A blasting safety protection method based on unstable roadway excavation

Abstract

The application discloses a kind of based on unstable roadway excavation blasting safety protection method, belong to the field of mine engineering and underground construction safety technology.The steps of the present application are:S1: multiple point strain gauge, MEMS sensor and fiber bragg grating sensor are arranged in the roadway, and strain value, stress value and bragg wavelength change data are collected in real time;S2: the data collected is cleaned, denoised and normalized;S3: use LSTM network to predict future strain value, stress value and bragg wavelength change;S4: use random forest model to classify LSTM prediction results and current monitoring data, and evaluate rock mass stability risk level;S5: according to the risk level output by the random forest model, the system automatically issues an early warning, and prompts to take corresponding physical protection measures.The present application significantly improves the safety and monitoring efficiency in the process of roadway excavation, reduces the hidden danger, and protects the safety of workers and equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mine engineering and underground construction safety, more particularly to a blasting safety protection method based on unstable roadway excavation. BACKGROUND

[0002] In recent years, with the development of sensor technology, wireless transmission technology and data processing technology, roadway monitoring schemes based on sensor networks have gradually been applied in mine safety monitoring. These systems use various sensors (such as strain gauges, accelerometers, laser scanners, etc.) for data collection, achieve real-time data transmission through wireless networks (such as ZigBee, LoRa and 5G), and perform centralized processing and analysis in the data center. Although these technologies have improved the monitoring accuracy and real-time performance to some extent, there are still some problems. First, the stability and reliability of data transmission is an important challenge. The mine environment is complex, and problems such as electromagnetic interference and signal obstruction often cause unstable wireless transmission signals, data loss and delay. Second, the existing data processing systems have limited processing capacity, making it difficult to effectively analyze and process a large amount of real-time monitoring data, resulting in lagging risk assessment results. Finally, the existing risk assessment models are often not perfect and cannot fully consider various influencing factors, resulting in insufficient risk warning accuracy. In addition, although some advanced monitoring technologies such as fiber Bragg grating sensors and MEMS sensors have been applied in some mines, their application range is still limited due to high cost, complex installation and maintenance, etc. At the same time, how to fuse and analyze the data of multiple sensors to form a comprehensive risk assessment result is also a problem to be solved. SUMMARY

[0003] The purpose of the present application is to provide a blasting safety protection method based on unstable roadway excavation, which realizes real-time monitoring of stress and displacement changes of the surrounding rock mass of the roadway by installing multi-point strain gauges, fiber Bragg grating sensors and MEMS sensors, high-precision and high-frequency data acquisition and wireless transmission using a data acquisition and transmission system, and comprehensive processing and risk assessment in the data center to solve the problems raised in the background technology.

[0004] To achieve the above purpose, the present application provides the following technical scheme:

[0005] A blasting safety protection method based on unstable roadway excavation, comprising the following steps:

[0006] S1: arranging multi-point strain gauges, MEMS sensors and fiber Bragg grating sensors in the roadway to collect strain values, stress values and Bragg wavelength change data in real time;

[0007] S2: cleaning, denoising and normalizing the collected data;

[0008] S3: Use LSTM network to predict future strain values, stress values ​​and Bragg wavelength changes;

[0009] S4: Use the random forest model to classify the LSTM prediction results and current monitoring data to assess the rock mass stability risk level;

[0010] S5: Based on the risk level output by the random forest model, the system automatically issues an early warning and prompts you to take appropriate physical protection measures.

[0011] Furthermore, in step S1, a multi-point strain gauge is installed every 5 meters on the roadway roof, one at a depth of 0.5 meters and one at a depth of 1 meter. The strain gauge reflects internal stress changes by detecting the deformation of the rock mass; a MEMS sensor is installed every 3 meters on both sides of the roadway, one at a depth of 0.5 meters and one at a depth of 1 meter. The MEMS sensor provides highly sensitive stress monitoring data by measuring tiny stress changes within the rock mass; and a fiber Bragg grating sensor is installed every 10 meters on the roadway floor, one at a depth of 1 meter and one at a depth of 2 meters. The fiber Bragg grating sensor accurately measures the stress distribution on the roadway floor and deep inside by changing the Bragg wavelength.

