Ultra-thin coal seam working face water seepage self-sensing system and method

Abstract

The invention discloses an ultra-thin coal seam working face water seepage self-sensing system and method, and belongs to the technical field of ultra-thin coal seam fully-mechanized mining. Hydrogeological conditions are divided into four types of geology according to the water content of the mining area of the ultra-thin coal seam, and a sensor grading sensing method is determined according to the four types of geology and comprises basic sensing, advanced sensing, enhanced sensing and comprehensive sensing. And recording data of each sensor, and after data processing, establishing a temperature-time chart according to time domain and positioning analysis. A regression integral moving average model is used for analyzing abnormal changes in historical data, a random forest model is used for recognizing abnormal points in residual errors and sudden change values in data such as corresponding temperatures, data weights are adjusted in real time through weighted average optimization, an optimal model is obtained through training, and the water seepage time and the water seepage level can be determined. The water seepage position can be obtained by combining the water seepage time and the speed of the scraper conveyor; a water seepage self-sensing model is established according to the steps, and the water seepage grade, time and position are automatically sensed.

Description

Technical Field

[0001] The present invention relates to the technical field of ultra-thin coal seam fully mechanized mining, and more specifically to an ultra-thin coal seam working face water seepage self-sensing system and method. Background Art

[0002] In recent years, with the reduction of thick and medium-thick coal seams and the increasing scarcity of resources, the mining of extremely thin coal seams has attracted great attention from the country. Extremely thin coal seams are widely distributed in my country, especially in southern mines, with a large number, considerable absolute reserves and excellent coal quality. Moreover, the distribution of extremely thin coal seams is complex, the space is narrow, and the faults are many, which will lead to more serious water seepage problems. During the mining process, water seepage in the working face will reduce the stability of the roof and floor of the mine, increase the risk of mine collapse and water seepage accidents; water seepage will make the tunnels and mining working faces wet, leading to a deterioration of the working environment; water seepage in the mine will corrode metal equipment and supports, shortening their service life; and the thickness of extremely thin coal seams makes the rock strata of the roof and floor more likely to move and deform, resulting in new cracks, which may become channels for water seepage. Unlike thick coal seams, extremely thin coal seams are inconvenient for workers to operate, and more automated technology needs to be set up. Therefore, intelligent perception of water seepage in extremely thin coal seams is one of the key technologies to be broken through.

[0003] At present, there are some shortcomings in the monitoring of water seepage in the fully mechanized mining working face of ultra-thin coal seams: the current water seepage and water hazard monitoring technologies are mostly aimed at medium and thick coal seams. Due to the narrow working face of ultra-thin coal seams, most mining areas are manually inspected, which not only reduces the detection efficiency, but also increases the danger of employees' work; in addition, the thickness of ultra-thin coal seams limits the selection of sensor equipment, and if there is basically no water seepage in the hydrogeological conditions of a certain ultra-thin coal seam mining area, then the installation method relying on a fixed water seepage system becomes no longer applicable; due to the thickness of the extremely thin coal seams, it is not suitable to deploy large or large numbers of detection equipment, which limits the layout of sensors; the complex geological conditions of the extremely thin coal seams may interfere with the sensor data, and there is also a limit on the number of sensors, so comprehensive data processing and data utilization efficiency are particularly important; the space limitation of the extremely thin coal seams, such as in the mining areas with more water seepage, the current monitoring of water seepage requires workers to move on the working face to find the water seepage location, and the space limitation of the extremely thin coal seams makes it inconvenient for workers to move. Summary of the invention

[0004] Since there are few water seepage detection methods currently applicable to extremely thin coal seam mining working faces, and they lack pertinence, the purpose of the invention concept designed in this application is to provide a unified, accurate and efficient detection method and system for the detection of water seepage under extremely thin coal seams, which is specifically embodied in the following aspects:

[0005] 1. The thickness of the extremely thin coal seam limits the choice of sensor equipment, and if there is basically no water seepage in the hydrogeological conditions of a certain extremely thin coal seam mining area, then the installation method relying on a fixed water seepage system becomes no longer applicable. This application aims to change the type of water seepage sensing sensor according to the water content in the specific hydrogeological conditions of the extremely thin coal seam mining area, reduce mining costs, and improve the accuracy of water seepage sensing.

[0006] 2. Due to the thickness of the extremely thin coal seam, it is not suitable to deploy large or large number of detection equipment, which limits the layout of sensors. This application aims to change the layout of sensors on the working face, select a small number of sensors to be deployed around the scraper conveyor, and deploy them at fixed positions, use the scraper conveyor to move back and forth, drive the sensors to patrol back and forth, reduce installation points, and improve the space utilization rate of the extremely thin coal seam mining working face.

[0007] 3. Due to the complex geological conditions of extremely thin coal seams, sensor data may be interfered with, and there is also a limit on the number of sensors, so comprehensive data processing and data utilization efficiency are particularly important. Therefore, this application proposes a series of data processing processes and establishes a water seepage self-sensing model to make full use of sensor data, increase data validity and improve data utilization.

[0008] 4. Due to the space limitation of extremely thin coal seams, such as in mining areas with more water seepage, the current monitoring of water seepage requires workers to move on the working face to find the water seepage location, and the space limitation of extremely thin coal seams makes it inconvenient for workers to move. This application aims to propose a variety of methods for determining the water seepage location, in which the specific location of the water seepage is obtained by calculation based on the back-and-forth movement of the scraper conveyor, the specific location of the water seepage is obtained by direct monitoring based on optical fiber sensing, and the specific water seepage point is obtained by back-and-forth movement scanning based on infrared sensing.

[0009] The present invention achieves the above-mentioned purpose through the following technical scheme: a method for self-sensing water seepage in an extremely thin coal seam working face, the method comprising the following steps:

[0010] According to the water content of the extremely thin coal seam mining area, the hydrogeological conditions are divided into four types of geology. The sensor layout is determined according to the hydrological conditions of the extremely thin coal seam mining area, including four types of ABCD;

[0011] The sensor layout is determined according to the hydrological conditions of the extremely thin coal seam mining area, and a hierarchical perception method is adopted: Class A geology adopts basic perception and temperature sensor monitoring; Class B geology adopts advanced perception and adds humidity sensors for monitoring; Class C geology adopts enhanced perception and adds infrared thermal imaging sensors for monitoring; Class D geology adopts comprehensive and integrated perception and adds optical fiber sensor monitoring;

[0012] Determine the spatial layout of sensors, use the back-and-forth movement of the scraper conveyor to drive the sensors to inspect back and forth, and select a small number of sensors to be laid on the scraper conveyor to cope with the narrow conditions of extremely thin coal seams;

[0013] Data is transmitted to the terminal through 5G communication, and noise points are removed through preprocessing, data correction, digital filtering and other operations, and the matching of time to sensor values ​​is completed; and time domain analysis and positioning analysis are performed using terminal processing equipment;

[0014] The regression integral moving average model is used to analyze abnormal changes in historical data. The random forest model is used to identify abnormal points in the residuals, corresponding to the sudden changes in temperature and other data. The data weights are adjusted in real time through weighted average optimization. The best model is trained to determine the seepage time and level. The seepage location can be obtained by combining the time of seepage and the speed of the scraper conveyor.

