Gas monitoring and early warning system for roadway tunneling construction

Abstract

The invention relates to the technical field of mine safety, and particularly discloses a gas monitoring and early warning system for roadway tunneling construction, which comprises a data acquisition module, a data processing module, a prediction module, an early warning analysis module, a sensor sensitivity compensation judgment module and a self-adaptive calibration module. Harmful gas concentration data and environment, geology and tunneling operation data are collected by constructing a three-level collaborative network of fixed nodes, an inspection robot and a tunneling machine micro sensor; carrying out data cleaning and fusing multi-source gas concentration data to obtain real-time gas concentration; predicting the gas concentration at the next moment through an LSTM network training model based on the real-time gas concentration; performing gradient early warning according to the predicted value and a safety threshold value, and linking ventilation, equipment and a risk avoiding system; sensitivity attenuation of the gas sensor is further judged and calibrated, full-domain real-time sensing, trend prediction and early warning and intelligent calibration of the sensor of the driving face gas are achieved, and monitoring timeliness, early warning scientificity and data reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of mine safety technology, specifically to a gas monitoring and early warning system for tunnel excavation. Background Technology

[0002] In underground engineering projects such as coal mines and non-coal mines, real-time monitoring and accurate early warning of hazardous gases are crucial for ensuring operational safety. As tunneling depth increases, geological conditions become more complex, and tunneling efficiency improves, the generation, emission, and diffusion patterns of hazardous gases become increasingly complex. Traditional gas monitoring and early warning technologies are no longer sufficient to meet the demands for real-time, comprehensive, accurate, and forward-looking safety management. Specific technical challenges are as follows:

[0003] 1. Traditional roadway gas monitoring relies heavily on fixed-node sensors, supplemented by regular manual inspections, which has significant drawbacks: Firstly, fixed nodes cannot cover dynamically changing areas where gas easily accumulates, such as in front of the tunneling machine head and at the top corners of the roadway. As the tunneling face continues to advance, fixed-point sensors cannot achieve "on-demand measurement," resulting in spatial monitoring blind spots. Secondly, the distance between fixed nodes and the tunneling face causes delays in concentration monitoring data transmission, missing the optimal window for emergency response. Furthermore, traditional systems are not deeply integrated with the tunneling machine and inspection equipment, lacking mobile monitoring capabilities. The monitoring data is difficult to reflect the real-time gas concentration at the tunneling face, resulting in extremely incomplete risk perception.

[0004] 2. Existing gas early warning systems mostly adopt the "current concentration threshold comparison" model. Due to the lack of intelligent fusion of multi-source data such as environmental, geological, and tunneling operation data, as well as long-term time-series trend prediction, there are obvious shortcomings: traditional systems only collect gas concentrations at fixed nodes without integrating related data, and single concentrations are prone to false alarms and missed alarms; some systems use simple empirical formulas or traditional time-series methods for prediction, but they are difficult to handle complex scenarios; and most of them are "passive alarms" without prediction before exceeding the limit, resulting in extremely short emergency response time and a serious lack of foresight.

[0005] 3. The sensitivity of gas concentration sensors is easily affected by the harsh downhole environment and long-term operation. However, traditional systems lack intelligent judgment and adaptive calibration mechanisms for sensitivity decay. On the one hand, decay judgment relies on periodic manual calibration with a fixed cycle, which cannot reflect the actual operating status of the sensor in real time. This can easily lead to situations where a significant drop in sensitivity is detected late and the monitoring data is distorted during the period. On the other hand, the calibration process has a low degree of automation. Manual operation is not only inefficient and poses safety risks, but parameter adjustment also depends on experience and lacks quantitative compensation for the internal and external environment of the sensor. The calibration accuracy is greatly affected by human factors. In addition, existing technologies lack dynamic compensation mechanisms and cannot correct sensitivity drift according to real-time changes in the sensor's operating environment. After calibration, the sensor still has detection deviations under actual working conditions, which seriously affects the reliability of monitoring data.

[0006] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0007] The purpose of this invention is to provide a gas monitoring and early warning system for tunnel excavation to solve the problems mentioned in the background.

[0008] The objective of this invention can be achieved through the following technical solution: a gas monitoring and early warning system for tunnel excavation, comprising:

[0009] The data acquisition module is used to collect real-time monitoring data on harmful gas concentrations, environmental data, geological data, and tunnel excavation operation data during the tunnel excavation process.

[0010] The data processing module is used to clean and normalize the real-time data collected during the tunnel excavation process to obtain the processed data. The processed real-time hazardous gas concentration data of each fixed point, each inspection and monitoring point, and each mobile monitoring point are then fused and analyzed to obtain the real-time monitoring hazardous gas concentration during the tunnel excavation process.

[0011] The prediction module is used to build a deep neural network model for predicting the concentration of harmful gases, and to predict the concentration of harmful gases at the next time point based on real-time monitoring of harmful gas concentrations.

[0012] The early warning analysis module is used to issue early warnings based on the predicted concentration of harmful gases at the next time point and the safe threshold for the concentration of harmful gases.

[0013] The sensor sensitivity compensation determination module is used to determine whether the sensitivity of each gas concentration sensor needs to be compensated based on the real-time monitoring data of the internal operating environment, the external operating environment, and the detection log data of each gas concentration sensor.

