Ultrathin coal seam collapse early warning system based on multi-field dynamic analysis

Abstract

The invention provides an ultra-thin coal seam collapse early warning system based on multi-field dynamic analysis and an optimization method thereof, and aims to carry out fusion and deep analysis on data by monitoring a plurality of physical data such as a temperature field, a confining pressure field and a gas field in a coal seam in real time and combining a parallel multi-grid analysis algorithm and an LSTM time sequence analysis technology so as to realize early warning of the collapse of an ultra-thin coal seam. Therefore, early warning of the ultra-thin coal seam collapse disaster is realized, and the safety of coal mine production is ensured. The system is composed of a monitoring sensor network, a data acquisition module, a data fusion and analysis module, an early warning decision module and a display and alarm module, can dynamically predict the dangerous state of a coal seam according to multi-field coupling data, and sends out an early warning signal in time when an abnormality is found. By accurately arranging the sensors and adopting an efficient multi-grid algorithm and a powerful LSTM analysis model, the system can realize intelligent monitoring and accurate early warning of the safety state of a coal mining area in a complex environment, and the safety and intelligent level of coal mine production are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal mining, and in particular to a very thin coal seam collapse early warning system and optimization method based on multi-field dynamic analysis, which is applicable to the real-time monitoring and early warning of very thin coal seam collapse during the coal mining process. Background Technique

[0002] With the increase of coal mining depth and the gradual thinning of coal seam thickness, the mining of very thin coal seams faces more complex and severe safety challenges. The mining of very thin coal seams is easily affected by various factors such as geological conditions, mining techniques, and support conditions, resulting in coal seam collapse or other sudden safety accidents. At present, the collapse early warning technology in coal mining mainly relies on single monitoring means, and the early warning system usually can only make judgments based on static data, lacking comprehensive analysis and response to dynamic change factors during the mining process, resulting in poor early warning timeliness and low accuracy, and being unable to effectively prevent safety hazards in the mining of very thin coal seams.

[0003] Therefore, there is an urgent need for an early warning system based on multi-field dynamic analysis, which can comprehensively consider various physical fields (such as temperature field, confining pressure field, gas field, etc.) and their interactions during the coal seam mining process, so as to realize the early warning and prevention and control of very thin coal seam collapse. Summary of the Invention

[0004] The purpose of the present invention is to provide a very thin coal seam collapse early warning system and optimization method based on multi-field dynamic analysis, which can monitor and analyze the dynamic changes of the coal seam in real time, and conduct comprehensive analysis by combining multi-field physical field data, so as to realize the early warning of coal seam collapse and ensure the safety of coal mine production.

[0005] The present invention realizes the above purpose through the following technical solutions: A very thin coal seam collapse early warning system based on multi-field dynamic analysis, and this system includes the following components

[0006] Monitoring sensor network: responsible for real-time collection of multi-field data such as temperature field, confining pressure field, gas field, etc. in the very thin coal mine mining area. According to different influence areas of the very thin coal seam, due to different degrees of physical field coupling influence in different influence areas, there are also differences in the arrangement of its sensors, and the arrangement situations at different positions are divided into five categories. The sensors used include temperature sensors, confining pressure sensors, gas sensors, etc.

[0007] Data acquisition module: used to receive and process the original data collected by the monitoring sensors, and convert it into digital signals available for analysis.

[0008] Data Fusion and Analysis Module: It is used to fuse, process, and analyze multi-field data. The parallel multi-grid analysis algorithm is adopted to fuse and analyze the collected multi-field data. Compared with conventional convergence algorithms such as Jacobi and Gauss-Seidel iteration methods, the parallel multi-grid algorithm can achieve rapid convergence of large errors with high frequencies in extremely thin coal seams, and can also quickly converge small errors caused by multi-field coupling effects such as temperature-confining pressure-gas in extremely thin coal seams. By using the parallel multi-grid algorithm to perform reciprocating iterations on the collected multi-field data, accurate time series data processed by the data fusion and analysis module is obtained, ensuring the accuracy of disaster warning under multi-field coupling effects such as temperature-confining pressure-gas in extremely thin coal seams.

