Method for eliminating influence of gas generated in underground paint spraying operation on precision of electrochemical sensor

Abstract

A method for eliminating the influence of gas generated in underground paint spraying operation on the precision of an electrochemical sensor comprises the following steps: firstly, designing an analysis experiment aiming at the influence degree of the gas generated in the paint spraying operation on an underground common electrochemical gas monitoring sensor, and supplementing the blank of the research; meanwhile, the influence quantity of gas generated in the paint spraying operation on the gas sensors along the road in the flowing path and process is monitored by using a mature underground tracer gas method for detecting the wind direction and the air leakage quantity; finally, on the basis of an artificial intelligence algorithm, the influenced gas sensors are corrected in a targeted manner, the authenticity of data of each gas monitoring sensor is improved, and the method has the advantages of correcting the monitoring data of the electrochemical gas sensors, comprehensively evaluating the safety condition of a coal mine and timely discovering and processing potential safety hazards. And the device plays an important role in ensuring the life health and safety of miners.

Description

Technical Field

[0001] The invention relates to a method for eliminating the influence of underground operation gas on the accuracy of underground sensors, and specifically to a method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors, belonging to the technical field of coal mine safety regulation. Background Art

[0002] In the special and dangerous working environment of coal mines, the monitoring of gases such as O2, CO, SO2, H2S, CH4, etc. has always been an indispensable part of coal mine safety production. However, these gas sensors are mostly made based on electrochemical principles, which will cause cross-interference in sensor monitoring. Cross-interference means that the sensor will also respond to interfering gases other than the target gas, resulting in reading errors (too high or too low). According to the characteristics of the interfering gas reaction in the sensor, it can be divided into negative cross-interference or positive cross-interference; positive cross-interference will cause false high readings and early response to produce false alarms; while negative cross-interference reduces the response signal of the target gas, resulting in delayed alarms, which is not conducive to ensuring environmental safety.

[0003] Gases such as formaldehyde, benzene, and ketones are produced during the production process of coal mines due to the need to spray paint on metal structures for corrosion protection, use spray paint to mark road signs and safety warnings, and spray paint maintenance on equipment. The content is relatively small and will not directly cause suffocation or even death. Therefore, coal mines pay little attention to these gases and generally do not install corresponding gas monitoring sensors. However, these gases will affect the monitoring accuracy of gas sensors based on electrochemical principles such as CO and H2S in normal coal mines, and it is generally difficult to remove these gases in coal mines.

[0004] At present, the detection of various types of gases in coal mines mainly relies on various instruments and equipment, but these existing technologies still have a series of problems and shortcomings. For dust detection, although the optical detection method has a good effect on detection accuracy, the existing optical detection equipment usually needs to transport dust particles to the detection instrument through a pipeline, which not only increases the complexity of the system, but more importantly, dust pollution and blockage are prone to occur inside the pipeline, resulting in deviations in the detection data and affecting the accuracy and stability of the detection results. In terms of methane detection, although the existing infrared spectroscopy and catalytic combustion methods can achieve relatively accurate detection, these methods usually require complex instruments and equipment and external power support, which increases the size and weight of the equipment and is not conducive to application in complex and narrow environments such as coal mines. In addition, these detection methods usually only target a single monitoring target and cannot achieve comprehensive monitoring of multiple hazardous gases, which to a certain extent limits their application effect in coal mine safety monitoring.

[0005] There are related studies on the cross-interference of electrochemical sensors in the prior art. The methods currently used are mainly through: 1) selecting characteristic electrode materials and improving catalysts to improve the selectivity of the sensor to the target gas; 2) applying a "bias" voltage so that gases with lower electrochemical reaction activity can also be oxidized or reduced, improving sensor sensitivity and response time to improve cross-interference; 3) built-in chemical filters, through chemical filters to react or adsorb non-target gases, selectively remove non-target gases to improve measurement accuracy; 4) built-in cross-interference coefficient automatic compensation methods to reduce or solve the sensor's response to non-target gases. However, these methods are suitable for instrument R&D personnel to calibrate sensor accuracy, and have certain technical barriers. These correction methods are difficult to implement in actual mine work. On the other hand, in the existing cross-interference research on gas sensors, most of them consider the research on CO electrochemical sensors being affected by H2S, NO2, H2, etc., and the research on formaldehyde electrochemical sensors being affected by NH3, alcohol, temperature and humidity, etc., but there are few studies on the mutual influence and degree of electrochemical sensors between NO2, CO, SO2, H2S, etc. and volatile substances produced by paint such as formaldehyde, benzene, ketones, etc., although it is known that formaldehyde sensors are indeed affected by gases such as H2S and CO.