[0014] The sensor adaptive calibration module is used to perform sensor sensitivity compensation analysis when the sensitivity of a gas concentration sensor needs to be compensated for, generate the sensor's calibration sensitivity, and feed back the calibration sensitivity to the corresponding sensor's calibration control terminal to perform corresponding calibration control and obtain the harmful gas concentration at the next time point.

[0015] The beneficial effects of this invention are:

[0016] This invention constructs a three-level collaborative hazardous gas concentration monitoring network consisting of fixed nodes, inspection robots, and micro sensors on the tunneling machine, forming a three-dimensional monitoring system that combines points, lines, and surfaces. This enables real-time perception of gas concentration throughout the entire process of dynamic advancement of the tunneling face, improving the timeliness and comprehensiveness of hazardous gas concentration monitoring during tunnel excavation.

[0017] This invention employs an LSTM network to construct a hazardous gas concentration prediction model. Leveraging its gating mechanism and long-term memory, the model input integrates historical concentration, environmental, geological, and tunneling operation data. Through dynamic adjustment of the forget gate, input gate, and output gate, it accurately learns correlation patterns, improving prediction accuracy and avoiding false alarms and missed alarms caused by prediction bias. Based on the model's output of the predicted hazardous gas concentration and safety threshold for the next monitoring time point, it provides tiered early warnings and triggers corresponding emergency measures, rather than only issuing warnings after the current concentration exceeds the limit. This breaks through the traditional passive mode of relying solely on current concentration alarms, reserving a critical time window for emergency response and significantly improving the accuracy and scientific rigor of early warnings, achieving a leap from threshold-based alarms to trend prediction-based alarms.

[0018] This invention analyzes and calculates the impact values ​​of the sensor's internal and external operating environments, as well as the detection deviation value, based on the sensor's internal operating environment data, external operating environment data, and detection log data. When any value exceeds a threshold, compensation is required. This eliminates the need for regular manual calibration, preventing sensor sensitivity degradation from going unnoticed. The invention calculates and corrects the guidance weight based on the cause of sensor attenuation and matches the corresponding sensitivity compensation value, ensuring that the sensor maintains high accuracy in complex underground mining environments and preventing data distortion from misleading safety decisions. Attached Figure Description

[0019] The invention will now be further described with reference to the accompanying drawings.

[0020] Figure 1 This is a system block diagram of the present invention.

[0021] Figure 2 This is a schematic diagram of the iterative training process of the LSTM network model of the present invention.

[0022] Figure 3 This is a schematic diagram illustrating the structural principle of the LSTM network model of this invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figure 1 As shown, this invention is a gas monitoring and early warning system for tunnel excavation, comprising:

[0025] The data acquisition module is used to collect real-time monitoring data on harmful gas concentrations, environmental data, geological data, and tunnel excavation operation data during the tunnel excavation process. The specific process is as follows:

[0026] Fixed nodes are set up at preset intervals along the tunnel, and each node includes a hazardous gas concentration monitoring sensor and an environmental monitoring sensor. The hazardous gas concentration monitoring sensor includes a gas sensor, a carbon monoxide sensor, etc., to detect the concentration of hazardous gases such as methane and carbon monoxide at each fixed point. The environmental monitoring sensor includes a dust concentration sensor, a temperature and humidity sensor, a wind speed sensor, etc., to detect environmental data such as dust concentration, temperature and humidity, and wind speed at each fixed point.

[0027] An explosion-proof tracked inspection robot is used, equipped with a laser gas sensor, an environmental monitoring sensor, and a TOFD ultrasonic positioning module. It performs periodic inspections of areas such as the top and corners of the tunnel that are not covered by fixed nodes, according to a preset route. The inspection robot obtains the position coordinates of the tunneling machine in real time and automatically adjusts the inspection route to ensure that the distance to the tunneling face is ≤5m and the route deviation is ≤±0.5m, eliminating blind spots and obtaining harmful gas concentration data and environmental data at each inspection and monitoring point.

[0028] Miniaturized, explosion-proof spectral gas detection modules and environmental monitoring modules are integrated into the body of the tunneling machine (especially the head section). The mobile monitoring node mounted on the tunneling machine moves with the tunneling machine and directly monitors the gas in front of and around the tunneling cutting face in real time and in situ, fundamentally solving the problems of monitoring lag and blind spots in front, and obtaining the concentration data of harmful gases and environmental data of each mobile monitoring point.

[0029] Geological data includes the degree of fracture development and the coal seam permeability coefficient. The process of obtaining the degree of fracture development is as follows: Images of the coal body within the working face are acquired using an intrinsically safe borehole inspection instrument. A U-Net semantic segmentation network is used to semantically segment the borehole images into fractured and intact coal bodies, distinguishing fractured regions. Then, through contour detection and morphological operations, the density and aperture of the fractures are extracted and compared with set first density thresholds, second density thresholds, first aperture thresholds, and second aperture thresholds. If the fracture density is less than the first density threshold and the fracture aperture is less than the first aperture threshold, the fractures are considered sparsely developed. If the fracture density is between the first and second density thresholds, or the fracture aperture is between the first and second aperture thresholds, the fractures are considered moderately developed. If the fracture density is greater than the second density threshold or the fracture aperture is greater than the second aperture threshold, the fractures are considered densely developed. The fracture development level is then matched with the fracture development degree corresponding to each set fracture development level to obtain the degree of fracture development.

