Tunnel construction blasting operation safety management and control method

Abstract

The invention relates to the technical field of tunnel construction, and discloses a tunnel construction blasting operation safety management and control method, which integrates an environmental data acquisition system, and comprises a vibration sensor, a noise sensor, a gas concentration monitor and an environmental parameter sensor, and monitoring and analyzing a plurality of key parameters of a tunnel construction blasting operation area in real time. By modeling the parameters into a Markov decision process, the method can judge whether the operation state is safe or not according to real-time data, and provides an accurate control instruction accordingly. For example, blast parameters are automatically adjusted to reduce shock effects, prompt to take noise reduction measures, or activate a ventilation system to dilute harmful gases. According to the method, a traditional safety management mode depending on experience is changed, scientific and systematic decision support is achieved, the safety management level of tunnel construction blasting operation is effectively improved, accident risks caused by misjudgment or improper control are reduced, and a powerful guarantee is provided for smooth proceeding of tunnel engineering.

Description

Technical Field

[0001] The invention relates to the technical field of tunnel construction, and in particular to a method for safety management and control of blasting operations in tunnel construction. Background Art

[0002] As an important technology in civil engineering, tunnel construction blasting operations are widely used in various types of tunnel projects. Due to the high risk and complexity of blasting operations themselves, their safety management and control have always been the focus and difficulty in the construction process. At present, although the safety management and control methods of tunnel construction blasting operations can ensure the smooth progress of construction to a certain extent, the existing safety management of blasting operations mainly relies on the experience of construction personnel and lacks scientific and systematic decision-making support. This management method is often difficult to make accurate and timely judgments and controls in the face of complex and changing construction environments, resulting in poor safety management effects. For example, in tunnel blasting operations, due to the construction personnel's misjudgment of the on-site environment and the failure to adjust the blasting parameters in time, the blasting vibration was too large, causing serious damage to the surrounding rock mass and causing a collapse accident. Summary of the invention

[0003] The object of the present invention is to provide a tunnel construction blasting operation safety management and control method to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solution: a tunnel construction blasting operation safety management and control method, the method comprising:

[0005] Integrated environmental data acquisition system, which consists of vibration sensors, noise sensors, gas concentration monitors and environmental parameter sensors, and is used to monitor and collect vibration waveforms, noise levels, harmful gas concentrations, and environmental temperature and humidity parameters in the tunnel construction blasting operation area in real time;

[0006] A safety assessment model for tunnel construction blasting operations is constructed, and the blasting operation safety management process is modeled as a Markov decision process. By analyzing the real-time environmental data of tunnel construction blasting operations and the working status of construction equipment, precise control instructions are provided for blasting operations.

[0007] Preferably, the method for constructing a security assessment model is:

[0008] Define the state space, including vibration waveforms, noise levels, harmful gas concentrations, ambient temperature and humidity parameters in the tunnel construction blasting operation area, and the working status of related equipment;

[0009] Define the action space, including the actions that the control equipment may take when dealing with abnormal blasting operations, including adjusting blasting parameters, activating warning systems, and activating emergency stop devices;

[0010] Design a reward function to evaluate the results of the action based on the safety, risk reduction, equipment loss and operating efficiency indicators of the blasting operation, and give corresponding rewards or penalties;

[0011] Formulate a policy, a rule for choosing an action in a given state, optimized through learning to maximize the cumulative reward;

[0012] Use deep reinforcement learning algorithms to train security analysis models.

[0013] Preferably, the reward function is designed as follows:

[0014] Safety reward items are set to evaluate the safety of blasting operations; safety thresholds are set based on real-time monitoring data of vibration waveforms, noise levels, and harmful gas concentrations; positive rewards are given when the monitoring data is below the safety threshold; negative rewards are given when the monitoring data exceeds the safety threshold, and the absolute value of the negative reward gradually increases as the degree of exceeding increases; the calculation formula for the safety reward item is expressed as: R safety =-k 1 *(max(0,vibration-vibration_threshold)+max(0,noise-noise_threshold)+max(0,gas_concentration-gas_threshold)), where k 1 is the safety reward coefficient, vibration, noise and gas_concentration are the real-time monitoring values ​​of vibration waveform, noise level and harmful gas concentration respectively, vibration_threshold, noise_threshold and gas_threshold are the corresponding safety thresholds respectively;