[0015] The established random forest and regression integral moving average models are used to train a model for automatic water seepage perception. After this stage, weighted average optimization is performed to obtain the weight of each sensor. According to the above steps, a water seepage self-perception model is established to determine the specific water seepage level, water seepage time, and water seepage location, thereby completing the automatic perception of water seepage on the working face.

[0016] Preferably, the sensor type and layout are determined according to the hydrogeological conditions of the extremely thin coal seam mining area, including the following steps:

[0017] Specifically, the hydrogeological conditions in the mining area are divided into four types of geology:

[0018] Class A geology, geological exploration before mining, the hydrological conditions of this extremely thin coal seam mining area are underground water content <0.01km 3 , the working face under the mine is relatively dry and is hardly affected by water damage, with occasional water seepage;

[0019] Class B geology, the hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is 0.01~0.1km 3 The humidity of the working face under the mine is normal, water seepage is common, and no water inrush will occur;

[0020] Class C geology: The hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is 0.1 to 0.5 km 3 In between, the working face under the mine is wet, with frequent water seepage and occasional water inrush;

[0021] Class D geology, the hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is >0.5km 3 , water inrush frequently occurs in the underground working face of the mine.

[0022] Preferably, the method further includes determining the sensor type according to the hydrogeological conditions of the extremely thin coal seam mining area, and adopting a hierarchical sensing method, including the following steps:

[0023] Specifically, basic sensing is used for Class A geology. Since Class A geology does not often seep water, sensors are not used frequently and only a single temperature sensor is used for monitoring;

[0024] Class B geology adopts advanced perception, using temperature sensors and humidity sensors for monitoring, to make up for the shortcomings of a single sensor and improve the reliability and progress of water seepage perception;

[0025] Class C geology uses enhanced perception, adding an infrared thermal imaging sensor on the basis of advanced perception. After roughly locating the water seepage location using the temperature and humidity sensor, the water seepage location is accurately located based on the infrared thermal imaging results.

[0026] Class D geology adopts comprehensive and integrated perception, increases the deployment of fiber optic sensors, monitors the amount and location of water seepage on the working face horizontally, and monitors the specific time of water seepage vertically, to make up for the lack of accurate monitoring in local areas, and integrates other sensors to accurately determine the amount and location of water seepage.

[0027] Preferably, determining the sensor deployment conditions according to the hydrogeological conditions of the extremely thin coal seam mining area includes the following steps:

[0028] Determine the sensor layout according to the hydrological conditions of the ultra-thin coal seam mining area, characterized in that it also includes the spatial layout of the sensors;

[0029] The fully mechanized mining working face of extremely thin coal seams is narrow, so it is not suitable to deploy large or large numbers of detection equipment. Specifically, a small number of sensors are selected for basic perception, advanced perception, and enhanced perception, and are deployed on the scraper conveyor at fixed intervals. The scraper conveyor is used to move back and forth to drive the sensors to patrol back and forth, reduce installation points, and realize continuous dynamic monitoring of the working face over a long distance.

[0030] The added fiber optic sensors for comprehensive and integrated perception are not laid out by drilling and landfilling due to the particularity of extremely thin coal seams. Instead, the mounting plate is fixed near the bottom plate of the scraper conveyor, and the fluidity of the scraper conveyor is used to arrange the sensors in sections on the mounting plate, which are in contact with the bottom plate. The rock strata on the bottom plate of the coal seam swell when it comes into contact with water, causing bottom bulging and cooling. The amount of water seepage can be comprehensively obtained based on the degree of bottom bulging and cooling.

[0031] Preferably, it also includes data collection, data processing, and data analysis in the ultra-thin coal seam water seepage self-sensing system, including the following steps:

[0032] Collect various types of data through the data collection terminal; use 5G communication to transmit to the terminal to ensure real-time monitoring;

[0033] At the data processing end, the data from different sensors are classified through preprocessing;

[0034] Digital filtering removes noise and interference to improve signal quality; data correction solves data misalignment problems;

[0035] Invalid data is removed through data cleaning, and missing values ​​are compensated through data interpolation;

[0036] After completing these processes, time domain analysis is performed, using temperature data to establish a temperature-time graph, and using humidity data to establish a humidity-time graph, to analyze and determine the time point of water seepage;

[0037] Then, positioning analysis is performed. The water seepage location is calculated based on the time when water seepage occurs in a single temperature and humidity sensor and the movement speed of the scraper conveyor. The water output volume evaluated by the optical fiber sensor at a certain moment is used to establish a water output volume-position diagram horizontally, and the specific water seepage location is determined in combination with the infrared thermal imaging sensor.

[0038] After completing the data analysis, the threshold value is judged and the corresponding threshold value is set. After completing the data analysis, the threshold value is judged and the corresponding threshold value is set. Finally, the model is established to obtain the perception model;

[0039] At the data analysis end, the multi-modal sensor data is integrated to perform multi-modal data weighting, assign a weight to each sensor, and perform weighted calculations to obtain the specific location and amount of water seepage.

[0040] Preferably, the data processing in the ultra-thin coal seam water seepage self-sensing system includes the following steps:

[0041] Specifically, digital filtering is implemented through Gaussian function:

[0042]

[0043] Where σ is the standard deviation of the Gaussian function, and x is the sensor value;

[0044] The dynamic time warping algorithm (DTW) is used to correct the data, match the sensor value and sampling time, and solve the data misalignment problem. Define a distance function d(i,j) to calculate the distance between corresponding points in two sequences, usually using the Euclidean distance:

[0045] d(i,j)=||x i -y j ||

[0046] Among them, x i and j are the points in the two sequences respectively;

[0047] Create a matrix D of size N×M, where N and M are the lengths of the two sequences respectively; fill the matrix using the following recursive formula:

[0048] D(i,j)=d(i,j)+Imin{D(i-1,j),D(i,j-1),D(i-1,j-1)}

[0049] Where D(i,j) represents the minimum cumulative distance between the i-th point of sequence X and the j-th point of sequence Y; the last element D(N,M) in the matrix D is the DTW time difference between the two sequences;

[0050] Data cleaning removes outliers and replaces missing values ​​with the mode in the data.

[0051] Time domain analysis and positioning analysis establish (1) a temperature-time diagram of a sensor at a certain location, (2) a humidity-time diagram of a sensor at a certain location, (3) a water output-location diagram of a fiber optic sensor at a certain time, and (4) infrared thermal imaging sensor for location determination, etc., which facilitates the subsequent analysis of water seepage level, water seepage time, and water seepage location. Based on the specific four types of perception, the subsequent transmission channels (2), (3), and (4) are closed.

[0052] Preferably, the water seepage self-sensing method data self-sensing model is established, comprising the following steps:

[0053] After completing the time domain and positioning analysis, the established ARF-WAA data self-perception model is used to improve the generalization ability of the model through multiple iterative training.