[0030] Based on Darcy's law of seepage, the permeability coefficient is calculated by relating borehole gas flow rate to pressure. This includes: drilling a borehole in the coal seam to be tested; sealing the borehole using a specialized sealing device (such as a capsule-viscose sealer) to ensure that gas only escapes from the coal seam through the borehole, preventing leakage; monitoring gas pressure changes within the borehole after sealing; and, once the pressure stabilizes, activating a borehole gas flow rate measuring device (such as a flow meter or a dynamic gas emission monitoring system) to continuously measure the gas emission rate Q at different times; and inputting borehole geometric parameters (borehole diameter) into the borehole flow rate measurement device. Hole depth Gas pressure gradient (gas pressure difference inside and outside the borehole) Substituting gas flow rate Q into the permeability coefficient calculation formula, and based on Darcy's law, the formula is derived using the radial flow formula. The coal seam permeability coefficient was calculated. ,in For gas dynamic viscosity, For coal seam thickness, To affect the radius, Gas pressure at different locations;

[0031] Tunnel excavation data includes excavation speed and cutting power;

[0032] It should be noted that this invention constructs a three-level collaborative hazardous gas concentration monitoring network consisting of fixed nodes, inspection robots, and micro sensors on the tunneling machine, forming a three-dimensional monitoring system that combines "points, lines, and surfaces." This enables real-time perception of gas concentration throughout the entire process of dynamic advancement of the tunneling face, greatly improving the timeliness and comprehensiveness of monitoring.

[0033] The data processing module is used to clean and normalize the real-time data collected during tunnel excavation, obtaining processed data. It then fuses and analyzes the processed real-time hazardous gas concentration data from fixed points, inspection monitoring points, and mobile monitoring points to obtain the real-time monitored hazardous gas concentration during tunnel excavation. The specific process is as follows:

[0034] Missing data in real-time data is detected and filled using the median. Outliers in the real-time data are then obtained using the Z-score method, deleted, and the real-time data is normalized.

[0035] The processed hazardous gas concentration data of each fixed point, each inspection and monitoring point, and each mobile monitoring point are acquired. The average values ​​of these data are calculated to obtain the average hazardous gas concentrations at the fixed points, inspection and monitoring points, and mobile monitoring points. A weighted average algorithm is then used to merge these average values, with the fixed point concentration having a weight of 0.3, the inspection and monitoring point concentration having a weight of 0.4, and the mobile monitoring point concentration having a weight of 0.3. This merged hazardous gas concentration is then used as the real-time monitoring hazardous gas concentration during the tunnel excavation process.

[0036] It should be noted that the weighting of the hazardous gas concentrations at fixed points, inspection monitoring points, and mobile monitoring points is based on the following criteria: Fixed monitoring nodes have stable installation locations and good long-term data stability, but they are far from the tunnel face, resulting in a delay in capturing instantaneous gas concentration changes at the face; therefore, their weight is set at 0.3. Inspection robots can dynamically approach the tunnel face, providing a more balanced combination of real-time data and accuracy. They are also equipped with laser sensors, offering strong resistance to dust and electromagnetic interference; therefore, their weight is set at 0.4. Miniature sensors are directly installed on the cutting section of the tunneling machine, capturing the original concentration at the moment of cutting; however, they are more susceptible to cutting vibration and dust interference, resulting in slightly weaker data stability; therefore, their weight is set at 0.3. By weighting these different parameters, the fusion result balances the capture of real-time concentrations at the tunnel face with the long-term stability of the data, avoiding the shortcomings of a single monitoring method and providing high-quality input for subsequent concentration prediction.

[0037] The prediction module is used to build a deep neural network model for predicting hazardous gas concentrations, and to predict the hazardous gas concentration at the next time point based on real-time monitoring. The specific process is as follows; please refer to [link / reference]. Figure 2 As shown:

[0038] A1. Obtain historical hazardous gas concentration data and tunnel excavation construction area characteristic data from the historical database. The tunnel excavation construction area characteristic data includes environmental data, geological data, and tunneling operation data. Filter the data to obtain a sample dataset.

[0039] A2. The sample datasets undergo preprocessing operations such as cleaning, standardization, and feature engineering (e.g., missing value imputation, outlier removal, normalization, and sequence segmentation) to obtain datasets for harmful gas concentration, environment, geology, and tunneling operations. These datasets are then divided into training, testing, and validation sets in a 6:2:2 ratio. It should be noted that the training set is used to train the deep LSTM network, allowing the model to learn data patterns; the testing set is used for final performance evaluation after model training (e.g., calculating test set error) to measure the model's generalization ability to unknown data; and the validation set is used during training to verify model performance (e.g., calculating validation set loss and evaluating generalization ability) and to assist in hyperparameter tuning.

[0040] A3. Construct a deep neural network model for predicting harmful gas concentration data. Initialize the model parameters and iteratively train the constructed deep neural network model using the training set. Starting from iteration number t=1, input the training data, and the network calculates the prediction output. Based on the error between the predicted value and the true value, adjust the network parameters in reverse to minimize the loss function. After each iteration, determine whether the deep neural network has converged. If not, proceed to t=t+1 and continue the next round of training iterations. If yes, calculate the test set error, use the pre-trained model to predict the test set data, and calculate the error between the predicted value and the true value (such as mean squared error, mean absolute error, etc.) to measure the model's generalization ability to unknown data. Determine whether the current test set error is the optimal value, i.e., the current model has the best prediction effect on the test set among all training iterations. If yes, the current model is the optimal model. If not, it means that the model parameters still need to be optimized. Readjust the model parameters, train and evaluate again, and finally obtain the trained deep neural network model for predicting harmful gas concentration.