[0009] Early Warning Decision Module: Based on the time series data obtained from the data fusion and analysis module, the LSTM and time series analysis algorithms are used to warn of disasters under multi-field coupling effects such as temperature-confining pressure-gas in extremely thin coal seams, and predict changes in multiple physical fields such as future temperature, confining pressure, and gas to achieve the warning function. Compared with traditional time series analysis methods such as ARIMA and GARCH, the LSTM and time series analysis algorithms can effectively analyze the complex, non-linear, and time-varying physical quantity changes in extremely thin coal seams and give predictions of future physical quantity changes.

[0010] Display and Alarm Module: According to the future prediction of the early warning decision module, it real-time displays the safety status of the coal seam mining area and gives an audible and visual alarm to prompt relevant personnel to take emergency measures in case of anomalies.

[0011] The real-time data of multi-field data (such as temperature, confining pressure, gas field, etc.) in the extremely thin coal mining area is sensed and monitored, and the monitored data is transmitted to the data acquisition module. Using the data acquisition module, the collected multi-field analog signals are converted into digital signals for analysis and transmitted to the data fusion and analysis module; in the data fusion and analysis module, the parallel multi-grid analysis algorithm is used to fuse and analyze the collected multi-field data. By using the parallel multi-grid algorithm to perform reciprocating iterations on the collected multi-field data, accurate time series data processed by the data fusion and analysis module is obtained. In the early warning decision module, the LSTM and time series analysis algorithms are used to warn of disasters under multi-field coupling effects such as temperature-confining pressure-gas in extremely thin coal seams, predict changes in multiple physical fields such as future temperature, confining pressure, and gas, and in the display and alarm module, the safety status of the extremely thin coal seam mining area is real-time displayed and an audible and visual alarm is given to prompt relevant personnel to take emergency measures in case of anomalies.

[0012] Through the method of this invention, the safety during coal mining can be effectively improved, potential collapse risks can be identified in advance, the accident incidence can be reduced, and more intelligent and precise safety guarantee is provided for coal production.

[0013] A very-thin coal seam collapse early warning system based on multi-field dynamic analysis. The system realizes disaster early warning for a very-thin coal seam under the coupling action of multiple fields such as temperature-confining pressure-gas, including the following steps:

[0014] Preferably, a very-thin coal seam collapse early warning system based on multi-field dynamic analysis - a monitoring sensor network;

[0015] In the field of very-thin coal seams, due to the small coal seam thickness, complex structure, and strong inhomogeneity, there are differences in the coupling degree of physical quantities involved in different areas of very-thin coal seams. Therefore, the arrangement of sensors needs to be more refined and targeted to accurately monitor the coupling relationship of physical quantities such as temperature, confining pressure, and gas. According to the placement of temperature-confining pressure-gas sensors in different areas of very-thin coal seams, it is divided into categories I, II, III, IV, and V.

[0016] For category I, at the fracture zone and fault zone of very-thin coal seams, mainly confining pressure sensors and gas sensors are used. The confining pressure sensors should be arranged at the upper and lower ends of the coal seam and near the fracture zone, especially at positions where the confining pressure changes greatly. The gas sensors should be arranged at the coal seam fracture zone or fissure zone to monitor the gas concentration change in real time. The temperature sensors can be arranged around the fissure area and fault area to monitor the influence of temperature rise on gas release.

[0017] For category II, at the mining influence area of very-thin coal seams, mainly confining pressure sensors, temperature sensors, and gas sensors are used: Due to the drastic changes in the coal seam stress field and temperature field caused by mining activities, the confining pressure, temperature, and gas concentration need to be comprehensively monitored. The confining pressure sensors should be arranged at different depths of the coal seam, especially near the mining face, goaf, and upper and lower end boundaries of the coal seam. The temperature sensors should be arranged in the middle of the coal seam and around the goaf to capture the influence of temperature change on gas desorption during the mining process. The gas sensors should be arranged in areas prone to leakage such as the mining face, goaf, and fissure zone.

[0018] For category III, at the area with uneven porosity of very-thin coal seams, due to the uneven porosity in very-thin coal seams, the occurrence and release characteristics of gas may vary significantly at different positions of the coal seam. Therefore, it is necessary to comprehensively arrange gas sensors and confining pressure sensors. The gas sensors should be arranged in areas with higher porosity and obvious porosity changes in the coal seam to monitor the gas concentration difference. The confining pressure sensors should be arranged at the upper and lower ends of the coal seam and positions with obvious porosity changes to ensure comprehensive monitoring of the confining pressure change. The temperature sensors should be arranged in different porosity areas inside and outside the coal seam, especially in areas with obvious porosity changes, to analyze the influence of temperature on gas desorption.