[0015] Set risk reduction rewards to encourage actions to reduce risk; set risk reduction targets based on historical data; give positive rewards when the actions taken make the current risk lower than the target risk; the calculation formula for risk reduction rewards is: R risk_reduction =k 2 *(risk_current-risk_target), where k 2 is the risk reduction reward coefficient, risk_current is the current risk value, and risk_target is the target risk value; the risk value is comprehensively calculated based on the real-time monitoring data of vibration waveform, noise level and harmful gas concentration;

[0016] Set equipment loss penalty items to evaluate equipment loss. Set equipment loss thresholds based on the actual service life and maintenance costs of the equipment. When equipment loss exceeds the threshold, a negative reward is given. The calculation formula for equipment loss penalty items is: R equipment_loss =-k 3 *max(0,equipment_loss-equipment_loss_threshold), where k 3 is the equipment loss penalty coefficient, equipment_loss is the equipment loss value, and equipment_loss_threshold is the equipment loss threshold;

[0017] Set work efficiency rewards to encourage the improvement of work efficiency; set work efficiency targets based on historical data; give positive rewards when work efficiency reaches or exceeds the target; the calculation formula for work efficiency rewards is: R efficiency =k 4 *max(0,efficiency-efficiency_target), where k 4 is the operating efficiency reward coefficient, efficiency is the current operating efficiency value, and efficiency_target is the operating efficiency target value;

[0018] Through the above rewards and penalties, we get the final reward function: R total =R safety +R risk_reduction -R equipment_loss +R efficiency .

[0019] Preferably, a deep Q network DQN algorithm is used to train the security analysis model.

[0020] Preferably, the step of training the security assessment model includes:

[0021] Step 1: Design the structure of a deep Q network to represent the strategy. The input of the network is the current state, and the output is the probability distribution of each action. Use the deep Q network to parameterize the strategy, denoted as π(a|s;θ), where s represents the state, a represents the action, and θ represents the parameters of the neural network.

[0022] Step 2: Start training, further including:

[0023] Step 201: For each training iteration, randomly extract a batch of state transition samples (s, a, r, s′) from the experience replay memory, where s is the current state, a is the action taken, r is the reward, and s′ is the next state;

[0024] Step 202: According to the Bellman equation, calculate the loss function L(θ), which is L(θ) = (r + γmax a′ Q(s′,a′;θ′)-Q(s,a;θ)) 2 , where γ is the discount factor, Q(s,a;θ) is the estimate of the state and action value function under the current strategy, and θ′ is the parameter of the target network;

[0025] Step 203: Update the neural network parameter θ using the gradient descent algorithm to minimize the loss function L(θ);

[0026] Step 3: During the training process, the actual environment data and equipment working status data of the tunnel construction blasting operation are continuously collected, and the new state transition samples are added to the experience replay memory for subsequent training iterations;

[0027] Step 4: Repeat steps 2 and 3 until the preset number of training rounds is reached;

[0028] Step 5: After the training is completed, the optimized safety assessment model is obtained, which is used to select the optimal action according to the current state to maximize the cumulative reward, thereby achieving safe management and control of tunnel construction blasting operations.

[0029] Preferably, in step 203, the formula for updating the parameters of the gradient descent algorithm is:

[0030]

[0031] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; α represents the learning rate, which is used to control the step size of parameter update; It represents the gradient of the loss function J(θ) with respect to the parameter θ, which is a vector pointing to the direction in which the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0032] Preferably, after the model training is completed, the safety assessment model is tested and evaluated, and the performance of the model recommendation scheme and the traditional safety management method in terms of risk identification speed, warning accuracy, equipment loss and operating efficiency indicators is compared through a combination of simulation testing and field testing.

[0033] Preferably, the method for performing simulation testing on the safety assessment model includes: using MATLAB / Simulink simulation software to perform simulation testing on the constructed model.

[0034] Preferably, in the simulation environment, different tunnel construction blasting operation scenarios are simulated, including different geological conditions, construction environments and blasting parameters.