[0012] The strain value detected by the multi-point strain gauge is ΔL is the deformed length, L is the original length, and the strain value is converted to the stress value by Hooke's law: σ=E·∈, where: σ is the stress value, and E is the elastic modulus of the material;

[0013] The stress value detected by the MEMS sensor is F is the applied force, A is the area under force;

[0014] The Bragg wavelength change detected by the fiber Bragg grating sensor is λ B is the initial Bragg wavelength, n eff is the effective refractive index of the optical fiber, ∈ is the strain value, P e is the photoelastic coefficient of the optical fiber;

[0015] The real-time assessment and early warning system based on LSTM and random forest uses LSTM networks to predict future strain values, stress values, and Bragg wavelength changes, and time series analysis:

[0016] LSTM unit formula: Forget gate: f t =σ(W f ·[h t-1 , x t ]+b f )

[0017] Among them, ft is the output of the forget gate, W f is the weight matrix of the forget gate, [h t-1 , x t ] is the concatenation of the previous hidden state and the current input, b f is the bias of the forget gate, and σ is the sigmoid activation function.

[0018] Input gate:

[0019] i t = o(W i · [h t-1 , x t ] + b i )

[0020]

[0021] where i t is the output of the input gate, is the candidate cell state, W i and WC are the corresponding weight matrices, b i and b C are the biases; σ is a symbol representing the sigmoid function, which is used to compress the input value to between 0 and 1; in the formula, represents the candidate memory cell state at time step t, tanh is a symbol representing the hyperbolic tangent function, which is used to compress the input value to between -1 and 1, W C is the weight matrix of the candidate memory cell state, h t-1 is the hidden state of the previous time step, x t is the input of the current time step, and b c is the bias of the candidate memory cell state.

[0022] Cell state update:

[0023] where C t is the cell state at the current time, and C t-1 is the cell state at the previous time.

[0024] Output gate:

[0025] o t = σ(W o · [h t-1 , x t ] + b o )

[0026] h t = o t * tanh(C t )

[0027] wherein o t is the output of the output gate, h t is the hidden state at the current time step, W o is the weight matrix of the output gate, b o is the bias;

[0028] The future strain value, stress value and Bragg wavelength change are predicted by the LSTM network, and the prediction formula is:

[0029] wherein LSTM is the calculation process of the unit, used to predict the future strain value stress value and Bragg wavelength change and respectively represent the predicted value at time step t+1, ∈ t ,σ t and Δλ B,t respectively represent the input value at time step t, and LSTM is a recurrent neural network used to process and predict time series data;

[0030] In step S4, the random forest classifier is a decision tree, and each decision tree is divided according to the input features to generate a class represented by a leaf node; the random forest is: wherein h(x) is the final prediction result of the random forest, h i (x) is the prediction result of the i-th decision tree, and N is the number of decision trees.

[0031] Training phase: the random forest model is trained using the strain value, stress value, Bragg wavelength change data and the corresponding risk level monitored in the history; the input features include the current and predicted strain value, stress value and Bragg wavelength change:

[0032]

[0033] Prediction phase: the current monitoring data and the LSTM prediction data are input into the trained random forest model to predict the risk level:

[0034]

[0035] Further, in step S5, if ; wherein the current stress value σ current is in the range of 0 to 5 MPa, and the stress change rate is less than 0.1 MPa / h, the difference |σ current -σ mean | between the current stress value and the average stress value is less than the standard deviation σstd , and the risk coefficient R is less than 1, that is, determined as low risk, then continue routine monitoring and inspection, maintain existing protection measures, and generate a monitoring report once a week.

[0036] Further, if σ std ≤σ current -σ mean <2σ std ; wherein the current stress value σ current is in the range of 5 to 7 MPa, and the stress change rate is between 0.1 to 0.5 MPa / h, the difference between the current stress value and the average stress value σ current -σ mean is greater than or equal to the standard deviation σ std and less than 2 times the standard deviation, that is, determined as medium risk, then increase the monitoring frequency and increase the support measures.

[0037] Further, if σ current >7MPa, |σ current -σ mean |≥2σ std , R≥2; wherein, in the case that the current stress value σ current is greater than 7 MPa, and the stress change rate is greater than 0.5 MPa / h, the difference between the current stress value and the average stress value |σ current -σ mean | is greater than or equal to 2 times the standard deviation σ std , and the risk coefficient R is greater than or equal to 2, that is, determined as high risk, emergency stop operation, evacuate personnel, implement emergency support and expert evaluation, develop a detailed reinforcement plan, and monitor stress data in real time.

