Intelligent blasting sequence control system for mixed loading explosives

Abstract

The invention discloses an intelligent blasting sequence control system for mixed loading explosives, which is provided with a data acquisition module, a data preprocessing module, a model construction module, a parameter optimization module, a multi-target control module and a data storage module to intelligently control the blasting sequence of the mixed loading explosives. Distance attenuation features, delay features, topographic features and blasting energy features of blasting data are constructed, the blasting prediction model is updated and iterated based on a neural network algorithm and historical blasting data information, and the performance of the blasting prediction model is evaluated. The number of hidden layers and the number of nodes in the blasting prediction model are updated and optimized through a control variable method, the blasting prediction model is optimized, when mixed explosive is adjusted and the blasting sequence is optimized, the optimal mixed explosive composition is searched through a search optimization algorithm, and the optimal mixed explosive composition is obtained through control signals in combination with real-time blasting data information. And the blasting sequence and delay are dynamically adjusted, so that the blasting of the mixed explosive is intelligently and accurately controlled in real time.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent blasting, and more specifically, to an intelligent blasting sequence control system for bulk explosives. Background Art

[0002] Blasting technology is widely used in fields such as mine exploitation, tunnel excavation, and building demolition. It is an efficient and economical method for rock fragmentation. The blasting effect directly affects project efficiency, cost, and safety. With the expansion of project scale and the improvement of environmental protection requirements, traditional blasting technologies face many challenges, such as vibration hazards, energy waste, and environmental impacts. In the blasting of bulk explosives, bulk explosives are composed of multiple explosive components (such as emulsion explosives, ammonium nitrate fuel oil explosives), which have the advantages of high energy, low cost, and strong adaptability. The energy release characteristics of bulk explosives are complex, and precise control of the blasting sequence and delay settings is required to achieve the best blasting effect.

[0003] Traditional blasting designs mainly rely on the experience of engineers and lack scientific basis. Blasting vibrations may cause harm to surrounding buildings, equipment, and the environment. It is difficult to precisely control the vibration speed by traditional methods. Unreasonable blasting sequences and delay settings may lead to energy waste and reduce the blasting effect. In the prior art, it is impossible to adjust the bulk explosive combination according to the real-time blasting situation to achieve the optimal intelligent blasting effect. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent blasting sequence control system for bulk explosives to solve the above problems existing in the prior art.

[0005] Specifically, the present application is as follows:

[0006] Provide an intelligent blasting sequence control system for bulk explosives, characterized in that the system includes: a data acquisition module, a data preprocessing module, a model construction module, a parameter optimization module, a multi-objective control module, and a data storage module;

[0007] The data acquisition module acquires historical blasting data information and current blasting data information. The current blasting data information includes basic data and dynamic data, and the dynamic data is collected in real time through the layout of sensors;

[0008] The data preprocessing module preprocesses the data information and constructs features to generate preprocessed blasting feature data;

[0009] The model construction module: constructs a blasting prediction model based on the blasting feature data. The blasting prediction model is updated and iterated based on the neural network algorithm and historical blasting data information. The historical blasting data information is input into the blasting prediction model for training to obtain a trained blasting prediction model. The current blasting data information is input into the trained blasting prediction model to obtain optimized control parameters, and a control parameter signal is generated and transmitted to the multi-objective control module;

[0010] The multi-objective control module: receives a control signal. The bulk explosive adjustment module adjusts the composition, charge amount, and charging structure of the explosive in real time, and the blasting sequence optimization module dynamically optimizes the blasting sequence to control the initiation sequence and delay time;

[0011] The data storage module: is connected to the data acquisition module, the data preprocessing module, the model construction module, the parameter optimization module, and the multi-objective control module, and stores the history and current blasting data information of the intelligent blasting sequence control of bulk explosives, the preprocessed blasting feature data, the output of the blasting prediction model, the model training optimization parameters, and the bulk explosive composition data.

[0012] The feature construction includes distance attenuation feature, delay feature, terrain feature, and blasting energy feature.