[0041] A4. The prediction of hazardous gas concentration is made by using a trained deep neural network model. The input is real-time monitored hazardous gas concentration data, environmental data, geological data and tunneling data, and the output is the predicted value of hazardous gas concentration at the next time point.

[0042] The deep neural network model for predicting harmful gas concentrations employs an LSTM network model, comprising an input layer, an LSTM layer, and an output layer. Each LSTM layer includes three gating units (forget gate, input gate, and output gate) and memory cells. Each gating unit outputs a value between 0 and 1 through a sigmoid activation function, representing the degree to which information is allowed to pass (0 indicates complete prohibition, 1 indicates complete permission). It should be noted that the memory cell state acts like a "highway" for information transmission, running throughout the entire LSTM network and responsible for transferring information between different time steps. Information can be stored long-term while being transferred within the memory cell state. The forget gate and input gate work together on the memory cell state; the forget gate determines which old information to delete, and the input gate determines which new information to add, thus updating the memory cell state. The output gate determines what information to output based on the current memory cell state. The input sequence of the LSTM network is... The input is fed into the LSTM layer, and the output sequence is... The output is from the LSTM layer, where... It is the forecast period. It is historical input data, including historical hazardous gas concentration data, environmental data, geological data, and tunneling operation data. It predicts the concentration of harmful gases; before the next monitoring time t+1 arrives, the input sequence will be... The input is fed into a trained deep neural network model for predicting harmful gas concentrations, and the output sequence is... This yields the predicted concentration of harmful gases at time t+1.

[0043] For the specific calculation process of the LSTM layer, please refer to [link / reference]. Figure 3 As shown:

[0044] B1. Assume that at time... The input is The hidden state at the previous moment was The cell state a moment ago was The forgetting gate decides to start from the previous cellular state. What information is forgotten in the process, and what information is received from the current input? Hidden state from the previous moment It outputs a vector between 0 and 1 through a sigmoid activation function. The formula is: ,in It represents the concatenated vector of the hidden state at the previous time step and the input at the current time step, integrating historical and current information; The weight matrix represents the forget gate. The bias term representing the forget gate is learned by the model during training and is used to fit the relationship between the input and the gate. This represents the sigmoid activation function, and its output range is... This indicates how much information is retained. The closer the value is to 1, the more "information corresponding to the previous cell state is retained". The closer the value is to 0, the more "information at that position is forgotten".

[0045] B2. The input gate filters how much new information from the current input needs to be added to the cell state (long-term memory). This includes the input gate layer and candidate cell states. The input gate layer determines which values ​​will be updated, using the following formula: ,in This represents the output vector of the input gate. The value of each element indicates "what proportion of the information at the corresponding position in the current input should be allowed to be added to the cell state". The closer the value is to 1, the more important the new information at that position is, and more should be added; the closer the value is to 0, the less important the information is, and less or no should be added. This represents the weight matrix of the input gate. This represents the bias term of the input gate; the candidate cell state generates a new candidate information vector, ready to be added to the cell state, as shown in the formula: ,in Represents the candidate cell state vector. The weight matrix representing the candidate cell state. Bias terms representing the candidate cell state This represents the hyperbolic tangent function, and its output range is... ;

[0046] B3. Cellular state It is the long-term memory carrier of LSTM, updated by combining the results of the forget gate and the input gate. The update formula is: ⊙ ⊙ , where ⊙ represents element-wise multiplication (Hadamard product). This represents the information to be retained from the old cell state. This indicates information newly added to the cell state; it should be noted that the cell state... It integrates "selected old memories" and "selected new information" to achieve dynamic updates of long-term memory. It does not lose key patterns in history, but can also incorporate changes in current operating conditions in a timely manner, providing long-term and real-time memory support for subsequent gas concentration prediction.

[0047] B4. The output gate determines which information from the cell state will be output as the hidden state at the current time step. The formula is:

[0048] ;

[0049] ⊙ ;

[0050] in The activation value of the output gate is a vector that controls which information in the cell state needs to be output as the current hidden state. The closer the value is to 1, the easier it is for the information of the corresponding cell state to be output. This represents the weight matrix of the output gate. This represents the bias term of the output gate.

[0051] The early warning analysis module is used to issue early warnings based on the predicted concentration of hazardous gases at the next time point and the safe threshold for hazardous gas concentration. The specific process is as follows:

[0052] C1. Extract the predicted value of the hazardous gas concentration at the next future time point and compare it with the safe threshold of the hazardous gas concentration. Determine whether the hazardous gas concentration at the next future monitoring time point is within the range of 80% of the safe threshold of the hazardous gas concentration. If so, generate a first-level warning signal; if it is less than 80% of the safe threshold of the hazardous gas concentration, do not take any action; if it is greater than the safe threshold of the hazardous gas concentration, jump to C2.

[0053] C2. Calculate the concentration difference of harmful gas by subtracting the concentration of harmful gas at the next monitoring time point from the safe threshold of harmful gas concentration. Determine whether the concentration difference of harmful gas is less than the threshold of harmful gas concentration. If yes, generate a level 2 warning signal; otherwise, generate a level 3 warning signal.

