Coal mine slope deformation monitoring system and method based on multi-source data fusion

Abstract

The present application relates to the technical field of coal mine slope monitoring, and discloses a coal mine slope deformation monitoring system and method based on multi-source data fusion. The system comprises data acquisition, pattern matching, action point positioning, matching degree analysis and weight calculation modules. The data acquisition module acquires real-time slope displacement time series, stress change gradient and microseismic energy distribution; the pattern matching module selects a deformation mode section matching the current deformation intensity characteristics from the historical database; the action point positioning module analyzes the displacement data to determine the displacement acceleration action point and calculates the time delay; the matching degree analysis module compares the stress and displacement data sequence trends before and after the acceleration action point to determine the deformation matching degree; and the weight calculation module determines the deformation weight of the mode section in combination with the time delay, deformation matching degree and deformation intensity difference. The system is associated with historical patterns through multi-source data fusion, and optimizes slope deformation monitoring and trend judgment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal mine slope monitoring, in particular to a coal mine slope deformation monitoring system and method based on multi-source data fusion. BACKGROUND

[0002] In coal mining operations, the stability of the slope structure is directly related to the normal progress of the mining operation. If the slope deformation is not detected in time and intervention measures are not taken, it may cause landslides, collapses and other safety accidents, which pose a serious threat to personnel safety and equipment property. In the current coal mine slope deformation monitoring technology, most monitoring systems rely on a single type of data collection method, such as obtaining slope position change information only through displacement sensors or capturing stress distribution only by relying on stress monitoring equipment. Such single data source monitoring mode has obvious limitations. Single data cannot fully reflect the complex process of slope deformation. For example, when a small stress concentration occurs in the slope, if there is no displacement data to verify, it is difficult to accurately judge the development trend of the deformation; and when only displacement changes are concerned, it is not possible to know whether the stress state has reached the critical value, which may lead to misjudgment or missed judgment of the deformation risk.

[0003] The existing monitoring system lacks an effective historical data correlation mechanism in the data processing and analysis link. Coal mine slope deformation often shows certain time sequence regularity. The deformation mode in different historical periods may be similar to the current deformation. By comparing and analyzing similar deformation modes, important references can be provided for current deformation trend prediction. However, the current system mostly only makes simple threshold judgments on real-time data, such as issuing an alarm when the displacement exceeds the set threshold. It fails to deeply mine the deformation mode information contained in the historical database and cannot use the development law of similar deformation processes in the historical data to assist in judging the possible trend of the current deformation. In addition, in determining the key nodes of slope deformation (such as the starting point of displacement acceleration change), the existing technology lacks a scientific analysis method, making it difficult to accurately locate the deformation acceleration point and thus unable to accurately calculate the time delay characteristics of the deformation development, which results in a large error in predicting the time window in which the deformation may cause an accident, affecting the efficiency of subsequent prevention and control measures. SUMMARY

[0004] The present application aims to provide a coal mine slope deformation monitoring system and method based on multi-source data fusion to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides a coal mine slope deformation monitoring system based on multi-source data fusion, which comprises:

[0006] The data acquisition module is used to acquire multi-source monitoring data of the slope in real time. The multi-source monitoring data includes displacement time series, stress change gradient and microseismic energy distribution.

[0007] The pattern matching module is used to filter deformation pattern segments from the historical database that match the deformation intensity characteristics of the current monitoring period.

[0008] The application point positioning module is used to analyze the displacement data changes of the matching deformation mode segment in time sequence, determine the displacement acceleration application point of the matching deformation mode segment, and use the time difference between the starting point of each matching deformation mode segment and the displacement acceleration application point as the time delay of each matching deformation mode segment.

[0009] The matching degree analysis module is used to compare the change trends of the stress change data sequence before the displacement acceleration point with the displacement data sequence after the displacement acceleration point to determine the deformation matching degree of each matching deformation mode segment.

[0010] The weight calculation module is used to determine the deformation weight of the matching deformation mode segment by combining the time delay, the deformation matching degree, and the difference in deformation intensity characteristics between the matching deformation mode segment and the current monitoring period.

[0011] Preferably, the pattern matching module includes:

[0012] The historical data segmentation unit is used to divide the displacement time series in the historical database into multiple displacement data segments and extract the deformation intensity characteristic value of each displacement data segment.

[0013] The strength comparison unit is used to calculate the absolute difference between the deformation strength characteristic value of the historical displacement data segment and the deformation strength characteristic value of the current monitoring period.

[0014] The negative correlation mapping unit is used to perform negative correlation mapping on the absolute difference values ​​to obtain the deformation intensity similarity.

[0015] The pattern filtering unit is used to filter out matching deformation pattern segments from historical displacement data segments based on the deformation intensity similarity.

[0016] Preferably, the action point positioning module specifically performs the following:

[0017] For any matching deformation mode segment, an adaptive sliding window decomposition algorithm is used to identify abrupt change points in the displacement time sequence;

[0018] The monitoring moment when the first abrupt change in displacement gradient occurs is marked as the displacement acceleration point.

[0019] Preferably, the matching degree analysis module includes:

[0020] The stress change analysis unit is used to calculate the average slope of adjacent data points in the stress change data sequence before the displacement acceleration point.