[0035] Preferably, a field test plan is formulated based on the results of the simulation test, including test site selection, test parameter setting, and test equipment preparation; actual tests are carried out according to the test plan at the selected tunnel construction blasting operation site. During the test, data on the risk identification speed, warning accuracy, equipment loss, and operation efficiency of the model recommendation plan and the traditional safety management method are recorded.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention models the safety management process of tunnel construction blasting operations as a Markov decision process, and uses the mathematical model to monitor and analyze the real-time status, which can quickly identify potential safety risks. By comparing the preset safety threshold, it can instantly determine whether the current status is within the safety range, providing accurate and timely information support for decision makers, enabling them to quickly make correct control instructions and effectively avoid the occurrence of safety accidents.

[0038] The present invention can automatically adjust blasting parameters, such as the amount of explosives and the arrangement of blasting holes, according to the analysis results of the safety assessment model, so as to reduce the vibration impact of the blasting operation on the surrounding rock mass. This can not only protect the stability of the tunnel structure, but also reduce the damage to the surrounding environment, reduce construction costs, and improve engineering benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A processing step diagram of a tunnel construction blasting operation safety management and control method according to the present invention;

[0040] Figure 2 This is the training flow chart of the security assessment model;

[0041] Figure 3 Design a schematic diagram for the reward function;

[0042] Figure 4 Flowchart for reward function calculation. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] See also Figure 1-4 The present invention provides a technical solution: a tunnel construction blasting operation safety management and control method, the method comprising:

[0045] Integrated environmental data acquisition system, which is mainly composed of vibration sensors, noise sensors, gas concentration monitors and environmental parameter sensors. The vibration sensor is used to monitor the vibration waveform in the tunnel construction blasting operation area to evaluate the impact of blasting on the surrounding rock mass; the noise sensor is used to monitor the noise level generated by the blasting operation in real time to ensure that it does not exceed the safety standard; the gas concentration monitor is used to monitor the concentration of harmful gases in the tunnel, such as carbon monoxide and nitrogen dioxide, to prevent excessive gas concentration from threatening the health of construction workers; the environmental parameter sensor is used to monitor the ambient temperature and humidity in the tunnel to provide necessary environmental parameter information for blasting operations.

[0046] The safety management process of tunnel construction blasting operations is modeled as a Markov decision process. The Markov decision process is a mathematical model used to describe how decision makers choose the best action to maximize expected benefits based on the current state in an uncertain environment. In this model, the state space of tunnel construction blasting operations is defined as a multivariate combination of parameters including vibration waveform, noise level, concentration of harmful gases, ambient temperature and humidity. The real-time status of tunnel construction blasting operations is monitored and analyzed using data obtained by the integrated environmental data acquisition system. By comparing the preset safety threshold, it is determined whether the current state is within the safety range.

[0047] According to the analysis results of the safety assessment model, precise control instructions are provided for blasting operations. For example, when abnormal vibration waveforms are detected, the model will automatically adjust blasting parameters, such as the amount of explosives and the layout of blasting holes, to reduce the impact of vibration on the surrounding rock mass; when the noise level is too high, the model will prompt the adoption of noise reduction measures; when the concentration of harmful gases exceeds the standard, the model will start the ventilation system to dilute the harmful gases.

[0048] This outbreak is further described below in conjunction with Examples 1 to 3:

[0049] Embodiment 1:

[0050] This embodiment is used to illustrate the steps of building and training a security assessment model, wherein the method of building a security assessment model includes:

[0051] Defining the state space: The state space is a collection of all possible states that describe the current status of the tunnel construction blasting operation. In the present invention, the state space includes the vibration waveform, noise level, harmful gas concentration, ambient temperature and humidity parameters in the tunnel construction blasting operation area, and the working status of related equipment (such as explosive initiators, ventilation systems, etc.).

[0052] Defining the action space: The action space is the set of all possible actions that the control device may take when dealing with abnormal blasting operations. In the present invention, the action space includes actions such as adjusting blasting parameters (such as explosive quantity, blasting hole layout, etc.), activating the early warning system (to notify construction personnel to take safety measures), and activating the emergency stop device (to stop blasting operations in extremely dangerous situations).

