Prediction method for mountain tunnel drilling and blasting method construction gun refilling event

Abstract

The invention provides a prediction method for a mountain tunnel drilling and blasting method construction gun refilling event, and the method comprises the steps: S1, data collection and preprocessing: building a data set for simulation prediction of the mountain tunnel drilling and blasting method construction gun refilling event, and carrying out the standardization of sample data, and enabling the standardized sample data to serve as the input of a simulation prediction model; s2, a machine learning logistic regression algorithm is utilized to construct a shot supplement event simulation prediction model; s3, training a shot filling event simulation prediction model by using the data set, and predicting the probability of the shot filling event; and S4, calculating the processing duration of the shot refilling event. The method can accurately predict the probability of the occurrence of the gun refilling event in the current construction cycle and the disposal duration after the occurrence of the gun refilling event. The construction blasting scheme can be reasonably adjusted according to the prediction result, the primary blasting effect is improved, and the gun refilling event is avoided as much as possible.

Description

Technical Field

[0001] The invention belongs to the field of construction organization information system, and in particular relates to a prediction method for supplementary blasting events in mountain tunnel drilling and blasting construction. Background Art

[0002] During the drilling and blasting construction of mountain tunnel projects, the initial blasting cannot achieve the ideal blasting effect because the surrounding rock of some operating sections is too hard, the parameters such as blasthole arrangement, charge amount, detonation sequence are set unreasonably, or the tunnel geological conditions have unfavorable geological bodies such as joints, fissures, fault fracture zones, etc., and it is necessary to add a supplementary blasting process and carry out blasting operations again on the blasted face to increase the blasting effect or achieve the blasting section required by the design.

[0003] The supplementary blasting incident is not within the scope of the construction plan and is a disturbance event in the tunnel drilling and blasting construction process. Traditional tunnel drilling and blasting construction usually draws on the experience of similar projects to formulate construction blasting plans. The lack of specific analysis and planning for the current project makes the construction blasting plan lack of specificity and makes it impossible to intervene in the supplementary blasting incident in advance, which can easily cause confusion in the succession of the cycle process and generate a large amount of abnormal construction time.

[0004] Based on the above analysis, the present invention proposes a prediction method for supplementary blasting events in mountain tunnel drilling and blasting construction. By establishing a function of the cumulative number of process cycles and supplementary blasting events, it is used to predict the probability of supplementary blasting events occurring in the current construction cycle and the duration of treatment after the occurrence. Construction personnel can reasonably adjust the construction blasting plan based on the prediction results, improve the initial blasting effect, and try to avoid the occurrence of supplementary blasting events. Summary of the invention

[0005] The present invention proposes a prediction method and system for supplementary blasting events in mountain tunnel construction by drilling and blasting method, which can predict the probability of supplementary blasting events occurring in the current construction cycle and the handling time after the occurrence.

[0006] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0007] A method for predicting supplementary blasting events in mountain tunnel construction using a drilling and blasting method, comprising:

[0008] S1. Data collection and preprocessing: Establish a data set for simulation prediction of supplementary blasting events in mountain tunnel construction by drilling and blasting method, and standardize sample data as input of simulation prediction model;

[0009] S2. Use machine learning logistic regression algorithm to build a simulation prediction model for supplementary artillery events;

[0010] S3, using the data set to train a simulation prediction model for supplementary shot events, and predicting the probability of supplementary shot events;

[0011] S4. Calculate the time required to handle the supplementary artillery event.

[0012] Furthermore, step S1 specifically includes:

[0013] S101, data description;

[0014] The cumulative number of process cycles and the cumulative number of supplementary blasting are selected as the initial variables for predicting the occurrence of supplementary blasting events in tunnel drilling and blasting construction;

[0015] S102, data preprocessing: using Min-Max standardization as preprocessing method;

[0016] S103, data set establishment; set C is the data set required for simulation prediction of supplementary blasting events in mountain tunnel drilling and blasting construction; C = {c i |i∈[1,a max ]}, where a max represents the maximum value of the cumulative number of cycles of the process when the supplementary shot event occurs, i represents the cumulative number of cycles of the process; c i =(i,t i ,d i ,r i ), t i Indicates the cumulative number of supplementary shots, d i Indicates the duration of the supplementary artillery disposal, r i Indicates whether a supplementary firing event occurs.