[0021] The displacement change analysis unit is used to calculate the average slope of adjacent data points in the displacement data sequence after the displacement acceleration point.

[0022] The trend difference mapping unit is used to perform negative correlation mapping on the difference between the output value of the stress change analysis unit and the output value of the displacement change analysis unit, and output the deformation matching degree.

[0023] Preferably, the weight calculation module includes:

[0024] The coordinate construction unit uses the stress value corresponding to the displacement acceleration point of each matching deformation mode segment as the abscissa and the time delay as the ordinate to construct a two-dimensional coordinate point set.

[0025] The curve fitting unit performs nonlinear curve fitting on the set of two-dimensional coordinate points to generate a fitted curve.

[0026] The residual calculation unit calculates the absolute value of the residual between each coordinate point and the fitted curve;

[0027] The weighted synthesis unit uses the deformation matching degree as the numerator and the product of the time delay and the absolute value of the residual as the denominator. It then weights and fuses the ratio result with the difference value of the deformation intensity feature to output the deformation weight.

[0028] Preferably, the system further includes: a topology mapping module, which constructs a slope rock stratum node topology network based on the rock stratum node positions corresponding to the displacement acceleration points of the matching deformation mode segments;

[0029] The topology mapping module specifically executes:

[0030] An initial topology network containing the locations of rock strata nodes and lithological parameters is constructed based on geological exploration data;

[0031] Map the coordinates of the displacement acceleration point of the matching deformation mode segment to the corresponding rock layer node of the initial topology network;

[0032] The node connection weights are updated by analyzing the stress transmission relationship of the rock strata to generate a rock strata node topology network.

[0033] Preferably, the system further includes: a probability inference module, used to input the deformation weights and the rock stratum node topology network into the deformation propagation inference model to generate an abnormal deformation probability distribution map of the slope area;

[0034] The probability deduction module includes:

[0035] The propagation simulation unit is used to simulate the propagation path of deformable waves in the rock strata based on tensor convolutional networks, using nodes carrying deformation weights as the initial propagation source.

[0036] The probability calculation unit counts the frequency of occurrence of each rock layer node in the propagation path and calculates the probability of abnormal deformation residence by combining lithological parameters.

[0037] The heat map generation unit generates an abnormal deformation probability distribution map covering the entire slope based on the abnormal deformation retention probability.

[0038] Preferably, the system further includes: a strategy generation module, which configures monitoring strategy parameters based on the gradient change characteristics of the abnormal deformation probability distribution map;

[0039] The strategy generation module specifically executes the following:

[0040] Extract the set of rock strata nodes whose probability values ​​exceed a preset threshold from the abnormal deformation probability distribution map;

[0041] Calculate the rate of change of the probability gradient between adjacent nodes;

[0042] Based on the probability gradient change rate, monitoring strategy parameters including monitoring frequency parameters and disturbance test parameters are generated.

[0043] Preferably, the system further includes a control execution module, which adjusts the acquisition mode of multi-source monitoring data based on the monitoring strategy parameters.

[0044] The control execution module includes:

[0045] The mode switching unit is used to increase the sampling frequency of the displacement sensor to 5 times the original frequency when the monitoring frequency parameter is triggered.

[0046] The disturbance execution unit is used to apply multi-frequency vibration excitation signals to designated rock stratum nodes according to the disturbance test parameters.

[0047] Preferably, the present invention also includes a coal mine slope deformation monitoring method based on multi-source data fusion, the method comprising all the modules and method flow of the coal mine slope deformation monitoring system based on multi-source data fusion as described above.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] By integrating three core monitoring data types—displacement time series, stress change gradient, and microseismic energy distribution—this approach breaks through the limitations of traditional single-source monitoring, enabling a multi-dimensional and comprehensive characterization of slope deformation processes. The multi-source data complements and validates each other. Displacement data directly reflects the slope's positional change trend, stress data reveals the stress state of the slope's internal structure, and microseismic energy distribution captures the energy release from minute internal fractures. The synergistic effect of these three data sources comprehensively presents the complete process of slope deformation, from internal stress accumulation to external displacement manifestation, and then to changes in microseismic activity. This avoids the problems of misjudgment or omission of deformation states caused by the limitations of single-source data, allowing monitoring personnel to more clearly and accurately grasp the true state of slope deformation.

[0050] The pattern matching module in the system establishes an effective correlation between real-time monitoring data and historical data by filtering deformation pattern segments that match the deformation intensity characteristics of the current monitoring period from historical databases. By leveraging the development patterns of similar historical deformation patterns, it can provide important references for analyzing the current slope deformation trend. For example, when the deformation intensity characteristics of the current monitoring data are similar to a certain historical deformation pattern segment, analyzing the subsequent deformation development path of that historical pattern segment can help determine the possible evolution direction of the current deformation. This changes the situation where traditional monitoring systems rely solely on real-time data for isolated analysis, improving the scientific rigor and rationality of deformation trend judgments.

[0051] The point of application positioning module can accurately analyze the displacement data changes of the matching deformation mode segment, determine the displacement acceleration point, and calculate the time delay. This function provides strong support for grasping the key development stages of slope deformation. The displacement acceleration point is an important node where slope deformation changes from slow development to rapid change. Accurately identifying this node and determining the time delay helps monitoring personnel clearly understand the time interval required from the onset of deformation acceleration to the potential risk state, allowing more reasonable preparation time for formulating prevention and control measures, and avoiding the problem of untimely response due to inaccurate identification of key nodes.