[0054] Specifically, ARF-WAA, namely, the comprehensive regression integrated moving average model (ARIMA), the random forest model (RF), and the weighted average optimization algorithm (WAA);

[0055] First, the data after the four types of sensing processing are trained by the regression integral moving average model. The regression integral moving average model is a model established by introducing a differential operation to process non-stationary time series, which can transform the non-stationary time series in the temperature, humidity and optical fiber sensing process into a stationary time series, and then regressing the dependent variable only on its lag value and the present value and lag value of the random error term. It combines the three components of difference (I), autoregression (AR) and moving average (MA); the ARIMA (p, d, q) model can be expressed as:

[0056]

[0057] Where, p: the order of the autoregressive term, that is, the number of lagged observations used in the model. d: the difference order, that is, how many times the original series needs to be differentiated to obtain a stationary series. q: the order of the moving average term, that is, the number of lagged values ​​of the error term in the model. yt is the observed value of the time series at time t. μ is the mean of the series. φ i is the number of families of autoregressive terms. ∈ t is the error term at time t, usually assumed to be white noise with mean 0. i is the coefficient of the moving average term;

[0058] Determine the parameters (p, d, q) of the ARIMA model; the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots can be used to assist in selecting the values ​​of p and q, and the ADF test can be used to determine the d value, that is, the order of difference;

[0059] Use the ARIMA model with selected parameters to fit the temperature and humidity data to obtain the model parameter estimates, linear prediction results and residual sequence. Use the fitted ARIMA model to analyze abnormal changes in historical data;

[0060] The features of the time series extracted from the ARIMA model are transferred to the random forest, which helps identify anomalies in the residuals, which may correspond to mutation values ​​in data such as temperature. The residual sequence of the ARIMA model is analyzed using the random forest model to capture nonlinear features and mutation points in temperature, humidity, and fiber optic data; random forests improve the accuracy and robustness of predictions by building multiple decision trees and combining their results;

[0061] Random forest can reduce the risk of overfitting of a single model through ensemble learning;

[0062] Specifically, the mutation score of the decision tree model:

[0063]

[0064] Among them, s i,j represents the mutation score of region i, x k,j represents the value of feature j of sample k, represents the average value of feature j in region i;

[0065] Mutation score of the sample:

[0066]

[0067] Among them, r k represents the mutation score of sample k, s i,j represents the mutation score of region i, represents the average anomaly score of feature j;

[0068] Mutation point threshold:

[0069]

[0070] Where T represents the abnormal interval value, α and β are adjustment parameters, usually set to 0.5 and 0.1;

[0071] Mutation Detection:

[0072]

[0073] Among them, y k Indicates whether sample k is an outlier, 1 indicates a mutation point, and 0 indicates a normal point;

[0074] Finally, the weight of each sensor data is adjusted in real time using the weighted average optimization algorithm (WAA) to determine the specific water seepage level, water seepage time, and water seepage location, thus completing the automatic perception of water seepage on the working face.

[0075] The weighted average optimization algorithm uses random initialization:

[0076] x i =LB i +rand×(UB i -LB i )

[0077] Among them, rand is a random number between 0 and 1, LB i is the i-th lower bound value of the given problem, UB i is the i-th upper bound value of the given problem;

[0078] The first step in the weighted average position calculation is to calculate the fitness of each participant and rearrange the population according to one of the two characteristics, that is, larger is better or smaller is better; after that, the first candidate is selected in the population and the weighted average position is calculated using the following formula:

[0079]

[0080] Where N is the number of vectors, x i is the position of the i-th vector, w i is the weight of the i-th vector; the AR-WAA data self-perception model is established by combining the regression integral moving average model, the random forest model and the weighted average optimization algorithm. Through the above steps for training, the temperature and humidity data are fitted to obtain a more stable and reliable model, reducing the risk of overfitting of a single model, so as to automatically perceive the water seepage level, specific time of water seepage and water seepage location.

[0081] Preferably, the water seepage self-sensing method for determining the water seepage location comprises the following steps:

[0082] In enhanced perception of water seepage measurement, when water penetrates through comprehensive mining equipment or rock formations, it causes temperature changes. Infrared thermal imagers can accurately capture these temperature changes, thus helping us locate the leak point;

[0083] First, basic perception and advanced perception determine the temperature and humidity sensor monitoring position. Based on the above-established (1) temperature-time graph of the sensor at a certain position and (2) humidity-time graph of the sensor at a certain position, calculate the water seepage position. Assume that the speed of the scraper conveyor is represented by v. The time from the start of monitoring to the discovery of the mutation point is t. x ;

[0084]

[0085] v(t) is the instantaneous speed of the scraper conveyor at time t. T is the time from the start of monitoring to the discovery of the abnormal point t x ;

[0086] Infrared thermal imaging sensors are added to enhanced perception. The water seepage location determined by basic perception and advanced perception temperature and humidity sensing monitoring can be combined with the above-established (4) infrared thermal imaging sensor map to determine the spatial location of the water seepage point.

[0087] The comprehensive and integrated optical fiber sensing is based on the above-established (3) optical fiber sensor water output-position diagram at a certain time. The maximum water output point (abnormal point) is obtained through monitoring. The specific position of the abnormal point relative to the working surface is determined according to the position of the optical fiber where the abnormal point is located. The water seepage position of the working surface is calculated by combining basic perception and advanced perception temperature and humidity sensing monitoring. Finally, the spatial position of the water seepage point is determined using the infrared thermal imaging sensor image.

[0088] Finally, the weight of each sensor data is adjusted in real time by using a weighted average optimization algorithm. If the weighted average data shows abnormal changes, this may indicate the occurrence of water seepage. The specific water seepage level, water seepage time, and water seepage location are determined to complete automatic perception of water seepage on the working face.

[0089] Preferably, it also includes a water seepage self-sensing system for an extremely thin coal seam working face, comprising the following steps:

[0090] The data acquisition module selects sensors according to the perception level determined by hydrological conditions, including temperature sensors, humidity sensors, infrared thermal imaging sensors, and optical fiber sensors, which are used to collect the temperature, humidity, and water output of the extremely thin coal seam mining working face, as well as to determine the specific location of water seepage; and is divided into four types of perception according to the hydrological conditions: basic, advanced, enhanced, and comprehensive;

[0091] The data transmission module is mainly composed of a 5G communication integrated board, etc., and is used to transmit the initial collected data to the terminal processing platform;

[0092] The data processing module is used to process various types of transmitted data, remove noise points through preprocessing, data correction, digital filtering and other operations, and complete the matching of time to sensor values. The terminal processing equipment is used to perform time domain analysis and positioning analysis;

[0093] The data analysis module is used to analyze the results of the processed data and use the established random forest and regression integral moving average to train a model for automatic perception of water seepage. After this stage, weighted average optimization is performed to obtain the weight of each sensor, determine the specific water seepage level, water seepage time, and water seepage location, and complete the automatic perception of water seepage on the working face.

[0094] Compared with the prior art, the present invention has the following beneficial effects:

[0095] In view of the particularity of extremely thin coal seams, the present invention changes the type of water seepage sensing sensor according to the water content in the specific hydrogeological conditions of the extremely thin coal seam mining area, reduces the mining cost and the accuracy of water seepage sensing; according to the change of the layout of the sensor on the working surface, a small number of sensors are selected to be laid around the scraper conveyor, and are laid out at fixed positions, and the scraper conveyor is used to move back and forth to drive the sensor to inspect back and forth, reduce the installation points, and improve the space utilization rate of the extremely thin coal seam mining working surface; then, the present invention also proposes a series of data processing processes and establishes a water seepage self-sensing model to make full use of sensor data, increase data validity and improve data utilization. Finally, the present invention also proposes a variety of water seepage location determination methods, in which the specific location of the water seepage is obtained by calculation according to the back and forth movement of the scraper conveyor, the specific location of the water seepage is obtained by direct monitoring based on optical fiber sensing, and the specific water seepage point is obtained by scanning the back and forth movement of infrared sensing.

