A coal mine disaster risk prediction and assessment method, system and medium

Abstract

The present invention relates to the field of coal mine safety technology, and in particular to a method, system and medium for predicting and evaluating coal mine disaster risks. The method comprises the following steps: deploying intelligent monitoring equipment on the coal mining face of the coal mine, and constructing an intelligent monitoring layout network; collecting multi-source vibration wave signals according to the intelligent monitoring layout network, and performing mechanical noise suppression processing to generate noise-reduced vibration wave signal data; distinguishing the rupture mode according to the noise-reduced vibration wave signal data to obtain roof rupture mode data; predicting the risk of roof falling according to the roof rupture mode data to obtain roof falling risk prediction data; performing risk warning processing according to the roof falling risk prediction data to generate coal mine disaster warning data. The present invention intelligently identifies the changing trend of coal mine roof activity and impact events, captures the early signs of coal rock mass rupture in advance, and realizes the high-precision prediction of coal mine roof collapse risk, which significantly improves the safety production level of coal mines.

Description

Technical Field

[0001] The present invention relates to the field of coal mine safety technology, and in particular to a coal mine disaster risk prediction and assessment method, system and medium. Background Art

[0002] With the continuous growth of global energy demand, coal mines, as one of the important energy sources, play an indispensable role in promoting economic development and meeting social needs. However, the operating environment of coal mines is complex and the safety risks they face are relatively high. Coal mine roof collapse is one of the most common and dangerous disasters in coal mining. The roof refers to the upper part of the working face and the surrounding rock formations during coal mining. The stability of the roof structure directly affects the safety of coal mining. In long-term coal mining operations, the roof deforms, cracks, and even collapses under stress, posing great safety hazards to miners and equipment. However, traditional coal mine roof collapse risk prediction and assessment methods mainly rely on experience judgment and manual monitoring, such as observing the cracks in the roof, knocking on the roof to listen to the sound, and measuring the delamination of the roof. These methods have the disadvantages of strong subjectivity, low accuracy, and low efficiency. They cannot capture the early signs of coal rock fracture in real time, and cannot dynamically reflect the changes in the stability of the roof, resulting in delayed prediction and failure to provide disaster warnings in time. Summary of the invention

[0003] Based on this, the present invention provides a coal mine disaster risk prediction and assessment method, system and storage medium to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a coal mine disaster risk prediction and assessment method comprises the following steps:

[0005] Step S1: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data; deploy intelligent monitoring equipment according to the coal mining field monitoring area data, thereby building an intelligent monitoring layout network;

[0006] Step S2: Adaptively adjust the monitoring frequency according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data;

[0007] Step S3: Processing the roof vibration trend according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively; evaluating the destruction degree index of the roof impact trend data through the roof activity change trend data, and distinguishing the rupture mode to obtain roof rupture mode data;

[0008] Step S4: Perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring timestamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis on the three-dimensional point cloud data of the mining face through the roof rupture mode data to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data;

[0009] Step S5: Perform weighted stability index processing based on the roof fall risk prediction data to generate a mining face top stability index; perform risk warning processing based on the mining face top stability index to generate coal mine disaster warning data.

[0010] The present invention can effectively identify potential dangerous areas in coal mining operations by dividing the key monitoring areas of the coal mining working face. According to the generation of the monitoring area data of the coal mining field, it is ensured that the characteristic data of different areas of the coal mining working face are fully considered, so that the monitoring work is more targeted. The deployment of intelligent monitoring equipment, especially in combination with the actual geological conditions and working face characteristics, constructs a comprehensive intelligent monitoring layout network. The network can not only realize multi-point and wide-coverage real-time monitoring in the region, but also ensure the stability and continuity of the monitoring data. According to the intelligent monitoring layout network, the adaptive monitoring frequency adjustment can optimize the working frequency of the sensor, so that the monitoring equipment can better adapt to the monitoring needs and working environment of different regions. This adaptive adjustment effectively improves the accuracy of signal acquisition and reduces the data loss or deviation during the monitoring process. Through the collection of multi-source vibration wave signals, the dynamic change information of the working face and the roof can be fully captured. At the same time, the mechanical noise suppression processing can effectively filter out the interference signals generated by the operation of mechanical equipment during coal mining, ensuring the purity of the signal. By processing the noise reduction vibration wave signal data, the roof activity change trend data and the roof impact trend data can be obtained, and these two trend data directly reflect the dynamic change of the roof. Roof activity change trend data can help identify the movement state of the roof, while roof impact trend data can capture the impact behavior of the roof when it is subjected to stress concentration. By combining roof activity change trend data with roof impact trend data to evaluate the degree of destruction index, the stress state and potential rupture risk of the roof can be accurately evaluated, and then the rupture mode of the roof can be identified, making the evaluation of roof stability more detailed and quantitative, which helps to identify the rupture point in advance and prevent sudden collapse of the roof. According to the coal mine monitoring timestamp data, three-dimensional laser scanning processing is performed, and the generated three-dimensional point cloud data of the mining face can accurately restore the three-dimensional structure of the coal mining face and the surrounding geological environment. By analyzing the three-dimensional point cloud data of the mining face, the dynamic working conditions of the roof activity can be grasped in detail, and the generated roof activity working condition data can help to fully understand the behavior mode of the roof under different stresses. Further prediction of the falling risk of the roof activity working condition data can accurately identify the roof falling area and generate roof falling risk prediction data. This prediction process provides timely and effective early warning signals for coal mine safety management and reduces the potential risk of roof collapse. 3D laser scanning and processing based on the coal mine monitoring timestamp data can accurately restore the 3D structure of the coal mining face and the surrounding geological environment. By analyzing the 3D point cloud data of the mining face, the dynamic working conditions of the roof activity can be grasped in detail, which helps to fully understand the behavior pattern of the roof under different stresses. Further prediction of the falling risk of the roof activity working condition data can accurately identify the risk of roof falling.According to the prediction data of roof fall risk, weighted stability index processing can be performed to comprehensively and quantitatively evaluate the stability risks of different regions. The stability index can more comprehensively reflect the stability of the roof of the entire coal mining face by weighted processing of the risks of different regions. Based on the risk warning processing of the top stability index of the mining face, timely measures can be taken before major damage to the roof occurs to avoid disasters, greatly improving the safety and production efficiency of coal mine operations. Therefore, a coal mine disaster risk prediction and assessment method of the present invention deploys high-frequency and low-frequency vibration sensors in different areas of the working face, collects the vibration signals of the roof in a comprehensive and real-time manner, and then performs noise reduction, time-frequency analysis, feature extraction and other processing, so as to identify the key information such as the frequency change trend of the roof activity, the characteristics of the impact event and the rupture mode, and capture the early signs of the rupture of the roof rock mass. On this basis, combined with the support working condition response analysis, a pre-trained long short-term memory network model is used to construct a roof risk prediction model to predict the probability of roof collapse in the future period of time, and to give an early warning according to the risk level, so as to achieve high-precision roof collapse risk prediction and significantly improve the safety production level of coal mines.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data, wherein the coal mining field monitoring area data includes the intersection of the working face tunnels, the weak support area, the top beam area of ​​the working face support, the vicinity of the coal wall cracks, the tunnel surrounding rock wall area and the coal rock surface area;

[0013] Step S12: Performing monitoring point layout analysis based on the coal mining field monitoring area data to obtain monitoring layout network data;

[0014] Step S13: classify the multi-source monitoring points of the monitoring layout network data, and mark the intersection of the working face and tunnel, the surrounding rock wall area of ​​the tunnel, and the weak area of ​​the support as far-field low-frequency monitoring point data;

[0015] Step S14: marking the top beam area of ​​the support of the working face, the surface area of ​​the coal and rock mass, and the vicinity of the coal wall cracks as near-field high-frequency monitoring point data;

[0016] Step S15: Use low-frequency vibration sensors to deploy far-field monitoring equipment for far-field low-frequency monitoring point data, and use high-frequency vibration sensors to deploy near-field monitoring equipment for near-field high-frequency monitoring point data, thereby building an intelligent monitoring layout network.

[0017] The present invention can effectively identify potential risk areas in the coal mining process by dividing the key monitoring areas of the coal mining face. By clearly dividing these areas, the important parts of the coal mining field can be fully and systematically covered to ensure the comprehensiveness and accuracy of the monitoring data. The layout analysis of the monitoring points based on the monitoring area data of the coal mining field can ensure the reasonable distribution of the monitoring equipment in the entire coal mining face and avoid the occurrence of monitoring blind spots. Through this layout analysis process, the layout position of the sensor can be optimized to ensure that the monitoring system can capture the dynamic change data of each key area comprehensively and in real time. By classifying the multi-source monitoring points of the monitoring layout network data, it is possible to achieve targeted division of the monitoring needs of different areas, which can not only ensure that different monitoring points are appropriately deployed according to their actual vibration characteristics, but also improve the accuracy of data acquisition through targeted monitoring, and reduce monitoring blind spots or false alarms caused by inappropriate equipment deployment. By using low-frequency vibration sensors to deploy far-field monitoring equipment for far-field low-frequency monitoring point data, we have built an intelligent monitoring layout network that covers the entire field and is highly targeted. This can comprehensively monitor vibration conditions within the coal mining face, which not only improves the efficiency and accuracy of monitoring, but also significantly improves the timeliness and effectiveness of coal mine disaster prediction and early warning.

[0018] Preferably, step S2 comprises the following steps:

[0019] Step S21: Initializing the monitoring frequency of the intelligent monitoring layout network through a preset initial sensor monitoring frequency to obtain initial monitoring frequency data;

[0020] Step S22: Counting the number of people working in the coal mining face in real time to generate real-time operator data;

[0021] Step S23: performing monitoring importance assessment according to the real-time operator data, and adaptively adjusting the initial monitoring frequency data to generate adaptive monitoring frequency data;

[0022] Step S24: performing sensor clock synchronization processing on the intelligent monitoring layout network, and collecting multi-source vibration wave signals according to the adaptive monitoring frequency data, to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data respectively;

[0023] Step S25: performing monitoring signal preprocessing on the original coal mining vibration wave signal data to obtain near-field vibration wave signal data and far-field vibration wave signal data respectively;

[0024] Step S26: using the intelligent monitoring layout network to perform mechanical noise suppression processing on the near-field vibration wave signal data and the far-field vibration wave signal data to generate noise-reduced vibration wave signal data.

