Intelligent identification and early warning method for hydraulic support instability state under dynamic load

Abstract

This invention relates to the field of safety monitoring and disaster early warning technology for fully mechanized coal mining faces, specifically to an intelligent identification and early warning method for the instability state of hydraulic supports under dynamic loads. The method includes the following steps: S1, collecting and preprocessing monitoring data of the support, rib spalling, and microseismic activity; S2, statistically analyzing resistance, valve opening rate, dynamic load coefficient, rib spalling depth, and microseismic characteristics according to time windows and constructing trend terms; S3, calculating the z-score based on the resistance rolling baseline to identify the pressure range; S4, calculating the rolling quantile threshold and screening candidate risk windows; S5, calculating the risk probability and classifying and issuing early warnings; S6, outputting the risk results and generating risk probability, event statistics, and coupling relationship diagrams. This invention achieves accurate identification and early warning of hydraulic support instability risks, while simultaneously improving on-site risk assessment and safety decision-making capabilities through visualized result output.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring and disaster early warning technology for fully mechanized coal mining faces, and in particular to an intelligent identification and early warning method for the instability state of hydraulic supports under dynamic loads. Background Technology

[0002] As fully mechanized coal mining operations develop towards higher mining heights and stronger mining dynamics, the phenomena of periodic roof pressure and dynamic load impacts are becoming increasingly prominent. Under the influence of roof fracture, surrounding rock instability, and strong mine pressure, the stress state of hydraulic supports fluctuates significantly, often manifesting as a sudden increase in support resistance, frequent valve opening, intensified coal wall spalling, and concentrated release of micro-vibration energy. As a crucial support equipment for roof control in fully mechanized mining faces, the stability of hydraulic supports directly affects the safe production of the working face. Therefore, real-time monitoring of the load state of hydraulic supports and timely identification of potential instability risks are essential technical means to ensure the safe operation of fully mechanized mining faces.

[0003] In existing technologies, the monitoring of the support status in fully mechanized mining faces largely relies on single monitoring indicators or fixed threshold alarm methods, such as simple judgments based on support resistance or valve opening status. However, in actual production processes, the stress state of the supports is affected by various factors such as mining intensity, roof structure, advance speed, and surrounding rock damage, exhibiting significant dynamic changes. A single indicator cannot comprehensively reflect the formation process of support instability; at the same time, fixed threshold methods are difficult to adapt to the changing working conditions of different working faces and different advance stages, easily leading to false alarms or missed alarms. Furthermore, phenomena such as abnormal support loading, coal wall spalling, and micro-vibration energy release often have obvious coupling relationships, but existing methods lack comprehensive utilization of multi-source monitoring information, making it difficult to achieve early identification and graded warning of instability risks. Summary of the Invention

[0004] This invention provides an intelligent identification and early warning method for the instability state of hydraulic supports under dynamic loads. By collecting support monitoring data, sidewall monitoring data, and microseismic monitoring data, multi-dimensional features are constructed based on time window statistics. Rolling baseline identification is used to identify pressure intervals, and rolling quantile thresholds are introduced to achieve adaptive threshold determination. Based on this, the risk probability is calculated and stable, critical, and unstable are classified. At the same time, the early warning result for the next time window is generated, thereby realizing intelligent identification and early warning of hydraulic support instability risk.

[0005] A method for intelligent identification and early warning of hydraulic support instability under dynamic load includes the following steps: S1. Collect and preprocess on-site data from the fully mechanized mining face. The on-site data includes support monitoring data, rib spall monitoring data, and microseismic monitoring data. The support monitoring data includes support timestamp, support number, valve opening mark, advance mileage, and support resistance. The rib spall monitoring data includes rib spall timestamp and rib spall depth. The microseismic monitoring data includes event time and event energy. S2, based on the set time window, perform time window alignment statistics on support monitoring data, sidewall monitoring data and microseismic monitoring data, calculate the mean resistance of the window, the maximum resistance of the window, the valve opening rate and the dynamic load coefficient respectively, and count the maximum sidewall depth, the number of microseismic events, the sum of microseismic energy, the maximum value of microseismic energy and the count of high-energy events within the window, and construct the trend terms of relevant features. S3, construct the rolling baseline mean and standard deviation based on the window mean resistance and calculate the z-score. When the z-score reaches the set threshold, it is marked as a candidate window for incoming pressure. A minimum continuous length constraint is applied to continuous candidate windows for incoming pressure in order to identify the incoming pressure interval. S4. Rolling quantile thresholds are calculated for dynamic load coefficient, valve opening rate and maximum spalling depth as adaptive thresholds. Candidate risk windows are determined by rule screening based on dynamic load coefficient, valve opening rate and maximum spalling depth and micro-seismic high-energy event conditions within the pressure range. S5 calculates the risk probability for candidate risk windows and classifies them into stable, critical, and unstable states based on the risk probability and preset threshold. When a critical or unstable state is reached, the current alarm is output, and the risk probability of the next time window is calculated based on trend characteristics to output the pre-alarm result. S6. Output the risk result file and visualize it. The risk result file includes the characteristic parameters, adaptive threshold, risk probability, risk status and alarm results corresponding to each time window, and generates a risk probability time series diagram, pressure event statistics diagram, dynamic load coefficient and valve opening rate relationship diagram and panel depth and micro-seismic energy coupling diagram.