[0038] Compared with the prior art, the beneficial effects of the present application are:

[0039] The present application can monitor the stress and displacement changes of the surrounding rock mass of the roadway in real time through the multi-point strain gauge, the fiber Bragg grating sensor and the MEMS sensor, the collected stress data has high precision and reliability, and the accuracy of the monitoring data is ensured; efficient data transmission and processing are achieved: advanced wireless transmission technology is adopted to realize efficient wireless transmission of the monitoring data, the data collector processes, digitizes and stores the sensor data, and transmits the data to the data center through the wireless transmission module, the data center pre-processes, stores and analyzes the received data, generates real-time risk assessment reports by using a machine learning model and a statistical analysis method; accurate risk assessment and early warning are achieved: through the calculation of the comprehensive risk index, the system can accurately assess the risk level of the roadway, and by setting reasonable risk thresholds, the system can timely warn different risk levels to guide the on-site personnel to take corresponding protective measures; according to the risk assessment results, the system provides specific physical protection methods for three risk levels of low, medium and high: low risk: continue routine monitoring and inspection, and maintain the existing protective measures; medium risk: increase the monitoring frequency and increase the supporting measures such as shotcrete and temporary support installation; high risk: emergency stop operation, evacuate personnel, implement emergency support and expert evaluation, and develop a detailed reinforcement scheme. The present application significantly improves the safety and monitoring efficiency in the process of roadway excavation, reduces safety hazards, and protects the safety of the operating personnel and equipment. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0042] Please refer to Figure 1 The present application provides a technical solution: a blasting safety protection method based on unstable roadway excavation, comprising the following steps:

[0043] S1: arranging a multi-point strain gauge, a MEMS sensor and a fiber Bragg grating sensor in the roadway to collect strain values, stress values and Bragg wavelength change data in real time;

[0044] A multi-point strain gauge is installed every 5 meters at the roadway roof, one at a depth of 0.5 meters and another at a depth of 1 meter. The strain gauge reflects the internal stress changes by detecting the deformation of the rock mass, helping to monitor the stability of the roadway roof, two sides and floor in real time, preventing rock mass instability and collapse. The strain gauge is installed at key positions on the roadway roof, sides and floor, and the strain value of the rock mass is collected at regular intervals. The strain gauge reflects the internal stress changes by measuring the deformation of the rock mass, and regular strain data recording helps to monitor the subsidence of the roof, the extrusion deformation of the two sides and the uplift of the floor, etc. Through the strain value, the response of the rock mass after being stressed is understood, and potential instability areas are identified.

[0045] A MEMS sensor is installed every 3 meters on both sides of the roadway, one at a depth of 0.5 meters and another at a depth of 1 meter. The MEMS sensor provides high-sensitivity stress monitoring data by measuring the small stress changes inside the rock mass, helping to discover local instability and crack development on both sides of the roadway in a timely manner. The MEMS sensor is installed on the surface and inside the roadway to monitor small stress changes and vibration conditions. These sensors have high sensitivity and high resolution, capable of capturing subtle stress changes inside the roadway. The MEMS sensor has a high data acquisition frequency, which can reflect the stress distribution and transmission inside the rock mass in real time, helping to identify local stress concentration and potential crack expansion risks.

[0046] A fiber Bragg grating sensor is installed every 10 meters at the roadway floor, one at a depth of 1 meter and another at a depth of 2 meters. The fiber Bragg grating sensor accurately measures the stress distribution at the roadway floor and deep inside by measuring the Bragg wavelength changes, providing high-precision stress data to ensure the long-term stability of the overall roadway structure. The fiber Bragg grating sensor is installed at the roadway floor and deep inside to measure Bragg wavelength changes, reflecting stress and deformation conditions by detecting Bragg wavelength changes. The fiber Bragg grating sensor has high precision and anti-interference capability, capable of stable operation in harsh environments, especially suitable for monitoring stress changes in deep rock mass, providing long-term monitoring data for the stability of the overall structure.