[0096] The present invention proposes a self-sensing method and system for water seepage applicable to extremely thin coal seams, in order to address the shortcoming that extremely thin coal seams are not suitable for deploying too many monitoring devices; it proposes to add a variety of monitoring means according to the hydrological conditions of the mine, in order to address the problem of the single existing monitoring means; it proposes a method that can automatically sense the specific location and level of water seepage, in order to reduce the inconvenience caused by the narrow space under the extremely thin coal seams. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a schematic diagram of the overall process structure of the present invention;

[0098] Figure 2 It is the classification diagram of hydrogeological conditions of the present invention;

[0099] Figure 3 This is an analysis diagram of the classification and selection of sensors and their location distribution in the present invention;

[0100] Figure 4 This is a framework diagram of the ultra-thin coal seam water seepage self-sensing system of the present invention;

[0101] Figure 5 This is a flow chart of the self-sensing method for water seepage in an ultra-thin coal seam according to the present invention;

[0102] Figure 6 This is a water seepage point positioning diagram for the ultra-thin coal seam working face of the present invention. DETAILED DESCRIPTION

[0103] The following will be combined with the attached embodiment of the present invention Figure 1-Figure 6 , clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0104] Example

[0105] like Figure 1 As shown, the present application proposes a method for self-sensing water seepage in an extremely thin coal seam working face, comprising the following steps:

[0106] S1: The hydrogeological conditions are divided into four types according to the water content of the extremely thin coal seam mining area. The sensor type selection and layout are determined according to the hydrological conditions of the extremely thin coal seam mining area, including four types of ABCD;

[0107] S2: Determine the sensor deployment according to the hydrological conditions of the extremely thin coal seam mining area, and adopt a graded perception method: Class A geology adopts basic perception and temperature sensor monitoring; Class B geology adopts advanced perception and adds humidity sensors for monitoring; Class C geology adopts enhanced perception and adds infrared thermal imaging sensors for monitoring; Class D geology adopts comprehensive and integrated perception and adds optical fiber sensor monitoring;

[0108] S3: Determine the spatial layout of sensors, use the back-and-forth movement of the scraper conveyor to drive the sensors to inspect back and forth, and select a small number of sensors to be deployed on the scraper conveyor to cope with the narrow conditions of extremely thin coal seams;

[0109] S4: Transmit data to the terminal through 5G communication, remove noise points through preprocessing, data correction, digital filtering and other operations, and complete the matching of time to sensor values; and use terminal processing equipment to perform time domain analysis and positioning analysis;

[0110] S5: Use the regression integral moving average model to analyze abnormal changes in historical data, identify abnormal points in the residuals through the random forest model, and the corresponding mutation values ​​in temperature and other data. Use weighted average optimization to adjust the data weights in real time and train the best model to determine the seepage time and level. Combined with the time of seepage and the speed of the scraper conveyor, the seepage location can be obtained.

[0111] S6: Use the established random forest and regression integral moving average models to train a model for automatic water seepage perception. After this stage, perform weighted average optimization to obtain the weight of each sensor. According to the above steps, establish a water seepage self-perception model, determine the specific water seepage level, water seepage time, and water seepage location, and complete the automatic perception of water seepage on the working face.

[0112] In step S1, the sensor type is determined according to the hydrogeological conditions of the extremely thin coal seam mining area, and the hydrogeological conditions of the mining area are divided into four types of geological conditions: Figure 2 As shown, including:

[0113] Class A geology, geological exploration before mining, the hydrological conditions of this extremely thin coal seam mining area are underground water content <0.01km 3 , the working face under the mine is relatively dry and is hardly affected by water damage, with occasional water seepage;

[0114] Class B geology, the hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is 0.01~0.1km 3 The humidity of the working face under the mine is normal, water seepage is common, and no water inrush will occur;

[0115] Class C geology: The hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is 0.1 to 0.5 km 3 In between, the working face under the mine is wet, with frequent water seepage and occasional water inrush;

[0116] Class D geology, the hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is >0.5km 3 , water inrush frequently occurs in the underground working face of the mine.

[0117] In step S2, the sensor type is determined according to the hydrogeological conditions of the extremely thin coal seam mining area, such as Figure 3 , using a hierarchical perception approach, including:

[0118] Class A geology adopts basic sensing. Since Class A geology does not often seep water, sensors are not used frequently and only a single temperature sensor is used for monitoring.

[0119] Class B geology adopts advanced perception, using temperature sensors and humidity sensors for monitoring, to make up for the shortcomings of a single sensor and improve the reliability and progress of water seepage perception;

[0120] Class C geology uses enhanced perception, adding an infrared thermal imaging sensor on the basis of advanced perception. After roughly locating the water seepage location using the temperature and humidity sensor, the water seepage location is accurately located based on the infrared thermal imaging results.

[0121] Class D geology adopts comprehensive and integrated perception, increases the deployment of fiber optic sensors, monitors the amount and location of water seepage on the working face horizontally, and monitors the specific time of water seepage vertically, to make up for the lack of accurate monitoring in local areas, and integrates other sensors to accurately determine the amount and location of water seepage.

[0122] In step S3, the hydrological conditions in the extremely thin coal seam mining area determine the sensor layout, such as Figure 3 , and also includes the spatial layout of sensors, including:

[0123] The fully mechanized mining working face of extremely thin coal seams is narrow, so it is not suitable to deploy large or large numbers of detection equipment. Specifically, a small number of sensors are selected for basic perception, advanced perception, and enhanced perception, and are deployed on the scraper conveyor at fixed intervals. The scraper conveyor is used to move back and forth to drive the sensors to patrol back and forth, reduce installation points, and realize continuous dynamic monitoring of the working face over a long distance.

[0124] The added fiber optic sensors for comprehensive and integrated perception are not laid out by drilling and landfilling due to the particularity of extremely thin coal seams. Instead, the mounting plate is fixed near the bottom plate of the scraper conveyor, and the fluidity of the scraper conveyor is used to arrange the sensors in sections on the mounting plate, which are in contact with the bottom plate. The rock strata on the bottom plate of the coal seam swell when it comes into contact with water, causing bottom bulging and cooling. The amount of water seepage can be comprehensively obtained based on the degree of bottom bulging and cooling.

[0125] In step S4, if Figure 4 , also includes data collection, data processing, and data analysis in the ultra-thin coal seam water seepage self-sensing system, specifically including:

[0126] Collect various types of data through the data collection terminal; use 5G communication to transmit to the terminal to ensure real-time monitoring;

[0127] At the data processing end, the data from different sensors are classified through preprocessing;

[0128] Digital filtering removes noise and interference to improve signal quality; data correction solves data misalignment problems;

[0129] Invalid data is removed through data cleaning, and missing values ​​are compensated through data interpolation;

[0130] After completing these processes, time domain analysis is performed, using temperature data to establish a temperature-time graph, and using humidity data to establish a humidity-time graph, to analyze and determine the time point of water seepage;

[0131] Then, positioning analysis is performed. The water seepage location is calculated based on the time when water seepage occurs in a single temperature and humidity sensor and the movement speed of the scraper conveyor. The water output volume evaluated by the optical fiber sensor at a certain moment is used to establish a water output volume-position diagram horizontally, and the specific water seepage location is determined in combination with the infrared thermal imaging sensor.