[0006] Optionally, S1 includes: S11, Deploy a support monitoring system on each hydraulic support of the fully mechanized mining face to collect support monitoring data in real time. The support monitoring data includes support timestamp, support number, valve opening mark, advance mileage and support resistance, which are used to characterize the stress state of each hydraulic support and the valve group opening status at different times. S12, A spalling monitoring device is installed in the coal wall area of ​​the fully mechanized mining face to collect spalling monitoring data in real time. The spalling monitoring data includes spalling timestamp and spalling depth, which are used to characterize the degree of coal wall spalling. S13, A microseismic monitoring system is deployed in the longwall mining face and surrounding rock area to record microseismic event information in real time and form microseismic monitoring data. The microseismic monitoring data includes event time and event energy, which are used to characterize the surrounding rock fracture and energy release. S14, When the support resistance is missing from the support monitoring data, the support resistance is calculated using the pressure of the left and right columns, combined with the effective pressure-bearing area of ​​the columns, and expressed as: ; in, for The resistance of the support at any moment , These are the pressure on the left column and the pressure on the right column, respectively. The effective pressure-bearing area of ​​a single column; S15, the acquired microseismic monitoring data is filtered according to a set energy threshold, retaining only microseismic events with event energy not lower than the microseismic energy filtering threshold, represented as: ; in, For event energy, This is the microseismic energy filtering threshold.

[0007] Optionally, S2 includes: S21, Set the time window width to... And construct the first based on continuous time series A time window is used to align and statistically analyze support monitoring data, sidewall monitoring data, and microseismic monitoring data under a unified time reference. The time window is represented as follows: ; in, For the first A time window interval, For the first The start time of each time window; S22, the monitoring data of the support structure are grouped and statistically analyzed according to time window and support structure number, and the average resistance within the window of a single support structure, the maximum resistance within the window of a single support structure, and the valve opening indicator within the window of a single support structure are calculated and represented as follows: ; in, For stent In the time window The mean resistance inside the single-bracket window. For time windows Internal support The number of effective sampling points, For stent At any moment Resistance; ; in, For stent In the time window The maximum resistance inside a single-bracket window; ; in, For stent In the time window The valve opening indicator inside the single-bracket window. For the frame At any moment Valve open indicator; S23 aggregates the support structures across the entire working face within the same time window, calculating the average resistance, maximum resistance, valve opening rate, and dynamic load coefficient of the working face window, expressed as: ; in, For time windows The average resistance of the working surface window inside, For time windows The number of stents effectively included in the statistics; ; in, For time windows The maximum resistance of the internal working surface window; ; in, For time windows Valve opening rate; ; in, This is the dynamic load factor; S24, the monitoring data of the film is divided into time windows. The maximum flake depth within each time window is calculated by aggregating the data and taking the maximum value. This maximum flake depth within the window is represented as: ; S25, statistical analysis of microseismic monitoring data is performed according to time windows, calculating the number of microseismic events, the sum of microseismic energies, the maximum microseismic energy, and the count of high-energy events, expressed as: ; in, For time windows The number of microseismic events within the area; ; in, For time windows Microseismic energy within; ; in, For time windows The maximum microseismic energy within; The high-energy event counting includes a time window. Internal event energy not less than Event counts and time windows Internal event energy not less than The event count is represented as: ; ; in, For time windows Internal event energy not less than Event count, For time windows Internal event energy not less than Event count; S26. Based on the average resistance, dynamic load coefficient, valve opening rate, micro-seismic energy, and maximum spalling depth of the working face obtained from statistics of each time window, construct trend terms for the corresponding features, including resistance trend, dynamic load coefficient trend, valve opening rate trend, micro-seismic energy trend, and spalling trend. Among them, the resistance trend is calculated by dividing the difference by the time window width, and the dynamic load coefficient trend, valve opening rate trend, micro-seismic energy trend, and spalling trend are calculated by the difference between adjacent time windows.

[0008] Optionally, the resistance trend is represented as: ; in, This is a resistance trend; The trend of the dynamic load factor is expressed as follows: ; in, The trend of dynamic load coefficient; The valve opening rate trend is expressed as follows: ; in, The trend of valve opening rate; The microseismic energy and trend are represented as follows: ; in, For microseismic energy and trends; The trend of the segmentation is represented as follows: ; in, This is a trend towards film blocs.

[0009] Optionally, S3 includes: S31, based on the average resistance of the working surface window in each time window, the length of the scrolling window is set to... , for the The average resistance of the working surface window in historical time windows prior to the first time window is statistically analyzed to obtain the first... The rolling baseline mean and rolling standard deviation for each time window are expressed as follows: ; ; in, The rolling baseline mean. For rolling standard deviation, For the first Mean resistance of the working surface window within a time window; S32, calculate the z-score using the mean resistance of the working face window in the current time window, the mean of the corresponding rolling baseline, and the rolling standard deviation, expressed as: ; in, To suppress z-score; S33, the first The pressure z-score and pressure threshold for each time window When comparing, At that time, mark the time window as a candidate window for pressure; S34, perform continuity detection on the incoming pressure candidate windows obtained in chronological order, form adjacent consecutive incoming pressure candidate windows into candidate incoming pressure segments, and apply a minimum continuity length to each candidate incoming pressure segment. Constraints are applied only to candidates whose pressure segments continuously satisfy the conditions and whose length is not less than [a certain value]. The candidate pressure segments are taken as the effective pressure intervals, thus obtaining the pressure interval marking. ; S35, assign event numbers to each valid pressure interval that satisfies the minimum continuous length constraint in chronological order, and generate corresponding pressure event numbers. .