[0006] The existing technology for monitoring and treating formaldehyde and the like still mainly targets the volatilization of formaldehyde and other gases in building spaces and coating materials. There are generally two methods for removing such gases. One is to remove such gases by adsorbing them using materials such as activated carbon. However, only a limited amount of gas can be adsorbed, and there is still a possibility that the gas will escape again after adsorption. The other is to remove them by photocatalyst method, which uses photocatalyst to produce hydroxyl radicals after being illuminated by natural light, and reacts with organic matter in the air to produce non-toxic inorganic substances. This method can effectively decompose formaldehyde, benzene, ammonia, etc., and convert them into H2O and CO2, which plays the role of air purification. However, this method requires natural light irradiation and is not suitable for underground coal mines. Although there is artificial natural light, that is, by using high-power halogen lamps, xenon lamps or LED arrays as light sources, it can emit a spectral distribution similar to that of sunlight to simulate the radiation characteristics of sunlight, but considering whether it will introduce new explosion hazards to the coal mine underground and whether it can pass the coal safety certification, this method is difficult to implement. Summary of the invention

[0007] The purpose of the present invention is to provide a method for eliminating the influence of gases generated by underground paint spraying operations on the accuracy of electrochemical sensors, which can eliminate the influence of gases such as formaldehyde generated by mine paint spraying operations on commonly used electrochemical gas monitoring sensors underground, and calibrate the electrochemical sensors in underground tunnels that are affected by gases such as formaldehyde to improve the safety of underground coal mines.

[0008] In order to achieve the above-mentioned purpose, the present invention provides a method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors. The hardware part for implementing the method includes an experimental box, which is sealed by a top box cover when in use, and a paint spraying board is arranged inside the experimental box and fits with the bottom surface, and the size of the paint spraying board is consistent with the size of the bottom surface of the experimental box; the air permeable mesh is detachably placed inside the experimental box and installed at a certain distance from the paint spraying board, and the size of the air permeable mesh is consistent with the size of the paint spraying board; a temperature and humidity sensor, a formaldehyde sensor, and a mining series sensor are installed on the air permeable mesh, and the temperature and humidity sensor, the formaldehyde sensor, and the mining series sensor are electrically connected to the STM32 development board outside the experimental box through the through holes on the side wall of the experimental box, and the STM32 development board is electrically connected to a PC computer and powered by a power module;

[0009] The method comprises the following steps:

[0010] Step 1: Design a gas monitoring experiment for painting operation to find out the influence of gas generated by painting operation on the accuracy of electrochemical sensor;

[0011] Step 2: Based on the tracer gas, the location of the gas sensor in the well that will be affected by the gas from the painting operation and the degree of its impact are known;

[0012] Step 3: Calibrate the corresponding electrochemical gas sensor by using a machine learning algorithm, etc.

[0013] The step 1 specifically comprises the following steps:

[0014] Step 1.1: Connect the above hardware parts and test whether they can work normally. Then calibrate the temperature and humidity sensor, formaldehyde sensor, and mining series sensor to zero point. When the formaldehyde concentration is 0, read the values ​​of the temperature and humidity sensor, formaldehyde sensor, and mining series sensor.

[0015] Step 1.2: Spray paint on the surface of the spray-painted board, and quickly place it at the bottom of the experimental box, and seal it with the box cover to create a closed environment;

[0016] Step 1.3: Use the STM32 development board to read the values ​​of the temperature and humidity sensor, formaldehyde sensor, and mining series sensor in real time and record them;

[0017] Step 1.4: Stop the experiment when there is no significant change in formaldehyde concentration or the change in concentration is less than 1% of the total concentration;

[0018] Step 1.5: Adjust the amount of paint sprayed on the spray paint board and conduct multiple experiments. Optionally, open the box cover to simulate the concentration change of formaldehyde in a natural volatilization state. The formaldehyde concentration change is measured in real time by a formaldehyde sensor.

[0019] The step 2 specifically includes the following steps:

[0020] Step 2.1: Before the experiment, based on the mine ventilation system, obtain or solve the basic mine information such as characteristic parameters, air volume, and airflow path of each ventilation tunnel, and analyze the location of gas sensors that will be affected by the paint spraying gas underground;

[0021] Step 2.2: Select a spray painting operation area in the mine and conduct a SF6 tracer gas release experiment using the constant release method;

[0022] Step 2.3: Use the constant release method to continuously release a certain concentration for 1 minute at the painting location. Arrange the SF6 tracer gas detector at the location of the gas sensor affected by the painting operation gas. The monitoring will obtain a concentration-time curve, which can indicate how long the location of the gas sensor affected by the painting operation gas will monitor the tracer gas and its corresponding concentration during the release of SF6 tracer gas.

[0023] Step 2.4: Record the location and initial concentration of the painting operation area, the location of each gas sensor affected by the painting operation gas, the distance between the gas sensor and the location where the SF6 tracer gas is released, and the time and concentration of the SF6 tracer gas detected;

[0024] Step 2.5: Based on the monitored data, the numerical simulation method can be used to show the complete process of the migration and distribution of the tracer gas in the mine. The fluent simulation can be used to calculate the position and distribution of the entire tracer gas flowing through the tunnel. Considering that the wind flow is generally turbulent during ventilation in coal mines, the time-averaged method is used to regard the turbulent flow as the superposition of an average fluid and a pulsating fluid. The standard k-ε equation can be used in fluent to solve it. In addition, the convective mass transfer of the tracer gas also needs to consider the component transport, and the mixed gas of SF6 tracer gas and air also needs the law of conservation of mass and momentum. Based on the above theoretical formulas, a model is established in the fluent software for solution, which can intuitively obtain the migration of SF6 tracer gas in the tunnel and the cross-sectional concentration distribution at the position of the gas sensor affected by the paint spraying gas.