[0052] The matching degree analysis module determines the deformation matching degree by comparing the trends of stress change data sequences and displacement data sequences before and after the displacement acceleration point, further improving the accuracy of judging the correlation between historical deformation modes and current deformation. The similarity between different historical deformation mode segments and the current deformation is not only reflected in the deformation intensity characteristics, but also in the synergy of stress and displacement change trends. By analyzing this synergy to determine the deformation matching degree, the historical mode segments most similar to the current deformation can be screened out, ensuring the effectiveness of the reference historical data and reducing interference caused by inaccurate mode matching in trend judgment.

[0053] The weight calculation module determines deformation weights by combining time delay, deformation matching degree, and differences in deformation intensity characteristics, enabling the scientific quantification of the reference value of different matching deformation mode segments. Different historical deformation mode segments exhibit varying degrees of correlation and temporal characteristics with the current deformation, resulting in different reference significance for current deformation analysis. Weight allocation highlights mode segments with higher reference value, allowing monitoring personnel to focus on key aspects when analyzing historical data, thus improving data interpretation efficiency and trend prediction accuracy. Overall, through the collaborative work of its modules, the system significantly enhances the comprehensiveness, accuracy, and trend prediction capabilities of coal mine slope deformation monitoring, providing a more reliable technical guarantee for coal mine slope safety control. Attached Figure Description

[0054] Figure 1 This is a time-series diagram of the coal mine slope deformation monitoring system based on multi-source data fusion as described in this invention.

[0055] Figure 2 A flowchart illustrating how the pattern matching module works;

[0056] Figure 3 A flowchart illustrating the operation of the matching analysis module;

[0057] Figure 4 A flowchart illustrating how the weight calculation module works;

[0058] Figure 5 A flowchart illustrating how the topology mapping module works. Detailed Implementation

[0059] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] Please see Figure 1 The present invention provides a coal mine slope deformation monitoring system based on multi-source data fusion. The system includes: a data acquisition module, a pattern matching module, an action point positioning module, a matching degree analysis module, and a weight calculation module.

[0061] The data acquisition module uses displacement sensors, stress sensors, and microseismic monitoring equipment deployed in the slope area to collect displacement time series, stress change gradients, and microseismic energy distribution in real time. This data is transmitted wirelessly to the central processing unit for storage and preprocessing. Specifically, the displacement time series is collected by displacement sensors deployed at different elevations and key geological interfaces on the slope, recording the three-dimensional positional changes of the slope surface and deep rock mass. Time-correlated displacement data sequences are continuously generated at preset sampling intervals, providing basic time series data for the pattern matching module to screen historical deformation pattern segments. The stress change gradient is obtained by stress sensors embedded in potential weak zones of the slope, collecting stress values ​​within the rock mass in real time. Time series analysis of continuous stress data is performed to calculate the amount and rate of stress change at adjacent monitoring times, forming stress change gradient data. The microseismic energy distribution is captured by distributed microseismic monitoring equipment. By receiving microseismic wave signals generated by rock mass fracturing within the slope, the location of the signal, the duration of vibration, and the energy magnitude are identified, generating spatial distribution data of microseismic energy covering the monitoring area. This data, along with displacement and stress data, constitutes a multi-source monitoring dataset, providing comprehensive data support for subsequent data analysis by various modules. When this data, together with displacement time series and stress change gradient data, constitutes a multi-source monitoring dataset, the collected microseismic wave signals are first filtered by distributed microseismic monitoring equipment to remove interference signals such as environmental vibration. Then, based on the correspondence between the propagation velocity of microseismic waves and the lithological parameters of rock strata, the precise location of the microseismic event is determined, forming a microseismic energy spatial distribution dataset containing the coordinates, occurrence time, and energy value of the microseismic event. This dataset can supplement monitoring information from the perspective of changes in the internal structure of the slope. For example, when displacement data shows minor positional changes on the slope surface and stress data shows localized stress concentration, the microseismic events in the corresponding area within the same time period can be queried from the spatial distribution data of microseismic energy. If there are microseismic events in the area with energy values ​​exceeding the set basic threshold, and the microseismic events occurred earlier than the surface displacement changes, it can be determined that the surface changes are related to internal rock mass fracture. If there are no obvious microseismic events in the area or the microseismic energy values ​​are lower than the basic threshold, the possibility of surface changes caused by internal rock mass fracture can be ruled out. This provides more comprehensive raw data support for the subsequent pattern matching module to screen historical deformation pattern segments, helping to accurately match historical deformation patterns that match the deformation intensity characteristics of the current monitoring period.

[0062] The pattern matching module retrieves displacement data segments from the historical database that are similar to the deformation intensity characteristics of the current monitoring period. The historical database contains slope monitoring data accumulated over many years, and the deformation intensity characteristics are characterized by calculating the variance and mean of the displacement data.

[0063] The action point localization module performs time-series analysis on the selected matching deformation mode segments, identifies the displacement acceleration action point, and calculates the time delay from the starting point to the action point.