[0025] The present invention initializes the monitoring frequency of the intelligent monitoring layout network through the preset initial sensor monitoring frequency, which can ensure that the monitoring frequency of all sensors is in the optimal state when the monitoring system is started, and can capture the vibration signal that meets the actual conditions of the coal mining face according to the preset parameters. According to the real-time number of operators in the coal mining face of the coal mine, the actual distribution of personnel in the current coal mining operation can be dynamically understood, and the priority and importance of monitoring can be further evaluated, and the changes in the working environment can be adapted. According to the real-time operator data, the importance of monitoring can be evaluated, and the focus of the monitoring system can be dynamically adjusted to ensure that high-risk areas and areas with a large number of operators can obtain higher frequency and more accurate monitoring. The initialization monitoring frequency data is adaptively adjusted to ensure that the monitoring frequency of the sensor matches the current working environment, and the generated adaptive monitoring frequency data reflects the real-time monitoring demand changes. By performing sensor clock synchronization processing on the intelligent monitoring layout network, the time data of different monitoring points can be ensured to be consistent, avoiding data errors caused by time differences. According to the adaptive monitoring frequency data, multi-source vibration wave signal acquisition can be performed to fully capture the vibration change information in the coal mining face. Preprocessing the original coal mining vibration wave signal data can effectively separate the vibration signal characteristics of different regions, reduce the interference of noise on the data, and analyze the vibration behavior of different regions more targeted, ensuring that high-frequency and low-frequency signals can be accurately captured and interpreted. Using the intelligent monitoring layout network to perform mechanical noise suppression on the near-field vibration wave signal data and the far-field vibration wave signal data can effectively eliminate the impact of external interference such as the operation of coal mining equipment on the vibration wave signal.

[0026] Preferably, step S26 comprises the following steps:

[0027] Step S261: using a bandpass filter to extract effective frequency bands from the near-field vibration wave signal data and the far-field vibration wave signal data, to obtain effective near-field vibration frequency band data and effective far-field vibration frequency band data respectively;

[0028] Step S262: performing frequency modal decomposition on the effective near-field vibration frequency band data and performing independent component analysis to obtain near-field frequency modal component data;

[0029] Step S263: using a preset mechanical noise classifier to identify mechanical noise components of near-field frequency modal component data, and generating mechanical interference source component data;

[0030] Step S264: using the mechanical interference source component data to remove mechanical noise from the near-field frequency modal component data, and reconstructing the near-field vibration signal to obtain noise-reduced near-field vibration signal data;

[0031] Step S265: using the intelligent monitoring layout network to perform signal space propagation attenuation analysis on the near-field vibration wave signal data and the far-field vibration wave signal data to obtain a signal attenuation adjustment factor;

[0032] Step S266: adjusting the interference source component of the mechanical interference source component data by using the signal attenuation adjustment factor, and performing mechanical noise suppression on the effective far-field vibration frequency band data to obtain noise-reduced far-field vibration signal data.

[0033] The present invention extracts effective frequency bands from near-field vibration wave signal data and far-field vibration wave signal data through a bandpass filter, which can eliminate noise and invalid information in non-target frequency bands, extract frequency signals with actual monitoring significance, and more accurately capture the vibration characteristics of coal rock mass in different areas. Frequency modal decomposition is performed on effective near-field vibration frequency band data to separate vibration characteristics under different modes, and to decompose complex vibration signals into component signals with physical significance. Near-field frequency modal component data is further decomposed through independent component analysis, so that the monitoring system can identify independent vibration sources hidden in complex signals. Mechanical noise component identification is performed on near-field frequency modal component data through a preset mechanical noise classifier, which can automatically filter out noise interference generated by the operation of mechanical equipment from near-field vibration signals, effectively improving the automation and intelligence level of signal analysis, and ensuring that the system can accurately distinguish effective signals from invalid interference signals in vibration signals. Mechanical noise elimination of near-field frequency modal component data using mechanical interference source component data can greatly reduce the interference effect of mechanical equipment operation on vibration signals. By using the intelligent monitoring layout network to perform signal space propagation attenuation analysis on near-field vibration wave signal data and far-field vibration wave signal data, the propagation characteristics of the vibration signal in space can be comprehensively evaluated, and the attenuation of the signal as the distance increases can be determined. By adjusting the mechanical interference source component data through the signal attenuation adjustment factor, the noise source can be effectively compensated and corrected according to the actual signal attenuation, thereby more accurately eliminating the mechanical noise in the far-field vibration wave signal.

[0034] Preferably, step S3 comprises the following steps:

[0035] Step S31: performing roof vibration trend processing according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively;

[0036] Step S32: performing trend energy characteristic analysis on the roof activity change trend data to generate roof activity energy characteristic data;

[0037] Step S33: dynamically correlating the roof impact trend data with the impact waveform through the roof activity energy characteristic data to generate correlated impact fusion data;

[0038] Step S34: deconstructing the main frequency mode according to the associated impact fusion data to obtain impact frequency mode data;

[0039] Step S35: performing multi-scale shock entropy calculation according to the associated shock fusion data to generate multi-scale shock entropy data;

[0040] Step S36: Calculate the entropy distribution density of the multi-scale impact entropy data, and perform entropy increment analysis to generate entropy increment feature data;

[0041] Step S37: evaluating the damage index of the roof impact trend data by using the entropy increment characteristic data to obtain damage index data;

[0042] Step S38: Based on the preset rupture mode library, the rupture mode of the damage degree index data is determined by the impact frequency mode data to obtain the roof rupture mode data.

[0043] The present invention can clearly show the dynamic activity and impact characteristics of the roof by processing the noise-reduced vibration wave signal data according to the roof vibration trend, reflecting the movement trajectory and vibration mode of the roof in different time periods, while the roof impact trend data reveals the energy release and impact wave characteristics of the roof during the coal-rock mass rupture process. Trend energy characteristic analysis of the roof activity change trend data can quantify the energy release of the roof during the vibration process, and can comprehensively reflect the energy distribution and change trend of the roof during vibration. By dynamically correlating the roof impact trend data with the roof activity energy characteristic data, the impact waveforms of different time periods and spatial positions can be fused, effectively improving the overall analytical ability of the impact waveform, so that the system can identify the internal connection and cumulative effect between multiple impact events, thereby more accurately assessing the risk of roof rupture. Deconstructing the main frequency mode based on the correlated impact fusion data can separate the representative frequency components in the roof impact process, effectively decompose the complex impact vibration signal, and extract the frequency mode that is most indicative of the risk of roof rupture. By calculating the multi-scale impact entropy of the associated impact fusion data, the complexity and uncertainty of the roof impact signal can be effectively described, so that the system can quantify the uncertainty and complexity changes of the roof in different impact events, and further reveal the implicit information and rupture risks contained in the impact process. By analyzing the trend of entropy value changes, the energy transfer and distribution conditions that occur during the roof rupture process are revealed. By evaluating the damage index of the roof impact trend data through the entropy increment feature data, the damage risk of the roof during the impact process can be accurately evaluated, and the damage index data can be obtained to quantify the severity of the roof rupture. Based on the preset rupture pattern library, the rupture pattern of the damage index data can be judged through the impact frequency pattern data, and the specific pattern of the roof rupture can be identified. Different types of roof rupture phenomena can be accurately distinguished, so that the system can not only identify the risk of rupture, but also predict the specific form of rupture.

[0044] Preferably, step S31 includes the following steps:

[0045] Step S311: extracting target signals from the noise reduction far-field vibration signal data using a preset roof activity frequency range to obtain roof activity vibration signal data;

[0046] Step S312: Calculate the low-frequency vibration characteristics of the top plate activity vibration signal data to generate low-frequency vibration characteristic data;

[0047] Step S313: performing time-frequency peak analysis on the roof activity vibration signal data through the low-frequency vibration characteristic data, and extracting the roof activity frequency to obtain the roof activity frequency data;

[0048] Step S314: identifying the frequency change trend according to the roof activity frequency data, and generating roof activity change trend data;

[0049] Step S315: performing signal energy mutation identification on the noise reduction near-field vibration signal data to obtain potential impact event point data;

[0050] Step S316: performing transient waveform interception on the noise reduction near-field vibration signal data through the potential impact event point data to generate transient waveform segment data;

[0051] Step S317: calculating the impact intensity of the transient waveform segment data and performing impact event recognition to obtain impact event recognition data;

[0052] Step S318: extracting shock vibration features according to the shock event identification data, and classifying shock events to generate shock event classification data;

[0053] Step S319: Perform impact intensity time series analysis on the impact event classification data to obtain roof impact trend data.

[0054] The present invention uses a preset roof activity frequency range to extract target signals from the noise reduction far-field vibration signal data, which can accurately separate the vibration signals related to the roof activity and effectively filter out other irrelevant vibration interferences. The low-frequency vibration feature calculation of the roof activity vibration signal data can quantify the vibration characteristics of the roof in a lower frequency range, capture the low-frequency motion law of the roof structure, and especially the change characteristics related to the slow activity and stress accumulation of the roof. By performing time-frequency peak analysis on the low-frequency vibration feature data, the significant frequency characteristics in the roof activity vibration signal can be identified, so that the system can extract the core frequency characteristics of the roof activity from the complex vibration signal. Frequency change trend identification is performed based on the roof activity frequency data to reveal the long-term change pattern of the roof activity frequency, which effectively reflects the change in the motion state of the roof in different time periods. Signal energy mutation identification is performed on the noise reduction near-field vibration signal data, which can accurately capture the moment when energy is suddenly released during the roof activity process. Transient waveform interception of the noise reduction near-field vibration signal data through potential impact event point data can capture short-term waveform changes related to the impact event and extract transient vibration characteristics closely related to the impact event. The impact intensity calculation of transient waveform segment data can quantify the energy released by each impact event, accurately identify the occurrence and severity of impact events, and ensure timely response to roof rupture and impact risks. The impact intensity time series analysis of impact event classification data can reveal the frequency and intensity change trend of impact events, and can fully grasp the evolution trend of roof impact risk.