[0132] After completing the data analysis, the threshold value is judged and the corresponding threshold value is set. After completing the data analysis, the threshold value is judged and the corresponding threshold value is set. Finally, the model is established to obtain the perception model;

[0133] At the data analysis end, the multi-modal sensor data is integrated to perform multi-modal data weighting, assign a weight to each sensor, and perform weighted calculations to obtain the specific location and amount of water seepage.

[0134] In step S4, data processing in the ultra-thin coal seam water seepage self-sensing system, the process, specifically, digital filtering is implemented by Gaussian function:

[0135]

[0136] Where σ is the standard deviation of the Gaussian function, and x is the sensor value;

[0137] The dynamic time warping algorithm (DTW) is used to correct the data, match the sensor value and sampling time, and solve the data misalignment problem. Define a distance function d(i,j) to calculate the distance between corresponding points in two sequences, usually using the Euclidean distance:

[0138] d(i,j)=||x i -y j ||

[0139] Among them, x i and j are the points in the two sequences respectively;

[0140] Create a matrix D of size N×M, where N and M are the lengths of the two sequences respectively; fill the matrix using the following recursive formula:

[0141] D(i,j)=d(i,j)+Imin{D(i-1,j),D(i,j-1),D(i-1,j-1)}

[0142] Where D(i,j) represents the minimum cumulative distance between the i-th point of sequence X and the j-th point of sequence Y; the last element D(N,M) in the matrix D is the DTW time difference between the two sequences;

[0143] Data cleaning removes outliers and replaces missing values ​​with the mode in the data.

[0144] Time domain analysis and positioning analysis establish (1) a temperature-time diagram of a sensor at a certain location, (2) a humidity-time diagram of a sensor at a certain location, (3) a water output-location diagram of a fiber optic sensor at a certain time, and (4) infrared thermal imaging sensor for location determination, etc., which facilitates the subsequent analysis of water seepage level, water seepage time, and water seepage location. Based on the specific four types of perception, the subsequent transmission channels (2), (3), and (4) are closed.

[0145] In steps S5 and S6, a water seepage self-sensing method data self-sensing model is established, which specifically includes:

[0146] After completing the time domain and positioning analysis, the established ARF-WAA data self-perception model is used to improve the generalization ability of the model through multiple iterative training;

[0147] Specifically, ARF-WAA, namely, the comprehensive regression integrated moving average model (ARIMA), the random forest model (RF), and the weighted average optimization algorithm (WAA);

[0148] First, the data after the four types of sensing processing are trained by the regression integral moving average model. The regression integral moving average model is a model established by introducing a differential operation to process non-stationary time series, which can transform the non-stationary time series in the temperature, humidity and optical fiber sensing process into a stationary time series, and then regressing the dependent variable only on its lag value and the present value and lag value of the random error term. It combines the three components of difference (I), autoregression (AR) and moving average (MA); the ARIMA (p, d, q) model can be expressed as:

[0149]

[0150] Where, p: the order of the autoregressive term, that is, the number of lagged observations used in the model. d: the difference order, that is, how many times the original series needs to be differentiated to obtain a stationary series. q: the order of the moving average term, that is, the number of lagged values ​​of the error term in the model. y t is the observed value of the time series at time t. μ is the mean of the series. φ i is the number of families of autoregressive terms. ∈ t is the error term at time t, usually assumed to be white noise with mean 0. i is the coefficient of the moving average term;

[0151] Determine the parameters (p, d, q) of the ARIMA model; the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots can be used to assist in selecting the values ​​of p and q, and the ADF test can be used to determine the d value, that is, the order of difference;

[0152] Use the ARIMA model with selected parameters to fit the temperature and humidity data to obtain the model parameter estimates, linear prediction results and residual sequence. Use the fitted ARIMA model to analyze abnormal changes in historical data;

[0153] The features of the time series extracted from the ARIMA model are transferred to the random forest, which helps identify anomalies in the residuals, which may correspond to mutation values ​​in data such as temperature. The residual sequence of the ARIMA model is analyzed using the random forest model to capture nonlinear features and mutation points in temperature, humidity, and fiber optic data; random forests improve the accuracy and robustness of predictions by building multiple decision trees and combining their results;

[0154] Random forest can reduce the risk of overfitting of a single model through ensemble learning;

[0155] Specifically, the mutation score of the decision tree model:

[0156]

[0157] Among them, s i,j represents the mutation score of region i, x k,j represents the value of feature j of sample k, represents the average value of feature j in region i;

[0158] Mutation score of the sample:

[0159]

[0160] Among them, r k represents the mutation score of sample k, s i,j represents the mutation score of region i, represents the average anomaly score of feature j;

[0161] Mutation point threshold:

[0162]

[0163] Where T represents the abnormal interval value, α and β are adjustment parameters, usually set to 0.5 and 0.1;

[0164] Mutation Detection:

[0165]

[0166] Among them, y k Indicates whether sample k is an outlier, 1 indicates a mutation point, and 0 indicates a normal point;

[0167] Finally, the weight of each sensor data is adjusted in real time using the weighted average optimization algorithm (WAA) to determine the specific water seepage level, water seepage time, and water seepage location, thus completing the automatic perception of water seepage on the working face.

[0168] The weighted average optimization algorithm uses random initialization:

[0169] x i =LB i +rand×(UB i -LB i )

[0170] Among them, rand is a random number between 0 and 1, LB i is the i-th lower bound value of the given problem, UB i is the i-th upper bound value of the given problem;

[0171] The first step in the weighted average position calculation is to calculate the fitness of each participant and rearrange the population according to one of the two characteristics, that is, larger is better or smaller is better; after that, the first candidate is selected in the population and the weighted average position is calculated using the following formula:

[0172]

[0173] Where N is the number of vectors, x i is the position of the i-th vector, w i is the weight of the i-th vector;

[0174] Combining the regression integral moving average model, random forest model and weighted average optimization algorithm, an AR-WAA data self-perception model is established. Through the above steps for training, data such as temperature and humidity are fitted to obtain a more stable and reliable model, reducing the risk of overfitting of a single model, thereby automatically perceiving the level of water seepage, the specific time of water seepage and the location of water seepage.

[0175] In step S6, the water seepage location is determined by the water seepage self-sensing method, and the process specifically includes:

[0176] In enhanced perception of water seepage measurement, when water penetrates through comprehensive mining equipment or rock formations, it causes temperature changes. Infrared thermal imagers can accurately capture these temperature changes, thus helping us locate the leak point;

[0177] First, basic perception and advanced perception determine the temperature and humidity sensor monitoring location. Figure 5 The water seepage location can be calculated from (1) the temperature-time diagram of a sensor at a certain location and (2) the humidity-time diagram of a sensor at a certain location. Assume that the speed of the scraper conveyor is represented by v. The time from the start of monitoring to the discovery of the mutation point is t. x ;

[0178]

[0179] v(t) is the instantaneous speed of the scraper conveyor at time t. T is the time from the start of monitoring to the discovery of the abnormal point t x ;

[0180] Infrared thermal imaging sensors are added to enhanced perception, and the water seepage location can be determined based on basic perception and advanced perception temperature and humidity sensor monitoring. Figure 5 (4) infrared thermal imaging sensor image to determine the spatial location of the water seepage point;

[0181] Fiber optic sensing for comprehensive perception is based on Figure 5 (3) The water output-position diagram of the optical fiber sensor at a certain time. The maximum water output point (abnormal point) is obtained through monitoring. The specific position of the abnormal point relative to the working surface is determined according to the position of the optical fiber where the abnormal point is located. The water seepage position of the working surface is calculated by combining basic perception and advanced perception temperature and humidity sensor monitoring. Finally, the spatial position of the water seepage point is determined using the infrared thermal imaging sensor image.