[0019] Category Ⅳ: In the area prone to gas outburst in extremely thin coal seams, the area prone to gas outburst may be accompanied by a sharp increase in gas concentration. Therefore, it is necessary to preferentially monitor the gas concentration and the change of confining pressure. Gas sensors should be arranged in the middle of the coal seam, the fracture-developed area and the area prone to gas outburst. Confining pressure sensors should be arranged at the upper and lower boundaries of the coal seam, the fracture zone and the positions where gas outburst may occur. Temperature sensors can be arranged in the gas outburst area, the bottom of the coal seam and near the mining face to monitor the influence of temperature on gas desorption.

[0020] Category Ⅴ: In the area with rising temperature in extremely thin coal seams (such as spontaneous combustion area or heat source influence area), the area with rising temperature is the key area where spontaneous combustion or external heat source affects the gas desorption of the coal seam. Therefore, temperature monitoring is crucial. Temperature sensors should be arranged in the middle, upper and lower parts of the coal seam in the spontaneous combustion area, especially where the coal seam is affected by the heat source. In addition, temperature sensors can also be arranged in the fracture zone and the surrounding area to monitor the influence of temperature on gas release and desorption. Confining pressure sensors can be arranged at various depths of the coal seam, especially in the area with rising temperature and the goaf. Gas sensors should be arranged in the area where the coal seam is prone to desorb gas to monitor the change of gas concentration.

[0021] By arranging sensors in different areas, it is possible to achieve precise monitoring of the multi-physical field coupling of temperature-confining pressure-gas in extremely thin coal seams by the monitoring sensor network module.

[0022] Preferably, an early warning system for roof fall in extremely thin coal seams based on multi-field dynamic analysis - data fusion and analysis module (parallel multi-grid analysis method);

[0023] Under the data monitoring of the multi-coupling of temperature-confining pressure-gas in extremely thin coal seams, most of the monitored physical quantities are in the low-frequency segment.

[0024] In extremely thin coal seams, the reservoir pressure ranges from 1.69 to 2.98 MPa, and the reservoir pressure gradient ranges from 0.35 to 0.58 MPa / 100 m, which is lower than the normal hydrostatic pressure gradient and belongs to an underpressure reservoir. The reservoir pressure and the burial depth are generally positively correlated. The overall geothermal gradient is 1.8℃ / 100 m, and the well temperature test data shows that the temperature of the thin coal seam ranges from 18.42 to 20.91℃. These data show the relatively stable changes of temperature and pressure, which can be regarded as low-frequency changes.

[0025] The research on the influence of multi-factor coupling and cyclic confining pressure on the CH4 permeability of coal and rock shows that the CH4 permeability is affected by the coupling of volumetric stress, temperature and gas pressure, among which the influence of gas pressure on the CH4 permeability is greater than that of volumetric stress and temperature. These data indicate that under the action of multiple couplings, the change of CH4 permeability may be relatively stable, belonging to low-frequency change. It can be seen that there are a large number of low-frequency segments in the physical quantities monitored in extremely thin coal seams, and relatively large low-frequency errors will be generated during the monitoring process.

[0026] By iterating on grids of different levels, the multigrid algorithm can quickly eliminate error components of all wavelengths, ensuring that the convergence rate does not decrease with the refinement of the grid. Jacobi and Gauss-Seidel iterations converge very slowly for low-frequency errors, while the multigrid method can solve this problem well. The multigrid algorithm effectively accelerates the decay of low-frequency errors through calculations on different grid levels.

[0027] In view of the working environment of extremely thin coal seams, by comparing whether the Jacobi and Guass-Seidel convergence methods are involved or a complete multi-field grid method is used under different matrix dimensions, it can be seen that the multigrid method has great superiority in dealing with low-frequency changes in extremely thin coal seams.