[0055] Preferably, step S4 comprises the following steps:

[0056] Step S41: performing three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face;

[0057] Step S42: Acquire historical mining face three-dimensional point cloud data; perform time series comparison and analysis on the mining face three-dimensional point cloud data using the historical mining face three-dimensional point cloud data to obtain motion deformation difference data;

[0058] Step S43: extracting motion feature parameters from the motion deformation difference data to generate deformation motion feature data;

[0059] Step S44: Based on the roof rupture mode data, the working condition simulation of the coal wall spalling and the mining face roof is performed through the deformation movement characteristic data to generate the roof activity working condition data;

[0060] Step S45: performing support condition response analysis according to the top plate activity condition data to generate support response characteristic data;

[0061] Step S46: using a preset long short-term memory network model to perform transfer learning on the roof activity condition data and the support response characteristic data to obtain a roof risk prediction model;

[0062] Step S47: Use the roof risk prediction model to predict the risk of roof falling and obtain roof falling risk prediction data.

[0063] The present invention uses coal mine monitoring timestamp data to perform three-dimensional laser scanning processing on the coal mining face, and can accurately capture the spatial form of the face at different times. By acquiring the three-dimensional point cloud data of the historical mining face, and using these data to perform time series comparison and analysis on the three-dimensional point cloud data of the current mining face, the deformation of the face at different time points can be accurately detected, and the cumulative deformation characteristics of the mining face during the long-term mining process can be revealed, providing key deformation parameters for identifying potential risks in the roof and coal wall areas. Extracting motion characteristic parameters from the motion deformation difference data to generate deformation motion characteristic data is helpful to analyze the dynamic change characteristics of the mining face, including parameters such as deformation rate and deformation amplitude, and can identify local abnormal deformation areas. Based on the roof rupture mode data, the working condition activity simulation of the coal wall spalling and the mining face roof is performed through the deformation motion characteristic data, and unstable areas can be identified. By analyzing the support response according to the roof activity data and generating support response characteristic data, we can fully understand the stress state and response characteristics of the support under different working conditions. By accurately calculating the stress state of the support structure, we can evaluate the stability of the support and effectively predict the potential failure risk of the support under different roof activity conditions. By using the characteristics of historical data and real-time data, we built a deep learning model for roof activity and support response, which enhanced the prediction accuracy of roof risk. The long short-term memory network provides an intelligent prediction method for predicting disaster risks by capturing the time series characteristics of roof activity and the dynamic changes of support response. The roof risk prediction model is used to predict the risk of roof falling. By analyzing the dynamic trend of roof activity, the stress state and deformation characteristics of the support, it can warn of potential roof falling risks in advance, reduce the possibility of accidents, and ensure the safety of coal mine operations.

[0064] Preferably, step S5 comprises the following steps:

[0065] Step S51: performing spatial risk probability mapping on the coal mining face of the coal mine through the roof fall risk prediction data to obtain the risk probability distribution data of the coal mining face;

[0066] Step S52: Calculate the weighted stability index based on the risk probability distribution data of the mining face to generate the top stability index of the mining face;

[0067] Step S53: performing comprehensive index summation processing on the top stability index of the mining face to obtain comprehensive assessment data of mining face risk;

[0068] Step S54: Risk levels are divided according to the comprehensive risk assessment data of the mining face, and risk warning processing is performed according to the mining face top stability index to generate coal mine disaster warning data.

[0069] The present invention maps the spatial risk probability of the coal mining face of the coal mine through the roof fall risk prediction data, can fully understand the risk level of different areas, can convert the roof fall risk data into a spatial distribution form, so that the spatial distribution of the risk becomes clear and intuitive. It is helpful to identify high-risk areas and low-risk areas. According to the risk probability distribution data of the mining face, a weighted stability index is calculated, and an index that comprehensively reflects the stability of the working face is calculated by combining the risk probability data with the actual stability index. The top stability index of the mining face can comprehensively consider the risk probability and stability level of different regions. The top stability index of the mining face is processed by comprehensive index summation, and the weighted results of multiple indicators are summarized to form a comprehensive risk assessment data, which can comprehensively reflect the overall risk status of the mining face and more accurately identify the overall risk level. According to the comprehensive assessment data of the mining face risk, the risk level division is carried out, and the risk level division can divide the mining face risk into different levels, clarify the risk situation corresponding to each level, and facilitate the formulation of specific response measures. The risk warning processing can send an alarm to relevant personnel in real time, prompt potential dangers, and suggest taking corresponding safety measures, effectively improve the early warning capability and response efficiency of coal mine disasters, and ensure the safety of coal mine operations.

[0070] Preferably, the present invention further provides a coal mine disaster risk prediction and assessment system, which executes the coal mine disaster risk prediction and assessment method as described above, and the coal mine disaster risk prediction and assessment system comprises:

[0071] The coal mine monitoring layout module is used to divide the key monitoring areas of the coal mining face of the coal mine and generate the monitoring area data of the coal mining field; intelligent monitoring equipment is deployed according to the monitoring area data of the coal mining field, thereby building an intelligent monitoring layout network;

[0072] The monitoring signal processing module is used to perform adaptive monitoring frequency adjustment according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data;

[0073] The rupture mode recognition module is used to process the roof vibration trend according to the noise reduction vibration wave signal data, and obtain the roof activity change trend data and the roof impact trend data respectively; the destruction degree index of the roof impact trend data is evaluated through the roof activity change trend data, and the rupture mode is identified to obtain the roof rupture mode data;

[0074] The roof fall risk prediction module is used to perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis based on the three-dimensional point cloud data of the mining face to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data;

[0075] The coal mine risk warning module is used to perform weighted stability index processing based on the roof fall risk prediction data to generate the mining face top stability index; perform risk warning processing based on the mining face top stability index to generate coal mine disaster warning data.

[0076] Preferably, a coal mine disaster risk prediction and assessment storage medium stores a computer program, and when the computer program is executed, a coal mine disaster risk prediction and assessment method as described in any one of the above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] Figure 1 It is a schematic diagram of the steps of the method for predicting and evaluating the risk of coal mine disasters of the present invention;

[0078] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0079] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0080] Figure 4 for Figure 1 Detailed implementation steps of step S5;

[0081] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0082] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0083] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0084] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0085] To achieve this, please refer to Figures 1 to 4 The present invention provides a method for predicting and evaluating coal mine disaster risks, comprising the following steps:

[0086] Step S1: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data; deploy intelligent monitoring equipment according to the coal mining field monitoring area data, thereby building an intelligent monitoring layout network;

[0087] Step S2: Adaptively adjust the monitoring frequency according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data;

[0088] Step S3: Processing the roof vibration trend according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively; evaluating the destruction degree index of the roof impact trend data through the roof activity change trend data, and distinguishing the rupture mode to obtain roof rupture mode data;

[0089] Step S4: Perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring timestamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis on the three-dimensional point cloud data of the mining face through the roof rupture mode data to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data;

[0090] Step S5: Perform weighted stability index processing based on the roof fall risk prediction data to generate a mining face top stability index; perform risk warning processing based on the mining face top stability index to generate coal mine disaster warning data.

[0091] In the embodiment of the present invention, reference Figure 1 The above is a schematic diagram of the steps of a method for predicting and evaluating coal mine disaster risks of the present invention. In this embodiment, the method for predicting and evaluating coal mine disaster risks includes the following steps:

[0092] Step S1: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data; deploy intelligent monitoring equipment according to the coal mining field monitoring area data, thereby building an intelligent monitoring layout network;

[0093] In an embodiment of the present invention, according to the coal mining process, geological conditions and historical monitoring data, the working face is divided into six key monitoring areas: the intersection of working face tunnels, the weak area of ​​the support, the top beam area of ​​the working face support, the vicinity of the coal wall fissure, the surrounding rock wall area of ​​the tunnel and the surface area of ​​the coal rock mass. Low-frequency vibration sensors are deployed at the intersection of working face tunnels, the surrounding rock wall area of ​​the tunnel and the weak area of ​​the support. For example, a low-frequency moving coil velocity sensor with a frequency range of 1Hz-50Hz is used and installed on the tunnel wall to monitor the vibration of the surrounding rock of the tunnel so as to capture long-distance, low-frequency vibration signals. High-frequency vibration sensors are deployed in the top beam area of ​​the working face support, the surface area of ​​the coal rock mass and the vicinity of the coal wall fissure. For example, a high-frequency piezoelectric acceleration sensor with a frequency range of 100Hz-10kHz is used and installed on the top beam of the support to monitor the microseismic signals of the roof so as to capture close-range, high-frequency vibration signals. All sensors are connected to the data acquisition and processing center through a wireless network to build an intelligent monitoring layout network covering the entire working face.

[0094] Step S2: Adaptively adjust the monitoring frequency according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data;

[0095] In an embodiment of the present invention, at the initial stage of system deployment, initial monitoring frequencies are preset for different types of sensors in the intelligent monitoring layout network based on experience and historical data. For far-field low-frequency monitoring point data, the initial monitoring frequency is set to 100Hz; for near-field high-frequency monitoring point data, the initial monitoring frequency is set to 1000Hz. The personnel positioning system is used to count the number of workers in the coal mining face of the coal mine in real time, and the importance of monitoring of different monitoring areas is evaluated based on the real-time worker data, and the initialization monitoring frequency data is adaptively adjusted. Multi-source vibration wave signals are collected using the deployed intelligent monitoring equipment, and mechanical noise suppression processing is performed on the collected original coal mining vibration wave signal data. For example, the wavelet transform method is used to reduce the noise of microseismic signals to remove the mechanical noise generated by equipment such as coal mining machines and scraper conveyors; the spectrum analysis method is used to reduce the noise of acoustic emission signals to remove environmental noise and equipment noise.