[0013] The model construction module includes using the preprocessed blasting feature data as the input unit of the neural network input layer. The input layer contains 4 input units, which respectively input the distance attenuation feature, delay feature, terrain feature, and blasting energy feature. The input layer node number is determined according to the unit number, the input feature parameters are read, and transmitted to the hidden layer;

[0014] The hidden layer consists of multiple neurons, uses a fully connected layer and a non-linear activation function to learn the relationship in the blasting feature parameters, and uses the preprocessed historical blasting data information to train the model;

[0015] Set a loss function and an optimization algorithm to minimize the loss function, adjust the weights and biases of the neural network, and add a Dropout layer to prevent overfitting of the data;

[0016] The output layer outputs 4 output units, namely blasting vibration velocity, safe blasting distance, effective throwing rate, and loosening coefficient, and generates a control signal to adjust the bulk explosive and optimize the blasting sequence.

[0017] The hidden layer consists of multiple neurons, uses a fully connected layer and a non-linear activation function to learn the relationship in the blasting feature parameters, and using the preprocessed historical blasting data information to train the model includes:

[0018] When the input feature parameters enter the hidden layer, first perform a linear transformation on the input feature parameters to calculate the value of the parameter entering the node, and then use a non-linear activation function to process the value entering the node to calculate the output activation value of the node. The non-linear activation function is the Tanh activation function, expressed as:

[0019]

[0020] where x represents the feature data input into the hidden layer node.

[0021] Setting the loss function and optimization algorithm to minimize the loss function, adjusting the weights and biases of the neural network, and adding a Dropout layer to prevent overfitting of the data includes:

[0022] The loss function is expressed as:

[0023]

[0024] where Loss represents the loss function, y i represents the true value in the blasting prediction model, represents the predicted value in the blasting prediction model, and n represents the number of samples in the model.

[0025] The optimization algorithm updates the parameters along the opposite direction of the gradient by calculating the gradients of the loss function with respect to the weights and biases, expressed as:

[0026]

[0027] where ω represents the weights, b represents the biases, L represents the loss function, and η represents the learning rate;

[0028] In each training iteration of the Dropout layer, all neurons participate in the prediction, and a part of the neurons are randomly discarded with probability p. By randomly discarding a part of the neurons during the training process, overfitting of the model is prevented.

[0029] Using the preprocessed historical blasting data information to train the model also includes:

[0030] If the performance of the blasting prediction model does not meet the requirements, optimize the training parameters and re-train and evaluate the model. Use the blasting data that has not participated in the training as the test set to evaluate the performance of the blasting prediction model. Adopt the method of controlling variables to update and optimize the number of hidden layers and nodes in the blasting prediction model. When the loss function used to evaluate the error between the predicted value and the actual value of the model for the test set tends to converge, the model training is completed, and an optimized blasting prediction model is obtained.

[0031] The mixed explosive adjustment module sets a blasting vibration velocity threshold. When the blasting vibration velocity is within the safety threshold, the optimal composition of the mixed explosive is searched through a search and optimization algorithm to achieve the best blasting effect;

[0032] The search and optimization algorithm first establishes a mixed explosive information matrix. The row vectors of the information matrix represent available mixed explosive combinations, the column vectors represent blasting requirements, and the information in the matrix is calculated from the loss values corresponding to the loss function of the blasting prediction model;

[0033] Based on the available mixed explosive composition and blasting requirements, the selection parameters for the optimal mixed explosive composition are calculated. The dynamic excitation factor is calculated according to the selection parameters, and the dynamic excitation factor is obtained through the operation of the mixed explosive adjustment coefficient and the selection parameters;

[0034] According to the state transition probability and under the condition of meeting the preset constraint conditions, an initial selection scheme for the mixed explosive composition is randomly generated, the search path is initialized, the loss value corresponding to the loss function of the blasting prediction model for each selection scheme is calculated, and the mixed explosive composition with the smallest loss value is recorded as the current optimal solution and stored;

[0035] Perform iterative search. Based on the state transition probability, a new selection scheme for the mixed explosive composition is determined. The state transition probability is calculated based on the information matrix and the dynamic excitation factor;

[0036] Through each round of iteration, a local search mechanism is used to optimize the current optimal solution. The global optimal solution is updated according to the local search results. When a global optimal solution better than the current optimal solution appears, the information matrix is updated and the local search times are reset. At the same time, the new global optimal solution is recorded. When the maximum number of iterations is reached, the iteration stops and the optimal mixed explosive composition scheme is output.

