convolutional neural network or a multi-layer perceptron. The goal is to distinguish whether the input data comes from the real data set or the fake data generated by the generator. The discriminator is a binary classification model, and outputs a probability between 0 and 1, indicating the probability that the input sample is real data.

[0041] Further, use the sample blasting feature prediction dataset and the sample generation blasting feature prediction dataset as training data to perform supervised training on the blasting feature data generation channel. During the training process of the generator, first, the generator generates a batch of data from random noise, and then sends the generated data into the discriminator. The discriminator will give the probability that the generated data is "real"; then calculate the loss function of the generator, and the goal is to make the data generated by the generator be judged as real by the discriminator; finally, update the weights of the generator so that the generator can generate more real data. During the training process of the discriminator, first sample a batch of data from the real dataset and generate a batch of fake data from random noise; then calculate the loss of the discriminator on these two sets of data and update the weights of the discriminator so that the discriminator can better distinguish between real data and generated data. The loss function of the discriminator is usually optimized through backpropagation so that it can accurately identify the difference between real data and generated data. The generator and the discriminator are alternately trained. The generator is continuously optimized to generate more real blasting feature data, and the discriminator is continuously optimized to better distinguish between real data and generated data. After multiple trainings, the generator can finally generate blasting feature data close to the real data. During the training process, the training progress is evaluated by monitoring the losses of the generator and the discriminator. When the losses of both tend to be stable and the quality of the generated data reaches the expected level, it is considered that the model has converged, and the training is stopped to obtain the trained blasting feature data generation channel.

[0042] By training the generator and the discriminator through a generative adversarial network to construct a data generation channel, a high-quality blasting feature prediction dataset can be generated. This method can expand the dataset, improve the prediction ability of the model, and can handle the problem of insufficient samples in traditional data collection methods.

[0043] S32: Obtain the first generated block size distribution set in the first generated blasting feature prediction dataset; S33: Obtain the real block size distribution set in the real blasting feature prediction dataset.

[0044] Specifically, extract multiple first generated block size distributions (including the proportion coefficients of multiple block size intervals) in the first generated blasting feature prediction dataset to construct the first generated block size distribution set; extract the real block size distribution (including the proportion coefficients of multiple block size intervals) in the real blasting feature prediction dataset to construct the real block size distribution set.

[0045] S34: Calculate the first block size distribution error between the first generated block size distribution set and the real block size distribution set, and perform error acceptance analysis and verification in combination with the transportation parameter interval to obtain the first error acceptance.

[0046] Furthermore, step S34 of the present invention further includes:

[0047] S341: Based on the first generated block size distribution set and the true block size distribution set, extract the mean of the proportion coefficients of each block size interval in the first generated block size distribution set and the true block size distribution set, obtaining a plurality of first generated proportion coefficients and a plurality of true proportion coefficients; S342: Calculate the ratio of the absolute difference between each first generated proportion coefficient and the true proportion coefficient to the true proportion coefficient, obtaining a plurality of first block size error coefficients; S343: Calculate the mean of the plurality of first block size error coefficients, obtaining a first average block size error coefficient; S344: Obtain the median value within the transportation parameter interval, and in combination with the minimum transportation parameter and the maximum transportation parameter within the transportation parameter interval, calculate to obtain a transportation parameter fluctuation coefficient; S345: Subtract the ratio of the first average block size error coefficient and the transportation parameter fluctuation coefficient from 1 to obtain a first error acceptance degree.

[0048] Specifically, first, based on the first generated block size distribution set and the true block size distribution set, calculate the mean of the proportion coefficients of each block size interval in the first generated block size distribution set. For example, assume that the proportion coefficients of a plurality of first generated block sizes in the first generated block size distribution set within the block size interval of 0 to 10 cm are 23%, 27%, 26%, and 25% respectively. Then the mean of the proportion coefficients of the first generated block size distribution set within the block size interval of 0 to 10 cm is (0.23 + 0.27 + 0.26 + 0.25) / 4 which is equal to 0.2525; and take the mean calculation result as the first generated proportion coefficient corresponding to the block size interval, obtaining a plurality of first generated proportion coefficients. On the other hand, calculate the mean of the proportion coefficients of each block size interval in the true block size distribution set, and set it as the true proportion coefficient corresponding to the block size interval, obtaining a plurality of true proportion coefficients. The proportion coefficient of each block size interval reflects the percentage of crushed stones within that interval.