[0096] Step S3: Processing the roof vibration trend according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively; evaluating the destruction degree index of the roof impact trend data through the roof activity change trend data, and distinguishing the rupture mode to obtain roof rupture mode data;

[0097] In an embodiment of the present invention, the noise-reduced vibration wave signal data is analyzed to extract indicators that can reflect the characteristics of roof activity, such as vibration amplitude, frequency, energy, etc. These indicators are subjected to time-frequency analysis using methods such as sliding windows to obtain roof activity change trend data and roof impact trend data, respectively. Based on the roof activity change trend data, the damage degree index of the roof impact event is calculated, for example, an energy-based indicator is used to quantify the degree of damage caused to the roof rock mass by each impact event. The rupture mode of the roof rock mass is analyzed by combining the damage degree index and the roof impact trend data. For example, based on the energy size, occurrence frequency, spatial distribution and other characteristics of the impact event, it is determined whether the roof rock mass is in a stable state, a micro-fracture state, or is about to collapse on a large scale. Finally, the roof rupture mode data is obtained.

[0098] Step S4: Perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring timestamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis on the three-dimensional point cloud data of the mining face through the roof rupture mode data to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data;

[0099] In an embodiment of the present invention, the coal mining face of the coal mine is subjected to three-dimensional laser scanning processing using the coal mine monitoring timestamp data. For example, a three-dimensional laser scanner is used to scan the face to obtain the face point cloud data, and the timestamp data is used to associate the data of each scan to construct a time series of the three-dimensional point cloud data of the mining face. The working condition activity analysis is performed based on the three-dimensional point cloud data of the mining face. For example, by comparing the point cloud data at different time points, the deformation, displacement and collapse of the roof are analyzed; the point cloud data is used to extract the roof crack information, and the distribution, expansion and penetration of the cracks are analyzed. A roof fall risk prediction model is constructed using machine learning algorithms, such as support vector machines, random forests, neural networks, etc. The input features of the model may include roof activity working condition data, roof rupture mode data, geological condition parameters, etc. The output of the model is roof fall risk prediction data, for example, the possibility of roof fall in the future period of time is expressed in the form of probability.

[0100] Step S5: Perform weighted stability index processing based on the roof fall risk prediction data to generate a mining face top stability index; perform risk warning processing based on the mining face top stability index to generate coal mine disaster warning data.

[0101] In an embodiment of the present invention, the weighted stability index of the coal mining face is calculated according to the prediction data of the risk of roof falling, combined with the weight coefficients of different risk levels. For example, the working face can be divided into four levels of safety, low risk, medium risk and high risk according to the risk prediction results, and different weight coefficients are assigned. The weighted stability index can comprehensively reflect the stability of the roof of the current working face and generate an intuitive top stability index of the mining face. For example, the index range can be set to 0-100, and the higher the value, the more stable the roof is, and vice versa, it means that there is a risk of collapse of the roof. The top stability index of the mining face is graded and warned according to the preset threshold. For example, when the index is lower than 20, a red warning is triggered, indicating that there is an extremely high risk of collapse of the roof, and it is necessary to stop work and evacuate people immediately; when the index is lower than 50, an orange warning is triggered, indicating that there is a high risk of collapse of the roof, and it is necessary to strengthen monitoring and take necessary support measures; when the index is lower than 80, a yellow warning is triggered, indicating that there is a certain risk of collapse of the roof, and it is necessary to pay attention to the changes in the roof. For example, when the top stability index of the mining face is lower than the orange warning threshold, the system automatically sends warning information to the mine dispatching room, safety monitoring center and underground workers.

[0102] Preferably, step S1 comprises the following steps:

[0103] Step S11: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data, wherein the coal mining field monitoring area data includes the intersection of the working face tunnels, the weak support area, the top beam area of ​​the working face support, the vicinity of the coal wall cracks, the tunnel surrounding rock wall area and the coal rock surface area;

[0104] Step S12: Performing monitoring point layout analysis based on the coal mining field monitoring area data to obtain monitoring layout network data;

[0105] Step S13: classify the multi-source monitoring points of the monitoring layout network data, and mark the intersection of the working face and tunnel, the surrounding rock wall area of ​​the tunnel, and the weak area of ​​the support as far-field low-frequency monitoring point data;

[0106] Step S14: marking the top beam area of ​​the support of the working face, the surface area of ​​the coal and rock mass, and the vicinity of the coal wall cracks as near-field high-frequency monitoring point data;

[0107] Step S15: Use low-frequency vibration sensors to deploy far-field monitoring equipment for far-field low-frequency monitoring point data, and use high-frequency vibration sensors to deploy near-field monitoring equipment for near-field high-frequency monitoring point data, thereby building an intelligent monitoring layout network.

[0108] In the embodiment of the present invention, according to the geological structure and mining technology of the working face, it is divided into six key monitoring areas: The intersection of working face tunnels: This area is prone to roof collapse due to the staggered tunnel excavation and support, stress concentration, and needs to be monitored. Weak support area: refers to the area where the support bearing capacity is insufficient due to changes in geological conditions or improper support, and monitoring needs to be strengthened. The top beam area of ​​the working face support: This area directly bears the roof pressure and is the direct occurrence area of ​​roof collapse, which requires high-frequency and real-time monitoring. Near the coal wall crack: The coal wall crack is an important precursor to the instability of the roof, and it needs to be closely monitored to detect the risk of roof collapse in time. The surrounding rock wall area of ​​the tunnel: The stability of the surrounding rock of the tunnel directly affects the stability of the roof, and the deformation and damage of the surrounding rock of the tunnel need to be monitored. The surface area of ​​the coal rock mass: By monitoring the stress changes, crack development, etc. in this area, the stability of the roof can be indirectly judged. Load the three-dimensional model of the working face and the monitoring area data of the coal mining field in the GIS system, and set the parameters such as the sensor monitoring range and the optimal monitoring distance. Then, using the spatial analysis function, candidate monitoring points are automatically generated, and they are screened and optimized according to factors such as the importance of the monitoring area, the accessibility of the monitoring points, and the reliability of signal transmission. Finally, the specific location coordinates of each monitoring point are determined, and the monitoring layout network data is generated. The intersection of the working face tunnels, the surrounding rock wall area of ​​the tunnels, and the weak support area are marked as far-field low-frequency monitoring point data. These areas are far away from the core area of ​​roof collapse, and mainly monitor the large-scale, low-frequency movement trend of the roof, such as slow sinking of the roof, large-area delamination, etc. The top beam area of ​​the working face support, the surface area of ​​the coal rock mass, and the vicinity of the coal wall cracks are marked as near-field high-frequency monitoring point data. These areas are close to the core area of ​​roof collapse, and mainly monitor the local, high-frequency dynamic changes of the roof, such as the generation and expansion of micro-cracks, the impact and vibration of the roof, etc. Low-frequency vibration sensors are used for monitoring. These sensors have high sensitivity and a wide monitoring range, and can capture long-distance, low-frequency vibration signals. For example, a low-frequency dynamic velocity sensor with a frequency range of 1Hz-50Hz is installed on the tunnel wall to monitor the vibration of the tunnel surrounding rock. High-frequency vibration sensors are used for monitoring. These sensors have high sampling frequency and high resolution and can capture close-range, high-frequency vibration signals. For example, a high-frequency piezoelectric acceleration sensor with a frequency range of 100Hz-10kHz is installed on the support top beam to monitor the microseismic signals of the roof.

[0109] Preferably, step S2 comprises the following steps:

[0110] Step S21: Initializing the monitoring frequency of the intelligent monitoring layout network through a preset initial sensor monitoring frequency to obtain initial monitoring frequency data;

[0111] Step S22: Counting the number of people working in the coal mining face in real time to generate real-time operator data;

[0112] Step S23: performing monitoring importance assessment according to the real-time operator data, and adaptively adjusting the initial monitoring frequency data to generate adaptive monitoring frequency data;

[0113] Step S24: performing sensor clock synchronization processing on the intelligent monitoring layout network, and collecting multi-source vibration wave signals according to the adaptive monitoring frequency data, to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data respectively;

[0114] Step S25: performing monitoring signal preprocessing on the original coal mining vibration wave signal data to obtain near-field vibration wave signal data and far-field vibration wave signal data respectively;

[0115] Step S26: using the intelligent monitoring layout network to perform mechanical noise suppression processing on the near-field vibration wave signal data and the far-field vibration wave signal data to generate noise-reduced vibration wave signal data.

[0116] As an example of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation steps of step S2 in the embodiment are shown in the flowchart. In this embodiment, step S2 includes:

[0117] Step S21: Initializing the monitoring frequency of the intelligent monitoring layout network through a preset initial sensor monitoring frequency to obtain initial monitoring frequency data;

[0118] In the embodiment of the present invention, at the initial stage of system deployment, the initial monitoring frequencies are preset for different types of sensors in the intelligent monitoring layout network based on experience and historical data. For example, for the far-field low-frequency monitoring point data, considering that its monitoring target is the large-scale, low-frequency roof movement trend, the initial monitoring frequency is set to 100Hz. For the near-field high-frequency monitoring point data, considering that its monitoring target is the local, high-frequency roof dynamic changes, the initial monitoring frequency is set to 1000Hz.

[0119] Step S22: Counting the number of people working in the coal mining face in real time to generate real-time operator data;

[0120] In the embodiment of the present invention, a personnel positioning system or a video monitoring system is used to count the number of people working in a coal mining face in real time to generate real-time personnel data. For example, a personnel positioning card reader can be installed at the entrance and exit of the coal mining face. When a person enters or leaves the face, the system automatically records the person's information and time, and counts the number of people in the current face.