[0182] Finally, the weight of each sensor data is adjusted in real time by using a weighted average optimization algorithm. If the weighted average data shows abnormal changes, this may indicate the occurrence of water seepage. The specific water seepage level, water seepage time, and water seepage location are determined to complete automatic perception of water seepage on the working face.

[0183] like Figure 3 As shown, the present application also proposes an extremely thin coal seam hydraulic support positioning system, comprising:

[0184] The data acquisition module selects sensors according to the perception level determined by hydrological conditions, including temperature sensors, humidity sensors, infrared thermal imaging sensors, and optical fiber sensors, which are used to collect the temperature, humidity, and water output of the extremely thin coal seam mining working face, as well as to determine the specific location of water seepage; and is divided into four types of perception according to the hydrological conditions: basic, advanced, enhanced, and comprehensive;

[0185] The data transmission module is mainly composed of a 5G communication integrated board, etc., and is used to transmit the initial collected data to the terminal processing platform;

[0186] The data processing module is used to process various types of transmitted data, remove noise points through preprocessing, data correction, digital filtering and other operations, and complete the matching of time to sensor values. The terminal processing equipment is used to perform time domain analysis and positioning analysis;

[0187] The data analysis module is used to analyze the results of the processed data and use the established random forest and regression integral moving average to train a model for automatic perception of water seepage. After this stage, weighted average optimization is performed to obtain the weight of each sensor, determine the specific water seepage level, water seepage time, and water seepage location, and complete the automatic perception of water seepage on the working face.

[0188] Specifically, in the data acquisition module, the temperature sensor is used to monitor the temperature of the working surface. A small number of sensors are selected and arranged on the scraper conveyor, and they are arranged at fixed positions. The scraper conveyor is used to move back and forth to drive the sensors to inspect back and forth, reducing the number of installation points.

[0189] Humidity sensors are used to monitor the absolute humidity of the working surface. A small number of sensors are selected and arranged on the scraper conveyor at fixed intervals. The scraper conveyor is used to move back and forth to drive the sensors to inspect back and forth, reducing the number of installation points.

[0190] Infrared thermal imaging sensors are used to determine the location of water seepage. When water penetrates through fully-mechanized mining equipment or rock formations, it causes temperature changes. Infrared thermal imaging cameras can accurately capture these temperature changes, thus helping us locate water leaks. A small number of sensors are selected and laid out on the scraper conveyor at fixed intervals. The scraper conveyor is used to move back and forth to drive the sensors to inspect back and forth, thus reducing the number of installation points.

[0191] The optical fiber sensor is used to monitor the water output of the working face. The installation plate is fixed near the bottom plate of the scraper conveyor. The scraper conveyor is used to arrange the installation plate in sections to contact the bottom plate. The rock layer of the coal seam bottom plate expands when it encounters water, causing bottom bulging and temperature drop. The water seepage amount can be obtained based on the degree of bottom bulging and temperature drop.

[0192] The data transmission module is mainly a 5G communication integrated board, which is responsible for transmitting the collected temperature and humidity, temperature, optical fiber sensor and other data from the data collection end to the data processing end. With its high speed, low latency and high bandwidth, the 5G network can achieve real-time or near real-time data transmission;

[0193] The data processing module includes preprocessing, digital filtering, data correction, cleaning and interpolation, positioning analysis, time domain analysis, threshold judgment and other units;

[0194] The digital filtering unit uses Gaussian function filtering. Specifically, digital filtering is implemented through Gaussian function:

[0195]

[0196] Where σ is the standard deviation of the Gaussian function, and x is the sensor value;

[0197] The data correction unit is mainly a dynamic time warping algorithm, which uses DTW to correct data, match sensor values ​​and sampling times, and solve the problem of data misalignment. Define a distance function d(i,j) to calculate the distance between corresponding points in two sequences, usually using Euclidean distance:

[0198] d(i,j)=||x i -y j ||

[0199] Among them, x i and j are the points in the two sequences respectively;

[0200] Create a matrix D of size N×M, where N and M are the lengths of the two sequences respectively. Use the following recursive formula to fill the matrix:

[0201] D(i,j)=d(i,j)+Imin{D(i-1,j),D(i,j-1),D(i-1,j-1)}

[0202] Where D(i,j) represents the minimum cumulative distance between the i-th point of sequence X and the j-th point of sequence Y;

[0203] The last element D(N,M) in the matrix D is the DTW time difference between the two sequences;

[0204] The main function of the data cleaning and interpolation unit is to remove outliers and replace missing values ​​with the mode in the data.

[0205] The model building unit is the ARF-WAA data self-sensing model, which integrates the regression integrated moving average model (ARIMA), random forest model (RF) and weighted average optimization algorithm (WAA);

[0206] The regression integral moving average model is a model that processes non-stationary time series by introducing differential operations. It can transform non-stationary time series in temperature, humidity and optical fiber sensing processes into stationary time series, and then regress the dependent variable only on its lagged value and the present value and lagged value of the random error term. It combines three components: differential (I), autoregression (AR) and moving average (MA); uses the ARIMA model with selected parameters to fit data such as temperature and humidity to obtain the model's parameter estimates, linear prediction results and residual series; uses the fitted ARIMA model to analyze abnormal changes in historical data;

[0207] Random forest can reduce the risk of overfitting of a single model through ensemble learning. It is used to identify outliers in residuals, which may correspond to mutation values ​​in data such as temperature. The basic idea is to improve the performance and accuracy of the overall model by building multiple decision trees. In a random forest, each tree will split the data to find the best split point. This process will be repeated until a complete decision tree is built. Each tree in a random forest will vote for outliers, and finally determine which points are outliers based on the voting results of the majority of trees.

[0208] The weighted average optimization algorithm (WAA) is a new meta-heuristic algorithm, the core idea of ​​which is to apply weighted average position to solve optimization problems. This weighted average position refers to a statistical measure that considers the weight (or importance) of each element in the data set when calculating the overall average position; in the WAA algorithm, the weighted average position represents the current population distribution, and is combined with the individual best position and the global best position to guide the next iteration of the position update to adjust the weight of each sensor data in real time;

[0209] The data analysis module includes units such as multi-mode data weighting, weighted calculation, water seepage location, and water seepage judgment.

[0210] In view of the particularity of extremely thin coal seams, the present invention changes the type of water seepage sensing sensor according to the water content in the specific hydrogeological conditions of the extremely thin coal seam mining area, reduces the mining cost and the accuracy of water seepage sensing; according to the change of the layout of the sensor on the working face, a small number of sensors are selected to be laid around the scraper conveyor, and are laid out at fixed positions, and the scraper conveyor is used to move back and forth to drive the sensor to inspect back and forth, reduce the installation points, and improve the space utilization rate of the extremely thin coal seam mining working face; then, the present invention also proposes a series of data processing processes and establishes a water seepage self-sensing model to make full use of sensor data, increase data validity and improve data utilization. Finally, the method of first performing geological modeling on geological exploration data and then fusing point cloud data and image data is adopted, which not only improves the speed of modeling, but also ensures the consistent fusion between data, and obtains a more comprehensive and accurate working face model.