[0025] Step 2.6: Change the concentration of the initially released SF6 tracer gas and conduct multiple experiments;

[0026] The step 3 performs calibration on the corresponding electrochemical gas sensor. Based on the above experiment on the influence of gas generated by paint spraying on the commonly used electrochemical gas monitoring sensor in the mine, the concentration of formaldehyde, temperature, humidity, and concentration readings of various gas sensors for mining are obtained. After data preprocessing (removing outliers, filling missing values, data normalization, etc.), the XGBoost concentration prediction model for different electrochemical gas sensors for mining is trained, and the data training time series prediction model is obtained according to the experiment of SF6 tracer gas. The specific processing process of the concentration prediction model is as follows:

[0027] Step 3.1: For different gas sensors, the formaldehyde concentration value, temperature, and humidity used to predict the concentration of the electrochemical gas sensor are used as independent variables, and the monitoring value of a certain gas sensor is used as the dependent variable. Three-quarters of the sample data are randomly selected to train the XGBoost concentration prediction model, and the remaining quarter of the sample data is used as the prediction sample data to evaluate and verify the constructed XGBoost concentration prediction model;

[0028] Step 3.2: Train the XGBoost concentration prediction model based on the training data, where the formaldehyde concentration value, temperature, and humidity of the training sample are used as independent variables, and the dust mass concentration value actually measured by each electrochemical gas sensor is used as the dependent variable. The relationship between the formaldehyde concentration value, temperature, humidity and the monitoring value of each gas sensor is solved, and each gas sensor corresponds to an XGBoost concentration prediction model;

[0029] Step 3.3: Input the formaldehyde concentration value, temperature, and humidity of the predicted sample as input variables into the established XGBoost concentration prediction model to obtain the corresponding predicted value monitored by the gas sensor;

[0030] Step 3.4: Combined with the actual measured gas sensor monitoring value, the predicted value of the gas sensor monitoring obtained in step 3.3 is evaluated, and the mean square error of the two is calculated. Assuming that the deviation between the two exceeds the acceptable range, the parameter value of the XGBoost concentration prediction model is adjusted and re-trained and predicted until it meets the acceptable range, thereby determining the XGBoost concentration prediction model and obtaining the relationship between formaldehyde concentration value, temperature, humidity and different gas sensor monitoring values.

[0031] In the present invention, the positions of the temperature and humidity sensor, the formaldehyde sensor and the mining series sensor on the air-permeable net are not fixed, so as to eliminate the measurement error caused by the change of the sensor position.

[0032] The present invention installs a humidifier on the side of the experimental box to spray the inside of the box, a small fan supplies air to accelerate air convection and help the air to mix evenly to simulate the humidity of the mine, and a heating plate is used to slightly heat the bottom of the experimental box to simulate the temperature of the mine.

[0033] The gas flow equation of SF6 tracer gas in the coal mine in step 2 of the present invention can be expressed as:

[0034]

[0035] Where: Q is the conserved variable, ρ is the mass density, ρu, ρv, ρw are the momentum densities in three directions respectively, ρE is the energy density, is the boundary of the control volume, F c is the inviscid flux, F v is the viscous flux;

[0036] in:

[0037] In step 2.5 of the present invention, when considering the ventilation of coal mine underground, the wind flow is generally in a turbulent state. The turbulent flow is regarded as the superposition of an average fluid and a pulsating fluid by using the time-averaged method, and the turbulent flow motion can be expressed as:

[0038]

[0039] The above are the transient continuity equation, the Reynolds time-averaged NS equation and the component transport equation;

[0040] where u i ' and φ' are the speed u i And the pulsation value of the scalar φ, which can be solved by the standard k-ε equation in FLUENT;

[0041] In addition, the convective mass transfer of SF6 tracer gas also needs to consider the component transport. The mixed gas of SF6 tracer gas and air also needs the law of conservation of mass and momentum. For components without chemical reactions, the law of conservation of mass can be expressed as:

[0042]

[0043] Where: c s is the concentration of SF6 tracer gas;

[0044] U is the fluid vector velocity, also known as wind speed;

[0045] D s is the diffusion coefficient of the gas;

[0046] S s is the source term;

[0047] The calculation of the diffusion coefficient of SF6 tracer gas in air in the experiment can be used:

[0048]

[0049] Where: r is the hydraulic diameter of the tunnel, m;

[0050] f is the roadway friction coefficient;

[0051] Based on the above theoretical formula, a model was established in the fluent software and solved, which can intuitively obtain the migration of SF6 tracer gas in the tunnel and the cross-sectional concentration distribution at the gas sensor position affected by the paint operation gas.