[0053] Design reward function: The reward function is used to evaluate the results of taking specific actions under specific conditions and give corresponding rewards or penalties. The present invention designs a reward function based on indicators such as the safety of blasting operations (such as whether there are accidents such as collapse and casualties), the degree of risk reduction (such as reducing the impact of vibration on the surrounding rock mass by adjusting blasting parameters), equipment loss (such as the wear of explosive detonators and ventilation systems) and operating efficiency (such as the progress of blasting operations). When the result of the action improves safety, reduces risk, reduces equipment loss or improves operating efficiency, a positive reward is given; otherwise, a negative reward (i.e., a penalty) is given.

[0054] Formulate strategy: A strategy is a rule for selecting actions under given conditions. The present invention adopts a learning optimization method to formulate a strategy to maximize the cumulative reward. Specifically, by continuously trying and adjusting actions, the strategy of selecting the best action under different conditions is learned. During the learning process, the reward function is used to evaluate the results of the actions, and the strategy is adjusted according to the evaluation results, so that the strategy gradually converges to the optimal solution.

[0055] The deep Q network DQN algorithm is used to train the security assessment model. The following describes the training steps of the security assessment model with Python code examples:

[0056] Step 1: Design the structure of a deep Q network to represent the strategy. The input of the network is the current state, and the output is the probability distribution of each action. Use the deep Q network to parameterize the strategy, denoted as π(a|s;θ), where s represents the state, a represents the action, and θ represents the parameters of the neural network. Use a multi-layer perceptron (MLP) structure, which includes an input layer, several hidden layers, and an output layer. The number of neurons in the input layer is the same as the dimension of the state space, the number of neurons in the output layer is the same as the dimension of the action space, and the number of neurons in the hidden layer and the activation function are selected according to the specific task.

[0057] Initialize the experience replay memory to store state transition samples.

[0058] Initialize the target network, which has the same structure as the current network but different parameters. The target network is used to calculate the target Q value to stabilize the training process.

[0059] Step 2: Start training, further including:

[0060] Step 201: For each training iteration, a batch of state transition samples (s, a, r, s′) are randomly extracted from the experience replay memory, where s is the current state, a is the action taken, r is the reward, and s′ is the next state.

[0061] Step 202: Calculate the Q value Q(s, a; θ) of each action a in the current state s according to the current network.

[0062] Step 203: Calculate the Q value Q(s′, a′; θ′) of each action a′ in the next state s′ according to the target network, and select the largest Q value as the target Q value.

[0063] Step 204: According to the Bellman equation, calculate the loss function L(θ), which is L(θ) = (r + γmax a′ Q(s′,a′;θ′)-Q(s,a;θ)) 2 , where γ is the discount factor, Q(s,a;θ) is the estimate of the state and action value function under the current strategy, and θ′ is the parameter of the target network.

[0064] Step 205: Use the gradient descent algorithm to update the neural network parameters θ so that the loss function L(θ) is minimized; the formula for updating the parameters of the gradient descent algorithm is:

[0065]

[0066] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; α represents the learning rate, which is used to control the step size of parameter update; It represents the gradient of the loss function J(θ) with respect to the parameter θ, which is a vector pointing to the direction in which the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0067] Step 206: Periodically copy the parameters of the current network to the target network to update the target network.

[0068] The python code example for training the model is as follows:

[0069]

[0070] Step 3: During the training process, the actual environmental data and equipment working status data of the tunnel construction blasting operation are continuously collected, and the new state transition samples are added to the experience replay memory for subsequent training iterations.

[0071] Step 4: Repeat steps 2 and 3 until the preset number of training rounds is reached.

[0072] Step 5: After training, the optimized safety assessment model is obtained. In actual application, real-time environmental data and the working status of related equipment are input into the model, and the model will output the probability distribution of each action. The optimal action is selected according to the probability distribution to maximize the cumulative reward, thereby achieving safe management and control of tunnel construction blasting operations.

[0073] Embodiment 2:

[0074] The design of the reward function aims to provide a clear optimization goal for the security assessment model by comprehensively considering multiple key factors. The following is the specific design of the reward function:

[0075] Design of safety reward items:

[0076] According to the characteristics of tunnel construction blasting operations, safety thresholds for vibration waveforms, noise levels and harmful gas concentrations are set. These thresholds are determined based on historical data, industry standards and safety regulations to ensure safety during operations.

