Early warning system and method based on multi-field information fusion

Abstract

The present invention provides an early warning system and method based on multi-field information fusion, which relates to the field of ground pressure detection technology, and includes an acquisition module, a processing module and a generation module; the acquisition module divides the monitoring units based on the physical information of the monitoring area, and obtains stress information, vibration information and energy information by directly acquiring and multiplexing the multi-field information of adjacent monitoring units; the processing module uses a sliding window mechanism to batch load the monitoring unit data, and calculates target features based on the data in the window, including statistical features, trend features and frequency features; the generation module dynamically adjusts the risk weights of multi-field information based on the target features, calculates a comprehensive risk assessment value in combination with the adjusted weights, and generates early warning information; the present invention can realize the dynamic acquisition and fusion of multi-field information, optimize data processing efficiency, improve the real-time and accuracy of early warning, and is suitable for safety monitoring and risk management in complex mine environments.

Description

Technical Field

[0001] The present disclosure relates to the field of ground pressure detection technology, and in particular, to an early warning system and method based on multi-field information fusion. Background Art

[0002] With the increase in the depth and intensity of coal mining, rock burst has become more and more serious. Its rapidity, suddenness and destructiveness have posed a huge threat to coal mine production safety. In order to cope with the risk of rock burst, data collection and analysis using multi-physical quantity monitoring methods has become an important early warning method. However, the traditional single signal monitoring method faces many limitations in practical applications, such as insufficient positioning accuracy, significant influence by external interference, and inability to fully reflect rock burst precursor information.

[0003] For example, the Chinese patent application with publication number CN117786597A discloses a coal mine rock burst data enhancement and deep fusion warning method under small sample conditions. This method builds a generative adversarial network to generate a data expansion small sample data set, and combines a hybrid deep neural network model to extract time series features to achieve efficient early warning of rock burst.

[0004] However, this technology mainly focuses on the problem of small sample data and fails to fully solve the problem of real-time collection and dynamic analysis of multi-physical field information of rock burst. In addition, this method cannot flexibly adjust the collection strategy for monitoring units in different areas, which may lead to low collection efficiency or data omissions under conditions of large monitoring areas or complex environments. Summary of the invention

[0005] In view of the deficiencies of the prior art, the present invention discloses an early warning system and method based on multi-field information fusion.

[0006] In a first aspect, the present invention discloses an early warning system based on multi-field information fusion, the system comprising:

[0007] A collection module collects multi-field information of a monitoring area using a plurality of preset monitoring units, wherein the collection area corresponding to the monitoring unit is divided based on the physical information of the monitoring area; the collection method of the multi-field information includes: direct collection and multiplexing of multi-field information of adjacent monitoring units; wherein the multi-field information includes: stress information, vibration information and energy information;

[0008] A processing module, using a sliding window mechanism, batch-loads the multiple fields of information in the multiple monitoring units, and determines target information features based on the multiple fields of information in the window; wherein the target information features include: statistical features, trend features, and frequency features;

[0009] A generation module dynamically adjusts the risk weights of different types of information in the multiple fields of information based on the target information characteristics; performs weighted calculation on the risk assessment values ​​corresponding to the multiple fields of information based on the adjusted risk weights to generate warning information.

[0010] As an optional implementation, the physical information includes: a topological structure; based on the physical information of the monitoring area, a plurality of monitoring units for collecting multi-field information of the monitoring area are divided, including:

[0011] Based on the topological structure of the monitoring area, a plurality of monitoring units for collecting multi-field information of the monitoring area are divided, wherein the topological structure includes: tunnels, working faces, and mining areas; each of the monitoring units corresponds to at least one of the main tunnels, branch tunnels, working faces, and mining areas.

[0012] As an optional implementation manner, the collecting multiple fields of information of the monitoring area includes:

[0013] In response to the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit being greater than or equal to the first fluctuation threshold, setting the collection mode of the target monitoring unit to direct collection;

[0014] In response to the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit being less than the first fluctuation threshold, the collection mode of the target monitoring unit is set to multiplexing the multi-field information of adjacent monitoring units.

[0015] As an optional implementation, the direct collection includes: collecting the multiple field information of the monitoring area corresponding to the target monitoring unit at a first frequency through multiple sensors deployed in the target monitoring unit;

[0016] The multiplexing of the multi-field information of the adjacent monitoring unit includes: determining the adjacent monitoring unit of the target monitoring unit based on the adjacency relationship of the target monitoring unit, and selecting at least one of the adjacent monitoring units;

[0017] Acquire multiple fields of information of the selected adjacent monitoring unit;

[0018] The multi-field information of the target monitoring unit is generated by performing interpolation calculation or direct reference on the multi-field information of the adjacent monitoring unit.

[0019] As an optional implementation manner, the determining the adjacent monitoring unit of the target monitoring unit based on the adjacency relationship of the target monitoring unit includes:

[0020] According to the position of the target monitoring unit in the topological structure, determining the monitoring unit directly connected to the target monitoring unit as an adjacent monitoring unit;

[0021] The adjacent monitoring units include monitoring units that share a common border with the target monitoring unit or are located in an adjacent area.