[0017] Furthermore, the model building in step S2 specifically includes:

[0018] S201, in the data set C (i,t i ) as data sample, set parameter structure; y i =σ(ω1i+ω2t i +β); where ω1 and ω2 are the cumulative number of process cycles i and the cumulative number of supplementary shots t, respectively. i The weight vector of , β is the bias parameter; y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; σ represents the sigmoid function;

[0019] S202, parameter initialization, initializing ω1 and ω2 to zero vectors, and initializing β to zero scalar;

[0020] S203, setting the loss function J, where y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; Represents a data sample (i,t i )’s true label;

[0021] S204. Use the improved stochastic gradient descent method for parameter learning. The model randomly selects a sample for training each time, and uses the learning rate α to adjust the parameters so that it moves along the negative gradient direction to reduce the loss until the change in the loss function is less than the set threshold.

[0022] Furthermore, step S3 specifically includes:

[0023] S301, dividing the data set into a training set, a validation set and a test set at a ratio of 3:1:1, inputting the training set into the supplementary shot event simulation prediction model for training, using the validation set for validation, continuously adjusting the model parameters, and finally selecting the model with the smallest loss value under the validation set as the final training result;

[0024] S302, calculate the total cumulative number of supplementary shots in each cycle in the training set D, D = ∑t i ; with t i The ratio of y to D is used as the judgment threshold. i If it is greater than or equal to the judgment threshold, it means that a supplementary shot event has occurred, otherwise it means that a supplementary shot event has not occurred.

[0025] Furthermore, step S4 specifically includes: establishing a time formula for handling the supplementary shot event d=γ×d x +δ×d a. / ; where d x is the duration of the supplementary shot disposal corresponding to the cumulative number of cycles x of the current process, d a. / is the arithmetic mean of all the time required for the replenishment; γ is d x The proportion weight, δ is d a. / The proportion weight of .

[0026] Another aspect of the present invention is to provide a prediction system for supplementary blasting events in mountain tunnel construction by drilling and blasting method, comprising:

[0027] Data module: data collection and preprocessing, establishing a data set for simulation prediction of supplementary blasting events in mountain tunnel construction by drilling and blasting method, and standardizing sample data as input for simulation prediction model;

[0028] Model building module: Use machine learning logistic regression algorithm to build a simulation prediction model for supplementary artillery events;

[0029] Prediction module: using the data set to train a simulation prediction model for supplementary gun events, and predicting the probability of supplementary gun events;

[0030] Duration calculation module: calculates the duration of handling the supplementary artillery event.

[0031] Furthermore, the data module includes:

[0032] Data description unit: The cumulative number of process cycles and the cumulative number of supplementary blasting are selected as the initial variables for predicting the occurrence of supplementary blasting events in tunnel drilling and blasting construction;

[0033] Data preprocessing unit: Min-Max standardization is used as the preprocessing method;

[0034] Data set establishment unit: Set C is the data set required for simulation prediction of supplementary blasting events in mountain tunnel drilling and blasting construction; C = {c i |i∈[1,a max ]}, where a max represents the maximum value of the cumulative number of cycles of the process when the supplementary shot event occurs, i represents the cumulative number of cycles of the process; c i =(i,t i ,d i ,r i ), t i Indicates the cumulative number of supplementary shots, d i Indicates the duration of the supplementary artillery disposal, r i Indicates whether a supplementary firing event occurs.

[0035] Furthermore, the model building module includes:

[0036] Parameter structure unit: based on the data set C (i, t i ) as data sample, set parameter structure; y i =σ(ω1i+ω2t i +β); where ω1 and ω2 are the cumulative number of process cycles i and the cumulative number of supplementary shots t, respectively. i The weight vector of , β is the bias parameter; y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; σ represents the sigmoid function;

[0037] Initialization unit: parameter initialization, initialize ω1 and ω2 to zero vectors, and initialize β to zero scalar;

[0038] Loss function unit: set the loss function J, where y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; Represents a data sample (i,t i )’s true label;

[0039] Parameter learning unit: The improved stochastic gradient descent method is used for parameter learning. The model randomly selects a sample for training each time, and uses the learning rate α to adjust the parameters so that it moves along the negative gradient direction to reduce the loss until the change in the loss function is less than the set threshold.