[0064] The matching degree analysis module compares the stress change data sequence and displacement data sequence before and after the displacement acceleration point, and evaluates the trend consistency through slope analysis.

[0065] The weight calculation module integrates the differences in time delay, deformation matching degree, and deformation intensity characteristics, and uses a weighted fusion algorithm to output deformation weights for subsequent deformation risk assessment.

[0066] Example 1: See Figure 1 This section details the specific operational procedures of the pattern matching module. The core function of this module is to filter displacement data segments from the historical database that match the deformation intensity characteristics of the current monitoring period; that is, to match deformation pattern segments. The historical data segmentation unit first processes the displacement time series in the historical database. This data originates from a long-term network of displacement sensors deployed on the slope, recording changes in slope surface displacement under different periods and geological conditions. The unit divides the continuous displacement time series data into multiple independent displacement data segments according to fixed time windows. The time length of each data segment is set to 24 hours, consistent with the cycle of daily monitoring reports, facilitating cross-time period comparative analysis. For each segmented displacement data, the unit extracts its deformation intensity characteristic value. This characteristic value is a composite index, obtained by calculating the root mean square error and peak amplitude of the displacement values ​​within the data segment. The root mean square error reflects the volatility and activity of the displacement data within that time period, while the peak amplitude indicates the maximum displacement occurring within that period. The combination of these two factors comprehensively characterizes the intensity features of slope deformation during that period. The extracted feature values, along with their corresponding timestamps and geological background information, are stored in the feature library to prepare for subsequent comparative analysis.

[0067] The intensity comparison unit is responsible for calculating the absolute difference between the deformation intensity characteristic values ​​of historical displacement data segments and those of the current monitoring period. The deformation intensity characteristic values ​​for the current monitoring period are calculated in real-time using the same algorithm, with data sourced from the operating displacement sensor. The calculation of the absolute difference is not a simple scalar subtraction, but rather a distance metric based on eigenvectors. Specifically, each eigenvalue is treated as a vector containing two components: root mean square error and peak amplitude. The absolute difference is obtained by calculating the Euclidean distance between the historical and current eigenvectors. This calculation method simultaneously considers the overall similarity of displacement fluctuations and the degree of matching in extreme cases. The calculated absolute difference is a non-negative real number; the smaller the value, the more similar the historical data segment and the current monitoring period are in terms of deformation intensity characteristics.

[0068] The negative correlation mapping unit receives the absolute difference values ​​output by the intensity comparison unit and converts them into deformation intensity similarity. Since a smaller absolute difference value indicates higher similarity (a negative correlation), it needs to be mathematically mapped to a positively correlated similarity index. This index must fall between zero and one for subsequent threshold determination. This conversion is achieved using a parameter-tuned sigmoid function. The center point of this function is determined based on the overall distribution of historical absolute difference values, allowing it to effectively distinguish between common and abnormal difference levels. After mapping, a larger absolute difference value outputs a similarity value close to zero, while a smaller absolute difference value outputs a similarity value close to one. Deformation intensity similarity intuitively quantifies the degree of matching between each historical data segment and the current situation.

[0069] The pattern selection unit filters matching deformation pattern segments from a large number of historical displacement data segments based on the deformation intensity similarity calculated by the negative correlation mapping unit. The selection process is based on a preset similarity threshold, which is usually set to 0.7. This means that only historical data segments that are highly similar to the deformation intensity characteristics of the current monitoring period are selected. All historical data segments with a similarity greater than or equal to this threshold are marked as matching deformation pattern segments and are completely extracted from the historical database, including their original displacement time series data, corresponding stress change data, and microseismic energy data. These selected matching segments are sent to a temporary buffer area, awaiting further in-depth analysis by the action point localization module. The entire pattern matching process is optimized through a distributed computing framework, enabling parallel retrieval and computation of massive historical data, thereby meeting the stringent processing speed requirements of the real-time monitoring system.

[0070] Example 2: See Figure 3This document details the operational procedures of the action point localization module and the matching degree analysis module. These two modules receive the output from the pattern matching module, perform in-depth analysis on the selected matching deformation pattern segments to determine their displacement acceleration action points, and evaluate their matching degree with the current monitoring data. The action point localization module processes each matching deformation pattern segment selected by the pattern matching module; these data segments contain historical displacement time series sequences. The module's primary task is to identify the points in the displacement data segment where abrupt changes occur, i.e., the displacement acceleration action points. To achieve this goal, the module employs an adaptive sliding window decomposition algorithm. This algorithm divides the input displacement time series sequence into a series of continuous and partially overlapping data windows. The window length is not fixed but adaptively adjusted according to the data sampling frequency and fluctuation characteristics. The initial window length is set to ten data points. If the standard deviation of the data within the window exceeds a certain level, the window is automatically reduced to capture finer changes; if the data is stable, the window is appropriately expanded to improve processing efficiency. Within each data window, the algorithm calculates the first-order difference of the displacement data, i.e., the difference between adjacent data points, thereby obtaining the displacement gradient sequence within that window. By analyzing the statistical characteristics of the gradient sequence and comparing it with a threshold trained on a large amount of historical data, the algorithm can determine whether a significant displacement abrupt change has occurred within the window. This threshold is typically set to twice the standard deviation of the average historical gradient data, thus filtering out ordinary fluctuations and focusing on genuine anomalous changes. When the displacement gradient within a data window first consistently exceeds this threshold, the monitoring time corresponding to the first data point within that window is marked as the displacement acceleration point. Once the point of action is determined, the module calculates the time elapsed from the start of the matching deformation pattern segment to the displacement acceleration point; this time difference is defined as the time delay of the data segment and recorded.