[0028] The parallel multigrid analysis method can utilize the parallelization and data synchronization strategies of this method to achieve the fusion and analysis of different grids. The steps are as follows:

[0029] 1. First, according to the multi-field dynamic data of extremely thin coal seams obtained by the data acquisition module, set reasonable initial values, and perform n iterations on the fine grid with a grid spacing of h to obtain calculate the error

[0030] 2. Convert the error r to the error r' on the coarse grid with a grid spacing of 2h; perform n iterations on A∈ = r' on the coarse grid to obtain ∈ n , calculate the error γ = r' - A∈ n ;

[0031] 3. Interpolate ∈ n on the coarse grid to obtain ∈' on the fine grid, and use as the initial value to iteratively solve After n iterations, a new is obtained, calculate the new error

[0032] Repeat steps 1, 2, and 3 n - 1 times for the collected data, and n iterations of the physical field data collected under the multi-field coupling of temperature-confining pressure-gas in extremely thin coal seams can be realized.

[0033] The use of parallel multigrid analysis method can accelerate the convergence rate of multi-field data collected in extremely thin coal seams. It has a significant effect on the convergence of low-frequency data, ensuring the accuracy of data fusion and analysis.

[0034] Preferably, an early warning system for roof fall in extremely thin coal seams based on multi-field dynamic analysis - an early warning decision-making module (LSTM and time series analysis algorithm);

[0035] LSTM has significant advantages in capturing long time series and complex non-linear relationships, and is suitable for dealing with the complex dynamic changes among various physical quantities in extremely thin coal seams.

[0036] LSTM and the time series analysis algorithm can effectively analyze the complex, non-linear and time-varying data changes under the coupling action of temperature-confining pressure-gas field in extremely thin coal seams for the time series data obtained by the data fusion and analysis module, and give the prediction of future physical quantity changes. The steps are as follows:

[0037] Initialization of weights and biases; Initialize multiple weight matrices and bias terms of LSTM. The time series data obtained by the data fusion and analysis module has an input dimension x t (dimension of the input vector), a hidden layer dimension h t of the LSTM cell.

[0038] Main weight matrices and bias terms: The weight matrix W f and bias b f of the forget gate; The weight matrix W i and bias b i of the input gate; The weight matrix W o and bias b o of the output gate; The weight matrix W c and bias b c used to calculate the candidate cell state;

[0039] Initialize these weight matrices and bias terms to random values in (-0.01 - 0.01); Initialize the cell state C o and the initial hidden state h o to zero vectors.

[0040] Forward propagation; The process where data is input from the input layer, passes through the hidden layer, and finally reaches the output layer. It includes the following steps:

[0041] Forget gate calculation: When new input enters the LSTM network, the forget gate decides which information should be forgotten or retained. The formula is as follows:

[0042] ft = σ(W fh ht-1 +W fx x t +b f )

[0043] where W f is the weight matrix of the forget gate, and b f is the bias term, x t is the input at the current time step, and h t-1 is the hidden state at the previous time step. σ is the sigmoid function that maps the input value to a probability between 0 and 1.

[0044] Input the hidden state h at the previous moment t-1 and the feature x at the current moment t , and output a number f between 0 and 1 through the sigmoid layer t , where 1 means "completely retain this content", and 0 means "completely discard this content". When f t is close to 1, the past information will be completely retained; when f t is close to 0, the past information will be completely forgotten.

[0045] Input gate and candidate memory cell calculation: When new input enters the LSTM network, the input gate determines which information should be retained and updated in the cell state.

[0046] The input gate uses a sigmoid function to determine which information needs to be retained. At each time step t, the calculation formula of the input gate is as follows:

[0047] i t = σ(W i · [h t-1 , x t + b i

[0048] where W i is the weight matrix of the input gate, b i is the bias term of the input gate, x t is the input at the current time step, and h t-1 is the hidden state at the previous time step. σ is the sigmoid function that maps the input value to a probability between 0 and 1.

[0049] The output value i of the input gate t is a value between 0 and 1, indicating which new input should be retained. When i t is close to 1, all new inputs will be completely retained; when i t is close to 0, all new inputs will be completely ignored.

[0050] Next, the LSTM calculates the candidate cell state It represents how much the new input at the current time step can affect the cell state. The formula for calculating the candidate cell state is as follows:

[0051]

[0052] where, W c is the weight matrix of the candidate cell state, b c is the bias term, x t is the input at the current time step, h t-1 is the hidden state at the previous time step, and tanh is the hyperbolic tangent function that maps the input value to a value between -1 and 1.

[0053] The role of the input gate is to control the weight of the new input at the current time step. Through the input gate, the LSTM can better process long sequence data, avoid the problems of gradient disappearance and gradient explosion, and thus improve the performance and stability of the model.