[0040] Furthermore, the prediction module includes:

[0041] Training unit: dividing the data set into a training set, a validation set and a test set at a ratio of 3:1:1, inputting the training set into the supplementary gun event simulation prediction model for training, using the validation set for validation, continuously adjusting the model parameters, and finally selecting the model with the smallest loss value under the validation set as the final training result;

[0042] Determination unit: Calculate the total cumulative number of supplementary shots in each cycle in the training set, D = ∑t i ; with t i The ratio of y to D is used as the judgment threshold. i If it is greater than or equal to the judgment threshold, it means that a supplementary shot event has occurred, otherwise it means that a supplementary shot event has not occurred.

[0043] Furthermore, the duration calculation module includes: establishing a time formula for handling the supplementary shot event d = γ × d x +δ×d a. / ; where d x is the duration of the supplementary shot disposal corresponding to the cumulative number of cycles x of the current process, d a. / is the arithmetic mean of all the time required for the replenishment; γ is d x The proportion weight, δ is d a. / The proportion weight of .

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The present invention can accurately predict the probability of supplementary blasting events in the current construction cycle and the duration of treatment after the occurrence by establishing a function of the cumulative number of process cycles and supplementary blasting events. Construction personnel can reasonably adjust the construction blasting plan based on the prediction results, improve the initial blasting effect, and try to avoid the occurrence of supplementary blasting events. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of the model training method of an embodiment of the present invention. DETAILED DESCRIPTION

[0047] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0048] In order to make the purpose and features of the present invention more obvious and understandable, the specific implementation methods of the present invention are further described below in conjunction with the accompanying drawings.

[0049] The method for predicting the supplementary blasting event in mountain tunnel construction by drilling and blasting method proposed in this embodiment includes the following steps:

[0050] S1 Data collection and preprocessing: Establish the data set required for simulation prediction of supplementary blasting events in mountain tunnel construction by drilling and blasting method, and standardize the sample data as the input of the simulation prediction model;

[0051] S2 uses machine learning logistic regression algorithm to build a simulation prediction model for supplementary artillery events;

[0052] S3 uses the data set to train a supplementary shot event simulation prediction model to predict the probability of the supplementary shot event;

[0053] S4 calculates the time required to handle the supplementary artillery event.

[0054] The details are as follows:

[0055] S1 Data collection and preprocessing:

[0056] Collect and prepare data related to supplementary blasting events in mountain tunnel construction using the drilling and blasting method, and establish the data set required for simulation prediction of supplementary blasting events in mountain tunnel construction using the drilling and blasting method.

[0057] (1) Data description:

[0058] Through the study of previous construction cases, it is found that under the same tunnel working conditions, the same construction team, within a certain construction mileage range, the occurrence of supplementary blasting events in the cyclic process operation is regularly distributed. In this embodiment, the cumulative number of process cycles and the cumulative number of supplementary blasting events are selected as the initial variables for predicting the occurrence of supplementary blasting events in tunnel drilling and blasting construction.

[0059] (2) Data preprocessing:

[0060] In order to reduce the negative impact of data size, characteristics, and distribution differences on model training, the present invention adopts Min-Max normalization as a preprocessing method, and the normalization formula is as follows:

[0061]

[0062] Where X 123 represents the standardized data, X represents the original data, and X mi1 Indicates the minimum value of the original data, X max Indicates the maximum value of the original data.

[0063] (3) Data set establishment:

[0064] Suppose a single data sample of the data set is represented as: c i =(i,t i ,d i ,r i ), where the cumulative number of process cycles i and the cumulative number of supplementary shots t i , treatment time d i 、Whether a supplementary gun event occurredi ; then the data set is represented as:

[0065] C={c i |i∈[1,a max ]}; a max is the maximum value of the cumulative number of cycles of the process,

[0066] Finally, the set C is the data set required for simulation prediction of supplementary blasting events in mountain tunnel construction using the drilling and blasting method.