[0121] Step S23: performing monitoring importance assessment according to the real-time operator data, and adaptively adjusting the initial monitoring frequency data to generate adaptive monitoring frequency data;

[0122] In an embodiment of the present invention, the monitoring importance of different monitoring areas is evaluated according to the real-time operator data, and the initial monitoring frequency data is adaptively adjusted. The monitoring importance level corresponding to different number of operators and the corresponding monitoring frequency adjustment strategy are preset. For example: when the real-time number of operators is less than 5, it is determined to be a low monitoring importance level, and all monitoring points maintain the initial monitoring frequency. When the real-time number of operators is between 5 and 10, it is determined to be a medium monitoring importance level, the monitoring frequency of the near-field high-frequency monitoring point data is doubled, and the monitoring frequency of the far-field low-frequency monitoring point data remains unchanged. When the real-time number of operators exceeds 10, it is determined to be a high monitoring importance level, and the monitoring frequency of all monitoring points is increased to four times the initial value. The system automatically determines the current monitoring importance level based on the real-time operator data, and modifies the sensor monitoring frequency of each monitoring point in real time according to the preset adjustment strategy. For example, the initial frequency is increased by 1.5 times or 2 times. Alternatively, its initial monitoring frequency can be maintained, or its monitoring frequency can be appropriately reduced to save system resources under the premise of ensuring safety.

[0123] Step S24: performing sensor clock synchronization processing on the intelligent monitoring layout network, and collecting multi-source vibration wave signals according to the adaptive monitoring frequency data, to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data respectively;

[0124] In an embodiment of the present invention, before data collection, it is necessary to perform clock synchronization processing on all sensors in the intelligent monitoring layout network to ensure that the data collected by all sensors have the same time reference. For example, the network time protocol (NTP) or GPS clock synchronization technology can be used to synchronize the clocks of all sensors to a unified time server or GPS satellite time. According to the adaptive monitoring frequency data, each sensor in the intelligent monitoring layout network is configured to collect data at a set frequency. For example, data acquisition software or an embedded program can be used to control the sampling frequency of each sensor, and multi-source vibration wave signals, including near-field high-frequency vibration signals and far-field low-frequency vibration signals, can be collected in real time. Record the coal mine monitoring timestamp data of each data collection point, such as a timestamp accurate to milliseconds.

[0125] Step S25: performing monitoring signal preprocessing on the original coal mining vibration wave signal data to obtain near-field vibration wave signal data and far-field vibration wave signal data respectively;

[0126] In an embodiment of the present invention, the original coal mining vibration wave signal data is preprocessed to distinguish near-field and far-field signals. According to the type of sensor and the division of the monitoring area, the preprocessed data is stored as near-field vibration wave signal data and far-field vibration wave signal data respectively. For example, the data collected by the near-field high-frequency vibration sensor can be stored as near-field vibration wave signal data, and the data collected by the far-field low-frequency vibration sensor can be stored as far-field vibration wave signal data. The near-field and far-field vibration wave signal data are respectively detrended to eliminate the baseline drift caused by factors such as sensor drift, and highlight the dynamic change information in the signal. For example, a high-pass filter is used to remove the low-frequency trend item in the signal and retain the high-frequency vibration information.

[0127] Step S26: using the intelligent monitoring layout network to perform mechanical noise suppression processing on the near-field vibration wave signal data and the far-field vibration wave signal data to generate noise-reduced vibration wave signal data.

[0128] In an embodiment of the present invention, an empirical mode decomposition (EMD) method is used to decompose the near-field vibration wave signal data, separate the intrinsic mode function (IMF) containing the main noise component, and remove it, retaining the IMF component containing the real top plate vibration information, thereby achieving noise suppression. An adaptive filtering method is used to process the far-field vibration wave signal data, and the correlation between multiple spatially distributed sensor signals is used to adaptively filter out random noise in the signal and enhance the real signal. For example, a least mean square (LMS) adaptive filtering algorithm is used to automatically adjust the filter coefficient according to the error between the reference signal (such as a sensor signal far from the noise source) and the input signal (such as a sensor signal close to the noise source) to achieve noise suppression.

[0129] Preferably, step S26 comprises the following steps:

[0130] Step S261: using a bandpass filter to extract effective frequency bands from the near-field vibration wave signal data and the far-field vibration wave signal data, to obtain effective near-field vibration frequency band data and effective far-field vibration frequency band data respectively;

[0131] Step S262: performing frequency modal decomposition on the effective near-field vibration frequency band data and performing independent component analysis to obtain near-field frequency modal component data;

[0132] Step S263: using a preset mechanical noise classifier to identify mechanical noise components of near-field frequency modal component data, and generating mechanical interference source component data;

[0133] Step S264: using the mechanical interference source component data to remove mechanical noise from the near-field frequency modal component data, and reconstructing the near-field vibration signal to obtain noise-reduced near-field vibration signal data;

[0134] Step S265: using the intelligent monitoring layout network to perform signal space propagation attenuation analysis on the near-field vibration wave signal data and the far-field vibration wave signal data to obtain a signal attenuation adjustment factor;

[0135] Step S266: adjusting the interference source component of the mechanical interference source component data by using the signal attenuation adjustment factor, and performing mechanical noise suppression on the effective far-field vibration frequency band data to obtain noise-reduced far-field vibration signal data.

[0136] In an embodiment of the present invention, according to the spectral characteristics of the coal mine roof vibration signal and common mechanical noise, appropriate bandpass filter parameters are set, the near-field and far-field vibration wave signal data are filtered, and the effective frequency band containing the main information components is extracted. For example, based on experience and historical data analysis, it is determined that the main energy of the roof vibration signal is concentrated in the range of 10Hz-500Hz, while the mechanical noise generated by equipment such as coal mining machines and conveyor belts is mainly concentrated in the range of 50Hz-200Hz. For the near-field vibration wave signal data, the passband range of the bandpass filter is set to 10Hz-500Hz, and the signal components below 10Hz and above 500Hz are filtered out to obtain effective near-field vibration frequency band data. For the far-field vibration wave signal data, the passband range of the bandpass filter is set to 10Hz-200Hz, and the signal components below 10Hz and above 200Hz are filtered out to obtain effective far-field vibration frequency band data. The empirical mode decomposition (EMD) method is used to decompose the effective near-field vibration frequency band data and decompose it into several inherent mode functions (IMFs) with different frequency characteristics. Each IMF component represents the local fluctuation characteristics of different scales in the original signal. Independent component analysis (ICA) is performed on the decomposed IMF components to separate the statistically independent signal components and obtain the near-field frequency modal component data. The near-field frequency modal component data is classified and identified using a pre-trained mechanical noise classifier. The classifier can be trained with a large amount of sample data based on machine learning algorithms such as support vector machines (SVM) and neural networks to learn the characteristics of different types of mechanical noise. According to the classification results, the IMF components identified as mechanical noise are marked as mechanical interference source component data. The mechanical interference source component data is removed from the near-field frequency modal component data, and only the IMF components representing the real top plate vibration information are retained. The remaining IMF components are linearly superimposed to reconstruct the noise-reduced near-field vibration signal data. The spatial position information of multiple sensors in the intelligent monitoring layout network is used to perform signal spatial propagation attenuation analysis on the near-field vibration wave signal data and the far-field vibration wave signal data. The mechanical interference source component data is adjusted according to the signal attenuation adjustment factor to match the energy attenuation degree of the far-field signal.For example, the three-dimensional coordinate information of each sensor is obtained from the intelligent monitoring layout network, and a near-field sensor is selected as a reference sensor. The spatial distance between the near-field sensor and the far-field sensor can be calculated using the Euclidean distance formula using the spatial coordinate information of each sensor. For each pair of near-field-far-field sensor combinations, the power spectrum density can be estimated using the Welch method, and then the ratio of the cross-power spectrum density and the self-power spectrum density of the two signals is calculated. A suitable signal propagation attenuation model, such as the spherical diffusion model and the cylindrical diffusion model, is selected for spatial attenuation modeling. According to the selected attenuation model and sensor spacing, the theoretical attenuation factor of the reference sensor signal propagating to each other sensor is calculated, and the energy ratio of the near-field and far-field signals received by the reference sensor and each other sensor is compared to calculate the actual attenuation factor. The actual attenuation factor is compared with the theoretical attenuation factor to obtain a correction coefficient for adjusting the attenuation degree of the mechanical noise component, and the attenuation adjustment factors of all sensor combinations are averaged to obtain a frequency-dependent attenuation adjustment factor. The signal attenuation adjustment factor is used to adjust the energy of the mechanical interference source component data. For each frequency component, the mechanical interference source component data is divided by the attenuation adjustment factor to obtain the adjusted mechanical interference source component data so that its energy attenuation degree matches the actual far-field signal. The adjusted mechanical interference source component data is used to perform adaptive filtering on the effective far-field vibration frequency band data to suppress the mechanical noise components in the far-field signal. For example, the adjusted mechanical interference source component data is used as a reference signal, and the effective far-field vibration frequency band data is used as an input signal. The mechanical noise in the far-field signal is filtered out using an adaptive filtering algorithm (such as an LMS algorithm) to obtain noise-reduced far-field vibration signal data.

[0137] Preferably, step S3 comprises the following steps:

[0138] Step S31: performing roof vibration trend processing according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively;

[0139] Step S32: performing trend energy characteristic analysis on the roof activity change trend data to generate roof activity energy characteristic data;

[0140] Step S33: dynamically correlating the roof impact trend data with the impact waveform through the roof activity energy characteristic data to generate correlated impact fusion data;

[0141] Step S34: deconstructing the main frequency mode according to the associated impact fusion data to obtain impact frequency mode data;

[0142] Step S35: performing multi-scale shock entropy calculation according to the associated shock fusion data to generate multi-scale shock entropy data;

[0143] Step S36: Calculate the entropy distribution density of the multi-scale impact entropy data, and perform entropy increment analysis to generate entropy increment feature data;

[0144] Step S37: evaluating the damage index of the roof impact trend data by using the entropy increment characteristic data to obtain damage index data;

[0145] Step S38: Based on the preset rupture mode library, the rupture mode of the damage degree index data is determined by the impact frequency mode data to obtain the roof rupture mode data.