[0054] Cell state - Memory cell update: The cell state can be regarded as the core of the entire LSTM network. It can store and transmit information, and at the same time can also control the flow and update of information.

[0055] The cell state of the LSTM will be updated and passed to the next time step. At each time step t, the update formula for the cell state is as follows:

[0056]

[0057] where, f t is the forget gate, representing the weight for forgetting the cell state; i t is the input gate, representing the weight for updating the cell state; is the candidate cell state at the current time step, representing how much the new input at the current time step can affect the cell state. When f t is close to 1, the past information will be completely retained; when f t is close to 0, the past information will be completely forgotten. When i t is close to 1, the new input will be completely retained; when i t is close to 0, the new input will be completely ignored.

[0058] Output gate and hidden state update: When it is necessary to pass the information at the current time step to the next layer or the output layer, the output gate is needed to control which information should be output.

[0059] The output gate uses a sigmoid function to determine which information needs to be output. At each time step t, the formula for the output gate is as follows:

[0060] o t = σ(W o [h t-1 , x t + b o )

[0061] where W o is the weight matrix of the output gate, b o is the bias term, x t is the input at the current time step, and h t-1 is the hidden state at the previous time step. σ is the sigmoid function that maps the input value to a probability value between 0 and 1.

[0062] The input o t to the input gate is a value between 0 and 1, indicating which information should be output. When o t is close to 1, all information will be fully retained; when o t is close to 0, all information will be fully blocked.

[0063] Similarly, the inputs h t-1 and x t pass through the sigmoid layer to output o t , which is then multiplied by the output value of C t passing through the tanh layer, and finally the hidden state h t at the current time is output:

[0064] h t = o t * tanh(C t )

[0065] where tanh is the hyperbolic tangent function that maps the input value to a value between -1 and 1.

[0066] Through the output gate, the LSTM can adaptively control the output of information based on the cell state and hidden state at the current time step. By applying the LSTM time series prediction method, important information in the temperature-confining pressure-gas multi-field coupling physical quantities in the monitored extremely thin coal seam can be automatically screened out and transmitted to the next layer or the output layer.

[0067] Backpropagation; the goal of backpropagation is to calculate the gradient of the loss function with respect to each weight and bias through the chain rule (chain differentiation). It includes the following steps:

[0068] Calculate the gradient of the loss function: Starting from the final output, calculate the gradient of the loss function with respect to h t and C t .

[0069] Backpropagation gradient: Gradients are propagated step by step backward from time step t, and the gradients of the forget gate, input gate, output gate, memory cell state, and hidden state are calculated in sequence.

[0070] Calculate the gradients of weights and biases: Using the gradients of each gate, combined with the hidden state and input of the previous time step, calculate the gradients of weights W f ,W i ,W o ,W c , and biases b f ,b i ,b o ,b c .

[0071] Update parameters: According to the calculated gradients, update the weights and biases through an optimization algorithm (such as gradient descent or Adam) to minimize the loss function.

[0072] Finally, use the current hidden state h t and the feature x at the next time step t+1 as inputs, and repeat the above steps until the entire sequence is processed.

[0073] In the monitoring of physical quantities of multi-field coupling of temperature-confining pressure-gas in extremely thin coal seams, compared with traditional time series analysis algorithms, using LSTM can be effectively applied to disaster warning in extremely thin coal seams, and LSTM has significant advantages in capturing long time series and complex non-linear relationships, and is suitable for dealing with the complex dynamic changes between various physical quantities in extremely thin coal seams. It can predict the changes of physical quantities in the future time period under the multi-field coupling of temperature-confining pressure-gas in extremely thin coal seams, and according to the predicted change data, the display and alarm module gives an early warning.

[0074] Preferably, an extremely thin coal seam collapse warning system based on multi-field dynamic analysis - display and alarm module;

[0075] The display and alarm module can monitor the physical quantities of multi-fields of temperature-confining pressure-gas in extremely thin coal seams in real time and display them in the form of charts; it alarms about the changes in the physical quantities detected in the future time period predicted by the early warning decision-making module. When the specified threshold is exceeded, the threshold alarm module is activated to send an alarm message, and the alarm message is synchronized to the cloud, and at the same time, data logging and viewing are carried out.