[0037] After receiving the control parameter signal, the blasting sequence optimization module combines real-time blasting data information, hole-to-hole delay, and row-to-row delay to allocate different preset weights, generates edge weight data representing the blasting sequence weight, and constructs a dynamic signal graph;

[0038] Based on the dynamic signal graph, preset conditions for meeting the blasting requirements are set, and the spanning tree that maximizes the sum of the edge weight data is calculated to obtain the optimal blasting topology structure;

[0039] According to the optimal blasting topology structure, the blasting sequence is dynamically adjusted, and combined with the set blasting safety threshold, the delay time between holes and rows is adjusted;

[0040] Through the control signal and combined with the real-time blasting data information, the blasting sequence and delay are dynamically adjusted, the hole-to-hole delay is optimized to reduce vibration superposition, the row-to-row delay is optimized, and the release of blasting energy is controlled.

[0041] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0042] 1. The present invention collects dynamic data during the blasting process of bulk explosives in real time through the layout of sensors, constructs and trains a blasting prediction model by combining historical blasting data information and basic blasting data information, and can accurately control the blasting sequence and delay setting in real time to achieve the best blasting effect;

[0043] 2. The present invention constructs distance attenuation characteristics, delay characteristics, terrain characteristics, and blasting energy characteristics of blasting data, which can better describe the internal relationship of blasting data, reduce data redundancy, improve the training efficiency of the blasting prediction model, reduce the sensitivity of the model to noise, and improve the generalization ability;

[0044] 3. The present invention updates and iterates the blasting prediction model based on the neural network algorithm and historical blasting data information, considers various factors affecting blasting in multiple aspects in the neural network algorithm, evaluates the performance of the blasting prediction model, and updates and optimizes the number of hidden layers and nodes in the blasting prediction model by using the control variable method to optimize the blasting prediction model and improve the accuracy of model prediction;

[0045] 4. When adjusting the bulk explosives and optimizing the blasting sequence, the optimal composition of the bulk explosives is searched through the search optimization algorithm to achieve the best blasting effect. By controlling the signal and combining the real-time blasting data information, the blasting sequence and delay are dynamically adjusted, the hole-to-hole delay is optimized to reduce vibration superposition, the row-to-row delay is optimized, and the release of blasting energy is controlled. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a schematic structural diagram of an intelligent blasting sequence control system for bulk explosives provided by an embodiment of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The present invention will be described in detail below with reference to the drawings.

[0048] As Figure 1 shown, the present invention provides an intelligent blasting sequence control system for bulk explosives, and the system includes: a data acquisition module, a data preprocessing module, a model construction module, a parameter optimization module, a multi-objective control module, and a data storage module;

[0049] The data acquisition module acquires historical blasting data information and current blasting data information, the current blasting data information includes basic data and dynamic data, and the dynamic data is collected in real time through the layout of sensors;

[0050] Specifically, the basic data includes: environmental parameters of the blasting area, rock and terrain geological parameters, and blasting design parameters. The blasting design parameters include the hole spacing, row spacing, hole depth, and minimum resistance line of the blast hole layout;

[0051] The dynamic data includes: real-time wind speed and direction, measured blasting vibration velocity, measured vibration frequency, and intensity of the air shock wave generated by the blasting at the measuring point;

[0052] The data preprocessing module preprocesses the data information and constructs features to generate preprocessed blasting feature data;

[0053] Specifically, the preprocessing includes removing outliers from the data information, interpolating missing values, standardizing, and dividing the training set and the test set;

[0054] The feature construction includes distance attenuation feature, delay feature, terrain feature, and blasting energy feature;

[0055] The distance attenuation feature is constructed according to the distance between the measuring point and the blast source, reflecting the attenuation law of blasting vibration with distance; the delay feature is constructed considering the influence of the delay between holes and the delay between rows on the vibration superposition to capture the dynamic characteristics of the vibration waveform; the terrain feature is constructed according to the height difference between the measuring point and the blast source to describe the influence of the terrain on the vibration propagation; the blasting energy feature is constructed considering the total energy released during the explosive explosion; by constructing features related to the problem, it can better meet the specific blasting requirements;