[0052] In step 3 of the present invention, the relationship between the formaldehyde concentration value, temperature, humidity and the monitoring values ​​of different gas sensors solved based on the XGBoost algorithm is used. The time and concentration at the gas sensor position affected by the paint spraying gas measured by the SF6 tracer gas experiment are used to establish a time series prediction model. The specific processing process of the time series prediction model is as follows:

[0053] ① Discretization of the formaldehyde release concentration curve. Considering that formaldehyde release is a continuous process, assuming that the concentration of formaldehyde release does not change in a short period of time, the total concentration curve of formaldehyde release is divided into many small segments, and the formaldehyde concentration in each small segment remains unchanged;

[0054] ②Data cleaning and preprocessing: According to the SF6 tracer gas experiment, the time and concentration of each small section of the concentration transmitted to the gas sensor position affected by the paint spraying gas are obtained, which are used to train the formaldehyde propagation model and clean the original data, including removing outliers, filling missing values, and normalizing the data;

[0055] ③ Establish an LSTM time series prediction model, divide the data into a training set and a prediction set at a ratio of 4:1, use LSTM to fit the concentration changes of formaldehyde at each gas sensor location, and verify it on the prediction set to evaluate the model performance. The size of each parameter of the model can be adjusted to achieve the optimal performance of the model;

[0056] ④ Based on the results of LSTM training, the predicted formaldehyde concentration value per second at each gas sensor location is obtained, and the aforementioned XGBoost model is called to predict the influence of formaldehyde on a certain gas at the gas sensor location.

[0057] Compared with the prior art, the present invention aims to fill the gap of the impact of gases such as formaldehyde generated by existing painting operations on the commonly used electrochemical gas monitoring sensors in underground tunnels, which has not been evaluated. Based on the results, the electrochemical sensors in underground tunnels that are affected by gases such as formaldehyde are corrected to improve the efficiency and accuracy of coal mine safety monitoring. First, an analytical experiment on the impact of gases generated by painting operations on commonly used electrochemical gas monitoring sensors in underground tunnels is designed to fill the gap in this research; at the same time, the mature method of measuring air leakage direction and air leakage volume with tracer gas in underground tunnels is used to monitor the impact of gases generated by painting operations on gas sensors along the path and process of flow; finally, based on artificial intelligence algorithms, the affected gas sensors are corrected in a targeted manner to improve the authenticity of the data of each gas monitoring sensor, which plays an important role in correcting the monitoring data of electrochemical gas sensors, comprehensively evaluating the safety status of coal mines, timely discovering and dealing with potential safety hazards, and ensuring the life, health and safety of miners. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a structural schematic diagram of the experimental box of the present invention.

[0059] In the figure: 1. Box cover, 2. Experimental box, 3. STM32 development board, 4. PC computer, 5. Breathable mesh, 6. Painted board, 7. Through hole, 8. Mining series sensor, 9. Formaldehyde sensor, 10. Temperature and humidity sensor. DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings.

[0061] like Figure 1 As shown, a method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensor, the hardware part of realizing this method includes an experimental box 2, the experimental box is sealed by a top box cover 1 when in use, a paint plate 6 is arranged inside the experimental box and fits with the bottom surface, and the size of the paint plate 6 is consistent with the size of the bottom surface of the experimental box; the air permeable mesh 5 is detachably placed inside the experimental box and installed at a certain distance from the paint plate 6, and the size of the air permeable mesh 5 is consistent with the size of the paint plate 6; a temperature and humidity sensor 10, a formaldehyde sensor 9, and a mining series sensor 8 are installed on the air permeable mesh 5, and the temperature and humidity sensor 10, the formaldehyde sensor 9, and the mining series sensor 8 are all electrically connected to the STM32 development board 3 outside the experimental box through the through hole 7 on the side wall of the experimental box. During the experiment, the through hole 7 is wrapped with a sealing ring to ensure that the experimental box 2 is in a sealed state, the STM32 development board 3 is electrically connected to the PC computer 4 and powered by the power module, and the STM32 development board uses its own signal transmission module to upload the data of each sensor to the PC computer 4;

[0062] The direct current power supply for supplying power to the STM32 development board of the present invention can be connected to a 220V household circuit through a 12V / 1A power adapter for power supply, and can be connected to a computer or other USB devices through a USB interface for power supply.

[0063] The method comprises the following steps:

[0064] Step 1: Design a gas monitoring experiment for painting operation to find out the influence of gas generated by painting operation on the accuracy of electrochemical sensor;

[0065] Step 2: Based on the tracer gas, the location of the gas sensor in the well that will be affected by the gas from the painting operation and the degree of its impact are known;

[0066] Step 3: Calibrate the corresponding electrochemical gas sensor by using a machine learning algorithm, etc.

[0067] The step 1 specifically comprises the following steps:

[0068] Step 1.1: Connect the above hardware parts and test whether they can work normally, then calibrate the temperature and humidity sensor 10, formaldehyde sensor 9, and mining series sensor 8 to zero point, and read the values ​​of the temperature and humidity sensor 10, formaldehyde sensor 9, and mining series sensor 8 when the formaldehyde concentration is 0;

[0069] Step 1.2: Spray paint on the surface of the spray paint board 6, and quickly place it at the bottom of the experimental box, and seal it with the box cover 1 to create a closed environment;

[0070] Step 1.3: Use the STM32 development board 3 to read the values ​​of the temperature and humidity sensor 10, the formaldehyde sensor 9, and the mining series sensor 8 in real time and record them;

[0071] Step 1.4: Stop the experiment when there is no significant change in formaldehyde concentration or the change in concentration is less than 1% of the total concentration;

[0072] Step 1.5: Adjust the amount of paint sprayed on the spray paint board 6 and conduct multiple experiments. Optionally, open the box cover 1 to simulate the concentration change of formaldehyde in a natural volatilization state. Opening the box cover 1 here can better simulate a formaldehyde concentration change environment under real air circulation. The formaldehyde concentration change of the present invention is measured in real time by the formaldehyde sensor 9;

[0073] During the above experiment, considering that indoor formaldehyde will cause people to feel obvious irritation when it exceeds 1.2 mg / L, the experiment to analyze the impact of gas generated by paint spraying on the commonly used electrochemical gas monitoring sensors underground is best carried out under a fume hood or in an open area with good ventilation.