[0047] The strain value detected by the multi-point strain gauge is ΔL is the deformation length, L is the original length, and the strain value is converted to stress value by Hooke's Law: σ = E·∈, where σ is the stress value and E is the elastic modulus of the material; The stress value detected by the MEMS sensor is F is the force and A is the force area; The Bragg wavelength change detected by the fiber Bragg grating sensor is λ B is the initial Bragg wavelength, n eff is the effective refractive index of the fiber, ∈ is the strain value, and P eis the photoelastic coefficient of the optical fiber. The data collection process not only includes real-time data acquisition, but also involves preliminary data processing and storage. The analog signals output by the sensor are filtered and amplified through a signal conditioning circuit to ensure signal stability and accuracy. The conditioned signals are then converted to digital signals by an analog-to-digital converter (ADC) and stored in the data collector. Each piece of data recorded by the data collector is accompanied by a timestamp and sensor location identifier for subsequent data analysis and processing. Real-time data is transmitted to the data center through a wireless transmission module (such as LoRa or 5G). The data transmission process needs to ensure the integrity and real-time nature of the data to avoid data loss and delay. After data transmission, the next step is data preprocessing, which provides basic data support for subsequent analysis and risk assessment.

[0048] S2: Clean, denoise and normalize the collected data; Data denoising: use a low-pass filter to remove high-frequency noise; Data normalization: standardize the data to a uniform scale range for subsequent processing; Data preprocessing is to convert the collected raw data into high-quality, suitable for analysis. Raw data often contains noise, outliers and inconsistent formats, and needs to be processed to ensure data accuracy and reliability.

[0049] First, data cleaning removes noise and outliers. Noise may come from environmental interference and device errors of the sensor, common methods include low-pass filtering and moving average filtering. Outlier detection and processing is to identify and correct sudden abnormal points in the data, which may be caused by sensor failure or external interference. Next, data denoising removes high-frequency noise in the data through filtering algorithms, common filtering algorithms include low-pass filtering, Gaussian filtering, etc. They can smooth the data curve, highlight the trend and pattern in the data, and reduce the influence of random fluctuations on the analysis results. After data denoising, data normalization is performed. Data normalization is to convert data of different dimensions to the same scale range, so that subsequent analysis models can treat each feature equally. Common normalization methods include min-max normalization, Z-score standardization, etc. The preprocessed data needs to be time-synchronized and format-unified. Data from different sensors may have different sampling frequencies and timestamps, time synchronization is to unify these data to the same time axis, ensuring that each data point has a consistent time reference. Format unification is to convert data from different sources to a unified format for subsequent analysis and processing. The ultimate goal of data preprocessing is to improve data quality and provide a reliable data foundation for subsequent analysis and modeling. Through data cleaning, denoising, normalization, time synchronization and format unification, the accuracy and consistency of the data are ensured, so that the next step of LSTM prediction and random forest classification can be carried out on the basis of high-quality data.

[0050] S3: Predicting future strain values, stress values, and Bragg wavelength changes using LSTM networks:

[0051] 1. LSTM unit formula: Forget gate: f t = σ(W f · [h t-1 , x t ] + b f )

[0052] Input gate: i t = σ(W i · [h t-1 , x t ] + b i )

[0053] Cell state update:

[0054] Output gate: o t = σ(W o · [h t-1 , x t ] + b o ), h t = o t *tanh(C t )

[0055] 2. Prediction formula:

[0056] LSTM (Long Short-Term Memory) is a type of recurrent neural network (RNN) specifically designed for processing and predicting time series data. In mine tunnel monitoring, LSTM networks can effectively capture the temporal dynamics of strain values, stress values, and Bragg wavelength changes, predicting future stress and deformation trends. LSTM networks control the flow of information through a series of gating mechanisms (forget gate, input gate, and output gate), allowing them to remember important information from long time series data while ignoring irrelevant information. Specifically, the forget gate determines which information in the current cell state needs to be discarded, the input gate determines which new information needs to be added to the cell state, and the output gate determines which parts of the current cell state will be passed as output to the next time step.