[0056] Specifically, the distance attenuation feature is

[0057]

[0058] where L represents the distance attenuation feature, q i represents the charge amount of the i-th blast hole, R i represents the distance between the i-th blast hole and the measuring point, and b represents the attenuation coefficient;

[0059] The delay feature is:

[0060]

[0061] where E(t) represents the delay feature, t i represents the delay of the i-th blast hole, δ(t - t i ) represents the delay function, and n represents the number of blast holes;

[0062] The terrain feature is:

[0063] ΔH = H K - H o

[0064] where ΔH is the elevation difference between the measurement point and the blasting source, representing the terrain feature, and H K represents the elevation of the measurement point, and H o represents the elevation of the blasting source;

[0065] The blasting energy feature is as follows:

[0066]

[0067] where θ represents the terrain slope, and f(ΔH,θ) represents the terrain correction function,

[0068] The model construction module: constructs a blasting prediction model according to the blasting characteristic data, and updates and iterates the blasting prediction model based on the neural network algorithm and the historical blasting data information; inputs the historical blasting data information into the blasting prediction model for training to obtain the trained blasting prediction model, inputs the current blasting data information into the trained blasting prediction model to obtain the optimized control parameters, and generates a control parameter signal to transmit to the multi-objective control module;

[0069] Uses the preprocessed blasting characteristic data as the input units of the input layer of the neural network. The input layer contains 4 input units, which respectively input the distance attenuation feature, the delay feature, the terrain feature, and the blasting energy feature. Determines the number of nodes in the input layer according to the number of units, reads the input feature parameters, and transmits them to the hidden layer;

[0070] The hidden layer consists of multiple neurons, uses the fully connected layer and the non-linear activation function to learn the relationship in the blasting characteristic parameters, and uses the preprocessed historical blasting data information to train the model;

[0071] Sets the loss function and the optimization algorithm to minimize the loss function, adjusts the weights and biases of the neural network, and adds a Dropout layer to prevent overfitting of the data;

[0072] The output layer outputs 4 output units, namely the blasting vibration velocity, the safe blasting distance, the effective throwing rate, and the loosening coefficient, and generates a control signal to adjust the bulk-loading explosive and optimize the blasting sequence;

[0073] The blasting vibration velocity refers to the maximum vibration velocity of the mass point when the seismic wave generated by the blasting propagates in the medium (such as rock, soil), and is expressed as:

[0074]

[0075] where PPV represents the blasting vibration velocity, and q i represents the charge amount of the i-th blast hole, and R iIt represents the distance between the i-th blast hole and the measuring point, and g, a, and c represent coefficients related to geological conditions and blasting methods;

[0076] Calculation formula for safe blasting distance:

[0077]

[0078] Among them, R s represents the safe blasting distance; q i represents the charge amount of the i-th blast hole; V represents the vibration velocity of the blasting safety point during operation; K represents the site coefficient; D represents the vertical distance from the goaf to the ground surface; α represents the coefficient related to rock characteristics; β represents the regional change influence factor;

[0079] The throwing rate refers to the percentage of the volume of flyrock that does not require secondary transfer in the directly thrown blasting area to the volume of the blasted rock and soil. The larger the throwing rate, the larger the amount of rock thrown to the blasting area, relatively reducing the operation volume of the dragline and lowering the stripping cost; the loose coefficient refers to the ratio of the loose volume after blasting to the volume of the rock before blasting, and the loose coefficient affects the loading efficiency of the dragline, the settlement height of the blasted pile, and the operation volume of auxiliary equipment;

[0080] The hidden layer is composed of multiple neurons, and the relationship in the blasting characteristic parameters is learned using a fully connected layer and a non-linear activation function. Using the preprocessed historical blasting data information to train the model includes:

[0081] When the input characteristic parameters enter the hidden layer, first perform a linear transformation on the input characteristic parameters, calculate the value of the parameter entering the node, and then use a non-linear activation function to process the value of the entering node to calculate the output activation value of the node. The non-linear activation function is the Tanh activation function, expressed as:

[0082]

[0083] Among them, x represents the characteristic data input into the hidden layer node;