[0022] As an optional implementation manner, the selecting at least one adjacent monitoring unit includes:

[0023] Based on the similarity between the historical multi-field information of the adjacent monitoring unit and the historical multi-field information of the target monitoring unit, the adjacent monitoring unit with the highest similarity is selected.

[0024] As an optional implementation manner, before collecting the multi-field information of the monitoring area, the method further includes:

[0025] Based on the historical multi-field information of the target monitoring unit, the real-time multi-field information of the target monitoring unit collected at the second frequency, and the real-time multi-field information of the adjacent monitoring units, a first prediction model is used to generate a predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit.

[0026] As an optional implementation, the sliding window mechanism includes: assigning different weights to the multiple fields of information collected by each monitoring unit based on the multiple fields of information collection mode of each monitoring unit;

[0027] Wherein, in response to the acquisition mode being direct acquisition, when determining the target information feature, a first weight is assigned to the multiple fields of information collected by the monitoring unit;

[0028] In response to the acquisition mode being multiplexing of multiple fields of information of adjacent monitoring units, when determining the target information feature, assigning a second weight to the multiple fields of information collected by the monitoring unit;

[0029] The first weight is greater than the second weight.

[0030] As an optional implementation manner, dynamically adjusting the risk weights of different types of information in the multiple fields of information based on the target information characteristics includes:

[0031] determining a mean and a variance of the stress information based on the statistical feature, and in response to the mean or the variance of the stress information exceeding a preset stress information threshold, increasing a risk weight of the stress information;

[0032] Based on the trend feature, determining a change trend of the vibration information, and in response to the vibration information presenting an upward trend, adjusting a risk weight of the vibration information based on a trend strength;

[0033] Based on the frequency characteristics, an abnormal frequency band in the energy information is identified, and in response to the occurrence frequency of the abnormal frequency band of the energy information exceeding a preset frequency threshold, a risk weight of the energy information is increased.

[0034] In a second aspect, the present invention further discloses an early warning method based on multi-field information fusion, the method comprising:

[0035] Using a plurality of preset monitoring units to collect multi-field information of a monitoring area, wherein the collection area corresponding to the monitoring unit is divided based on the physical information of the monitoring area; the collection method of the multi-field information includes: direct collection and multiplexing of multi-field information of adjacent monitoring units; wherein the multi-field information includes: stress information, vibration information and energy information;

[0036] Using a sliding window mechanism, the multiple fields of information in the multiple monitoring units are batch loaded, and target information features are determined based on the multiple fields of information in the window; wherein the target information features include: statistical features, trend features, and frequency features;

[0037] Based on the target information characteristics, the risk weights of different types of information in the multiple fields of information are dynamically adjusted; based on the adjusted risk weights, the risk assessment values ​​corresponding to the multiple fields of information are weighted calculated to generate early warning information.

[0038] Compared with the prior art, the beneficial effects of the present invention are: through the fusion of multi-field information and dynamic collection strategy, comprehensive monitoring and early warning of coal mine rock burst are realized. The collection method is dynamically adjusted according to the predicted value of the fluctuation amplitude of multi-field information of the target monitoring unit, and direct collection or data reuse of adjacent monitoring units are flexibly selected, which not only improves the collection accuracy, but also significantly optimizes the resource utilization efficiency.

[0039] At the same time, the sliding window mechanism is used to efficiently process the collected data, and the weights of data collected in different ways are dynamically allocated to highlight the dominant role of directly collected data, which not only reduces the computational complexity but also ensures the reliability of the analysis results. During the data analysis process, the present invention dynamically adjusts the risk weights of multiple fields of information based on the characteristics of target information, and highlights key risk information according to the changes in real-time data, thereby achieving a more accurate comprehensive risk assessment.

[0040] In addition, the present invention divides the monitoring units according to the topological structure of the monitoring area, so that it can flexibly adapt to the complex mine environment and show superior performance in scenarios with large-scale monitoring and limited resources. Through the dynamic collection, weight optimization and real-time processing strategy of multi-field information, the present invention effectively solves the shortcomings of the existing technology in balancing resource consumption and real-time performance, and provides efficient and reliable technical support for the real-time early warning of coal mine rock burst. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of an early warning system based on multi-field information fusion provided by an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of a monitoring area provided by an embodiment of the present invention;

[0043] Figure 3 A flowchart of a multi-field information multiplexing method provided by an embodiment of the present invention;

[0044] Figure 4 A flowchart of an early warning method based on multi-field information fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0046] See also Figure 1 , Figure 1 A schematic diagram of an early warning system based on multi-field information fusion provided by an embodiment of the present invention. An embodiment of the present invention provides an early warning system based on multi-field information fusion. The system includes a collection module 10, a processing module 20 and a generation module 30, and each module cooperates with each other to achieve real-time monitoring and early warning of coal mine rock burst risks. Among them:

[0047] The acquisition module 10 uses a plurality of preset monitoring units to acquire multi-field information of the monitoring area, wherein the acquisition area corresponding to the monitoring unit is divided based on the physical information of the monitoring area; the acquisition method of the multi-field information includes: direct acquisition and multiplexing of multi-field information of adjacent monitoring units; wherein the multi-field information includes: stress information, vibration information and energy information;

[0048] The processing module 20 uses a sliding window mechanism to batch load the multiple field information in the multiple monitoring units, and determines the target information characteristics based on the multiple field information in the window; wherein the target information characteristics include: statistical characteristics, trend characteristics and frequency characteristics;

[0049] The generation module 30 dynamically adjusts the risk weights of different types of information in the multiple fields of information based on the target information characteristics; performs weighted calculation on the risk assessment values ​​corresponding to the multiple fields of information based on the adjusted risk weights to generate warning information.