[0057] In implementation, the LSTM network first needs to be trained. By inputting a large amount of historical monitoring data (strain value, stress value and Bragg wavelength change), the LSTM network learns the time pattern and rule of these data, adjusts the internal parameters to minimize the prediction error. After training, the LSTM network takes the current monitoring data as input to predict the strain value, stress value and Bragg wavelength change in the future few time points. The output of the LSTM network is the predicted future strain value, stress value and Bragg wavelength change, which reflects the stress and deformation trend of the roadway rock mass in the future period of time. Through the prediction of the LSTM network, the construction personnel can understand the potential risks in advance and take preventive measures. The results predicted by the LSTM network will be used as input for the random forest model to assess and classify the risks, and the application of the LSTM network not only improves the utilization rate of monitoring data, but also provides early warning of future stress and deformation, making the safety management of mine roadway more proactive and efficient.

[0058] S4: Use the random forest model to classify the LSTM prediction results and current monitoring data, and evaluate the rock mass stability risk level:

[0059] The random forest classifier is a decision tree, and each decision tree divides the input features to generate leaf nodes representing the class. The random forest: where h(x) is the final prediction result of the random forest, h i (x) is the prediction result of the i-th decision tree, and N is the number of decision trees.

[0060] Training phase: Use the historical monitoring strain value, stress value, Bragg wavelength change data and their corresponding risk level to train the random forest model; the input features include the current and predicted strain value, stress value and Bragg wavelength change:

[0061]

[0062] Prediction phase: input the current monitoring data and LSTM prediction data into the trained random forest model to predict the risk level:

[0063]

[0064] Random forest is an ensemble learning method based on decision trees, widely used in classification and regression tasks. In mine roadway monitoring, the random forest uses the future strain value, stress value and Bragg wavelength change predicted by the LSTM network, as well as the current monitoring data, to classify the rock mass stability and safety. The random forest improves the stability and accuracy of the model by constructing multiple decision trees. Each decision tree is trained by randomly sampling from the training data set and selecting the best split point through a random subset of features to generate multiple different decision trees.

[0065] In classification, random forest takes the prediction results of multiple decision trees as votes and finally outputs the category. By integrating the prediction results of multiple models, random forest can effectively reduce the overfitting problem of a single decision tree and improve the generalization ability of the model. In practical application, first, the historical monitoring data (including strain value, stress value, Bragg wavelength change and corresponding risk level) are used to train the random forest model. The feature vector of the training data includes the current and predicted strain value, stress value and Bragg wavelength change, and the target variable is the corresponding risk level (low risk, medium risk and high risk). After training, the random forest model takes the current monitoring data and the LSTM prediction result as input for real-time risk classification. Specifically, the input feature vector includes the current strain value, stress value and Bragg wavelength change, as well as the future strain value, stress value and Bragg wavelength change predicted by the LSTM network. The random forest model passes these input features to each decision tree, and the decision tree classifies according to the splitting rules learned in the training process. Finally, the random forest determines the risk level by majority voting.

[0066] S5: According to the risk level output by the random forest model, the system automatically issues a warning and prompts to take corresponding physical protection measures:

[0067] In the case of low risk, the current stress value and stress change rate are within the safe range, but monitoring and preventive measures still need to be maintained. The specific physical protection methods are as follows: routine monitoring measures: continue to use the existing sensor system for real-time monitoring, purpose: to ensure that potential stress changes can be discovered in time; Regular calibration of sensors: regularly calibrate multi-point strain gauges, fiber Bragg grating sensors and MEMS sensors, purpose: to ensure the accuracy of measurement data; Routine support measures: maintain the existing anchor rod, anchor cable and support system, purpose: to provide basic structural stability; Regular inspection measures: regularly check the integrity of the roadway structure, including the roof, two sides and the floor, purpose: to discover and repair small-scale cracks and deformations in time; Monitoring report measures: generate a monitoring report once a week, purpose: to record and analyze the current stress state, ensure long-term accumulation and analysis of monitoring data.

[0068] In medium-risk situations, the current stress value and stress change rate are close to but have not exceeded the safety threshold, and monitoring and support measures need to be strengthened. The specific physical protection methods are as follows: Measures to increase monitoring frequency: Increase the monitoring frequency from once an hour to once every 30 minutes, with the purpose of capturing stress changes more timely and preventing sudden risks; Measures to strengthen support: Increase the density of anchor rods and anchor cables, especially in stress concentration areas, with the purpose of enhancing the overall stability of the structure and preventing local instability; Measures to spray concrete: Spray a layer of concrete on the tunnel surface to increase the support strength, with the purpose of preventing rock spalling and small-scale collapse; Measures to install temporary supports: Install temporary steel supports in high-stress areas, with the purpose of providing additional support and reducing rock stress; Measures for detailed inspections: Conduct detailed inspections every day, focusing on stress concentration and deformation areas, with the purpose of promptly detecting and dealing with potential structural problems; Measures for detailed monitoring reports: Generate a detailed monitoring report every day, with the purpose of analyzing current stress changes and formulating further protective measures; Measures to prepare emergency plans: Develop and prepare emergency plans, including evacuation routes and emergency support materials, with the purpose of ensuring a rapid response when risks escalate.