[0054] C3. Generate early warning information based on the early warning signals, which include Level 1, Level 2, and Level 3 early warning signals;

[0055] Based on the warning signal, early warning information is generated, including: when a level one warning signal is received, the concentration of generated gas is showing an upward trend and is about to approach the safety threshold. Please check the ventilation equipment and adjust the tunneling pace. At the same time, linkage measures are taken: trigger the audible and visual alarm of the ground monitoring platform and push the prompt information to the intrinsically safe mining mobile phone and positioning card terminal of the workers to warn of risks in advance.

[0056] When a Level 2 warning signal is received, the concentration of generated gas has slightly exceeded the standard. The system has increased the ventilation volume and cut off the power to the tunneling machine. The alarm message asks on-site personnel to cooperate in handling the situation. At the same time, the following linkage measures are taken: the underground audible and visual alarm is activated, the air volume of the ventilation system is increased through the frequency converter, the tunneling machine control system cuts off the power to the cutting motor and transfer motor, and locks the operating handle to prevent accidental start.

[0057] When a Level 3 warning signal is received, indicating that the concentration of generated gas is severely exceeded and the diffusion range is large, an emergency alarm message is issued to immediately evacuate to the refuge chamber. Simultaneously, coordinated measures are implemented, including all measures for a Level 2 warning. The electromagnetic valves of the refuge chamber are automatically opened, the oxygen supply system is activated, and the oxygen concentration inside the chamber is maintained at 18.5%-23% to ensure the breathing needs of those seeking refuge are met. The optimal evacuation route based on the electronic map of the tunnel and real-time positioning is pushed to the underground display screen and personnel terminals.

[0058] It should be noted that by using an LSTM network to construct a hazardous gas concentration prediction model, and leveraging its gating mechanism and long-term memory characteristics, the model input integrates historical concentration, environmental, geological, and tunneling operation data. Through dynamic adjustment of the forget gate, input gate, and output gate, it accurately learns correlation patterns, improves prediction accuracy, and avoids false alarms and missed alarms caused by prediction bias. Based on the predicted hazardous gas concentration and safety threshold of the next monitoring time point output by the model, it provides graded early warnings and links corresponding emergency measures, rather than only issuing warnings after the current concentration exceeds the standard. This breaks through the traditional passive mode of relying solely on current concentration alarms, reserves a critical time window for emergency response, significantly improves the accuracy and scientific nature of early warnings, and achieves a leap from threshold alarms to trend prediction alarms.

[0059] The sensor sensitivity compensation determination module is used to determine whether the sensitivity of each gas concentration sensor needs compensation based on the real-time monitoring data of the internal operating environment, the external operating environment, and the detection log data. The specific process is as follows:

[0060] It should be noted that the gas concentration sensors include fixed-point gas sensors, inspection and monitoring point gas sensors, and mobile monitoring point gas sensors.

[0061] The system acquires real-time internal and external environmental data from various gas concentration sensors. The internal environmental data includes current, voltage, and temperature, while the external environmental data includes temperature, humidity, air pressure, dust concentration, wind speed, and tunneling vibration.

[0062] Using the monitoring time point as the horizontal axis and current, voltage and temperature as the vertical axis, we constructed the real-time monitoring current fluctuation curve, voltage fluctuation curve and temperature fluctuation curve of each gas concentration sensor, and extracted the number of current fluctuations, voltage fluctuations and temperature fluctuations from them respectively.

[0063] The ratio of the number of current fluctuations to the number of monitoring time points is calculated to generate the current fluctuation ratio; similarly, the voltage fluctuation ratio and temperature fluctuation ratio are obtained through analysis.

[0064] The average difference between the current corresponding to each current fluctuation number and the standard current corresponding to the sensor is calculated to generate the average difference value of current fluctuation; similarly, the average difference value of voltage fluctuation and the average difference value of temperature fluctuation are obtained.

[0065] The maximum value among the proportions of current fluctuation, voltage fluctuation, and temperature fluctuation is added to the maximum value among the average differences of current fluctuation, voltage fluctuation, and temperature fluctuation to generate the internal environmental impact value.

[0066] The maximum values ​​of each environmental index of each gas concentration sensor are extracted from the external environmental data of each monitoring time point in the real-time monitoring of each gas concentration sensor, and used as the representative values ​​of each environmental index.

[0067] The representative values ​​of each environmental indicator are compared with the corresponding reference values ​​for the detection interference of harmful gas concentrations. If the representative value of an environmental indicator is greater than its corresponding reference value for the detection interference of harmful gas concentrations, the environmental indicator is determined to be an environmental interference indicator. The number of environmental interference indicators is counted accordingly. The ratio of the number of environmental interference indicators to the number of environmental indicators is calculated to generate the proportion of environmental interference indicators. The interference degree corresponding to each environmental interference indicator is extracted (the difference between the representative value of the environmental interference indicator and its corresponding detection interference reference value). The cumulative result of the interference degrees corresponding to all environmental interference indicators is taken as the environmental interference degree.

[0068] The weighted sum of the proportion of environmental interference indicators and the degree of environmental interference is used to generate the value of external environmental impact of operation.

[0069] The detection gas concentration values ​​of each test are obtained from the real-time monitoring data of harmful gas concentrations by each gas concentration sensor, and compared with the set gas concentration ranges to obtain the detection values ​​of each gas concentration range.