[0084] Setting the loss function and optimization algorithm to minimize the loss function, adjusting the weights and biases of the neural network, and adding a Dropout layer to prevent overfitting of the data includes:

[0085] The loss function is expressed as:

[0086]

[0087] Among them, Loss represents the loss function, y i represents the true value in the blasting prediction model, represents the predicted value in the blasting prediction model, and n represents the number of samples in the model;

[0088] The optimization algorithm updates the parameters along the opposite direction of the gradient by calculating the gradients of the loss function with respect to the weights and biases, which is expressed as:

[0089]

[0090] where ω represents the weights, b represents the biases, L represents the loss function, and η represents the learning rate;

[0091] In each training iteration of the Dropout layer, all neurons participate in the prediction, and a portion of neurons are randomly discarded with probability p. By randomly discarding a portion of neurons during the training process, overfitting of the model is prevented;

[0092] The training of the model using the preprocessed historical blasting data information further includes:

[0093] If the performance of the blasting prediction model does not meet the requirements, the training parameters are optimized and the model is retrained and evaluated. The blasting data not involved in the training is used as the test set to evaluate the performance of the blasting prediction model. The number of hidden layers and the number of nodes in the blasting prediction model are updated and optimized using the method of controlling variables. When the loss function used to evaluate the error between the predicted value and the actual value of the model for the test set tends to converge, the model training is completed, and an optimized blasting prediction model is obtained;

[0094] Specifically, the performance of the blasting prediction model can be measured by metrics such as root mean square error (RMSE), mean absolute error (MAE), etc. to measure the prediction accuracy of the model,

[0095] The multi-objective control module: receives control signals, the mixed explosive adjustment module adjusts the composition, charge amount, and charge structure of the explosive in real time, and the blasting sequence optimization module dynamically optimizes the blasting sequence to control the initiation sequence and delay time;

[0096] The mixed explosive adjustment module sets a blasting vibration velocity threshold. When the blasting vibration velocity is within the safety threshold, the optimal mixed explosive composition is searched through a search optimization algorithm to achieve the best blasting effect;

[0097] The search optimization algorithm first establishes a mixed explosive information matrix. The row vectors of the information matrix represent available mixed explosive combinations, the column vectors represent blasting requirements, and the information in the matrix is calculated from the loss values corresponding to the loss function of the blasting prediction model;

[0098] Based on the available mixed explosive composition and blasting requirements, the selection parameters for the optimal mixed explosive composition are calculated, and the dynamic excitation factor is calculated according to the selection parameters. The dynamic excitation factor is obtained through the operation of the mixed explosive adjustment coefficient and the selection parameters;

[0099] Randomly generate an initial selection scheme for the composition of the bulk explosive under the condition of the state transition probability and meeting the preset constraint conditions, initialize the search path, calculate the loss value corresponding to the loss function of the blasting prediction model for each selection scheme, record the composition of the bulk explosive with the minimum loss value as the current optimal solution, and store it;

[0100] Execute iterative search, and determine a new selection scheme for the composition of the bulk explosive based on the state transition probability, where the state transition probability is calculated based on the information matrix and the dynamic excitation factor;

[0101] Through each round of iteration, adopt a local search mechanism to optimize the current optimal solution, update the global optimal solution according to the local search results. When a global optimal solution better than the current optimal solution appears, update the information matrix and reset the local search times, and at the same time record the new global optimal solution. When the maximum number of iterations is reached, stop the iteration and output the optimal composition scheme of the bulk explosive.

[0102] After receiving the control parameter signal, the blasting sequence optimization module combines the real-time blasting data information, the hole-by-hole delay, and the row-by-row delay to allocate different preset weights, generates edge weight data representing the blasting sequence weights, and constructs a dynamic signal graph;

[0103] Based on the dynamic signal graph, set the preset conditions that meet the blasting requirements, calculate the spanning tree that maximizes the sum of the edge weight data, and obtain the optimal blasting topology structure;

[0104] Dynamically adjust the blasting sequence according to the optimal blasting topology structure, and combine the set blasting safety threshold to adjust the delay time between holes and rows;