[0050] For the above acquisition module 10:

[0051] The acquisition module 10 is a core component of the early warning system based on multi-field information fusion of the present invention, and is mainly responsible for dividing the monitoring units based on the physical information of the monitoring area and collecting multi-field information in each monitoring unit.

[0052] See also Figure 2, Figure 2 A schematic diagram of a monitoring area provided by an embodiment of the present invention. As an optional implementation, the physical information includes: a topological structure; based on the physical information of the monitoring area, a plurality of monitoring units for collecting multi-field information of the monitoring area are divided, including:

[0053] Based on the topological structure of the monitoring area, a plurality of monitoring units for collecting multi-field information of the monitoring area are divided, wherein the topological structure includes: tunnels, working faces, and mining areas; each of the monitoring units corresponds to at least one of the following areas: main tunnels, branch tunnels, working faces, and mining areas.

[0054] In some embodiments, the physical information used to divide the monitoring units may include, in addition to the topological structure, other physical information, such as geological characteristics, coal seam structure, rock stress distribution, coal mining process parameters, and environmental temperature and humidity characteristics, etc. These information can reflect the characteristics of the monitoring area to varying degrees and affect the division strategy of the monitoring units.

[0055] It should be noted that, compared with the other physical information mentioned above, the selection of topological structure as the main basis for the division of monitoring units in this embodiment has many advantages. First, the topological structure is a stable and relatively fixed physical information, which is not significantly affected by coal mining activities or environmental changes, and therefore has high reliability. Secondly, it is relatively easy to obtain topological structure data, which can be directly obtained through coal mine geological modeling databases, engineering design drawings or scanning equipment. In addition, the geometric characteristics of the topological structure (such as the length, width, and branch position of the tunnel) are highly consistent with the division rules of the monitoring unit, and the scope of the monitoring unit can be directly determined by simple geometric rules without relying on complex dynamic calculations or high-frequency data input. This feature not only reduces the implementation complexity of the system, but also ensures the consistency of the monitoring unit division results.

[0056] In specific implementation, the topological structure data of the tunnels and related areas can be directly imported from the geological modeling database of the coal mine project. These data usually include the direction of the main tunnel, the branch nodes of the branch tunnel, the dynamic position of the working face, and the boundary range of the mining area. The geometric morphology of the tunnel is scanned regularly using coal mine monitoring equipment (such as laser scanners) to update the boundaries and shapes of the mined areas.

[0057] Furthermore, according to the topological structure, the monitoring area is divided into multiple monitoring units. The specific rules can be as follows:

[0058] For the main lane:

[0059] The main tunnel is the main transportation channel of the coal mine, with a large cross-section and a long straight-line distance. Therefore, the main tunnel can be divided into several monitoring units at fixed intervals (such as every 100 meters), and each monitoring unit covers a linear segment of the main tunnel.

[0060] For branch lanes:

[0061] A branch tunnel is a branch passage connected to the main tunnel, usually used for auxiliary transportation or local coal mining activities. Therefore, due to the complex geometric structure of the branch tunnel, the monitoring unit can be divided into smaller intervals (such as every 50 meters) to ensure accurate data collection.

[0062] For the working surface:

[0063] The working face is the core area of ​​coal mining operations and usually has dynamically changing boundaries. Therefore, each working face can be divided into one or more monitoring units according to its actual area. For example, for a rectangular working face, multiple monitoring units are divided into 50×50 meter grids.

[0064] For the mining area:

[0065] The mining area is an area that has been mined out, and its boundary changes over time. Therefore, the mining area can be divided into several monitoring units based on the real-time updated mining area boundary information, and each unit covers a local range of the mining area.

[0066] As an optional implementation manner, the collecting multiple fields of information of the monitoring area includes:

[0067] In response to the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit being greater than or equal to the first fluctuation threshold, setting the collection mode of the target monitoring unit to direct collection;

[0068] In response to the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit being less than the first fluctuation threshold, the collection mode of the target monitoring unit is set to multiplexing the multi-field information of adjacent monitoring units.

[0069] As an optional implementation manner, before collecting multiple field information of the monitoring area, the method further includes:

[0070] Generate a predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit using the first prediction model based on the historical multi-field information of the target monitoring unit, the real-time multi-field information of the target monitoring unit collected at the second frequency, and the real-time multi-field information of the adjacent monitoring units;

[0071] The second frequency is less than the first frequency during direct acquisition.

[0072] In order to efficiently collect multiple fields of information in the monitoring area in the early warning system, while taking into account the real-time performance of data collection and resource utilization efficiency, this embodiment adopts a dynamic collection strategy based on the predicted value of the fluctuation amplitude of multiple fields of information of the target monitoring unit. Specifically, before collecting multiple fields of information, the system generates the predicted value of the fluctuation amplitude of multiple fields of information by combining the historical data and low-frequency real-time data of the target monitoring unit using the first prediction model, and dynamically adjusts the collection mode to direct collection or multiple fields of information multiplexing according to the prediction results.