[0049] Next, according to a plurality of block size intervals, calculate the ratio of the absolute difference between the first generated proportion coefficient and the true proportion coefficient of each block size interval to the true proportion coefficient respectively as the first block size error coefficient. For example, assume that within the block size interval of 0 to 10 cm, the first generated proportion coefficient is 25.25% and the true proportion coefficient is 23%. Then the first block size error coefficient is (25.25% - 23%) / 23% which is approximately equal to 9.8%, obtaining a plurality of first block size error coefficients corresponding to a plurality of block size intervals. Further calculate the mean of the plurality of first block size error coefficients to obtain a first average block size error coefficient, which reflects the overall trend of the errors of all block size intervals.

[0050] Then, obtain the median value within the range of the transportation parameters, that is, the average value of the minimum carrying capacity and the maximum carrying capacity, to obtain the average carrying capacity; then calculate the error margin between the minimum transportation parameter (minimum carrying capacity) and the average carrying capacity or the error margin between the maximum transportation parameter (maximum carrying capacity) and the average carrying capacity as the transportation parameter fluctuation coefficient. Here, the error margin is the ratio of the absolute value of the difference between the minimum transportation parameter (minimum carrying capacity) and the average carrying capacity to the average carrying capacity. For example, assume the minimum carrying capacity is 15 tons and the maximum carrying capacity is 20 tons, then the average carrying capacity is 17.5 tons, and the error margin is (20 - 17.5) / 17.5, approximately equal to 14.3%, so the transportation parameter fluctuation coefficient is 14.3%. Finally, subtract the ratio of the first average block size error coefficient to the transportation parameter fluctuation coefficient from 1 to obtain the first error acceptance degree. For example, assume the first average block size error coefficient is 9.8% and the transportation parameter fluctuation coefficient is 14.3%, then the first error acceptance degree is 1 - 0.098 / 0.143, approximately equal to 31.5%; where, if the error coefficient is small and the transportation parameter fluctuation coefficient is large, the error acceptance degree will be high, indicating that the generated data meets the expectations; if the error coefficient is large and the transportation parameter fluctuation coefficient is small, the error acceptance degree will be low, indicating that the quality of the generated data is poor and needs further optimization.

[0051] By calculating the error acceptance degree of the generated data, a more comprehensive evaluation of the quality of the generated data can be carried out to reduce the difference between the generated data and the real data.

[0052] S35: Continue to optimize the generation of the blasting characteristic prediction data until convergence, and obtain the optimal generated blasting characteristic prediction data set with the maximum error acceptance degree.

[0053] Specifically, then randomly select the real blasting characteristic prediction data again in the real blasting characteristic prediction data set according to the data division ratio (such as 10%) to obtain the second real blasting characteristic prediction data subset, and output the second generated blasting characteristic prediction data set through the blasting characteristic data generation channel, and calculate the second error acceptance degree of the second generated blasting characteristic prediction data set; continue to perform iterative generation and error acceptance degree calculation of the blasting characteristic prediction data using the same method until the predetermined selection times (such as 100 times) are met, to obtain multiple generated blasting characteristic prediction data sets and multiple error acceptance degrees. Finally, select the generated blasting characteristic prediction data set with the maximum error acceptance degree as the optimal generated blasting characteristic prediction data set. This iterative generation and optimization process gradually improves the quality of the generated data by continuously generating and evaluating multiple data sets, using the error acceptance degree as a measurement standard. Each iteration selects a subset and generates the corresponding prediction data, and judges the quality of the generated data by calculating the error acceptance degree. Finally, the data set with the maximum error acceptance degree is selected as the optimal generated data set, so as to achieve the purpose of optimizing the blasting characteristic prediction data.