[0069] In high-risk situations, the current stress value and stress change rate have exceeded the safety threshold, and emergency measures need to be taken to prevent accidents. The specific physical protection methods are as follows: Emergency stop operation measures: immediately stop all excavation and blasting operations, the purpose is to prevent further increase in rock stress and avoid causing collapse; evacuation measures: immediately evacuate all workers to ensure their safety, the purpose is to avoid casualties; implement emergency support measures: install steel arches and spray concrete, especially in high stress and deformation areas, the purpose is to quickly stabilize the rock mass and prevent large-scale collapse; organize expert evaluation measures: organize geological and engineering experts to conduct on-site evaluation Assessment, purpose: to formulate detailed reinforcement and repair plans; real-time monitoring measures: add real-time monitoring equipment to ensure that stress changes are monitored at any time, purpose: to provide real-time data support and guide emergency and repair work; risk assessment report update measures: update the risk assessment report in real time based on the latest monitoring data, purpose: to provide the latest risk analysis and suggestions; construction prohibition measures: prohibit any construction activities before the high risk is lifted, purpose: to ensure the safety of personnel and equipment and avoid secondary risks; formulate detailed reinforcement plan measures: formulate and implement a detailed reinforcement plan based on the expert assessment results, purpose: to stabilize the rock mass in the long term and restore the safety of the tunnel.

[0070] In summary, the application is a more efficient and reliable monitoring and protection technology, which improves the safety and efficiency in the process of roadway excavation, realizes real-time monitoring and evaluation of the stress and displacement inside the roadway through integration of various sensor technologies, advanced data acquisition and transmission systems and scientific risk evaluation models, so that effective protection measures can be taken in time to ensure construction safety; through specific physical protection methods for different risk levels, it can effectively respond to various risks in time during the process of roadway excavation, and ensure the safety and efficiency of operation.

[0071] The above describes the application and its embodiments in a schematic manner, which is not restrictive, and the embodiments shown in the drawings are only one of the embodiments of the application, and the actual structure is not limited thereto. Therefore, if a person skilled in the art is inspired by it, without departing from the purpose of the application, similar structure and embodiments can be designed without creativity, which shall belong to the protection scope of the application.

Claims

1. A blasting safety protection method based on unstable tunnel excavation, characterized by: The following steps are involved: S1: Multi-point strain gauges, MEMS sensors, and fiber Bragg grating sensors are deployed in the tunnel to collect strain values, stress values, and Bragg wavelength change data in real time; S2: Clean, denoise and normalize the collected data; S3: Use LSTM network to predict future strain values, stress values ​​and Bragg wavelength changes; S4: Use the random forest model to classify the LSTM prediction results and current monitoring data to assess the rock mass stability risk level; S5: Based on the risk level output by the random forest model, the system automatically issues an early warning and prompts the user to take appropriate physical protective measures. In step S5, if |σ current -σ mean |<σ std ,R<1; where the current stress value σ current In the range of 0 to 5 MPa, and the stress change rate When the stress value is less than 0.1 MPa / h, the difference between the current stress value and the average stress value |σ current -σ mean |Less than the standard deviation σ std , and the risk coefficient R is less than 1, that is, it is judged to be low risk, then continue routine monitoring and inspections, maintain existing protective measures, and generate a monitoring report once a week; In step S5, if σ std ≤|σ current -σ mean |<2σ std ; Among them, the current stress value σ current In the range of 5 to 7 MPa, and the stress change rate In the case of 0.1 to 0.5 MPa / h, the difference between the current stress value and the average stress value |σ current -σ mean |greater than or equal to the standard deviation σ std If the risk is less than 2 times the standard deviation, it is considered to be medium risk, and the monitoring frequency should be increased and support measures should be added; In step S5, |σ current -σ mean |≥2σ std ,R≥2; where, at the current stress value σ current Greater than 7MPa, and the stress change rate When the stress is greater than 0.5 MPa / h, the difference between the current stress value and the average stress value |σ current -σ mean | Greater than or equal to 2 times the standard deviation σ std , and the risk coefficient R is greater than or equal to 2, which means it is judged as high risk, and the operation is stopped immediately, personnel are evacuated, emergency support and expert evaluation are implemented, a detailed reinforcement plan is formulated, and stress data is monitored in real time.