[0067] S2 builds a simulation prediction model for supplementary gun events:

[0068] The logistic regression algorithm has a clear model, strong noise resistance, and can handle multicollinearity problems well. It is often used in binary classification problems. It maps the fitted prediction value on the Sigmoid function, converts it into probability, and realizes classification. Therefore, the basic assumption of using logistic regression for classification is that the samples are linearly separable.

[0069] Assuming that the data samples of the supplementary gun event are linearly separable, a logistic regression model is established to predict whether a supplementary gun event will occur. i ) forms an N-dimensional array, where N represents the number of training set data. incident is an N-dimensional {0,1} array, where 1 represents True, indicating that a supplementary shot event has occurred, and 0 represents False, indicating that a supplementary shot event has not occurred. The model construction process includes four steps: parameter structure, initialization process, loss function setting, and parameter learning process based on stochastic gradient descent. Specifically, it includes:

[0070] (1) Parameter structure:

[0071] The parameter structure of the logistic regression model includes the weight vector ω and the bias β. During the model training process, the model parameters ω and β are adjusted iteratively to reduce the value of the loss function and achieve the purpose of model optimization. i Sample of a supplementary shot event with two features, the model's predicted output y i It is expressed as:

[0072] y i =σ(ω1i+ω2t i +β);

[0073] Where ω1 and ω2 are the cumulative number of process cycles i and the cumulative number of supplementary shots t, respectively. i The weight vector of , β is the bias parameter; y i is the model prediction output, which indicates the probability of the occurrence of the supplementary gun event category; σ represents the sigmoid function; through the Sigmoid function, the output of the logistic regression model is mapped to a probability value between 0 and 1, which is convenient for classification judgment. σ(x) is defined as:

[0074] (2) Initialization process:

[0075] Initialize the parameters, initialize ω1 and ω2 to zero vectors, and initialize β to zero scalar.

[0076] (3) Loss function setting:

[0077] The loss function uses the cross entropy loss function. For the sample (i, t i ) and the corresponding model output y i , and its cross entropy loss function is expressed as:

[0078] where y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; Represents a data sample (i,t i )’s true label; if no supplementary shot event occurs, the value is 0; if a disturbance event occurs, the value is 1.

[0079] (4) Parameter learning based on stochastic gradient descent:

[0080] The improved stochastic gradient descent method is used for parameter learning. The model randomly selects a sample for training each time, and uses the learning rate α to adjust the parameters so that they move along the negative gradient direction to reduce the loss. The parameter update process is as follows:

[0081]

[0082] Repeat the above steps until the change in the loss function is less than the set threshold, and the training ends.

[0083] S3 supplementary gun event probability simulation prediction:

[0084] The data set of drilling and blasting construction supplementary blasting events is divided into training set, validation set and test set at a ratio of 3:1:1. The training set is input into the model, and after encoding and decoding, calculating the reconstruction error, etc., a certain round of training is completed, and the validation set is used for validation, and the model parameters are continuously adjusted. Finally, the model with the smallest loss value under the validation set is selected as the final training result.

[0085] The model training process is as follows Figure 1 shown.

[0086] The test set data is input into the trained prediction model, and the logistic regression algorithm obtains a value through linear fitting, and the probability of the occurrence of the supplementary gun event is obtained through the Sigmoid function.

[0087] Calculate the total number of cumulative supplementary shots in each cycle in the training set. Define a set of cycles as starting from 1 cumulative cycle and ending with the cumulative number of cycles when a supplementary shot event occurs. Therefore, this value is equivalent to the number of cycle groups, denoted as D, and its calculation formula is as follows:

[0088] D=∑t i ;

[0089] Take the current cumulative number of cycles as x and the corresponding cumulative number of supplementary shots in the training set t x , calculate its ratio to the number of cycle groups D as the judgment threshold under the current cumulative number of cycles.

[0090]

[0091] If y i ≥ratio, it is True, indicating that a supplementary shot event has occurred; otherwise, it is False, indicating that a supplementary shot event has not occurred.