[0010] Optionally, S4 includes: S41, perform rolling quantile statistics on the dynamic load coefficient, valve opening rate, and maximum sheet depth within the rolling window before each time window to obtain the adaptive threshold for the corresponding time window, expressed as: ; ; ; in, For time windows The dynamic load coefficient adaptive threshold, For time windows The valve opening rate adaptive threshold, For time windows Maximum edge depth adaptive threshold To Find the quantiles. To Find the quantiles. To Find the quantiles; S42, when the number of historical time windows used to calculate the rolling quantile is insufficient, a preset backoff threshold is applied. , , As alternative thresholds for dynamic load factor, valve opening rate and maximum flaring depth; S43, only within the identified pressure range, performs rule-based judgments on the dynamic load factor, valve opening rate, maximum spalling depth, and micro-seismic high-energy events of the current time window. The judgment is made when the dynamic load factor exceeds the adaptive threshold, the valve opening rate exceeds the adaptive threshold, the maximum spalling depth exceeds the adaptive threshold, or the energy of an event within the time window is not less than [a certain threshold]. When calculating the number of microseismic events, this time window is retained as a candidate risk window; S44, combining the future pressure interval constraint and the rule judgment result, yields the candidate risk window criterion, expressed as: ; in, This is an indicator of the candidate risk window. For logical AND, For logical symbols OR; S45 marks the time window that meets the candidate risk window criterion as a candidate risk window and outputs the candidate risk window indicator.

[0011] Optionally, S5 includes: S51, for the selected candidate risk windows, extract the dynamic load coefficient, valve opening rate, maximum spalling depth, microseismic energy, high-energy event count and trend items corresponding to the current time window; S52, normalize the dynamic load coefficient, valve opening rate, and maximum spalling depth of the current time window according to their respective adaptive thresholds to generate dynamic load over-threshold normalization terms, valve opening over-threshold normalization terms, and spalling over-threshold normalization terms, as follows: ; ; ; in, , , These are the normalized terms for dynamic load over-threshold, valve opening over-threshold, and plate over-threshold, respectively. S53, logarithmically compress the microseismic energy sum of the current time window to form a logarithmic compression term for microseismic energy, and construct a high-energy event weighting term by combining it with high-energy event counting, expressed as: ; ; in, This is the logarithmic compression term of the microseismic energy. Weighting for high-energy events; S54 integrates the trends of dynamic load coefficient, valve opening rate, sidewall, and microseismic energy and trend, and compresses them using a hyperbolic tangent function to obtain the trend fusion term. The microseismic energy and trend are normalized using the microseismic energy and mean over the past 30 time windows, and are expressed as: ; ; in, The data represents the microseismic energy and mean over a period of approximately 30 time windows. For trend fusion items, It is the hyperbolic tangent function; S55 linearly weights the dynamic load overthreshold normalization term, valve opening overthreshold normalization term, spalling overthreshold normalization term, microseismic energy logarithmic compression term, high-energy event weighting term, and trend fusion term to obtain a linear fusion score, which is then mapped to the risk probability of the current time window using the Sigmoid function, expressed as: ; ; in, For linear fusion scores, For risk probability; S56, the risk probability of the current time window. By comparing with stability and instability thresholds, the corresponding risk status is determined, including stable, critical, and unstable. S57, determine the risk status corresponding to the candidate risk window. If the current time window is determined to be critical or unstable, output the current alarm flag. For non-candidate risk windows, force the output to be stable and do not trigger the current alarm. S58, construct a prediction enhancement factor based on the trend fusion term of the current time window, which is used to enhance and correct the risk probability of the next time window. Calculate the risk probability of the next time window based on the risk probability of the current time window and the prediction enhancement factor, and output a pre-alarm flag for the next time window based on whether the risk probability of the next time window reaches the instability threshold, expressed as: ; ; ; in, To predict the enhancement factor, The probability of risk in the next time window. This is a truncation function. Pre-alarm sign for the next window; S59 outputs the risk probability, risk status, current alarm flag, risk probability of the next time window, and pre-alarm flag of the next window for each candidate risk window.

[0012] Optionally, the risk status is represented as: ; in, To stabilize the threshold, This is the instability threshold.

[0013] Optionally, S6 includes: S61, for each time window Construct a risk outcome record, including time window number, average resistance of working face window, maximum resistance of working face window, valve opening rate, dynamic load coefficient, maximum spalling depth, number of micro-seismic events, sum of micro-seismic energy, maximum micro-seismic energy, high-energy event count, adaptive threshold for dynamic load coefficient, adaptive threshold for valve opening rate, adaptive threshold for maximum spalling depth, risk probability, risk status, current alarm flag, and pre-alarm flag for the next window; S62, organize the risk results records in chronological order, and write the characteristic parameters, adaptive thresholds, risk probabilities, risk status and alarm results corresponding to each time window into the risk results file; S63, based on the risk probability, stability threshold and instability threshold in the risk result file, visualize the graph by plotting the curve of risk probability changing with time window on the time axis, and overlaying the stability threshold, instability threshold, current alarm flag and next window pre-alarm flag to generate a risk probability time series graph; S64. Based on the pressure interval marking and the corresponding pressure event number, summarize the statistical characteristics of each pressure event, including the average resistance of the working face window, valve opening rate, dynamic load coefficient, maximum rib depth, number of micro-seismic events, micro-seismic energy and risk probability, and generate a statistical plot of the pressure event using the pressure event number as the unit. S65. Based on the dynamic load factor and valve opening rate in the risk results file, construct a two-dimensional scatter plot. By drawing the scatter plot of the synergistic relationship between the dynamic load factor and valve opening rate, show the coupling relationship between the support impact load and the valve assembly unloading behavior, thus forming a dynamic load factor and valve opening rate relationship diagram. S66, construct a coupling relationship diagram using the maximum spalling depth and microseismic energy from the risk results file, and generate a spalling depth and microseismic energy coupling diagram by plotting the coupling distribution relationship between the maximum spalling depth and microseismic energy. S67 saves the generated risk probability time series diagram, pressure event statistics diagram, dynamic load coefficient and valve opening rate relationship diagram, and spalling depth and microseismic energy coupling diagram as graphic files, and outputs them together with the risk results file.