[0105] Through the control signal and combined with the real-time blasting data information, dynamically adjust the blasting sequence and delay, optimize the hole-by-hole delay, reduce the vibration superposition, optimize the row-by-row delay, and control the release of blasting energy;

[0106] The data storage module: is connected to the data acquisition module, the data preprocessing module, the model construction module, the parameter optimization module, and the multi-objective control module, and stores the history and current blasting data information, the preprocessed blasting feature data, the output of the blasting prediction model, the model training optimization parameters, and the composition data of the bulk explosive during the intelligent blasting sequence control of the bulk explosive;

[0107] In the specification provided here, a large number of specific details are described. However, it can be understood that the embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and technologies are not shown in detail so as not to obscure the understanding of this specification.

[0108] Similarly, it should be understood that, for the purpose of streamlining the present disclosure and aiding in the understanding of one or more of the various inventive aspects, in the foregoing description of the exemplary embodiments of the present invention, the various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be construed as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, the inventive aspects lie in less than all of the features of the single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate embodiment of the present invention.

[0109] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments and not others, combinations of features of different embodiments are meant to be within the scope of the present invention and form different embodiments. For example, in the following claims, any one of the claimed embodiments can be used in any combination.

Claims

1. An intelligent blasting sequence control system for mixed explosives, characterized in that: The system includes: a data acquisition module, a data preprocessing module, a model building module, a parameter optimization module, a multi-objective control module, and a data storage module; The data acquisition module acquires historical blasting data information and current blasting data information, wherein the current blasting data information includes basic data and dynamic data, and the dynamic data is collected in real time through the deployment of sensors; The data preprocessing module preprocesses the data information and performs feature construction to generate preprocessed blasting feature data; The model building module: builds a blasting prediction model according to blasting feature data, the blasting prediction model is updated and iterated based on a neural network algorithm and historical blasting data information, the historical blasting data information is input into the blasting prediction model for training, a trained blasting prediction model is obtained, the current blasting data information is input into the trained blasting prediction model, an optimized control parameter is obtained, and a control parameter signal is generated and transmitted to the multi-objective control module; The multi-objective control module receives the control signal, the mixed explosive adjustment module adjusts the composition, charge amount and charge structure of the explosives in real time, and the blasting sequence optimization module dynamically optimizes the blasting sequence and controls the detonation sequence and delay time; The data storage module is connected to the data acquisition module, the data preprocessing module, the model building module, the parameter optimization module, and the multi-objective control module, and stores the history of intelligent blasting sequence control of mixed explosives and current blasting data information, preprocessed blasting feature data, output of blasting prediction model, model training optimization parameters, and mixed explosive composition data.

2. The intelligent blasting sequence control system for mixed explosives according to claim 1 is characterized in that: The characteristic structure includes distance attenuation characteristics, time delay characteristics, terrain characteristics and blasting energy characteristics.

3. The intelligent blasting sequence control system for mixed explosives according to claim 1 is characterized in that: The model building module includes using the preprocessed blasting feature data as an input unit of the neural network input layer, the input layer includes 4 input units, respectively inputting distance attenuation features, delay features, terrain features and blasting energy features, determining the number of input layer nodes according to the number of units, reading the input feature parameters, and transmitting them to the hidden layer; The hidden layer is composed of multiple neurons, and the relationship among the burst feature parameters is learned by using a fully connected layer and a nonlinear activation function, and the model is trained using preprocessed historical burst data information; Set the loss function and optimization algorithm to minimize the loss function, adjust the weights and biases of the neural network, and add a Dropout layer to prevent data overfitting; The output layer outputs 4 output units, namely blasting vibration speed, safe blasting distance, effective throwing rate, loose coefficient, and generates control signals to adjust mixed explosives and optimize blasting sequence.

4. The intelligent blasting sequence control system for mixed explosives according to claim 3 is characterized in that: The hidden layer is composed of multiple neurons, and the relationship among the burst feature parameters is learned by using a fully connected layer and a nonlinear activation function. The model is trained using the preprocessed historical burst data information, including: When the input feature parameters enter the hidden layer, the input feature parameters are first linearly transformed to calculate the value of the parameter entering the node, and then the value entering the node is processed by a nonlinear activation function to calculate the output activation value of the node. The nonlinear activation function is a Tanh activation function, which is expressed as: Among them, x represents the feature data input into the hidden layer node.