[0074] The step 2 specifically includes the following steps:

[0075] Step 2.1: Before the experiment, based on the mine ventilation system, obtain or solve the basic mine information such as characteristic parameters, air volume, and airflow path of each ventilation tunnel. The mine ventilation system can be consulted with the coal mine management personnel to know the mine air volume, the direction of the mine airflow, and whether it is positive pressure or negative pressure, and analyze the location of the gas sensor that will be affected by the paint spraying gas underground;

[0076] Step 2.2: Select the spray painting operation area in the mine to conduct the SF6 tracer gas release experiment. There are two types of SF6 tracer gas: constant release method and instantaneous release method. Since the air volume at the working face of the mine is large, the wind flow is often in a turbulent state, and the diffusion coefficient of the tracer gas is also large. The concentration received by the instantaneous release method is generally small; at the same time, due to the sampling frequency of the instrument, the concentration-time curve obtained by the instantaneous release method is often incomplete, so the constant release method is selected for the experiment;

[0077] Step 2.3: Use the constant release method to continuously release a certain concentration for 1 minute at the painting location. Arrange the SF6 tracer gas detector at the location of the gas sensor affected by the painting operation gas. The monitoring will obtain a concentration-time curve, which can indicate how long the location of the gas sensor affected by the painting operation gas will monitor the tracer gas and its corresponding concentration during the release of SF6 tracer gas.

[0078] Step 2.4: Record the location and initial concentration of the painting operation area (SF6 tracer gas release), the location of each gas sensor affected by the painting operation gas, the distance between the gas sensor and the SF6 tracer gas release location, and the time and concentration of the SF6 tracer gas detected;

[0079] Step 2.5: Based on the monitored data, the numerical simulation method can be used to show the complete process of the migration and distribution of the tracer gas in the mine. The fluent simulation can be used to calculate the position and distribution of the entire tracer gas flowing through the tunnel. Considering that the wind flow is generally turbulent during ventilation in coal mines, the time-averaged method is used to regard the turbulent flow as the superposition of an average fluid and a pulsating fluid. The standard k-ε equation can be used in fluent to solve it. In addition, the convective mass transfer of the tracer gas also needs to consider the component transport, and the mixed gas of SF6 tracer gas and air also needs the law of conservation of mass and momentum. Based on the above theoretical formulas, a model is established in the fluent software for solution, which can intuitively obtain the migration of SF6 tracer gas in the tunnel and the cross-sectional concentration distribution at the position of the gas sensor affected by the paint spraying gas.

[0080] Step 2.6: Change the concentration of the initially released SF6 tracer gas and conduct multiple experiments. Through multiple independent experiments, multiple experimental results are obtained to expand the data set for subsequent data analysis.

[0081] The step 3 calibrates the corresponding electrochemical gas sensor. Based on the above experiment on the influence of gas generated by paint spraying on the commonly used electrochemical gas monitoring sensor in the mine, the concentration of formaldehyde, temperature, humidity, and concentration readings of various gas sensors for mining are obtained. After data preprocessing (including removing outliers, filling missing values, data normalization, etc.), the XGBoost concentration prediction model for different electrochemical gas sensors for mining is trained, and the data training time series prediction model is obtained according to the experiment of SF6 tracer gas. The specific processing process of the concentration prediction model is as follows:

[0082] Step 3.1: For different gas sensors, the formaldehyde concentration value, temperature, and humidity used to predict the concentration of the electrochemical gas sensor are used as independent variables, and the monitoring value of a certain gas sensor is used as the dependent variable. Three-quarters of the sample data are randomly selected to train the XGBoost concentration prediction model, and the remaining quarter of the sample data is used as the prediction sample data to evaluate and verify the constructed XGBoost concentration prediction model;

[0083] Step 3.2: Train the XGBoost concentration prediction model based on the training data, where the formaldehyde concentration value, temperature, and humidity of the training sample are used as independent variables, and the dust mass concentration value actually measured by each electrochemical gas sensor is used as the dependent variable. The relationship between the formaldehyde concentration value, temperature, humidity and the monitoring value of each gas sensor is solved, and each gas sensor corresponds to an XGBoost concentration prediction model;

[0084] Step 3.3: Input the formaldehyde concentration value, temperature, and humidity of the predicted sample as input variables into the established XGBoost concentration prediction model to obtain the corresponding predicted value monitored by the gas sensor;

[0085] Step 3.4: Combined with the actual measured gas sensor monitoring value, the predicted value of the gas sensor monitoring obtained in step 3.3 is evaluated, and the mean square error of the two is calculated. Assuming that the deviation between the two exceeds the acceptable range, the parameter value of the XGBoost concentration prediction model is adjusted and re-trained and predicted until it meets the acceptable range, thereby determining the XGBoost concentration prediction model and obtaining the relationship between formaldehyde concentration value, temperature, humidity and different gas sensor monitoring values.