[0073] In specific implementation, the stress, vibration and energy data of the target monitoring unit in the past period of time can be obtained through the built-in storage module of each monitoring unit. These historical data include time series, change trends and statistical characteristics, which can reflect the long-term change rules of the monitoring unit. Real-time data collected from the target monitoring unit at a second frequency (such as once every 5 minutes) is used to make up for the lack of timeliness of historical data. Low-frequency real-time data mainly reflects the stress, vibration and energy status of the target monitoring unit at the current time point.

[0074] Furthermore, the historical data and low-frequency real-time data of the target monitoring unit are used as input data, and the first prediction model is used to predict the fluctuation amplitude of multiple fields of information. The first prediction model can adopt a statistical model based on time series analysis (such as ARIMA) or a machine learning model (such as LSTM, GRU). For example, the LSTM network can predict the future range of changes in stress, vibration or energy by capturing the long-term and short-term dependencies of time series data. The first prediction model generates the predicted value of the fluctuation amplitude of multiple fields of information of the target monitoring unit, which is used to dynamically adjust the collection method.

[0075] In a specific implementation, when the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit is greater than or equal to the first fluctuation threshold, it is determined that the multi-field information of the target monitoring unit may have changed significantly and high-precision data needs to be acquired in real time. At this time, the acquisition module 10 sets the acquisition mode to direct acquisition.

[0076] As an optional implementation, the direct collection includes: collecting the multiple fields of information of the monitoring area corresponding to the target monitoring unit at a first frequency through multiple sensors deployed in the target monitoring unit.

[0077] In the direct collection mode, the collection module 10 calls multiple sensors deployed in the target monitoring unit, such as stress sensors, vibration sensors, and energy detection sensors, to collect multiple field information at a first frequency (e.g., once every 1 minute). The real-time collected data is directly uploaded to the processing module 20 for subsequent data analysis and risk assessment.

[0078] Among them, the stress sensor collects the instantaneous and average values ​​of rock stress, which are used to analyze the changes in the stress state of the rock in the target monitoring unit; the vibration sensor records the amplitude and frequency of the vibration, which are used to detect the intensity and spectral distribution characteristics of the vibration; the energy detection sensor captures the energy release rate and accumulated energy, which are used to evaluate the fracture energy characteristics of the rock.

[0079] When the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit is less than the first fluctuation threshold, it is determined that the change of the multi-field information of the target monitoring unit is small. At this time, the multi-field information multiplexing method is selected to reduce the data collection burden.

[0080] See also Figure 3 , Figure 3 A flowchart of a method for multiplexing multi-field information provided by an embodiment of the present invention, wherein the multiplexing of multi-field information of adjacent monitoring units includes S101 to S103, wherein:

[0081] S101: determining adjacent monitoring units of the target monitoring unit based on the adjacency relationship of the target monitoring unit, and selecting at least one of the adjacent monitoring units;

[0082] S102: Acquire multiple fields of information of the selected adjacent monitoring unit;

[0083] S103: Generate multi-field information of the target monitoring unit by performing interpolation calculation or direct reference on the multi-field information of the adjacent monitoring units.

[0084] In the multi-field information multiplexing mode, the acquisition module 10 obtains multi-field information from adjacent monitoring units according to the adjacency relationship of the target monitoring unit, and generates the data of the target monitoring unit by interpolation calculation or weighted average. For example, for vibration information, the middle value of the adjacent monitoring unit data can be calculated by linear interpolation to generate the vibration characteristics of the target monitoring unit.

[0085] In a specific implementation, the acquisition module 10 transmits the collected real-time data to the processing module 20 through the wireless communication module or the wired communication module. The directly collected high-frequency data provides a reliable real-time monitoring basis for the system, which is particularly suitable for key areas with large fluctuations.

[0086] When the fluctuation amplitude of the multi-field information of the target monitoring unit is small or the sensor is unavailable, the system generates the data of the target monitoring unit by multiplexing the data of adjacent monitoring units.

[0087] In a specific implementation, according to the position of the target monitoring unit in the topological structure, the monitoring unit directly connected to it is determined as the adjacent monitoring unit. Among them, the adjacent monitoring units include:

[0088] Monitoring units that share a common boundary: for example, two endpoints or faces of a lane are shared with the target monitoring unit.

[0089] Monitoring units located in adjacent areas: For example, when the target monitoring unit is located in a main lane, its adjacent monitoring unit may be a branch lane monitoring unit connected to the main lane.

[0090] Furthermore, by analyzing the historical multi-field information of the target monitoring unit and the adjacent monitoring unit, the similarity between the two is calculated, and the adjacent monitoring unit with the highest similarity is selected. For example, for stress information, the similarity score is calculated by comparing the historical stress mean, variance, and change trend of the target monitoring unit and the adjacent monitoring unit; for vibration information, the correlation of the vibration frequency distribution of the two is compared through spectrum analysis; for energy information, the similarity is evaluated by comparing the change curves of the accumulated energy and the energy release rate. The adjacent monitoring unit with the highest similarity is selected for multiplexing of the multi-field information of the target monitoring unit.