[0092] S4 treatment time simulation prediction:

[0093] According to the current cumulative number of cycles x, take the corresponding supplementary shot disposal time d in the data set x , calculate the arithmetic mean of the processing time of all supplementary artillery events d a. / Finally, the formula for handling the replenishment event is established:

[0094] d=γ×d x +δ×d avg ;

[0095] Among them, γ = 0.8, δ = 0.2. γ and δ represent the weights of the processing time of the supplementary gun event under the current cumulative number of cycles and the arithmetic mean of the processing time of all supplementary gun events.

[0096] In this embodiment, the trained model can be evaluated by evaluation indicators, such as accuracy, precision, recall, and F1.

[0097] Table 1 is a confusion matrix. In the table, FP (False Positives) indicates false positives, i.e., samples that actually did not have a supplementary shot event are mistakenly predicted to have a supplementary shot event; TN (True Negatives) indicates true negatives, i.e., samples that actually did not have a supplementary shot event are correctly predicted to have not had a supplementary shot event; TP (True Positives) indicates true negatives, i.e., samples that actually had a supplementary shot event are correctly predicted to have had a supplementary shot event; FN (False Negatives) indicates false negatives, i.e., samples that actually had a supplementary shot event are mistakenly predicted to have not had a supplementary shot event.

[0098] Table 1 Confusion matrix

[0099]

[0100] (1) Accuracy refers to the ratio of correctly classified samples to the total number of samples, which can be expressed as:

[0101]

[0102] (2) The precision rate refers to the proportion of samples predicted to have supplementary firing events that actually have supplementary firing events, and the expression is:

[0103]

[0104] (3) The recall rate refers to the proportion of samples that are predicted to have occurred supplementary firing events among samples that actually occurred supplementary firing events. The expression is:

[0105]

[0106] (4) F1 represents the harmonic mean of precision and recall, expressed as:

[0107]

[0108] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting supplementary blasting events in mountain tunnel construction using drill and blast method, characterized in that: include: S1. Data collection and preprocessing: Establish a data set for simulation prediction of supplementary blasting events in mountain tunnel construction by drilling and blasting method, and standardize sample data as input of simulation prediction model; S2. Use machine learning logistic regression algorithm to build a simulation prediction model for supplementary artillery events; S3, using the data set to train a simulation prediction model for supplementary shot events, and predicting the probability of supplementary shot events; S4. Calculate the time required to handle the supplementary artillery event.

2. The method for predicting supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 1 is characterized in that: Step S1 specifically includes: S101, data description; The cumulative number of process cycles and the cumulative number of supplementary blasting are selected as the initial variables for predicting the occurrence of supplementary blasting events in tunnel drilling and blasting construction; S102, data preprocessing: using Min-Max standardization as preprocessing method; S103, data set establishment; set C is the data set required for simulation prediction of supplementary blasting events in mountain tunnel drilling and blasting construction; C = {c i |i∈[1,a max ]}, where a max represents the maximum value of the cumulative number of cycles of the process when the supplementary shot event occurs, i represents the cumulative number of cycles of the process; c i =(i,t i ,d i ,r i ), t i Indicates the cumulative number of supplementary shots, d i Indicates the duration of the supplementary artillery disposal, r i Indicates whether a supplementary firing event occurs.

3. The method for predicting supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 1 is characterized in that: The model construction in step S2 specifically includes: S201, taking the data set C (i,t i ) as data sample, set the parameter structure; y i =σ(ω1i+ω2t i +β); where ω1 and ω2 are the cumulative number of process cycles i and the cumulative number of supplementary shots t, respectively. i The weight vector of , β is the bias parameter; y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; σ represents the sigmoid function; S202, parameter initialization, initializing ω1, ω2 to 0 vectors, and initializing β to 0 scalar; S203, setting the loss function J, where y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; Represents a data sample (i,t i )’s true label; S204. Use the improved stochastic gradient descent method for parameter learning. The model randomly selects a sample for training each time, and uses the learning rate α to adjust the parameters so that it moves along the negative gradient direction to reduce the loss until the change in the loss function is less than the set threshold.

4. The method for predicting supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 1 is characterized in that: Step S3 specifically includes: S301, dividing the data set into a training set, a validation set and a test set at a ratio of 3:1:1, inputting the training set into the supplementary shot event simulation prediction model for training, using the validation set for validation, continuously adjusting the model parameters, and finally selecting the model with the smallest loss value under the validation set as the final training result; S302, calculate the total cumulative number of supplementary shots in each cycle in the training set D, D = ∑t i ; with t i The ratio of y to D is used as the judgment threshold. i If it is greater than or equal to the judgment threshold, it indicates that a supplementary shot event has occurred, otherwise it indicates that a supplementary shot event has not occurred.