[0071] The matching degree analysis module further explores the intrinsic relationship between data before and after the displacement acceleration point. This module believes that a truly valuable matching pattern should exhibit a predictive correlation between stress changes before the acceleration point and displacement changes afterward. The module first extracts historical stress data for the corresponding time period from the data storage unit. The stress change analysis unit specifically processes the historical stress data sequence before the displacement acceleration point. This unit calculates the average slope of the changes between all adjacent data points in this sequence. This average slope value reflects the overall rate and trend of stress accumulation or release within the rock mass before the displacement acceleration occurs. Similarly, the displacement change analysis unit processes the historical displacement data sequence after the displacement acceleration point, using the same algorithm to calculate the average slope of the displacement changes between adjacent data points. This value characterizes the speed and direction of slope deformation development after the displacement abrupt change.

[0072] The trend difference mapping unit receives the calculation results from the first two units: the average slope of stress change and the average slope of displacement change. The core function of this unit is to quantify the consistency of these two slopes in terms of trend. It calculates the absolute difference between the two slope values, which intuitively reflects the degree of deviation between the stress change trend and the displacement change trend. A smaller difference means that the stress change well predicts subsequent displacement development, indicating that the historical pattern has high reference value. To transform this difference into a standardized and easily understood matching index, the unit uses an exponential decay function for negative correlation mapping. This function maps the input absolute difference to a deformation matching degree value between zero and one. When the absolute difference is zero, the deformation matching degree output is one, indicating a perfect match; as the absolute difference increases, the deformation matching degree value non-linearly decreases to zero. In this way, each matched deformation pattern segment is assigned a deformation matching degree, which provides a key basis for its importance in subsequent weight calculations.

[0073] Example 3: See Figure 4 This section details the specific operational procedures of the weight calculation module. The core function of this module is to calculate a quantified deformation weight for each matched deformation pattern segment by comprehensively considering multiple factors such as time delay, deformation matching degree, and differences in deformation intensity characteristics. This weight is used to assess the importance of the historical pattern for predicting the current slope condition. The coordinate construction unit first processes data from the action point positioning module and the pattern matching module. For each selected matched deformation pattern segment, this unit reads the stress value corresponding to its displacement acceleration action point. This value originates from the historical stress sensor records at the action point, reflecting the stress level experienced by the rock mass when significant displacement acceleration occurs. Simultaneously, the unit reads the time delay of the pattern segment, i.e., the time elapsed from the start of deformation to the occurrence of acceleration. Using the stress value as the abscissa and the time delay as the ordinate, a two-dimensional coordinate point is constructed for each matched deformation pattern segment. The coordinate points of all pattern segments together constitute a two-dimensional coordinate point set, which implicitly contains the potential relationship between stress level and deformation development rate.

[0074] The curve fitting unit receives the aforementioned two-dimensional coordinate point set and performs curve fitting on it using a nonlinear least squares method. The goal of the fitting is to find a mathematical expression that describes the overall trend between the stress value and the time delay. The unit attempts various models, such as quadratic curves and exponential decay curves, and selects the optimal fitting model by calculating the sum of squared residuals. Assume the final fitted curve equation is:

[0075]

[0076] in: It represents the time delay, and its physical meaning is the length of time required for deformation to go from the start to acceleration. Numerically, it is equal to the difference between the point of application and the start of the segment. The stress value representing the point of displacement acceleration is a direct measurement of the stress experienced by the rock mass at the critical moment. (Parameter) , ,and These are coefficients determined during the fitting process; they collectively define the shape and position of the curve, describing the overall pattern of time delay as a function of stress. This curve reflects the general temporal pattern of slope deformation development to the acceleration phase under different stress levels in historical data.

[0077] The residual calculation unit then evaluates the degree of deviation of each specific matching deformation pattern segment from the aforementioned general law. For each coordinate point in the point set... The unit will display the actual value of its time delay. At the same stress value as the fitted curve Predicted time delay at location Compare them. Calculate the absolute value of the difference between the two. This value is the absolute value of the residual. A large absolute value of the residual indicates that the development rate of this historical model segment deviates significantly from the normal pattern. Its cause may involve special geological conditions or external disturbances. Therefore, its importance may be reduced when used as a reference for the current state.

[0078] The weighting synthesis unit ultimately performs the calculation of deformation weights, which comprehensively considers three factors: deformation matching degree. Time delay and absolute value of residuals Deformation matching degree Provided by the matching degree analysis module, it quantifies the consistency of the trends in stress and displacement changes in this historical pattern. The element first calculates the ratio. The implication of this ratio is: the higher the trend consistency (larger S), the slower the development speed (…). The larger the historical pattern (and the more it conforms to general laws, the smaller the R value), the higher its reference value, and the larger the resulting ratio. Subsequently, the difference between this ratio and the deformation intensity characteristic is... Perform weighted fusion. Difference values. Derived from the pattern matching module, this represents the difference in overall deformation intensity between historical and current patterns. The fusion process employs a linear weighting method, calculated by acknowledging the reference value of historical patterns (represented by the ratio) while also considering their overall similarity to the current situation (represented by D). The final output deformation weight is a normalized value, which will be used in subsequent probabilistic extrapolation to assess the stability risk of the current slope.