[0091] In addition, when the historical data of the adjacent monitoring unit is highly similar to that of the target monitoring unit, the real-time multi-field information of the adjacent monitoring unit is directly quoted as the data of the target monitoring unit. When there is a certain difference between the historical data of the adjacent monitoring unit and the target monitoring unit, the real-time data of the adjacent monitoring unit is used for interpolation calculation.

[0092] For example, the linear interpolation method is used to generate the stress information of the target monitoring unit based on the stress data of the adjacent monitoring units; for vibration information, the multivariate interpolation method is used to fuse the frequency characteristics of multiple adjacent monitoring units to generate the data of the target monitoring unit.

[0093] For example, in a coal mine monitoring area, the target monitoring unit A is located in the main tunnel, which extends in a straight line, and each monitoring unit covers a spatial range of 100 meters. The sensor of the target monitoring unit A fails to collect real-time data due to equipment failure. The system generates multiple field information of the target monitoring unit A according to the data reuse strategy of the adjacent monitoring units.

[0094] First, based on the topological structure of the monitoring area, the adjacent monitoring units of the target monitoring unit A are determined to be monitoring unit B and monitoring unit C. Monitoring unit B is located in the upstream direction of the target monitoring unit A, and monitoring unit C is located in its downstream direction, and both share a common boundary with the target monitoring unit A. Subsequently, multi-field information is obtained from monitoring units B and C in real time, including stress data, vibration data, and energy data.

[0095] In order to select a suitable data reuse source, the similarity of the historical multi-field information of monitoring units B and C and the target monitoring unit A is calculated. Taking stress information as an example, the stress mean, change range and change trend of the three in the past 24 hours are analyzed, and the stress similarity between monitoring unit B and target monitoring unit A is calculated to be 0.85, and the similarity of monitoring unit C is 0.72. Since the similarity of monitoring unit B is higher, monitoring unit B is preferred as the main source of data reuse.

[0096] After obtaining the real-time stress data of monitoring units B and C, the data are interpolated. Specifically, the real-time stress values ​​of monitoring units B and C (12.5 MPa and 11.8 MPa, respectively) are used, combined with the spatial position of the target monitoring unit A, to calculate the stress data of the target monitoring unit A through linear interpolation, and the result is 12.2 MPa. Similarly, for vibration information, the vibration amplitude of the target monitoring unit A is generated as 0.7 mm / s by performing multivariate interpolation on the vibration amplitudes of monitoring units B and C (0.8 mm / s and 0.6 mm / s, respectively). In addition, the energy data of monitoring unit B is directly quoted as the energy value of the target monitoring unit A, because historical data analysis shows that the energy release characteristics of monitoring unit B are highly consistent with those of target monitoring unit A.

[0097] In this way, the multiplexed multi-field information (stress value 12.2 MPa, vibration amplitude 0.7 mm / s, energy value 5.4 kJ) of the target monitoring unit A is transmitted to the processing module 20 for subsequent feature calculation and risk assessment. Through this multiplexing strategy, even if the sensor of the target monitoring unit A fails, its multi-field information can still be accurately generated, ensuring the integrity and real-time performance of data acquisition.

[0098] In the specific implementation, a minimum direct collection ratio needs to be set (for example, 10% of the total number of monitoring units corresponding to a certain area). When the proportion of monitoring units covered by the reuse mode is too high, the system will randomly select a certain number of monitoring units for direct collection. Even if the fluctuation amplitude prediction value of these units is lower than the first fluctuation threshold, the direct collection mode will still be used.

[0099] It should be noted that the first fluctuation threshold is used to determine the switching of the acquisition mode and is set based on the statistical analysis of historical monitoring data or experimental results. For example, for stress information, the system can set the 95% percentile value of the historical stress fluctuation amplitude as the threshold; for vibration information, the standard deviation range of the abnormal frequency band can be calculated through spectrum analysis, and the amplitude change of the frequency band can be set as the threshold.

[0100] It should be emphasized that the second frequency is significantly lower than the first frequency of direct acquisition. For example, the first frequency in the direct acquisition mode is once every 1 minute, while the second frequency used in the calculation of the fluctuation range forecast value can be set to once every 5 minutes or longer. The data collected at the second frequency is mainly used for model input to reduce the collection and calculation burden of the system.

[0101] In this way, by combining historical data and low-frequency real-time data to generate fluctuation amplitude prediction values ​​and dynamically adjusting the collection method, this system can significantly optimize the allocation of data collection resources while ensuring data real-time and accuracy. The direct collection method is suitable for high-fluctuation areas to ensure data accuracy and real-time performance; while the multi-field information reuse method is suitable for low-fluctuation areas to reduce the frequency of sensor calls and improve system operation efficiency. This collection strategy is applicable to a variety of coal mine environments, especially under conditions of limited resources or complex environments, and can significantly improve the reliability of rock burst warnings.

[0102] For processing module 20:

[0103] In the present disclosure, the processing module 20 batch-loads and processes the multi-field information provided by the acquisition module 10 through a sliding window mechanism, and dynamically adjusts the weights based on the acquisition method, thereby optimizing the calculation results of the target information features.