5. The method for predicting supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 1, characterized in that: Step S4 specifically includes: establishing a time formula for handling the supplementary shot event d = γ × d x +δ×d a. / ; where d x is the duration of the supplementary shot disposal corresponding to the cumulative number of cycles x of the current process, d a. / is the arithmetic mean of all the time required for the replenishment; γ is d x The proportion weight, δ is d a. / The proportion weight of .

6. A prediction system for supplementary blasting events in mountain tunnel construction by drill and blast method, characterized in that: include: Data module: data collection and preprocessing, establishing a data set for simulation prediction of supplementary blasting events in mountain tunnel construction by drilling and blasting method, and standardizing sample data as input for simulation prediction model; Model building module: Use machine learning logistic regression algorithm to build a simulation prediction model for supplementary artillery events; Prediction module: using the data set to train a simulation prediction model for supplementary gun events, and predicting the probability of supplementary gun events; Duration calculation module: calculates the duration of handling the supplementary artillery event.

7. The prediction system for supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 6 is characterized in that: The data modules include: Data description unit: The cumulative number of process cycles and the cumulative number of supplementary blasting are selected as the initial variables for predicting the occurrence of supplementary blasting events in tunnel drilling and blasting construction; Data preprocessing unit: Min-Max standardization is used as the preprocessing method; Data set establishment unit: Set C is the data set required for simulation prediction of supplementary blasting events in mountain tunnel drilling and blasting construction; C = {c i |i∈[1,a max ]}, where a max represents the maximum value of the cumulative number of cycles of the process when the supplementary shot event occurs, i represents the cumulative number of cycles of the process; c i =(i,t i ,d i ,r i ), t i Indicates the cumulative number of supplementary shots, d i Indicates the duration of the supplementary artillery disposal, r i Indicates whether a supplementary firing event occurs.

8. The prediction system for supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 6 is characterized in that: Model building modules include: Parameter structure unit: based on the data set C (i, t i ) as data sample, set the parameter structure; y i =σ(ω1i+ω2t i +β); where ω1 and ω2 are the cumulative number of process cycles i and the cumulative number of supplementary shots t, respectively. i The weight vector of , β is the bias parameter; y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; σ represents the sigmoid function; Initialization unit: parameter initialization, initializing ω1 and ω2 to zero vectors, and initializing β to zero scalar; Loss function unit: set the loss function J, where y i is the model prediction output, indicating the probability of the occurrence of the supplementary shot event category; Represents a data sample (i,t i )’s true label; Parameter learning unit: The improved stochastic gradient descent method is used for parameter learning. The model randomly selects a sample for training each time, and uses the learning rate α to adjust the parameters so that it moves along the negative gradient direction to reduce the loss until the change in the loss function is less than the set threshold.

9. The prediction system for supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 6, characterized in that: The prediction module includes: Training unit: dividing the data set into a training set, a validation set and a test set at a ratio of 3:1:1, inputting the training set into the supplementary gun event simulation prediction model for training, using the validation set for validation, continuously adjusting the model parameters, and finally selecting the model with the smallest loss value under the validation set as the final training result; Determination unit: Calculate the total cumulative number of supplementary shots in each cycle in the training set, D = ∑t i ; with t i The ratio of y to D is used as the judgment threshold. i If it is greater than or equal to the judgment threshold, it indicates that a supplementary shot event has occurred, otherwise it indicates that a supplementary shot event has not occurred.

10. The prediction system for supplementary blasting events in mountain tunnel construction by drilling and blasting method according to claim 6, characterized in that: The duration calculation module includes: establishing the time formula for handling the supplementary gun event d = γ × d x +δ×d a. / ; where d x is the duration of the supplementary shot disposal corresponding to the cumulative number of cycles x of the current process, d a. / is the arithmetic mean of all the time required for the replenishment; γ is d x The proportion weight, δ is d a. / The proportion weight of .

Images