[0076] Preferably, an extremely thin coal seam collapse warning system based on multi-field dynamic analysis mainly consists of the following parts:

[0077] Monitoring sensor network: This network is responsible for collecting multi-field data such as temperature, confining pressure, and gas in the extremely thin coal seam mining area in real time. Since the degree of physical field coupling influence in different areas is different, the sensor layout is correspondingly divided into five categories. The sensors used include temperature sensors, confining pressure sensors, gas sensors, etc.

[0078] Data acquisition module: This module receives and processes the raw data collected by the monitoring sensors, and converts it into digital signals for analysis.

[0079] Data fusion and analysis module: This module is responsible for fusing, processing, and analyzing multi-field data. Using the parallel multi-grid analysis algorithm, compared with traditional convergence algorithms such as Jacobi and Gauss-Seidel iteration methods, it can converge the high-frequency errors of the extremely thin coal seam and the small errors caused by the multi-field coupling of temperature, confining pressure, and gas more quickly. By performing multiple iterations on the data through the parallel multi-grid algorithm, accurate time series data is obtained, thus ensuring the disaster warning accuracy under the multi-field coupling of the extremely thin coal seam.

[0080] Early warning decision module: Based on the time series data obtained by the data fusion and analysis module, using LSTM and time series analysis algorithms, it warns of disasters in the extremely thin coal seam under multi-field coupling. Compared with traditional time series analysis methods such as ARIMA and GARCH, LSTM and time series analysis algorithms can more effectively analyze the complex, non-linear, and time-varying physical quantity changes in the extremely thin coal seam and predict future physical quantity changes.

[0081] Display and alarm module: According to the prediction results of the early warning decision module, it displays the safety status of the coal seam mining area in real time, and gives an audible and visual alarm to prompt relevant personnel to take emergency measures when abnormalities are detected.

[0082] Compared with the prior art, the beneficial effects of the present invention are:

[0083] Aiming at the particularity of the extremely thin coal seam, according to the degree of disasters caused by the physical quantity coupling in different areas of the extremely thin coal seam, the layout categories of sensors are divided according to the different degrees of physical quantity coupling influence in the extremely thin coal seam.

[0084] Aiming at the large low-frequency errors under the multi-field coupling of temperature-confining pressure-gas in the extremely thin coal seam, using the parallel multi-network analysis method, compared with traditional convergence algorithms such as Jacobi and Gauss-Seidel iteration methods, it can achieve rapid convergence of the low-frequency errors of the extremely thin coal seam, obtain accurate time series data, and ensure the accuracy of early warning.

[0085] Subsequently, considering the complexity, non-linearity, and time-varying nature of the monitored physical quantities in extremely thin coal seams, LSTM and time series analysis are used to predict the changes in the monitored physical quantities in future time periods. When the monitored physical quantities exceed the threshold in the future time period, the display and warning module will display and alarm the monitored physical quantities in real time.

[0086] Thus, the present invention realizes the rapid and accurate monitoring and real-time warning of temperature-confining pressure-gas physical quantities in the working environment of extremely thin coal seams.

[0087] A collapse warning system for extremely thin coal seams based on multi-field dynamic analysis proposed by the present invention aims at the disadvantages such as large low-frequency errors under the multi-field coupling of temperature-confining pressure-gas in extremely thin coal seams, complex, non-linear, and strong time-varying nature of the monitored physical quantities in extremely thin coal seams. It is proposed to use the parallel multi-grid analysis method to achieve rapid convergence of the large low-frequency errors of the monitored physical quantities in extremely thin coal seams, so as to ensure the rapidity and accuracy of monitoring. It is proposed to use LSTM and time series analysis to predict the monitored physical quantities in future time periods, and realize the accurate monitoring and real-time alarm functions under the multi-field action of extremely thin coal seams. Brief Description of the Drawings Attached Figure 1 is the warning flow chart of the present invention; Attached Figure 2 is the schematic diagram of the main components of the present invention; Attached Figure 3 is the distribution diagram of sensors in each area of the extremely thin coal seam in the present invention; Attached Figure 4 is the flow chart of the parallel multi-grid analysis method in the present invention; Attached Figure 5 is the comparison diagram of the parallel multi-network analysis method and the traditional method in low dimensions in the present invention; Attached Figure 6 is the flow chart of the LSTM and time series algorithms in the present invention.