[0104] Among them, the sliding window mechanism is an efficient data processing method. By limiting the data range of each calculation, only new data units are updated when the window moves, and the existing data in the window is reused, the computational burden of data processing is significantly reduced.

[0105] In the specific implementation, first, the spatial range of the sliding window needs to be defined, for example, a window covering a 3×3 grid, initially containing 9 monitoring units. The processing module 20 loads the multi-field information of these monitoring units from the acquisition module 10, including directly acquired data and multiplexed generated data. The data loading in the window is dynamic. Each time the window moves, only the data of the newly added monitoring unit is loaded, while the existing data of other monitoring units in the window is retained.

[0106] In the sliding window, the processing module 20 calculates the target information features of the multi-field information, including statistical features, trend features and frequency features. Statistical features such as mean, variance and extreme value can reflect the overall state of the monitoring area, such as the mean of rock stress and the fluctuation range of vibration amplitude. Trend feature analysis is used to capture the changing trend of multi-field information, such as the rising and falling trend of rock stress or the increasing trend of vibration frequency. Frequency features extract the frequency distribution information of vibration and energy data through spectrum analysis, which is used to identify abnormal frequency bands and potential risk signals.

[0107] As an optional implementation, the sliding window mechanism includes: assigning different weights to the multiple fields of information collected by each monitoring unit based on the multiple fields of information collection mode of each monitoring unit;

[0108] Wherein, in response to the acquisition mode being direct acquisition, when determining the target information feature, a first weight is assigned to the multiple fields of information collected by the monitoring unit;

[0109] In response to the acquisition mode being multiplexing of multiple fields of information of adjacent monitoring units, when determining the target information feature, assigning a second weight to the multiple fields of information collected by the monitoring unit;

[0110] The first weight is greater than the second weight.

[0111] In a specific implementation, in order to improve the accuracy of feature calculation, the processing module 20 can assign different weights to the data of the monitoring unit within the sliding window according to the collection method. For directly collected data, because its source is reliable, a higher first weight, such as 1.0, is assigned; and for multiplexed data, because it is generated based on adjacent monitoring units, a lower second weight, such as 0.5, is assigned. When calculating the target information features, the processing module 20 performs weighted processing on the data according to the weights, for example, the weighted mean is calculated by multiplying the data value by the corresponding weight and then summing them up, to ensure that the directly collected data has a greater influence in the feature calculation.

[0112] In addition, when the proportion of directly collected data in the sliding window is low, for example, less than 30% of the total data volume, the weight of the reused data is further reduced to reduce its deviation from the feature calculation results. In this way, even when the direct collection ratio is low, the accuracy of the calculation results can still be guaranteed by optimizing the weight distribution.

[0113] For example, in a coal mine area, the processing module 20 processes a sliding window containing 9 monitoring units, of which 6 monitoring units obtain data through direct acquisition and 3 monitoring units generate data through multiplexing. After the processing module 20 loads these data, it calculates the target information features for stress information, vibration information and energy information respectively. For example, when calculating the mean of stress information, the directly acquired data is processed with a weight of 1.0, and the multiplexed data is processed with a weight of 0.5, and finally a weighted mean is obtained. When analyzing the vibration frequency distribution, the directly acquired data contributes more to the spectral characteristics, while the multiplexed data has a smaller impact on the results due to its lower weight.

[0114] When the sliding window moves forward, the data of a new monitoring unit is added, and the data of a monitoring unit is removed from the window. The processing module 20 only needs to load the new data and update the feature calculation results in the window, without reprocessing the data of the entire area. This incremental calculation method not only reduces the calculation complexity, but also can update the target information features in real time, providing accurate data support for the generation module 30.

[0115] By combining the sliding window mechanism with the weight-based optimization strategy, the processing module 20 can reduce the computational complexity while highlighting the leading role of directly collected data in feature calculation, ensuring that the system can still provide high-precision risk assessment results in the case of multiple information collection methods. This mechanism is particularly suitable for complex coal mine environments with high data collection frequency and wide regional coverage.

[0116] For the above generation module 30:

[0117] In this embodiment, the generation module 30 first receives the target information features provided by the processing module 20, including statistical features, trend features, frequency features, etc. Based on these features, the generation module 30 dynamically adjusts the risk weights of different types of information in the multi-field information, and performs weighted calculation on the risk assessment values ​​of various types of multi-field information, thereby generating a comprehensive risk assessment value.

[0118] In the process of dynamically adjusting the risk weight, the generation module 30 combines the different properties of the target information features to independently adjust the weights of stress information, vibration information and energy information. For statistical features, the generation module 30 calculates the mean and variance of the stress information of the target monitoring unit and compares them with the preset stress information threshold based on the statistical analysis of historical monitoring data.

[0119] For example, when the mean of stress information is higher than 10 MPa, or the variance exceeds 2 MPa, it is determined that the stress information is significantly abnormal, and the risk weight of the stress information is increased. For example, the weight can be adjusted from 0.6 to 0.8, thereby highlighting the influence of stress information on comprehensive risk assessment.