[0146] In an embodiment of the present invention, a time-frequency analysis method such as wavelet transform or Hilbert-Huang transform (HHT) is used to convert the noise reduction vibration wave signal data from the time domain to the time-frequency domain to obtain the energy distribution of the signal in different time scales and frequency ranges. The change trend of the signal energy over time is analyzed to identify the active period and relatively quiet period of the roof activity, and generate the roof activity change trend data. A window function (such as a Hamming window) and a window length can be preset to perform frame processing on the roof activity change trend data, and then Fourier transform each frame of data to obtain the spectrum corresponding to the frame. By sliding the window, the time-frequency representation of the entire signal can be obtained. The energy distribution at each moment is calculated in the time-frequency domain. The spectrum amplitude at each moment can be squared to obtain the energy spectrum at that moment. The energy spectra at all moments are spliced ​​together to form a two-dimensional matrix, which is the energy distribution diagram of the roof activity change trend data. The marginal distribution of the energy distribution diagram can be calculated, that is, the energy spectrum at each moment is summed to obtain the curve of the total energy changing with time; and the energy corresponding to each frequency is summed to obtain the curve of the total energy changing with frequency, and the peak value of this curve is detected to find the moment of energy mutation as the dangerous period of roof impact. The curve can also be segmented and counted, such as calculating the average energy of each minute to obtain a step curve of energy change, reflecting the long-term trend of roof activity energy. Some statistical features such as mean, variance, skewness, kurtosis, etc. are extracted from the marginal distribution curve as characteristic parameters of roof activity energy; the peak point of the marginal distribution curve can also be found, and the time, frequency and amplitude corresponding to the peak point are extracted as the main mode of roof activity. The extracted characteristic parameters and the main mode as well as the dangerous period and long-term trend are combined into a characteristic vector as the representation of roof activity energy characteristic data. The active period of roof activity identified in the energy characteristic analysis is associated with the occurrence time of the impact signal, and the impact signal closely related to the roof activity is screened out to generate associated impact fusion data. Fast Fourier transform (FFT) and other methods are used to convert the associated shock fusion data from the time domain to the frequency domain, analyze the spectral structure of the signal, identify the main frequency components and their corresponding energy values, and generate shock frequency pattern data. The multiscale entropy algorithm can analyze the complexity of the signal at different time scales, and can better reflect the dynamic change characteristics of the signal than traditional methods such as sample entropy. Calculate the sample entropy values ​​at different scales to generate multiscale shock entropy data. Use kernel density estimation and other methods to smooth the multiscale shock entropy data to obtain the probability density function of the entropy value, and analyze its peak position, peak width, skewness and other characteristics. Calculate the difference between the entropy values ​​at adjacent scales as the entropy increment, and analyze the change trend of the entropy increment. For example, the increase in the entropy value at each scale relative to the entropy value at the previous scale can be calculated, or the growth rate of the entropy value at each scale can be calculated.A mapping relationship between the entropy increment characteristic data and the degree of roof damage is established. For example, linear regression, neural network and other methods can be used for modeling. Using the established model, the degree of roof damage is predicted based on the entropy increment characteristic data, and quantified as a damage degree index. Based on the pre-established rupture pattern library, the impact frequency pattern data and damage degree index data are used to identify the rupture pattern to obtain the roof rupture pattern data. The rupture pattern library contains typical impact frequency patterns and damage degree index ranges corresponding to different types of roof ruptures. For example, tensile failure usually corresponds to a lower frequency component and a smaller damage degree index, while shear failure corresponds to a higher frequency component and a larger damage degree index. Based on the impact frequency pattern data and the damage degree index data, the system searches for the best matching rupture pattern in the rupture pattern library and uses it as the prediction result of the roof rupture pattern.

[0147] Preferably, step S31 includes the following steps:

[0148] Step S311: extracting target signals from the noise reduction far-field vibration signal data using a preset roof activity frequency range to obtain roof activity vibration signal data;

[0149] Step S312: Calculate the low-frequency vibration characteristics of the top plate activity vibration signal data to generate low-frequency vibration characteristic data;

[0150] Step S313: performing time-frequency peak analysis on the roof activity vibration signal data through the low-frequency vibration characteristic data, and extracting the roof activity frequency to obtain the roof activity frequency data;

[0151] Step S314: identifying the frequency change trend according to the roof activity frequency data, and generating roof activity change trend data;

[0152] Step S315: performing signal energy mutation identification on the noise reduction near-field vibration signal data to obtain potential impact event point data;

[0153] Step S316: performing transient waveform interception on the noise reduction near-field vibration signal data through the potential impact event point data to generate transient waveform segment data;

[0154] Step S317: calculating the impact intensity of the transient waveform segment data and performing impact event recognition to obtain impact event recognition data;

[0155] Step S318: extracting shock vibration features according to the shock event identification data, and classifying shock events to generate shock event classification data;

[0156] Step S319: Perform impact intensity time series analysis on the impact event classification data to obtain roof impact trend data.

[0157] In an embodiment of the present invention, according to a preset roof activity frequency range, a bandpass filter is used to filter the noise reduction far-field vibration signal data to extract the target signal. For example, based on experience and historical data analysis, it is determined that the main frequency range of roof activity is 2Hz-20Hz. The passband range of the bandpass filter is set to 2Hz-20Hz, and the signal components below 2Hz and above 20Hz are filtered out to obtain the roof activity vibration signal data. The low-frequency vibration characteristics of the roof activity vibration signal data are calculated to extract the characteristic parameters that can reflect the roof activity state. For example, the energy, main frequency, frequency bandwidth, kurtosis, kurtosis and other characteristic parameters of the signal can be calculated to generate low-frequency vibration characteristic data. The time-frequency analysis method such as short-time Fourier transform (STFT) or wavelet transform is used to perform time-frequency analysis on the roof activity vibration signal data to obtain the energy distribution of the signal in different time periods and frequency ranges. The peak position in the time-frequency spectrum is analyzed, the main frequency components of the roof activity are extracted, and the roof activity frequency data is generated. According to the roof activity frequency data, the frequency change trend over time is identified to generate the roof activity change trend data. For example, the changing trend of the main frequency, the changing trend of the frequency bandwidth, the frequency of the frequency peak, etc. can be analyzed to determine the activity and trend of the roof activity. The signal energy of the noise reduction near-field vibration signal data is calculated, and the energy mutation threshold is set to identify potential impact event points. For example, the signal energy of the noise reduction near-field vibration signal data can be calculated every 0.1 seconds in a sliding window manner. The difference between the signal energy values ​​of two adjacent time points is compared. If the difference exceeds the preset threshold (for example, the threshold is set to 5 times the average value of the signal energy in the previous 1 minute), it is considered that a signal energy mutation has occurred, and the moment is recorded as the potential impact event point data. With each potential impact event point as the center, 0.2 seconds of signal data are intercepted forward and 0.8 seconds of signal data are intercepted backward to form a transient waveform segment data with a length of 1 second. For each transient waveform segment data, its peak amplitude, that is, the maximum value of the absolute value of the signal, is calculated as a measure of the impact intensity. According to the impact intensity calculation results, combined with the preset impact event classification criteria, the impact events are classified to generate impact event classification data. For example, according to the different levels of impact intensity, impact events can be divided into weak impact, slight impact, medium impact and strong impact. By observing the changing trend of these statistics over time, the activity and intensity change pattern of roof impact can be analyzed. For example, if the number of medium-intensity and strong impact events increases per minute, and the average impact intensity shows an upward trend, it indicates that the impact damage of the roof is increasing.

[0158] Preferably, step S4 comprises the following steps:

[0159] Step S41: performing three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face;

[0160] Step S42: Acquire historical mining face three-dimensional point cloud data; perform time series comparison and analysis on the mining face three-dimensional point cloud data using the historical mining face three-dimensional point cloud data to obtain motion deformation difference data;

[0161] Step S43: extracting motion feature parameters from the motion deformation difference data to generate deformation motion feature data;

[0162] Step S44: Based on the roof rupture mode data, the working condition simulation of the coal wall spalling and the mining face roof is performed through the deformation movement characteristic data to generate the roof activity working condition data;

[0163] Step S45: performing support condition response analysis according to the top plate activity condition data to generate support response characteristic data;

[0164] Step S46: using a preset long short-term memory network model to perform transfer learning on the roof activity condition data and the support response characteristic data to obtain a roof risk prediction model;

[0165] Step S47: Use the roof risk prediction model to predict the risk of roof falling and obtain roof falling risk prediction data.

[0166] As an example of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation steps of step S4 in the embodiment are shown in the flowchart. In this embodiment, step S4 includes:

[0167] Step S41: performing three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face;

[0168] In the embodiment of the present invention, a reasonable scanning path is planned according to the shape and size of the working surface to ensure that all key areas are covered, and a suitable scanning density is set to obtain sufficient spatial detail information. According to the planned path, the three-dimensional laser scanner is controlled to scan the working surface line by line, and the spatial coordinate information of each laser point and additional information such as reflection intensity are recorded. The collected laser point data is imported into the point cloud processing software, and pre-processing operations such as splicing, denoising, and filtering are performed to generate three-dimensional point cloud data of the mining surface.

[0169] Step S42: Acquire historical mining face three-dimensional point cloud data; perform time series comparison and analysis on the mining face three-dimensional point cloud data using the historical mining face three-dimensional point cloud data to obtain motion deformation difference data;

[0170] In an embodiment of the present invention, the three-dimensional point cloud data of the mining surface at a historical moment is obtained, and the data adjacent to the current time point or with a short time interval is selected as the comparison object. Since there is a difference in the coordinate system of the point cloud data obtained by the two scans, it is necessary to first perform point cloud registration to unify the two sets of point cloud data into the same coordinate system. A feature-based registration method, such as the ICP (Iterative Closest Point) algorithm, can be used to iteratively calculate the minimum distance between the corresponding point pairs in the two sets of point clouds, continuously optimize the transformation matrix, and finally achieve accurate matching of the two sets of point clouds. After completing the point cloud registration, the difference information such as the three-dimensional distance and normal vector angle between the corresponding points can be calculated to generate motion deformation difference data. For example, the displacement vector, distance change value, normal vector change angle, etc. of each point between the two scans can be calculated and stored as the attribute information of the point cloud.