Claims

1. An extremely thin coal seam collapse early warning system based on multi-field dynamic analysis, characterized in that: The system consists of the following parts: Monitoring sensor network: responsible for real-time collection of temperature field, confining pressure field, gas field and other field data in the ultra-thin coal mining area. The arrangement of sensors is divided into five categories according to the different degrees of physical field coupling influence in different areas of the coal seam, including temperature sensors, confining pressure sensors and gas sensors. Data acquisition module: used to receive and process the raw data collected by the monitoring sensors and convert them into digital signals for analysis. Data fusion and analysis module: used to fuse, process and analyze multi-field data, using parallel multi-grid analysis algorithm and multiple iterative processing to obtain accurate time series data. Early warning decision module: Based on the time series data of the data fusion and analysis module, LSTM and time series analysis algorithms are used to warn of disasters under the coupling of multiple fields and predict future changes in multiple physical fields such as temperature, confining pressure, and gas. Display and alarm module: Based on the prediction results of the early warning decision module, the safety status of the coal mining area is displayed in real time. When an abnormality occurs, sound and light alarms are used to prompt relevant personnel to take emergency measures.

2. The ultra-thin coal seam collapse early warning system based on multi-field dynamic analysis according to claim 1, wherein: The sensor arrangement of the monitoring sensor network is divided into the following five categories according to the physical field coupling in different areas of the extremely thin coal seam: Category I: Mainly arranged in extremely thin coal seam fracture zones and fault zones, using confining pressure sensors and gas sensors. Category II: Arranged in the area affected by extremely thin coal seam mining, using confining pressure sensors, temperature sensors and gas sensors. Category III: Arranged in the area of ​​uneven porosity in extremely thin coal seams, using gas sensors and confining pressure sensors. Category IV: It is arranged in areas prone to gas outbursts and mainly monitors changes in gas concentration and confining pressure. Category V: Arranged in temperature rising areas, such as spontaneous combustion areas or heat source affected areas, mainly monitoring temperature changes.

3. The ultra-thin coal seam collapse early warning system based on multi-field dynamic analysis according to claim 1, wherein: The data fusion and analysis module uses a parallel multi-grid analysis algorithm, which implements data processing through the following steps: Perform preliminary calculations on the multi-field dynamic data obtained by the data acquisition module and calculate the errors. The error is converted into error on a coarse grid and further iterated to obtain accurate time series data.

4. The ultra-thin coal seam collapse early warning system based on multi-field dynamic analysis according to claim 1, wherein: The early warning decision module uses LSTM and time series analysis algorithm, which includes the following steps: Initialize the weights and biases in the LSTM network. Forward propagation calculation, update of forget gate, input gate, output gate and cell state. Based on the time series data of the data fusion and analysis module, the future changes in physical quantities such as temperature, confining pressure, and gas are predicted.

5. The ultra-thin coal seam collapse early warning system based on multi-field dynamic analysis according to claim 1, wherein: The display and alarm module displays the safety status of the coal mining area in real time according to the prediction results of the early warning decision module, and prompts relevant personnel to take emergency measures through sound and light alarms when abnormalities occur.

6. A very thin coal seam collapse early warning method based on multi-field dynamic analysis, characterized in that: The following steps are involved: Real-time monitoring of temperature, confining pressure, gas and other physical quantities of extremely thin coal seams to generate monitoring data. Use the data acquisition module to convert the monitoring data into digital signals for analysis. In the data fusion and analysis module, a parallel multi-grid algorithm is used to iterate the data multiple times to obtain accurate time series data. In the early warning decision module, LSTM and time series analysis algorithms are used to analyze and predict future changes in multiple physical quantities to provide disaster warnings. In the display and alarm module, the safety status of the coal seam is displayed in real time and an alarm is issued based on the early warning results.

7. The early warning method according to claim 6, wherein: The data fusion and analysis module uses a parallel multi-grid analysis algorithm, which processes the data through the following steps: Perform preliminary calculations on the collected multi-field data and obtain the errors. The error is transferred to the coarse grid and further iterated to accelerate convergence.

8. The early warning method according to claim 6, wherein: LSTM and time series analysis algorithms analyze time series data to predict future changes in physical quantities such as temperature, confining pressure, and gas in extremely thin coal seams, and issue disaster warnings based on the prediction results.

9. The early warning method according to claim 6, wherein: The display and alarm module displays the safety status of the coal mining area in real time according to the prediction results, and sends out an alarm signal when an abnormal situation exceeding the threshold is detected.

Images