[0077] Design a calculation formula for the safety reward item. When the real-time monitoring data is lower than the safety threshold, a positive reward is given; when the monitoring data exceeds the safety threshold, a negative reward is given, and the absolute value of the negative reward gradually increases with the degree of exceeding, in order to motivate the model to take actions to maintain the safety of the operation. The calculation formula for the safety reward item is expressed as:

[0078] R safety =-k 1 *(max(0,vibration-vibration_threshold)+max(0,noise

[0079] -noise_threshold)+max(0,gas_concentration-gas_threshold))

[0080] Among them, k 1 is the safety reward coefficient, vibration, noise and gas_concentration are the real-time monitoring values ​​of vibration waveform, noise level and harmful gas concentration respectively, and vibration_threshold, noise_threshold and gas_threshold are the corresponding safety thresholds respectively.

[0081] Design of risk reduction incentives:

[0082] According to historical data, risk reduction targets can be set. Risk values ​​can be comprehensively calculated based on real-time monitoring data such as vibration waveforms, noise levels, and harmful gas concentrations.

[0083] Design a calculation formula for the risk reduction reward item to encourage the model to take actions to reduce risk. When the action taken makes the current risk lower than the target risk, a positive reward is given. The calculation formula for the risk reduction reward item is expressed as:

[0084] R risk_reduction =k 2 *(risk_current - risk_target)

[0085] Among them, k 2 is the risk reduction reward coefficient, risk_current is the current risk value, and risk_target is the target risk value.

[0086] Design of equipment loss penalty item:

[0087] According to the actual service life and maintenance cost of the equipment, set the equipment loss threshold. Design the calculation formula of the equipment loss penalty item to evaluate the equipment loss situation. When the equipment loss exceeds the threshold, a negative reward is given. The calculation formula of the equipment loss penalty item is expressed as:

[0088] R equipment_loss =-k 3 *max(0,equipment_loss-equipment_loss_threshold)

[0089] Among them, k 3 is the equipment loss penalty coefficient, equipment_loss is the equipment loss value, and equipment_loss_threshold is the equipment loss threshold.

[0090] Design of work efficiency reward items:

[0091] According to historical data and experience, set the work efficiency target. Design the calculation formula of the work efficiency reward item to encourage the improvement of work efficiency. When the work efficiency reaches or exceeds the target, give positive rewards. The calculation formula of the work efficiency reward item is expressed as:

[0092] R efficiency =k 4 *max(0,efficiency-efficiency_target)

[0093] Among them, k 4 is the operating efficiency reward coefficient, efficiency is the current operating efficiency value, and efficiency_target is the operating efficiency target value.

[0094] Finally, the above rewards and penalties are combined to obtain the final reward function: R total=R safety +R risk_reduction -R equipment_loss +R efficiency This function comprehensively considers the safety, risk reduction, equipment loss and operation efficiency indicators of blasting operations, and provides a clear optimization goal for the safety assessment model.

[0095] Embodiment 3:

[0096] After completing the training of the security assessment model, in order to comprehensively evaluate its performance, this embodiment uses a combination of simulation testing and field testing to compare and analyze the model recommendation scheme and traditional security management methods on multiple key indicators. The following are the specific implementation steps:

[0097] S1: Simulation test

[0098] S1.1. Simulation environment and tools: MATLAB / Simulink is selected as the simulation test platform, which can provide powerful numerical computing capabilities and an intuitive graphical interface to facilitate the simulation of complex tunnel construction blasting operation scenarios.

[0099] S1.2, Model construction and parameter setting: In the Simulink environment, according to the actual situation of tunnel construction blasting operations, a simulation model including geological conditions, construction environment and blasting parameters is constructed. Specific parameters include but are not limited to rock hardness, groundwater level, blasting explosive quantity, blasthole arrangement, etc., to ensure the comprehensiveness and accuracy of the simulation test.

[0100] S1.3. Scenario simulation: Use the constructed simulation model to simulate tunnel construction blasting operation scenarios under different geological conditions (such as hard rock formations, soft rock formations, karst development areas, etc.), different construction environments (such as dry, humid, high temperature, etc.) and different blasting parameters (such as explosive type, charge amount, detonation method, etc.).

[0101] S1.4, Data collection and analysis: During the simulation test, record the key indicators of the model's recommended solution, such as risk identification speed and warning accuracy. At the same time, through comparative analysis, preliminarily evaluate the performance of the model in different scenarios to provide a basis for the formulation of subsequent field test plans.