[0120] For trend characteristics, the generation module 30 analyzes the changing trend of vibration information, for example, the growth rate and frequency change of vibration amplitude. When the vibration amplitude increases from 0.5 mm / s to 1.2 mm / s, the growth rate reaches 150%, and exceeds the set 100% threshold, it is determined that there is a significant change in the vibration information, and the risk weight of the vibration information is increased. The adjustment range of the weight is related to the trend strength. For example, the higher the growth rate, the greater the increase in the adjusted weight, so that the vibration information occupies a larger proportion in the comprehensive assessment.

[0121] For frequency characteristics, the generation module 30 identifies abnormal frequency bands in energy information through spectrum analysis. When the energy value of a specific frequency band appears 6 times within 10 minutes, exceeding the 5-time frequency threshold set in the experiment, the system increases the risk weight of the energy information. For example, the weight is increased from 0.4 to 0.6 to ensure that energy anomalies are given enough attention in the comprehensive assessment.

[0122] It is understandable that the specific adjustment values ​​of the above weights can also be obtained through data analysis, model processing, etc.

[0123] After completing the dynamic adjustment of the risk weight, the generation module 30 performs weighted calculation on the risk assessment values ​​of multiple fields of information based on the adjusted weight to obtain a comprehensive risk assessment value. The calculation formula is:

[0124] ;

[0125] in, , , are the dynamic risk weights of stress information, vibration information and energy information, , , are the risk assessment values ​​corresponding to stress information, vibration information and energy information. For example, in a coal mining area, the adjusted risk weights are 0.8, 0.7 and 0.6, and the corresponding risk assessment values ​​are 3.0, 2.5 and 1.8, respectively. The comprehensive risk assessment value is calculated as:

[0126] ;

[0127] The generation module 30 compares the comprehensive risk assessment value with the warning threshold. When the comprehensive risk assessment value exceeds the warning threshold, for example, when the warning threshold is 5.0, a high-risk warning message is generated, and the target monitoring unit is marked as a high-risk area. This information can be displayed through the system interface or sent to the mobile device of the mine manager, prompting the on-site staff to promptly check the rock stability and vibration source of the area and take necessary prevention and control measures.

[0128] In addition, the above-mentioned weight adjustment method can also be applied to machine learning models, making the warning information generated by the model more accurate.

[0129] In this way, by dynamically adjusting the risk weights, the generation module 30 can timely highlight key risk features according to changes in real-time monitoring data, thereby improving the accuracy and real-time performance of the comprehensive risk assessment. In addition, the weighted calculation method of the comprehensive risk assessment value enables the system to flexibly adapt to different scenarios and multi-field information characteristics, significantly improving the reliability and efficiency of rock burst warning in complex coal mine environments.

[0130] Based on the same inventive concept, this embodiment also provides a method corresponding to the above-mentioned early warning system based on multi-field information fusion. Since the principle of solving the problem by the method in this embodiment is similar to the above-mentioned early warning system based on multi-field information fusion in this embodiment, the implementation of the method can refer to the implementation of the system, and the repeated parts will not be repeated.

[0131] See also Figure 4 , Figure 4 A flowchart of an early warning method based on multi-field information fusion provided by an embodiment of the present invention includes S201 to S203, wherein:

[0132] S201: using a plurality of preset monitoring units to collect multi-field information of a monitoring area, wherein the collection area corresponding to the monitoring unit is divided based on the physical information of the monitoring area; the collection method of the multi-field information includes: direct collection and multiplexing of multi-field information of adjacent monitoring units; wherein the multi-field information includes: stress information, vibration information and energy information;

[0133] S202: using a sliding window mechanism, batch loading the multiple fields of information in the multiple monitoring units, and determining target information features based on the multiple fields of information in the window; wherein the target information features include: statistical features, trend features, and frequency features;

[0134] S203: Based on the target information characteristics, dynamically adjust the risk weights of different types of information in the multiple fields of information; perform weighted calculation on the risk assessment values ​​corresponding to the multiple fields of information based on the adjusted risk weights to generate warning information.

[0135] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center through a wired network or / and a wireless network. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0136] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0137] It should be understood that determining B based on A does not mean determining B only based on A. B can also be determined based on A and / or other information.