[0070] The response time of each gas concentration sensor in real-time monitoring is obtained from the detection logs of each gas concentration sensor, and then the response time of each detection in each gas concentration range is filtered.

[0071] By plotting the detection order on the x-axis and the response time on the y-axis, a curve showing the change in detection response time for each gas concentration range is constructed, and the slope of the curve is extracted from the curve. and the length of the detection response time variation curve for each gas concentration range , where i is the number of each gas concentration interval, i=1,2,...a, and a is the total number of gas concentration intervals. The length of the curve located above the corresponding reference detection response time for each gas concentration interval is extracted from the database. According to the formula The detection deviation values ​​of each gas concentration sensor were calculated. , This represents an exponential function with the natural constant e as the base. The floor symbol indicates the rounding down symbol. This indicates the preset detection response delay growth rate;

[0072] The real-time monitoring values ​​of the internal and external environmental influences and the detection deviation values ​​of each gas concentration sensor are compared with the set threshold values ​​for the internal and external environmental influences and the detection deviation values.

[0073] If the value of the internal environmental impact monitored in real time by a gas concentration sensor is greater than the set threshold value of the internal environmental impact, an internal environmental impact command will be generated.

[0074] If the value of the external environmental impact monitored in real time by a gas concentration sensor exceeds the set threshold value, an external environmental impact command will be generated.

[0075] If the detection deviation value monitored in real time by a gas concentration sensor is greater than the set detection deviation value threshold, a detection deviation command is generated.

[0076] If any of the instructions is present, it is determined that the sensitivity of the gas concentration sensor needs to be compensated for.

[0077] If none of the instructions are available, it is determined that the gas concentration sensor sensitivity is not compatible and requires no compensation.

[0078] The sensor adaptive calibration module is used to perform sensor sensitivity compensation analysis when the sensitivity of a gas concentration sensor needs compensation due to attenuation. It generates the calibration sensitivity of the sensor, feeds it back to the corresponding sensor's calibration control terminal, performs appropriate calibration control, and obtains the harmful gas concentration at the next time point. The specific process is as follows:

[0079] When the sensitivity of a sensor decreases and compensation is needed, the influence value of the sensor's internal operating environment is extracted. External environmental impact values and detection deviation value Each of these is associated with a threshold value corresponding to the internal operating environment. Threshold for external environmental impact values and detection deviation threshold Comparison:

[0080] If only exists At that time, the difference between the threshold value of the internal environmental impact of the sensor and the actual internal environmental impact value is calculated to generate the difference in the internal environmental impact value. The ratio of the difference in the internal environmental impact value to the threshold value of the internal environmental impact is then calculated to generate the degree of difference in the internal environmental impact value. According to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. , This indicates the baseline correction guidance weight corresponding to the set internal operating environment;

[0081] If only exists At that time, the difference between the threshold value of the external environmental impact of the sensor and the actual value of the external environmental impact is calculated to generate the difference in the external environmental impact value. The ratio of the difference in the external environmental impact value to the threshold value of the external environmental impact is then calculated to generate the degree of difference in the external environmental impact value. According to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. , This represents the baseline correction guidance weight corresponding to the set external operating environment;

[0082] If only exists At that time, the difference between the sensor's detection deviation threshold and the detection deviation value is calculated to generate the detection deviation difference. The ratio between the detection deviation value and the detection deviation threshold is then calculated to generate the detection deviation difference degree. According to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. , This indicates the baseline correction guidance weight corresponding to the set detection deviation;

[0083] If it exists and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ;

[0084] If it exists and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ;

[0085] If it exists and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ;

[0086] If it exists , and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ;

[0087] Based on this analysis, the correction guidance weights for the calibration sensitivity of each gas concentration sensor are obtained. ,in or or or or or or .

[0088] The calibration sensitivity correction guide weight of the sensor is matched with the sensitivity compensation value corresponding to each set correction guide weight to obtain the calibration sensitivity compensation value of the sensor. The sum of the real-time monitoring set sensitivity of the sensor and the calibration sensitivity compensation value is used as the calibration sensitivity of the sensor.

[0089] It should be noted that by analyzing and calculating the internal and external environmental data and detection log data of the sensor, the impact values ​​of the sensor's internal and external environments, as well as the detection deviation value, are calculated. When any value exceeds the threshold, it is determined that "compensation is required". There is no need to rely on manual periodic calibration, which avoids the situation where the sensor sensitivity declines unnoticed. The correction guide weight is calculated according to the cause of sensor attenuation and the corresponding sensitivity compensation value is matched to ensure that the sensor maintains high accuracy in the complex environment of the mine, and avoids misleading safety decisions due to data distortion.

[0090] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0091] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A gas monitoring and early warning system for tunnel excavation, characterized in that, include: The prediction module is used to build a deep neural network model for predicting the concentration of harmful gases, and to predict the concentration of harmful gases at the next time point based on real-time monitoring of harmful gas concentrations. The early warning analysis module is used to issue early warnings based on the predicted concentration of harmful gases at the next time point and the safe threshold for the concentration of harmful gases. The sensor sensitivity compensation determination module is used to determine whether the sensitivity of each gas concentration sensor needs to be compensated based on the real-time monitoring data of the internal operating environment, the external operating environment, and the detection log data of each gas concentration sensor. The sensor adaptive calibration module is used to perform sensor sensitivity compensation analysis when the sensitivity of a gas concentration sensor needs to be compensated for, generate the sensor's calibration sensitivity, and feed back the calibration sensitivity to the corresponding sensor's calibration control terminal to perform corresponding calibration control and obtain the harmful gas concentration at the next time point.