[0014] The beneficial effects of this invention are: This invention constructs adaptive thresholds for dynamic load coefficient, valve opening rate, and maximum sidewall depth by introducing a rolling quantile threshold based on time window statistics. It also combines a pressure interval identification mechanism to constrain and filter abnormal features, enabling the thresholds to be dynamically updated as the working face advances and working conditions change. Compared with fixed threshold judgment methods, this can effectively reduce false alarms and false alarms, and improve the adaptability and accuracy of hydraulic support instability risk identification.

[0015] This invention integrates support monitoring data, rock spalling monitoring data, and microseismic monitoring data to jointly analyze multi-source information such as dynamic load impact, valve unloading behavior, rock spalling failure, and microseismic energy release. It also constructs characteristics and trend terms such as dynamic load coefficient, valve opening rate, rock spalling depth, and microseismic energy to achieve a multi-dimensional characterization of the dynamic load-induced disaster process in fully mechanized mining faces. This allows for a more comprehensive reflection of the coupling relationship between the support loading state and the surrounding rock fracturing process, improving the reliability of instability state identification.

[0016] This invention calculates the risk probability of candidate risk windows and classifies them into stable, critical, and unstable categories. It also calculates the risk probability of the next time window based on trend characteristics to achieve early warning. Based on the generated risk result file, it outputs a risk probability time series diagram, a pressure event statistics diagram, a dynamic load coefficient and valve opening rate relationship diagram, and a panel depth and microseismic energy coupling diagram. This not only enables early warning of support instability risk, but also visually displays the risk evolution process, facilitating rapid assessment and safety decision-making by on-site dispatchers. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the identification and early warning method according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] like Figure 1 As shown, a method for intelligent identification and early warning of hydraulic support instability under dynamic load includes the following steps: S1 collects and preprocesses on-site data from the fully mechanized mining face. The on-site data includes support monitoring data, rib spall monitoring data, and microseismic monitoring data. Among them, support monitoring data includes support timestamp, support number, valve opening mark, advance mileage, and support resistance; rib spall monitoring data includes rib spall timestamp and rib spall depth; and microseismic monitoring data includes event time and event energy. S2, based on the set time window, perform time window alignment statistics on support monitoring data, sidewall monitoring data and microseismic monitoring data, calculate the mean resistance of the window, the maximum resistance of the window, the valve opening rate and the dynamic load coefficient respectively, and count the maximum sidewall depth, the number of microseismic events, the sum of microseismic energy, the maximum value of microseismic energy and the count of high-energy events within the window, and construct the trend terms of relevant features. S3, construct the rolling baseline mean and standard deviation based on the window mean resistance and calculate the z-score. When the z-score reaches the set threshold, it is marked as a candidate window for incoming pressure. A minimum continuous length constraint is applied to continuous candidate windows for incoming pressure in order to identify the incoming pressure interval. S4. Rolling quantile thresholds are calculated for dynamic load coefficient, valve opening rate and maximum spalling depth as adaptive thresholds. Candidate risk windows are determined by rule screening based on dynamic load coefficient, valve opening rate and maximum spalling depth and micro-seismic high-energy event conditions within the pressure range. S5 calculates the risk probability for candidate risk windows and classifies them into stable, critical, and unstable states based on the risk probability and preset threshold. When a critical or unstable state is reached, the current alarm is output, and the risk probability of the next time window is calculated based on trend characteristics to output the pre-alarm result. S6 outputs a risk result file and displays it visually. The risk result file includes the characteristic parameters, adaptive threshold, risk probability, risk status and alarm results corresponding to each time window, and generates a risk probability time series diagram, pressure event statistics diagram, dynamic load coefficient and valve opening rate relationship diagram, and panel depth and microseismic energy coupling diagram.

[0021] S1 includes: S11, deploy a support monitoring system on each hydraulic support of the fully mechanized mining face to collect support monitoring data in real time. The support monitoring data includes support timestamp, support number, valve opening mark, advance mileage and support resistance, which are used to characterize the stress state of each hydraulic support and the valve group opening status at different times. S12, a spalling monitoring device is installed in the coal wall area of ​​the fully mechanized mining face to collect spalling monitoring data in real time. The spalling monitoring data includes spalling timestamp and spalling depth, which are used to characterize the degree of coal wall spalling. S13, a microseismic monitoring system is deployed in the longwall mining face and surrounding rock area to record microseismic event information in real time and form microseismic monitoring data. The microseismic monitoring data includes event time and event energy, which are used to characterize the surrounding rock fracture and energy release. S14, When the support resistance is missing from the support monitoring data, the support resistance is calculated using the pressure of the left and right columns, combined with the effective pressure-bearing area of ​​the columns, and expressed as: ; in, for The resistance of the support at any moment , These are the pressure on the left column and the pressure on the right column, respectively. The effective pressure-bearing area of ​​a single column; S15, the acquired microseismic monitoring data is filtered according to a set energy threshold, retaining only microseismic events with event energy not lower than the microseismic energy filtering threshold, represented as: ; in, For event energy, This is the microseismic energy filtering threshold.