[0138] In the description of this specification, the description with reference to the terms "in a specific implementation", "exemplary", "for example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0139] The preferred embodiments of the present invention disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details in detail, nor do they limit the present application to specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can understand and use the present application well. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. An early warning system based on multi-field information fusion, characterized in that: include: The acquisition module uses a plurality of preset monitoring units to acquire multi-field information of the monitoring area, wherein the acquisition area corresponding to the monitoring unit is divided based on the physical information of the monitoring area; the acquisition method of the multi-field information includes: direct acquisition and multiplexing of multi-field information of adjacent monitoring units; wherein the multi-field information includes: stress information, vibration information and energy information; the physical information includes: topological structure; The collecting of the multi-field information of the detection area includes: in response to the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit being less than the first fluctuation threshold, setting the collection mode of the target monitoring unit to multiplexing the multi-field information of adjacent monitoring units; The multiplexing of the multi-field information of the adjacent monitoring unit includes: determining the adjacent monitoring unit of the target monitoring unit based on the adjacency relationship of the target monitoring unit, and selecting at least one of the adjacent monitoring units; Acquire multiple fields of information of the selected adjacent monitoring unit; Generate multi-field information of the target monitoring unit by performing interpolation calculation or direct reference on the multi-field information of the adjacent monitoring unit; The determining, based on the adjacency relationship of the target monitoring unit, the adjacent monitoring unit of the target monitoring unit comprises: According to the position of the target monitoring unit in the topological structure, determining the monitoring unit directly connected to the target monitoring unit as an adjacent monitoring unit; The adjacent monitoring units include monitoring units that share a common boundary with the target monitoring unit or are located in an adjacent area; The selecting at least one of the adjacent monitoring units comprises: Based on the similarity between the historical multi-field information of the adjacent monitoring unit and the historical multi-field information of the target monitoring unit, selecting the adjacent monitoring unit with the highest similarity; The selecting of the adjacent monitoring unit with the highest similarity includes: calculating the similarity score by comparing the historical stress mean, variance and change trend of the target monitoring unit and the adjacent monitoring unit; comparing the correlation of the vibration frequency distribution of the target monitoring unit and the adjacent monitoring unit by spectrum analysis; and evaluating the similarity of the target monitoring unit and the adjacent monitoring unit by comparing the change curves of the accumulated energy and the energy release rate; A processing module, using a sliding window mechanism, batch-loads the multiple fields of information in the multiple monitoring units, and determines target information features based on the multiple fields of information in the window; wherein the target information features include: statistical features, trend features, and frequency features; A generation module dynamically adjusts the risk weights of different types of information in the multiple fields of information based on the target information characteristics; performs weighted calculation on the risk assessment values ​​corresponding to the multiple fields of information based on the adjusted risk weights to generate warning information.

2. The early warning system based on multi-field information fusion according to claim 1 is characterized in that: Based on the physical information of the monitoring area, a plurality of monitoring units for collecting multi-field information of the monitoring area are divided, including: Based on the topological structure of the monitoring area, a plurality of monitoring units for collecting multi-field information of the monitoring area are divided, wherein the topological structure includes: tunnels, working faces, and mining areas; each of the monitoring units corresponds to at least one of the main tunnels, branch tunnels, working faces, and mining areas.

3. The early warning system based on multi-field information fusion according to claim 2 is characterized in that: The collecting of multiple fields of information of the monitoring area includes: In response to the predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit being greater than or equal to the first fluctuation threshold, the collection mode of the target monitoring unit is set to direct collection.

4. The early warning system based on multi-field information fusion according to claim 3 is characterized in that: The direct collection includes: collecting the multiple fields of information of the monitoring area corresponding to the target monitoring unit at a first frequency through multiple sensors deployed in the target monitoring unit.

5. The early warning system based on multi-field information fusion according to claim 4 is characterized in that: Before collecting the multiple field information of the monitoring area, the method further includes: Based on the historical multi-field information of the target monitoring unit, the real-time multi-field information of the target monitoring unit collected at the second frequency, and the real-time multi-field information of the adjacent monitoring units, a first prediction model is used to generate a predicted value of the fluctuation amplitude of the multi-field information of the target monitoring unit.

6. The early warning system based on multi-field information fusion according to claim 5 is characterized in that: The sliding window mechanism includes: assigning different weights to the multiple fields of information collected by each monitoring unit based on the multiple fields of information collection method of each monitoring unit; Wherein, in response to the acquisition mode being direct acquisition, when determining the target information feature, a first weight is assigned to the multiple fields of information collected by the monitoring unit; In response to the acquisition mode being multiplexing of multiple fields of information of adjacent monitoring units, when determining the target information feature, assigning a second weight to the multiple fields of information collected by the monitoring unit; The first weight is greater than the second weight.

7. The early warning system based on multi-field information fusion according to claim 6 is characterized in that: The dynamically adjusting the risk weights of different types of information in the multiple fields of information based on the target information characteristics includes: determining a mean and a variance of the stress information based on the statistical feature, and in response to the mean or the variance of the stress information exceeding a preset stress information threshold, increasing a risk weight of the stress information; Based on the trend feature, determining a change trend of the vibration information, and in response to the vibration information presenting an upward trend, adjusting a risk weight of the vibration information based on a trend strength; Based on the frequency characteristics, an abnormal frequency band in the energy information is identified, and in response to the occurrence frequency of the abnormal frequency band of the energy information exceeding a preset frequency threshold, a risk weight of the energy information is increased.

8. An early warning method based on multi-field information fusion, implemented based on an early warning system based on multi-field information fusion according to any one of claims 1 to 7, characterized in that: include: Using a plurality of preset monitoring units to collect multi-field information of a monitoring area, wherein the collection area corresponding to the monitoring unit is divided based on the physical information of the monitoring area; the collection method of the multi-field information includes: direct collection and multiplexing of multi-field information of adjacent monitoring units; wherein the multi-field information includes: stress information, vibration information and energy information; Using a sliding window mechanism, the multiple fields of information in the multiple monitoring units are batch loaded, and target information features are determined based on the multiple fields of information in the window; wherein the target information features include: statistical features, trend features, and frequency features; Based on the target information characteristics, the risk weights of different types of information in the multiple fields of information are dynamically adjusted; based on the adjusted risk weights, the risk assessment values ​​corresponding to the multiple fields of information are weighted calculated to generate early warning information.

Images