[0102] S2: Field test

[0103] S2.1. Test site selection: Based on the simulation test results, a representative tunnel construction blasting operation site is selected as the test site. Ensure that the test site can cover a variety of geological conditions, construction environments and blasting parameters to fully verify the actual application effect of the model.

[0104] S2.2, Test parameter setting: According to the actual situation of the test site, set reasonable test parameters, including the specific time, location, type of explosives, charge amount, etc. of the blasting operation, to ensure the controllability and safety of the test process.

[0105] S2.3, Test equipment preparation: Prepare necessary test equipment, such as sensors, data acquisition instruments, surveillance cameras, etc., to monitor and record various data during the blasting operation in real time. At the same time, ensure the compatibility of the test equipment with the model recommendation scheme and traditional safety management methods.

[0106] S2.4. Actual testing: Conduct actual testing at the selected test site according to the preset test plan. During the test, run the model recommendation plan and the traditional safety management method at the same time, and record the data of the two methods in terms of risk identification speed, warning accuracy, equipment loss and operating efficiency. Ensure the fairness and objectivity of the test process.

[0107] S2.5, Data collection and analysis: After the test is completed, collect and organize the test data. Through comparative analysis, evaluate the advantages and disadvantages of the model recommendation scheme and traditional safety management methods in various indicators. Specific analysis methods include but are not limited to statistical analysis, chart display, etc., to ensure the accuracy and intuitiveness of the analysis results.

[0108] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0109] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A tunnel construction blasting operation safety management and control method, characterized in that: The method comprises: Integrated environmental data acquisition system, which consists of vibration sensors, noise sensors, gas concentration monitors and environmental parameter sensors, and is used to monitor and collect vibration waveforms, noise levels, harmful gas concentrations, and environmental temperature and humidity parameters in the tunnel construction blasting operation area in real time; A safety assessment model for tunnel construction blasting operations is constructed, and the blasting operation safety management process is modeled as a Markov decision process. By analyzing the real-time environmental data of tunnel construction blasting operations and the working status of construction equipment, precise control instructions are provided for blasting operations.

2. A tunnel construction blasting operation safety management and control method according to claim 1, characterized in that: The method of building a security assessment model is as follows: Define the state space, including vibration waveforms, noise levels, harmful gas concentrations, ambient temperature and humidity parameters in the tunnel construction blasting operation area, and the working status of related equipment; Define the action space, including the actions that the control equipment may take when dealing with abnormal blasting operations, including adjusting blasting parameters, activating warning systems, and activating emergency stop devices; Design a reward function to evaluate the results of the action based on the safety, risk reduction, equipment loss and operating efficiency indicators of the blasting operation, and give corresponding rewards or penalties; Formulate a policy, a rule for choosing an action in a given state, optimized through learning to maximize the cumulative reward; Use deep reinforcement learning algorithms to train security analysis models.

3. A tunnel construction blasting operation safety management and control method according to claim 2, characterized in that: The reward function is designed as follows: Safety reward items are set to evaluate the safety of blasting operations; safety thresholds are set based on real-time monitoring data of vibration waveforms, noise levels, and harmful gas concentrations; positive rewards are given when the monitoring data is below the safety threshold; negative rewards are given when the monitoring data exceeds the safety threshold, and the absolute value of the negative reward gradually increases as the degree of exceeding increases; the calculation formula for the safety reward item is expressed as: R safety =-k1*(max(0,vibration-vibration_threshold)+max(0,noise-noise_threshold)+max(0,gas_concentration-gas_threshold)), where k1 is the safety bonus coefficient, vibration, noise and gas_concentration are the real-time monitoring values ​​of vibration waveform, noise level and harmful gas concentration respectively, and vibration_threshold, noise_threshold and gas_threshold are the corresponding safety thresholds respectively; Set risk reduction rewards to encourage actions to reduce risk; set risk reduction targets based on historical data; give positive rewards when the actions taken make the current risk lower than the target risk; the calculation formula for risk reduction rewards is: R risk_reduction =k2*(risk_current-risk_target), where k2 is the risk reduction reward coefficient, risk_current is the current risk value, and risk_target is the target risk value; the risk value is comprehensively calculated based on the real-time monitoring data of vibration waveform, noise level and harmful gas concentration; Set equipment loss penalty items to evaluate equipment loss. Set equipment loss thresholds based on the actual service life and maintenance costs of the equipment. When equipment loss exceeds the threshold, a negative reward is given. The calculation formula for equipment loss penalty items is: R equipment_loss =-k3*max(0,equipment_loss-equipment_loss_threshold), where k3 is the equipment loss penalty coefficient, equipment_loss is the equipment loss value, and equipment_loss_threshold is the equipment loss threshold; Set work efficiency rewards to encourage the improvement of work efficiency; set work efficiency targets based on historical data; give positive rewards when work efficiency reaches or exceeds the target; the calculation formula for work efficiency rewards is: R efficiency =k4*max(0,efficiency-efficiency_target), where k4 is the operating efficiency reward coefficient, efficiency is the current operating efficiency value, and efficiency_target is the operating efficiency target value; Through the above rewards and penalties, we get the final reward function: R total =R safety +R risk_reduction -R equipment_loss +R efficiency .