[0022] S2 includes: S21, Set the time window width to... And construct the first based on continuous time series A time window is used to align and statistically analyze support monitoring data, sidewall monitoring data, and microseismic monitoring data under a unified time reference. The time window is represented as follows: ; in, For the first A time window interval, For the first The start time of each time window; S22, the monitoring data of the support structure are grouped and statistically analyzed according to time window and support structure number, and the average resistance within the window of a single support structure, the maximum resistance within the window of a single support structure, and the valve opening indicator within the window of a single support structure are calculated and represented as follows: ; in, For stent In the time window Mean resistance inside a single-bracket window. For time windows Internal support The number of effective sampling points, For stent At any moment Resistance; ; in, For stent In the time window The maximum resistance inside a single-bracket window; ; in, For stent In the time window The valve opening indicator inside the single-bracket window. For the frame At any moment Valve open indicator; S23 aggregates the support structures across the entire working face within the same time window, calculating the average resistance, maximum resistance, valve opening rate, and dynamic load coefficient of the working face window, expressed as: ; in, For time windows The average resistance of the working surface window inside, For time windows The number of stents effectively included in the statistics; ; in, For time windows The maximum resistance of the internal working surface window; ; in, For time windows Valve opening rate; ; in, This is the dynamic load factor; S24, the monitoring data of the film is divided into time windows. The maximum flake depth within each time window is calculated by aggregating the data and taking the maximum value. This maximum flake depth within the window is represented as: ; S25, statistical analysis of microseismic monitoring data is performed according to time windows, calculating the number of microseismic events, the sum of microseismic energies, the maximum microseismic energy, and the count of high-energy events, expressed as: ; in, For time windows The number of microseismic events within the area; ; in, For time windows Microseismic energy within; ; in, For time windows The maximum microseismic energy within; High-energy event counting includes time windows Internal event energy not less than Event counts and time windows Internal event energy not less than The event count is represented as: ; ; in, For time windows Internal event energy not less than Event count, For time windows Internal event energy not less than Event count; S26. Based on the average resistance, dynamic load coefficient, valve opening rate, micro-seismic energy, and maximum spalling depth of the working face obtained from statistics of each time window, construct trend terms for the corresponding features, including resistance trend, dynamic load coefficient trend, valve opening rate trend, micro-seismic energy trend, and spalling trend. Among them, the resistance trend is calculated by dividing the difference by the time window width, and the dynamic load coefficient trend, valve opening rate trend, micro-seismic energy trend, and spalling trend are calculated by the difference between adjacent time windows.

[0023] Resistance trends are represented as: ; in, This is a resistance trend; The trend of dynamic load factor is expressed as follows: ; in, The trend of dynamic load coefficient; The valve opening rate trend is expressed as follows: ; in, The trend of valve opening rate; Microseismic energy and trends are represented as follows: ; in, For microseismic energy and trends; The trend of film gangs is represented as follows: ; in, This is a trend towards film blocs.

[0024] S3 includes: S31, based on the average resistance of the working surface window in each time window, the length of the scrolling window is set to... , for the The average resistance of the working surface window in historical time windows prior to the first time window is statistically analyzed to obtain the first... The rolling baseline mean and rolling standard deviation for each time window are expressed as follows: ; ; in, The rolling baseline mean. For rolling standard deviation, For the first Mean resistance of the working surface window within a time window; S32 calculates the z-score using the mean resistance of the working face window in the current time window, the mean of the corresponding rolling baseline, and the rolling standard deviation. This z-score measures the degree of deviation of the current resistance level from the historical baseline and is expressed as: ; in, To suppress z-score; S33, the first The pressure z-score and pressure threshold for each time window When comparing, At that time, mark the time window as a candidate window for pressure; S34, perform continuity detection on the incoming pressure candidate windows obtained in chronological order, form adjacent consecutive incoming pressure candidate windows into candidate incoming pressure segments, and apply a minimum continuity length to each candidate incoming pressure segment. Constraints are applied only to candidates whose pressure segments continuously satisfy the conditions and whose length is not less than [a certain value]. The candidate pressure segments are taken as the effective pressure intervals, thus obtaining the pressure interval marking. ; S35, assign event numbers to each valid pressure interval that satisfies the minimum continuous length constraint in chronological order, and generate corresponding pressure event numbers. .

[0025] S4 includes: S41, perform rolling quantile statistics on the dynamic load coefficient, valve opening rate, and maximum sheet depth within the rolling window before each time window to obtain the adaptive threshold for the corresponding time window, expressed as: ; ; ; in, For time windows The dynamic load coefficient adaptive threshold, For time windows The valve opening rate adaptive threshold, For time windows Maximum edge depth adaptive threshold To Find the quantiles. To Find the quantiles. To Find the quantiles; S42, When the number of historical time windows used to calculate the rolling quantile is insufficient, the rolling quantile statistics are no longer used, and the preset backoff threshold is used instead. , , As alternative thresholds for dynamic load factor, valve opening rate and maximum flaring depth; S43, only within the identified pressure range, performs rule-based judgments on the dynamic load factor, valve opening rate, maximum spalling depth, and micro-seismic high-energy events of the current time window. The judgment is made when the dynamic load factor exceeds the adaptive threshold, the valve opening rate exceeds the adaptive threshold, the maximum spalling depth exceeds the adaptive threshold, or the energy of an event within the time window is not less than [a certain threshold]. When calculating the number of microseismic events, this time window is retained as a candidate risk window; S44, combining the future pressure interval constraint and the rule judgment result, yields the candidate risk window criterion, expressed as: ; in, This is an indicator of the candidate risk window. For logical AND, For logical symbols OR; S45 marks the time window that meets the candidate risk window criterion as a candidate risk window and outputs the candidate risk window indicator.