[0054] S40: Using the real blasting feature prediction dataset and the optimal generated blasting feature prediction dataset, and adopting artificial intelligence to train a rock mass blasting predictor. Input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain a predicted fragment size distribution as the predicted rock mass blasting feature data.

[0055] Furthermore, step S40 of the present invention further includes:

[0056] S41: Adopting artificial intelligence to construct a rock mass blasting predictor; S42: Using the real blasting feature prediction dataset and the optimal generated blasting feature prediction dataset as supervised training data and test data to conduct supervised training on the rock mass blasting predictor until the accuracy rate meets the prediction requirements; S43: Input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain a predicted fragment size distribution as the predicted rock mass blasting feature data.

[0057] Specifically, adopting artificial intelligence to construct a rock mass blasting predictor. For example, constructing a rock mass blasting predictor based on a BP neural network, which is used to predict the characteristics (fragment size distribution) after rock mass blasting. The BP neural network is a common deep learning method widely used in pattern recognition and prediction problems. It can automatically adjust weights and biases by learning historical blasting data to minimize the prediction error. The rock mass blasting predictor includes an input layer, multiple hidden layers, and an output layer. Among them, the input data of the input layer is the blasting feature prediction data, and the output data of the output layer is the fragment size distribution.

[0058] Next, use the real blasting feature prediction dataset and the optimal generated blasting feature prediction dataset as training data, and divide them into a training set and a test set according to a predetermined ratio. Usually, the training set accounts for 80% to 90%, and the test set accounts for 10% to 20%. Further, take the blasting basic data of the blasting feature prediction data as the input and the fragmentation distribution of the blasting feature prediction data as the supervision, and use the training set and the test set to supervise the training and testing of the rock mass blasting predictor respectively. During the training process, first, use a preset network architecture (for example, input layer, hidden layer, output layer), select appropriate activation functions (such as ReLU activation function) and optimization algorithms (such as Adam optimizer); then, for each piece of data in the training set, input the blasting basic data into the neural network. The data passes through the input layer and the hidden layer and finally reaches the output layer to generate the predicted fragmentation distribution; then calculate the error between the predicted value (the output of the neural network) and the true value (the fragmentation distribution in the training set); further calculate the gradient of each layer according to the error, and pass the error from the output layer back to the input layer through the backpropagation algorithm to adjust the weights and biases of each node. The backpropagation uses the gradient descent method or its variants (such as Adam optimizer) to update the weights and biases in the network to minimize the loss function; repeat the steps of forward propagation, error calculation, and backpropagation to iteratively train the network until the loss function converges to a preset threshold or reaches the maximum number of training times. After the training is completed, use the test set to evaluate the performance of the model. Use the input data (blasting basic data) in the test set, input it into the trained neural network, and obtain the predicted fragmentation distribution after passing through the input layer and the hidden layer; compare the difference between the prediction result of the neural network on the test set and the actual result, calculate the error. The performance of the test set can be used to verify the generalization ability of the model. If the error on the test set is large, it may indicate that the model has overfitted, resulting in poor prediction performance on unseen data. At this time, the model can be improved by adjusting the network structure, increasing the regularization term, increasing the training data, etc. until the expected requirements (such as meeting the expected accuracy) are met, and at this time, the trained rock mass blasting predictor is obtained.

[0059] Finally, input the actual blasting basic data into the rock mass blasting predictor for prediction, and output the predicted fragmentation distribution as the predicted rock mass blasting feature data. By using artificial intelligence to construct the rock mass blasting predictor, the intelligent level of the prediction of rock mass blasting feature data can be significantly improved, thereby improving the data prediction efficiency and accuracy.