5. The intelligent blasting sequence control system for mixed explosives according to claim 3 is characterized in that: The setting of the loss function and the optimization algorithm to minimize the loss function, adjusting the weights and biases of the neural network, and adding the Dropout layer to prevent data overfitting include: The loss function is expressed as: Among them, Loss represents the loss function, y i represents the true value in the burst prediction model, represents the predicted value in the burst prediction model, and n represents the number of samples in the model.

6. The intelligent blasting sequence control system for mixed explosives according to claim 5, characterized in that: The optimization algorithm calculates the gradient of the loss function with respect to the weights and biases and updates the parameters in the opposite direction of the gradient, which can be expressed as: Among them, ω represents weight, b represents bias, L represents loss function, and η represents learning rate; In each training iteration of the Dropout layer, all neurons participate in prediction, and a part of neurons is randomly discarded with probability p. By randomly discarding a part of neurons during the training process, the model is prevented from overfitting.

7. The intelligent blasting sequence control system for mixed explosives according to claim 3 is characterized in that: The use of the pre-processed historical blasting data information to train the model also includes: If the performance of the burst prediction model does not meet the requirements, optimize the training parameters and retrain and evaluate the model. Use the burst data that did not participate in the training as the test set to evaluate the performance of the burst prediction model. Use the control variable method to update and optimize the number of hidden layers and nodes in the burst prediction model. When the loss function used to evaluate the error between the predicted value and the actual value of the model for the test set tends to converge, the model training is completed and the optimized burst prediction model is obtained.

8. The intelligent blasting sequence control system for mixed explosives according to claim 1, characterized in that: The mixed explosive adjustment module sets a blasting vibration speed threshold. When the blasting vibration speed is within the safety threshold, the optimal mixed explosive composition is searched through a search optimization algorithm to achieve the best blasting effect. The search optimization algorithm first establishes a mixed explosive information matrix, wherein the row vectors of the information matrix represent the available mixed explosive combinations, and the column vectors represent the blasting requirements. The information in the matrix is ​​calculated by the loss value corresponding to the loss function of the blasting prediction model. Calculate the selection parameters of the optimal mixed explosive composition based on the available mixed explosive composition and the blasting requirements, and calculate the dynamic excitation factor according to the selection parameters, wherein the dynamic excitation factor is obtained by calculating the mixed explosive adjustment coefficient and the selection parameters; An initial mixed explosive composition selection scheme is randomly generated according to the state transition probability and under the condition that preset constraints are met, and a search path is initialized, and a loss value corresponding to the blasting prediction model loss function corresponding to each selection scheme is calculated, and the mixed explosive composition with the smallest loss value is recorded as the current optimal solution and stored; Performing iterative search to determine a new mixed explosive composition selection scheme based on a state transition probability, wherein the state transition probability is calculated based on the information matrix and a dynamic excitation factor; Through each round of iteration, the local search mechanism is used to optimize the current optimal solution, and the global optimal solution is updated according to the local search results. When a global optimal solution that is better than the current optimal solution appears, the information matrix is ​​updated and the number of local searches is reset. At the same time, the new global optimal solution is recorded. When the maximum number of iterations is reached, the iteration is stopped and the optimal mixed explosive composition plan is output.

9. The intelligent blasting sequence control system for mixed explosives according to claim 1, characterized in that: After receiving the control parameter signal, the blasting sequence optimization module combines the real-time blasting data information, the delay between holes, and the delay between rows to assign different preset weights, generates edge weight data representing the blasting sequence weight, and constructs a dynamic signal graph; Based on the dynamic signal graph, set the preset conditions that meet the blasting requirements, calculate the spanning tree that maximizes the sum of edge weight data, and obtain the optimal blasting topology structure; Dynamically adjust the blasting sequence according to the optimal blasting topology structure, and adjust the delay time between holes and rows in combination with the set blasting safety threshold; By controlling the signal and combining it with real-time blasting data information, the blasting sequence and delay are dynamically adjusted, the delay between holes is optimized, the vibration superposition is reduced, the delay between rows is optimized, and the release of blasting energy is controlled.

Images