[0211] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0212] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for self-sensing water seepage in an extremely thin coal seam working face, characterized in that: The following steps are involved: According to the water content of the extremely thin coal seam mining area, the hydrogeological conditions are divided into four types of geology. The sensor layout is determined according to the hydrological conditions of the extremely thin coal seam mining area, including four types of ABCD; The sensor layout is determined according to the hydrological conditions of the extremely thin coal seam mining area, and a hierarchical perception method is adopted: Class A geology adopts basic perception and temperature sensor monitoring; Class B geology adopts advanced perception and adds humidity sensors for monitoring; Class C geology adopts enhanced perception and adds infrared thermal imaging sensors for monitoring; Class D geology adopts comprehensive and integrated perception and adds optical fiber sensor monitoring; Determine the spatial layout of sensors, use the back-and-forth movement of the scraper conveyor to drive the sensors to inspect back and forth, and select a small number of sensors to be laid on the scraper conveyor to cope with the narrow conditions of extremely thin coal seams; Data is transmitted to the terminal through 5G communication, and noise points are removed through preprocessing, data correction, digital filtering and other operations, and the matching of time to sensor values ​​is completed; and time domain analysis and positioning analysis are performed using terminal processing equipment; The regression integral moving average model is used to analyze abnormal changes in historical data. The random forest model is used to identify abnormal points in the residuals, corresponding to the sudden changes in temperature and other data. The data weights are adjusted in real time through weighted average optimization. The best model is trained to determine the seepage time and level. The seepage location can be obtained by combining the time of seepage and the speed of the scraper conveyor. The established random forest and regression integral moving average models are used to train a model for automatic water seepage perception. After this stage, weighted average optimization is performed to obtain the weight of each sensor. According to the above steps, a water seepage self-perception model is established to determine the specific water seepage level, water seepage time, and water seepage location, thereby completing the automatic perception of water seepage on the working face.

2. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 1 is characterized in that: The following steps are involved: The sensor type and layout are determined according to the hydrogeological conditions of the extremely thin coal seam mining area. Specifically, the hydrogeological conditions of the mining area are divided into four types of geology: Class A geology, geological exploration before mining, the hydrological conditions of this extremely thin coal seam mining area are underground water content <0.01km 3 , the working face under the mine is relatively dry and is hardly affected by water damage, with occasional water seepage; Class B geology, the hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is 0.01~0.1km 3 The humidity of the working face under the mine is normal, water seepage is common, and no water inrush will occur; Class C geology: The hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is 0.1 to 0.5 km 3 In between, the working face under the mine is wet, with frequent water seepage and occasional water inrush; Class D geology, the hydrological conditions in the extremely thin coal seam mining area are that the underground water content in the mining area is >0.5km 3 , water inrush frequently occurs in the underground working face of the mine.

3. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 2 is characterized in that: It also includes determining the sensor type according to the hydrogeological conditions of the extremely thin coal seam mining area, and adopting a hierarchical perception method, including: Specifically, basic sensing is used for Class A geology. Since Class A geology does not often seep water, sensors are not used frequently and only a single temperature sensor is used for monitoring; Class B geology adopts advanced perception, using temperature sensors and humidity sensors for monitoring, to make up for the shortcomings of a single sensor and improve the reliability and progress of water seepage perception; Class C geology uses enhanced perception, adding an infrared thermal imaging sensor on the basis of advanced perception. After roughly locating the water seepage location using the temperature and humidity sensor, the water seepage location is accurately located based on the infrared thermal imaging results. Class D geology adopts comprehensive and integrated perception, increases the deployment of fiber optic sensors, monitors the amount and location of water seepage on the working face horizontally, and monitors the specific time of water seepage vertically, to make up for the lack of accurate monitoring in local areas, and integrates other sensors to accurately determine the amount and location of water seepage.

4. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 3 is characterized in that: The sensor layout is determined according to the hydrogeological conditions of the extremely thin coal seam mining area, including: Determine the sensor layout according to the hydrological conditions of the ultra-thin coal seam mining area, characterized in that it also includes the spatial layout of the sensors; The fully mechanized mining working face of extremely thin coal seams is narrow, so it is not suitable to deploy large or large numbers of detection equipment. Specifically, a small number of sensors are selected for basic perception, advanced perception, and enhanced perception, and are deployed on the scraper conveyor at fixed intervals. The scraper conveyor is used to move back and forth to drive the sensors to patrol back and forth, reduce installation points, and realize continuous dynamic monitoring of the working face over a long distance. The added fiber optic sensors for comprehensive and integrated perception are not laid out by drilling and landfilling due to the particularity of extremely thin coal seams. Instead, the mounting plate is fixed near the bottom plate of the scraper conveyor, and the fluidity of the scraper conveyor is used to arrange the sensors in sections on the mounting plate, which are in contact with the bottom plate. The rock strata on the bottom plate of the coal seam swell when it comes into contact with water, causing bottom bulging and cooling. The amount of water seepage can be comprehensively obtained based on the degree of bottom bulging and cooling.

5. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 4 is characterized in that: It also includes data collection, data processing, and data analysis in the ultra-thin coal seam water seepage self-sensing system. The specific steps include: Collect various types of data through the data collection terminal; use 5G communication to transmit to the terminal to ensure real-time monitoring; At the data processing end, the data from different sensors are classified through preprocessing; Digital filtering removes noise and interference to improve signal quality; data correction solves data misalignment problems; Invalid data is removed through data cleaning, and missing values ​​are compensated through data interpolation; After completing these processes, time domain analysis is performed, using temperature data to establish a temperature-time graph, and using humidity data to establish a humidity-time graph, to analyze and determine the time point of water seepage; Then, positioning analysis is performed. The water seepage location is calculated based on the time when water seepage occurs in a single temperature and humidity sensor and the movement speed of the scraper conveyor. The water output volume evaluated by the optical fiber sensor at a certain moment is used to establish a water output volume-position diagram horizontally, and the specific water seepage location is determined in combination with the infrared thermal imaging sensor. After completing the data analysis, the threshold value is judged and the corresponding threshold value is set. After completing the data analysis, the threshold value is judged and the corresponding threshold value is set. Finally, the model is established to obtain the perception model; At the data analysis end, the multi-modal sensor data is integrated to perform multi-modal data weighting, assign a weight to each sensor, and perform weighted calculations to obtain the specific location and amount of water seepage.

6. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 5 is characterized in that: The data processing in the ultra-thin coal seam water seepage self-sensing system specifically includes the following steps: Specifically, digital filtering is implemented through Gaussian function: Where σ is the standard deviation of the Gaussian function, and x is the sensor value; The dynamic time warping algorithm (DTW) is used to correct the data, match the sensor value and sampling time, and solve the data misalignment problem. Define a distance function d(i, j) to calculate the distance between corresponding points in two sequences, usually using the Euclidean distance: d(i,j)=||x i -y j || Among them, x i and i are the points in the two sequences respectively; Create a matrix D of size N×M, where N and M are the lengths of the two sequences respectively; fill the matrix using the following recursive formula: D(i,j)=d(i,j)+Imin{D(i-1,j),D(i,j-1),D(i-1,j-1)} Where D(i, j) represents the minimum cumulative distance between the i-th point of sequence X and the j-th point of sequence Y; the last element D(N, M) in the matrix D is the DTW time difference between the two sequences; Data cleaning removes outliers and replaces missing values ​​with the mode in the data. Time domain analysis and positioning analysis establish (1) a temperature-time diagram of a sensor at a certain location, (2) a humidity-time diagram of a sensor at a certain location, (3) a water output-location diagram of a fiber optic sensor at a certain time, and (4) infrared thermal imaging sensor for location determination, etc., which facilitates the subsequent analysis of water seepage level, water seepage time, and water seepage location. Based on the specific four types of perception, the subsequent transmission channels (2), (3), and (4) are closed.

7. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 6 is characterized in that: The data self-sensing model of water seepage self-sensing method is established, which specifically includes the following steps: After completing the time domain and positioning analysis, the established ARF-WAA data self-perception model is used to improve the generalization ability of the model through multiple iterative training. Specifically, ARF-WAA, namely, the comprehensive regression integrated moving average model (ARIMA), the random forest model (RF), and the weighted average optimization algorithm (WAA); First, the data after the four types of sensing processing are trained by the regression integral moving average model. The regression integral moving average model is a model established by introducing a differential operation to process non-stationary time series, which can transform the non-stationary time series in the temperature, humidity and optical fiber sensing process into a stationary time series, and then regressing the dependent variable only on its lag value and the present value and lag value of the random error term. It combines the three components of difference (I), autoregression (AR) and moving average (MA); the ARIMA (p, d, q) model can be expressed as: Where, p: the order of the autoregressive term, that is, the number of lagged observations used in the model. d: the difference order, that is, how many times the original series needs to be differentiated to obtain a stationary series. q: the order of the moving average term, that is, the number of lagged values ​​of the error term in the model. y t is the observed value of the time series at time t. μ is the mean of the series. φ i is the number of families of autoregressive terms. ∈ t is the error term at time t, usually assumed to be white noise with mean 0. i is the coefficient of the moving average term; Determine the parameters (p, d, q) of the ARIMA model; the autocorrelation function (ACF) and partial autocorrelation function (PACF) plots can be used to assist in selecting the values ​​of p and q, and the ADF test can be used to determine the d value, that is, the order of difference; Use the ARIMA model with selected parameters to fit the temperature and humidity data to obtain the model parameter estimates, linear prediction results and residual sequence. Use the fitted ARIMA model to analyze abnormal changes in historical data; The features of the time series extracted from the ARIMA model are transferred to the random forest, which helps identify anomalies in the residuals, which may correspond to mutation values ​​in data such as temperature. The residual sequence of the ARIMA model is analyzed using the random forest model to capture nonlinear features and mutation points in temperature, humidity, and fiber optic data; random forests improve the accuracy and robustness of predictions by building multiple decision trees and combining their results; Random forest can reduce the risk of overfitting of a single model through ensemble learning; Specifically, the mutation score of the decision tree model: Among them, s i，j represents the mutation score of region i, x k，j represents the value of feature j of sample k, represents the average value of feature j in region i; Mutation score of the sample: Among them, r k represents the mutation score of sample k, s i，j represents the mutation score of region i, represents the average anomaly score of feature j; Mutation point threshold: Where T represents the abnormal interval value, α and β are adjustment parameters, usually set to 0.5 and 0.1; Mutation Detection: Among them, y k Indicates whether sample k is an outlier, 1 indicates a mutation point, and 0 indicates a normal point; Finally, the weight of each sensor data is adjusted in real time using the weighted average optimization algorithm (WAA) to determine the specific water seepage level, water seepage time, and water seepage location, thus completing the automatic perception of water seepage on the working face. The weighted average optimization algorithm uses random initialization: x i =LB i +rand×(UB i -LB i ) Among them, rand is a random number between 0 and 1, LB i is the i-th lower bound value of the given problem, UB i is the i-th upper bound value of the given problem; The first step in the weighted average position calculation is to calculate the fitness of each participant and rearrange the population according to one of the two characteristics, that is, larger is better or smaller is better; after that, the first candidate is selected in the population and the weighted average position is calculated using the following formula: Where N is the number of vectors, x i is the position of the i-th vector, w i is the weight of the i-th vector; Combining the regression integral moving average model, random forest model and weighted average optimization algorithm, an AR-WAA data self-perception model is established. Through the above steps for training, data such as temperature and humidity are fitted to obtain a more stable and reliable model, reducing the risk of overfitting of a single model, thereby automatically perceiving the level of water seepage, the specific time of water seepage and the location of water seepage.

8. The method for self-sensing water seepage in an ultra-thin coal seam working face according to claim 7 is characterized in that: The process of determining the water seepage location by the water seepage self-sensing method specifically includes: In enhanced perception of water seepage measurement, when water penetrates through comprehensive mining equipment or rock formations, it causes temperature changes. Infrared thermal imagers can accurately capture these temperature changes, thus helping us locate the leak point; First, basic perception and advanced perception determine the temperature and humidity sensor monitoring position. According to (1) the temperature-time diagram of a sensor at a certain position and (2) the humidity-time diagram of a sensor at a certain position in claim 6, it is assumed that the speed of the scraper conveyor is represented by v; the time from the start of monitoring to the discovery of the mutation point is t x ; v(t) is the instantaneous speed of the scraper conveyor at time t. T is the time from the start of monitoring to the discovery of the abnormal point t x ; Infrared thermal imaging sensors are added to enhanced perception, and the spatial location of the water seepage point can be determined based on the water seepage location determined by basic perception and advanced perception temperature and humidity sensing monitoring, combined with the infrared thermal imaging sensor image (4) in claim 6; The optical fiber sensing with comprehensive and integrated perception is based on the optical fiber sensor water output-position diagram at a certain time in claim 6 (3), and the maximum water output point (abnormal point) is obtained through monitoring. The specific position of the abnormal point relative to the working surface is determined according to the position of the optical fiber where the abnormal point is located, and the water seepage position of the working surface is calculated by combining basic perception and advanced perception temperature and humidity sensing monitoring. Finally, the spatial position of the water seepage point is determined using the infrared thermal imaging sensor image; Finally, the weight of each sensor data is adjusted in real time by using a weighted average optimization algorithm. If the weighted average data shows abnormal changes, this may indicate the occurrence of water seepage. The specific water seepage level, water seepage time, and water seepage location are determined to complete automatic perception of water seepage on the working face.

9. A water seepage self-sensing system for an ultra-thin coal seam working face based on claims 1 to 8, characterized in that: The data acquisition module selects sensors according to the perception level determined by hydrological conditions, including temperature sensors, humidity sensors, infrared thermal imaging sensors, and optical fiber sensors, which are used to collect the temperature, humidity, and water output of the extremely thin coal seam mining working face, as well as to determine the specific location of water seepage; and is divided into four types of perception according to the hydrological conditions: basic, advanced, enhanced, and comprehensive; The data transmission module is mainly composed of a 5G communication integrated board, etc., and is used to transmit the initial collected data to the terminal processing platform; The data processing module is used to process various types of transmitted data, remove noise points through preprocessing, data correction, digital filtering and other operations, and complete the matching of time to sensor values. The terminal processing equipment is used to perform time domain analysis and positioning analysis; The data analysis module is used to analyze the results of the processed data and use the established random forest and regression integral moving average to train a model for automatic perception of water seepage. After this stage, weighted average optimization is performed to obtain the weight of each sensor, determine the specific water seepage level, water seepage time, and water seepage location, and complete the automatic perception of water seepage on the working face.

Images