[0060] In summary, the method for predicting rock mass blasting feature data using artificial intelligence provided by the present invention has the following technical effects:

[0061] A true blasting feature prediction dataset for predicting rock mass blasting feature data through collection. Each true blasting feature prediction data includes true blasting basic data and true fragment size distribution. Then, collect the actual blasting basic data of the current blasting and the transport parameter range of the current blasting transportation. Further, according to the true blasting feature prediction dataset, optimize the generation of blasting feature prediction data to obtain an optimal generated blasting feature prediction dataset. Calculate the error between the generated fragment size distribution and the true fragment size distribution in the generated blasting feature prediction data, and conduct error acceptance analysis and verification in combination with the transport parameter range for optimization. Then, use the true blasting feature prediction dataset and the optimal generated blasting feature prediction dataset, and use artificial intelligence to train a rock mass blasting predictor. Finally, input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain a predicted fragment size distribution as the predicted rock mass blasting feature data. That is to say, by using a generative adversarial network to generate new training data based on limited true blasting data and optimizing in combination with the current transport parameters, the diversity and reliability of the training data can be significantly improved, thereby greatly enhancing the accuracy, generalization ability, and robustness of the prediction of rock mass blasting feature data, and at the same time effectively reducing the subsequent crushing and transportation costs.

[0062] Embodiment 2. Based on the same inventive concept as the method for predicting rock mass blasting feature data using artificial intelligence in the foregoing embodiment, the present invention also provides a system for predicting rock mass blasting feature data using artificial intelligence. Please refer to the appendix Figure 2 , including: a blasting prediction data collection module 11 for collecting a true blasting feature prediction dataset for predicting rock mass blasting feature data. Each true blasting feature prediction data includes true blasting basic data and true fragment size distribution; an actual blasting data collection module 12 for collecting the actual blasting basic data of the current blasting and the transport parameter range of the current blasting transportation; a prediction data generation optimization module 13 for optimizing the generation of blasting feature prediction data according to the true blasting feature prediction dataset to obtain an optimal generated blasting feature prediction dataset. Calculate the error between the generated fragment size distribution and the true fragment size distribution in the generated blasting feature prediction data, and conduct error acceptance analysis and verification in combination with the transport parameter range for optimization; a predicted fragment size distribution output module 14 for using the true blasting feature prediction dataset and the optimal generated blasting feature prediction dataset, using artificial intelligence to train a rock mass blasting predictor, inputting the actual blasting basic data into the rock mass blasting predictor, and predicting and outputting to obtain a predicted fragment size distribution as the predicted rock mass blasting feature data.

[0063] Furthermore, the rock mass blasting characteristic data prediction system using artificial intelligence is also used for: collecting a set of real blasting basic data from the blasting record data within a preset historical time range, where each piece of real blasting basic data includes rock mass characteristic data and blasting characteristic data; collecting the real block size distribution after blasting for each piece of real blasting basic data to obtain a set of real block size distributions, where each real block size distribution includes a plurality of proportion coefficients for a plurality of block size intervals; combining the set of real blasting basic data and the set of real block size distributions to obtain a real blasting characteristic prediction data set.

[0064] Furthermore, the rock mass blasting characteristic data prediction system using artificial intelligence is also used for: collecting the actual blasting basic data for the current blasting; collecting the transportation parameter interval for the rock blocks transportation after the current blasting, where the transportation parameter interval includes the minimum transportation parameter and the maximum transportation parameter.

[0065] Furthermore, the rock mass blasting characteristic data prediction system using artificial intelligence is also used for: randomly dividing a preset proportion of the first real blasting characteristic prediction data subset from the real blasting characteristic prediction data set, inputting it into the pre-trained blasting characteristic data generation channel to generate and output the first generated blasting characteristic prediction data set; obtaining the first generated block size distribution set within the first generated blasting characteristic prediction data set; obtaining the real block size distribution set within the real blasting characteristic prediction data set; calculating the first block size distribution error between the first generated block size distribution set and the real block size distribution set, and conducting error acceptance analysis and verification in combination with the transportation parameter interval to obtain the first error acceptance; continuing to optimize the generation of blasting characteristic prediction data until convergence, and obtaining the optimal generated blasting characteristic prediction data set with the maximum error acceptance.

[0066] Furthermore, the rock mass blasting characteristic data prediction system using artificial intelligence is also used for: collecting a sample blasting characteristic prediction data set from the blasting data records within the historical time, and adjusting and generating it to obtain a sample generated blasting characteristic prediction data set; constructing a generator and a discriminator based on the generative adversarial network to obtain a blasting characteristic data generation channel; using the sample blasting characteristic prediction data set and the sample generated blasting characteristic prediction data set to conduct supervised training on the blasting characteristic data generation channel until both the generator and the discriminator converge to complete the pre-training.