[0079] Example 4: SeeFigure 5 This involves the specific operational procedures of the topology mapping module and the probability inference module. The function of these two modules is to use the calculation results of the aforementioned modules to construct a network model reflecting the internal relationships of the rock strata, and based on this, simulate deformation propagation, ultimately generating a spatial probability distribution map that can intuitively display the potential risk areas of the slope.

[0080] The topology mapping module first constructs an initial topology network of slope rock strata nodes based on detailed exploration data provided by the geological exploration department. Microseismic energy spatial distribution data can assist in optimizing the initial topology network and updating node connection weights. In the initial topology network construction stage, which includes the locations of rock strata nodes and lithological parameters based on geological exploration data, the coordinate information of all microseismic events in the historical microseismic energy spatial distribution data is extracted. The frequency of occurrence and cumulative energy value of microseismic events per unit area in different regions are statistically analyzed. Areas with a higher frequency of occurrence and a higher cumulative energy value than the average are marked as key areas of interest. Subsequently, during the initial topology network construction process, the number of rock strata nodes in the key areas of interest is increased and the node spacing is reduced according to preset subdivision rules, achieving a more detailed division of rock strata nodes in these areas prone to rock fracturing and concentrated energy release. The coordinates of the displacement acceleration points are mapped to the initial topology network. Subsequently, when updating the node connection weights based on the stress transmission relationship of rock strata, the propagation path data of microseismic waves in the corresponding rock strata node areas are first extracted from the microseismic energy spatial distribution data. The propagation time and energy attenuation of microseismic waves between different nodes are analyzed. Nodes with short propagation times and small energy attenuation indicate low stress transmission resistance and high efficiency, and are assigned higher weight values ​​when updating node connection weights. Nodes with long propagation times and large energy attenuation indicate high stress transmission resistance and low efficiency, and are assigned lower weight values. This method more accurately judges the actual efficiency of stress transmission between different rock strata nodes, making the generated rock strata node topology network more closely match the actual deformation propagation characteristics inside the slope. These data come from previous borehole exploration, ground-penetrating radar scanning, and core sample laboratory analysis, and include key parameters such as rock strata interface coordinates, lithological classification, joint development degree, elastic modulus, Poisson's ratio, and density. The initial network discretizes the slope rock mass into a series of interconnected nodes, each node representing a rock mass unit at a specific location (defined by three-dimensional coordinates X, Y, Z) and accompanied by its lithological parameters. The connection relationships (edges) between nodes are established based on geological structural features (such as bedding planes and fault lines) and the principle of spatial proximity. The initial connection weights are usually set based on the Euclidean distance and lithological similarity between nodes.

[0081] The module maps the displacement acceleration point information corresponding to each matched deformation pattern segment selected by the pattern matching module onto this initial network. Each displacement acceleration point is associated with a geographic coordinate, obtained through a high-precision GPS positioning system. The module employs a spatial nearest neighbor matching algorithm to find the nearest rock stratum node for each point coordinate in the initial topology network. Once a corresponding node is found, it is marked as a "deformation source," and its associated attribute data is updated, incorporating the deformation weight value of that historical pattern segment from the weight calculation module. This weight value indicates the historical credibility of the node as a deformation origin point and its impact potential.

[0082] After mapping, the module does not maintain static network connections; instead, it dynamically updates the connection weights between nodes based on the stress transmission principles in rock mechanics. The connection weights no longer depend solely on the initial distance and lithology but also take into account the stress transmission efficiency of the rock layers. For example, a connection passing through a hard, intact sandstone layer will have a higher weight than a connection passing through a weak, fractured mudstone layer. Through this dynamic updating, a rock layer node topology network that more closely reflects the actual geomechanical environment is generated. This network quantifies the ease with which deformation and stress propagate along different paths within the rock mass.

[0083] The probabilistic inference module receives this dynamic topological network carrying deformation weight information and inputs it into the deformation propagation inference model. The core of this model is a convolutional network based on graph theory and tensor operations, used to simulate the process of deformation waves propagating through the rock strata network from various "deformation source" nodes. The propagation simulation unit performs this calculation, treating the deformation weight of each source node as initial energy. The deformation wave propagates along the network edges, and its "energy" is reduced at each node and connection it passes through, based on the attenuation coefficient determined by the connection weight and the node's lithological parameters. This process simulates the propagation of multiple waves, recording the deformation wave's arrival at each node throughout the network and its intensity.

[0084] The probability calculation unit then processes this propagation path data, counting the frequency with which each rock stratum node is traversed by the deformation wave in multiple simulated propagations. A higher frequency indicates a greater likelihood that the node is located on a potential deformation propagation path. This frequency is then combined with the node's own lithological parameters; for example, weak rock strata nodes are more prone to sustained deformation after disturbance, thus receiving a higher residence probability multiplier. Through calculation, each node is ultimately assigned an anomalous deformation residence probability value, which comprehensively reflects its likelihood of being influenced by historical deformation sources and its own lithological vulnerability.