[0086] The positions of the temperature and humidity sensor 10, the formaldehyde sensor 9, and the mining series sensor 8 on the air permeable mesh 5 are not fixed to eliminate measurement errors caused by changes in the sensor positions.

[0087] A humidifier is installed on the side of the experimental box to spray the inside of the box. A small fan supplies air to accelerate air convection and help the air mix evenly to simulate the humidity of the mine. A heating plate is used to slightly heat the bottom of the experimental box to simulate the temperature of the mine.

[0088] The gas flow equation of SF6 tracer gas in coal mine in step 2 can be expressed as:

[0089]

[0090] Where: Q is the conserved variable, ρ is the mass density, ρu, ρv, ρw are the momentum densities in three directions respectively, ρE is the energy density, is the boundary of the control volume, F c is the inviscid flux, F v is the viscous flux;

[0091] in:

[0092] In step 2.5, when considering the ventilation of coal mine underground, the wind flow is generally in a turbulent state. The time-averaged method is used to regard the turbulent flow as the superposition of an average fluid and a pulsating fluid, and the turbulent flow motion can be expressed as:

[0093]

[0094] The above are the transient continuity equation, the Reynolds time-averaged NS equation and the component transport equation;

[0095] where u i ' and φ' are the speed u i And the pulsation value of the scalar φ, which can be solved by the standard k-ε equation in FLUENT;

[0096] In addition, the convective mass transfer of SF6 tracer gas also needs to consider the component transport. The mixed gas of SF6 tracer gas and air also needs the law of conservation of mass and momentum. For components without chemical reactions, the law of conservation of mass can be expressed as:

[0097]

[0098] Where: c s is the concentration of SF6 tracer gas;

[0099] U is the fluid vector velocity, also known as wind speed;

[0100] D s is the diffusion coefficient of the gas;

[0101] S s is the source term;

[0102] The calculation of the diffusion coefficient of SF6 tracer gas in air in the experiment can be used:

[0103]

[0104] Where: r is the hydraulic diameter of the tunnel, m;

[0105] f is the roadway friction coefficient;

[0106] Based on the above theoretical formula, a model was established in the fluent software and solved, which can intuitively obtain the migration of SF6 tracer gas in the tunnel and the cross-sectional concentration distribution at the gas sensor position affected by the paint operation gas.

[0107] In step 3, the relationship between formaldehyde concentration, temperature, humidity and monitoring values ​​of different gas sensors solved by XGBoost algorithm is used. The time and concentration of the gas sensor position affected by the paint spraying gas measured by SF6 tracer gas experiment are used to establish a time series prediction model. The specific processing process of the time series prediction model is as follows:

[0108] ① Discretization of the formaldehyde release concentration curve. Considering that formaldehyde release is a continuous process, assuming that the concentration of formaldehyde release does not change in a short period of time, the total concentration curve of formaldehyde release is divided into many small segments, and the formaldehyde concentration in each small segment remains unchanged;

[0109] ②Data cleaning and preprocessing: According to the SF6 tracer gas experiment, the time and concentration of each small section of the concentration transmitted to the gas sensor position affected by the paint spraying gas are obtained, which are used to train the formaldehyde propagation model and clean the original data, including removing outliers, filling missing values, and normalizing the data;

[0110] ③ Establish an LSTM time series prediction model, divide the data into a training set and a prediction set at a ratio of 4:1, use LSTM to fit the concentration changes of formaldehyde at each gas sensor location, and verify it on the prediction set to evaluate the model performance. The size of each parameter of the model can be adjusted to achieve the optimal performance of the model;

[0111] ④ Based on the results of LSTM training, the predicted formaldehyde concentration value per second at each gas sensor location is obtained, and the aforementioned XGBoost model is called to predict the influence of formaldehyde on a certain gas at the gas sensor location.

[0112] Optionally, the XGBoost algorithm of the present invention may also be replaced with other machine learning algorithms such as random forest, LightGBM, etc., or other deep learning algorithms, and the calculation accuracy and loss of all feasible algorithms may be comprehensively compared to ultimately determine which algorithm to use.

[0113] Alternatively, the LSTM algorithm can also be replaced by other time series models such as convolutional neural networks (CNN).

[0114] For example, the XGBoost model predicts that the measurement accuracy of the CO sensor will be affected by formaldehyde as follows:

[0115] When the formaldehyde concentration is 1 mg / m 3 When the CO sensor reads +0.01 mg / m 3 LSTM predicts that when formaldehyde releases 10 mg at the initial point and reaches a certain CO sensor installation location 20 seconds later, the predicted formaldehyde concentration at this time is 1 mg / m 3 , the XGBoost model is called to calculate that the CO sensor reading impact value is +0.01mg / m 3 If the actual CO sensor reading is 0.06 mg / m 3 , then the actual CO concentration here is considered to be 0.05 mg / m 3 .