2. The gas monitoring and early warning system for tunnel excavation as described in claim 1, characterized in that, The specific process of predicting the harmful gas concentration at the next time point based on real-time monitoring of harmful gas concentration is as follows: A1. Obtain historical hazardous gas concentration data and tunnel excavation construction area characteristic data from the historical database. The tunnel excavation construction area characteristic data includes environmental data, geological data, and tunneling operation data. Filter the data to obtain a sample dataset. A2. Perform preprocessing operations such as cleaning, standardization and feature engineering on the sample dataset to obtain the harmful gas concentration dataset, environmental dataset, geological dataset and tunneling operation dataset, and divide them into training set, test set and validation set with a ratio of 6:2:

2. A3. Construct a deep neural network model for predicting harmful gas concentration data. Initialize the model parameters and iteratively train the constructed deep neural network model using the training set. Starting from iteration number t=1, input the training data, and the network calculates the prediction output. Based on the error between the predicted value and the true value, adjust the network parameters in reverse to minimize the loss function. After each iteration, determine whether the deep neural network has converged. If not, proceed to t=t+1 and continue the next round of training iterations. If yes, calculate the test set error, use the pre-trained model to predict the test set data, calculate the error between the predicted value and the true value, and determine whether the current test set error is the optimal value. If yes, the current model is the optimal model. If not, readjust the model parameters, train and evaluate again, and finally obtain the trained deep neural network model for predicting harmful gas concentration. A4. The prediction is made by using a trained deep neural network model for predicting the concentration of harmful gases. The inputs are real-time monitored data on the concentration of harmful gases, environmental data, geological data, and tunneling data. The output is the predicted value of the concentration of harmful gases at the next time point.

3. The gas monitoring and early warning system for tunnel excavation as described in claim 2, characterized in that, The deep neural network model for predicting harmful gas concentrations uses an LSTM network model, which includes an input layer, an LSTM layer, and an output layer. The LSTM layer includes a forget gate, an input gate, an output gate, and a memory cell. The input sequence of the LSTM network is The input is fed into the LSTM layer, and the output sequence is... The output is from the LSTM layer, where... It is the forecast period. It is historical input data, including historical hazardous gas concentration data, environmental data, geological data, and tunneling operation data. It predicts the concentration of harmful gases; before the next monitoring time t+1 arrives, the input sequence will be... The input is fed into a trained deep neural network model for predicting harmful gas concentrations, and the output sequence is... This yields the predicted concentration of harmful gases at time t+1.

4. The gas monitoring and early warning system for tunnel excavation as described in claim 1, characterized in that, The process of issuing an early warning based on the predicted concentration of harmful gases at the next time point and the safe threshold for harmful gas concentration is as follows: C1. Extract the predicted value of the hazardous gas concentration at the next future time point and compare it with the safe threshold of the hazardous gas concentration. Determine whether the hazardous gas concentration at the next future monitoring time point is within the range of 80% of the safe threshold of the hazardous gas concentration. If so, generate a first-level warning signal; if it is less than 80% of the safe threshold of the hazardous gas concentration, do not take any action; if it is greater than the safe threshold of the hazardous gas concentration, jump to C2. C2. Calculate the concentration difference of harmful gas by subtracting the concentration of harmful gas at the next monitoring time point from the safe threshold of harmful gas concentration. Determine whether the concentration difference of harmful gas is less than the threshold of harmful gas concentration. If yes, generate a level 2 warning signal; otherwise, generate a level 3 warning signal. C3. Generate early warning information based on the early warning signal.

5. A gas monitoring and early warning system for tunnel excavation as described in claim 1, characterized in that, The specific process for determining whether the sensitivity of each gas concentration sensor needs compensation is as follows: Acquire real-time operating internal and external environmental data monitored by each gas concentration sensor. Construct real-time monitoring curves of current, voltage, and temperature fluctuations for each gas concentration sensor, extract the number of current fluctuations, voltage fluctuations, and temperature fluctuations, and calculate and generate the proportion of current fluctuations, the proportion of voltage fluctuations, and the proportion of temperature fluctuations, as well as the average difference of current fluctuations, the average difference of voltage fluctuations, and the average difference of temperature fluctuations. The maximum value among the proportions of current fluctuation, voltage fluctuation, and temperature fluctuation is added to the maximum value among the average differences of current fluctuation, voltage fluctuation, and temperature fluctuation to generate the internal environmental impact value. The maximum values ​​of each environmental indicator of each gas concentration sensor are extracted from the real-time monitoring data of the external environment of each gas concentration sensor and used as the representative values ​​of each environmental indicator. The representative values ​​of each environmental indicator are compared with the reference values ​​for the detection interference of harmful gas concentrations corresponding to each environmental indicator to determine the number of environmental interference indicators and calculate the proportion of environmental interference indicators. The interference degree corresponding to each environmental interference indicator is extracted and the cumulative result is used as the environmental interference degree. The proportion of environmental interference indicators and the environmental interference degree are weighted and summed to generate the value of external environmental impact of operation. The detection gas concentration values ​​of each detection are obtained from the real-time monitoring data of harmful gas concentrations by each gas concentration sensor, and compared with the set gas concentration ranges to obtain the detection values ​​of each gas concentration range; the response time of each detection in each gas concentration range is obtained, the detection response time variation curve of each gas concentration range is constructed, the slope of the curve, the length of the detection response time variation curve of each gas concentration range, and the length of the curve above the corresponding reference detection response time of each gas concentration range are extracted to generate the detection deviation value of each gas concentration sensor. If the internal environmental impact value monitored in real time by a gas concentration sensor is greater than the set internal environmental impact value threshold, an internal environmental impact command is generated; if the external environmental impact value is greater than the set external environmental impact value threshold, an external environmental impact command is generated; if the detection deviation value is greater than the set detection deviation value threshold, a detection deviation command is generated. If any of the instructions is present, it is determined that the sensitivity of the gas concentration sensor needs to be compensated for. If none of the instructions are available, it is determined that the gas concentration sensor sensitivity is not compatible and requires no compensation.