2. The blasting safety protection method based on unstable tunnel excavation according to claim 1 is characterized by: In step S1, a multi-point strain gauge is installed every 5 meters on the tunnel roof, one at a depth of 0.5 meters and one at a depth of 1 meter. The strain gauge reflects internal stress changes by detecting the deformation of the rock mass. A MEMS sensor is installed every 3 meters on both sides of the tunnel, one at a depth of 0.5 meters and one at a depth of 1 meter. The MEMS sensor provides highly sensitive stress monitoring data by measuring tiny stress changes within the rock mass. A fiber Bragg grating sensor is installed every 10 meters on the tunnel floor, one at a depth of 1 meter and one at a depth of 2 meters. The fiber Bragg grating sensor accurately measures the stress distribution on the tunnel floor and deep inside by changing the Bragg wavelength. The strain value detected by the multi-point strain gauge is ∈, ΔL is the deformed length, L is the original length, and the strain value is converted to the stress value by Hooke's law: σ=E·∈, where: σ is the stress value, and E is the elastic modulus of the material; The stress value detected by the MEMS sensor is σ, F is the applied force, A is the area under force; The Bragg wavelength change detected by the fiber Bragg grating sensor is Δλ B , λ B is the initial Bragg wavelength, n eff is the effective refractive index of the optical fiber, ∈ is the strain value, P e is the photoelastic coefficient of the optical fiber; The real-time assessment and early warning system of LSTM and random forest uses the LSTM network to predict future strain values, stress values, and Bragg wavelength changes: LSTM unit formula: Forget gate: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t is the output of the forget gate, W f is the weight matrix of the forget gate, [h t-1 ,x t ], is the concatenation of the hidden state of the previous moment and the current input, b f is the bias of the forget gate, σ is the sigmoid activation function; Input Gate: i t =σ(W i ·[h t-1 ,x t ]+b i ) Among them, i t is the output of the input gate, is the candidate cell state, W i and W C is the corresponding weight matrix, b i and b C is the bias value; σ is a symbol representing the sigmoid function, which is used to compress the input value to between 0 and 1; Represents the candidate memory cell state at time step t, tanh is a symbol representing the hyperbolic tangent function, and the tanh function is used to compress the input value to between -1 and 1, W c is the weight matrix of the candidate memory cell state, h t-1 is the hidden state of the previous time step, x t is the input of the current time step, b c is the bias of the candidate memory cell state; Cell status update: Among them, C t is the cell state at the current moment, C t-1 is the cell state at the previous moment; Output gate: the t =σ(W o ·[h t-1 ,x t ]+b o ) h t =o t *fishy(C) t ) Among them, t is the output of the output gate, h t is the hidden state at the current moment, W o is the weight matrix of the output gate, b o is bias; The LSTM network is used to predict future strain values, stress values, and Bragg wavelength changes. The prediction formula is: Among them, LSTM is the calculation process of the unit, which is used to predict the future strain value Stress value and Bragg wavelength variation and Represent the predicted value at time step t+1, ∈ t ,σ t and Δλ B,t They represent the input values ​​at time step t respectively. LSTM is a recurrent neural network used to process and predict time series data. In step S4, the random forest classifier is a decision tree. Each decision tree is divided into rows according to the input features and generates categories represented by leaf nodes. Random forest: Among them, h(x) is the final prediction result of random forest, h i (x) is the prediction result of the i-th decision tree, and N is the number of decision trees; Training phase: Use historically monitored strain values, stress values, Bragg wavelength change data and their corresponding risk levels to train the random forest model; input features include current and predicted strain values, stress values, and Bragg wavelength changes: Prediction stage: Input the current monitoring data and LSTM prediction data into the trained random forest model to predict the risk level:

Images