[0067] Further, the rock mass blasting characteristic data prediction system using artificial intelligence is further configured to: according to the first generated fragmentation distribution set and the true fragmentation distribution set, extract the mean of the proportion coefficients of each fragmentation interval in the first generated fragmentation distribution set and the true fragmentation distribution set, to obtain a plurality of first generated proportion coefficients and a plurality of true proportion coefficients; calculate the ratio of the absolute difference between each first generated proportion coefficient and the true proportion coefficient to the true proportion coefficient, to obtain a plurality of first fragmentation error coefficients; calculate the mean of the plurality of first fragmentation error coefficients, to obtain a first average fragmentation error coefficient; obtain the median value within the transportation parameter interval, and in combination with the minimum transportation parameter and the maximum transportation parameter within the transportation parameter interval, calculate to obtain a transportation parameter fluctuation coefficient; use 1 minus the ratio of the first average fragmentation error coefficient to the transportation parameter fluctuation coefficient, to obtain a first error acceptance degree.

[0068] Further, the rock mass blasting characteristic data prediction system using artificial intelligence is further configured to: use artificial intelligence to construct a rock mass blasting predictor; use the true blasting characteristic prediction data set and the optimal generated blasting characteristic prediction data set as supervised training data and test data, and perform supervised training on the rock mass blasting predictor until the accuracy rate meets the prediction requirements; input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain a predicted fragmentation distribution, as the predicted rock mass blasting characteristic data.

[0069] The various embodiments in this specification are described in a progressive manner, and the key points of each embodiment are the differences from other embodiments. The method and specific examples of a rock mass blasting characteristic data prediction method using artificial intelligence in the foregoing Embodiment 1 are equally applicable to the rock mass blasting characteristic data prediction system using artificial intelligence in this embodiment. Through the foregoing detailed description of a rock mass blasting characteristic data prediction method using artificial intelligence, those skilled in the art can clearly know the rock mass blasting characteristic data prediction system in this embodiment. Therefore, for the sake of brevity of the specification, it will not be elaborated herein. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0070] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0071] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A method for predicting rock mass blasting characteristic data using artificial intelligence, characterized in that, The method includes collecting a real blasting feature prediction dataset for predicting rock mass blasting feature data, where each real blasting feature prediction data includes real blasting basic data and real fragment size distribution; collecting the actual blasting basic data of the current blasting, and collecting the transportation parameter range of the current blasting transportation; According to the real blasting feature prediction dataset, perform generation optimization of blasting feature prediction data to obtain an optimal generated blasting feature prediction dataset. Among them, calculate the error between the generated fragment size distribution and the real fragment size distribution in the generated blasting feature prediction data, and combine the transportation parameter range to perform error acceptance analysis and verification. The optimization includes: Randomly divide a preset proportion of the first real blasting feature prediction data subset in the real blasting feature prediction dataset, input it into the pre-trained blasting feature data generation channel, and generate and output the first generated blasting feature prediction dataset; Obtain the first generated fragment size distribution set in the first generated blasting feature prediction dataset; Obtain the real fragment size distribution set in the real blasting feature prediction dataset; Calculate the first fragment size distribution error between the first generated fragment size distribution set and the real fragment size distribution set, and combine the transportation parameter range to perform error acceptance analysis and verification to obtain the first error acceptance, including: According to the first generated fragment size distribution set and the real fragment size distribution set, extract the mean of the proportion coefficients of each fragment size interval in the first generated fragment size distribution set and the real fragment size distribution set to obtain a plurality of first generated proportion coefficients and a plurality of real proportion coefficients; Calculate the ratio of the absolute difference between each first generated proportion coefficient and the real proportion coefficient to the real proportion coefficient to obtain a plurality of first fragment size error coefficients; Calculate the mean of the plurality of first fragment size error coefficients to obtain the first average fragment size error coefficient; Obtain the median value within the transportation parameter range, and combine the minimum transportation parameter and the maximum transportation parameter within the transportation parameter range to calculate the transportation parameter fluctuation coefficient; Subtract the ratio of the first average fragment size error coefficient and the transportation parameter fluctuation coefficient from 1 to obtain the first error acceptance; Continue to perform generation optimization of blasting feature prediction data until convergence, and obtain the optimal generated blasting feature prediction dataset with the largest error acceptance; Use the real blasting feature prediction dataset and the optimal generated blasting feature prediction dataset, and use artificial intelligence to train a rock mass blasting predictor. Input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain the predicted fragment size distribution as the predicted rock mass blasting feature data; Among them, collecting the real blasting feature prediction dataset for predicting rock mass blasting feature data includes: In the blasting record data within a preset historical time range, collect the real blasting basic data set, where each real blasting basic data includes rock mass characteristic data and blasting feature data; Collect the real fragment size distribution after blasting for each real blasting basic data to obtain the real fragment size distribution set, where each real fragment size distribution includes a plurality of proportion coefficients of a plurality of fragment size intervals; Combine the real blasting basic data set and the real fragment size distribution set to obtain the real blasting feature prediction dataset.