6. A gas monitoring and early warning system for tunnel excavation as described in claim 1, characterized in that, The specific process for generating the calibration sensitivity of the sensor is as follows: When the sensitivity of a sensor decreases and compensation is needed, the influence value of the sensor's internal operating environment is extracted. External environmental impact values and detection deviation value Each of these is associated with a threshold value corresponding to the internal operating environment. Threshold for external environmental impact values and detection deviation threshold Comparison: If only exists At that time, calculate the difference in the internal environmental impact values ​​generated during operation. According to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. , This indicates the baseline correction guidance weight corresponding to the set internal operating environment; If only exists At that time, calculate the difference in the external environmental impact values ​​generated during operation. According to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. , This represents the baseline correction guidance weight corresponding to the set external operating environment; If only exists At that time, the difference in the generated detection deviation value is calculated. According to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. , This indicates the baseline correction guidance weight corresponding to the set detection deviation; If it exists and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ; If it exists and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ; If it exists and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ; If it exists , and At that time, according to the formula The corrected steering weights for the sensor's calibration sensitivity are calculated. ; Based on this analysis, the correction guidance weights for the calibration sensitivity of each gas concentration sensor are obtained. ,in or or or or or or ; The calibration sensitivity correction guide weight of the sensor is matched with the sensitivity compensation value corresponding to each set correction guide weight to obtain the calibration sensitivity compensation value of the sensor. The sum of the real-time monitoring set sensitivity of the sensor and the calibration sensitivity compensation value is used as the calibration sensitivity of the sensor.

7. A gas monitoring and early warning system for tunnel excavation as described in claim 1, characterized in that, It also includes a data acquisition module and a data processing module. The data acquisition module is used to collect real-time monitoring data on harmful gas concentration, environmental data, geological data, and tunnel excavation operation data during the tunnel excavation process. The geological data includes the degree of fracture development and the coal seam permeability coefficient. The tunnel excavation operation data includes the excavation speed and cutting power. The process of acquiring the harmful gas concentration data and environmental data is as follows: fixed nodes are set up along the roadway at preset distance intervals, and each node contains a harmful gas concentration monitoring sensor and an environmental monitoring sensor to collect the harmful gas concentration data and environmental data at each fixed point. The tracked explosion-proof inspection robot is equipped with a laser gas sensor, an environmental monitoring sensor and a TOFD ultrasonic positioning module. It performs periodic inspections of areas not covered by fixed nodes according to a preset route. The inspection robot obtains the location coordinates of the tunneling machine in real time, automatically adjusts the inspection route, and obtains harmful gas concentration data and environmental data at each inspection and monitoring point. The tunneling machine integrates a miniaturized, explosion-proof spectral gas detection module and an environmental monitoring module. This mobile monitoring node, mounted on the tunneling machine, moves with the machine and directly monitors the gas in front of and around the tunneling cutting face in real time and in situ, obtaining harmful gas concentration data and environmental data at each mobile monitoring point.

8. A gas monitoring and early warning system for tunnel excavation as described in claim 7, characterized in that, The specific process for acquiring the geological data is as follows: Images of the coal face were acquired using an intrinsically safe borehole inspection instrument. A U-Net semantic segmentation network was then used to semantically segment the borehole images into fractured and intact coal bodies, distinguishing fractured regions. Contour detection and morphological operations were then performed to extract fracture density and aperture, which were compared with set first density thresholds, second density thresholds, first aperture thresholds, and second aperture thresholds to determine the fracture development level. Fracture development levels included sparse fracture development, moderate fracture development, and dense fracture development. Finally, the fracture development level was matched with the fracture development degree corresponding to each level to obtain the fracture development degree. When drilling in the coal seam to be tested, a special sealing device is used to seal the borehole. After sealing, the gas pressure change in the borehole is monitored. Once the pressure stabilizes, the borehole gas flow measurement device is turned on to continuously measure the gas emission at different times. By substituting borehole diameter, borehole depth, gas pressure gradient, and gas flow rate into the permeability coefficient calculation formula, and deriving it based on Darcy's law, the coal seam permeability coefficient is obtained through the radial flow formula.

9. A gas monitoring and early warning system for tunnel excavation as described in claim 8, characterized in that, The data processing module is used to clean and normalize the real-time data collected during the tunnel excavation process to obtain the processed data. The processed real-time hazardous gas concentration data of each fixed point, each inspection and monitoring point, and each mobile monitoring point are then fused and analyzed to obtain the real-time monitoring hazardous gas concentration during the tunnel excavation process.

Images