[0171] Step S43: extracting motion feature parameters from the motion deformation difference data to generate deformation motion feature data;

[0172] In an embodiment of the present invention, statistics such as the average displacement, maximum displacement, standard deviation of displacement, and the distribution of displacement directions of all points are calculated to determine the overall movement trend and deformation degree of the working surface. For example, if the average displacement is large and mainly concentrated in the vertical direction, it indicates that the working surface has an obvious settlement trend. If the standard deviation of the displacement is large, it indicates that the deformation degree of each area of ​​the working surface is uneven and there are local deformation areas. The working surface is divided into several areas. For example, it can be divided according to the support structure on the working surface, and the deformation characteristic parameters of each area are calculated separately to identify areas with more severe deformation. For example, the average displacement, maximum displacement, displacement gradient and other parameters of the points in each area can be calculated. Different color mapping relationships can be set according to the size of the displacement, and the deformation information can be visualized on the point cloud model to intuitively display the deformation distribution of the working surface.

[0173] Step S44: Based on the roof rupture mode data, the working condition simulation of the coal wall spalling and the mining face roof is performed through the deformation movement characteristic data to generate the roof activity working condition data;

[0174] In an embodiment of the present invention, a three-dimensional numerical model is established according to the actual situation of the mining working face, including the stratum structure, rock mass parameters, support system, mining boundary conditions, etc. The roof rupture mode data (for example, rupture type, rupture range, rupture degree, etc.) is imported into the numerical model to simulate the mechanical behavior and deformation characteristics of the roof under different rupture modes. The deformation motion characteristic data (for example, displacement, strain, stress, etc.) are applied to the numerical model as boundary conditions to simulate the dynamic change process of the roof during the mining process. The numerical simulation program is run to calculate the stress distribution, deformation characteristics, rupture development trend, etc. of the roof under different working conditions (for example, different mining speeds, different support strengths, etc.), and generate roof activity working condition data.

[0175] Step S45: performing support condition response analysis according to the top plate activity condition data to generate support response characteristic data;

[0176] In an embodiment of the present invention, information such as stress and displacement at the contact position with the bracket is extracted from the top plate activity working condition data to calculate the stress state of the bracket, such as the axial force, bending moment, shear force, etc. of the bracket. The deformation characteristics of the bracket are calculated based on the material properties and structural parameters of the bracket, such as the compression, curvature, inclination angle, etc. of the bracket. The stress and deformation characteristics of the bracket are compared with the bearing capacity and stability index of the bracket to evaluate the stability and reliability of the bracket system. Key parameters that can reflect the response characteristics of the bracket system are extracted, such as the maximum axial force of the bracket, the maximum deformation, the overall stability coefficient of the bracket system, etc., to generate bracket response characteristic data.

[0177] Step S46: using a preset long short-term memory network model to perform transfer learning on the roof activity condition data and the support response characteristic data to obtain a roof risk prediction model;

[0178] In the embodiment of the present invention, the roof activity condition data and the support response characteristic data are normalized, dimensionally reduced and other preprocessing operations are performed to eliminate the influence of data dimension and redundant information. The parameters of the pre-trained LSTM model are used as initial parameters, and the model is fine-tuned using the historical monitoring data of the current working face so that the model can adapt to the geological conditions and mining environment of the current working face. The pre-processed roof activity condition data and the support response characteristic data are input into the LSTM model, and the roof fall events recorded in the historical monitoring data are used as labels to train the model, optimize the model parameters, and improve the prediction accuracy of the model.

[0179] Step S47: Use the roof risk prediction model to predict the risk of roof falling and obtain roof falling risk prediction data.

[0180] In an embodiment of the present invention, the roof activity condition data and support response characteristic data of the current working face are input into the roof risk prediction model. Based on the input data, the model calculates the probability of the roof falling in a future period of time (for example, the next hour, the next day, etc.), and generates roof falling risk prediction data. For example: the model predicts that within the next hour, the probability of the roof falling in the middle area of ​​the coal mining working face is 80%, and the risk level is "high". The model predicts that within the next hour, the probability of the roof falling in the boundary area of ​​the coal mining working face is 10%, and the risk level is "low".

[0181] Preferably, step S5 comprises the following steps:

[0182] Step S51: performing spatial risk probability mapping on the coal mining face of the coal mine through the roof fall risk prediction data to obtain the risk probability distribution data of the coal mining face;

[0183] Step S52: Calculate the weighted stability index based on the risk probability distribution data of the mining face to generate the top stability index of the mining face;

[0184] Step S53: performing comprehensive index summation processing on the top stability index of the mining face to obtain comprehensive assessment data of mining face risk;

[0185] Step S54: Risk levels are divided according to the comprehensive risk assessment data of the mining face, and risk warning processing is performed according to the mining face top stability index to generate coal mine disaster warning data.

[0186] As an example of the present invention, refer to Figure 4 As shown, Figure 1 Detailed implementation steps of step S5 in the flowchart, in this example, step S5 includes:

[0187] Step S51: performing spatial risk probability mapping on the coal mining face of the coal mine through the roof fall risk prediction data to obtain the risk probability distribution data of the coal mining face;

[0188] In an embodiment of the present invention, the coal mining face is divided into a number of regular grid units. For example, the face can be divided into 1 meter × 1 meter grids. According to the roof fall risk prediction data, a corresponding risk probability value is assigned to each grid unit. For example, if the predicted probability of roof fall in a certain grid unit is 60%, the risk probability value of the grid unit is set to 0.6. According to the size of the risk probability value, different color mapping relationships are set. For example, a gradient color from green to red can be used. The lower the risk probability value, the closer the color is to green, and the higher the risk probability value, the closer the color is to red. The risk probability distribution after color mapping is superimposed on the three-dimensional spatial model of the coal mining face to generate the risk probability distribution data of the mining face.

[0189] Step S52: Calculate the weighted stability index based on the risk probability distribution data of the mining face to generate the top stability index of the mining face;

[0190] In the embodiment of the present invention, a weighted stability index is calculated based on the risk probability distribution data of the mining face. For example, the risk probability value of each grid unit can be multiplied by a weight coefficient, and the weighted risk probability values ​​of all grid units are summed to obtain the mining face top stability index. The weight coefficient can be set according to the importance or risk sensitivity of the grid unit. For example, a grid unit close to the working face or a densely populated area can be given a higher weight coefficient. The mining face top stability index can be calculated using the formula: mining face top stability index = .

[0191] Step S53: performing comprehensive index summation processing on the top stability index of the mining face to obtain comprehensive assessment data of mining face risk;

[0192] In an embodiment of the present invention, a statistical analysis is performed on the top stability index of the mining face, for example, the mean value, standard deviation, risk coefficient, etc. are calculated, and combined with actual conditions, a comprehensive assessment is performed on the overall risk level of the coal mining face to generate comprehensive risk assessment data for the mining face.

[0193] Step S54: Risk levels are divided according to the comprehensive risk assessment data of the mining face, and risk warning processing is performed according to the mining face top stability index to generate coal mine disaster warning data.

[0194] In an embodiment of the present invention, the risk level of the coal mining face is divided into different levels such as safety, low risk, medium risk, and high risk according to a pre-set risk assessment standard. For example: Safety: The top stability index of the mining face is greater than 0.8. Low risk: The top stability index of the mining face is between 0.6-0.8. Medium risk: The top stability index of the mining face is between 0.4-0.6. High risk: The top stability index of the mining face is less than 0.4. According to different risk levels, corresponding early warning measures are set, and coal mine disaster early warning data is generated. For example: Safety: No special measures are required. Low risk: Strengthen monitoring and pay close attention to changes in the roof. Medium risk: Take measures such as strengthening support and reducing mining intensity, and prepare for the evacuation of personnel. High risk: Immediately stop operations, evacuate all personnel, and take emergency measures.

[0195] Preferably, the present invention further provides a coal mine disaster risk prediction and assessment system, which executes the coal mine disaster risk prediction and assessment method as described above, and the coal mine disaster risk prediction and assessment system comprises:

[0196] The coal mine monitoring layout module is used to divide the key monitoring areas of the coal mining face of the coal mine and generate the monitoring area data of the coal mining field; intelligent monitoring equipment is deployed according to the monitoring area data of the coal mining field, thereby building an intelligent monitoring layout network;

[0197] The monitoring signal processing module is used to perform adaptive monitoring frequency adjustment according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data;

[0198] The rupture mode recognition module is used to process the roof vibration trend according to the noise reduction vibration wave signal data, and obtain the roof activity change trend data and the roof impact trend data respectively; the destruction degree index of the roof impact trend data is evaluated through the roof activity change trend data, and the rupture mode is identified to obtain the roof rupture mode data;

[0199] The roof fall risk prediction module is used to perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis based on the three-dimensional point cloud data of the mining face to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data;

[0200] The coal mine risk warning module is used to perform weighted stability index processing based on the roof fall risk prediction data to generate the mining face top stability index; perform risk warning processing based on the mining face top stability index to generate coal mine disaster warning data.

[0201] Preferably, a coal mine disaster risk prediction and assessment storage medium stores a computer program, and when the computer program is executed, a coal mine disaster risk prediction and assessment method as described in any one of the above is implemented.

[0202] The beneficial effect of this application is that by dividing the coal mining face into key areas, important areas in the coal mining face are calibrated, such as the intersection of the working face tunnels, the weak area of ​​the support, the top beam area of ​​the working face support, the vicinity of the coal wall fissure, the surrounding rock wall area of ​​the tunnel, and the surface area of ​​the coal rock mass. Next, based on the data of these monitoring areas, intelligent monitoring equipment, including low-frequency and high-frequency vibration sensors, are deployed to ensure coverage of all key areas. When installing the equipment, an automated equipment layout optimization tool is used to adjust the layout density and location of the monitoring points according to the actual mining environment and coal mining progress, thereby building a comprehensive intelligent monitoring layout network. All-round and multi-point monitoring of key monitoring areas is carried out, and the monitoring frequency is adaptively adjusted according to the monitoring data and the preset threshold value to achieve real-time dynamic monitoring of the roof status. For example, when microseismic activity in a certain area is frequent or the roof delamination exceeds the preset threshold, the sampling frequency of the monitoring equipment in the area is automatically increased to capture the roof activity information more finely. Comprehensive analysis of vibration wave signal data, 3D laser scanning data, etc. can capture early signs of roof rock fracture, such as abnormal changes in microseismic energy, accelerated development of roof delamination, etc. The roof risk prediction model can predict the risk of roof collapse in advance, and trigger corresponding warning levels according to the risk level, such as yellow warning, orange warning and red warning, etc. Warning information is sent to relevant personnel in a variety of ways, such as text messages, voice broadcasts, sound and light alarms, etc., effectively improving the prediction and warning capabilities of coal mine disasters and ensuring the safety of miners.