Claims

1. A method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors, wherein the hardware part for implementing the method comprises an experimental box (2), wherein the experimental box (2) is sealed by a top box cover (1) when in use, a paint spraying board (6) is arranged inside the experimental box and fits with the bottom surface, and the size of the paint spraying board (6) is consistent with the size of the bottom surface of the experimental box; a ventilating mesh (5) is detachably placed inside the experimental box and installed at a certain distance from the paint spraying board (6), and the size of the ventilating mesh (5) is consistent with the size of the paint spraying board (6); a temperature and humidity sensor (10), a formaldehyde sensor (9), and a mining series sensor (8) are installed on the ventilating mesh (5), and the temperature and humidity sensor (10), the formaldehyde sensor (9), and the mining series sensor (8) are all electrically connected to an STM32 development board (3) outside the experimental box through a through hole on the side wall of the experimental box, and the STM32 development board (3) is electrically connected to a PC (4) and powered by a power module; It is characterized in that The method comprises the following steps: Step 1: Design a gas monitoring experiment for painting operation to find out the influence of gas generated by painting operation on the accuracy of electrochemical sensor; Step 2: Based on the tracer gas, the location of the gas sensor in the well that will be affected by the gas from the painting operation and the degree of its impact are known; Step 3: Calibrate the corresponding electrochemical gas sensor by using a machine learning algorithm, etc. The step 1 specifically comprises the following steps: Step 1.1: Connect the above hardware parts and test whether they can work normally, then calibrate the temperature and humidity sensor (10), formaldehyde sensor (9), and mining series sensor (8) to zero point, and read the values ​​of the temperature and humidity sensor (10), formaldehyde sensor (9), and mining series sensor (8) when the formaldehyde concentration is 0; Step 1.2: spray paint on the surface of the spray paint plate (6), and quickly place it at the bottom of the experimental box, and seal it with the box cover (1); Step 1.3: Use the STM32 development board (3) to read the values ​​of the temperature and humidity sensor (10), formaldehyde sensor (9), and mining series sensor (8) in real time and record them; Step 1.4: Stop the experiment when there is no significant change in formaldehyde concentration or the change in concentration is less than 1% of the total concentration; Step 1.5: adjusting the amount of paint sprayed on the spray paint panel (6) and conducting multiple experiments, and optionally, opening the box cover (1) to simulate the concentration change of formaldehyde in a natural volatilization state, and the formaldehyde concentration change is measured in real time by the formaldehyde sensor (9); The step 2 specifically includes the following steps: Step 2.1: Before the experiment, based on the mine ventilation system, obtain or solve the characteristic parameters, air volume, and basic mine information of the airflow path of each ventilation tunnel, and analyze the location of the gas sensors that will be affected by the paint spraying gas in the mine; Step 2.2: Select a spray painting operation area in the mine and conduct SF6 tracer gas release experiment by constant release method; Step 2.3: Use the constant release method to continuously release a certain concentration for 1 minute at the painting location. Arrange the SF6 tracer gas detector at the location of the gas sensor affected by the painting operation gas. The monitoring will obtain a concentration-time curve, which can indicate how long the location of the gas sensor affected by the painting operation gas will monitor the tracer gas and its corresponding concentration during the release of SF6 tracer gas. Step 2.4: Record the location and initial concentration of the painting operation area, the location of each gas sensor affected by the painting operation gas, the distance between the gas sensor and the location where the SF6 tracer gas is released, and the time and concentration of the SF6 tracer gas detected; Step 2.5: Based on the monitored data, numerical simulation is used to demonstrate the complete process of tracer gas migration and distribution in the mine. Fluent simulation is used to calculate the position and distribution of the entire tracer gas flowing through the tunnel. Considering that the wind flow is generally turbulent during ventilation in coal mines, the time-averaged method is used to regard the turbulent flow as the superposition of an average fluid and a pulsating fluid. The standard k-ε equation is used in Fluent to solve it. In addition, the convective mass transfer of the tracer gas also needs to consider the component transport, and the mixed gas of SF6 tracer gas and air also needs the laws of conservation of mass and momentum. Based on the above theoretical formulas, a model is established in the Fluent software for solution, which can intuitively obtain the migration of SF6 tracer gas in the tunnel and the cross-sectional concentration distribution at the gas sensor position affected by the paint spraying gas. Step 2.6: Change the concentration of the initially released SF6 tracer gas and conduct multiple experiments; The step 3 calibrates the corresponding electrochemical gas sensor. Based on the above experiment on the influence of the gas generated by the painting operation on the commonly used electrochemical gas monitoring sensor in the mine, the formaldehyde concentration, temperature, humidity, and concentration readings of various gas sensors for mining are obtained. After preprocessing the data, the XGBoost concentration prediction model for different electrochemical gas sensors for mining is trained, and the data training time series prediction model is obtained according to the experiment of SF6 tracer gas. The specific processing process of the concentration prediction model is as follows: Step 3.1: For different gas sensors, the formaldehyde concentration value, temperature, and humidity used to predict the concentration of the electrochemical gas sensor are used as independent variables, and the monitoring value of a certain gas sensor is used as the dependent variable. Three-quarters of the sample data are randomly selected to train the XGBoost concentration prediction model, and the remaining quarter of the sample data is used as the prediction sample data to evaluate and verify the constructed XGBoost concentration prediction model; Step 3.2: Train the XGBoost concentration prediction model based on the training data, where the formaldehyde concentration value, temperature, and humidity of the training sample are used as independent variables, and the dust mass concentration value actually measured by each electrochemical gas sensor is used as the dependent variable. The relationship between the formaldehyde concentration value, temperature, humidity and the monitoring value of each gas sensor is solved, and each gas sensor corresponds to an XGBoost concentration prediction model; Step 3.3: Input the formaldehyde concentration value, temperature, and humidity of the predicted sample as input variables into the established XGBoost concentration prediction model to obtain the corresponding predicted value monitored by the gas sensor; Step 3.4: Combined with the actual measured gas sensor monitoring value, the predicted value of the gas sensor monitoring obtained in step 3.3 is evaluated, and the mean square error of the two is calculated. Assuming that the deviation between the two exceeds the acceptable range, the parameter value of the XGBoost concentration prediction model is adjusted and re-trained and predicted until it meets the acceptable range, thereby determining the XGBoost concentration prediction model and obtaining the relationship between formaldehyde concentration value, temperature, humidity and different gas sensor monitoring values.