2. The method for predicting rock mass blasting characteristic data using artificial intelligence according to claim 1, characterized in that, Collect the actual blasting basic data of the current blasting, and collect the transportation parameter range of the current blasting transportation, including: Collect the actual blasting basic data of the current blasting; Collect the transportation parameter range of the rock block transportation after the current blasting, wherein the transportation parameter range includes the minimum transportation parameter and the maximum transportation parameter.

3. The method for predicting rock mass blasting characteristic data using artificial intelligence according to claim 1, characterized in that The pre-training steps of the blasting feature data generation channel include: In the blasting data records within the historical time, collect the sample blasting feature prediction data set, and perform adjustment and generation to obtain the sample generated blasting feature prediction data set; Based on the generative adversarial network, construct a generator and a discriminator to obtain the blasting feature data generation channel; Use the sample blasting feature prediction data set and the sample generated blasting feature prediction data set to perform supervised training on the blasting feature data generation channel until both the generator and the discriminator converge, and complete the pre-training.

4. The method for predicting rock mass blasting characteristic data using artificial intelligence according to claim 1, characterized in that Use the real blasting feature prediction data set and the optimal generated blasting feature prediction data set, and use artificial intelligence to train the rock mass blasting predictor. Input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain the predicted block size distribution as the predicted rock mass blasting feature data, including: Use artificial intelligence to construct a rock mass blasting predictor; Use the real blasting feature prediction data set and the optimal generated blasting feature prediction data set as the supervised training data and test data to perform supervised training on the rock mass blasting predictor until the accuracy meets the prediction requirements; Input the actual blasting basic data into the rock mass blasting predictor, and predict and output to obtain the predicted block size distribution as the predicted rock mass blasting feature data.

5. A prediction system for rock mass blasting characteristic data using artificial intelligence, characterized in that, The steps for implementing the method for predicting rock mass blasting feature data using artificial intelligence according to any one of claims 1 to 4 include A blasting prediction data collection module for collecting the real blasting feature prediction data set for predicting rock mass blasting feature data, wherein each real blasting feature prediction data includes real blasting basic data and real block size distribution; An actual blasting data collection module for collecting the actual blasting basic data of the current blasting and the transportation parameter range of the current blasting transportation; A prediction data generation and optimization module for generating and optimizing the blasting feature prediction data according to the real blasting feature prediction data set to obtain the optimal generated blasting feature prediction data set, wherein calculate the error between the generated block size distribution and the real block size distribution in the generated blasting feature prediction data, and perform error acceptance analysis and verification in combination with the transportation parameter range for optimization; A predicted block size distribution output module for using the real blasting feature prediction data set and the optimal generated blasting feature prediction data set, using artificial intelligence to train the rock mass blasting predictor, inputting the actual blasting basic data into the rock mass blasting predictor, and predicting and outputting to obtain the predicted block size distribution as the predicted rock mass blasting feature data.

Images