[0026] S5 includes: S51, for the selected candidate risk windows, extract the dynamic load coefficient, valve opening rate, maximum spalling depth, microseismic energy, high-energy event count and trend items corresponding to the current time window; S52, normalize the dynamic load coefficient, valve opening rate, and maximum spalling depth of the current time window according to their respective adaptive thresholds to generate dynamic load over-threshold normalization terms, valve opening over-threshold normalization terms, and spalling over-threshold normalization terms, as follows: ; ; ; in, , , These are the normalized terms for dynamic load over-threshold, valve opening over-threshold, and plate over-threshold, respectively. S53, logarithmically compress the microseismic energy sum of the current time window to form a logarithmic compression term for microseismic energy, and construct a high-energy event weighting term by combining it with high-energy event counting, expressed as: ; ; in, This is the logarithmic compression term of the microseismic energy. Weighting for high-energy events; S54 integrates the trends of dynamic load coefficient, valve opening rate, sidewall, and microseismic energy and trend, and compresses them using a hyperbolic tangent function to obtain the trend fusion term. The microseismic energy and trend are normalized using the microseismic energy and mean over the past 30 time windows, and are expressed as: ; ; in, The data represents the microseismic energy and mean over a period of approximately 30 time windows. For trend fusion items, It is the hyperbolic tangent function; S55 linearly weights the dynamic load overthreshold normalization term, valve opening overthreshold normalization term, spalling overthreshold normalization term, microseismic energy logarithmic compression term, high-energy event weighting term, and trend fusion term to obtain a linear fusion score, which is then mapped to the risk probability of the current time window using the Sigmoid function, expressed as: ; ; in, For linear fusion scores, For risk probability; S56, the risk probability of the current time window. By comparing with stability and instability thresholds, the corresponding risk status is determined, including stable, critical, and unstable. S57, determine the risk status corresponding to the candidate risk window. If the current time window is determined to be critical or unstable, output the current alarm flag. For non-candidate risk windows, force the output to be stable and do not trigger the current alarm. S58, construct a prediction enhancement factor based on the trend fusion term of the current time window, which is used to enhance and correct the risk probability of the next time window. Calculate the risk probability of the next time window based on the risk probability of the current time window and the prediction enhancement factor, and output a pre-alarm flag for the next time window based on whether the risk probability of the next time window reaches the instability threshold, expressed as: ; ; ; in, To predict the enhancement factor, The probability of risk in the next time window. This is a truncation function. Pre-alarm sign for the next window; S59 outputs the risk probability, risk status, current alarm flag, risk probability of the next time window, and pre-alarm flag of the next window for each candidate risk window.

[0027] Risk status is represented as: ; in, To stabilize the threshold, This is the instability threshold.

[0028] S6 includes: S61, for each time window Construct a risk outcome record, including time window number, average resistance of working face window, maximum resistance of working face window, valve opening rate, dynamic load coefficient, maximum spalling depth, number of micro-seismic events, sum of micro-seismic energy, maximum micro-seismic energy, high-energy event count, adaptive threshold for dynamic load coefficient, adaptive threshold for valve opening rate, adaptive threshold for maximum spalling depth, risk probability, risk status, current alarm flag, and pre-alarm flag for the next window; S62, organize the risk results records in chronological order, and write the characteristic parameters, adaptive thresholds, risk probabilities, risk status and alarm results corresponding to each time window into the risk results file; S63, based on the risk probability, stability threshold and instability threshold in the risk result file, visualize the graph by plotting the curve of risk probability changing with time window on the time axis, and overlaying the stability threshold, instability threshold, current alarm flag and next window pre-alarm flag to generate a risk probability time series graph; S64. Based on the pressure interval marking and the corresponding pressure event number, summarize the statistical characteristics of each pressure event, including the average resistance of the working face window, valve opening rate, dynamic load coefficient, maximum rib depth, number of micro-seismic events, micro-seismic energy and risk probability, and generate a statistical plot of the pressure event using the pressure event number as the unit. S65. Based on the dynamic load factor and valve opening rate in the risk results file, construct a two-dimensional scatter plot. By drawing the scatter plot of the synergistic relationship between the dynamic load factor and valve opening rate, show the coupling relationship between the support impact load and the valve assembly unloading behavior, thus forming a dynamic load factor and valve opening rate relationship diagram. S66, construct a coupling relationship diagram using the maximum spalling depth and microseismic energy from the risk results file, and generate a spalling depth and microseismic energy coupling diagram by plotting the coupling distribution relationship between the maximum spalling depth and microseismic energy. S67 saves the generated risk probability time series diagram, pressure event statistics diagram, dynamic load coefficient and valve opening rate relationship diagram, and spalling depth and microseismic energy coupling diagram as graphic files, and outputs them together with the risk results file.

[0029] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0030] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent identification and early warning of instability state of hydraulic supports under dynamic load, characterized in that, Includes the following steps: S1 collects and preprocesses data from the support monitoring, rib spalling monitoring, and microseismic monitoring of the fully mechanized mining face. The support monitoring data includes timestamps, support numbers, valve opening signs, advance mileage, and support resistance. The rib spalling monitoring data includes timestamps and rib spalling depth. The microseismic monitoring data includes event time and event energy. S2, based on the set time window, perform alignment statistics on the above data, calculate the window mean resistance, maximum resistance, valve opening rate and dynamic load coefficient, and count the maximum spalling depth, number of micro-seismic events, energy sum, maximum energy value and high-energy event count, and construct trend terms. S3, calculate the rolling baseline mean and standard deviation based on the window mean resistance and obtain the z-score. When the z-score reaches the threshold, it is marked as a candidate window for pressure, and the pressure interval is identified by the minimum continuous length constraint. S4 calculates the rolling quantile threshold as an adaptive threshold based on the dynamic load coefficient, valve opening rate and maximum sheet depth, and combines micro-seismic high-energy events to screen candidate risk windows within the pressure range. S5 calculates the risk probability for candidate risk windows and classifies them into stable, critical, and unstable states based on thresholds. When the critical or unstable state is reached, the current alarm is output, and the risk probability of the next time window is calculated to achieve pre-alarm. S6 outputs the risk results file and generates a risk probability time series diagram, a pressure event statistics diagram, a dynamic load coefficient and valve opening rate relationship diagram, and a panel depth and microseismic energy coupling diagram.

2. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 1, characterized in that, S1 includes: S11, Deploy a support monitoring system on each hydraulic support of the fully mechanized mining face to collect support monitoring data in real time. The support monitoring data includes support timestamp, support number, valve opening mark, advance mileage and support resistance, which are used to characterize the stress state of each hydraulic support and the valve group opening status at different times. S12, A spalling monitoring device is installed in the coal wall area of ​​the fully mechanized mining face to collect spalling monitoring data in real time. The spalling monitoring data includes spalling timestamp and spalling depth, which are used to characterize the degree of coal wall spalling. S13, A microseismic monitoring system is deployed in the longwall mining face and surrounding rock area to record microseismic event information in real time and form microseismic monitoring data. The microseismic monitoring data includes event time and event energy, which are used to characterize the surrounding rock fracture and energy release. S14, When the support resistance is missing from the support monitoring data, the support resistance is calculated using the pressure of the left and right columns, combined with the effective pressure-bearing area of ​​the columns, and expressed as: ； in, for The resistance of the support at any moment , These are the pressure on the left column and the pressure on the right column, respectively. The effective pressure-bearing area of ​​a single column; S15, the acquired microseismic monitoring data is filtered according to a set energy threshold, retaining only microseismic events with event energy not lower than the microseismic energy filtering threshold, represented as: ； in, For event energy, This is the microseismic energy filtering threshold.

3. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 2, characterized in that, S2 includes: S21, Set the time window width to... And construct the first based on continuous time series A time window is used to align and statistically analyze support monitoring data, sidewall monitoring data, and microseismic monitoring data under a unified time reference. The time window is represented as follows: ； in, For the first A time window interval, For the first The start time of each time window; S22, the monitoring data of the support structure are grouped and statistically analyzed according to time window and support structure number, and the average resistance within the window of a single support structure, the maximum resistance within the window of a single support structure, and the valve opening indicator within the window of a single support structure are calculated and represented as follows: ； in, For stent In the time window Mean resistance inside a single-bracket window. For time windows Internal support The number of effective sampling points, For stent At any moment Resistance; ； in, For stent In the time window The maximum resistance inside a single-bracket window; ； in, For stent In the time window The valve opening indicator inside the single-bracket window. For the frame At any moment Valve open indicator; S23 aggregates the support structures across the entire working face within the same time window, calculating the average resistance, maximum resistance, valve opening rate, and dynamic load coefficient of the working face window, expressed as: ； in, For time windows The average resistance of the working surface window inside, For time windows The number of stents effectively included in the statistics; ； in, For time windows The maximum resistance of the internal working surface window; ； in, For time windows Valve opening rate; ； in, This is the dynamic load factor; S24, the monitoring data of the film is divided into time windows. The maximum flake depth within each time window is calculated by aggregating the data and taking the maximum value. This maximum flake depth within the window is represented as: ； S25, statistical analysis of microseismic monitoring data is performed according to time windows, calculating the number of microseismic events, the sum of microseismic energies, the maximum microseismic energy, and the count of high-energy events, expressed as: ； in, For time windows The number of microseismic events within the area; ； in, For time windows Microseismic energy within; ； in, For time windows The maximum microseismic energy within; The high-energy event counting includes a time window. Internal event energy not less than Event counts and time windows Internal event energy not less than The event count is represented as: ； ； in, For time windows Internal event energy not less than Event count, For time windows Internal event energy not less than Event count; S26. Based on the average resistance, dynamic load coefficient, valve opening rate, micro-seismic energy, and maximum spalling depth of the working face obtained from statistics of each time window, construct trend terms for the corresponding features, including resistance trend, dynamic load coefficient trend, valve opening rate trend, micro-seismic energy trend, and spalling trend. Among them, the resistance trend is calculated by dividing the difference by the time window width, and the dynamic load coefficient trend, valve opening rate trend, micro-seismic energy trend, and spalling trend are calculated by the difference between adjacent time windows.

4. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 3, characterized in that, The resistance trend is represented as follows: ； in, This is a resistance trend; The trend of the dynamic load factor is expressed as follows: ； in, The trend of dynamic load coefficient; The valve opening rate trend is expressed as follows: ； in, The trend of valve opening rate; The microseismic energy and trend are represented as follows: ； in, For microseismic energy and trends; The trend of the segmentation is represented as follows: ； in, This is a trend towards film blocs.

5. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 4, characterized in that, S3 includes: S31, based on the average resistance of the working surface window in each time window, the length of the scrolling window is set to... , for the The average resistance of the working surface window in historical time windows prior to the first time window is statistically analyzed to obtain the first... The rolling baseline mean and rolling standard deviation for each time window are expressed as follows: ； ； in, The rolling baseline mean. For rolling standard deviation, For the first Mean resistance of the working surface window within a time window; S32, calculate the z-score using the mean resistance of the working face window in the current time window, the mean of the corresponding rolling baseline, and the rolling standard deviation, expressed as: ； in, To suppress z-score; S33, the first The pressure z-score and pressure threshold for each time window When comparing, At that time, mark the time window as a candidate window for pressure; S34, perform continuity detection on the incoming pressure candidate windows obtained in chronological order, form adjacent consecutive incoming pressure candidate windows into candidate incoming pressure segments, and apply a minimum continuity length to each candidate incoming pressure segment. Constraints are applied only to candidates whose pressure segments continuously satisfy the conditions and whose length is not less than [a certain value]. The candidate pressure segments are taken as the effective pressure intervals, thus obtaining the pressure interval marking. ; S35, assign event numbers to each valid pressure interval that satisfies the minimum continuous length constraint in chronological order, and generate corresponding pressure event numbers. .

6. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 5, characterized in that, S4 includes: S41, perform rolling quantile statistics on the dynamic load coefficient, valve opening rate, and maximum sheet depth within the rolling window before each time window to obtain the adaptive threshold for the corresponding time window, expressed as: ； ； ； in, For time windows The dynamic load coefficient adaptive threshold, For time windows The valve opening rate adaptive threshold, For time windows Maximum edge depth adaptive threshold, To Find the quantiles. To Find the quantiles. To Find the quantiles; S42, when the number of historical time windows used to calculate the rolling quantile is insufficient, a preset backoff threshold is applied. , , As alternative thresholds for dynamic load factor, valve opening rate and maximum flaring depth; S43, only within the identified pressure range, performs rule-based judgments on the dynamic load factor, valve opening rate, maximum spalling depth, and micro-seismic high-energy events of the current time window. The judgment is made when the dynamic load factor exceeds the adaptive threshold, the valve opening rate exceeds the adaptive threshold, the maximum spalling depth exceeds the adaptive threshold, or the energy of an event within the time window is not less than [a certain threshold]. When calculating the number of microseismic events, this time window is retained as a candidate risk window; S44, combining the future pressure interval constraint and the rule judgment result, yields the candidate risk window criterion, expressed as: ； in, This is an indicator of the candidate risk window. For logical AND, For logical symbols OR; S45 marks the time window that meets the candidate risk window criterion as a candidate risk window and outputs the candidate risk window indicator.

7. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 6, characterized in that, S5 includes: S51, for the selected candidate risk windows, extract the dynamic load coefficient, valve opening rate, maximum spalling depth, microseismic energy, high-energy event count and trend items corresponding to the current time window; S52, normalize the dynamic load coefficient, valve opening rate, and maximum spalling depth of the current time window according to their respective adaptive thresholds to generate dynamic load over-threshold normalization terms, valve opening over-threshold normalization terms, and spalling over-threshold normalization terms, as follows: ； ； ； in, , , These are the normalized terms for dynamic load over-threshold, valve opening over-threshold, and plate over-threshold, respectively. S53, logarithmically compress the microseismic energy sum of the current time window to form a logarithmic compression term for microseismic energy, and construct a high-energy event weighting term by combining it with high-energy event counting, expressed as: ； ； in, This is the logarithmic compression term of the microseismic energy. Weighting for high-energy events; S54 integrates the trends of dynamic load coefficient, valve opening rate, sidewall, and microseismic energy and trend, and compresses them using a hyperbolic tangent function to obtain the trend fusion term. The microseismic energy and trend are normalized using the microseismic energy and mean over the past 30 time windows, and are expressed as: ； ； in, The data represents the microseismic energy and mean over a period of approximately 30 time windows. For trend fusion items, It is the hyperbolic tangent function; S55 linearly weights the dynamic load overthreshold normalization term, valve opening overthreshold normalization term, spalling overthreshold normalization term, microseismic energy logarithmic compression term, high-energy event weighting term, and trend fusion term to obtain a linear fusion score, which is then mapped to the risk probability of the current time window using the Sigmoid function, expressed as: ； ； in, For linear fusion scores, Risk probability; S56, the risk probability of the current time window. By comparing with stability and instability thresholds, the corresponding risk status is determined, including stable, critical, and unstable. S57, determine the risk status corresponding to the candidate risk window. If the current time window is determined to be critical or unstable, output the current alarm flag. For non-candidate risk windows, force the output to be stable and do not trigger the current alarm. S58, construct a prediction enhancement factor based on the trend fusion term of the current time window, which is used to enhance and correct the risk probability of the next time window. Calculate the risk probability of the next time window based on the risk probability of the current time window and the prediction enhancement factor, and output a pre-alarm flag for the next time window based on whether the risk probability of the next time window reaches the instability threshold, expressed as: ； ； ； in, To predict the enhancement factor, The probability of risk in the next time window. This is a truncation function. Pre-alarm sign for the next window; S59 outputs the risk probability, risk status, current alarm flag, risk probability of the next time window, and pre-alarm flag of the next window for each candidate risk window.

8. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 7, characterized in that, The risk status is represented as follows: ； in, To stabilize the threshold, This is the instability threshold.

9. The intelligent identification and early warning method for the instability state of a hydraulic support under dynamic load as described in claim 8, characterized in that, S6 includes: S61, for each time window Construct a risk outcome record, including time window number, average resistance of working face window, maximum resistance of working face window, valve opening rate, dynamic load coefficient, maximum spalling depth, number of micro-seismic events, sum of micro-seismic energy, maximum micro-seismic energy, high-energy event count, adaptive threshold for dynamic load coefficient, adaptive threshold for valve opening rate, adaptive threshold for maximum spalling depth, risk probability, risk status, current alarm flag, and pre-alarm flag for the next window; S62, organize the risk results records in chronological order, and write the characteristic parameters, adaptive thresholds, risk probabilities, risk status and alarm results corresponding to each time window into the risk results file; S63, based on the risk probability, stability threshold and instability threshold in the risk result file, visualize the graph by plotting the curve of risk probability changing with time window on the time axis, and overlaying the stability threshold, instability threshold, current alarm flag and next window pre-alarm flag to generate a risk probability time series graph; S64. Based on the pressure interval marking and the corresponding pressure event number, summarize the statistical characteristics of each pressure event, including the average resistance of the working face window, valve opening rate, dynamic load coefficient, maximum rib depth, number of micro-seismic events, micro-seismic energy and risk probability, and generate a statistical plot of the pressure event using the pressure event number as the unit. S65. Based on the dynamic load factor and valve opening rate in the risk results file, construct a two-dimensional scatter plot. By drawing the scatter plot of the synergistic relationship between the dynamic load factor and valve opening rate, show the coupling relationship between the support impact load and the valve assembly unloading behavior, thus forming a dynamic load factor and valve opening rate relationship diagram. S66, construct a coupling relationship diagram using the maximum spalling depth and microseismic energy from the risk results file, and generate a spalling depth and microseismic energy coupling diagram by plotting the coupling distribution relationship between the maximum spalling depth and microseismic energy. S67 saves the generated risk probability time series diagram, pressure event statistics diagram, dynamic load coefficient and valve opening rate relationship diagram, and spalling depth and microseismic energy coupling diagram as graphic files, and outputs them together with the risk results file.

Images