[0085] The heatmap generation unit ultimately converts these discrete node probability values ​​into a continuous probability distribution map covering the entire slope area. It employs spatial interpolation techniques to diffuse the probability value of each node to its surrounding area, forming smooth probability contour lines. The final output abnormal deformation probability distribution map is a color heatmap, typically using blue to represent low-probability areas and red to represent high-probability areas, thus visually identifying potential hazard areas requiring close monitoring.

[0086] Table 1: Topological network attributes of rock strata nodes.

[0087]

[0088] Table 1 shows the properties of some nodes in the topology network. Node N102 is identified as a deformation source with a deformation weight of 0.72. Its connection weight with node N101 (both belonging to sandstone) is relatively high (0.85), while its connection weight with node N105 (belonging to mudstone) is relatively low (0.60), reflecting the influence of lithology on stress propagation. Although nodes N105 and N110 are not direct source nodes, they may be affected by stress propagation through network connections.

[0089] Example 5: Specific operational procedures involving the strategy generation module and the control execution module. These two modules automatically generate targeted monitoring instructions based on the risk visualization results produced by the aforementioned probability inference module, and drive the hardware system to execute corresponding data acquisition mode adjustments, thereby achieving closed-loop control from risk analysis to proactive monitoring. The strategy generation module processes the abnormal deformation probability distribution map output by the probability inference module. This module first analyzes this heat map covering the entire slope area, extracting the coordinates of all rock strata nodes with probability values ​​exceeding a preset threshold, forming a high-probability node set. This threshold is a configurable parameter, typically set based on inversion analysis of historical landslide events or domain expert experience, used to filter out core areas requiring priority attention. Subsequently, the module analyzes the spatial relationships between these high-probability nodes in depth. It calculates the probability gradient change rate between adjacent nodes, reflecting the steepness of the probability value change in space. A sharp gradient change indicates a rapid increase or decrease in risk over a short distance, often marking the location boundary of a potential sliding surface or geologically weak zone. Based on these spatial analysis results, the module automatically configures two sets of core monitoring strategy parameters. The first set of parameters is the monitoring frequency, which is positively correlated with the calculated rate of change of the probability gradient. For areas with drastic gradient changes, it indicates that the rock mass may be at an unstable critical point, thus requiring a significant increase in monitoring frequency to capture potential rapid deformation. The second set is the perturbation test parameters, which define in detail the location of the rock strata nodes to be artificially disturbed, the frequency range of the vibration signal, the signal amplitude, and the duration. These parameters are set based on probability distribution patterns and geological lithology data, with the aim of detecting the dynamic response characteristics of suspected weak areas through active excitation.

[0090] The control execution module receives instruction parameters from the strategy generation module and translates them into specific hardware control commands. This module contains two functional units, each responsible for different execution tasks. The mode switching unit is specifically responsible for adjusting the sampling frequency of the displacement sensors. When the monitoring frequency parameter is triggered, this unit sends instructions via industrial fieldbus or wireless communication network to the group of displacement sensors deployed in a designated high-probability area, increasing their data acquisition frequency from the conventional base frequency to a higher level, such as five times the original frequency. This increase ensures that the system can capture possible transient or high-frequency deformation signals, providing a higher-resolution data foundation for analyzing the dynamic stability of slopes.

[0091] The disturbance execution unit is responsible for performing more complex active detection tasks. Based on the received disturbance test parameters, this unit controls hydraulic or electric vibrators installed at specific locations on the slope. These vibrators can apply precisely controlled multi-frequency vibration excitation signals to designated rock strata nodes. The frequency range of the excitation signals covers the spectrum from low to mid-frequency, designed to excite structural surfaces and their resonant characteristics at different depths of the rock mass. The amplitude and duration of the signals are carefully designed to ensure both a sufficiently measurable response signal and energy control within a safe range to avoid human interference with slope stability. Simultaneously with the excitation, the system activates high-speed acquisition modes of all relevant sensors to record the rock mass's response data to artificial vibration. This data will be used for subsequent analysis of the rock mass's structural integrity and potential weak zones.

[0092] Through the collaborative work of the strategy generation module and the control execution module, the entire monitoring system has upgraded from passive perception to active detection. The system no longer merely records deformation data under natural conditions, but can intelligently adjust the allocation of monitoring resources based on real-time risk assessments and conduct targeted, detailed detection of key suspicious areas. This dynamic monitoring strategy based on risk assessment feedback significantly improves the system's ability to detect potential hazards and the timeliness of early warnings.