[0203] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0204] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for predicting and assessing coal mine disaster risk, characterized in that: The following steps are involved: Step S1: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data; deploy intelligent monitoring equipment according to the coal mining field monitoring area data, thereby building an intelligent monitoring layout network; Step S2: Adaptively adjust the monitoring frequency according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data; Step S3: Processing the roof vibration trend according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively; evaluating the destruction degree index of the roof impact trend data through the roof activity change trend data, and distinguishing the rupture mode to obtain roof rupture mode data; Step S4: Perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring timestamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis on the three-dimensional point cloud data of the mining face through the roof rupture mode data to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data; Step S5: Perform weighted stability index processing according to the roof fall risk prediction data to generate a mining face top stability index; perform risk warning processing according to the mining face top stability index to generate coal mine disaster warning data; Step S5 includes the following steps: Step S51: performing spatial risk probability mapping on the coal mining face of the coal mine through the roof fall risk prediction data to obtain the risk probability distribution data of the coal mining face; Step S52: Calculate the weighted stability index based on the risk probability distribution data of the mining face to generate the top stability index of the mining face; Step S53: performing comprehensive index summation processing on the top stability index of the mining face to obtain comprehensive assessment data of mining face risk; Step S54: Risk levels are divided according to the comprehensive risk assessment data of the mining face, and risk warning processing is performed according to the mining face top stability index to generate coal mine disaster warning data.

2. The method for predicting and assessing coal mine disaster risk according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Divide the coal mining face into key monitoring areas and generate coal mining field monitoring area data, wherein the coal mining field monitoring area data includes the intersection of the working face tunnels, the weak support area, the top beam area of ​​the working face support, the vicinity of the coal wall cracks, the tunnel surrounding rock wall area and the coal rock surface area; Step S12: Performing monitoring point layout analysis based on the coal mining field monitoring area data to obtain monitoring layout network data; Step S13: classify the multi-source monitoring points of the monitoring layout network data, and mark the intersection of the working face and tunnel, the surrounding rock wall area of ​​the tunnel, and the weak area of ​​the support as far-field low-frequency monitoring point data; Step S14: marking the top beam area of ​​the support of the working face, the surface area of ​​the coal and rock mass, and the vicinity of the coal wall cracks as near-field high-frequency monitoring point data; Step S15: Use low-frequency vibration sensors to deploy far-field monitoring equipment for far-field low-frequency monitoring point data, and use high-frequency vibration sensors to deploy near-field monitoring equipment for near-field high-frequency monitoring point data, thereby building an intelligent monitoring layout network.

3. The method for predicting and assessing coal mine disaster risk according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Initializing the monitoring frequency of the intelligent monitoring layout network through a preset initial sensor monitoring frequency to obtain initial monitoring frequency data; Step S22: Counting the number of people working in the coal mining face in real time to generate real-time operator data; Step S23: performing monitoring importance assessment according to the real-time operator data, and adaptively adjusting the initial monitoring frequency data to generate adaptive monitoring frequency data; Step S24: performing sensor clock synchronization processing on the intelligent monitoring layout network, and collecting multi-source vibration wave signals according to the adaptive monitoring frequency data, to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data respectively; Step S25: performing monitoring signal preprocessing on the original coal mining vibration wave signal data to obtain near-field vibration wave signal data and far-field vibration wave signal data respectively; Step S26: using the intelligent monitoring layout network to perform mechanical noise suppression processing on the near-field vibration wave signal data and the far-field vibration wave signal data to generate noise-reduced vibration wave signal data.

4. The method for predicting and assessing coal mine disaster risk according to claim 3, characterized in that: The noise reduction vibration wave signal data includes noise reduction near-field vibration signal data and noise reduction far-field vibration signal data, and step S26 includes the following steps: Step S261: using a bandpass filter to extract effective frequency bands from the near-field vibration wave signal data and the far-field vibration wave signal data, to obtain effective near-field vibration frequency band data and effective far-field vibration frequency band data respectively; Step S262: performing frequency modal decomposition on the effective near-field vibration frequency band data and performing independent component analysis to obtain near-field frequency modal component data; Step S263: using a preset mechanical noise classifier to identify mechanical noise components of near-field frequency modal component data, and generating mechanical interference source component data; Step S264: using the mechanical interference source component data to remove mechanical noise from the near-field frequency modal component data, and reconstructing the near-field vibration signal to obtain noise-reduced near-field vibration signal data; Step S265: using the intelligent monitoring layout network to perform signal space propagation attenuation analysis on the near-field vibration wave signal data and the far-field vibration wave signal data to obtain a signal attenuation adjustment factor; Step S266: adjusting the interference source component of the mechanical interference source component data by using the signal attenuation adjustment factor, and performing mechanical noise suppression on the effective far-field vibration frequency band data to obtain noise-reduced far-field vibration signal data.

5. The method for predicting and assessing coal mine disaster risk according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing roof vibration trend processing according to the noise reduction vibration wave signal data to obtain roof activity change trend data and roof impact trend data respectively; Step S32: performing trend energy characteristic analysis on the roof activity change trend data to generate roof activity energy characteristic data; Step S33: dynamically correlating the roof impact trend data with the impact waveform through the roof activity energy characteristic data to generate correlated impact fusion data; Step S34: deconstructing the main frequency mode according to the associated impact fusion data to obtain impact frequency mode data; Step S35: performing multi-scale shock entropy calculation according to the associated shock fusion data to generate multi-scale shock entropy data; Step S36: Calculate the entropy distribution density of the multi-scale impact entropy data, and perform entropy increment analysis to generate entropy increment feature data; Step S37: evaluating the damage index of the roof impact trend data by using the entropy increment characteristic data to obtain damage index data; Step S38: Based on the preset rupture mode library, the rupture mode of the damage degree index data is determined by the impact frequency mode data to obtain the roof rupture mode data.

6. The method for predicting and assessing coal mine disaster risk according to claim 5, characterized in that: Step S31 includes the following steps: Step S311: extracting target signals from the noise reduction far-field vibration signal data using a preset roof activity frequency range to obtain roof activity vibration signal data; Step S312: Calculate the low-frequency vibration characteristics of the top plate activity vibration signal data to generate low-frequency vibration characteristic data; Step S313: performing time-frequency peak analysis on the roof activity vibration signal data through the low-frequency vibration characteristic data, and extracting the roof activity frequency to obtain the roof activity frequency data; Step S314: identifying the frequency change trend according to the roof activity frequency data, and generating roof activity change trend data; Step S315: performing signal energy mutation identification on the noise reduction near-field vibration signal data to obtain potential impact event point data; Step S316: performing transient waveform interception on the noise reduction near-field vibration signal data through the potential impact event point data to generate transient waveform segment data; Step S317: calculating the impact intensity of the transient waveform segment data and identifying the impact event to obtain the impact event identification data; Step S318: extracting shock vibration features according to the shock event identification data, and classifying shock events to generate shock event classification data; Step S319: Perform impact intensity time series analysis on the impact event classification data to obtain roof impact trend data.

7. The method for predicting and assessing coal mine disaster risk according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face; Step S42: Acquire historical mining face three-dimensional point cloud data; perform time series comparison and analysis on the mining face three-dimensional point cloud data using the historical mining face three-dimensional point cloud data to obtain motion deformation difference data; Step S43: extracting motion feature parameters from the motion deformation difference data to generate deformation motion feature data; Step S44: Based on the roof rupture mode data, the working condition simulation of the coal wall spalling and the mining face roof is performed through the deformation movement characteristic data to generate the roof activity working condition data; Step S45: performing support condition response analysis according to the top plate activity condition data to generate support response characteristic data; Step S46: using a preset long short-term memory network model to perform transfer learning on the roof activity condition data and the support response characteristic data to obtain a roof risk prediction model; Step S47: Use the roof risk prediction model to predict the risk of roof falling and obtain roof falling risk prediction data.

8. A coal mine disaster risk prediction and assessment system, characterized in that: Used to execute the coal mine disaster risk prediction and assessment method as claimed in any one of claims 1 to 7, the coal mine disaster risk prediction and assessment system comprises: The coal mine monitoring layout module is used to divide the key monitoring areas of the coal mining face of the coal mine and generate the monitoring area data of the coal mining field; intelligent monitoring equipment is deployed according to the monitoring area data of the coal mining field, thereby building an intelligent monitoring layout network; The monitoring signal processing module is used to perform adaptive monitoring frequency adjustment according to the intelligent monitoring layout network, and collect multi-source vibration wave signals to obtain original coal mining vibration wave signal data and coal mine monitoring timestamp data; perform mechanical noise suppression processing on the original coal mining vibration wave signal data to generate noise-reduced vibration wave signal data; The rupture mode recognition module is used to process the roof vibration trend according to the noise reduction vibration wave signal data, and obtain the roof activity change trend data and the roof impact trend data respectively; the destruction degree index of the roof impact trend data is evaluated through the roof activity change trend data, and the rupture mode is identified to obtain the roof rupture mode data; The roof fall risk prediction module is used to perform three-dimensional laser scanning processing on the coal mining face of the coal mine through the coal mine monitoring time stamp data to generate three-dimensional point cloud data of the mining face; perform working condition activity analysis based on the three-dimensional point cloud data of the mining face to generate roof activity working condition data; perform roof fall risk prediction on the roof activity working condition data to obtain roof fall risk prediction data; The coal mine risk warning module is used to perform weighted stability index processing based on the roof fall risk prediction data to generate the mining face top stability index; perform risk warning processing based on the mining face top stability index to generate coal mine disaster warning data.

9. A storage medium for predicting and evaluating coal mine disaster risk, storing a computer program, characterized in that: When the computer program is executed, a method for predicting and evaluating coal mine disaster risks as described in any one of claims 1 to 7 is implemented.

Images