4. A tunnel construction blasting operation safety management and control method according to claim 3, characterized in that: The deep Q network DQN algorithm is used to train the security analysis model.

5. A tunnel construction blasting operation safety management and control method according to claim 4, characterized in that: The steps to train a security assessment model include: Step 1: Design the structure of a deep Q network to represent the strategy. The input of the network is the current state, and the output is the probability distribution of each action. Use the deep Q network to parameterize the strategy, denoted as π(a|s;θ), where s represents the state, a represents the action, and θ represents the parameters of the neural network. Step 2: Start training, further including: Step 201: For each training iteration, randomly extract a batch of state transition samples (s, a, r, s′) from the experience replay memory, where s is the current state, a is the action taken, r is the reward, and s′ is the next state; Step 202: According to the Bellman equation, calculate the loss function L(θ), which is L(θ) = (r + γmax a′ Q(s′,a′;θ′)-Q(s,a;θ)) 2 , where γ is the discount factor, Q(s,a;θ) is the estimate of the state and action value function under the current strategy, and θ′ is the parameter of the target network; Step 203: Update the neural network parameter θ using the gradient descent algorithm to minimize the loss function L(θ); Step 3: During the training process, the actual environment data and equipment working status data of the tunnel construction blasting operation are continuously collected, and the new state transition samples are added to the experience replay memory for subsequent training iterations; Step 4: Repeat steps 2 and 3 until the preset number of training rounds is reached; Step 5: After the training is completed, the optimized safety assessment model is obtained, which is used to select the optimal action according to the current state to maximize the cumulative reward, thereby achieving safe management and control of tunnel construction blasting operations.

6. A tunnel construction blasting operation safety management and control method according to claim 5, characterized in that: In step 203, the formula for updating the parameters of the gradient descent algorithm is: Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; α represents the learning rate, which is used to control the step size of parameter update; It represents the gradient of the loss function J(θ) with respect to the parameter θ, which is a vector pointing to the direction in which the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

7. A tunnel construction blasting operation safety management and control method according to claim 6, characterized in that: After the model training is completed, the safety assessment model is tested and evaluated. By combining simulation testing and field testing, the model recommendation scheme is compared with the traditional safety management method in terms of risk identification speed, warning accuracy, equipment loss and operating efficiency indicators.

8. A tunnel construction blasting operation safety management and control method according to claim 7, characterized in that: The method of performing simulation test on the safety assessment model includes: using MATLAB / Simulink simulation software to perform simulation test on the constructed model.

9. A tunnel construction blasting operation safety management and control method according to claim 8, characterized in that: In the simulation environment, different tunnel construction blasting operation scenarios are simulated, including different geological conditions, construction environments and blasting parameters.

10. A tunnel construction blasting operation safety management and control method according to claim 9, characterized in that: According to the results of the simulation test, a field test plan was formulated, including the selection of the test site, the setting of test parameters and the preparation of test equipment; at the selected tunnel construction blasting operation site, actual tests were carried out according to the test plan. During the test, data on the risk identification speed, early warning accuracy, equipment loss and operation efficiency of the model recommendation plan and the traditional safety management method were recorded.

Images