2. A method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors according to claim 1, characterized in that: The positions of the temperature and humidity sensor (10), the formaldehyde sensor (9), and the mining series sensor (8) on the air permeable mesh (5) are not fixed.

3. A method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors according to claim 1, characterized in that: A humidifier is installed on the side of the experimental box to spray the inside of the box. A small fan supplies air to accelerate air convection and help the air mix evenly to simulate the humidity of the mine. A heating plate is used to slightly heat the bottom of the experimental box to simulate the temperature of the mine.

4. A method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors according to claim 1, characterized in that: The gas flow equation of SF6 tracer gas in the coal mine in step 2 is expressed as: Where: Q is the conserved variable, ρ is the mass density, ρu, ρv, ρw are the momentum densities in three directions respectively, ρE is the energy density, is the boundary of the control volume, F c is the inviscid flux, F v is the viscous flux; in:

5. A method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors according to claim 4, characterized in that: In step 2.5, when considering the ventilation of coal mine underground, the wind flow is generally in a turbulent state. The time-averaged method is used to regard the turbulent flow as the superposition of an average fluid and a pulsating fluid, and the turbulent flow motion is expressed as: The above are the transient continuity equation, the Reynolds time-averaged NS equation and the component transport equation; where u i ' and φ' are the speed u i The pulsation value of the scalar φ is solved by using the standard k-ε equation in FLUENT; In addition, the convective mass transfer of SF6 tracer gas also needs to consider the component transport. The mixed gas of SF6 tracer gas and air also needs the law of conservation of mass and momentum. For components without chemical reaction, the law of conservation of mass is expressed as: Where: c s is the concentration of SF6 tracer gas; U is the fluid vector velocity; D s is the diffusion coefficient of the gas; S s is the source term; In the experiment, the diffusion coefficient of SF6 tracer gas in air was calculated using: Where: r is the hydraulic diameter of the tunnel, m; f is the roadway friction coefficient; Based on the above theoretical formula, a model was established in the fluent software and solved, which can intuitively obtain the migration of SF6 tracer gas in the tunnel and the cross-sectional concentration distribution at the gas sensor position affected by the paint operation gas.

6. A method for eliminating the influence of gas generated by underground painting operation on the accuracy of electrochemical sensors according to claim 5, characterized in that: In step 3, the relationship between formaldehyde concentration, temperature, humidity and monitoring values ​​of different gas sensors solved by XGBoost algorithm is used. The time and concentration of the gas sensor position affected by the paint spraying gas measured by SF6 tracer gas experiment are used to establish a time series prediction model. The specific processing process of the time series prediction model is as follows: ① Discretization of the formaldehyde release concentration curve. Considering that formaldehyde release is a continuous process, assuming that the concentration of formaldehyde release does not change in a short period of time, the total concentration curve of formaldehyde release is divided into many small segments, and the formaldehyde concentration in each small segment remains unchanged; ②Data cleaning and preprocessing: According to the SF6 tracer gas experiment, the time and concentration of each small section of the concentration transmitted to the gas sensor position affected by the paint spraying gas are obtained, which are used to train the formaldehyde propagation model and clean the original data, including removing outliers, filling missing values, and normalizing the data; ③ Establish an LSTM time series prediction model, divide the data into a training set and a prediction set at a ratio of 4:1, use LSTM to fit the concentration changes of formaldehyde at each gas sensor location, and verify it on the prediction set to evaluate the model performance. The size of each parameter of the model can be adjusted to achieve the optimal performance of the model; ④ Based on the results of LSTM training, the predicted formaldehyde concentration value per second at each gas sensor location is obtained, and the aforementioned XGBoost model is called to predict the influence of formaldehyde on a certain gas at the gas sensor location.

Images