[0093] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0094] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A coal mine slope deformation monitoring system based on multi-source data fusion, characterized in that, include: The data acquisition module is used to acquire multi-source monitoring data of the slope in real time. The multi-source monitoring data includes displacement time series, stress change gradient and microseismic energy distribution. The pattern matching module is used to filter deformation pattern segments from the historical database that match the deformation intensity characteristics of the current monitoring period. The application point positioning module is used to analyze the displacement data changes of the matching deformation mode segment in time and determine the displacement acceleration application point of the matching deformation mode segment. The time difference between the starting point of each matching deformation mode segment and the point of displacement acceleration is used as the time delay of each matching deformation mode segment. The matching degree analysis module is used to compare the change trends of the stress change data sequence before the displacement acceleration point with the displacement data sequence after the displacement acceleration point to determine the deformation matching degree of each matching deformation mode segment. The weight calculation module is used to determine the deformation weight of the matching deformation mode segment by combining the time delay, the deformation matching degree, and the difference between the deformation intensity characteristics of the matching deformation mode segment and the current monitoring period. The pattern matching module includes: The historical data segmentation unit is used to divide the displacement time series in the historical database into multiple displacement data segments and extract the deformation intensity characteristic value of each displacement data segment. The strength comparison unit is used to calculate the absolute difference between the deformation strength characteristic value of the historical displacement data segment and the deformation strength characteristic value of the current monitoring period. The negative correlation mapping unit is used to perform negative correlation mapping on the absolute difference values ​​to obtain the deformation intensity similarity. The pattern filtering unit is used to filter out matching deformation pattern segments from historical displacement data segments based on the deformation intensity similarity. The matching degree analysis module includes: The stress change analysis unit is used to calculate the average slope of adjacent data points in the stress change data sequence before the displacement acceleration point. The displacement change analysis unit is used to calculate the average slope of adjacent data points in the displacement data sequence after the displacement acceleration point. The trend difference mapping unit is used to perform negative correlation mapping on the difference between the output value of the stress change analysis unit and the output value of the displacement change analysis unit, and output the deformation matching degree. The weight calculation module includes: The coordinate construction unit uses the stress value corresponding to the displacement acceleration point of each matching deformation mode segment as the abscissa and the time delay as the ordinate to construct a two-dimensional coordinate point set. The curve fitting unit performs nonlinear curve fitting on the set of two-dimensional coordinate points to generate a fitted curve. The residual calculation unit calculates the absolute value of the residual between each coordinate point and the fitted curve; The weighted synthesis unit uses the deformation matching degree as the numerator and the product of the time delay and the absolute value of the residual as the denominator. It then weights and fuses the ratio result with the difference value of the deformation intensity feature to output the deformation weight.

2. The coal mine slope deformation monitoring system based on multi-source data fusion according to claim 1, characterized in that, The action point positioning module specifically performs the following: For any matching deformation mode segment, an adaptive sliding window decomposition algorithm is used to identify abrupt change points in the displacement time sequence; The monitoring moment when the first abrupt change in displacement gradient occurs is marked as the displacement acceleration point.

3. The coal mine slope deformation monitoring system based on multi-source data fusion according to claim 1, characterized in that, Also includes: The topology mapping module constructs a slope rock stratum node topology network based on the rock stratum node positions corresponding to the displacement acceleration points of the matched deformation mode segments. The topology mapping module specifically executes: An initial topology network containing the locations of rock strata nodes and lithological parameters is constructed based on geological exploration data; Map the coordinates of the displacement acceleration point of the matching deformation mode segment to the corresponding rock layer node of the initial topology network; The node connection weights are updated by analyzing the stress transmission relationship of the rock strata to generate a rock strata node topology network.

4. The coal mine slope deformation monitoring system based on multi-source data fusion according to claim 3, characterized in that, Also includes: The probability inference module is used to input the deformation weights and the rock stratum node topology network into the deformation propagation inference model to generate an abnormal deformation probability distribution map of the slope area. The probability deduction module includes: The propagation simulation unit is used to simulate the propagation path of deformable waves in the rock strata based on tensor convolutional networks, using nodes carrying deformation weights as the initial propagation source. The probability calculation unit counts the frequency of occurrence of each rock layer node in the propagation path and calculates the probability of abnormal deformation residence by combining lithological parameters. The heat map generation unit generates an abnormal deformation probability distribution map covering the entire slope based on the abnormal deformation retention probability.

5. The coal mine slope deformation monitoring system based on multi-source data fusion according to claim 4, characterized in that, Also includes: The strategy generation module configures monitoring strategy parameters based on the gradient change characteristics of the abnormal deformation probability distribution map. The strategy generation module specifically executes the following: Extract the set of rock strata nodes whose probability values ​​exceed a preset threshold from the abnormal deformation probability distribution map; Calculate the rate of change of the probability gradient between adjacent nodes; Based on the probability gradient change rate, monitoring strategy parameters including monitoring frequency parameters and disturbance test parameters are generated.

6. The coal mine slope deformation monitoring system based on multi-source data fusion according to claim 5, characterized in that, Also includes: The control execution module adjusts the acquisition mode of multi-source monitoring data based on the monitoring strategy parameters; The control execution module includes: The mode switching unit is used to increase the sampling frequency of the displacement sensor to 5 times the original frequency when the monitoring frequency parameter is triggered. The disturbance execution unit is used to apply multi-frequency vibration excitation signals to designated rock stratum nodes according to the disturbance test parameters.

7. A method for monitoring coal mine slope deformation based on multi-source data fusion, characterized in that, It includes all modules and method flows of the coal mine slope deformation monitoring system based on multi-source data fusion as described in any one of claims 1 to 6.

Images